blob: 8c8dc16839a79473e976d9de2132cc6df911e418 [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>
Ben Hutchings5188cd42014-10-30 18:27:17 +000068#include <net/ipv6.h>
Eric W. Biederman881d9662007-09-17 11:56:21 -070069#include <net/net_namespace.h>
Pavel Emelyanov79d17602008-04-16 00:40:46 -070070#include <net/netns/generic.h>
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -080071#include <net/rtnetlink.h>
Herbert Xu33dccbb2009-02-05 21:25:32 -080072#include <net/sock.h>
Masatake YAMATO93e14b62014-01-29 16:43:31 +090073#include <linux/seq_file.h>
Herbert Xue0b46d02014-11-07 21:22:23 +080074#include <linux/uio.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <asm/uaccess.h>
77
Rusty Russell14daa022008-04-12 18:48:58 -070078/* Uncomment to enable debugging */
79/* #define TUN_DEBUG 1 */
80
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#ifdef TUN_DEBUG
82static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070083
Joe Perches6b8a66e2011-03-02 07:18:10 +000084#define tun_debug(level, tun, fmt, args...) \
85do { \
86 if (tun->debug) \
87 netdev_printk(level, tun->dev, fmt, ##args); \
88} while (0)
89#define DBG1(level, fmt, args...) \
90do { \
91 if (debug == 2) \
92 printk(level fmt, ##args); \
93} while (0)
Rusty Russell14daa022008-04-12 18:48:58 -070094#else
Joe Perches6b8a66e2011-03-02 07:18:10 +000095#define tun_debug(level, tun, fmt, args...) \
96do { \
97 if (0) \
98 netdev_printk(level, tun->dev, fmt, ##args); \
99} while (0)
100#define DBG1(level, fmt, args...) \
101do { \
102 if (0) \
103 printk(level fmt, ##args); \
104} while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#endif
106
Michael S. Tsirkin031f5e032014-11-19 14:44:40 +0200107/* TUN device flags */
108
109/* IFF_ATTACH_QUEUE is never stored in device flags,
110 * overload it to mean fasync when stored there.
111 */
112#define TUN_FASYNC IFF_ATTACH_QUEUE
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200113/* High bits in flags field are unused. */
114#define TUN_VNET_LE 0x80000000
Michael S. Tsirkin031f5e032014-11-19 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 Krasnyanskyf271b2cc2008-07-14 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
stephen hemminger92d4ea62013-12-05 20:42:58 -0800127/* DEFAULT_MAX_NUM_RSS_QUEUES were chosen to let the rx/tx queues allocated for
Jason Wangedfb6a12013-01-23 03:59:12 +0000128 * the netdevice to be fit in one page. So we can make sure the success of
129 * memory allocation. TODO: increase the limit. */
130#define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES
Jason Wangb8732fb2013-01-23 03:59:13 +0000131#define MAX_TAP_FLOWS 4096
Jason Wangc8d68e62012-10-31 19:46:00 +0000132
Jason Wang96442e422012-10-31 19:46:02 +0000133#define TUN_FLOW_EXPIRE (3 * HZ)
134
Jason Wang54f968d2012-10-31 19:45:57 +0000135/* A tun_file connects an open character device to a tuntap netdevice. It
stephen hemminger92d4ea62013-12-05 20:42:58 -0800136 * also contains all socket related structures (except sock_fprog and tap_filter)
Jason Wang54f968d2012-10-31 19:45:57 +0000137 * to serve as one transmit queue for tuntap device. The sock_fprog and
138 * tap_filter were kept in tun_struct since they were used for filtering for the
Rami Rosen36fe8c092012-11-25 22:07:40 +0000139 * netdevice not for a specific queue (at least I didn't see the requirement for
Jason Wang54f968d2012-10-31 19:45:57 +0000140 * this).
Jason Wang6e914fc2012-10-31 19:45:58 +0000141 *
142 * RCU usage:
Rami Rosen36fe8c092012-11-25 22:07:40 +0000143 * The tun_file and tun_struct are loosely coupled, the pointer from one to the
Jason Wang6e914fc2012-10-31 19:45:58 +0000144 * other can only be read while rcu_read_lock or rtnl_lock is held.
Jason Wang54f968d2012-10-31 19:45:57 +0000145 */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000146struct tun_file {
Jason Wang54f968d2012-10-31 19:45:57 +0000147 struct sock sk;
148 struct socket socket;
149 struct socket_wq wq;
Jason Wang6e914fc2012-10-31 19:45:58 +0000150 struct tun_struct __rcu *tun;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +0000151 struct net *net;
Jason Wang54f968d2012-10-31 19:45:57 +0000152 struct fasync_struct *fasync;
153 /* only used for fasnyc */
154 unsigned int flags;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +0400155 union {
156 u16 queue_index;
157 unsigned int ifindex;
158 };
Jason Wang4008e972012-12-13 23:53:30 +0000159 struct list_head next;
160 struct tun_struct *detached;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000161};
162
Jason Wang96442e422012-10-31 19:46:02 +0000163struct tun_flow_entry {
164 struct hlist_node hash_link;
165 struct rcu_head rcu;
166 struct tun_struct *tun;
167
168 u32 rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800169 u32 rps_rxhash;
Jason Wang96442e422012-10-31 19:46:02 +0000170 int queue_index;
171 unsigned long updated;
172};
173
174#define TUN_NUM_FLOW_ENTRIES 1024
175
Jason Wang54f968d2012-10-31 19:45:57 +0000176/* Since the socket were moved to tun_file, to preserve the behavior of persist
Rami Rosen36fe8c092012-11-25 22:07:40 +0000177 * device, socket filter, sndbuf and vnet header size were restore when the
Jason Wang54f968d2012-10-31 19:45:57 +0000178 * file were attached to a persist device.
179 */
Rusty Russell14daa022008-04-12 18:48:58 -0700180struct tun_struct {
Jason Wangc8d68e62012-10-31 19:46:00 +0000181 struct tun_file __rcu *tfiles[MAX_TAP_QUEUES];
182 unsigned int numqueues;
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700183 unsigned int flags;
Eric W. Biederman0625c882012-02-07 16:48:55 -0800184 kuid_t owner;
185 kgid_t group;
Rusty Russell14daa022008-04-12 18:48:58 -0700186
Rusty Russell14daa022008-04-12 18:48:58 -0700187 struct net_device *dev;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000188 netdev_features_t set_features;
Michał Mirosław88255372011-04-19 06:13:10 +0000189#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
Ben Hutchings3d0ad092014-10-30 18:27:12 +0000190 NETIF_F_TSO6)
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200191
192 int vnet_hdr_sz;
Jason Wang54f968d2012-10-31 19:45:57 +0000193 int sndbuf;
194 struct tap_filter txflt;
195 struct sock_fprog fprog;
196 /* protected by rtnl lock */
197 bool filter_attached;
Rusty Russell14daa022008-04-12 18:48:58 -0700198#ifdef TUN_DEBUG
199 int debug;
200#endif
Jason Wang96442e422012-10-31 19:46:02 +0000201 spinlock_t lock;
Jason Wang96442e422012-10-31 19:46:02 +0000202 struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
203 struct timer_list flow_gc_timer;
204 unsigned long ageing_time;
Jason Wang4008e972012-12-13 23:53:30 +0000205 unsigned int numdisabled;
206 struct list_head disabled;
Paul Moore5dbbaf22013-01-14 07:12:19 +0000207 void *security;
Jason Wangb8732fb2013-01-23 03:59:13 +0000208 u32 flow_count;
Rusty Russell14daa022008-04-12 18:48:58 -0700209};
210
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300211static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val)
212{
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200213 return __virtio16_to_cpu(tun->flags & TUN_VNET_LE, val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300214}
215
216static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val)
217{
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200218 return __cpu_to_virtio16(tun->flags & TUN_VNET_LE, val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300219}
220
Jason Wang96442e422012-10-31 19:46:02 +0000221static inline u32 tun_hashfn(u32 rxhash)
222{
223 return rxhash & 0x3ff;
224}
225
226static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash)
227{
228 struct tun_flow_entry *e;
Jason Wang96442e422012-10-31 19:46:02 +0000229
Sasha Levinb67bfe02013-02-27 17:06:00 -0800230 hlist_for_each_entry_rcu(e, head, hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000231 if (e->rxhash == rxhash)
232 return e;
233 }
234 return NULL;
235}
236
237static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun,
238 struct hlist_head *head,
239 u32 rxhash, u16 queue_index)
240{
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000241 struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC);
242
Jason Wang96442e422012-10-31 19:46:02 +0000243 if (e) {
244 tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n",
245 rxhash, queue_index);
246 e->updated = jiffies;
247 e->rxhash = rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800248 e->rps_rxhash = 0;
Jason Wang96442e422012-10-31 19:46:02 +0000249 e->queue_index = queue_index;
250 e->tun = tun;
251 hlist_add_head_rcu(&e->hash_link, head);
Jason Wangb8732fb2013-01-23 03:59:13 +0000252 ++tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000253 }
254 return e;
255}
256
Jason Wang96442e422012-10-31 19:46:02 +0000257static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e)
258{
259 tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n",
260 e->rxhash, e->queue_index);
Tom Herbert9bc88932013-12-22 18:54:32 +0800261 sock_rps_reset_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000262 hlist_del_rcu(&e->hash_link);
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000263 kfree_rcu(e, rcu);
Jason Wangb8732fb2013-01-23 03:59:13 +0000264 --tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000265}
266
267static void tun_flow_flush(struct tun_struct *tun)
268{
269 int i;
270
271 spin_lock_bh(&tun->lock);
272 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
273 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800274 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000275
Sasha Levinb67bfe02013-02-27 17:06:00 -0800276 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
Jason Wang96442e422012-10-31 19:46:02 +0000277 tun_flow_delete(tun, e);
278 }
279 spin_unlock_bh(&tun->lock);
280}
281
282static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
283{
284 int i;
285
286 spin_lock_bh(&tun->lock);
287 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
288 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800289 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000290
Sasha Levinb67bfe02013-02-27 17:06:00 -0800291 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000292 if (e->queue_index == queue_index)
293 tun_flow_delete(tun, e);
294 }
295 }
296 spin_unlock_bh(&tun->lock);
297}
298
299static void tun_flow_cleanup(unsigned long data)
300{
301 struct tun_struct *tun = (struct tun_struct *)data;
302 unsigned long delay = tun->ageing_time;
303 unsigned long next_timer = jiffies + delay;
304 unsigned long count = 0;
305 int i;
306
307 tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n");
308
309 spin_lock_bh(&tun->lock);
310 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
311 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800312 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000313
Sasha Levinb67bfe02013-02-27 17:06:00 -0800314 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000315 unsigned long this_timer;
316 count++;
317 this_timer = e->updated + delay;
318 if (time_before_eq(this_timer, jiffies))
319 tun_flow_delete(tun, e);
320 else if (time_before(this_timer, next_timer))
321 next_timer = this_timer;
322 }
323 }
324
325 if (count)
326 mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer));
327 spin_unlock_bh(&tun->lock);
328}
329
Eric Dumazet49974422012-12-12 19:22:57 +0000330static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
Jason Wang9e857222013-01-28 01:05:19 +0000331 struct tun_file *tfile)
Jason Wang96442e422012-10-31 19:46:02 +0000332{
333 struct hlist_head *head;
334 struct tun_flow_entry *e;
335 unsigned long delay = tun->ageing_time;
Jason Wang9e857222013-01-28 01:05:19 +0000336 u16 queue_index = tfile->queue_index;
Jason Wang96442e422012-10-31 19:46:02 +0000337
338 if (!rxhash)
339 return;
340 else
341 head = &tun->flows[tun_hashfn(rxhash)];
342
343 rcu_read_lock();
344
Jason Wang9e857222013-01-28 01:05:19 +0000345 /* We may get a very small possibility of OOO during switching, not
346 * worth to optimize.*/
347 if (tun->numqueues == 1 || tfile->detached)
Jason Wang96442e422012-10-31 19:46:02 +0000348 goto unlock;
349
350 e = tun_flow_find(head, rxhash);
351 if (likely(e)) {
352 /* TODO: keep queueing to old queue until it's empty? */
353 e->queue_index = queue_index;
354 e->updated = jiffies;
Tom Herbert9bc88932013-12-22 18:54:32 +0800355 sock_rps_record_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000356 } else {
357 spin_lock_bh(&tun->lock);
Jason Wangb8732fb2013-01-23 03:59:13 +0000358 if (!tun_flow_find(head, rxhash) &&
359 tun->flow_count < MAX_TAP_FLOWS)
Jason Wang96442e422012-10-31 19:46:02 +0000360 tun_flow_create(tun, head, rxhash, queue_index);
361
362 if (!timer_pending(&tun->flow_gc_timer))
363 mod_timer(&tun->flow_gc_timer,
364 round_jiffies_up(jiffies + delay));
365 spin_unlock_bh(&tun->lock);
366 }
367
368unlock:
369 rcu_read_unlock();
370}
371
Tom Herbert9bc88932013-12-22 18:54:32 +0800372/**
373 * Save the hash received in the stack receive path and update the
374 * flow_hash table accordingly.
