blob: a74661690a11798c39cb35d945b34f39d78862c1 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
16 */
17
18/*
19 * Changes:
20 *
Mike Kershawff4cc3a2005-09-01 17:40:05 -070021 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
23 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 * Mark Smith <markzzzsmith@yahoo.com.au>
Joe Perches344dc8e2012-07-12 19:33:09 +000025 * Use eth_random_addr() for tap MAC address.
Linus Torvalds1da177e2005-04-16 15:20:36 -070026 *
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
30 * Added ethtool API.
31 * Minor cleanups
32 *
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
35 */
36
Joe Perches6b8a66e2011-03-02 07:18:10 +000037#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#define DRV_NAME "tun"
40#define DRV_VERSION "1.6"
41#define DRV_DESCRIPTION "Universal TUN/TAP device driver"
42#define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/module.h>
45#include <linux/errno.h>
46#include <linux/kernel.h>
47#include <linux/major.h>
48#include <linux/slab.h>
49#include <linux/poll.h>
50#include <linux/fcntl.h>
51#include <linux/init.h>
52#include <linux/skbuff.h>
53#include <linux/netdevice.h>
54#include <linux/etherdevice.h>
55#include <linux/miscdevice.h>
56#include <linux/ethtool.h>
57#include <linux/rtnetlink.h>
Arnd Bergmann50857e22009-11-06 22:52:32 -080058#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/if.h>
60#include <linux/if_arp.h>
61#include <linux/if_ether.h>
62#include <linux/if_tun.h>
Jason Wang6680ec62013-07-25 13:00:33 +080063#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064#include <linux/crc32.h>
Pavel Emelyanovd647a592008-04-16 00:41:16 -070065#include <linux/nsproxy.h>
Rusty Russellf43798c2008-07-03 03:48:02 -070066#include <linux/virtio_net.h>
Michael S. Tsirkin99405162010-02-14 01:01:10 +000067#include <linux/rcupdate.h>
Eric W. Biederman881d9662007-09-17 11:56:21 -070068#include <net/net_namespace.h>
Pavel Emelyanov79d17602008-04-16 00:40:46 -070069#include <net/netns/generic.h>
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -080070#include <net/rtnetlink.h>
Herbert Xu33dccbb2009-02-05 21:25:32 -080071#include <net/sock.h>
Masatake YAMATO93e14b62014-01-29 16:43:31 +090072#include <linux/seq_file.h>
Herbert Xue0b46d02014-11-07 21:22:23 +080073#include <linux/uio.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <asm/uaccess.h>
76
Rusty Russell14daa022008-04-12 18:48:58 -070077/* Uncomment to enable debugging */
78/* #define TUN_DEBUG 1 */
79
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#ifdef TUN_DEBUG
81static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070082
Joe Perches6b8a66e2011-03-02 07:18:10 +000083#define tun_debug(level, tun, fmt, args...) \
84do { \
85 if (tun->debug) \
86 netdev_printk(level, tun->dev, fmt, ##args); \
87} while (0)
88#define DBG1(level, fmt, args...) \
89do { \
90 if (debug == 2) \
91 printk(level fmt, ##args); \
92} while (0)
Rusty Russell14daa022008-04-12 18:48:58 -070093#else
Joe Perches6b8a66e2011-03-02 07:18:10 +000094#define tun_debug(level, tun, fmt, args...) \
95do { \
96 if (0) \
97 netdev_printk(level, tun->dev, fmt, ##args); \
98} while (0)
99#define DBG1(level, fmt, args...) \
100do { \
101 if (0) \
102 printk(level fmt, ##args); \
103} while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#endif
105
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +0200106/* TUN device flags */
107
108/* IFF_ATTACH_QUEUE is never stored in device flags,
109 * overload it to mean fasync when stored there.
110 */
111#define TUN_FASYNC IFF_ATTACH_QUEUE
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200112/* High bits in flags field are unused. */
113#define TUN_VNET_LE 0x80000000
Greg Kurz8b8e6582015-04-24 14:50:36 +0200114#define TUN_VNET_BE 0x40000000
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +0200115
116#define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200117 IFF_MULTI_QUEUE)
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000118#define GOODCOPY_LEN 128
119
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700120#define FLT_EXACT_COUNT 8
121struct tap_filter {
122 unsigned int count; /* Number of addrs. Zero means disabled */
123 u32 mask[2]; /* Mask of the hashed addrs */
124 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
125};
126
Pankaj Guptabaf71c52015-01-12 11:41:29 +0530127/* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal
128 * to max number of VCPUs in guest. */
129#define MAX_TAP_QUEUES 256
Jason Wangb8732fb2013-01-23 03:59:13 +0000130#define MAX_TAP_FLOWS 4096
Jason Wangc8d68e62012-10-31 19:46:00 +0000131
Jason Wang96442e422012-10-31 19:46:02 +0000132#define TUN_FLOW_EXPIRE (3 * HZ)
133
Jason Wang54f968d2012-10-31 19:45:57 +0000134/* A tun_file connects an open character device to a tuntap netdevice. It
stephen hemminger92d4ea62013-12-05 20:42:58 -0800135 * also contains all socket related structures (except sock_fprog and tap_filter)
Jason Wang54f968d2012-10-31 19:45:57 +0000136 * to serve as one transmit queue for tuntap device. The sock_fprog and
137 * tap_filter were kept in tun_struct since they were used for filtering for the
Rami Rosen36fe8c092012-11-25 22:07:40 +0000138 * netdevice not for a specific queue (at least I didn't see the requirement for
Jason Wang54f968d2012-10-31 19:45:57 +0000139 * this).
Jason Wang6e914fc2012-10-31 19:45:58 +0000140 *
141 * RCU usage:
Rami Rosen36fe8c092012-11-25 22:07:40 +0000142 * The tun_file and tun_struct are loosely coupled, the pointer from one to the
Jason Wang6e914fc2012-10-31 19:45:58 +0000143 * other can only be read while rcu_read_lock or rtnl_lock is held.
Jason Wang54f968d2012-10-31 19:45:57 +0000144 */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000145struct tun_file {
Jason Wang54f968d2012-10-31 19:45:57 +0000146 struct sock sk;
147 struct socket socket;
148 struct socket_wq wq;
Jason Wang6e914fc2012-10-31 19:45:58 +0000149 struct tun_struct __rcu *tun;
Jason Wang54f968d2012-10-31 19:45:57 +0000150 struct fasync_struct *fasync;
151 /* only used for fasnyc */
152 unsigned int flags;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +0400153 union {
154 u16 queue_index;
155 unsigned int ifindex;
156 };
Jason Wang4008e972012-12-13 23:53:30 +0000157 struct list_head next;
158 struct tun_struct *detached;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000159};
160
Jason Wang96442e422012-10-31 19:46:02 +0000161struct tun_flow_entry {
162 struct hlist_node hash_link;
163 struct rcu_head rcu;
164 struct tun_struct *tun;
165
166 u32 rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800167 u32 rps_rxhash;
Jason Wang96442e422012-10-31 19:46:02 +0000168 int queue_index;
169 unsigned long updated;
170};
171
172#define TUN_NUM_FLOW_ENTRIES 1024
173
Jason Wang54f968d2012-10-31 19:45:57 +0000174/* Since the socket were moved to tun_file, to preserve the behavior of persist
Rami Rosen36fe8c092012-11-25 22:07:40 +0000175 * device, socket filter, sndbuf and vnet header size were restore when the
Jason Wang54f968d2012-10-31 19:45:57 +0000176 * file were attached to a persist device.
