blob: b1878faea3974f6ae984763109edae78c4cd3342 [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>
Jason Wang6680ec62013-07-25 13:00:33 +080063#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064#include <linux/crc32.h>
Pavel Emelyanovd647a592008-04-16 00:41:16 -070065#include <linux/nsproxy.h>
Rusty Russellf43798c2008-07-03 03:48:02 -070066#include <linux/virtio_net.h>
Michael S. Tsirkin99405162010-02-14 01:01:10 +000067#include <linux/rcupdate.h>
Eric W. Biederman881d9662007-09-17 11:56:21 -070068#include <net/net_namespace.h>
Pavel Emelyanov79d17602008-04-16 00:40:46 -070069#include <net/netns/generic.h>
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -080070#include <net/rtnetlink.h>
Herbert Xu33dccbb2009-02-05 21:25:32 -080071#include <net/sock.h>
Masatake YAMATO93e14b62014-01-29 16:43:31 +090072#include <linux/seq_file.h>
Herbert Xue0b46d02014-11-07 21:22:23 +080073#include <linux/uio.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <asm/uaccess.h>
76
Rusty Russell14daa022008-04-12 18:48:58 -070077/* Uncomment to enable debugging */
78/* #define TUN_DEBUG 1 */
79
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#ifdef TUN_DEBUG
81static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070082
Joe Perches6b8a66e2011-03-02 07:18:10 +000083#define tun_debug(level, tun, fmt, args...) \
84do { \
85 if (tun->debug) \
86 netdev_printk(level, tun->dev, fmt, ##args); \
87} while (0)
88#define DBG1(level, fmt, args...) \
89do { \
90 if (debug == 2) \
91 printk(level fmt, ##args); \
92} while (0)
Rusty Russell14daa022008-04-12 18:48:58 -070093#else
Joe Perches6b8a66e2011-03-02 07:18:10 +000094#define tun_debug(level, tun, fmt, args...) \
95do { \
96 if (0) \
97 netdev_printk(level, tun->dev, fmt, ##args); \
98} while (0)
99#define DBG1(level, fmt, args...) \
100do { \
101 if (0) \
102 printk(level fmt, ##args); \
103} while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#endif
105
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +0200106/* TUN device flags */
107
108/* IFF_ATTACH_QUEUE is never stored in device flags,
109 * overload it to mean fasync when stored there.
110 */
111#define TUN_FASYNC IFF_ATTACH_QUEUE
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200112/* High bits in flags field are unused. */
113#define TUN_VNET_LE 0x80000000
Greg Kurz8b8e6582015-04-24 14:50:36 +0200114#define TUN_VNET_BE 0x40000000
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +0200115
116#define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200117 IFF_MULTI_QUEUE)
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000118#define GOODCOPY_LEN 128
119
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700120#define FLT_EXACT_COUNT 8
121struct tap_filter {
122 unsigned int count; /* Number of addrs. Zero means disabled */
123 u32 mask[2]; /* Mask of the hashed addrs */
124 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
125};
126
Pankaj Guptabaf71c52015-01-12 11:41:29 +0530127/* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal
128 * to max number of VCPUs in guest. */
129#define MAX_TAP_QUEUES 256
Jason Wangb8732fb2013-01-23 03:59:13 +0000130#define MAX_TAP_FLOWS 4096
Jason Wangc8d68e62012-10-31 19:46:00 +0000131
Jason Wang96442e422012-10-31 19:46:02 +0000132#define TUN_FLOW_EXPIRE (3 * HZ)
133
Jason Wang54f968d2012-10-31 19:45:57 +0000134/* A tun_file connects an open character device to a tuntap netdevice. It
stephen hemminger92d4ea62013-12-05 20:42:58 -0800135 * also contains all socket related structures (except sock_fprog and tap_filter)
Jason Wang54f968d2012-10-31 19:45:57 +0000136 * to serve as one transmit queue for tuntap device. The sock_fprog and
137 * tap_filter were kept in tun_struct since they were used for filtering for the
Rami Rosen36fe8c092012-11-25 22:07:40 +0000138 * netdevice not for a specific queue (at least I didn't see the requirement for
Jason Wang54f968d2012-10-31 19:45:57 +0000139 * this).
Jason Wang6e914fc2012-10-31 19:45:58 +0000140 *
141 * RCU usage:
Rami Rosen36fe8c092012-11-25 22:07:40 +0000142 * The tun_file and tun_struct are loosely coupled, the pointer from one to the
Jason Wang6e914fc2012-10-31 19:45:58 +0000143 * other can only be read while rcu_read_lock or rtnl_lock is held.
Jason Wang54f968d2012-10-31 19:45:57 +0000144 */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000145struct tun_file {
Jason Wang54f968d2012-10-31 19:45:57 +0000146 struct sock sk;
147 struct socket socket;
148 struct socket_wq wq;
Jason Wang6e914fc2012-10-31 19:45:58 +0000149 struct tun_struct __rcu *tun;
Jason Wang54f968d2012-10-31 19:45:57 +0000150 struct fasync_struct *fasync;
151 /* only used for fasnyc */
152 unsigned int flags;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +0400153 union {
154 u16 queue_index;
155 unsigned int ifindex;
156 };
Jason Wang4008e972012-12-13 23:53:30 +0000157 struct list_head next;
158 struct tun_struct *detached;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000159};
160
Jason Wang96442e422012-10-31 19:46:02 +0000161struct tun_flow_entry {
162 struct hlist_node hash_link;
163 struct rcu_head rcu;
164 struct tun_struct *tun;
165
166 u32 rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800167 u32 rps_rxhash;
Jason Wang96442e422012-10-31 19:46:02 +0000168 int queue_index;
169 unsigned long updated;
170};
171
172#define TUN_NUM_FLOW_ENTRIES 1024
173
Jason Wang54f968d2012-10-31 19:45:57 +0000174/* Since the socket were moved to tun_file, to preserve the behavior of persist
Rami Rosen36fe8c092012-11-25 22:07:40 +0000175 * device, socket filter, sndbuf and vnet header size were restore when the
Jason Wang54f968d2012-10-31 19:45:57 +0000176 * file were attached to a persist device.
177 */
Rusty Russell14daa022008-04-12 18:48:58 -0700178struct tun_struct {
Jason Wangc8d68e62012-10-31 19:46:00 +0000179 struct tun_file __rcu *tfiles[MAX_TAP_QUEUES];
180 unsigned int numqueues;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700181 unsigned int flags;
Eric W. Biederman0625c882012-02-07 16:48:55 -0800182 kuid_t owner;
183 kgid_t group;
Rusty Russell14daa022008-04-12 18:48:58 -0700184
Rusty Russell14daa022008-04-12 18:48:58 -0700185 struct net_device *dev;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000186 netdev_features_t set_features;
Michał Mirosław88255372011-04-19 06:13:10 +0000187#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
Vlad Yaseviche3e3c422015-02-03 16:36:17 -0500188 NETIF_F_TSO6|NETIF_F_UFO)
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200189
190 int vnet_hdr_sz;
Jason Wang54f968d2012-10-31 19:45:57 +0000191 int sndbuf;
192 struct tap_filter txflt;
193 struct sock_fprog fprog;
194 /* protected by rtnl lock */
195 bool filter_attached;
Rusty Russell14daa022008-04-12 18:48:58 -0700196#ifdef TUN_DEBUG
197 int debug;
198#endif
Jason Wang96442e422012-10-31 19:46:02 +0000199 spinlock_t lock;
Jason Wang96442e422012-10-31 19:46:02 +0000200 struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
201 struct timer_list flow_gc_timer;
202 unsigned long ageing_time;
Jason Wang4008e972012-12-13 23:53:30 +0000203 unsigned int numdisabled;
204 struct list_head disabled;
Paul Moore5dbbaf22013-01-14 07:12:19 +0000205 void *security;
Jason Wangb8732fb2013-01-23 03:59:13 +0000206 u32 flow_count;
Rusty Russell14daa022008-04-12 18:48:58 -0700207};
208
Greg Kurz8b8e6582015-04-24 14:50:36 +0200209#ifdef CONFIG_TUN_VNET_CROSS_LE
210static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
211{
212 return tun->flags & TUN_VNET_BE ? false :
213 virtio_legacy_is_little_endian();
214}
215
216static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
217{
218 int be = !!(tun->flags & TUN_VNET_BE);
219
220 if (put_user(be, argp))
221 return -EFAULT;
222
223 return 0;
224}
225
226static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
227{
228 int be;
229
230 if (get_user(be, argp))
231 return -EFAULT;
232
233 if (be)
234 tun->flags |= TUN_VNET_BE;
235 else
236 tun->flags &= ~TUN_VNET_BE;
237
238 return 0;
239}
240#else
241static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
242{
243 return virtio_legacy_is_little_endian();
244}
245
246static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
247{
248 return -EINVAL;
249}
250
251static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
252{
253 return -EINVAL;
254}
255#endif /* CONFIG_TUN_VNET_CROSS_LE */
256
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200257static inline bool tun_is_little_endian(struct tun_struct *tun)
258{
Greg Kurz7d824102015-04-24 14:26:24 +0200259 return tun->flags & TUN_VNET_LE ||
Greg Kurz8b8e6582015-04-24 14:50:36 +0200260 tun_legacy_is_little_endian(tun);
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200261}
262
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300263static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val)
264{
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200265 return __virtio16_to_cpu(tun_is_little_endian(tun), val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300266}
267
268static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val)
269{
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200270 return __cpu_to_virtio16(tun_is_little_endian(tun), val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300271}
272
Jason Wang96442e422012-10-31 19:46:02 +0000273static inline u32 tun_hashfn(u32 rxhash)
274{
275 return rxhash & 0x3ff;
276}
277
278static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash)
279{
280 struct tun_flow_entry *e;
Jason Wang96442e422012-10-31 19:46:02 +0000281
Sasha Levinb67bfe02013-02-27 17:06:00 -0800282 hlist_for_each_entry_rcu(e, head, hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000283 if (e->rxhash == rxhash)
284 return e;
285 }
286 return NULL;
287}
288
289static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun,
290 struct hlist_head *head,
291 u32 rxhash, u16 queue_index)
292{
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000293 struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC);
294
Jason Wang96442e422012-10-31 19:46:02 +0000295 if (e) {
296 tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n",
297 rxhash, queue_index);
298 e->updated = jiffies;
299 e->rxhash = rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800300 e->rps_rxhash = 0;
Jason Wang96442e422012-10-31 19:46:02 +0000301 e->queue_index = queue_index;
302 e->tun = tun;
303 hlist_add_head_rcu(&e->hash_link, head);
Jason Wangb8732fb2013-01-23 03:59:13 +0000304 ++tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000305 }
306 return e;
307}
308
Jason Wang96442e422012-10-31 19:46:02 +0000309static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e)
310{
311 tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n",
312 e->rxhash, e->queue_index);
313 hlist_del_rcu(&e->hash_link);
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000314 kfree_rcu(e, rcu);
Jason Wangb8732fb2013-01-23 03:59:13 +0000315 --tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000316}
317
318static void tun_flow_flush(struct tun_struct *tun)
319{
320 int i;
321
322 spin_lock_bh(&tun->lock);
323 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
324 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800325 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000326
Sasha Levinb67bfe02013-02-27 17:06:00 -0800327 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
Jason Wang96442e422012-10-31 19:46:02 +0000328 tun_flow_delete(tun, e);
329 }
330 spin_unlock_bh(&tun->lock);
331}
332
333static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
334{
335 int i;
336
337 spin_lock_bh(&tun->lock);
338 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
339 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800340 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000341
Sasha Levinb67bfe02013-02-27 17:06:00 -0800342 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000343 if (e->queue_index == queue_index)
344 tun_flow_delete(tun, e);
345 }
346 }
347 spin_unlock_bh(&tun->lock);
348}
349
350static void tun_flow_cleanup(unsigned long data)
351{
352 struct tun_struct *tun = (struct tun_struct *)data;
353 unsigned long delay = tun->ageing_time;
354 unsigned long next_timer = jiffies + delay;
355 unsigned long count = 0;
356 int i;
357
358 tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n");
359
360 spin_lock_bh(&tun->lock);
361 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
362 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800363 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000364
Sasha Levinb67bfe02013-02-27 17:06:00 -0800365 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000366 unsigned long this_timer;
367 count++;
368 this_timer = e->updated + delay;
369 if (time_before_eq(this_timer, jiffies))
370 tun_flow_delete(tun, e);
371 else if (time_before(this_timer, next_timer))
372 next_timer = this_timer;
373 }
374 }
375
376 if (count)
377 mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer));
378 spin_unlock_bh(&tun->lock);
379}
380
Eric Dumazet49974422012-12-12 19:22:57 +0000381static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
Jason Wang9e857222013-01-28 01:05:19 +0000382 struct tun_file *tfile)
Jason Wang96442e422012-10-31 19:46:02 +0000383{
384 struct hlist_head *head;
385 struct tun_flow_entry *e;
386 unsigned long delay = tun->ageing_time;
Jason Wang9e857222013-01-28 01:05:19 +0000387 u16 queue_index = tfile->queue_index;
Jason Wang96442e422012-10-31 19:46:02 +0000388
389 if (!rxhash)
390 return;
391 else
392 head = &tun->flows[tun_hashfn(rxhash)];
393
394 rcu_read_lock();
395
Jason Wang9e857222013-01-28 01:05:19 +0000396 /* We may get a very small possibility of OOO during switching, not
397 * worth to optimize.*/
398 if (tun->numqueues == 1 || tfile->detached)
Jason Wang96442e422012-10-31 19:46:02 +0000399 goto unlock;
400
401 e = tun_flow_find(head, rxhash);
402 if (likely(e)) {
403 /* TODO: keep queueing to old queue until it's empty? */
404 e->queue_index = queue_index;
405 e->updated = jiffies;
Tom Herbert9bc88932013-12-22 18:54:32 +0800406 sock_rps_record_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000407 } else {
408 spin_lock_bh(&tun->lock);
Jason Wangb8732fb2013-01-23 03:59:13 +0000409 if (!tun_flow_find(head, rxhash) &&
410 tun->flow_count < MAX_TAP_FLOWS)
Jason Wang96442e422012-10-31 19:46:02 +0000411 tun_flow_create(tun, head, rxhash, queue_index);
412
413 if (!timer_pending(&tun->flow_gc_timer))
414 mod_timer(&tun->flow_gc_timer,
415 round_jiffies_up(jiffies + delay));
416 spin_unlock_bh(&tun->lock);
417 }
418
419unlock:
420 rcu_read_unlock();
421}
422
Tom Herbert9bc88932013-12-22 18:54:32 +0800423/**
424 * Save the hash received in the stack receive path and update the
425 * flow_hash table accordingly.