375 */
376static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash)
377{
378 if (unlikely(e->rps_rxhash != hash)) {
379 sock_rps_reset_flow_hash(e->rps_rxhash);
380 e->rps_rxhash = hash;
381 }
382}
383
Jason Wangc8d68e62012-10-31 19:46:00 +0000384/* We try to identify a flow through its rxhash first. The reason that
stephen hemminger92d4ea62013-12-05 20:42:58 -0800385 * we do not check rxq no. is because some cards(e.g 82599), chooses
Jason Wangc8d68e62012-10-31 19:46:00 +0000386 * the rxq based on the txq where the last packet of the flow comes. As
387 * the userspace application move between processors, we may get a
388 * different rxq no. here. If we could not get rxhash, then we would
389 * hope the rxq no. may help here.
390 */
Jason Wangf663dd92014-01-10 16:18:26 +0800391static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100392 void *accel_priv, select_queue_fallback_t fallback)
Jason Wangc8d68e62012-10-31 19:46:00 +0000393{
394 struct tun_struct *tun = netdev_priv(dev);
Jason Wang96442e422012-10-31 19:46:02 +0000395 struct tun_flow_entry *e;
Jason Wangc8d68e62012-10-31 19:46:00 +0000396 u32 txq = 0;
397 u32 numqueues = 0;
398
399 rcu_read_lock();
Jason Wang92bb73e2013-06-05 16:44:57 +0800400 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000401
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800402 txq = skb_get_hash(skb);
Jason Wangc8d68e62012-10-31 19:46:00 +0000403 if (txq) {
Jason Wang96442e422012-10-31 19:46:02 +0000404 e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
Tom Herbert9bc88932013-12-22 18:54:32 +0800405 if (e) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800406 tun_flow_save_rps_rxhash(e, txq);
Zhi Yong Wufbe4d452014-01-02 13:24:28 +0800407 txq = e->queue_index;
Tom Herbert9bc88932013-12-22 18:54:32 +0800408 } else
Jason Wang96442e422012-10-31 19:46:02 +0000409 /* use multiply and shift instead of expensive divide */
410 txq = ((u64)txq * numqueues) >> 32;
Jason Wangc8d68e62012-10-31 19:46:00 +0000411 } else if (likely(skb_rx_queue_recorded(skb))) {
412 txq = skb_get_rx_queue(skb);
413 while (unlikely(txq >= numqueues))
414 txq -= numqueues;
415 }
416
417 rcu_read_unlock();
418 return txq;
419}
420
Jason Wangcde8b152012-10-31 19:46:01 +0000421static inline bool tun_not_capable(struct tun_struct *tun)
422{
423 const struct cred *cred = current_cred();
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000424 struct net *net = dev_net(tun->dev);
Jason Wangcde8b152012-10-31 19:46:01 +0000425
426 return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
427 (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000428 !ns_capable(net->user_ns, CAP_NET_ADMIN);
Jason Wangcde8b152012-10-31 19:46:01 +0000429}
430
Jason Wangc8d68e62012-10-31 19:46:00 +0000431static void tun_set_real_num_queues(struct tun_struct *tun)
432{
433 netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
434 netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
435}
436
Jason Wang4008e972012-12-13 23:53:30 +0000437static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
438{
439 tfile->detached = tun;
440 list_add_tail(&tfile->next, &tun->disabled);
441 ++tun->numdisabled;
442}
443
Jason Wangd32649d2012-12-18 11:00:27 +0800444static struct tun_struct *tun_enable_queue(struct tun_file *tfile)
Jason Wang4008e972012-12-13 23:53:30 +0000445{
446 struct tun_struct *tun = tfile->detached;
447
448 tfile->detached = NULL;
449 list_del_init(&tfile->next);
450 --tun->numdisabled;
451 return tun;
452}
453
Jason Wang4bfb0512013-09-05 17:53:59 +0800454static void tun_queue_purge(struct tun_file *tfile)
455{
456 skb_queue_purge(&tfile->sk.sk_receive_queue);
457 skb_queue_purge(&tfile->sk.sk_error_queue);
458}
459
Jason Wangc8d68e62012-10-31 19:46:00 +0000460static void __tun_detach(struct tun_file *tfile, bool clean)
461{
462 struct tun_file *ntfile;
463 struct tun_struct *tun;
Jason Wangc8d68e62012-10-31 19:46:00 +0000464
Jason Wangb8deabd2013-01-11 16:59:32 +0000465 tun = rtnl_dereference(tfile->tun);
466
Jason Wang9e857222013-01-28 01:05:19 +0000467 if (tun && !tfile->detached) {
Jason Wangc8d68e62012-10-31 19:46:00 +0000468 u16 index = tfile->queue_index;
469 BUG_ON(index >= tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000470
471 rcu_assign_pointer(tun->tfiles[index],
472 tun->tfiles[tun->numqueues - 1]);
Jason Wangb8deabd2013-01-11 16:59:32 +0000473 ntfile = rtnl_dereference(tun->tfiles[index]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000474 ntfile->queue_index = index;
475
476 --tun->numqueues;
Jason Wang9e857222013-01-28 01:05:19 +0000477 if (clean) {
Monam Agarwalc9566742014-03-24 00:02:32 +0530478 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang4008e972012-12-13 23:53:30 +0000479 sock_put(&tfile->sk);
Jason Wang9e857222013-01-28 01:05:19 +0000480 } else
Jason Wang4008e972012-12-13 23:53:30 +0000481 tun_disable_queue(tun, tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000482
483 synchronize_net();
Jason Wang96442e422012-10-31 19:46:02 +0000484 tun_flow_delete_by_queue(tun, tun->numqueues + 1);
Jason Wangc8d68e62012-10-31 19:46:00 +0000485 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800486 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000487 tun_set_real_num_queues(tun);
Jason Wangdd38bd82013-01-11 16:59:34 +0000488 } else if (tfile->detached && clean) {
Jason Wang4008e972012-12-13 23:53:30 +0000489 tun = tun_enable_queue(tfile);
Jason Wangdd38bd82013-01-11 16:59:34 +0000490 sock_put(&tfile->sk);
491 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000492
493 if (clean) {
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000494 if (tun && tun->numqueues == 0 && tun->numdisabled == 0) {
495 netif_carrier_off(tun->dev);
496
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200497 if (!(tun->flags & IFF_PERSIST) &&
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000498 tun->dev->reg_state == NETREG_REGISTERED)
Jason Wang4008e972012-12-13 23:53:30 +0000499 unregister_netdevice(tun->dev);
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000500 }
Jason Wang4008e972012-12-13 23:53:30 +0000501
Jason Wangc8d68e62012-10-31 19:46:00 +0000502 BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED,
503 &tfile->socket.flags));
504 sk_release_kernel(&tfile->sk);
505 }
506}
507
508static void tun_detach(struct tun_file *tfile, bool clean)
509{
510 rtnl_lock();
511 __tun_detach(tfile, clean);
512 rtnl_unlock();
513}
514
515static void tun_detach_all(struct net_device *dev)
516{
517 struct tun_struct *tun = netdev_priv(dev);
Jason Wang4008e972012-12-13 23:53:30 +0000518 struct tun_file *tfile, *tmp;
Jason Wangc8d68e62012-10-31 19:46:00 +0000519 int i, n = tun->numqueues;
520
521 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000522 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000523 BUG_ON(!tfile);
Xi Wang9e641bd2014-05-16 15:11:48 -0700524 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530525 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wangc8d68e62012-10-31 19:46:00 +0000526 --tun->numqueues;
527 }
Jason Wang9e857222013-01-28 01:05:19 +0000528 list_for_each_entry(tfile, &tun->disabled, next) {
Xi Wang9e641bd2014-05-16 15:11:48 -0700529 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530530 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang9e857222013-01-28 01:05:19 +0000531 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000532 BUG_ON(tun->numqueues != 0);
533
534 synchronize_net();
535 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000536 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000537 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800538 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000539 sock_put(&tfile->sk);
540 }
Jason Wang4008e972012-12-13 23:53:30 +0000541 list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
542 tun_enable_queue(tfile);
Jason Wang4bfb0512013-09-05 17:53:59 +0800543 tun_queue_purge(tfile);
Jason Wang4008e972012-12-13 23:53:30 +0000544 sock_put(&tfile->sk);
545 }
546 BUG_ON(tun->numdisabled != 0);
Jason Wangdd38bd82013-01-11 16:59:34 +0000547
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200548 if (tun->flags & IFF_PERSIST)
Jason Wangdd38bd82013-01-11 16:59:34 +0000549 module_put(THIS_MODULE);
Jason Wangc8d68e62012-10-31 19:46:00 +0000550}
551
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400552static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter)
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000553{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000554 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000555 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000556
Paul Moore5dbbaf22013-01-14 07:12:19 +0000557 err = security_tun_dev_attach(tfile->socket.sk, tun->security);
558 if (err < 0)
559 goto out;
560
Eric W. Biederman38231b72009-01-20 11:02:28 +0000561 err = -EINVAL;
Jason Wang9e857222013-01-28 01:05:19 +0000562 if (rtnl_dereference(tfile->tun) && !tfile->detached)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000563 goto out;
564
565 err = -EBUSY;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200566 if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1)
Jason Wangc8d68e62012-10-31 19:46:00 +0000567 goto out;
568
569 err = -E2BIG;
Jason Wang4008e972012-12-13 23:53:30 +0000570 if (!tfile->detached &&
571 tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000572 goto out;
573
574 err = 0;
Jason Wang54f968d2012-10-31 19:45:57 +0000575
stephen hemminger92d4ea62013-12-05 20:42:58 -0800576 /* Re-attach the filter to persist device */
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400577 if (!skip_filter && (tun->filter_attached == true)) {
Jason Wang54f968d2012-10-31 19:45:57 +0000578 err = sk_attach_filter(&tun->fprog, tfile->socket.sk);
579 if (!err)
580 goto out;
581 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000582 tfile->queue_index = tun->numqueues;
Jason Wang6e914fc2012-10-31 19:45:58 +0000583 rcu_assign_pointer(tfile->tun, tun);
Jason Wangc8d68e62012-10-31 19:46:00 +0000584 rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000585 tun->numqueues++;
586
Jason Wang4008e972012-12-13 23:53:30 +0000587 if (tfile->detached)
588 tun_enable_queue(tfile);
589 else
590 sock_hold(&tfile->sk);
591
Jason Wangc8d68e62012-10-31 19:46:00 +0000592 tun_set_real_num_queues(tun);
593
Jason Wangc8d68e62012-10-31 19:46:00 +0000594 /* device is allowed to go away first, so no need to hold extra
595 * refcnt.