177 */
Rusty Russell14daa022008-04-12 18:48:58 -0700178struct tun_struct {
Jason Wangc8d68e62012-10-31 19:46:00 +0000179 struct tun_file __rcu *tfiles[MAX_TAP_QUEUES];
180 unsigned int numqueues;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700181 unsigned int flags;
Eric W. Biederman0625c882012-02-07 16:48:55 -0800182 kuid_t owner;
183 kgid_t group;
Rusty Russell14daa022008-04-12 18:48:58 -0700184
Rusty Russell14daa022008-04-12 18:48:58 -0700185 struct net_device *dev;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000186 netdev_features_t set_features;
Michał Mirosław88255372011-04-19 06:13:10 +0000187#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
Vlad Yaseviche3e3c422015-02-03 16:36:17 -0500188 NETIF_F_TSO6|NETIF_F_UFO)
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200189
Paolo Abenieaea34b2016-02-26 10:45:40 +0100190 int align;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200191 int vnet_hdr_sz;
Jason Wang54f968d2012-10-31 19:45:57 +0000192 int sndbuf;
193 struct tap_filter txflt;
194 struct sock_fprog fprog;
195 /* protected by rtnl lock */
196 bool filter_attached;
Rusty Russell14daa022008-04-12 18:48:58 -0700197#ifdef TUN_DEBUG
198 int debug;
199#endif
Jason Wang96442e422012-10-31 19:46:02 +0000200 spinlock_t lock;
Jason Wang96442e422012-10-31 19:46:02 +0000201 struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
202 struct timer_list flow_gc_timer;
203 unsigned long ageing_time;
Jason Wang4008e972012-12-13 23:53:30 +0000204 unsigned int numdisabled;
205 struct list_head disabled;
Paul Moore5dbbaf22013-01-14 07:12:19 +0000206 void *security;
Jason Wangb8732fb2013-01-23 03:59:13 +0000207 u32 flow_count;
Rusty Russell14daa022008-04-12 18:48:58 -0700208};
209
Greg Kurz8b8e6582015-04-24 14:50:36 +0200210#ifdef CONFIG_TUN_VNET_CROSS_LE
211static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
212{
213 return tun->flags & TUN_VNET_BE ? false :
214 virtio_legacy_is_little_endian();
215}
216
217static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
218{
219 int be = !!(tun->flags & TUN_VNET_BE);
220
221 if (put_user(be, argp))
222 return -EFAULT;
223
224 return 0;
225}
226
227static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
228{
229 int be;
230
231 if (get_user(be, argp))
232 return -EFAULT;
233
234 if (be)
235 tun->flags |= TUN_VNET_BE;
236 else
237 tun->flags &= ~TUN_VNET_BE;
238
239 return 0;
240}
241#else
242static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
243{
244 return virtio_legacy_is_little_endian();
245}
246
247static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
248{
249 return -EINVAL;
250}
251
252static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
253{
254 return -EINVAL;
255}
256#endif /* CONFIG_TUN_VNET_CROSS_LE */
257
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200258static inline bool tun_is_little_endian(struct tun_struct *tun)
259{
Greg Kurz7d824102015-04-24 14:26:24 +0200260 return tun->flags & TUN_VNET_LE ||
Greg Kurz8b8e6582015-04-24 14:50:36 +0200261 tun_legacy_is_little_endian(tun);
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200262}
263
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300264static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val)
265{
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200266 return __virtio16_to_cpu(tun_is_little_endian(tun), val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300267}
268
269static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val)
270{
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200271 return __cpu_to_virtio16(tun_is_little_endian(tun), val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300272}
273
Jason Wang96442e422012-10-31 19:46:02 +0000274static inline u32 tun_hashfn(u32 rxhash)
275{
276 return rxhash & 0x3ff;
277}
278
279static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash)
280{
281 struct tun_flow_entry *e;
Jason Wang96442e422012-10-31 19:46:02 +0000282
Sasha Levinb67bfe02013-02-27 17:06:00 -0800283 hlist_for_each_entry_rcu(e, head, hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000284 if (e->rxhash == rxhash)
285 return e;
286 }
287 return NULL;
288}
289
290static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun,
291 struct hlist_head *head,
292 u32 rxhash, u16 queue_index)
293{
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000294 struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC);
295
Jason Wang96442e422012-10-31 19:46:02 +0000296 if (e) {
297 tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n",
298 rxhash, queue_index);
299 e->updated = jiffies;
300 e->rxhash = rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800301 e->rps_rxhash = 0;
Jason Wang96442e422012-10-31 19:46:02 +0000302 e->queue_index = queue_index;
303 e->tun = tun;
304 hlist_add_head_rcu(&e->hash_link, head);
Jason Wangb8732fb2013-01-23 03:59:13 +0000305 ++tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000306 }
307 return e;
308}
309
Jason Wang96442e422012-10-31 19:46:02 +0000310static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e)
311{
312 tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n",
313 e->rxhash, e->queue_index);
314 hlist_del_rcu(&e->hash_link);
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000315 kfree_rcu(e, rcu);
Jason Wangb8732fb2013-01-23 03:59:13 +0000316 --tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000317}
318
319static void tun_flow_flush(struct tun_struct *tun)
320{
321 int i;
322
323 spin_lock_bh(&tun->lock);
324 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
325 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800326 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000327
Sasha Levinb67bfe02013-02-27 17:06:00 -0800328 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
Jason Wang96442e422012-10-31 19:46:02 +0000329 tun_flow_delete(tun, e);
330 }
331 spin_unlock_bh(&tun->lock);
332}
333
334static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
335{
336 int i;
337
338 spin_lock_bh(&tun->lock);
339 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
340 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800341 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000342
Sasha Levinb67bfe02013-02-27 17:06:00 -0800343 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000344 if (e->queue_index == queue_index)
345 tun_flow_delete(tun, e);
346 }
347 }
348 spin_unlock_bh(&tun->lock);
349}
350
351static void tun_flow_cleanup(unsigned long data)
352{
353 struct tun_struct *tun = (struct tun_struct *)data;
354 unsigned long delay = tun->ageing_time;
355 unsigned long next_timer = jiffies + delay;
356 unsigned long count = 0;
357 int i;
358
359 tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n");
360
361 spin_lock_bh(&tun->lock);
362 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
363 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800364 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000365
Sasha Levinb67bfe02013-02-27 17:06:00 -0800366 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000367 unsigned long this_timer;
368 count++;
369 this_timer = e->updated + delay;
370 if (time_before_eq(this_timer, jiffies))
371 tun_flow_delete(tun, e);
372 else if (time_before(this_timer, next_timer))
373 next_timer = this_timer;
374 }
375 }
376
377 if (count)
378 mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer));
379 spin_unlock_bh(&tun->lock);
380}
381
Eric Dumazet49974422012-12-12 19:22:57 +0000382static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
Jason Wang9e857222013-01-28 01:05:19 +0000383 struct tun_file *tfile)
Jason Wang96442e422012-10-31 19:46:02 +0000384{
385 struct hlist_head *head;
386 struct tun_flow_entry *e;
387 unsigned long delay = tun->ageing_time;
Jason Wang9e857222013-01-28 01:05:19 +0000388 u16 queue_index = tfile->queue_index;
Jason Wang96442e422012-10-31 19:46:02 +0000389
390 if (!rxhash)
391 return;
392 else
393 head = &tun->flows[tun_hashfn(rxhash)];
394
395 rcu_read_lock();
396
Jason Wang9e857222013-01-28 01:05:19 +0000397 /* We may get a very small possibility of OOO during switching, not
398 * worth to optimize.*/
399 if (tun->numqueues == 1 || tfile->detached)
Jason Wang96442e422012-10-31 19:46:02 +0000400 goto unlock;
401
402 e = tun_flow_find(head, rxhash);
403 if (likely(e)) {
404 /* TODO: keep queueing to old queue until it's empty? */
405 e->queue_index = queue_index;
406 e->updated = jiffies;
Tom Herbert9bc88932013-12-22 18:54:32 +0800407 sock_rps_record_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000408 } else {
409 spin_lock_bh(&tun->lock);
Jason Wangb8732fb2013-01-23 03:59:13 +0000410 if (!tun_flow_find(head, rxhash) &&
411 tun->flow_count < MAX_TAP_FLOWS)
Jason Wang96442e422012-10-31 19:46:02 +0000412 tun_flow_create(tun, head, rxhash, queue_index);
413
414 if (!timer_pending(&tun->flow_gc_timer))
415 mod_timer(&tun->flow_gc_timer,
416 round_jiffies_up(jiffies + delay));
417 spin_unlock_bh(&tun->lock);
418 }
419
420unlock:
421 rcu_read_unlock();
422}
423
Tom Herbert9bc88932013-12-22 18:54:32 +0800424/**
425 * Save the hash received in the stack receive path and update the
426 * flow_hash table accordingly.
427 */
428static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash)
429{
Eric Dumazet567e4b72015-02-06 12:59:01 -0800430 if (unlikely(e->rps_rxhash != hash))
Tom Herbert9bc88932013-12-22 18:54:32 +0800431 e->rps_rxhash = hash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800432}
433
Jason Wangc8d68e62012-10-31 19:46:00 +0000434/* We try to identify a flow through its rxhash first. The reason that
stephen hemminger92d4ea62013-12-05 20:42:58 -0800435 * we do not check rxq no. is because some cards(e.g 82599), chooses
Jason Wangc8d68e62012-10-31 19:46:00 +0000436 * the rxq based on the txq where the last packet of the flow comes. As
437 * the userspace application move between processors, we may get a
438 * different rxq no. here. If we could not get rxhash, then we would
439 * hope the rxq no. may help here.
440 */
Jason Wangf663dd92014-01-10 16:18:26 +0800441static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100442 void *accel_priv, select_queue_fallback_t fallback)
Jason Wangc8d68e62012-10-31 19:46:00 +0000443{
444 struct tun_struct *tun = netdev_priv(dev);
Jason Wang96442e422012-10-31 19:46:02 +0000445 struct tun_flow_entry *e;
Jason Wangc8d68e62012-10-31 19:46:00 +0000446 u32 txq = 0;
447 u32 numqueues = 0;
448
449 rcu_read_lock();
Jason Wang92bb73e2013-06-05 16:44:57 +0800450 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000451
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800452 txq = skb_get_hash(skb);
Jason Wangc8d68e62012-10-31 19:46:00 +0000453 if (txq) {
Jason Wang96442e422012-10-31 19:46:02 +0000454 e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
Tom Herbert9bc88932013-12-22 18:54:32 +0800455 if (e) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800456 tun_flow_save_rps_rxhash(e, txq);
Zhi Yong Wufbe4d452014-01-02 13:24:28 +0800457 txq = e->queue_index;
Tom Herbert9bc88932013-12-22 18:54:32 +0800458 } else
Jason Wang96442e422012-10-31 19:46:02 +0000459 /* use multiply and shift instead of expensive divide */
460 txq = ((u64)txq * numqueues) >> 32;
Jason Wangc8d68e62012-10-31 19:46:00 +0000461 } else if (likely(skb_rx_queue_recorded(skb))) {
462 txq = skb_get_rx_queue(skb);
463 while (unlikely(txq >= numqueues))
464 txq -= numqueues;