426 */
427static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash)
428{
Eric Dumazet567e4b72015-02-06 12:59:01 -0800429 if (unlikely(e->rps_rxhash != hash))
Tom Herbert9bc88932013-12-22 18:54:32 +0800430 e->rps_rxhash = hash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800431}
432
Jason Wangc8d68e62012-10-31 19:46:00 +0000433/* We try to identify a flow through its rxhash first. The reason that
stephen hemminger92d4ea62013-12-05 20:42:58 -0800434 * we do not check rxq no. is because some cards(e.g 82599), chooses
Jason Wangc8d68e62012-10-31 19:46:00 +0000435 * the rxq based on the txq where the last packet of the flow comes. As
436 * the userspace application move between processors, we may get a
437 * different rxq no. here. If we could not get rxhash, then we would
438 * hope the rxq no. may help here.
439 */
Jason Wangf663dd92014-01-10 16:18:26 +0800440static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100441 void *accel_priv, select_queue_fallback_t fallback)
Jason Wangc8d68e62012-10-31 19:46:00 +0000442{
443 struct tun_struct *tun = netdev_priv(dev);
Jason Wang96442e422012-10-31 19:46:02 +0000444 struct tun_flow_entry *e;
Jason Wangc8d68e62012-10-31 19:46:00 +0000445 u32 txq = 0;
446 u32 numqueues = 0;
447
448 rcu_read_lock();
Jason Wang92bb73e2013-06-05 16:44:57 +0800449 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000450
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800451 txq = skb_get_hash(skb);
Jason Wangc8d68e62012-10-31 19:46:00 +0000452 if (txq) {
Jason Wang96442e422012-10-31 19:46:02 +0000453 e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
Tom Herbert9bc88932013-12-22 18:54:32 +0800454 if (e) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800455 tun_flow_save_rps_rxhash(e, txq);
Zhi Yong Wufbe4d452014-01-02 13:24:28 +0800456 txq = e->queue_index;
Tom Herbert9bc88932013-12-22 18:54:32 +0800457 } else
Jason Wang96442e422012-10-31 19:46:02 +0000458 /* use multiply and shift instead of expensive divide */
459 txq = ((u64)txq * numqueues) >> 32;
Jason Wangc8d68e62012-10-31 19:46:00 +0000460 } else if (likely(skb_rx_queue_recorded(skb))) {
461 txq = skb_get_rx_queue(skb);
462 while (unlikely(txq >= numqueues))
463 txq -= numqueues;
464 }
465
466 rcu_read_unlock();
467 return txq;
468}
469
Jason Wangcde8b152012-10-31 19:46:01 +0000470static inline bool tun_not_capable(struct tun_struct *tun)
471{
472 const struct cred *cred = current_cred();
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000473 struct net *net = dev_net(tun->dev);
Jason Wangcde8b152012-10-31 19:46:01 +0000474
475 return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
476 (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000477 !ns_capable(net->user_ns, CAP_NET_ADMIN);
Jason Wangcde8b152012-10-31 19:46:01 +0000478}
479
Jason Wangc8d68e62012-10-31 19:46:00 +0000480static void tun_set_real_num_queues(struct tun_struct *tun)
481{
482 netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
483 netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
484}
485
Jason Wang4008e972012-12-13 23:53:30 +0000486static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
487{
488 tfile->detached = tun;
489 list_add_tail(&tfile->next, &tun->disabled);
490 ++tun->numdisabled;
491}
492
Jason Wangd32649d2012-12-18 11:00:27 +0800493static struct tun_struct *tun_enable_queue(struct tun_file *tfile)
Jason Wang4008e972012-12-13 23:53:30 +0000494{
495 struct tun_struct *tun = tfile->detached;
496
497 tfile->detached = NULL;
498 list_del_init(&tfile->next);
499 --tun->numdisabled;
500 return tun;
501}
502
Jason Wang4bfb0512013-09-05 17:53:59 +0800503static void tun_queue_purge(struct tun_file *tfile)
504{
505 skb_queue_purge(&tfile->sk.sk_receive_queue);
506 skb_queue_purge(&tfile->sk.sk_error_queue);
507}
508
Jason Wangc8d68e62012-10-31 19:46:00 +0000509static void __tun_detach(struct tun_file *tfile, bool clean)
510{
511 struct tun_file *ntfile;
512 struct tun_struct *tun;
Jason Wangc8d68e62012-10-31 19:46:00 +0000513
Jason Wangb8deabd2013-01-11 16:59:32 +0000514 tun = rtnl_dereference(tfile->tun);
515
Jason Wang9e857222013-01-28 01:05:19 +0000516 if (tun && !tfile->detached) {
Jason Wangc8d68e62012-10-31 19:46:00 +0000517 u16 index = tfile->queue_index;
518 BUG_ON(index >= tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000519
520 rcu_assign_pointer(tun->tfiles[index],
521 tun->tfiles[tun->numqueues - 1]);
Jason Wangb8deabd2013-01-11 16:59:32 +0000522 ntfile = rtnl_dereference(tun->tfiles[index]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000523 ntfile->queue_index = index;
524
525 --tun->numqueues;
Jason Wang9e857222013-01-28 01:05:19 +0000526 if (clean) {
Monam Agarwalc9566742014-03-24 00:02:32 +0530527 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang4008e972012-12-13 23:53:30 +0000528 sock_put(&tfile->sk);
Jason Wang9e857222013-01-28 01:05:19 +0000529 } else
Jason Wang4008e972012-12-13 23:53:30 +0000530 tun_disable_queue(tun, tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000531
532 synchronize_net();
Jason Wang96442e422012-10-31 19:46:02 +0000533 tun_flow_delete_by_queue(tun, tun->numqueues + 1);
Jason Wangc8d68e62012-10-31 19:46:00 +0000534 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800535 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000536 tun_set_real_num_queues(tun);
Jason Wangdd38bd82013-01-11 16:59:34 +0000537 } else if (tfile->detached && clean) {
Jason Wang4008e972012-12-13 23:53:30 +0000538 tun = tun_enable_queue(tfile);
Jason Wangdd38bd82013-01-11 16:59:34 +0000539 sock_put(&tfile->sk);
540 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000541
542 if (clean) {
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000543 if (tun && tun->numqueues == 0 && tun->numdisabled == 0) {
544 netif_carrier_off(tun->dev);
545
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200546 if (!(tun->flags & IFF_PERSIST) &&
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000547 tun->dev->reg_state == NETREG_REGISTERED)
Jason Wang4008e972012-12-13 23:53:30 +0000548 unregister_netdevice(tun->dev);
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000549 }
Eric W. Biederman140e807da2015-05-08 21:07:08 -0500550 sock_put(&tfile->sk);
Jason Wangc8d68e62012-10-31 19:46:00 +0000551 }
552}
553
554static void tun_detach(struct tun_file *tfile, bool clean)
555{
556 rtnl_lock();
557 __tun_detach(tfile, clean);
558 rtnl_unlock();
559}
560
561static void tun_detach_all(struct net_device *dev)
562{
563 struct tun_struct *tun = netdev_priv(dev);
Jason Wang4008e972012-12-13 23:53:30 +0000564 struct tun_file *tfile, *tmp;
Jason Wangc8d68e62012-10-31 19:46:00 +0000565 int i, n = tun->numqueues;
566
567 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000568 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000569 BUG_ON(!tfile);
Xi Wang9e641bd2014-05-16 15:11:48 -0700570 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530571 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wangc8d68e62012-10-31 19:46:00 +0000572 --tun->numqueues;
573 }
Jason Wang9e857222013-01-28 01:05:19 +0000574 list_for_each_entry(tfile, &tun->disabled, next) {
Xi Wang9e641bd2014-05-16 15:11:48 -0700575 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530576 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang9e857222013-01-28 01:05:19 +0000577 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000578 BUG_ON(tun->numqueues != 0);
579
580 synchronize_net();
581 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000582 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000583 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800584 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000585 sock_put(&tfile->sk);
586 }
Jason Wang4008e972012-12-13 23:53:30 +0000587 list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
588 tun_enable_queue(tfile);
Jason Wang4bfb0512013-09-05 17:53:59 +0800589 tun_queue_purge(tfile);
Jason Wang4008e972012-12-13 23:53:30 +0000590 sock_put(&tfile->sk);
591 }
592 BUG_ON(tun->numdisabled != 0);
Jason Wangdd38bd82013-01-11 16:59:34 +0000593
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200594 if (tun->flags & IFF_PERSIST)
Jason Wangdd38bd82013-01-11 16:59:34 +0000595 module_put(THIS_MODULE);
Jason Wangc8d68e62012-10-31 19:46:00 +0000596}
597
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400598static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter)
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000599{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000600 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000601 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000602
Paul Moore5dbbaf22013-01-14 07:12:19 +0000603 err = security_tun_dev_attach(tfile->socket.sk, tun->security);
604 if (err < 0)
605 goto out;
606
Eric W. Biederman38231b72009-01-20 11:02:28 +0000607 err = -EINVAL;
Jason Wang9e857222013-01-28 01:05:19 +0000608 if (rtnl_dereference(tfile->tun) && !tfile->detached)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000609 goto out;
610
611 err = -EBUSY;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200612 if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1)
Jason Wangc8d68e62012-10-31 19:46:00 +0000613 goto out;
614
615 err = -E2BIG;
Jason Wang4008e972012-12-13 23:53:30 +0000616 if (!tfile->detached &&
617 tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000618 goto out;
619
620 err = 0;
Jason Wang54f968d2012-10-31 19:45:57 +0000621
stephen hemminger92d4ea62013-12-05 20:42:58 -0800622 /* Re-attach the filter to persist device */
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400623 if (!skip_filter && (tun->filter_attached == true)) {
Jason Wang54f968d2012-10-31 19:45:57 +0000624 err = sk_attach_filter(&tun->fprog, tfile->socket.sk);
625 if (!err)
626 goto out;
627 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000628 tfile->queue_index = tun->numqueues;
Jason Wang6e914fc2012-10-31 19:45:58 +0000629 rcu_assign_pointer(tfile->tun, tun);
Jason Wangc8d68e62012-10-31 19:46:00 +0000630 rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000631 tun->numqueues++;
632
Jason Wang4008e972012-12-13 23:53:30 +0000633 if (tfile->detached)
634 tun_enable_queue(tfile);
635 else
636 sock_hold(&tfile->sk);
637
Jason Wangc8d68e62012-10-31 19:46:00 +0000638 tun_set_real_num_queues(tun);
639
Jason Wangc8d68e62012-10-31 19:46:00 +0000640 /* device is allowed to go away first, so no need to hold extra
641 * refcnt.