596 */
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000597
Eric W. Biederman38231b72009-01-20 11:02:28 +0000598out:
Eric W. Biederman38231b72009-01-20 11:02:28 +0000599 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000600}
601
Eric W. Biederman631ab462009-01-20 11:00:40 +0000602static struct tun_struct *__tun_get(struct tun_file *tfile)
603{
Jason Wang6e914fc2012-10-31 19:45:58 +0000604 struct tun_struct *tun;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000605
Jason Wang6e914fc2012-10-31 19:45:58 +0000606 rcu_read_lock();
607 tun = rcu_dereference(tfile->tun);
608 if (tun)
609 dev_hold(tun->dev);
610 rcu_read_unlock();
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000611
612 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000613}
614
615static struct tun_struct *tun_get(struct file *file)
616{
617 return __tun_get(file->private_data);
618}
619
620static void tun_put(struct tun_struct *tun)
621{
Jason Wang6e914fc2012-10-31 19:45:58 +0000622 dev_put(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000623}
624
Joe Perches6b8a66e2011-03-02 07:18:10 +0000625/* TAP filtering */
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700626static void addr_hash_set(u32 *mask, const u8 *addr)
627{
628 int n = ether_crc(ETH_ALEN, addr) >> 26;
629 mask[n >> 5] |= (1 << (n & 31));
630}
631
632static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
633{
634 int n = ether_crc(ETH_ALEN, addr) >> 26;
635 return mask[n >> 5] & (1 << (n & 31));
636}
637
638static int update_filter(struct tap_filter *filter, void __user *arg)
639{
640 struct { u8 u[ETH_ALEN]; } *addr;
641 struct tun_filter uf;
642 int err, alen, n, nexact;
643
644 if (copy_from_user(&uf, arg, sizeof(uf)))
645 return -EFAULT;
646
647 if (!uf.count) {
648 /* Disabled */
649 filter->count = 0;
650 return 0;
651 }
652
653 alen = ETH_ALEN * uf.count;
654 addr = kmalloc(alen, GFP_KERNEL);
655 if (!addr)
656 return -ENOMEM;
657
658 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
659 err = -EFAULT;
660 goto done;
661 }
662
663 /* The filter is updated without holding any locks. Which is
664 * perfectly safe. We disable it first and in the worst
665 * case we'll accept a few undesired packets. */
666 filter->count = 0;
667 wmb();
668
669 /* Use first set of addresses as an exact filter */
670 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
671 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
672
673 nexact = n;
674
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800675 /* Remaining multicast addresses are hashed,
676 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700677 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800678 for (; n < uf.count; n++) {
679 if (!is_multicast_ether_addr(addr[n].u)) {
680 err = 0; /* no filter */
681 goto done;
682 }
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700683 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800684 }
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700685
686 /* For ALLMULTI just set the mask to all ones.
687 * This overrides the mask populated above. */
688 if ((uf.flags & TUN_FLT_ALLMULTI))
689 memset(filter->mask, ~0, sizeof(filter->mask));
690
691 /* Now enable the filter */
692 wmb();
693 filter->count = nexact;
694
695 /* Return the number of exact filters */
696 err = nexact;
697
698done:
699 kfree(addr);
700 return err;
701}
702
703/* Returns: 0 - drop, !=0 - accept */
704static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
705{
706 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
707 * at this point. */
708 struct ethhdr *eh = (struct ethhdr *) skb->data;
709 int i;
710
711 /* Exact match */
712 for (i = 0; i < filter->count; i++)
Joe Perches2e42e472012-05-09 17:17:46 +0000713 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700714 return 1;
715
716 /* Inexact match (multicast only) */
717 if (is_multicast_ether_addr(eh->h_dest))
718 return addr_hash_test(filter->mask, eh->h_dest);
719
720 return 0;
721}
722
723/*
724 * Checks whether the packet is accepted or not.
725 * Returns: 0 - drop, !=0 - accept
726 */
727static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
728{
729 if (!filter->count)
730 return 1;
731
732 return run_filter(filter, skb);
733}
734
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735/* Network device part of the driver */
736
Jeff Garzik7282d492006-09-13 14:30:00 -0400737static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000739/* Net device detach from fd. */
740static void tun_net_uninit(struct net_device *dev)
741{
Jason Wangc8d68e62012-10-31 19:46:00 +0000742 tun_detach_all(dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000743}
744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745/* Net device open. */
746static int tun_net_open(struct net_device *dev)
747{
Jason Wangc8d68e62012-10-31 19:46:00 +0000748 netif_tx_start_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 return 0;
750}
751
752/* Net device close. */
753static int tun_net_close(struct net_device *dev)
754{
Jason Wangc8d68e62012-10-31 19:46:00 +0000755 netif_tx_stop_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 return 0;
757}
758
759/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000760static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761{
762 struct tun_struct *tun = netdev_priv(dev);
Jason Wangc8d68e62012-10-31 19:46:00 +0000763 int txq = skb->queue_mapping;
Jason Wang6e914fc2012-10-31 19:45:58 +0000764 struct tun_file *tfile;
Dominic Curranfa358642014-01-22 03:03:23 +0000765 u32 numqueues = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
Jason Wang6e914fc2012-10-31 19:45:58 +0000767 rcu_read_lock();
Jason Wangc8d68e62012-10-31 19:46:00 +0000768 tfile = rcu_dereference(tun->tfiles[txq]);
Dominic Curranfa358642014-01-22 03:03:23 +0000769 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 /* Drop packet if interface is not attached */
Dominic Curranfa358642014-01-22 03:03:23 +0000772 if (txq >= numqueues)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 goto drop;
774
Dominic Curranfa358642014-01-22 03:03:23 +0000775 if (numqueues == 1) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800776 /* Select queue was not called for the skbuff, so we extract the
777 * RPS hash and save it into the flow_table here.
778 */
779 __u32 rxhash;
780
781 rxhash = skb_get_hash(skb);
782 if (rxhash) {
783 struct tun_flow_entry *e;
784 e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)],
785 rxhash);
786 if (e)
787 tun_flow_save_rps_rxhash(e, rxhash);
788 }
789 }
790
Jason Wang6e914fc2012-10-31 19:45:58 +0000791 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
792
Jason Wangc8d68e62012-10-31 19:46:00 +0000793 BUG_ON(!tfile);
794
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700795 /* Drop if the filter does not like it.
796 * This is a noop if the filter is disabled.
797 * Filter can be enabled only for the TAP devices. */
798 if (!check_filter(&tun->txflt, skb))
799 goto drop;
800
Jason Wang54f968d2012-10-31 19:45:57 +0000801 if (tfile->socket.sk->sk_filter &&
802 sk_filter(tfile->socket.sk, skb))
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000803 goto drop;
804
Rami Rosen36fe8c092012-11-25 22:07:40 +0000805 /* Limit the number of packets queued by dividing txq length with the
Jason Wangc8d68e62012-10-31 19:46:00 +0000806 * number of queues.
807 */
Dominic Curranfa358642014-01-22 03:03:23 +0000808 if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues
809 >= dev->tx_queue_len)
Michael S. Tsirkin5d097102012-12-03 10:07:14 +0000810 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Jason Wang7bf66302013-09-05 17:54:00 +0800812 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
813 goto drop;
814
Richard Cochraneda29772013-07-19 19:40:10 +0200815 if (skb->sk) {
816 sock_tx_timestamp(skb->sk, &skb_shinfo(skb)->tx_flags);
817 sw_tx_timestamp(skb);
818 }
819
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000820 /* Orphan the skb - required as we might hang on to it
Jason Wang7bf66302013-09-05 17:54:00 +0800821 * for indefinite time.
822 */
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000823 skb_orphan(skb);
824
Eric Dumazetf8af75f2013-03-06 11:02:37 +0000825 nf_reset(skb);
826
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700827 /* Enqueue packet */
Jason Wang54f968d2012-10-31 19:45:57 +0000828 skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
830 /* Notify and wake up reader process */
Jason Wang54f968d2012-10-31 19:45:57 +0000831 if (tfile->flags & TUN_FASYNC)
832 kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
Xi Wang9e641bd2014-05-16 15:11:48 -0700833 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Jason Wang6e914fc2012-10-31 19:45:58 +0000834
835 rcu_read_unlock();
Patrick McHardy6ed10652009-06-23 06:03:08 +0000836 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837
838drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700839 dev->stats.tx_dropped++;
Michael S. Tsirkin149d36f2012-11-01 09:16:32 +0000840 skb_tx_error(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 kfree_skb(skb);
Jason Wang6e914fc2012-10-31 19:45:58 +0000842 rcu_read_unlock();
Jason Wangbaeabab2014-11-18 13:20:41 +0800843 return NET_XMIT_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844}
845
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700846static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847{
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700848 /*
849 * This callback is supposed to deal with mc filter in
850 * _rx_ path and has nothing to do with the _tx_ path.
851 * In rx path we always accept everything userspace gives us.
852 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853}
854
Ed Swierk4885a502007-09-16 12:21:38 -0700855#define MIN_MTU 68
856#define MAX_MTU 65535
857
858static int
859tun_net_change_mtu(struct net_device *dev, int new_mtu)
860{
861 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
862 return -EINVAL;
863 dev->mtu = new_mtu;
864 return 0;
865}
866
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000867static netdev_features_t tun_net_fix_features(struct net_device *dev,
868 netdev_features_t features)
Michał Mirosław88255372011-04-19 06:13:10 +0000869{
870 struct tun_struct *tun = netdev_priv(dev);
871
872 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
873}
Neil Hormanbebd0972011-06-15 05:25:01 +0000874#ifdef CONFIG_NET_POLL_CONTROLLER
875static void tun_poll_controller(struct net_device *dev)
876{
877 /*
878 * Tun only receives frames when:
879 * 1) the char device endpoint gets data from user space
880 * 2) the tun socket gets a sendmsg call from user space
stephen hemminger92d4ea62013-12-05 20:42:58 -0800881 * Since both of those are synchronous operations, we are guaranteed
Neil Hormanbebd0972011-06-15 05:25:01 +0000882 * never to have pending data when we poll for it
stephen hemminger92d4ea62013-12-05 20:42:58 -0800883 * so there is nothing to do here but return.