465 }
466
467 rcu_read_unlock();
468 return txq;
469}
470
Jason Wangcde8b152012-10-31 19:46:01 +0000471static inline bool tun_not_capable(struct tun_struct *tun)
472{
473 const struct cred *cred = current_cred();
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000474 struct net *net = dev_net(tun->dev);
Jason Wangcde8b152012-10-31 19:46:01 +0000475
476 return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
477 (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000478 !ns_capable(net->user_ns, CAP_NET_ADMIN);
Jason Wangcde8b152012-10-31 19:46:01 +0000479}
480
Jason Wangc8d68e62012-10-31 19:46:00 +0000481static void tun_set_real_num_queues(struct tun_struct *tun)
482{
483 netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
484 netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
485}
486
Jason Wang4008e972012-12-13 23:53:30 +0000487static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
488{
489 tfile->detached = tun;
490 list_add_tail(&tfile->next, &tun->disabled);
491 ++tun->numdisabled;
492}
493
Jason Wangd32649d2012-12-18 11:00:27 +0800494static struct tun_struct *tun_enable_queue(struct tun_file *tfile)
Jason Wang4008e972012-12-13 23:53:30 +0000495{
496 struct tun_struct *tun = tfile->detached;
497
498 tfile->detached = NULL;
499 list_del_init(&tfile->next);
500 --tun->numdisabled;
501 return tun;
502}
503
Jason Wang4bfb0512013-09-05 17:53:59 +0800504static void tun_queue_purge(struct tun_file *tfile)
505{
506 skb_queue_purge(&tfile->sk.sk_receive_queue);
507 skb_queue_purge(&tfile->sk.sk_error_queue);
508}
509
Jason Wangc8d68e62012-10-31 19:46:00 +0000510static void __tun_detach(struct tun_file *tfile, bool clean)
511{
512 struct tun_file *ntfile;
513 struct tun_struct *tun;
Jason Wangc8d68e62012-10-31 19:46:00 +0000514
Jason Wangb8deabd2013-01-11 16:59:32 +0000515 tun = rtnl_dereference(tfile->tun);
516
Jason Wang9e857222013-01-28 01:05:19 +0000517 if (tun && !tfile->detached) {
Jason Wangc8d68e62012-10-31 19:46:00 +0000518 u16 index = tfile->queue_index;
519 BUG_ON(index >= tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000520
521 rcu_assign_pointer(tun->tfiles[index],
522 tun->tfiles[tun->numqueues - 1]);
Jason Wangb8deabd2013-01-11 16:59:32 +0000523 ntfile = rtnl_dereference(tun->tfiles[index]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000524 ntfile->queue_index = index;
525
526 --tun->numqueues;
Jason Wang9e857222013-01-28 01:05:19 +0000527 if (clean) {
Monam Agarwalc9566742014-03-24 00:02:32 +0530528 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang4008e972012-12-13 23:53:30 +0000529 sock_put(&tfile->sk);
Jason Wang9e857222013-01-28 01:05:19 +0000530 } else
Jason Wang4008e972012-12-13 23:53:30 +0000531 tun_disable_queue(tun, tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000532
533 synchronize_net();
Jason Wang96442e422012-10-31 19:46:02 +0000534 tun_flow_delete_by_queue(tun, tun->numqueues + 1);
Jason Wangc8d68e62012-10-31 19:46:00 +0000535 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800536 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000537 tun_set_real_num_queues(tun);
Jason Wangdd38bd82013-01-11 16:59:34 +0000538 } else if (tfile->detached && clean) {
Jason Wang4008e972012-12-13 23:53:30 +0000539 tun = tun_enable_queue(tfile);
Jason Wangdd38bd82013-01-11 16:59:34 +0000540 sock_put(&tfile->sk);
541 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000542
543 if (clean) {
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000544 if (tun && tun->numqueues == 0 && tun->numdisabled == 0) {
545 netif_carrier_off(tun->dev);
546
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200547 if (!(tun->flags & IFF_PERSIST) &&
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000548 tun->dev->reg_state == NETREG_REGISTERED)
Jason Wang4008e972012-12-13 23:53:30 +0000549 unregister_netdevice(tun->dev);
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000550 }
Eric W. Biederman140e807da2015-05-08 21:07:08 -0500551 sock_put(&tfile->sk);
Jason Wangc8d68e62012-10-31 19:46:00 +0000552 }
553}
554
555static void tun_detach(struct tun_file *tfile, bool clean)
556{
557 rtnl_lock();
558 __tun_detach(tfile, clean);
559 rtnl_unlock();
560}
561
562static void tun_detach_all(struct net_device *dev)
563{
564 struct tun_struct *tun = netdev_priv(dev);
Jason Wang4008e972012-12-13 23:53:30 +0000565 struct tun_file *tfile, *tmp;
Jason Wangc8d68e62012-10-31 19:46:00 +0000566 int i, n = tun->numqueues;
567
568 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000569 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000570 BUG_ON(!tfile);
Xi Wang9e641bd2014-05-16 15:11:48 -0700571 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530572 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wangc8d68e62012-10-31 19:46:00 +0000573 --tun->numqueues;
574 }
Jason Wang9e857222013-01-28 01:05:19 +0000575 list_for_each_entry(tfile, &tun->disabled, next) {
Xi Wang9e641bd2014-05-16 15:11:48 -0700576 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530577 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang9e857222013-01-28 01:05:19 +0000578 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000579 BUG_ON(tun->numqueues != 0);
580
581 synchronize_net();
582 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000583 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000584 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800585 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000586 sock_put(&tfile->sk);
587 }
Jason Wang4008e972012-12-13 23:53:30 +0000588 list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
589 tun_enable_queue(tfile);
Jason Wang4bfb0512013-09-05 17:53:59 +0800590 tun_queue_purge(tfile);
Jason Wang4008e972012-12-13 23:53:30 +0000591 sock_put(&tfile->sk);
592 }
593 BUG_ON(tun->numdisabled != 0);
Jason Wangdd38bd82013-01-11 16:59:34 +0000594
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200595 if (tun->flags & IFF_PERSIST)
Jason Wangdd38bd82013-01-11 16:59:34 +0000596 module_put(THIS_MODULE);
Jason Wangc8d68e62012-10-31 19:46:00 +0000597}
598
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400599static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter)
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000600{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000601 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000602 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000603
Paul Moore5dbbaf22013-01-14 07:12:19 +0000604 err = security_tun_dev_attach(tfile->socket.sk, tun->security);
605 if (err < 0)
606 goto out;
607
Eric W. Biederman38231b72009-01-20 11:02:28 +0000608 err = -EINVAL;
Jason Wang9e857222013-01-28 01:05:19 +0000609 if (rtnl_dereference(tfile->tun) && !tfile->detached)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000610 goto out;
611
612 err = -EBUSY;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200613 if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1)
Jason Wangc8d68e62012-10-31 19:46:00 +0000614 goto out;
615
616 err = -E2BIG;
Jason Wang4008e972012-12-13 23:53:30 +0000617 if (!tfile->detached &&
618 tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000619 goto out;
620
621 err = 0;
Jason Wang54f968d2012-10-31 19:45:57 +0000622
stephen hemminger92d4ea62013-12-05 20:42:58 -0800623 /* Re-attach the filter to persist device */
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400624 if (!skip_filter && (tun->filter_attached == true)) {
Hannes Frederic Sowa8ced4252016-04-05 17:10:16 +0200625 lock_sock(tfile->socket.sk);
626 err = sk_attach_filter(&tun->fprog, tfile->socket.sk);
627 release_sock(tfile->socket.sk);
Jason Wang54f968d2012-10-31 19:45:57 +0000628 if (!err)
629 goto out;
630 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000631 tfile->queue_index = tun->numqueues;
Jason Wang6e914fc2012-10-31 19:45:58 +0000632 rcu_assign_pointer(tfile->tun, tun);
Jason Wangc8d68e62012-10-31 19:46:00 +0000633 rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000634 tun->numqueues++;
635
Jason Wang4008e972012-12-13 23:53:30 +0000636 if (tfile->detached)
637 tun_enable_queue(tfile);
638 else
639 sock_hold(&tfile->sk);
640
Jason Wangc8d68e62012-10-31 19:46:00 +0000641 tun_set_real_num_queues(tun);
642
Jason Wangc8d68e62012-10-31 19:46:00 +0000643 /* device is allowed to go away first, so no need to hold extra
644 * refcnt.
645 */
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000646
Eric W. Biederman38231b72009-01-20 11:02:28 +0000647out:
Eric W. Biederman38231b72009-01-20 11:02:28 +0000648 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000649}
650
Eric W. Biederman631ab462009-01-20 11:00:40 +0000651static struct tun_struct *__tun_get(struct tun_file *tfile)
652{
Jason Wang6e914fc2012-10-31 19:45:58 +0000653 struct tun_struct *tun;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000654
Jason Wang6e914fc2012-10-31 19:45:58 +0000655 rcu_read_lock();
656 tun = rcu_dereference(tfile->tun);
657 if (tun)
658 dev_hold(tun->dev);
659 rcu_read_unlock();
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000660
661 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000662}
663
664static struct tun_struct *tun_get(struct file *file)
665{
666 return __tun_get(file->private_data);
667}
668
669static void tun_put(struct tun_struct *tun)
670{
Jason Wang6e914fc2012-10-31 19:45:58 +0000671 dev_put(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000672}
673
Joe Perches6b8a66e2011-03-02 07:18:10 +0000674/* TAP filtering */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700675static void addr_hash_set(u32 *mask, const u8 *addr)
676{
677 int n = ether_crc(ETH_ALEN, addr) >> 26;
678 mask[n >> 5] |= (1 << (n & 31));
679}
680
681static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
682{
683 int n = ether_crc(ETH_ALEN, addr) >> 26;
684 return mask[n >> 5] & (1 << (n & 31));
685}
686
687static int update_filter(struct tap_filter *filter, void __user *arg)
688{
689 struct { u8 u[ETH_ALEN]; } *addr;
690 struct tun_filter uf;
691 int err, alen, n, nexact;
692
693 if (copy_from_user(&uf, arg, sizeof(uf)))
694 return -EFAULT;
695
696 if (!uf.count) {
697 /* Disabled */
698 filter->count = 0;
699 return 0;
700 }
701
702 alen = ETH_ALEN * uf.count;
703 addr = kmalloc(alen, GFP_KERNEL);
704 if (!addr)
705 return -ENOMEM;
706
707 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
708 err = -EFAULT;
709 goto done;
710 }
711
712 /* The filter is updated without holding any locks. Which is
713 * perfectly safe. We disable it first and in the worst
714 * case we'll accept a few undesired packets. */
715 filter->count = 0;
716 wmb();
717
718 /* Use first set of addresses as an exact filter */
719 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
720 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
721
722 nexact = n;
723
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800724 /* Remaining multicast addresses are hashed,
725 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700726 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800727 for (; n < uf.count; n++) {
728 if (!is_multicast_ether_addr(addr[n].u)) {
729 err = 0; /* no filter */
730 goto done;
731 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700732 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800733 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700734
735 /* For ALLMULTI just set the mask to all ones.