642 */
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000643
Eric W. Biederman38231b72009-01-20 11:02:28 +0000644out:
Eric W. Biederman38231b72009-01-20 11:02:28 +0000645 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000646}
647
Eric W. Biederman631ab462009-01-20 11:00:40 +0000648static struct tun_struct *__tun_get(struct tun_file *tfile)
649{
Jason Wang6e914fc2012-10-31 19:45:58 +0000650 struct tun_struct *tun;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000651
Jason Wang6e914fc2012-10-31 19:45:58 +0000652 rcu_read_lock();
653 tun = rcu_dereference(tfile->tun);
654 if (tun)
655 dev_hold(tun->dev);
656 rcu_read_unlock();
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000657
658 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000659}
660
661static struct tun_struct *tun_get(struct file *file)
662{
663 return __tun_get(file->private_data);
664}
665
666static void tun_put(struct tun_struct *tun)
667{
Jason Wang6e914fc2012-10-31 19:45:58 +0000668 dev_put(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000669}
670
Joe Perches6b8a66e2011-03-02 07:18:10 +0000671/* TAP filtering */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700672static void addr_hash_set(u32 *mask, const u8 *addr)
673{
674 int n = ether_crc(ETH_ALEN, addr) >> 26;
675 mask[n >> 5] |= (1 << (n & 31));
676}
677
678static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
679{
680 int n = ether_crc(ETH_ALEN, addr) >> 26;
681 return mask[n >> 5] & (1 << (n & 31));
682}
683
684static int update_filter(struct tap_filter *filter, void __user *arg)
685{
686 struct { u8 u[ETH_ALEN]; } *addr;
687 struct tun_filter uf;
688 int err, alen, n, nexact;
689
690 if (copy_from_user(&uf, arg, sizeof(uf)))
691 return -EFAULT;
692
693 if (!uf.count) {
694 /* Disabled */
695 filter->count = 0;
696 return 0;
697 }
698
699 alen = ETH_ALEN * uf.count;
700 addr = kmalloc(alen, GFP_KERNEL);
701 if (!addr)
702 return -ENOMEM;
703
704 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
705 err = -EFAULT;
706 goto done;
707 }
708
709 /* The filter is updated without holding any locks. Which is
710 * perfectly safe. We disable it first and in the worst
711 * case we'll accept a few undesired packets. */
712 filter->count = 0;
713 wmb();
714
715 /* Use first set of addresses as an exact filter */
716 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
717 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
718
719 nexact = n;
720
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800721 /* Remaining multicast addresses are hashed,
722 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700723 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800724 for (; n < uf.count; n++) {
725 if (!is_multicast_ether_addr(addr[n].u)) {
726 err = 0; /* no filter */
727 goto done;
728 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700729 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800730 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700731
732 /* For ALLMULTI just set the mask to all ones.
733 * This overrides the mask populated above. */
734 if ((uf.flags & TUN_FLT_ALLMULTI))
735 memset(filter->mask, ~0, sizeof(filter->mask));
736
737 /* Now enable the filter */
738 wmb();
739 filter->count = nexact;
740
741 /* Return the number of exact filters */
742 err = nexact;
743
744done:
745 kfree(addr);
746 return err;
747}
748
749/* Returns: 0 - drop, !=0 - accept */
750static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
751{
752 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
753 * at this point. */
754 struct ethhdr *eh = (struct ethhdr *) skb->data;
755 int i;
756
757 /* Exact match */
758 for (i = 0; i < filter->count; i++)
Joe Perches2e42e472012-05-09 17:17:46 +0000759 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700760 return 1;
761
762 /* Inexact match (multicast only) */
763 if (is_multicast_ether_addr(eh->h_dest))
764 return addr_hash_test(filter->mask, eh->h_dest);
765
766 return 0;
767}
768
769/*
770 * Checks whether the packet is accepted or not.
771 * Returns: 0 - drop, !=0 - accept
772 */
773static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
774{
775 if (!filter->count)
776 return 1;
777
778 return run_filter(filter, skb);
779}
780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781/* Network device part of the driver */
782
Jeff Garzik7282d492006-09-13 14:30:00 -0400783static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000785/* Net device detach from fd. */
786static void tun_net_uninit(struct net_device *dev)
787{
Jason Wangc8d68e62012-10-31 19:46:00 +0000788 tun_detach_all(dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000789}
790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791/* Net device open. */
792static int tun_net_open(struct net_device *dev)
793{
Jason Wangc8d68e62012-10-31 19:46:00 +0000794 netif_tx_start_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 return 0;
796}
797
798/* Net device close. */
799static int tun_net_close(struct net_device *dev)
800{
Jason Wangc8d68e62012-10-31 19:46:00 +0000801 netif_tx_stop_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 return 0;
803}
804
805/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000806static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807{
808 struct tun_struct *tun = netdev_priv(dev);
Jason Wangc8d68e62012-10-31 19:46:00 +0000809 int txq = skb->queue_mapping;
Jason Wang6e914fc2012-10-31 19:45:58 +0000810 struct tun_file *tfile;
Dominic Curranfa358642014-01-22 03:03:23 +0000811 u32 numqueues = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Jason Wang6e914fc2012-10-31 19:45:58 +0000813 rcu_read_lock();
Jason Wangc8d68e62012-10-31 19:46:00 +0000814 tfile = rcu_dereference(tun->tfiles[txq]);
Dominic Curranfa358642014-01-22 03:03:23 +0000815 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000816
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 /* Drop packet if interface is not attached */
Dominic Curranfa358642014-01-22 03:03:23 +0000818 if (txq >= numqueues)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 goto drop;
820
Dominic Curranfa358642014-01-22 03:03:23 +0000821 if (numqueues == 1) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800822 /* Select queue was not called for the skbuff, so we extract the
823 * RPS hash and save it into the flow_table here.
824 */
825 __u32 rxhash;
826
827 rxhash = skb_get_hash(skb);
828 if (rxhash) {
829 struct tun_flow_entry *e;
830 e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)],
831 rxhash);
832 if (e)
833 tun_flow_save_rps_rxhash(e, rxhash);
834 }
835 }
836
Jason Wang6e914fc2012-10-31 19:45:58 +0000837 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
838
Jason Wangc8d68e62012-10-31 19:46:00 +0000839 BUG_ON(!tfile);
840
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700841 /* Drop if the filter does not like it.
842 * This is a noop if the filter is disabled.
843 * Filter can be enabled only for the TAP devices. */
844 if (!check_filter(&tun->txflt, skb))
845 goto drop;
846
Jason Wang54f968d2012-10-31 19:45:57 +0000847 if (tfile->socket.sk->sk_filter &&
848 sk_filter(tfile->socket.sk, skb))
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000849 goto drop;
850
Rami Rosen36fe8c092012-11-25 22:07:40 +0000851 /* Limit the number of packets queued by dividing txq length with the
Jason Wangc8d68e62012-10-31 19:46:00 +0000852 * number of queues.
853 */
Dominic Curranfa358642014-01-22 03:03:23 +0000854 if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues
855 >= dev->tx_queue_len)
Michael S. Tsirkin5d097102012-12-03 10:07:14 +0000856 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857
Jason Wang7bf66302013-09-05 17:54:00 +0800858 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
859 goto drop;
860
Eric Dumazet5fcd2d82015-10-09 15:42:21 -0700861 if (skb->sk && sk_fullsock(skb->sk)) {
Richard Cochraneda29772013-07-19 19:40:10 +0200862 sock_tx_timestamp(skb->sk, &skb_shinfo(skb)->tx_flags);
863 sw_tx_timestamp(skb);
864 }
865
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000866 /* Orphan the skb - required as we might hang on to it
Jason Wang7bf66302013-09-05 17:54:00 +0800867 * for indefinite time.
868 */
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000869 skb_orphan(skb);
870
Eric Dumazetf8af75f2013-03-06 11:02:37 +0000871 nf_reset(skb);
872
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700873 /* Enqueue packet */
Jason Wang54f968d2012-10-31 19:45:57 +0000874 skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875
876 /* Notify and wake up reader process */
Jason Wang54f968d2012-10-31 19:45:57 +0000877 if (tfile->flags & TUN_FASYNC)
878 kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
Xi Wang9e641bd2014-05-16 15:11:48 -0700879 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Jason Wang6e914fc2012-10-31 19:45:58 +0000880
881 rcu_read_unlock();
Patrick McHardy6ed10652009-06-23 06:03:08 +0000882 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
884drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700885 dev->stats.tx_dropped++;
Michael S. Tsirkin149d36f2012-11-01 09:16:32 +0000886 skb_tx_error(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 kfree_skb(skb);
Jason Wang6e914fc2012-10-31 19:45:58 +0000888 rcu_read_unlock();
Jason Wangbaeabab2014-11-18 13:20:41 +0800889 return NET_XMIT_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890}
891
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700892static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700894 /*
895 * This callback is supposed to deal with mc filter in
896 * _rx_ path and has nothing to do with the _tx_ path.