Neil Hormanbebd0972011-06-15 05:25:01 +0000884 * We need this though so netpoll recognizes us as an interface that
885 * supports polling, which enables bridge devices in virt setups to
886 * still use netconsole
887 */
888 return;
889}
890#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800891static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000892 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800893 .ndo_open = tun_net_open,
894 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800895 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800896 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000897 .ndo_fix_features = tun_net_fix_features,
Jason Wangc8d68e62012-10-31 19:46:00 +0000898 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000899#ifdef CONFIG_NET_POLL_CONTROLLER
900 .ndo_poll_controller = tun_poll_controller,
901#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800902};
903
904static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000905 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800906 .ndo_open = tun_net_open,
907 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800908 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800909 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000910 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +0000911 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800912 .ndo_set_mac_address = eth_mac_addr,
913 .ndo_validate_addr = eth_validate_addr,
Jason Wangc8d68e62012-10-31 19:46:00 +0000914 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000915#ifdef CONFIG_NET_POLL_CONTROLLER
916 .ndo_poll_controller = tun_poll_controller,
917#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800918};
919
Pavel Emelyanov944a1372013-06-11 17:01:08 +0400920static void tun_flow_init(struct tun_struct *tun)
Jason Wang96442e422012-10-31 19:46:02 +0000921{
922 int i;
923
Jason Wang96442e422012-10-31 19:46:02 +0000924 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
925 INIT_HLIST_HEAD(&tun->flows[i]);
926
927 tun->ageing_time = TUN_FLOW_EXPIRE;
928 setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun);
929 mod_timer(&tun->flow_gc_timer,
930 round_jiffies_up(jiffies + tun->ageing_time));
Jason Wang96442e422012-10-31 19:46:02 +0000931}
932
933static void tun_flow_uninit(struct tun_struct *tun)
934{
935 del_timer_sync(&tun->flow_gc_timer);
936 tun_flow_flush(tun);
Jason Wang96442e422012-10-31 19:46:02 +0000937}
938
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939/* Initialize net device. */
940static void tun_net_init(struct net_device *dev)
941{
942 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400943
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200945 case IFF_TUN:
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800946 dev->netdev_ops = &tun_netdev_ops;
947
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 /* Point-to-Point TUN Device */
949 dev->hard_header_len = 0;
950 dev->addr_len = 0;
951 dev->mtu = 1500;
952
953 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400954 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
956 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
957 break;
958
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200959 case IFF_TAP:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -0800960 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 /* Ethernet TAP Device */
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -0700962 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +0000963 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
stephen hemmingera6768472012-12-10 15:16:00 +0000964 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
Danny Kukawkaf2cedb62012-02-15 06:45:39 +0000966 eth_hw_addr_random(dev);
Brian Braunstein36226a82007-04-26 01:00:55 -0700967
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
969 break;
970 }
971}
972
973/* Character device part */
974
975/* Poll */
Jason Wangc8d68e62012-10-31 19:46:00 +0000976static unsigned int tun_chr_poll(struct file *file, poll_table *wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400977{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +0000978 struct tun_file *tfile = file->private_data;
979 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000980 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800981 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
983 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +0000984 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
Jason Wang54f968d2012-10-31 19:45:57 +0000986 sk = tfile->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000987
Joe Perches6b8a66e2011-03-02 07:18:10 +0000988 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
Xi Wang9e641bd2014-05-16 15:11:48 -0700990 poll_wait(file, sk_sleep(sk), wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400991
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000992 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 mask |= POLLIN | POLLRDNORM;
994
Herbert Xu33dccbb2009-02-05 21:25:32 -0800995 if (sock_writeable(sk) ||
996 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
997 sock_writeable(sk)))
998 mask |= POLLOUT | POLLWRNORM;
999
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001000 if (tun->dev->reg_state != NETREG_REGISTERED)
1001 mask = POLLERR;
1002
Eric W. Biederman631ab462009-01-20 11:00:40 +00001003 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 return mask;
1005}
1006
Rusty Russellf42157c2008-08-15 15:15:10 -07001007/* prepad is the amount to reserve at front. len is length after that.
1008 * linear is a hint as to how much to copy (usually headers). */
Jason Wang54f968d2012-10-31 19:45:57 +00001009static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001010 size_t prepad, size_t len,
1011 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -07001012{
Jason Wang54f968d2012-10-31 19:45:57 +00001013 struct sock *sk = tfile->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -07001014 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001015 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -07001016
1017 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -07001018 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001019 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -07001020
Herbert Xu33dccbb2009-02-05 21:25:32 -08001021 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
Eric Dumazet28d64272013-08-08 14:38:47 -07001022 &err, 0);
Rusty Russellf42157c2008-08-15 15:15:10 -07001023 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001024 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -07001025
1026 skb_reserve(skb, prepad);
1027 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001028 skb->data_len = len - linear;
1029 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -07001030
1031 return skb;
1032}
1033
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034/* Get packet from user space buffer */
Jason Wang54f968d2012-10-31 19:45:57 +00001035static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
Al Virof5ff53b2014-06-19 15:36:49 -04001036 void *msg_control, struct iov_iter *from,
1037 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038{
Harvey Harrison09640e632009-02-01 00:45:17 -08001039 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 struct sk_buff *skb;
Al Virof5ff53b2014-06-19 15:36:49 -04001041 size_t total_len = iov_iter_count(from);
Jason Wang3dd5c332013-07-10 13:43:27 +08001042 size_t len = total_len, align = NET_SKB_PAD, linear;
Rusty Russellf43798c2008-07-03 03:48:02 -07001043 struct virtio_net_hdr gso = { 0 };
Jason Wang96f8d9e2013-11-13 14:00:39 +08001044 int good_linear;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001045 int copylen;
1046 bool zerocopy = false;
1047 int err;
Eric Dumazet49974422012-12-12 19:22:57 +00001048 u32 rxhash;
Al Virof5ff53b2014-06-19 15:36:49 -04001049 ssize_t n;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001051 if (!(tun->flags & IFF_NO_PI)) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001052 if (len < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001054 len -= sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
Al Virof5ff53b2014-06-19 15:36:49 -04001056 n = copy_from_iter(&pi, sizeof(pi), from);
1057 if (n != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 return -EFAULT;
1059 }
1060
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001061 if (tun->flags & IFF_VNET_HDR) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001062 if (len < tun->vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001063 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001064 len -= tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001065
Al Virof5ff53b2014-06-19 15:36:49 -04001066 n = copy_from_iter(&gso, sizeof(gso), from);
1067 if (n != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001068 return -EFAULT;
1069
Herbert Xu49091222009-06-08 00:20:01 -07001070 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001071 tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len))
1072 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 -07001073
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001074 if (tun16_to_cpu(tun, gso.hdr_len) > len)
Rusty Russellf43798c2008-07-03 03:48:02 -07001075 return -EINVAL;
Herbert Xud8febb72014-11-30 18:03:31 +08001076 iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001077 }
1078
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001079 if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) {
stephen hemmingera504b862011-06-08 14:33:07 +00001080 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -07001081 if (unlikely(len < ETH_HLEN ||
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001082 (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -07001083 return -EINVAL;
1084 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001085
Jason Wang96f8d9e2013-11-13 14:00:39 +08001086 good_linear = SKB_MAX_HEAD(align);
1087
Jason Wang88529172013-07-18 10:55:15 +08001088 if (msg_control) {
Al Virof5ff53b2014-06-19 15:36:49 -04001089 struct iov_iter i = *from;
1090
Jason Wang88529172013-07-18 10:55:15 +08001091 /* There are 256 bytes to be copied in skb, so there is
1092 * enough room for skb expand head in case it is used.
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001093 * The rest of the buffer is mapped from userspace.
1094 */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001095 copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001096 if (copylen > good_linear)
1097 copylen = good_linear;
Jason Wang3dd5c332013-07-10 13:43:27 +08001098 linear = copylen;
Al Virof5ff53b2014-06-19 15:36:49 -04001099 iov_iter_advance(&i, copylen);
1100 if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS)
Jason Wang88529172013-07-18 10:55:15 +08001101 zerocopy = true;
1102 }
1103
1104 if (!zerocopy) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001105 copylen = len;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001106 if (tun16_to_cpu(tun, gso.hdr_len) > good_linear)
Jason Wang96f8d9e2013-11-13 14:00:39 +08001107 linear = good_linear;
1108 else
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001109 linear = tun16_to_cpu(tun, gso.hdr_len);
Jason Wang3dd5c332013-07-10 13:43:27 +08001110 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001111
Jason Wang3dd5c332013-07-10 13:43:27 +08001112 skb = tun_alloc_skb(tfile, align, copylen, linear, noblock);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001113 if (IS_ERR(skb)) {
1114 if (PTR_ERR(skb) != -EAGAIN)
1115 tun->dev->stats.rx_dropped++;
1116 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 }
1118
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001119 if (zerocopy)
Al Virof5ff53b2014-06-19 15:36:49 -04001120 err = zerocopy_sg_from_iter(skb, from);
Jason Wang88529172013-07-18 10:55:15 +08001121 else {
Al Virof5ff53b2014-06-19 15:36:49 -04001122 err = skb_copy_datagram_from_iter(skb, 0, from, len);
Jason Wang88529172013-07-18 10:55:15 +08001123 if (!err && msg_control) {
1124 struct ubuf_info *uarg = msg_control;
1125 uarg->callback(uarg, false);
1126 }
1127 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001128
1129 if (err) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001130 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -08001131 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -08001133 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134
Rusty Russellf43798c2008-07-03 03:48:02 -07001135 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001136 if (!skb_partial_csum_set(skb, tun16_to_cpu(tun, gso.csum_start),
1137 tun16_to_cpu(tun, gso.csum_offset))) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001138 tun->dev->stats.rx_frame_errors++;
1139 kfree_skb(skb);
1140 return -EINVAL;
1141 }
Michał Mirosław88255372011-04-19 06:13:10 +00001142 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001143
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001145 case IFF_TUN:
1146 if (tun->flags & IFF_NO_PI) {
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -07001147 switch (skb->data[0] & 0xf0) {
1148 case 0x40:
1149 pi.proto = htons(ETH_P_IP);
1150 break;
1151 case 0x60:
1152 pi.proto = htons(ETH_P_IPV6);
1153 break;
1154 default:
1155 tun->dev->stats.rx_dropped++;
1156 kfree_skb(skb);
1157 return -EINVAL;
1158 }
1159 }
1160
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001161 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -07001163 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001165 case IFF_TAP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 skb->protocol = eth_type_trans(skb, tun->dev);