736 * This overrides the mask populated above. */
737 if ((uf.flags & TUN_FLT_ALLMULTI))
738 memset(filter->mask, ~0, sizeof(filter->mask));
739
740 /* Now enable the filter */
741 wmb();
742 filter->count = nexact;
743
744 /* Return the number of exact filters */
745 err = nexact;
746
747done:
748 kfree(addr);
749 return err;
750}
751
752/* Returns: 0 - drop, !=0 - accept */
753static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
754{
755 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
756 * at this point. */
757 struct ethhdr *eh = (struct ethhdr *) skb->data;
758 int i;
759
760 /* Exact match */
761 for (i = 0; i < filter->count; i++)
Joe Perches2e42e472012-05-09 17:17:46 +0000762 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700763 return 1;
764
765 /* Inexact match (multicast only) */
766 if (is_multicast_ether_addr(eh->h_dest))
767 return addr_hash_test(filter->mask, eh->h_dest);
768
769 return 0;
770}
771
772/*
773 * Checks whether the packet is accepted or not.
774 * Returns: 0 - drop, !=0 - accept
775 */
776static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
777{
778 if (!filter->count)
779 return 1;
780
781 return run_filter(filter, skb);
782}
783
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784/* Network device part of the driver */
785
Jeff Garzik7282d492006-09-13 14:30:00 -0400786static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000788/* Net device detach from fd. */
789static void tun_net_uninit(struct net_device *dev)
790{
Jason Wangc8d68e62012-10-31 19:46:00 +0000791 tun_detach_all(dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000792}
793
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794/* Net device open. */
795static int tun_net_open(struct net_device *dev)
796{
Jason Wangc8d68e62012-10-31 19:46:00 +0000797 netif_tx_start_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 return 0;
799}
800
801/* Net device close. */
802static int tun_net_close(struct net_device *dev)
803{
Jason Wangc8d68e62012-10-31 19:46:00 +0000804 netif_tx_stop_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 return 0;
806}
807
808/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000809static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
811 struct tun_struct *tun = netdev_priv(dev);
Jason Wangc8d68e62012-10-31 19:46:00 +0000812 int txq = skb->queue_mapping;
Jason Wang6e914fc2012-10-31 19:45:58 +0000813 struct tun_file *tfile;
Dominic Curranfa358642014-01-22 03:03:23 +0000814 u32 numqueues = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815
Jason Wang6e914fc2012-10-31 19:45:58 +0000816 rcu_read_lock();
Jason Wangc8d68e62012-10-31 19:46:00 +0000817 tfile = rcu_dereference(tun->tfiles[txq]);
Dominic Curranfa358642014-01-22 03:03:23 +0000818 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000819
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 /* Drop packet if interface is not attached */
Dominic Curranfa358642014-01-22 03:03:23 +0000821 if (txq >= numqueues)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 goto drop;
823
Dominic Curranfa358642014-01-22 03:03:23 +0000824 if (numqueues == 1) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800825 /* Select queue was not called for the skbuff, so we extract the
826 * RPS hash and save it into the flow_table here.
827 */
828 __u32 rxhash;
829
830 rxhash = skb_get_hash(skb);
831 if (rxhash) {
832 struct tun_flow_entry *e;
833 e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)],
834 rxhash);
835 if (e)
836 tun_flow_save_rps_rxhash(e, rxhash);
837 }
838 }
839
Jason Wang6e914fc2012-10-31 19:45:58 +0000840 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
841
Jason Wangc8d68e62012-10-31 19:46:00 +0000842 BUG_ON(!tfile);
843
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700844 /* Drop if the filter does not like it.
845 * This is a noop if the filter is disabled.
846 * Filter can be enabled only for the TAP devices. */
847 if (!check_filter(&tun->txflt, skb))
848 goto drop;
849
Jason Wang54f968d2012-10-31 19:45:57 +0000850 if (tfile->socket.sk->sk_filter &&
851 sk_filter(tfile->socket.sk, skb))
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000852 goto drop;
853
Rami Rosen36fe8c092012-11-25 22:07:40 +0000854 /* Limit the number of packets queued by dividing txq length with the
Jason Wangc8d68e62012-10-31 19:46:00 +0000855 * number of queues.
856 */
Dominic Curranfa358642014-01-22 03:03:23 +0000857 if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues
858 >= dev->tx_queue_len)
Michael S. Tsirkin5d097102012-12-03 10:07:14 +0000859 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
Jason Wang7bf66302013-09-05 17:54:00 +0800861 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
862 goto drop;
863
Eric Dumazet5fcd2d82015-10-09 15:42:21 -0700864 if (skb->sk && sk_fullsock(skb->sk)) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400865 sock_tx_timestamp(skb->sk, skb->sk->sk_tsflags,
866 &skb_shinfo(skb)->tx_flags);
Richard Cochraneda29772013-07-19 19:40:10 +0200867 sw_tx_timestamp(skb);
868 }
869
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000870 /* Orphan the skb - required as we might hang on to it
Jason Wang7bf66302013-09-05 17:54:00 +0800871 * for indefinite time.
872 */
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000873 skb_orphan(skb);
874
Eric Dumazetf8af75f2013-03-06 11:02:37 +0000875 nf_reset(skb);
876
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700877 /* Enqueue packet */
Jason Wang54f968d2012-10-31 19:45:57 +0000878 skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
880 /* Notify and wake up reader process */
Jason Wang54f968d2012-10-31 19:45:57 +0000881 if (tfile->flags & TUN_FASYNC)
882 kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
Xi Wang9e641bd2014-05-16 15:11:48 -0700883 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Jason Wang6e914fc2012-10-31 19:45:58 +0000884
885 rcu_read_unlock();
Patrick McHardy6ed10652009-06-23 06:03:08 +0000886 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
888drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700889 dev->stats.tx_dropped++;
Michael S. Tsirkin149d36f2012-11-01 09:16:32 +0000890 skb_tx_error(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 kfree_skb(skb);
Jason Wang6e914fc2012-10-31 19:45:58 +0000892 rcu_read_unlock();
Jason Wangbaeabab2014-11-18 13:20:41 +0800893 return NET_XMIT_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894}
895
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700896static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700898 /*
899 * This callback is supposed to deal with mc filter in
900 * _rx_ path and has nothing to do with the _tx_ path.
901 * In rx path we always accept everything userspace gives us.
902 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903}
904
Ed Swierk4885a502007-09-16 12:21:38 -0700905#define MIN_MTU 68
906#define MAX_MTU 65535
907
908static int
909tun_net_change_mtu(struct net_device *dev, int new_mtu)
910{
911 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
912 return -EINVAL;
913 dev->mtu = new_mtu;
914 return 0;
915}
916
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000917static netdev_features_t tun_net_fix_features(struct net_device *dev,
918 netdev_features_t features)
Michał Mirosław88255372011-04-19 06:13:10 +0000919{
920 struct tun_struct *tun = netdev_priv(dev);
921
922 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
923}
Neil Hormanbebd0972011-06-15 05:25:01 +0000924#ifdef CONFIG_NET_POLL_CONTROLLER
925static void tun_poll_controller(struct net_device *dev)
926{
927 /*
928 * Tun only receives frames when:
929 * 1) the char device endpoint gets data from user space
930 * 2) the tun socket gets a sendmsg call from user space
stephen hemminger92d4ea62013-12-05 20:42:58 -0800931 * Since both of those are synchronous operations, we are guaranteed
Neil Hormanbebd0972011-06-15 05:25:01 +0000932 * never to have pending data when we poll for it
stephen hemminger92d4ea62013-12-05 20:42:58 -0800933 * so there is nothing to do here but return.
Neil Hormanbebd0972011-06-15 05:25:01 +0000934 * We need this though so netpoll recognizes us as an interface that
935 * supports polling, which enables bridge devices in virt setups to
936 * still use netconsole
937 */
938 return;
939}
940#endif
Paolo Abenieaea34b2016-02-26 10:45:40 +0100941
942static void tun_set_headroom(struct net_device *dev, int new_hr)
943{
944 struct tun_struct *tun = netdev_priv(dev);
945
946 if (new_hr < NET_SKB_PAD)
947 new_hr = NET_SKB_PAD;
948
949 tun->align = new_hr;
950}
951
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800952static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000953 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800954 .ndo_open = tun_net_open,
955 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800956 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800957 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000958 .ndo_fix_features = tun_net_fix_features,
Jason Wangc8d68e62012-10-31 19:46:00 +0000959 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000960#ifdef CONFIG_NET_POLL_CONTROLLER
961 .ndo_poll_controller = tun_poll_controller,
962#endif
Paolo Abenieaea34b2016-02-26 10:45:40 +0100963 .ndo_set_rx_headroom = tun_set_headroom,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800964};
965
966static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000967 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800968 .ndo_open = tun_net_open,
969 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800970 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800971 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000972 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +0000973 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800974 .ndo_set_mac_address = eth_mac_addr,
975 .ndo_validate_addr = eth_validate_addr,
Jason Wangc8d68e62012-10-31 19:46:00 +0000976 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000977#ifdef CONFIG_NET_POLL_CONTROLLER
978 .ndo_poll_controller = tun_poll_controller,
979#endif
Toshiaki Makita5e527962015-07-31 15:03:27 +0900980 .ndo_features_check = passthru_features_check,
Paolo Abenieaea34b2016-02-26 10:45:40 +0100981 .ndo_set_rx_headroom = tun_set_headroom,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800982};
983
Pavel Emelyanov944a1372013-06-11 17:01:08 +0400984static void tun_flow_init(struct tun_struct *tun)
Jason Wang96442e422012-10-31 19:46:02 +0000985{
986 int i;
987
Jason Wang96442e422012-10-31 19:46:02 +0000988 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
989 INIT_HLIST_HEAD(&tun->flows[i]);
990
991 tun->ageing_time = TUN_FLOW_EXPIRE;
992 setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun);
993 mod_timer(&tun->flow_gc_timer,
994 round_jiffies_up(jiffies + tun->ageing_time));
Jason Wang96442e422012-10-31 19:46:02 +0000995}
996
997static void tun_flow_uninit(struct tun_struct *tun)
998{
999 del_timer_sync(&tun->flow_gc_timer);
1000 tun_flow_flush(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001001}
1002
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003/* Initialize net device. */
1004static void tun_net_init(struct net_device *dev)
1005{
1006 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001007
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001009 case IFF_TUN:
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001010 dev->netdev_ops = &tun_netdev_ops;
1011
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 /* Point-to-Point TUN Device */
1013 dev->hard_header_len = 0;
1014 dev->addr_len = 0;
1015 dev->mtu = 1500;
1016
1017 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001018 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 break;
1021
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001022 case IFF_TAP:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -08001023 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001025 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +00001026 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
stephen hemmingera6768472012-12-10 15:16:00 +00001027 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
Danny Kukawkaf2cedb62012-02-15 06:45:39 +00001029 eth_hw_addr_random(dev);
Brian Braunstein36226a82007-04-26 01:00:55 -07001030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 break;
1032 }
1033}
1034
1035/* Character device part */
1036
1037/* Poll */
Jason Wangc8d68e62012-10-31 19:46:00 +00001038static unsigned int tun_chr_poll(struct file *file, poll_table *wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001039{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +00001040 struct tun_file *tfile = file->private_data;
1041 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +00001042 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001043 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
1045 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +00001046 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
Jason Wang54f968d2012-10-31 19:45:57 +00001048 sk = tfile->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +00001049
Joe Perches6b8a66e2011-03-02 07:18:10 +00001050 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
Xi Wang9e641bd2014-05-16 15:11:48 -07001052 poll_wait(file, sk_sleep(sk), wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001053
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +00001054 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 mask |= POLLIN | POLLRDNORM;
1056
Herbert Xu33dccbb2009-02-05 21:25:32 -08001057 if (sock_writeable(sk) ||
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001058 (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
Herbert Xu33dccbb2009-02-05 21:25:32 -08001059 sock_writeable(sk)))
1060 mask |= POLLOUT | POLLWRNORM;
1061
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001062 if (tun->dev->reg_state != NETREG_REGISTERED)
1063 mask = POLLERR;
1064
Eric W. Biederman631ab462009-01-20 11:00:40 +00001065 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 return mask;
1067}
1068
Rusty Russellf42157c2008-08-15 15:15:10 -07001069/* prepad is the amount to reserve at front. len is length after that.