897 * In rx path we always accept everything userspace gives us.
898 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899}
900
Ed Swierk4885a502007-09-16 12:21:38 -0700901#define MIN_MTU 68
902#define MAX_MTU 65535
903
904static int
905tun_net_change_mtu(struct net_device *dev, int new_mtu)
906{
907 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
908 return -EINVAL;
909 dev->mtu = new_mtu;
910 return 0;
911}
912
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000913static netdev_features_t tun_net_fix_features(struct net_device *dev,
914 netdev_features_t features)
Michał Mirosław88255372011-04-19 06:13:10 +0000915{
916 struct tun_struct *tun = netdev_priv(dev);
917
918 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
919}
Neil Hormanbebd0972011-06-15 05:25:01 +0000920#ifdef CONFIG_NET_POLL_CONTROLLER
921static void tun_poll_controller(struct net_device *dev)
922{
923 /*
924 * Tun only receives frames when:
925 * 1) the char device endpoint gets data from user space
926 * 2) the tun socket gets a sendmsg call from user space
stephen hemminger92d4ea62013-12-05 20:42:58 -0800927 * Since both of those are synchronous operations, we are guaranteed
Neil Hormanbebd0972011-06-15 05:25:01 +0000928 * never to have pending data when we poll for it
stephen hemminger92d4ea62013-12-05 20:42:58 -0800929 * so there is nothing to do here but return.
Neil Hormanbebd0972011-06-15 05:25:01 +0000930 * We need this though so netpoll recognizes us as an interface that
931 * supports polling, which enables bridge devices in virt setups to
932 * still use netconsole
933 */
934 return;
935}
936#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800937static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000938 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800939 .ndo_open = tun_net_open,
940 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800941 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800942 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000943 .ndo_fix_features = tun_net_fix_features,
Jason Wangc8d68e62012-10-31 19:46:00 +0000944 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000945#ifdef CONFIG_NET_POLL_CONTROLLER
946 .ndo_poll_controller = tun_poll_controller,
947#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800948};
949
950static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000951 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800952 .ndo_open = tun_net_open,
953 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800954 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800955 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000956 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +0000957 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800958 .ndo_set_mac_address = eth_mac_addr,
959 .ndo_validate_addr = eth_validate_addr,
Jason Wangc8d68e62012-10-31 19:46:00 +0000960 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000961#ifdef CONFIG_NET_POLL_CONTROLLER
962 .ndo_poll_controller = tun_poll_controller,
963#endif
Toshiaki Makita5e527962015-07-31 15:03:27 +0900964 .ndo_features_check = passthru_features_check,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800965};
966
Pavel Emelyanov944a1372013-06-11 17:01:08 +0400967static void tun_flow_init(struct tun_struct *tun)
Jason Wang96442e422012-10-31 19:46:02 +0000968{
969 int i;
970
Jason Wang96442e422012-10-31 19:46:02 +0000971 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
972 INIT_HLIST_HEAD(&tun->flows[i]);
973
974 tun->ageing_time = TUN_FLOW_EXPIRE;
975 setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun);
976 mod_timer(&tun->flow_gc_timer,
977 round_jiffies_up(jiffies + tun->ageing_time));
Jason Wang96442e422012-10-31 19:46:02 +0000978}
979
980static void tun_flow_uninit(struct tun_struct *tun)
981{
982 del_timer_sync(&tun->flow_gc_timer);
983 tun_flow_flush(tun);
Jason Wang96442e422012-10-31 19:46:02 +0000984}
985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986/* Initialize net device. */
987static void tun_net_init(struct net_device *dev)
988{
989 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200992 case IFF_TUN:
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800993 dev->netdev_ops = &tun_netdev_ops;
994
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 /* Point-to-Point TUN Device */
996 dev->hard_header_len = 0;
997 dev->addr_len = 0;
998 dev->mtu = 1500;
999
1000 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001001 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
1003 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
1004 break;
1005
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001006 case IFF_TAP:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -08001007 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001009 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +00001010 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
stephen hemmingera6768472012-12-10 15:16:00 +00001011 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012
Danny Kukawkaf2cedb62012-02-15 06:45:39 +00001013 eth_hw_addr_random(dev);
Brian Braunstein36226a82007-04-26 01:00:55 -07001014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
1016 break;
1017 }
1018}
1019
1020/* Character device part */
1021
1022/* Poll */
Jason Wangc8d68e62012-10-31 19:46:00 +00001023static unsigned int tun_chr_poll(struct file *file, poll_table *wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001024{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +00001025 struct tun_file *tfile = file->private_data;
1026 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +00001027 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001028 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029
1030 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +00001031 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Jason Wang54f968d2012-10-31 19:45:57 +00001033 sk = tfile->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +00001034
Joe Perches6b8a66e2011-03-02 07:18:10 +00001035 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
Xi Wang9e641bd2014-05-16 15:11:48 -07001037 poll_wait(file, sk_sleep(sk), wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001038
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001039 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 mask |= POLLIN | POLLRDNORM;
1041
Herbert Xu33dccbb2009-02-05 21:25:32 -08001042 if (sock_writeable(sk) ||
1043 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
1044 sock_writeable(sk)))
1045 mask |= POLLOUT | POLLWRNORM;
1046
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001047 if (tun->dev->reg_state != NETREG_REGISTERED)
1048 mask = POLLERR;
1049
Eric W. Biederman631ab462009-01-20 11:00:40 +00001050 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 return mask;
1052}
1053
Rusty Russellf42157c2008-08-15 15:15:10 -07001054/* prepad is the amount to reserve at front. len is length after that.
1055 * linear is a hint as to how much to copy (usually headers). */
Jason Wang54f968d2012-10-31 19:45:57 +00001056static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001057 size_t prepad, size_t len,
1058 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -07001059{
Jason Wang54f968d2012-10-31 19:45:57 +00001060 struct sock *sk = tfile->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -07001061 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001062 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -07001063
1064 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -07001065 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001066 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -07001067
Herbert Xu33dccbb2009-02-05 21:25:32 -08001068 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
Eric Dumazet28d64272013-08-08 14:38:47 -07001069 &err, 0);
Rusty Russellf42157c2008-08-15 15:15:10 -07001070 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001071 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -07001072
1073 skb_reserve(skb, prepad);
1074 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001075 skb->data_len = len - linear;
1076 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -07001077
1078 return skb;
1079}
1080
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081/* Get packet from user space buffer */
Jason Wang54f968d2012-10-31 19:45:57 +00001082static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
Al Virof5ff53b2014-06-19 15:36:49 -04001083 void *msg_control, struct iov_iter *from,
1084 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085{
Harvey Harrison09640e62009-02-01 00:45:17 -08001086 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 struct sk_buff *skb;
Al Virof5ff53b2014-06-19 15:36:49 -04001088 size_t total_len = iov_iter_count(from);
Jason Wang3dd5c332013-07-10 13:43:27 +08001089 size_t len = total_len, align = NET_SKB_PAD, linear;
Rusty Russellf43798c2008-07-03 03:48:02 -07001090 struct virtio_net_hdr gso = { 0 };
Jason Wang96f8d9e2013-11-13 14:00:39 +08001091 int good_linear;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001092 int copylen;
1093 bool zerocopy = false;
1094 int err;
Eric Dumazet49974422012-12-12 19:22:57 +00001095 u32 rxhash;
Al Virof5ff53b2014-06-19 15:36:49 -04001096 ssize_t n;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001098 if (!(tun->flags & IFF_NO_PI)) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001099 if (len < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001101 len -= sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
Al Virof5ff53b2014-06-19 15:36:49 -04001103 n = copy_from_iter(&pi, sizeof(pi), from);
1104 if (n != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 return -EFAULT;
1106 }
1107
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001108 if (tun->flags & IFF_VNET_HDR) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001109 if (len < tun->vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001110 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001111 len -= tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001112
Al Virof5ff53b2014-06-19 15:36:49 -04001113 n = copy_from_iter(&gso, sizeof(gso), from);
1114 if (n != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001115 return -EFAULT;
1116
Herbert Xu49091222009-06-08 00:20:01 -07001117 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001118 tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len))
1119 gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2);
Herbert Xu49091222009-06-08 00:20:01 -07001120
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001121 if (tun16_to_cpu(tun, gso.hdr_len) > len)
Rusty Russellf43798c2008-07-03 03:48:02 -07001122 return -EINVAL;
Herbert Xud8febb72014-11-30 18:03:31 +08001123 iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001124 }
1125
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001126 if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) {
stephen hemmingera504b862011-06-08 14:33:07 +00001127 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -07001128 if (unlikely(len < ETH_HLEN ||
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001129 (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -07001130 return -EINVAL;
1131 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001132
Jason Wang96f8d9e2013-11-13 14:00:39 +08001133 good_linear = SKB_MAX_HEAD(align);
1134
Jason Wang88529172013-07-18 10:55:15 +08001135 if (msg_control) {
Al Virof5ff53b2014-06-19 15:36:49 -04001136 struct iov_iter i = *from;
1137
Jason Wang88529172013-07-18 10:55:15 +08001138 /* There are 256 bytes to be copied in skb, so there is
1139 * enough room for skb expand head in case it is used.
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001140 * The rest of the buffer is mapped from userspace.