1167 break;
Joe Perches6403eab2011-06-03 11:51:20 +00001168 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
Ben Hutchings5188cd42014-10-30 18:27:17 +00001170 skb_reset_network_header(skb);
1171
Rusty Russellf43798c2008-07-03 03:48:02 -07001172 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
1173 pr_debug("GSO!\n");
1174 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
1175 case VIRTIO_NET_HDR_GSO_TCPV4:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001176 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
Rusty Russellf43798c2008-07-03 03:48:02 -07001177 break;
1178 case VIRTIO_NET_HDR_GSO_TCPV6:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001179 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
Rusty Russellf43798c2008-07-03 03:48:02 -07001180 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001181 case VIRTIO_NET_HDR_GSO_UDP:
Ben Hutchings3d0ad092014-10-30 18:27:12 +00001182 {
1183 static bool warned;
1184
1185 if (!warned) {
1186 warned = true;
1187 netdev_warn(tun->dev,
1188 "%s: using disabled UFO feature; please fix this program\n",
1189 current->comm);
1190 }
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001191 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Ben Hutchings5188cd42014-10-30 18:27:17 +00001192 if (skb->protocol == htons(ETH_P_IPV6))
1193 ipv6_proxy_select_ident(skb);
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001194 break;
Ben Hutchings3d0ad092014-10-30 18:27:12 +00001195 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001196 default:
1197 tun->dev->stats.rx_frame_errors++;
1198 kfree_skb(skb);
1199 return -EINVAL;
1200 }
1201
1202 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001203 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
Rusty Russellf43798c2008-07-03 03:48:02 -07001204
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001205 skb_shinfo(skb)->gso_size = tun16_to_cpu(tun, gso.gso_size);
Rusty Russellf43798c2008-07-03 03:48:02 -07001206 if (skb_shinfo(skb)->gso_size == 0) {
1207 tun->dev->stats.rx_frame_errors++;
1208 kfree_skb(skb);
1209 return -EINVAL;
1210 }
1211
1212 /* Header must be checked, and gso_segs computed. */
1213 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
1214 skb_shinfo(skb)->gso_segs = 0;
1215 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001216
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001217 /* copy skb_ubuf_info for callback when skb has no error */
1218 if (zerocopy) {
1219 skb_shinfo(skb)->destructor_arg = msg_control;
1220 skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001221 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001222 }
1223
Jason Wang40893fd2013-03-26 23:11:22 +00001224 skb_probe_transport_header(skb, 0);
Jason Wang38502af2013-03-25 20:19:56 +00001225
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001226 rxhash = skb_get_hash(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001228
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001229 tun->dev->stats.rx_packets++;
1230 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231
Jason Wang9e857222013-01-28 01:05:19 +00001232 tun_flow_update(tun, rxhash, tfile);
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001233 return total_len;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001234}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
Al Virof5ff53b2014-06-19 15:36:49 -04001236static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001238 struct file *file = iocb->ki_filp;
Herbert Xuab46d7792009-02-14 20:46:39 -08001239 struct tun_struct *tun = tun_get(file);
Jason Wang54f968d2012-10-31 19:45:57 +00001240 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001241 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
1243 if (!tun)
1244 return -EBADFD;
1245
Al Virof5ff53b2014-06-19 15:36:49 -04001246 result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001247
1248 tun_put(tun);
1249 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250}
1251
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +00001253static ssize_t tun_put_user(struct tun_struct *tun,
Jason Wang54f968d2012-10-31 19:45:57 +00001254 struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001255 struct sk_buff *skb,
Herbert Xue0b46d02014-11-07 21:22:23 +08001256 struct iov_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257{
1258 struct tun_pi pi = { 0, skb->protocol };
Herbert Xue0b46d02014-11-07 21:22:23 +08001259 ssize_t total;
Jason Wang8c847d22014-11-13 16:54:14 +08001260 int vlan_offset = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001261 int vlan_hlen = 0;
Herbert Xu2eb783c2014-11-03 04:30:14 +08001262 int vnet_hdr_sz = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001263
1264 if (vlan_tx_tag_present(skb))
1265 vlan_hlen = VLAN_HLEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001267 if (tun->flags & IFF_VNET_HDR)
Herbert Xu2eb783c2014-11-03 04:30:14 +08001268 vnet_hdr_sz = tun->vnet_hdr_sz;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269
Herbert Xue0b46d02014-11-07 21:22:23 +08001270 total = skb->len + vlan_hlen + vnet_hdr_sz;
1271
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001272 if (!(tun->flags & IFF_NO_PI)) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001273 if (iov_iter_count(iter) < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 return -EINVAL;
1275
Herbert Xue0b46d02014-11-07 21:22:23 +08001276 total += sizeof(pi);
1277 if (iov_iter_count(iter) < total) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 /* Packet will be striped */
1279 pi.flags |= TUN_PKT_STRIP;
1280 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001281
Herbert Xue0b46d02014-11-07 21:22:23 +08001282 if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 return -EFAULT;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001284 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285
Herbert Xu2eb783c2014-11-03 04:30:14 +08001286 if (vnet_hdr_sz) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001287 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Herbert Xue0b46d02014-11-07 21:22:23 +08001288 if (iov_iter_count(iter) < vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001289 return -EINVAL;
1290
1291 if (skb_is_gso(skb)) {
1292 struct skb_shared_info *sinfo = skb_shinfo(skb);
1293
1294 /* This is a hint as to how much should be linear. */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001295 gso.hdr_len = cpu_to_tun16(tun, skb_headlen(skb));
1296 gso.gso_size = cpu_to_tun16(tun, sinfo->gso_size);
Rusty Russellf43798c2008-07-03 03:48:02 -07001297 if (sinfo->gso_type & SKB_GSO_TCPV4)
1298 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
1299 else if (sinfo->gso_type & SKB_GSO_TCPV6)
1300 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001301 else {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001302 pr_err("unexpected GSO type: "
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001303 "0x%x, gso_size %d, hdr_len %d\n",
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001304 sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size),
1305 tun16_to_cpu(tun, gso.hdr_len));
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001306 print_hex_dump(KERN_ERR, "tun: ",
1307 DUMP_PREFIX_NONE,
1308 16, 1, skb->head,
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001309 min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true);
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001310 WARN_ON_ONCE(1);
1311 return -EINVAL;
1312 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001313 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
1314 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
1315 } else
1316 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
1317
1318 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1319 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001320 gso.csum_start = cpu_to_tun16(tun, skb_checksum_start_offset(skb) +
1321 vlan_hlen);
1322 gso.csum_offset = cpu_to_tun16(tun, skb->csum_offset);
Jason Wang10a8d942011-06-10 00:56:17 +00001323 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
1324 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
Rusty Russellf43798c2008-07-03 03:48:02 -07001325 } /* else everything is zero */
1326
Herbert Xue0b46d02014-11-07 21:22:23 +08001327 if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001328 return -EFAULT;
Jason Wang8c847d22014-11-13 16:54:14 +08001329
1330 iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001331 }
1332
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001333 if (vlan_hlen) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001334 int ret;
Jason Wang6680ec62013-07-25 13:00:33 +08001335 struct {
1336 __be16 h_vlan_proto;
1337 __be16 h_vlan_TCI;
1338 } veth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
Jason Wang6680ec62013-07-25 13:00:33 +08001340 veth.h_vlan_proto = skb->vlan_proto;
1341 veth.h_vlan_TCI = htons(vlan_tx_tag_get(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
Jason Wang6680ec62013-07-25 13:00:33 +08001343 vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto);
Jason Wang6680ec62013-07-25 13:00:33 +08001344
Herbert Xue0b46d02014-11-07 21:22:23 +08001345 ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset);
1346 if (ret || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001347 goto done;
1348
Herbert Xue0b46d02014-11-07 21:22:23 +08001349 ret = copy_to_iter(&veth, sizeof(veth), iter);
1350 if (ret != sizeof(veth) || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001351 goto done;
1352 }
1353
Herbert Xue0b46d02014-11-07 21:22:23 +08001354 skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset);
Jason Wang6680ec62013-07-25 13:00:33 +08001355
1356done:
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001357 tun->dev->stats.tx_packets++;
Herbert Xue0b46d02014-11-07 21:22:23 +08001358 tun->dev->stats.tx_bytes += skb->len + vlan_hlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
1360 return total;
1361}
1362
Jason Wang54f968d2012-10-31 19:45:57 +00001363static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
Al Viro9b067032014-11-07 13:52:07 -05001364 struct iov_iter *to,
1365 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 struct sk_buff *skb;
Al Viro9b067032014-11-07 13:52:07 -05001368 ssize_t ret;
Xi Wang9e641bd2014-05-16 15:11:48 -07001369 int peeked, err, off = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370
Rami Rosen3872baf2012-11-25 22:07:41 +00001371 tun_debug(KERN_INFO, tun, "tun_do_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
Al Viro9b067032014-11-07 13:52:07 -05001373 if (!iov_iter_count(to))
1374 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Xi Wang9e641bd2014-05-16 15:11:48 -07001376 if (tun->dev->reg_state != NETREG_REGISTERED)
1377 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378
Xi Wang9e641bd2014-05-16 15:11:48 -07001379 /* Read frames from queue */
1380 skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0,
1381 &peeked, &off, &err);
Herbert Xue0b46d02014-11-07 21:22:23 +08001382 if (!skb)
Al Viro9b067032014-11-07 13:52:07 -05001383 return 0;
Herbert Xue0b46d02014-11-07 21:22:23 +08001384
Al Viro9b067032014-11-07 13:52:07 -05001385 ret = tun_put_user(tun, tfile, skb, to);
Jason Wangf51a5e82014-12-01 16:53:15 +08001386 if (unlikely(ret < 0))
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -07001387 kfree_skb(skb);
Jason Wangf51a5e82014-12-01 16:53:15 +08001388 else
1389 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001391 return ret;
1392}
1393
Al Viro9b067032014-11-07 13:52:07 -05001394static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001395{
1396 struct file *file = iocb->ki_filp;
1397 struct tun_file *tfile = file->private_data;
1398 struct tun_struct *tun = __tun_get(tfile);
Al Viro9b067032014-11-07 13:52:07 -05001399 ssize_t len = iov_iter_count(to), ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001400
1401 if (!tun)
1402 return -EBADFD;
Al Viro9b067032014-11-07 13:52:07 -05001403 ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK);
David S. Miller42404c02013-12-10 22:05:45 -05001404 ret = min_t(ssize_t, ret, len);
Zhi Yong Wud0b7da8a2013-12-06 14:16:51 +08001405 if (ret > 0)
1406 iocb->ki_pos = ret;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001407 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 return ret;
1409}
1410
Jason Wang96442e422012-10-31 19:46:02 +00001411static void tun_free_netdev(struct net_device *dev)
1412{
1413 struct tun_struct *tun = netdev_priv(dev);
1414
Jason Wang4008e972012-12-13 23:53:30 +00001415 BUG_ON(!(list_empty(&tun->disabled)));
Jason Wang96442e422012-10-31 19:46:02 +00001416 tun_flow_uninit(tun);
Paul Moore5dbbaf22013-01-14 07:12:19 +00001417 security_tun_dev_free_security(tun->security);
Jason Wang96442e422012-10-31 19:46:02 +00001418 free_netdev(dev);
1419}
1420
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421static void tun_setup(struct net_device *dev)
1422{
1423 struct tun_struct *tun = netdev_priv(dev);
1424
Eric W. Biederman0625c882012-02-07 16:48:55 -08001425 tun->owner = INVALID_UID;
1426 tun->group = INVALID_GID;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 dev->ethtool_ops = &tun_ethtool_ops;
Jason Wang96442e422012-10-31 19:46:02 +00001429 dev->destructor = tun_free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430}
1431
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001432/* Trivial set of netlink ops to allow deleting tun or tap
1433 * device with netlink.