1070 * linear is a hint as to how much to copy (usually headers). */
Jason Wang54f968d2012-10-31 19:45:57 +00001071static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001072 size_t prepad, size_t len,
1073 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -07001074{
Jason Wang54f968d2012-10-31 19:45:57 +00001075 struct sock *sk = tfile->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -07001076 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001077 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -07001078
1079 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -07001080 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001081 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -07001082
Herbert Xu33dccbb2009-02-05 21:25:32 -08001083 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
Eric Dumazet28d64272013-08-08 14:38:47 -07001084 &err, 0);
Rusty Russellf42157c2008-08-15 15:15:10 -07001085 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001086 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -07001087
1088 skb_reserve(skb, prepad);
1089 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001090 skb->data_len = len - linear;
1091 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -07001092
1093 return skb;
1094}
1095
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096/* Get packet from user space buffer */
Jason Wang54f968d2012-10-31 19:45:57 +00001097static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
Al Virof5ff53b2014-06-19 15:36:49 -04001098 void *msg_control, struct iov_iter *from,
1099 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100{
Harvey Harrison09640e62009-02-01 00:45:17 -08001101 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 struct sk_buff *skb;
Al Virof5ff53b2014-06-19 15:36:49 -04001103 size_t total_len = iov_iter_count(from);
Paolo Abenieaea34b2016-02-26 10:45:40 +01001104 size_t len = total_len, align = tun->align, linear;
Rusty Russellf43798c2008-07-03 03:48:02 -07001105 struct virtio_net_hdr gso = { 0 };
Jason Wang96f8d9e2013-11-13 14:00:39 +08001106 int good_linear;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001107 int copylen;
1108 bool zerocopy = false;
1109 int err;
Eric Dumazet49974422012-12-12 19:22:57 +00001110 u32 rxhash;
Al Virof5ff53b2014-06-19 15:36:49 -04001111 ssize_t n;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Eric Dumazet1bd4978a2015-12-16 08:57:37 -08001113 if (!(tun->dev->flags & IFF_UP))
1114 return -EIO;
1115
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001116 if (!(tun->flags & IFF_NO_PI)) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001117 if (len < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001119 len -= sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120
Al Virof5ff53b2014-06-19 15:36:49 -04001121 n = copy_from_iter(&pi, sizeof(pi), from);
1122 if (n != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 return -EFAULT;
1124 }
1125
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001126 if (tun->flags & IFF_VNET_HDR) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001127 if (len < tun->vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001128 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001129 len -= tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001130
Al Virof5ff53b2014-06-19 15:36:49 -04001131 n = copy_from_iter(&gso, sizeof(gso), from);
1132 if (n != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001133 return -EFAULT;
1134
Herbert Xu49091222009-06-08 00:20:01 -07001135 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001136 tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len))
1137 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 -07001138
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001139 if (tun16_to_cpu(tun, gso.hdr_len) > len)
Rusty Russellf43798c2008-07-03 03:48:02 -07001140 return -EINVAL;
Herbert Xud8febb72014-11-30 18:03:31 +08001141 iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001142 }
1143
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001144 if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) {
stephen hemmingera504b862011-06-08 14:33:07 +00001145 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -07001146 if (unlikely(len < ETH_HLEN ||
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001147 (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -07001148 return -EINVAL;
1149 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001150
Jason Wang96f8d9e2013-11-13 14:00:39 +08001151 good_linear = SKB_MAX_HEAD(align);
1152
Jason Wang88529172013-07-18 10:55:15 +08001153 if (msg_control) {
Al Virof5ff53b2014-06-19 15:36:49 -04001154 struct iov_iter i = *from;
1155
Jason Wang88529172013-07-18 10:55:15 +08001156 /* There are 256 bytes to be copied in skb, so there is
1157 * enough room for skb expand head in case it is used.
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001158 * The rest of the buffer is mapped from userspace.
1159 */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001160 copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001161 if (copylen > good_linear)
1162 copylen = good_linear;
Jason Wang3dd5c332013-07-10 13:43:27 +08001163 linear = copylen;
Al Virof5ff53b2014-06-19 15:36:49 -04001164 iov_iter_advance(&i, copylen);
1165 if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS)
Jason Wang88529172013-07-18 10:55:15 +08001166 zerocopy = true;
1167 }
1168
1169 if (!zerocopy) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001170 copylen = len;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001171 if (tun16_to_cpu(tun, gso.hdr_len) > good_linear)
Jason Wang96f8d9e2013-11-13 14:00:39 +08001172 linear = good_linear;
1173 else
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001174 linear = tun16_to_cpu(tun, gso.hdr_len);
Jason Wang3dd5c332013-07-10 13:43:27 +08001175 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001176
Jason Wang3dd5c332013-07-10 13:43:27 +08001177 skb = tun_alloc_skb(tfile, align, copylen, linear, noblock);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001178 if (IS_ERR(skb)) {
1179 if (PTR_ERR(skb) != -EAGAIN)
1180 tun->dev->stats.rx_dropped++;
1181 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 }
1183
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001184 if (zerocopy)
Al Virof5ff53b2014-06-19 15:36:49 -04001185 err = zerocopy_sg_from_iter(skb, from);
Jason Wang88529172013-07-18 10:55:15 +08001186 else {
Al Virof5ff53b2014-06-19 15:36:49 -04001187 err = skb_copy_datagram_from_iter(skb, 0, from, len);
Jason Wang88529172013-07-18 10:55:15 +08001188 if (!err && msg_control) {
1189 struct ubuf_info *uarg = msg_control;
1190 uarg->callback(uarg, false);
1191 }
1192 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001193
1194 if (err) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001195 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -08001196 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -08001198 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Rusty Russellf43798c2008-07-03 03:48:02 -07001200 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001201 if (!skb_partial_csum_set(skb, tun16_to_cpu(tun, gso.csum_start),
1202 tun16_to_cpu(tun, gso.csum_offset))) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001203 tun->dev->stats.rx_frame_errors++;
1204 kfree_skb(skb);
1205 return -EINVAL;
1206 }
Michał Mirosław88255372011-04-19 06:13:10 +00001207 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001208
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001210 case IFF_TUN:
1211 if (tun->flags & IFF_NO_PI) {
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -07001212 switch (skb->data[0] & 0xf0) {
1213 case 0x40:
1214 pi.proto = htons(ETH_P_IP);
1215 break;
1216 case 0x60:
1217 pi.proto = htons(ETH_P_IPV6);
1218 break;
1219 default:
1220 tun->dev->stats.rx_dropped++;
1221 kfree_skb(skb);
1222 return -EINVAL;
1223 }
1224 }
1225
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001226 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -07001228 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001230 case IFF_TAP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 skb->protocol = eth_type_trans(skb, tun->dev);
1232 break;
Joe Perches6403eab2011-06-03 11:51:20 +00001233 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
Rusty Russellf43798c2008-07-03 03:48:02 -07001235 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
1236 pr_debug("GSO!\n");
1237 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
1238 case VIRTIO_NET_HDR_GSO_TCPV4:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001239 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
Rusty Russellf43798c2008-07-03 03:48:02 -07001240 break;
1241 case VIRTIO_NET_HDR_GSO_TCPV6:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001242 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
Rusty Russellf43798c2008-07-03 03:48:02 -07001243 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001244 case VIRTIO_NET_HDR_GSO_UDP:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001245 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001246 break;
Rusty Russellf43798c2008-07-03 03:48:02 -07001247 default:
1248 tun->dev->stats.rx_frame_errors++;
1249 kfree_skb(skb);
1250 return -EINVAL;
1251 }
1252
1253 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001254 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
Rusty Russellf43798c2008-07-03 03:48:02 -07001255
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001256 skb_shinfo(skb)->gso_size = tun16_to_cpu(tun, gso.gso_size);
Rusty Russellf43798c2008-07-03 03:48:02 -07001257 if (skb_shinfo(skb)->gso_size == 0) {
1258 tun->dev->stats.rx_frame_errors++;
1259 kfree_skb(skb);
1260 return -EINVAL;
1261 }
1262
1263 /* Header must be checked, and gso_segs computed. */
1264 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
1265 skb_shinfo(skb)->gso_segs = 0;
1266 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001267
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001268 /* copy skb_ubuf_info for callback when skb has no error */
1269 if (zerocopy) {
1270 skb_shinfo(skb)->destructor_arg = msg_control;
1271 skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001272 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001273 }
1274
Vlad Yasevich72f65102015-02-03 16:36:16 -05001275 skb_reset_network_header(skb);
Jason Wang40893fd2013-03-26 23:11:22 +00001276 skb_probe_transport_header(skb, 0);
Jason Wang38502af2013-03-25 20:19:56 +00001277
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001278 rxhash = skb_get_hash(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001280
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001281 tun->dev->stats.rx_packets++;
1282 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283