1141 */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001142 copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001143 if (copylen > good_linear)
1144 copylen = good_linear;
Jason Wang3dd5c332013-07-10 13:43:27 +08001145 linear = copylen;
Al Virof5ff53b2014-06-19 15:36:49 -04001146 iov_iter_advance(&i, copylen);
1147 if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS)
Jason Wang88529172013-07-18 10:55:15 +08001148 zerocopy = true;
1149 }
1150
1151 if (!zerocopy) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001152 copylen = len;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001153 if (tun16_to_cpu(tun, gso.hdr_len) > good_linear)
Jason Wang96f8d9e2013-11-13 14:00:39 +08001154 linear = good_linear;
1155 else
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001156 linear = tun16_to_cpu(tun, gso.hdr_len);
Jason Wang3dd5c332013-07-10 13:43:27 +08001157 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001158
Jason Wang3dd5c332013-07-10 13:43:27 +08001159 skb = tun_alloc_skb(tfile, align, copylen, linear, noblock);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001160 if (IS_ERR(skb)) {
1161 if (PTR_ERR(skb) != -EAGAIN)
1162 tun->dev->stats.rx_dropped++;
1163 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 }
1165
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001166 if (zerocopy)
Al Virof5ff53b2014-06-19 15:36:49 -04001167 err = zerocopy_sg_from_iter(skb, from);
Jason Wang88529172013-07-18 10:55:15 +08001168 else {
Al Virof5ff53b2014-06-19 15:36:49 -04001169 err = skb_copy_datagram_from_iter(skb, 0, from, len);
Jason Wang88529172013-07-18 10:55:15 +08001170 if (!err && msg_control) {
1171 struct ubuf_info *uarg = msg_control;
1172 uarg->callback(uarg, false);
1173 }
1174 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001175
1176 if (err) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001177 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -08001178 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -08001180 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
Rusty Russellf43798c2008-07-03 03:48:02 -07001182 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001183 if (!skb_partial_csum_set(skb, tun16_to_cpu(tun, gso.csum_start),
1184 tun16_to_cpu(tun, gso.csum_offset))) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001185 tun->dev->stats.rx_frame_errors++;
1186 kfree_skb(skb);
1187 return -EINVAL;
1188 }
Michał Mirosław88255372011-04-19 06:13:10 +00001189 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001192 case IFF_TUN:
1193 if (tun->flags & IFF_NO_PI) {
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -07001194 switch (skb->data[0] & 0xf0) {
1195 case 0x40:
1196 pi.proto = htons(ETH_P_IP);
1197 break;
1198 case 0x60:
1199 pi.proto = htons(ETH_P_IPV6);
1200 break;
1201 default:
1202 tun->dev->stats.rx_dropped++;
1203 kfree_skb(skb);
1204 return -EINVAL;
1205 }
1206 }
1207
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001208 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -07001210 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001212 case IFF_TAP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 skb->protocol = eth_type_trans(skb, tun->dev);
1214 break;
Joe Perches6403eab2011-06-03 11:51:20 +00001215 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216
Rusty Russellf43798c2008-07-03 03:48:02 -07001217 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
1218 pr_debug("GSO!\n");
1219 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
1220 case VIRTIO_NET_HDR_GSO_TCPV4:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001221 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
Rusty Russellf43798c2008-07-03 03:48:02 -07001222 break;
1223 case VIRTIO_NET_HDR_GSO_TCPV6:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001224 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
Rusty Russellf43798c2008-07-03 03:48:02 -07001225 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001226 case VIRTIO_NET_HDR_GSO_UDP:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001227 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001228 break;
Rusty Russellf43798c2008-07-03 03:48:02 -07001229 default:
1230 tun->dev->stats.rx_frame_errors++;
1231 kfree_skb(skb);
1232 return -EINVAL;
1233 }
1234
1235 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001236 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
Rusty Russellf43798c2008-07-03 03:48:02 -07001237
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001238 skb_shinfo(skb)->gso_size = tun16_to_cpu(tun, gso.gso_size);
Rusty Russellf43798c2008-07-03 03:48:02 -07001239 if (skb_shinfo(skb)->gso_size == 0) {
1240 tun->dev->stats.rx_frame_errors++;
1241 kfree_skb(skb);
1242 return -EINVAL;
1243 }
1244
1245 /* Header must be checked, and gso_segs computed. */
1246 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
1247 skb_shinfo(skb)->gso_segs = 0;
1248 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001249
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001250 /* copy skb_ubuf_info for callback when skb has no error */
1251 if (zerocopy) {
1252 skb_shinfo(skb)->destructor_arg = msg_control;
1253 skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001254 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001255 }
1256
Vlad Yasevich72f65102015-02-03 16:36:16 -05001257 skb_reset_network_header(skb);
Jason Wang40893fd2013-03-26 23:11:22 +00001258 skb_probe_transport_header(skb, 0);
Jason Wang38502af2013-03-25 20:19:56 +00001259
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001260 rxhash = skb_get_hash(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001262
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001263 tun->dev->stats.rx_packets++;
1264 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
Jason Wang9e857222013-01-28 01:05:19 +00001266 tun_flow_update(tun, rxhash, tfile);
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001267 return total_len;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001268}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269
Al Virof5ff53b2014-06-19 15:36:49 -04001270static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001272 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -08001273 struct tun_struct *tun = tun_get(file);
Jason Wang54f968d2012-10-31 19:45:57 +00001274 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001275 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
1277 if (!tun)
1278 return -EBADFD;
1279
Al Virof5ff53b2014-06-19 15:36:49 -04001280 result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001281
1282 tun_put(tun);
1283 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284}
1285
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +00001287static ssize_t tun_put_user(struct tun_struct *tun,
Jason Wang54f968d2012-10-31 19:45:57 +00001288 struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001289 struct sk_buff *skb,
Herbert Xue0b46d02014-11-07 21:22:23 +08001290 struct iov_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291{
1292 struct tun_pi pi = { 0, skb->protocol };
Herbert Xue0b46d02014-11-07 21:22:23 +08001293 ssize_t total;
Jason Wang8c847d22014-11-13 16:54:14 +08001294 int vlan_offset = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001295 int vlan_hlen = 0;
Herbert Xu2eb783c2014-11-03 04:30:14 +08001296 int vnet_hdr_sz = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001297
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001298 if (skb_vlan_tag_present(skb))
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001299 vlan_hlen = VLAN_HLEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001301 if (tun->flags & IFF_VNET_HDR)
Herbert Xu2eb783c2014-11-03 04:30:14 +08001302 vnet_hdr_sz = tun->vnet_hdr_sz;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303
Herbert Xue0b46d02014-11-07 21:22:23 +08001304 total = skb->len + vlan_hlen + vnet_hdr_sz;
1305
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001306 if (!(tun->flags & IFF_NO_PI)) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001307 if (iov_iter_count(iter) < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 return -EINVAL;
1309
Herbert Xue0b46d02014-11-07 21:22:23 +08001310 total += sizeof(pi);
1311 if (iov_iter_count(iter) < total) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 /* Packet will be striped */
1313 pi.flags |= TUN_PKT_STRIP;
1314 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001315
Herbert Xue0b46d02014-11-07 21:22:23 +08001316 if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 return -EFAULT;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001318 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
Herbert Xu2eb783c2014-11-03 04:30:14 +08001320 if (vnet_hdr_sz) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001321 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Herbert Xue0b46d02014-11-07 21:22:23 +08001322 if (iov_iter_count(iter) < vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001323 return -EINVAL;
1324
1325 if (skb_is_gso(skb)) {
1326 struct skb_shared_info *sinfo = skb_shinfo(skb);
1327
1328 /* This is a hint as to how much should be linear. */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001329 gso.hdr_len = cpu_to_tun16(tun, skb_headlen(skb));
1330 gso.gso_size = cpu_to_tun16(tun, sinfo->gso_size);
Rusty Russellf43798c2008-07-03 03:48:02 -07001331 if (sinfo->gso_type & SKB_GSO_TCPV4)
1332 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
1333 else if (sinfo->gso_type & SKB_GSO_TCPV6)
1334 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Vlad Yaseviche3e3c422015-02-03 16:36:17 -05001335 else if (sinfo->gso_type & SKB_GSO_UDP)
1336 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001337 else {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001338 pr_err("unexpected GSO type: "
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001339 "0x%x, gso_size %d, hdr_len %d\n",
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001340 sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size),
1341 tun16_to_cpu(tun, gso.hdr_len));
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001342 print_hex_dump(KERN_ERR, "tun: ",
1343 DUMP_PREFIX_NONE,
1344 16, 1, skb->head,
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001345 min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true);
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001346 WARN_ON_ONCE(1);
1347 return -EINVAL;
1348 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001349 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
1350 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
1351 } else
1352 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
1353
1354 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1355 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001356 gso.csum_start = cpu_to_tun16(tun, skb_checksum_start_offset(skb) +
1357 vlan_hlen);
1358 gso.csum_offset = cpu_to_tun16(tun, skb->csum_offset);
Jason Wang10a8d942011-06-10 00:56:17 +00001359 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
1360 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
Rusty Russellf43798c2008-07-03 03:48:02 -07001361 } /* else everything is zero */
1362
Herbert Xue0b46d02014-11-07 21:22:23 +08001363 if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001364 return -EFAULT;
Jason Wang8c847d22014-11-13 16:54:14 +08001365
1366 iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001367 }
1368
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001369 if (vlan_hlen) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001370 int ret;
Jason Wang6680ec62013-07-25 13:00:33 +08001371 struct {
1372 __be16 h_vlan_proto;
1373 __be16 h_vlan_TCI;
1374 } veth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Jason Wang6680ec62013-07-25 13:00:33 +08001376 veth.h_vlan_proto = skb->vlan_proto;
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001377 veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378
Jason Wang6680ec62013-07-25 13:00:33 +08001379 vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto);
Jason Wang6680ec62013-07-25 13:00:33 +08001380
Herbert Xue0b46d02014-11-07 21:22:23 +08001381 ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset);
1382 if (ret || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001383 goto done;
1384
Herbert Xue0b46d02014-11-07 21:22:23 +08001385 ret = copy_to_iter(&veth, sizeof(veth), iter);
1386 if (ret != sizeof(veth) || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001387 goto done;
1388 }
1389
Herbert Xue0b46d02014-11-07 21:22:23 +08001390 skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset);
Jason Wang6680ec62013-07-25 13:00:33 +08001391
1392done:
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001393 tun->dev->stats.tx_packets++;
Herbert Xue0b46d02014-11-07 21:22:23 +08001394 tun->dev->stats.tx_bytes += skb->len + vlan_hlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395
1396 return total;
1397}
1398
Jason Wang54f968d2012-10-31 19:45:57 +00001399static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
Al Viro9b067032014-11-07 13:52:07 -05001400 struct iov_iter *to,
1401 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 struct sk_buff *skb;
Al Viro9b067032014-11-07 13:52:07 -05001404 ssize_t ret;
Xi Wang9e641bd2014-05-16 15:11:48 -07001405 int peeked, err, off = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406
Rami Rosen3872baf2012-11-25 22:07:41 +00001407 tun_debug(KERN_INFO, tun, "tun_do_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
Al Viro9b067032014-11-07 13:52:07 -05001409 if (!iov_iter_count(to))
1410 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411
Xi Wang9e641bd2014-05-16 15:11:48 -07001412 if (tun->dev->reg_state != NETREG_REGISTERED)
1413 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414
Xi Wang9e641bd2014-05-16 15:11:48 -07001415 /* Read frames from queue */
1416 skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0,
1417 &peeked, &off, &err);
Herbert Xue0b46d02014-11-07 21:22:23 +08001418 if (!skb)
Alex Gartrell957f0942014-12-25 23:22:49 -08001419 return err;
Herbert Xue0b46d02014-11-07 21:22:23 +08001420
Al Viro9b067032014-11-07 13:52:07 -05001421 ret = tun_put_user(tun, tfile, skb, to);
Jason Wangf51a5e82014-12-01 16:53:15 +08001422 if (unlikely(ret < 0))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001423 kfree_skb(skb);
Jason Wangf51a5e82014-12-01 16:53:15 +08001424 else
1425 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001427 return ret;
1428}
1429
Al Viro9b067032014-11-07 13:52:07 -05001430static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001431{
1432 struct file *file = iocb->ki_filp;
1433 struct tun_file *tfile = file->private_data;
1434 struct tun_struct *tun = __tun_get(tfile);
Al Viro9b067032014-11-07 13:52:07 -05001435 ssize_t len = iov_iter_count(to), ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001436
1437 if (!tun)
1438 return -EBADFD;
Al Viro9b067032014-11-07 13:52:07 -05001439 ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK);
David S. Miller42404c02013-12-10 22:05:45 -05001440 ret = min_t(ssize_t, ret, len);
Zhi Yong Wud0b7da8a2013-12-06 14:16:51 +08001441 if (ret > 0)
1442 iocb->ki_pos = ret;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001443 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 return ret;
1445}
1446
Jason Wang96442e422012-10-31 19:46:02 +00001447static void tun_free_netdev(struct net_device *dev)
1448{
1449 struct tun_struct *tun = netdev_priv(dev);
1450
Jason Wang4008e972012-12-13 23:53:30 +00001451 BUG_ON(!(list_empty(&tun->disabled)));
Jason Wang96442e422012-10-31 19:46:02 +00001452 tun_flow_uninit(tun);
Paul Moore5dbbaf22013-01-14 07:12:19 +00001453 security_tun_dev_free_security(tun->security);
Jason Wang96442e422012-10-31 19:46:02 +00001454 free_netdev(dev);
1455}
1456
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457static void tun_setup(struct net_device *dev)
1458{
1459 struct tun_struct *tun = netdev_priv(dev);
1460
Eric W. Biederman0625c882012-02-07 16:48:55 -08001461 tun->owner = INVALID_UID;
1462 tun->group = INVALID_GID;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 dev->ethtool_ops = &tun_ethtool_ops;
Jason Wang96442e422012-10-31 19:46:02 +00001465 dev->destructor = tun_free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466}
1467
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001468/* Trivial set of netlink ops to allow deleting tun or tap
1469 * device with netlink.