1434 */
1435static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
1436{
1437 return -EINVAL;
1438}
1439
1440static struct rtnl_link_ops tun_link_ops __read_mostly = {
1441 .kind = DRV_NAME,
1442 .priv_size = sizeof(struct tun_struct),
1443 .setup = tun_setup,
1444 .validate = tun_validate,
1445};
1446
Herbert Xu33dccbb2009-02-05 21:25:32 -08001447static void tun_sock_write_space(struct sock *sk)
1448{
Jason Wang54f968d2012-10-31 19:45:57 +00001449 struct tun_file *tfile;
Eric Dumazet43815482010-04-29 11:01:49 +00001450 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001451
1452 if (!sock_writeable(sk))
1453 return;
1454
Herbert Xu33dccbb2009-02-05 21:25:32 -08001455 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
1456 return;
1457
Eric Dumazet43815482010-04-29 11:01:49 +00001458 wqueue = sk_sleep(sk);
1459 if (wqueue && waitqueue_active(wqueue))
1460 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001461 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -07001462
Jason Wang54f968d2012-10-31 19:45:57 +00001463 tfile = container_of(sk, struct tun_file, sk);
1464 kill_fasync(&tfile->fasync, SIGIO, POLL_OUT);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001465}
1466
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001467static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
1468 struct msghdr *m, size_t total_len)
1469{
Jason Wang54f968d2012-10-31 19:45:57 +00001470 int ret;
1471 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1472 struct tun_struct *tun = __tun_get(tfile);
1473
1474 if (!tun)
1475 return -EBADFD;
Al Virof5ff53b2014-06-19 15:36:49 -04001476
Al Viroc0371da2014-11-24 10:42:55 -05001477 ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04001478 m->msg_flags & MSG_DONTWAIT);
Jason Wang54f968d2012-10-31 19:45:57 +00001479 tun_put(tun);
1480 return ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001481}
1482
1483static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
1484 struct msghdr *m, size_t total_len,
1485 int flags)
1486{
Jason Wang54f968d2012-10-31 19:45:57 +00001487 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1488 struct tun_struct *tun = __tun_get(tfile);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001489 int ret;
Jason Wang54f968d2012-10-31 19:45:57 +00001490
1491 if (!tun)
1492 return -EBADFD;
1493
Richard Cochraneda29772013-07-19 19:40:10 +02001494 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) {
Gao feng3811ae72013-04-24 21:59:23 +00001495 ret = -EINVAL;
1496 goto out;
1497 }
Richard Cochraneda29772013-07-19 19:40:10 +02001498 if (flags & MSG_ERRQUEUE) {
1499 ret = sock_recv_errqueue(sock->sk, m, total_len,
1500 SOL_PACKET, TUN_TX_TIMESTAMP);
1501 goto out;
1502 }
Al Viroc0371da2014-11-24 10:42:55 -05001503 ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT);
David S. Miller42404c02013-12-10 22:05:45 -05001504 if (ret > total_len) {
1505 m->msg_flags |= MSG_TRUNC;
1506 ret = flags & MSG_TRUNC ? ret : total_len;
1507 }
Gao feng3811ae72013-04-24 21:59:23 +00001508out:
Jason Wang54f968d2012-10-31 19:45:57 +00001509 tun_put(tun);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001510 return ret;
1511}
1512
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001513static int tun_release(struct socket *sock)
1514{
1515 if (sock->sk)
1516 sock_put(sock->sk);
1517 return 0;
1518}
1519
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001520/* Ops structure to mimic raw sockets with tun */
1521static const struct proto_ops tun_socket_ops = {
1522 .sendmsg = tun_sendmsg,
1523 .recvmsg = tun_recvmsg,
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001524 .release = tun_release,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001525};
1526
Herbert Xu33dccbb2009-02-05 21:25:32 -08001527static struct proto tun_proto = {
1528 .name = "tun",
1529 .owner = THIS_MODULE,
Jason Wang54f968d2012-10-31 19:45:57 +00001530 .obj_size = sizeof(struct tun_file),
Herbert Xu33dccbb2009-02-05 21:25:32 -08001531};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001532
David Woodhouse980c9e82009-05-09 22:54:21 -07001533static int tun_flags(struct tun_struct *tun)
1534{
Michael S. Tsirkin031f5e032014-11-19 14:44:40 +02001535 return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP);
David Woodhouse980c9e82009-05-09 22:54:21 -07001536}
1537
1538static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1539 char *buf)
1540{
1541 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1542 return sprintf(buf, "0x%x\n", tun_flags(tun));
1543}
1544
1545static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1546 char *buf)
1547{
1548 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001549 return uid_valid(tun->owner)?
1550 sprintf(buf, "%u\n",
1551 from_kuid_munged(current_user_ns(), tun->owner)):
1552 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001553}
1554
1555static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1556 char *buf)
1557{
1558 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001559 return gid_valid(tun->group) ?
1560 sprintf(buf, "%u\n",
1561 from_kgid_munged(current_user_ns(), tun->group)):
1562 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001563}
1564
1565static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1566static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1567static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1568
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001569static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570{
1571 struct tun_struct *tun;
Jason Wang54f968d2012-10-31 19:45:57 +00001572 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 struct net_device *dev;
1574 int err;
1575
Jason Wang7c0c3b12013-01-11 16:59:33 +00001576 if (tfile->detached)
1577 return -EINVAL;
1578
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001579 dev = __dev_get_by_name(net, ifr->ifr_name);
1580 if (dev) {
David Woodhousef85ba782009-04-27 03:23:54 -07001581 if (ifr->ifr_flags & IFF_TUN_EXCL)
1582 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001583 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1584 tun = netdev_priv(dev);
1585 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1586 tun = netdev_priv(dev);
1587 else
1588 return -EINVAL;
1589
Jason Wang8e6d91a2013-05-28 18:32:11 +00001590 if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) !=
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001591 !!(tun->flags & IFF_MULTI_QUEUE))
Jason Wang8e6d91a2013-05-28 18:32:11 +00001592 return -EINVAL;
1593
Jason Wangcde8b152012-10-31 19:46:01 +00001594 if (tun_not_capable(tun))
Paul Moore2b980db2009-08-28 18:12:43 -04001595 return -EPERM;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001596 err = security_tun_dev_open(tun->security);
Paul Moore2b980db2009-08-28 18:12:43 -04001597 if (err < 0)
1598 return err;
1599
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001600 err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001601 if (err < 0)
1602 return err;
Jason Wang4008e972012-12-13 23:53:30 +00001603
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001604 if (tun->flags & IFF_MULTI_QUEUE &&
Jason Wange8dbad62013-04-22 20:40:39 +00001605 (tun->numqueues + tun->numdisabled > 1)) {
1606 /* One or more queue has already been attached, no need
1607 * to initialize the device again.
1608 */
1609 return 0;
1610 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001611 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 else {
1613 char *name;
1614 unsigned long flags = 0;
Jason Wangedfb6a12013-01-23 03:59:12 +00001615 int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
1616 MAX_TAP_QUEUES : 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617
Eric W. Biedermanc260b772012-11-18 21:34:11 +00001618 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001619 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001620 err = security_tun_dev_create();
1621 if (err < 0)
1622 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001623
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 /* Set dev type */
1625 if (ifr->ifr_flags & IFF_TUN) {
1626 /* TUN device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001627 flags |= IFF_TUN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 name = "tun%d";
1629 } else if (ifr->ifr_flags & IFF_TAP) {
1630 /* TAP device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001631 flags |= IFF_TAP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001633 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001634 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001635
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 if (*ifr->ifr_name)
1637 name = ifr->ifr_name;
1638
Jason Wangc8d68e62012-10-31 19:46:00 +00001639 dev = alloc_netdev_mqs(sizeof(struct tun_struct), name,
Tom Gundersenc835a672014-07-14 16:37:24 +02001640 NET_NAME_UNKNOWN, tun_setup, queues,
1641 queues);
Jason Wangedfb6a12013-01-23 03:59:12 +00001642
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 if (!dev)
1644 return -ENOMEM;
1645
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001646 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001647 dev->rtnl_link_ops = &tun_link_ops;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001648 dev->ifindex = tfile->ifindex;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001649
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 tun = netdev_priv(dev);
1651 tun->dev = dev;
1652 tun->flags = flags;
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -07001653 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001654 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655
Jason Wang54f968d2012-10-31 19:45:57 +00001656 tun->filter_attached = false;
1657 tun->sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001658
Jason Wang96442e422012-10-31 19:46:02 +00001659 spin_lock_init(&tun->lock);
1660
Paul Moore5dbbaf22013-01-14 07:12:19 +00001661 err = security_tun_dev_alloc_security(&tun->security);
1662 if (err < 0)
1663 goto err_free_dev;
Paul Moore2b980db2009-08-28 18:12:43 -04001664
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 tun_net_init(dev);
Pavel Emelyanov944a1372013-06-11 17:01:08 +04001666 tun_flow_init(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001667
Michał Mirosław88255372011-04-19 06:13:10 +00001668 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
Jason Wang6680ec62013-07-25 13:00:33 +08001669 TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
1670 NETIF_F_HW_VLAN_STAG_TX;
Michał Mirosław88255372011-04-19 06:13:10 +00001671 dev->features = dev->hw_features;
Fernando Luis Vazquez Cao6671b222014-02-18 21:20:09 +09001672 dev->vlan_features = dev->features &
1673 ~(NETIF_F_HW_VLAN_CTAG_TX |
1674 NETIF_F_HW_VLAN_STAG_TX);
Michał Mirosław88255372011-04-19 06:13:10 +00001675
Jason Wang4008e972012-12-13 23:53:30 +00001676 INIT_LIST_HEAD(&tun->disabled);
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001677 err = tun_attach(tun, file, false);
Jason Wangeb0fb362012-12-02 17:19:45 +00001678 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001679 goto err_free_flow;
Jason Wangeb0fb362012-12-02 17:19:45 +00001680
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 err = register_netdevice(tun->dev);
1682 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001683 goto err_detach;
Herbert Xu9c3fea62009-04-18 14:15:52 +00001684
David Woodhouse980c9e82009-05-09 22:54:21 -07001685 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1686 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1687 device_create_file(&tun->dev->dev, &dev_attr_group))
Joe Perches6b8a66e2011-03-02 07:18:10 +00001688 pr_err("Failed to create tun sysfs files\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 }
1690
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +00001691 netif_carrier_on(tun->dev);
1692
Joe Perches6b8a66e2011-03-02 07:18:10 +00001693 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
Michael S. Tsirkin031f5e032014-11-19 14:44:40 +02001695 tun->flags = (tun->flags & ~TUN_FEATURES) |
1696 (ifr->ifr_flags & TUN_FEATURES);
Jason Wangc8d68e62012-10-31 19:46:00 +00001697
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001698 /* Make sure persistent devices do not get stuck in
1699 * xoff state.