Jason Wang9e857222013-01-28 01:05:19 +00001284 tun_flow_update(tun, rxhash, tfile);
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001285 return total_len;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001286}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287
Al Virof5ff53b2014-06-19 15:36:49 -04001288static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001290 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -08001291 struct tun_struct *tun = tun_get(file);
Jason Wang54f968d2012-10-31 19:45:57 +00001292 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001293 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294
1295 if (!tun)
1296 return -EBADFD;
1297
Al Virof5ff53b2014-06-19 15:36:49 -04001298 result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001299
1300 tun_put(tun);
1301 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302}
1303
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +00001305static ssize_t tun_put_user(struct tun_struct *tun,
Jason Wang54f968d2012-10-31 19:45:57 +00001306 struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001307 struct sk_buff *skb,
Herbert Xue0b46d02014-11-07 21:22:23 +08001308 struct iov_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309{
1310 struct tun_pi pi = { 0, skb->protocol };
Herbert Xue0b46d02014-11-07 21:22:23 +08001311 ssize_t total;
Jason Wang8c847d22014-11-13 16:54:14 +08001312 int vlan_offset = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001313 int vlan_hlen = 0;
Herbert Xu2eb783c2014-11-03 04:30:14 +08001314 int vnet_hdr_sz = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001315
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001316 if (skb_vlan_tag_present(skb))
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001317 vlan_hlen = VLAN_HLEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001319 if (tun->flags & IFF_VNET_HDR)
Herbert Xu2eb783c2014-11-03 04:30:14 +08001320 vnet_hdr_sz = tun->vnet_hdr_sz;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
Herbert Xue0b46d02014-11-07 21:22:23 +08001322 total = skb->len + vlan_hlen + vnet_hdr_sz;
1323
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001324 if (!(tun->flags & IFF_NO_PI)) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001325 if (iov_iter_count(iter) < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 return -EINVAL;
1327
Herbert Xue0b46d02014-11-07 21:22:23 +08001328 total += sizeof(pi);
1329 if (iov_iter_count(iter) < total) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 /* Packet will be striped */
1331 pi.flags |= TUN_PKT_STRIP;
1332 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001333
Herbert Xue0b46d02014-11-07 21:22:23 +08001334 if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 return -EFAULT;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001336 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
Herbert Xu2eb783c2014-11-03 04:30:14 +08001338 if (vnet_hdr_sz) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001339 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Herbert Xue0b46d02014-11-07 21:22:23 +08001340 if (iov_iter_count(iter) < vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001341 return -EINVAL;
1342
1343 if (skb_is_gso(skb)) {
1344 struct skb_shared_info *sinfo = skb_shinfo(skb);
1345
1346 /* This is a hint as to how much should be linear. */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001347 gso.hdr_len = cpu_to_tun16(tun, skb_headlen(skb));
1348 gso.gso_size = cpu_to_tun16(tun, sinfo->gso_size);
Rusty Russellf43798c2008-07-03 03:48:02 -07001349 if (sinfo->gso_type & SKB_GSO_TCPV4)
1350 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
1351 else if (sinfo->gso_type & SKB_GSO_TCPV6)
1352 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Vlad Yaseviche3e3c422015-02-03 16:36:17 -05001353 else if (sinfo->gso_type & SKB_GSO_UDP)
1354 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001355 else {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001356 pr_err("unexpected GSO type: "
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001357 "0x%x, gso_size %d, hdr_len %d\n",
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001358 sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size),
1359 tun16_to_cpu(tun, gso.hdr_len));
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001360 print_hex_dump(KERN_ERR, "tun: ",
1361 DUMP_PREFIX_NONE,
1362 16, 1, skb->head,
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001363 min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true);
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001364 WARN_ON_ONCE(1);
1365 return -EINVAL;
1366 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001367 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
1368 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
1369 } else
1370 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
1371
1372 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1373 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001374 gso.csum_start = cpu_to_tun16(tun, skb_checksum_start_offset(skb) +
1375 vlan_hlen);
1376 gso.csum_offset = cpu_to_tun16(tun, skb->csum_offset);
Jason Wang10a8d942011-06-10 00:56:17 +00001377 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
1378 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
Rusty Russellf43798c2008-07-03 03:48:02 -07001379 } /* else everything is zero */
1380
Herbert Xue0b46d02014-11-07 21:22:23 +08001381 if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001382 return -EFAULT;
Jason Wang8c847d22014-11-13 16:54:14 +08001383
1384 iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001385 }
1386
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001387 if (vlan_hlen) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001388 int ret;
Jason Wang6680ec62013-07-25 13:00:33 +08001389 struct {
1390 __be16 h_vlan_proto;
1391 __be16 h_vlan_TCI;
1392 } veth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
Jason Wang6680ec62013-07-25 13:00:33 +08001394 veth.h_vlan_proto = skb->vlan_proto;
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001395 veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396
Jason Wang6680ec62013-07-25 13:00:33 +08001397 vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto);
Jason Wang6680ec62013-07-25 13:00:33 +08001398
Herbert Xue0b46d02014-11-07 21:22:23 +08001399 ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset);
1400 if (ret || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001401 goto done;
1402
Herbert Xue0b46d02014-11-07 21:22:23 +08001403 ret = copy_to_iter(&veth, sizeof(veth), iter);
1404 if (ret != sizeof(veth) || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001405 goto done;
1406 }
1407
Herbert Xue0b46d02014-11-07 21:22:23 +08001408 skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset);
Jason Wang6680ec62013-07-25 13:00:33 +08001409
1410done:
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001411 tun->dev->stats.tx_packets++;
Herbert Xue0b46d02014-11-07 21:22:23 +08001412 tun->dev->stats.tx_bytes += skb->len + vlan_hlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413
1414 return total;
1415}
1416
Jason Wang54f968d2012-10-31 19:45:57 +00001417static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
Al Viro9b067032014-11-07 13:52:07 -05001418 struct iov_iter *to,
1419 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 struct sk_buff *skb;
Al Viro9b067032014-11-07 13:52:07 -05001422 ssize_t ret;
Xi Wang9e641bd2014-05-16 15:11:48 -07001423 int peeked, err, off = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
Rami Rosen3872baf2012-11-25 22:07:41 +00001425 tun_debug(KERN_INFO, tun, "tun_do_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Al Viro9b067032014-11-07 13:52:07 -05001427 if (!iov_iter_count(to))
1428 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Xi Wang9e641bd2014-05-16 15:11:48 -07001430 if (tun->dev->reg_state != NETREG_REGISTERED)
1431 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432
Xi Wang9e641bd2014-05-16 15:11:48 -07001433 /* Read frames from queue */
1434 skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0,
1435 &peeked, &off, &err);
Herbert Xue0b46d02014-11-07 21:22:23 +08001436 if (!skb)
Alex Gartrell957f0942014-12-25 23:22:49 -08001437 return err;
Herbert Xue0b46d02014-11-07 21:22:23 +08001438
Al Viro9b067032014-11-07 13:52:07 -05001439 ret = tun_put_user(tun, tfile, skb, to);
Jason Wangf51a5e82014-12-01 16:53:15 +08001440 if (unlikely(ret < 0))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001441 kfree_skb(skb);
Jason Wangf51a5e82014-12-01 16:53:15 +08001442 else
1443 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001445 return ret;
1446}
1447
Al Viro9b067032014-11-07 13:52:07 -05001448static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001449{
1450 struct file *file = iocb->ki_filp;
1451 struct tun_file *tfile = file->private_data;
1452 struct tun_struct *tun = __tun_get(tfile);
Al Viro9b067032014-11-07 13:52:07 -05001453 ssize_t len = iov_iter_count(to), ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001454
1455 if (!tun)
1456 return -EBADFD;
Al Viro9b067032014-11-07 13:52:07 -05001457 ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK);
David S. Miller42404c02013-12-10 22:05:45 -05001458 ret = min_t(ssize_t, ret, len);
Zhi Yong Wud0b7da8a2013-12-06 14:16:51 +08001459 if (ret > 0)
1460 iocb->ki_pos = ret;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001461 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 return ret;
1463}
1464
Jason Wang96442e422012-10-31 19:46:02 +00001465static void tun_free_netdev(struct net_device *dev)
1466{
1467 struct tun_struct *tun = netdev_priv(dev);
1468
Jason Wang4008e972012-12-13 23:53:30 +00001469 BUG_ON(!(list_empty(&tun->disabled)));
Jason Wang96442e422012-10-31 19:46:02 +00001470 tun_flow_uninit(tun);
Paul Moore5dbbaf22013-01-14 07:12:19 +00001471 security_tun_dev_free_security(tun->security);
Jason Wang96442e422012-10-31 19:46:02 +00001472 free_netdev(dev);
1473}
1474
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475static void tun_setup(struct net_device *dev)
1476{
1477 struct tun_struct *tun = netdev_priv(dev);
1478
Eric W. Biederman0625c882012-02-07 16:48:55 -08001479 tun->owner = INVALID_UID;
1480 tun->group = INVALID_GID;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 dev->ethtool_ops = &tun_ethtool_ops;
Jason Wang96442e422012-10-31 19:46:02 +00001483 dev->destructor = tun_free_netdev;
Jason Wang016adb72016-04-08 13:26:48 +08001484 /* We prefer our own queue length */
1485 dev->tx_queue_len = TUN_READQ_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486}
1487
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001488/* Trivial set of netlink ops to allow deleting tun or tap
1489 * device with netlink.