1470 */
1471static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
1472{
1473 return -EINVAL;
1474}
1475
1476static struct rtnl_link_ops tun_link_ops __read_mostly = {
1477 .kind = DRV_NAME,
1478 .priv_size = sizeof(struct tun_struct),
1479 .setup = tun_setup,
1480 .validate = tun_validate,
1481};
1482
Herbert Xu33dccbb2009-02-05 21:25:32 -08001483static void tun_sock_write_space(struct sock *sk)
1484{
Jason Wang54f968d2012-10-31 19:45:57 +00001485 struct tun_file *tfile;
Eric Dumazet43815482010-04-29 11:01:49 +00001486 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001487
1488 if (!sock_writeable(sk))
1489 return;
1490
Herbert Xu33dccbb2009-02-05 21:25:32 -08001491 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
1492 return;
1493
Eric Dumazet43815482010-04-29 11:01:49 +00001494 wqueue = sk_sleep(sk);
1495 if (wqueue && waitqueue_active(wqueue))
1496 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001497 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -07001498
Jason Wang54f968d2012-10-31 19:45:57 +00001499 tfile = container_of(sk, struct tun_file, sk);
1500 kill_fasync(&tfile->fasync, SIGIO, POLL_OUT);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001501}
1502
Ying Xue1b784142015-03-02 15:37:48 +08001503static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001504{
Jason Wang54f968d2012-10-31 19:45:57 +00001505 int ret;
1506 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1507 struct tun_struct *tun = __tun_get(tfile);
1508
1509 if (!tun)
1510 return -EBADFD;
Al Virof5ff53b2014-06-19 15:36:49 -04001511
Al Viroc0371da2014-11-24 10:42:55 -05001512 ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04001513 m->msg_flags & MSG_DONTWAIT);
Jason Wang54f968d2012-10-31 19:45:57 +00001514 tun_put(tun);
1515 return ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001516}
1517
Ying Xue1b784142015-03-02 15:37:48 +08001518static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001519 int flags)
1520{
Jason Wang54f968d2012-10-31 19:45:57 +00001521 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1522 struct tun_struct *tun = __tun_get(tfile);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001523 int ret;
Jason Wang54f968d2012-10-31 19:45:57 +00001524
1525 if (!tun)
1526 return -EBADFD;
1527
Richard Cochraneda29772013-07-19 19:40:10 +02001528 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) {
Gao feng3811ae72013-04-24 21:59:23 +00001529 ret = -EINVAL;
1530 goto out;
1531 }
Richard Cochraneda29772013-07-19 19:40:10 +02001532 if (flags & MSG_ERRQUEUE) {
1533 ret = sock_recv_errqueue(sock->sk, m, total_len,
1534 SOL_PACKET, TUN_TX_TIMESTAMP);
1535 goto out;
1536 }
Al Viroc0371da2014-11-24 10:42:55 -05001537 ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT);
Alex Gartrell87897932014-12-25 23:05:03 -08001538 if (ret > (ssize_t)total_len) {
David S. Miller42404c02013-12-10 22:05:45 -05001539 m->msg_flags |= MSG_TRUNC;
1540 ret = flags & MSG_TRUNC ? ret : total_len;
1541 }
Gao feng3811ae72013-04-24 21:59:23 +00001542out:
Jason Wang54f968d2012-10-31 19:45:57 +00001543 tun_put(tun);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001544 return ret;
1545}
1546
1547/* Ops structure to mimic raw sockets with tun */
1548static const struct proto_ops tun_socket_ops = {
1549 .sendmsg = tun_sendmsg,
1550 .recvmsg = tun_recvmsg,
1551};
1552
Herbert Xu33dccbb2009-02-05 21:25:32 -08001553static struct proto tun_proto = {
1554 .name = "tun",
1555 .owner = THIS_MODULE,
Jason Wang54f968d2012-10-31 19:45:57 +00001556 .obj_size = sizeof(struct tun_file),
Herbert Xu33dccbb2009-02-05 21:25:32 -08001557};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001558
David Woodhouse980c9e82009-05-09 22:54:21 -07001559static int tun_flags(struct tun_struct *tun)
1560{
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001561 return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP);
David Woodhouse980c9e82009-05-09 22:54:21 -07001562}
1563
1564static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1565 char *buf)
1566{
1567 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1568 return sprintf(buf, "0x%x\n", tun_flags(tun));
1569}
1570
1571static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1572 char *buf)
1573{
1574 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001575 return uid_valid(tun->owner)?
1576 sprintf(buf, "%u\n",
1577 from_kuid_munged(current_user_ns(), tun->owner)):
1578 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001579}
1580
1581static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1582 char *buf)
1583{
1584 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001585 return gid_valid(tun->group) ?
1586 sprintf(buf, "%u\n",
1587 from_kgid_munged(current_user_ns(), tun->group)):
1588 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001589}
1590
1591static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1592static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1593static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1594
Takashi Iwaic4d33e22015-02-04 14:37:34 +01001595static struct attribute *tun_dev_attrs[] = {
1596 &dev_attr_tun_flags.attr,
1597 &dev_attr_owner.attr,
1598 &dev_attr_group.attr,
1599 NULL
1600};
1601
1602static const struct attribute_group tun_attr_group = {
1603 .attrs = tun_dev_attrs
1604};
1605
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001606static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607{
1608 struct tun_struct *tun;
Jason Wang54f968d2012-10-31 19:45:57 +00001609 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 struct net_device *dev;
1611 int err;
1612
Jason Wang7c0c3b12013-01-11 16:59:33 +00001613 if (tfile->detached)
1614 return -EINVAL;
1615
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001616 dev = __dev_get_by_name(net, ifr->ifr_name);
1617 if (dev) {
David Woodhousef85ba782009-04-27 03:23:54 -07001618 if (ifr->ifr_flags & IFF_TUN_EXCL)
1619 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001620 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1621 tun = netdev_priv(dev);
1622 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1623 tun = netdev_priv(dev);
1624 else
1625 return -EINVAL;
1626
Jason Wang8e6d91a2013-05-28 18:32:11 +00001627 if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) !=
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001628 !!(tun->flags & IFF_MULTI_QUEUE))
Jason Wang8e6d91a2013-05-28 18:32:11 +00001629 return -EINVAL;
1630
Jason Wangcde8b152012-10-31 19:46:01 +00001631 if (tun_not_capable(tun))
Paul Moore2b980db2009-08-28 18:12:43 -04001632 return -EPERM;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001633 err = security_tun_dev_open(tun->security);
Paul Moore2b980db2009-08-28 18:12:43 -04001634 if (err < 0)
1635 return err;
1636
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001637 err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001638 if (err < 0)
1639 return err;
Jason Wang4008e972012-12-13 23:53:30 +00001640
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001641 if (tun->flags & IFF_MULTI_QUEUE &&
Jason Wange8dbad62013-04-22 20:40:39 +00001642 (tun->numqueues + tun->numdisabled > 1)) {
1643 /* One or more queue has already been attached, no need
1644 * to initialize the device again.
1645 */
1646 return 0;
1647 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001648 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 else {
1650 char *name;
1651 unsigned long flags = 0;
Jason Wangedfb6a12013-01-23 03:59:12 +00001652 int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
1653 MAX_TAP_QUEUES : 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654
Eric W. Biedermanc260b772012-11-18 21:34:11 +00001655 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001656 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001657 err = security_tun_dev_create();
1658 if (err < 0)
1659 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001660
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 /* Set dev type */
1662 if (ifr->ifr_flags & IFF_TUN) {
1663 /* TUN device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001664 flags |= IFF_TUN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 name = "tun%d";
1666 } else if (ifr->ifr_flags & IFF_TAP) {
1667 /* TAP device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001668 flags |= IFF_TAP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001670 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001671 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001672
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 if (*ifr->ifr_name)
1674 name = ifr->ifr_name;
1675
Jason Wangc8d68e62012-10-31 19:46:00 +00001676 dev = alloc_netdev_mqs(sizeof(struct tun_struct), name,
Tom Gundersenc835a672014-07-14 16:37:24 +02001677 NET_NAME_UNKNOWN, tun_setup, queues,
1678 queues);
Jason Wangedfb6a12013-01-23 03:59:12 +00001679
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 if (!dev)
1681 return -ENOMEM;
1682
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001683 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001684 dev->rtnl_link_ops = &tun_link_ops;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001685 dev->ifindex = tfile->ifindex;
Takashi Iwaic4d33e22015-02-04 14:37:34 +01001686 dev->sysfs_groups[0] = &tun_attr_group;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001687
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 tun = netdev_priv(dev);
1689 tun->dev = dev;
1690 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001691 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001692 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693
Jason Wang54f968d2012-10-31 19:45:57 +00001694 tun->filter_attached = false;
1695 tun->sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001696
Jason Wang96442e422012-10-31 19:46:02 +00001697 spin_lock_init(&tun->lock);
1698
Paul Moore5dbbaf22013-01-14 07:12:19 +00001699 err = security_tun_dev_alloc_security(&tun->security);
1700 if (err < 0)
1701 goto err_free_dev;
Paul Moore2b980db2009-08-28 18:12:43 -04001702
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 tun_net_init(dev);
Pavel Emelyanov944a1372013-06-11 17:01:08 +04001704 tun_flow_init(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001705
Michał Mirosław88255372011-04-19 06:13:10 +00001706 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
Jason Wang6680ec62013-07-25 13:00:33 +08001707 TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
1708 NETIF_F_HW_VLAN_STAG_TX;
Michał Mirosław88255372011-04-19 06:13:10 +00001709 dev->features = dev->hw_features;
Fernando Luis Vazquez Cao6671b222014-02-18 21:20:09 +09001710 dev->vlan_features = dev->features &
1711 ~(NETIF_F_HW_VLAN_CTAG_TX |
1712 NETIF_F_HW_VLAN_STAG_TX);
Michał Mirosław88255372011-04-19 06:13:10 +00001713
Jason Wang4008e972012-12-13 23:53:30 +00001714 INIT_LIST_HEAD(&tun->disabled);
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001715 err = tun_attach(tun, file, false);
Jason Wangeb0fb362012-12-02 17:19:45 +00001716 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001717 goto err_free_flow;
Jason Wangeb0fb362012-12-02 17:19:45 +00001718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 err = register_netdevice(tun->dev);
1720 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001721 goto err_detach;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 }
1723
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +00001724 netif_carrier_on(tun->dev);
1725
Joe Perches6b8a66e2011-03-02 07:18:10 +00001726 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001728 tun->flags = (tun->flags & ~TUN_FEATURES) |
1729 (ifr->ifr_flags & TUN_FEATURES);
Jason Wangc8d68e62012-10-31 19:46:00 +00001730
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001731 /* Make sure persistent devices do not get stuck in
1732 * xoff state.