1700 */
1701 if (netif_running(tun->dev))
Jason Wangc8d68e62012-10-31 19:46:00 +00001702 netif_tx_wake_all_queues(tun->dev);
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001703
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 strcpy(ifr->ifr_name, tun->dev->name);
1705 return 0;
1706
Jason Wang662ca432013-09-11 18:09:48 +08001707err_detach:
1708 tun_detach_all(dev);
1709err_free_flow:
1710 tun_flow_uninit(tun);
1711 security_tun_dev_free_security(tun->security);
1712err_free_dev:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 return err;
1715}
1716
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001717static void tun_get_iff(struct net *net, struct tun_struct *tun,
Herbert Xu876bfd42009-08-06 14:22:44 +00001718 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001719{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001720 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001721
1722 strcpy(ifr->ifr_name, tun->dev->name);
1723
David Woodhouse980c9e82009-05-09 22:54:21 -07001724 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001725
Mark McLoughline3b99552008-08-15 15:09:56 -07001726}
1727
Rusty Russell5228ddc2008-07-03 03:46:16 -07001728/* This is like a cut-down ethtool ops, except done via tun fd so no
1729 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001730static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001731{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001732 netdev_features_t features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001733
1734 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001735 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001736 arg &= ~TUN_F_CSUM;
1737
1738 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1739 if (arg & TUN_F_TSO_ECN) {
1740 features |= NETIF_F_TSO_ECN;
1741 arg &= ~TUN_F_TSO_ECN;
1742 }
1743 if (arg & TUN_F_TSO4)
1744 features |= NETIF_F_TSO;
1745 if (arg & TUN_F_TSO6)
1746 features |= NETIF_F_TSO6;
1747 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1748 }
1749 }
1750
1751 /* This gives the user a way to test for new features in future by
1752 * trying to set them. */
1753 if (arg)
1754 return -EINVAL;
1755
Michał Mirosław88255372011-04-19 06:13:10 +00001756 tun->set_features = features;
1757 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001758
1759 return 0;
1760}
1761
Jason Wangc8d68e62012-10-31 19:46:00 +00001762static void tun_detach_filter(struct tun_struct *tun, int n)
1763{
1764 int i;
1765 struct tun_file *tfile;
1766
1767 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001768 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001769 sk_detach_filter(tfile->socket.sk);
1770 }
1771
1772 tun->filter_attached = false;
1773}
1774
1775static int tun_attach_filter(struct tun_struct *tun)
1776{
1777 int i, ret = 0;
1778 struct tun_file *tfile;
1779
1780 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001781 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001782 ret = sk_attach_filter(&tun->fprog, tfile->socket.sk);
1783 if (ret) {
1784 tun_detach_filter(tun, i);
1785 return ret;
1786 }
1787 }
1788
1789 tun->filter_attached = true;
1790 return ret;
1791}
1792
1793static void tun_set_sndbuf(struct tun_struct *tun)
1794{
1795 struct tun_file *tfile;
1796 int i;
1797
1798 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001799 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001800 tfile->socket.sk->sk_sndbuf = tun->sndbuf;
1801 }
1802}
1803
Jason Wangcde8b152012-10-31 19:46:01 +00001804static int tun_set_queue(struct file *file, struct ifreq *ifr)
1805{
1806 struct tun_file *tfile = file->private_data;
1807 struct tun_struct *tun;
Jason Wangcde8b152012-10-31 19:46:01 +00001808 int ret = 0;
1809
1810 rtnl_lock();
1811
1812 if (ifr->ifr_flags & IFF_ATTACH_QUEUE) {
Jason Wang4008e972012-12-13 23:53:30 +00001813 tun = tfile->detached;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001814 if (!tun) {
Jason Wangcde8b152012-10-31 19:46:01 +00001815 ret = -EINVAL;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001816 goto unlock;
1817 }
1818 ret = security_tun_dev_attach_queue(tun->security);
1819 if (ret < 0)
1820 goto unlock;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001821 ret = tun_attach(tun, file, false);
Jason Wang4008e972012-12-13 23:53:30 +00001822 } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001823 tun = rtnl_dereference(tfile->tun);
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001824 if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached)
Jason Wang4008e972012-12-13 23:53:30 +00001825 ret = -EINVAL;
1826 else
1827 __tun_detach(tfile, false);
1828 } else
Jason Wangcde8b152012-10-31 19:46:01 +00001829 ret = -EINVAL;
1830
Paul Moore5dbbaf22013-01-14 07:12:19 +00001831unlock:
Jason Wangcde8b152012-10-31 19:46:01 +00001832 rtnl_unlock();
1833 return ret;
1834}
1835
Arnd Bergmann50857e22009-11-06 22:52:32 -08001836static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1837 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001839 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001840 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841 void __user* argp = (void __user*)arg;
1842 struct ifreq ifr;
Eric W. Biederman0625c882012-02-07 16:48:55 -08001843 kuid_t owner;
1844 kgid_t group;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001845 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001846 int vnet_hdr_sz;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001847 unsigned int ifindex;
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02001848 int le;
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -07001849 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
Jason Wangcde8b152012-10-31 19:46:01 +00001851 if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) {
Arnd Bergmann50857e22009-11-06 22:52:32 -08001852 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 return -EFAULT;
David S. Miller8bbb1812012-07-30 14:52:48 -07001854 } else {
Mathias Krausea117dac2012-07-29 19:45:14 +00001855 memset(&ifr, 0, sizeof(ifr));
David S. Miller8bbb1812012-07-30 14:52:48 -07001856 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001857 if (cmd == TUNGETFEATURES) {
1858 /* Currently this just means: "what IFF flags are valid?".
1859 * This is needed because we never checked for invalid flags on
Michael S. Tsirkin031f5e032014-11-19 14:44:40 +02001860 * TUNSETIFF.
1861 */
1862 return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES,
Eric W. Biederman631ab462009-01-20 11:00:40 +00001863 (unsigned int __user*)argp);
Jason Wangcde8b152012-10-31 19:46:01 +00001864 } else if (cmd == TUNSETQUEUE)
1865 return tun_set_queue(file, &ifr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001866
Jason Wangc8d68e62012-10-31 19:46:00 +00001867 ret = 0;
Herbert Xu876bfd42009-08-06 14:22:44 +00001868 rtnl_lock();
1869
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001870 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1873
Herbert Xu876bfd42009-08-06 14:22:44 +00001874 ret = tun_set_iff(tfile->net, file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875
Herbert Xu876bfd42009-08-06 14:22:44 +00001876 if (ret)
1877 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878
Arnd Bergmann50857e22009-11-06 22:52:32 -08001879 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001880 ret = -EFAULT;
1881 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 }
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001883 if (cmd == TUNSETIFINDEX) {
1884 ret = -EPERM;
1885 if (tun)
1886 goto unlock;
1887
1888 ret = -EFAULT;
1889 if (copy_from_user(&ifindex, argp, sizeof(ifindex)))
1890 goto unlock;
1891
1892 ret = 0;
1893 tfile->ifindex = ifindex;
1894 goto unlock;
1895 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896
Herbert Xu876bfd42009-08-06 14:22:44 +00001897 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001899 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900
Jason Wang1e588332012-10-31 19:45:56 +00001901 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902
Eric W. Biederman631ab462009-01-20 11:00:40 +00001903 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001905 case TUNGETIFF:
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001906 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001907
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001908 if (tfile->detached)
1909 ifr.ifr_flags |= IFF_DETACH_QUEUE;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001910 if (!tfile->socket.sk->sk_filter)
1911 ifr.ifr_flags |= IFF_NOFILTER;
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001912
Arnd Bergmann50857e22009-11-06 22:52:32 -08001913 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001914 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001915 break;
1916
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 case TUNSETNOCSUM:
1918 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
Michał Mirosław88255372011-04-19 06:13:10 +00001920 /* [unimplemented] */
1921 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00001922 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 break;
1924
1925 case TUNSETPERSIST:
Jason Wang54f968d2012-10-31 19:45:57 +00001926 /* Disable/Enable persist mode. Keep an extra reference to the
1927 * module to prevent the module being unprobed.
1928 */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001929 if (arg && !(tun->flags & IFF_PERSIST)) {
1930 tun->flags |= IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001931 __module_get(THIS_MODULE);
Jason Wangdd38bd82013-01-11 16:59:34 +00001932 }
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001933 if (!arg && (tun->flags & IFF_PERSIST)) {
1934 tun->flags &= ~IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001935 module_put(THIS_MODULE);
1936 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937
Joe Perches6b8a66e2011-03-02 07:18:10 +00001938 tun_debug(KERN_INFO, tun, "persist %s\n",
1939 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 break;
1941
1942 case TUNSETOWNER:
1943 /* Set owner of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001944 owner = make_kuid(current_user_ns(), arg);
1945 if (!uid_valid(owner)) {
1946 ret = -EINVAL;
1947 break;
1948 }
1949 tun->owner = owner;
Jason Wang1e588332012-10-31 19:45:56 +00001950 tun_debug(KERN_INFO, tun, "owner set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08001951 from_kuid(&init_user_ns, tun->owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 break;
1953
Guido Guenther8c644622007-07-02 22:50:25 -07001954 case TUNSETGROUP:
1955 /* Set group of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001956 group = make_kgid(current_user_ns(), arg);
1957 if (!gid_valid(group)) {
1958 ret = -EINVAL;
1959 break;
1960 }
1961 tun->group = group;
Jason Wang1e588332012-10-31 19:45:56 +00001962 tun_debug(KERN_INFO, tun, "group set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08001963 from_kgid(&init_user_ns, tun->group));
Guido Guenther8c644622007-07-02 22:50:25 -07001964 break;
1965
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001966 case TUNSETLINK:
1967 /* Only allow setting the type when the interface is down */
1968 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001969 tun_debug(KERN_INFO, tun,
1970 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07001971 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001972 } else {
1973 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00001974 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1975 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07001976 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001977 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001978 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001979
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980#ifdef TUN_DEBUG
1981 case TUNSETDEBUG:
1982 tun->debug = arg;
1983 break;
1984#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07001985 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00001986 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001987 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001988
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -07001989 case TUNSETTXFILTER:
1990 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00001991 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001992 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Eric W. Biederman631ab462009-01-20 11:00:40 +00001993 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07001994 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001995 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996
1997 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04001998 /* Get hw address */
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -07001999 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
2000 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08002001 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00002002 ret = -EFAULT;
2003 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004
2005 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2cc2008-07-14 22:18:19 -07002006 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00002007 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
2008 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08002009
Kim B. Heino40102372008-02-29 12:26:21 -08002010 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002011 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002012
2013 case TUNGETSNDBUF:
Jason Wang54f968d2012-10-31 19:45:57 +00002014 sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002015 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
2016 ret = -EFAULT;
2017 break;
2018
2019 case TUNSETSNDBUF:
2020 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
2021 ret = -EFAULT;
2022 break;
2023 }
2024
Jason Wangc8d68e62012-10-31 19:46:00 +00002025 tun->sndbuf = sndbuf;
2026 tun_set_sndbuf(tun);
Herbert Xu33dccbb2009-02-05 21:25:32 -08002027 break;
2028
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02002029 case TUNGETVNETHDRSZ:
2030 vnet_hdr_sz = tun->vnet_hdr_sz;
2031 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
2032 ret = -EFAULT;
2033 break;
2034
2035 case TUNSETVNETHDRSZ:
2036 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
2037 ret = -EFAULT;
2038 break;
2039 }
2040 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
2041 ret = -EINVAL;
2042 break;
2043 }
2044
2045 tun->vnet_hdr_sz = vnet_hdr_sz;
2046 break;
2047
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02002048 case TUNGETVNETLE:
2049 le = !!(tun->flags & TUN_VNET_LE);
2050 if (put_user(le, (int __user *)argp))
2051 ret = -EFAULT;
2052 break;
2053
2054 case TUNSETVNETLE:
2055 if (get_user(le, (int __user *)argp)) {
2056 ret = -EFAULT;
2057 break;
2058 }
2059 if (le)
2060 tun->flags |= TUN_VNET_LE;
2061 else
2062 tun->flags &= ~TUN_VNET_LE;
2063 break;
2064
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002065 case TUNATTACHFILTER:
2066 /* Can be set only for TAPs */
2067 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002068 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002069 break;
2070 ret = -EFAULT;
Jason Wang54f968d2012-10-31 19:45:57 +00002071 if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog)))
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002072 break;
2073
Jason Wangc8d68e62012-10-31 19:46:00 +00002074 ret = tun_attach_filter(tun);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002075 break;
2076
2077 case TUNDETACHFILTER:
2078 /* Can be set only for TAPs */
2079 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002080 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002081 break;
Jason Wangc8d68e62012-10-31 19:46:00 +00002082 ret = 0;
2083 tun_detach_filter(tun, tun->numqueues);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002084 break;
2085
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002086 case TUNGETFILTER:
2087 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002088 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002089 break;
2090 ret = -EFAULT;
2091 if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog)))
2092 break;
2093 ret = 0;
2094 break;
2095
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00002097 ret = -EINVAL;
2098 break;
Joe Perchesee289b62010-05-17 22:47:34 -07002099 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100
Herbert Xu876bfd42009-08-06 14:22:44 +00002101unlock:
2102 rtnl_unlock();
2103 if (tun)
2104 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002105 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106}
2107
Arnd Bergmann50857e22009-11-06 22:52:32 -08002108static long tun_chr_ioctl(struct file *file,
2109 unsigned int cmd, unsigned long arg)
2110{
2111 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
2112}
2113
2114#ifdef CONFIG_COMPAT
2115static long tun_chr_compat_ioctl(struct file *file,
2116 unsigned int cmd, unsigned long arg)
2117{
2118 switch (cmd) {
2119 case TUNSETIFF:
2120 case TUNGETIFF:
2121 case TUNSETTXFILTER:
2122 case TUNGETSNDBUF:
2123 case TUNSETSNDBUF:
2124 case SIOCGIFHWADDR:
2125 case SIOCSIFHWADDR:
2126 arg = (unsigned long)compat_ptr(arg);
2127 break;
2128 default:
2129 arg = (compat_ulong_t)arg;
2130 break;
2131 }
2132
2133 /*
2134 * compat_ifreq is shorter than ifreq, so we must not access beyond
2135 * the end of that structure. All fields that are used in this
2136 * driver are compatible though, we don't need to convert the
2137 * contents.