1490 */
1491static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
1492{
1493 return -EINVAL;
1494}
1495
1496static struct rtnl_link_ops tun_link_ops __read_mostly = {
1497 .kind = DRV_NAME,
1498 .priv_size = sizeof(struct tun_struct),
1499 .setup = tun_setup,
1500 .validate = tun_validate,
1501};
1502
Herbert Xu33dccbb2009-02-05 21:25:32 -08001503static void tun_sock_write_space(struct sock *sk)
1504{
Jason Wang54f968d2012-10-31 19:45:57 +00001505 struct tun_file *tfile;
Eric Dumazet43815482010-04-29 11:01:49 +00001506 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001507
1508 if (!sock_writeable(sk))
1509 return;
1510
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001511 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags))
Herbert Xu33dccbb2009-02-05 21:25:32 -08001512 return;
1513
Eric Dumazet43815482010-04-29 11:01:49 +00001514 wqueue = sk_sleep(sk);
1515 if (wqueue && waitqueue_active(wqueue))
1516 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001517 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -07001518
Jason Wang54f968d2012-10-31 19:45:57 +00001519 tfile = container_of(sk, struct tun_file, sk);
1520 kill_fasync(&tfile->fasync, SIGIO, POLL_OUT);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001521}
1522
Ying Xue1b784142015-03-02 15:37:48 +08001523static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001524{
Jason Wang54f968d2012-10-31 19:45:57 +00001525 int ret;
1526 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1527 struct tun_struct *tun = __tun_get(tfile);
1528
1529 if (!tun)
1530 return -EBADFD;
Al Virof5ff53b2014-06-19 15:36:49 -04001531
Al Viroc0371da2014-11-24 10:42:55 -05001532 ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04001533 m->msg_flags & MSG_DONTWAIT);
Jason Wang54f968d2012-10-31 19:45:57 +00001534 tun_put(tun);
1535 return ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001536}
1537
Ying Xue1b784142015-03-02 15:37:48 +08001538static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001539 int flags)
1540{
Jason Wang54f968d2012-10-31 19:45:57 +00001541 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1542 struct tun_struct *tun = __tun_get(tfile);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001543 int ret;
Jason Wang54f968d2012-10-31 19:45:57 +00001544
1545 if (!tun)
1546 return -EBADFD;
1547
Richard Cochraneda29772013-07-19 19:40:10 +02001548 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) {
Gao feng3811ae72013-04-24 21:59:23 +00001549 ret = -EINVAL;
1550 goto out;
1551 }
Richard Cochraneda29772013-07-19 19:40:10 +02001552 if (flags & MSG_ERRQUEUE) {
1553 ret = sock_recv_errqueue(sock->sk, m, total_len,
1554 SOL_PACKET, TUN_TX_TIMESTAMP);
1555 goto out;
1556 }
Al Viroc0371da2014-11-24 10:42:55 -05001557 ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT);
Alex Gartrell87897932014-12-25 23:05:03 -08001558 if (ret > (ssize_t)total_len) {
David S. Miller42404c02013-12-10 22:05:45 -05001559 m->msg_flags |= MSG_TRUNC;
1560 ret = flags & MSG_TRUNC ? ret : total_len;
1561 }
Gao feng3811ae72013-04-24 21:59:23 +00001562out:
Jason Wang54f968d2012-10-31 19:45:57 +00001563 tun_put(tun);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001564 return ret;
1565}
1566
1567/* Ops structure to mimic raw sockets with tun */
1568static const struct proto_ops tun_socket_ops = {
1569 .sendmsg = tun_sendmsg,
1570 .recvmsg = tun_recvmsg,
1571};
1572
Herbert Xu33dccbb2009-02-05 21:25:32 -08001573static struct proto tun_proto = {
1574 .name = "tun",
1575 .owner = THIS_MODULE,
Jason Wang54f968d2012-10-31 19:45:57 +00001576 .obj_size = sizeof(struct tun_file),
Herbert Xu33dccbb2009-02-05 21:25:32 -08001577};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001578
David Woodhouse980c9e82009-05-09 22:54:21 -07001579static int tun_flags(struct tun_struct *tun)
1580{
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001581 return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP);
David Woodhouse980c9e82009-05-09 22:54:21 -07001582}
1583
1584static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1585 char *buf)
1586{
1587 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1588 return sprintf(buf, "0x%x\n", tun_flags(tun));
1589}
1590
1591static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1592 char *buf)
1593{
1594 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001595 return uid_valid(tun->owner)?
1596 sprintf(buf, "%u\n",
1597 from_kuid_munged(current_user_ns(), tun->owner)):
1598 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001599}
1600
1601static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1602 char *buf)
1603{
1604 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001605 return gid_valid(tun->group) ?
1606 sprintf(buf, "%u\n",
1607 from_kgid_munged(current_user_ns(), tun->group)):
1608 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001609}
1610
1611static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1612static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1613static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1614
Takashi Iwaic4d33e22015-02-04 14:37:34 +01001615static struct attribute *tun_dev_attrs[] = {
1616 &dev_attr_tun_flags.attr,
1617 &dev_attr_owner.attr,
1618 &dev_attr_group.attr,
1619 NULL
1620};
1621
1622static const struct attribute_group tun_attr_group = {
1623 .attrs = tun_dev_attrs
1624};
1625
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001626static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627{
1628 struct tun_struct *tun;
Jason Wang54f968d2012-10-31 19:45:57 +00001629 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 struct net_device *dev;
1631 int err;
1632
Jason Wang7c0c3b12013-01-11 16:59:33 +00001633 if (tfile->detached)
1634 return -EINVAL;
1635
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001636 dev = __dev_get_by_name(net, ifr->ifr_name);
1637 if (dev) {
David Woodhousef85ba782009-04-27 03:23:54 -07001638 if (ifr->ifr_flags & IFF_TUN_EXCL)
1639 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001640 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1641 tun = netdev_priv(dev);
1642 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1643 tun = netdev_priv(dev);
1644 else
1645 return -EINVAL;
1646
Jason Wang8e6d91a2013-05-28 18:32:11 +00001647 if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) !=
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001648 !!(tun->flags & IFF_MULTI_QUEUE))
Jason Wang8e6d91a2013-05-28 18:32:11 +00001649 return -EINVAL;
1650
Jason Wangcde8b152012-10-31 19:46:01 +00001651 if (tun_not_capable(tun))
Paul Moore2b980db2009-08-28 18:12:43 -04001652 return -EPERM;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001653 err = security_tun_dev_open(tun->security);
Paul Moore2b980db2009-08-28 18:12:43 -04001654 if (err < 0)
1655 return err;
1656
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001657 err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001658 if (err < 0)
1659 return err;
Jason Wang4008e972012-12-13 23:53:30 +00001660
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001661 if (tun->flags & IFF_MULTI_QUEUE &&
Jason Wange8dbad62013-04-22 20:40:39 +00001662 (tun->numqueues + tun->numdisabled > 1)) {
1663 /* One or more queue has already been attached, no need
1664 * to initialize the device again.
1665 */
1666 return 0;
1667 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001668 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 else {
1670 char *name;
1671 unsigned long flags = 0;
Jason Wangedfb6a12013-01-23 03:59:12 +00001672 int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
1673 MAX_TAP_QUEUES : 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674
Eric W. Biedermanc260b772012-11-18 21:34:11 +00001675 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001676 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001677 err = security_tun_dev_create();
1678 if (err < 0)
1679 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001680
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 /* Set dev type */
1682 if (ifr->ifr_flags & IFF_TUN) {
1683 /* TUN device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001684 flags |= IFF_TUN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 name = "tun%d";
1686 } else if (ifr->ifr_flags & IFF_TAP) {
1687 /* TAP device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001688 flags |= IFF_TAP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001690 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001691 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001692
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 if (*ifr->ifr_name)
1694 name = ifr->ifr_name;
1695
Jason Wangc8d68e62012-10-31 19:46:00 +00001696 dev = alloc_netdev_mqs(sizeof(struct tun_struct), name,
Tom Gundersenc835a672014-07-14 16:37:24 +02001697 NET_NAME_UNKNOWN, tun_setup, queues,
1698 queues);
Jason Wangedfb6a12013-01-23 03:59:12 +00001699
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 if (!dev)
1701 return -ENOMEM;
1702
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001703 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001704 dev->rtnl_link_ops = &tun_link_ops;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001705 dev->ifindex = tfile->ifindex;
Takashi Iwaic4d33e22015-02-04 14:37:34 +01001706 dev->sysfs_groups[0] = &tun_attr_group;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001707
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 tun = netdev_priv(dev);
1709 tun->dev = dev;
1710 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001711 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001712 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Paolo Abenieaea34b2016-02-26 10:45:40 +01001714 tun->align = NET_SKB_PAD;
Jason Wang54f968d2012-10-31 19:45:57 +00001715 tun->filter_attached = false;
1716 tun->sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001717
Jason Wang96442e422012-10-31 19:46:02 +00001718 spin_lock_init(&tun->lock);
1719
Paul Moore5dbbaf22013-01-14 07:12:19 +00001720 err = security_tun_dev_alloc_security(&tun->security);
1721 if (err < 0)
1722 goto err_free_dev;
Paul Moore2b980db2009-08-28 18:12:43 -04001723
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 tun_net_init(dev);
Pavel Emelyanov944a1372013-06-11 17:01:08 +04001725 tun_flow_init(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001726
Michał Mirosław88255372011-04-19 06:13:10 +00001727 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
Jason Wang6680ec62013-07-25 13:00:33 +08001728 TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
1729 NETIF_F_HW_VLAN_STAG_TX;
Michał Mirosław88255372011-04-19 06:13:10 +00001730 dev->features = dev->hw_features;
Fernando Luis Vazquez Cao6671b222014-02-18 21:20:09 +09001731 dev->vlan_features = dev->features &
1732 ~(NETIF_F_HW_VLAN_CTAG_TX |
1733 NETIF_F_HW_VLAN_STAG_TX);
Michał Mirosław88255372011-04-19 06:13:10 +00001734
Jason Wang4008e972012-12-13 23:53:30 +00001735 INIT_LIST_HEAD(&tun->disabled);
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001736 err = tun_attach(tun, file, false);
Jason Wangeb0fb362012-12-02 17:19:45 +00001737 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001738 goto err_free_flow;
Jason Wangeb0fb362012-12-02 17:19:45 +00001739
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 err = register_netdevice(tun->dev);
1741 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001742 goto err_detach;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 }
1744
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +00001745 netif_carrier_on(tun->dev);
1746
Joe Perches6b8a66e2011-03-02 07:18:10 +00001747 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001749 tun->flags = (tun->flags & ~TUN_FEATURES) |
1750 (ifr->ifr_flags & TUN_FEATURES);
Jason Wangc8d68e62012-10-31 19:46:00 +00001751
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001752 /* Make sure persistent devices do not get stuck in
1753 * xoff state.