1733 */
1734 if (netif_running(tun->dev))
Jason Wangc8d68e62012-10-31 19:46:00 +00001735 netif_tx_wake_all_queues(tun->dev);
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001736
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 strcpy(ifr->ifr_name, tun->dev->name);
1738 return 0;
1739
Jason Wang662ca432013-09-11 18:09:48 +08001740err_detach:
1741 tun_detach_all(dev);
1742err_free_flow:
1743 tun_flow_uninit(tun);
1744 security_tun_dev_free_security(tun->security);
1745err_free_dev:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 return err;
1748}
1749
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001750static void tun_get_iff(struct net *net, struct tun_struct *tun,
Herbert Xu876bfd42009-08-06 14:22:44 +00001751 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001752{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001753 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001754
1755 strcpy(ifr->ifr_name, tun->dev->name);
1756
David Woodhouse980c9e82009-05-09 22:54:21 -07001757 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001758
Mark McLoughline3b99552008-08-15 15:09:56 -07001759}
1760
Rusty Russell5228ddc2008-07-03 03:46:16 -07001761/* This is like a cut-down ethtool ops, except done via tun fd so no
1762 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001763static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001764{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001765 netdev_features_t features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001766
1767 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001768 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001769 arg &= ~TUN_F_CSUM;
1770
1771 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1772 if (arg & TUN_F_TSO_ECN) {
1773 features |= NETIF_F_TSO_ECN;
1774 arg &= ~TUN_F_TSO_ECN;
1775 }
1776 if (arg & TUN_F_TSO4)
1777 features |= NETIF_F_TSO;
1778 if (arg & TUN_F_TSO6)
1779 features |= NETIF_F_TSO6;
1780 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1781 }
Vlad Yaseviche3e3c422015-02-03 16:36:17 -05001782
1783 if (arg & TUN_F_UFO) {
1784 features |= NETIF_F_UFO;
1785 arg &= ~TUN_F_UFO;
1786 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001787 }
1788
1789 /* This gives the user a way to test for new features in future by
1790 * trying to set them. */
1791 if (arg)
1792 return -EINVAL;
1793
Michał Mirosław88255372011-04-19 06:13:10 +00001794 tun->set_features = features;
1795 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001796
1797 return 0;
1798}
1799
Jason Wangc8d68e62012-10-31 19:46:00 +00001800static void tun_detach_filter(struct tun_struct *tun, int n)
1801{
1802 int i;
1803 struct tun_file *tfile;
1804
1805 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001806 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001807 sk_detach_filter(tfile->socket.sk);
1808 }
1809
1810 tun->filter_attached = false;
1811}
1812
1813static int tun_attach_filter(struct tun_struct *tun)
1814{
1815 int i, ret = 0;
1816 struct tun_file *tfile;
1817
1818 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001819 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001820 ret = sk_attach_filter(&tun->fprog, tfile->socket.sk);
1821 if (ret) {
1822 tun_detach_filter(tun, i);
1823 return ret;
1824 }
1825 }
1826
1827 tun->filter_attached = true;
1828 return ret;
1829}
1830
1831static void tun_set_sndbuf(struct tun_struct *tun)
1832{
1833 struct tun_file *tfile;
1834 int i;
1835
1836 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001837 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001838 tfile->socket.sk->sk_sndbuf = tun->sndbuf;
1839 }
1840}
1841
Jason Wangcde8b152012-10-31 19:46:01 +00001842static int tun_set_queue(struct file *file, struct ifreq *ifr)
1843{
1844 struct tun_file *tfile = file->private_data;
1845 struct tun_struct *tun;
Jason Wangcde8b152012-10-31 19:46:01 +00001846 int ret = 0;
1847
1848 rtnl_lock();
1849
1850 if (ifr->ifr_flags & IFF_ATTACH_QUEUE) {
Jason Wang4008e972012-12-13 23:53:30 +00001851 tun = tfile->detached;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001852 if (!tun) {
Jason Wangcde8b152012-10-31 19:46:01 +00001853 ret = -EINVAL;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001854 goto unlock;
1855 }
1856 ret = security_tun_dev_attach_queue(tun->security);
1857 if (ret < 0)
1858 goto unlock;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001859 ret = tun_attach(tun, file, false);
Jason Wang4008e972012-12-13 23:53:30 +00001860 } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001861 tun = rtnl_dereference(tfile->tun);
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001862 if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached)
Jason Wang4008e972012-12-13 23:53:30 +00001863 ret = -EINVAL;
1864 else
1865 __tun_detach(tfile, false);
1866 } else
Jason Wangcde8b152012-10-31 19:46:01 +00001867 ret = -EINVAL;
1868
Paul Moore5dbbaf22013-01-14 07:12:19 +00001869unlock:
Jason Wangcde8b152012-10-31 19:46:01 +00001870 rtnl_unlock();
1871 return ret;
1872}
1873
Arnd Bergmann50857e22009-11-06 22:52:32 -08001874static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1875 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001877 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001878 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 void __user* argp = (void __user*)arg;
1880 struct ifreq ifr;
Eric W. Biederman0625c882012-02-07 16:48:55 -08001881 kuid_t owner;
1882 kgid_t group;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001883 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001884 int vnet_hdr_sz;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001885 unsigned int ifindex;
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02001886 int le;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001887 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888
Jason Wangcde8b152012-10-31 19:46:01 +00001889 if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) {
Arnd Bergmann50857e22009-11-06 22:52:32 -08001890 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 return -EFAULT;
David S. Miller8bbb1812012-07-30 14:52:48 -07001892 } else {
Mathias Krausea117dac2012-07-29 19:45:14 +00001893 memset(&ifr, 0, sizeof(ifr));
David S. Miller8bbb1812012-07-30 14:52:48 -07001894 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001895 if (cmd == TUNGETFEATURES) {
1896 /* Currently this just means: "what IFF flags are valid?".
1897 * This is needed because we never checked for invalid flags on
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001898 * TUNSETIFF.
1899 */
1900 return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES,
Eric W. Biederman631ab462009-01-20 11:00:40 +00001901 (unsigned int __user*)argp);
Jason Wangcde8b152012-10-31 19:46:01 +00001902 } else if (cmd == TUNSETQUEUE)
1903 return tun_set_queue(file, &ifr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001904
Jason Wangc8d68e62012-10-31 19:46:00 +00001905 ret = 0;
Herbert Xu876bfd42009-08-06 14:22:44 +00001906 rtnl_lock();
1907
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001908 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1911
Eric W. Biederman140e807da2015-05-08 21:07:08 -05001912 ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Herbert Xu876bfd42009-08-06 14:22:44 +00001914 if (ret)
1915 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916
Arnd Bergmann50857e22009-11-06 22:52:32 -08001917 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001918 ret = -EFAULT;
1919 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 }
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001921 if (cmd == TUNSETIFINDEX) {
1922 ret = -EPERM;
1923 if (tun)
1924 goto unlock;
1925
1926 ret = -EFAULT;
1927 if (copy_from_user(&ifindex, argp, sizeof(ifindex)))
1928 goto unlock;
1929
1930 ret = 0;
1931 tfile->ifindex = ifindex;
1932 goto unlock;
1933 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934
Herbert Xu876bfd42009-08-06 14:22:44 +00001935 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001937 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938
Jason Wang1e588332012-10-31 19:45:56 +00001939 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940
Eric W. Biederman631ab462009-01-20 11:00:40 +00001941 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001943 case TUNGETIFF:
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001944 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001945
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001946 if (tfile->detached)
1947 ifr.ifr_flags |= IFF_DETACH_QUEUE;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001948 if (!tfile->socket.sk->sk_filter)
1949 ifr.ifr_flags |= IFF_NOFILTER;
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001950
Arnd Bergmann50857e22009-11-06 22:52:32 -08001951 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001952 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001953 break;
1954
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 case TUNSETNOCSUM:
1956 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
Michał Mirosław88255372011-04-19 06:13:10 +00001958 /* [unimplemented] */
1959 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00001960 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 break;
1962
1963 case TUNSETPERSIST:
Jason Wang54f968d2012-10-31 19:45:57 +00001964 /* Disable/Enable persist mode. Keep an extra reference to the
1965 * module to prevent the module being unprobed.