2138 */
2139 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
2140}
2141#endif /* CONFIG_COMPAT */
2142
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143static int tun_chr_fasync(int fd, struct file *file, int on)
2144{
Jason Wang54f968d2012-10-31 19:45:57 +00002145 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 int ret;
2147
Jason Wang54f968d2012-10-31 19:45:57 +00002148 if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06002149 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002150
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 if (on) {
Jeff Laytone0b93ed2014-08-22 11:27:32 -04002152 __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Jason Wang54f968d2012-10-31 19:45:57 +00002153 tfile->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002154 } else
Jason Wang54f968d2012-10-31 19:45:57 +00002155 tfile->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06002156 ret = 0;
2157out:
Jonathan Corbet9d319522008-06-19 15:50:37 -06002158 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159}
2160
2161static int tun_chr_open(struct inode *inode, struct file * file)
2162{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002163 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07002164
Joe Perches6b8a66e2011-03-02 07:18:10 +00002165 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00002166
Jason Wang54f968d2012-10-31 19:45:57 +00002167 tfile = (struct tun_file *)sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL,
2168 &tun_proto);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002169 if (!tfile)
2170 return -ENOMEM;
Monam Agarwalc9566742014-03-24 00:02:32 +05302171 RCU_INIT_POINTER(tfile->tun, NULL);
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00002172 tfile->net = get_net(current->nsproxy->net_ns);
Jason Wang54f968d2012-10-31 19:45:57 +00002173 tfile->flags = 0;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002174 tfile->ifindex = 0;
Jason Wang54f968d2012-10-31 19:45:57 +00002175
Jason Wang54f968d2012-10-31 19:45:57 +00002176 init_waitqueue_head(&tfile->wq.wait);
Xi Wang9e641bd2014-05-16 15:11:48 -07002177 RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq);
Jason Wang54f968d2012-10-31 19:45:57 +00002178
2179 tfile->socket.file = file;
2180 tfile->socket.ops = &tun_socket_ops;
2181
2182 sock_init_data(&tfile->socket, &tfile->sk);
2183 sk_change_net(&tfile->sk, tfile->net);
2184
2185 tfile->sk.sk_write_space = tun_sock_write_space;
2186 tfile->sk.sk_sndbuf = INT_MAX;
2187
Eric W. Biederman631ab462009-01-20 11:00:40 +00002188 file->private_data = tfile;
Jason Wang54f968d2012-10-31 19:45:57 +00002189 set_bit(SOCK_EXTERNALLY_ALLOCATED, &tfile->socket.flags);
Jason Wang4008e972012-12-13 23:53:30 +00002190 INIT_LIST_HEAD(&tfile->next);
Jason Wang54f968d2012-10-31 19:45:57 +00002191
Jason Wang19a6afb2013-06-08 14:17:41 +08002192 sock_set_flag(&tfile->sk, SOCK_ZEROCOPY);
2193
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 return 0;
2195}
2196
2197static int tun_chr_close(struct inode *inode, struct file *file)
2198{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002199 struct tun_file *tfile = file->private_data;
Jason Wang54f968d2012-10-31 19:45:57 +00002200 struct net *net = tfile->net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201
Jason Wangc8d68e62012-10-31 19:46:00 +00002202 tun_detach(tfile, true);
Jason Wang54f968d2012-10-31 19:45:57 +00002203 put_net(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204
2205 return 0;
2206}
2207
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002208#ifdef CONFIG_PROC_FS
Joe Perchesa3816ab2014-09-29 16:08:25 -07002209static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f)
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002210{
2211 struct tun_struct *tun;
2212 struct ifreq ifr;
2213
2214 memset(&ifr, 0, sizeof(ifr));
2215
2216 rtnl_lock();
2217 tun = tun_get(f);
2218 if (tun)
2219 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
2220 rtnl_unlock();
2221
2222 if (tun)
2223 tun_put(tun);
2224
Joe Perchesa3816ab2014-09-29 16:08:25 -07002225 seq_printf(m, "iff:\t%s\n", ifr.ifr_name);
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002226}
2227#endif
2228
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08002229static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002230 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 .llseek = no_llseek,
Al Viro9b067032014-11-07 13:52:07 -05002232 .read = new_sync_read,
Al Virof5ff53b2014-06-19 15:36:49 -04002233 .write = new_sync_write,
Al Viro9b067032014-11-07 13:52:07 -05002234 .read_iter = tun_chr_read_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04002235 .write_iter = tun_chr_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08002237 .unlocked_ioctl = tun_chr_ioctl,
2238#ifdef CONFIG_COMPAT
2239 .compat_ioctl = tun_chr_compat_ioctl,
2240#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 .open = tun_chr_open,
2242 .release = tun_chr_close,
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002243 .fasync = tun_chr_fasync,
2244#ifdef CONFIG_PROC_FS
2245 .show_fdinfo = tun_chr_show_fdinfo,
2246#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247};
2248
2249static struct miscdevice tun_miscdev = {
2250 .minor = TUN_MINOR,
2251 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02002252 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254};
2255
2256/* ethtool interface */
2257
2258static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
2259{
2260 cmd->supported = 0;
2261 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00002262 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263 cmd->duplex = DUPLEX_FULL;
2264 cmd->port = PORT_TP;
2265 cmd->phy_address = 0;
2266 cmd->transceiver = XCVR_INTERNAL;
2267 cmd->autoneg = AUTONEG_DISABLE;
2268 cmd->maxtxpkt = 0;
2269 cmd->maxrxpkt = 0;
2270 return 0;
2271}
2272
2273static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
2274{
2275 struct tun_struct *tun = netdev_priv(dev);
2276
Rick Jones33a5ba12011-11-15 14:59:53 +00002277 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
2278 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279
2280 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002281 case IFF_TUN:
Rick Jones33a5ba12011-11-15 14:59:53 +00002282 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002284 case IFF_TAP:
Rick Jones33a5ba12011-11-15 14:59:53 +00002285 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 break;
2287 }
2288}
2289
2290static u32 tun_get_msglevel(struct net_device *dev)
2291{
2292#ifdef TUN_DEBUG
2293 struct tun_struct *tun = netdev_priv(dev);
2294 return tun->debug;
2295#else
2296 return -EOPNOTSUPP;
2297#endif
2298}
2299
2300static void tun_set_msglevel(struct net_device *dev, u32 value)
2301{
2302#ifdef TUN_DEBUG
2303 struct tun_struct *tun = netdev_priv(dev);
2304 tun->debug = value;
2305#endif
2306}
2307
Jeff Garzik7282d492006-09-13 14:30:00 -04002308static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 .get_settings = tun_get_settings,
2310 .get_drvinfo = tun_get_drvinfo,
2311 .get_msglevel = tun_get_msglevel,
2312 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00002313 .get_link = ethtool_op_get_link,
Richard Cochraneda29772013-07-19 19:40:10 +02002314 .get_ts_info = ethtool_op_get_ts_info,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315};
2316
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002317
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318static int __init tun_init(void)
2319{
2320 int ret = 0;
2321
Joe Perches6b8a66e2011-03-02 07:18:10 +00002322 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
2323 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002325 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002326 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002327 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002328 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002329 }
2330
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002332 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002333 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002334 goto err_misc;
2335 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002336 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002337err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002338 rtnl_link_unregister(&tun_link_ops);
2339err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340 return ret;
2341}
2342
2343static void tun_cleanup(void)
2344{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002345 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002346 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347}
2348
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002349/* Get an underlying socket object from tun file. Returns error unless file is
2350 * attached to a device. The returned object works like a packet socket, it
2351 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
2352 * holding a reference to the file for as long as the socket is in use. */
2353struct socket *tun_get_socket(struct file *file)
2354{
Jason Wang6e914fc2012-10-31 19:45:58 +00002355 struct tun_file *tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002356 if (file->f_op != &tun_fops)
2357 return ERR_PTR(-EINVAL);
Jason Wang6e914fc2012-10-31 19:45:58 +00002358 tfile = file->private_data;
2359 if (!tfile)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002360 return ERR_PTR(-EBADFD);
Jason Wang54f968d2012-10-31 19:45:57 +00002361 return &tfile->socket;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002362}
2363EXPORT_SYMBOL_GPL(tun_get_socket);
2364
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365module_init(tun_init);
2366module_exit(tun_cleanup);
2367MODULE_DESCRIPTION(DRV_DESCRIPTION);
2368MODULE_AUTHOR(DRV_COPYRIGHT);
2369MODULE_LICENSE("GPL");
2370MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02002371MODULE_ALIAS("devname:net/tun");