1966 */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001967 if (arg && !(tun->flags & IFF_PERSIST)) {
1968 tun->flags |= IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001969 __module_get(THIS_MODULE);
Jason Wangdd38bd82013-01-11 16:59:34 +00001970 }
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001971 if (!arg && (tun->flags & IFF_PERSIST)) {
1972 tun->flags &= ~IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001973 module_put(THIS_MODULE);
1974 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975
Joe Perches6b8a66e2011-03-02 07:18:10 +00001976 tun_debug(KERN_INFO, tun, "persist %s\n",
1977 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 break;
1979
1980 case TUNSETOWNER:
1981 /* Set owner of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001982 owner = make_kuid(current_user_ns(), arg);
1983 if (!uid_valid(owner)) {
1984 ret = -EINVAL;
1985 break;
1986 }
1987 tun->owner = owner;
Jason Wang1e588332012-10-31 19:45:56 +00001988 tun_debug(KERN_INFO, tun, "owner set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08001989 from_kuid(&init_user_ns, tun->owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 break;
1991
Guido Guenther8c644622007-07-02 22:50:25 -07001992 case TUNSETGROUP:
1993 /* Set group of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001994 group = make_kgid(current_user_ns(), arg);
1995 if (!gid_valid(group)) {
1996 ret = -EINVAL;
1997 break;
1998 }
1999 tun->group = group;
Jason Wang1e588332012-10-31 19:45:56 +00002000 tun_debug(KERN_INFO, tun, "group set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08002001 from_kgid(&init_user_ns, tun->group));
Guido Guenther8c644622007-07-02 22:50:25 -07002002 break;
2003
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002004 case TUNSETLINK:
2005 /* Only allow setting the type when the interface is down */
2006 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002007 tun_debug(KERN_INFO, tun,
2008 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07002009 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002010 } else {
2011 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00002012 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
2013 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07002014 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002015 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00002016 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002017
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018#ifdef TUN_DEBUG
2019 case TUNSETDEBUG:
2020 tun->debug = arg;
2021 break;
2022#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07002023 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00002024 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002025 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07002026
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002027 case TUNSETTXFILTER:
2028 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00002029 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002030 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Eric W. Biederman631ab462009-01-20 11:00:40 +00002031 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07002032 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002033 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
2035 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04002036 /* Get hw address */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002037 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
2038 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08002039 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00002040 ret = -EFAULT;
2041 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042
2043 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002044 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00002045 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
2046 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08002047
Kim B. Heino40102372008-02-29 12:26:21 -08002048 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002049 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002050
2051 case TUNGETSNDBUF:
Jason Wang54f968d2012-10-31 19:45:57 +00002052 sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002053 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
2054 ret = -EFAULT;
2055 break;
2056
2057 case TUNSETSNDBUF:
2058 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
2059 ret = -EFAULT;
2060 break;
2061 }
2062
Jason Wangc8d68e62012-10-31 19:46:00 +00002063 tun->sndbuf = sndbuf;
2064 tun_set_sndbuf(tun);
Herbert Xu33dccbb2009-02-05 21:25:32 -08002065 break;
2066
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02002067 case TUNGETVNETHDRSZ:
2068 vnet_hdr_sz = tun->vnet_hdr_sz;
2069 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
2070 ret = -EFAULT;
2071 break;
2072
2073 case TUNSETVNETHDRSZ:
2074 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
2075 ret = -EFAULT;
2076 break;
2077 }
2078 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
2079 ret = -EINVAL;
2080 break;
2081 }
2082
2083 tun->vnet_hdr_sz = vnet_hdr_sz;
2084 break;
2085
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02002086 case TUNGETVNETLE:
2087 le = !!(tun->flags & TUN_VNET_LE);
2088 if (put_user(le, (int __user *)argp))
2089 ret = -EFAULT;
2090 break;
2091
2092 case TUNSETVNETLE:
2093 if (get_user(le, (int __user *)argp)) {
2094 ret = -EFAULT;
2095 break;
2096 }
2097 if (le)
2098 tun->flags |= TUN_VNET_LE;
2099 else
2100 tun->flags &= ~TUN_VNET_LE;
2101 break;
2102
Greg Kurz8b8e6582015-04-24 14:50:36 +02002103 case TUNGETVNETBE:
2104 ret = tun_get_vnet_be(tun, argp);
2105 break;
2106
2107 case TUNSETVNETBE:
2108 ret = tun_set_vnet_be(tun, argp);
2109 break;
2110
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002111 case TUNATTACHFILTER:
2112 /* Can be set only for TAPs */
2113 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002114 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002115 break;
2116 ret = -EFAULT;
Jason Wang54f968d2012-10-31 19:45:57 +00002117 if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog)))
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002118 break;
2119
Jason Wangc8d68e62012-10-31 19:46:00 +00002120 ret = tun_attach_filter(tun);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002121 break;
2122
2123 case TUNDETACHFILTER:
2124 /* Can be set only for TAPs */
2125 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002126 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002127 break;
Jason Wangc8d68e62012-10-31 19:46:00 +00002128 ret = 0;
2129 tun_detach_filter(tun, tun->numqueues);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002130 break;
2131
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002132 case TUNGETFILTER:
2133 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002134 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002135 break;
2136 ret = -EFAULT;
2137 if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog)))
2138 break;
2139 ret = 0;
2140 break;
2141
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00002143 ret = -EINVAL;
2144 break;
Joe Perchesee289b62010-05-17 22:47:34 -07002145 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146
Herbert Xu876bfd42009-08-06 14:22:44 +00002147unlock:
2148 rtnl_unlock();
2149 if (tun)
2150 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002151 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152}
2153
Arnd Bergmann50857e22009-11-06 22:52:32 -08002154static long tun_chr_ioctl(struct file *file,
2155 unsigned int cmd, unsigned long arg)
2156{
2157 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
2158}
2159
2160#ifdef CONFIG_COMPAT
2161static long tun_chr_compat_ioctl(struct file *file,
2162 unsigned int cmd, unsigned long arg)
2163{
2164 switch (cmd) {
2165 case TUNSETIFF:
2166 case TUNGETIFF:
2167 case TUNSETTXFILTER:
2168 case TUNGETSNDBUF:
2169 case TUNSETSNDBUF:
2170 case SIOCGIFHWADDR:
2171 case SIOCSIFHWADDR:
2172 arg = (unsigned long)compat_ptr(arg);
2173 break;
2174 default:
2175 arg = (compat_ulong_t)arg;
2176 break;
2177 }
2178
2179 /*
2180 * compat_ifreq is shorter than ifreq, so we must not access beyond
2181 * the end of that structure. All fields that are used in this
2182 * driver are compatible though, we don't need to convert the
2183 * contents.
2184 */
2185 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
2186}
2187#endif /* CONFIG_COMPAT */
2188
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189static int tun_chr_fasync(int fd, struct file *file, int on)
2190{
Jason Wang54f968d2012-10-31 19:45:57 +00002191 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 int ret;
2193
Jason Wang54f968d2012-10-31 19:45:57 +00002194 if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06002195 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002196
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197 if (on) {
Jeff Laytone0b93ed2014-08-22 11:27:32 -04002198 __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Jason Wang54f968d2012-10-31 19:45:57 +00002199 tfile->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002200 } else
Jason Wang54f968d2012-10-31 19:45:57 +00002201 tfile->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06002202 ret = 0;
2203out:
Jonathan Corbet9d319522008-06-19 15:50:37 -06002204 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205}
2206
2207static int tun_chr_open(struct inode *inode, struct file * file)
2208{
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002209 struct net *net = current->nsproxy->net_ns;
Eric W. Biederman631ab462009-01-20 11:00:40 +00002210 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07002211
Joe Perches6b8a66e2011-03-02 07:18:10 +00002212 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00002213
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002214 tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05002215 &tun_proto, 0);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002216 if (!tfile)
2217 return -ENOMEM;
Monam Agarwalc9566742014-03-24 00:02:32 +05302218 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang54f968d2012-10-31 19:45:57 +00002219 tfile->flags = 0;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002220 tfile->ifindex = 0;
Jason Wang54f968d2012-10-31 19:45:57 +00002221
Jason Wang54f968d2012-10-31 19:45:57 +00002222 init_waitqueue_head(&tfile->wq.wait);
Xi Wang9e641bd2014-05-16 15:11:48 -07002223 RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq);
Jason Wang54f968d2012-10-31 19:45:57 +00002224
2225 tfile->socket.file = file;
2226 tfile->socket.ops = &tun_socket_ops;
2227
2228 sock_init_data(&tfile->socket, &tfile->sk);
Jason Wang54f968d2012-10-31 19:45:57 +00002229
2230 tfile->sk.sk_write_space = tun_sock_write_space;
2231 tfile->sk.sk_sndbuf = INT_MAX;
2232
Eric W. Biederman631ab462009-01-20 11:00:40 +00002233 file->private_data = tfile;
Jason Wang4008e972012-12-13 23:53:30 +00002234 INIT_LIST_HEAD(&tfile->next);
Jason Wang54f968d2012-10-31 19:45:57 +00002235
Jason Wang19a6afb2013-06-08 14:17:41 +08002236 sock_set_flag(&tfile->sk, SOCK_ZEROCOPY);
2237
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238 return 0;
2239}
2240
2241static int tun_chr_close(struct inode *inode, struct file *file)
2242{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002243 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244
Jason Wangc8d68e62012-10-31 19:46:00 +00002245 tun_detach(tfile, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246
2247 return 0;
2248}
2249
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002250#ifdef CONFIG_PROC_FS
Joe Perchesa3816ab2014-09-29 16:08:25 -07002251static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f)
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002252{
2253 struct tun_struct *tun;
2254 struct ifreq ifr;
2255
2256 memset(&ifr, 0, sizeof(ifr));
2257
2258 rtnl_lock();
2259 tun = tun_get(f);
2260 if (tun)
2261 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
2262 rtnl_unlock();
2263
2264 if (tun)
2265 tun_put(tun);
2266
Joe Perchesa3816ab2014-09-29 16:08:25 -07002267 seq_printf(m, "iff:\t%s\n", ifr.ifr_name);
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002268}
2269#endif
2270
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08002271static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002272 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002273 .llseek = no_llseek,
Al Viro9b067032014-11-07 13:52:07 -05002274 .read_iter = tun_chr_read_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04002275 .write_iter = tun_chr_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08002277 .unlocked_ioctl = tun_chr_ioctl,
2278#ifdef CONFIG_COMPAT
2279 .compat_ioctl = tun_chr_compat_ioctl,
2280#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281 .open = tun_chr_open,
2282 .release = tun_chr_close,
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002283 .fasync = tun_chr_fasync,
2284#ifdef CONFIG_PROC_FS
2285 .show_fdinfo = tun_chr_show_fdinfo,
2286#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287};
2288
2289static struct miscdevice tun_miscdev = {
2290 .minor = TUN_MINOR,
2291 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02002292 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294};
2295
2296/* ethtool interface */
2297
2298static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
2299{
2300 cmd->supported = 0;
2301 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00002302 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303 cmd->duplex = DUPLEX_FULL;
2304 cmd->port = PORT_TP;
2305 cmd->phy_address = 0;
2306 cmd->transceiver = XCVR_INTERNAL;
2307 cmd->autoneg = AUTONEG_DISABLE;
2308 cmd->maxtxpkt = 0;
2309 cmd->maxrxpkt = 0;
2310 return 0;
2311}
2312
2313static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
2314{
2315 struct tun_struct *tun = netdev_priv(dev);
2316
Rick Jones33a5ba12011-11-15 14:59:53 +00002317 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
2318 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319
2320 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002321 case IFF_TUN:
Rick Jones33a5ba12011-11-15 14:59:53 +00002322 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002324 case IFF_TAP:
Rick Jones33a5ba12011-11-15 14:59:53 +00002325 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 break;
2327 }
2328}
2329
2330static u32 tun_get_msglevel(struct net_device *dev)
2331{
2332#ifdef TUN_DEBUG
2333 struct tun_struct *tun = netdev_priv(dev);
2334 return tun->debug;
2335#else
2336 return -EOPNOTSUPP;
2337#endif
2338}
2339
2340static void tun_set_msglevel(struct net_device *dev, u32 value)
2341{
2342#ifdef TUN_DEBUG
2343 struct tun_struct *tun = netdev_priv(dev);
2344 tun->debug = value;
2345#endif
2346}
2347
Jeff Garzik7282d492006-09-13 14:30:00 -04002348static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002349 .get_settings = tun_get_settings,
2350 .get_drvinfo = tun_get_drvinfo,
2351 .get_msglevel = tun_get_msglevel,
2352 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00002353 .get_link = ethtool_op_get_link,
Richard Cochraneda29772013-07-19 19:40:10 +02002354 .get_ts_info = ethtool_op_get_ts_info,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355};
2356
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002357
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358static int __init tun_init(void)
2359{
2360 int ret = 0;
2361
Joe Perches6b8a66e2011-03-02 07:18:10 +00002362 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
2363 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002365 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002366 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002367 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002368 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002369 }
2370
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002372 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002373 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002374 goto err_misc;
2375 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002376 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002377err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002378 rtnl_link_unregister(&tun_link_ops);
2379err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380 return ret;
2381}
2382
2383static void tun_cleanup(void)
2384{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002385 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002386 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387}
2388
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002389/* Get an underlying socket object from tun file. Returns error unless file is
2390 * attached to a device. The returned object works like a packet socket, it
2391 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
2392 * holding a reference to the file for as long as the socket is in use. */
2393struct socket *tun_get_socket(struct file *file)
2394{
Jason Wang6e914fc2012-10-31 19:45:58 +00002395 struct tun_file *tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002396 if (file->f_op != &tun_fops)
2397 return ERR_PTR(-EINVAL);
Jason Wang6e914fc2012-10-31 19:45:58 +00002398 tfile = file->private_data;
2399 if (!tfile)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002400 return ERR_PTR(-EBADFD);
Jason Wang54f968d2012-10-31 19:45:57 +00002401 return &tfile->socket;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002402}
2403EXPORT_SYMBOL_GPL(tun_get_socket);
2404
Linus Torvalds1da177e2005-04-16 15:20:36 -07002405module_init(tun_init);
2406module_exit(tun_cleanup);
2407MODULE_DESCRIPTION(DRV_DESCRIPTION);
2408MODULE_AUTHOR(DRV_COPYRIGHT);
2409MODULE_LICENSE("GPL");
2410MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02002411MODULE_ALIAS("devname:net/tun");