1754 */
1755 if (netif_running(tun->dev))
Jason Wangc8d68e62012-10-31 19:46:00 +00001756 netif_tx_wake_all_queues(tun->dev);
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001757
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 strcpy(ifr->ifr_name, tun->dev->name);
1759 return 0;
1760
Jason Wang662ca432013-09-11 18:09:48 +08001761err_detach:
1762 tun_detach_all(dev);
1763err_free_flow:
1764 tun_flow_uninit(tun);
1765 security_tun_dev_free_security(tun->security);
1766err_free_dev:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 return err;
1769}
1770
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001771static void tun_get_iff(struct net *net, struct tun_struct *tun,
Herbert Xu876bfd42009-08-06 14:22:44 +00001772 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001773{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001774 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001775
1776 strcpy(ifr->ifr_name, tun->dev->name);
1777
David Woodhouse980c9e82009-05-09 22:54:21 -07001778 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001779
Mark McLoughline3b99552008-08-15 15:09:56 -07001780}
1781
Rusty Russell5228ddc2008-07-03 03:46:16 -07001782/* This is like a cut-down ethtool ops, except done via tun fd so no
1783 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001784static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001785{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001786 netdev_features_t features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001787
1788 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001789 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001790 arg &= ~TUN_F_CSUM;
1791
1792 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1793 if (arg & TUN_F_TSO_ECN) {
1794 features |= NETIF_F_TSO_ECN;
1795 arg &= ~TUN_F_TSO_ECN;
1796 }
1797 if (arg & TUN_F_TSO4)
1798 features |= NETIF_F_TSO;
1799 if (arg & TUN_F_TSO6)
1800 features |= NETIF_F_TSO6;
1801 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1802 }
Vlad Yaseviche3e3c422015-02-03 16:36:17 -05001803
1804 if (arg & TUN_F_UFO) {
1805 features |= NETIF_F_UFO;
1806 arg &= ~TUN_F_UFO;
1807 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001808 }
1809
1810 /* This gives the user a way to test for new features in future by
1811 * trying to set them. */
1812 if (arg)
1813 return -EINVAL;
1814
Michał Mirosław88255372011-04-19 06:13:10 +00001815 tun->set_features = features;
1816 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001817
1818 return 0;
1819}
1820
Jason Wangc8d68e62012-10-31 19:46:00 +00001821static void tun_detach_filter(struct tun_struct *tun, int n)
1822{
1823 int i;
1824 struct tun_file *tfile;
1825
1826 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001827 tfile = rtnl_dereference(tun->tfiles[i]);
Hannes Frederic Sowa8ced4252016-04-05 17:10:16 +02001828 lock_sock(tfile->socket.sk);
1829 sk_detach_filter(tfile->socket.sk);
1830 release_sock(tfile->socket.sk);
Jason Wangc8d68e62012-10-31 19:46:00 +00001831 }
1832
1833 tun->filter_attached = false;
1834}
1835
1836static int tun_attach_filter(struct tun_struct *tun)
1837{
1838 int i, ret = 0;
1839 struct tun_file *tfile;
1840
1841 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001842 tfile = rtnl_dereference(tun->tfiles[i]);
Hannes Frederic Sowa8ced4252016-04-05 17:10:16 +02001843 lock_sock(tfile->socket.sk);
1844 ret = sk_attach_filter(&tun->fprog, tfile->socket.sk);
1845 release_sock(tfile->socket.sk);
Jason Wangc8d68e62012-10-31 19:46:00 +00001846 if (ret) {
1847 tun_detach_filter(tun, i);
1848 return ret;
1849 }
1850 }
1851
1852 tun->filter_attached = true;
1853 return ret;
1854}
1855
1856static void tun_set_sndbuf(struct tun_struct *tun)
1857{
1858 struct tun_file *tfile;
1859 int i;
1860
1861 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001862 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001863 tfile->socket.sk->sk_sndbuf = tun->sndbuf;
1864 }
1865}
1866
Jason Wangcde8b152012-10-31 19:46:01 +00001867static int tun_set_queue(struct file *file, struct ifreq *ifr)
1868{
1869 struct tun_file *tfile = file->private_data;
1870 struct tun_struct *tun;
Jason Wangcde8b152012-10-31 19:46:01 +00001871 int ret = 0;
1872
1873 rtnl_lock();
1874
1875 if (ifr->ifr_flags & IFF_ATTACH_QUEUE) {
Jason Wang4008e972012-12-13 23:53:30 +00001876 tun = tfile->detached;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001877 if (!tun) {
Jason Wangcde8b152012-10-31 19:46:01 +00001878 ret = -EINVAL;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001879 goto unlock;
1880 }
1881 ret = security_tun_dev_attach_queue(tun->security);
1882 if (ret < 0)
1883 goto unlock;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001884 ret = tun_attach(tun, file, false);
Jason Wang4008e972012-12-13 23:53:30 +00001885 } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001886 tun = rtnl_dereference(tfile->tun);
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001887 if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached)
Jason Wang4008e972012-12-13 23:53:30 +00001888 ret = -EINVAL;
1889 else
1890 __tun_detach(tfile, false);
1891 } else
Jason Wangcde8b152012-10-31 19:46:01 +00001892 ret = -EINVAL;
1893
Paul Moore5dbbaf22013-01-14 07:12:19 +00001894unlock:
Jason Wangcde8b152012-10-31 19:46:01 +00001895 rtnl_unlock();
1896 return ret;
1897}
1898
Arnd Bergmann50857e22009-11-06 22:52:32 -08001899static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1900 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001902 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001903 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 void __user* argp = (void __user*)arg;
1905 struct ifreq ifr;
Eric W. Biederman0625c882012-02-07 16:48:55 -08001906 kuid_t owner;
1907 kgid_t group;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001908 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001909 int vnet_hdr_sz;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001910 unsigned int ifindex;
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02001911 int le;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001912 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Jason Wangcde8b152012-10-31 19:46:01 +00001914 if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) {
Arnd Bergmann50857e22009-11-06 22:52:32 -08001915 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 return -EFAULT;
David S. Miller8bbb1812012-07-30 14:52:48 -07001917 } else {
Mathias Krausea117dac2012-07-29 19:45:14 +00001918 memset(&ifr, 0, sizeof(ifr));
David S. Miller8bbb1812012-07-30 14:52:48 -07001919 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001920 if (cmd == TUNGETFEATURES) {
1921 /* Currently this just means: "what IFF flags are valid?".
1922 * This is needed because we never checked for invalid flags on
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001923 * TUNSETIFF.
1924 */
1925 return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES,
Eric W. Biederman631ab462009-01-20 11:00:40 +00001926 (unsigned int __user*)argp);
Jason Wangcde8b152012-10-31 19:46:01 +00001927 } else if (cmd == TUNSETQUEUE)
1928 return tun_set_queue(file, &ifr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001929
Jason Wangc8d68e62012-10-31 19:46:00 +00001930 ret = 0;
Herbert Xu876bfd42009-08-06 14:22:44 +00001931 rtnl_lock();
1932
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001933 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1936
Eric W. Biederman140e807da2015-05-08 21:07:08 -05001937 ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938
Herbert Xu876bfd42009-08-06 14:22:44 +00001939 if (ret)
1940 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941
Arnd Bergmann50857e22009-11-06 22:52:32 -08001942 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001943 ret = -EFAULT;
1944 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 }
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001946 if (cmd == TUNSETIFINDEX) {
1947 ret = -EPERM;
1948 if (tun)
1949 goto unlock;
1950
1951 ret = -EFAULT;
1952 if (copy_from_user(&ifindex, argp, sizeof(ifindex)))
1953 goto unlock;
1954
1955 ret = 0;
1956 tfile->ifindex = ifindex;
1957 goto unlock;
1958 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959
Herbert Xu876bfd42009-08-06 14:22:44 +00001960 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001962 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963
Jason Wang1e588332012-10-31 19:45:56 +00001964 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965
Eric W. Biederman631ab462009-01-20 11:00:40 +00001966 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001968 case TUNGETIFF:
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001969 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001970
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001971 if (tfile->detached)
1972 ifr.ifr_flags |= IFF_DETACH_QUEUE;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001973 if (!tfile->socket.sk->sk_filter)
1974 ifr.ifr_flags |= IFF_NOFILTER;
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001975
Arnd Bergmann50857e22009-11-06 22:52:32 -08001976 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001977 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001978 break;
1979
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 case TUNSETNOCSUM:
1981 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982
Michał Mirosław88255372011-04-19 06:13:10 +00001983 /* [unimplemented] */
1984 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00001985 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 break;
1987
1988 case TUNSETPERSIST:
Jason Wang54f968d2012-10-31 19:45:57 +00001989 /* Disable/Enable persist mode. Keep an extra reference to the
1990 * module to prevent the module being unprobed.
1991 */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001992 if (arg && !(tun->flags & IFF_PERSIST)) {
1993 tun->flags |= IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001994 __module_get(THIS_MODULE);
Jason Wangdd38bd82013-01-11 16:59:34 +00001995 }
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001996 if (!arg && (tun->flags & IFF_PERSIST)) {
1997 tun->flags &= ~IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001998 module_put(THIS_MODULE);
1999 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000
Joe Perches6b8a66e2011-03-02 07:18:10 +00002001 tun_debug(KERN_INFO, tun, "persist %s\n",
2002 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 break;
2004
2005 case TUNSETOWNER:
2006 /* Set owner of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08002007 owner = make_kuid(current_user_ns(), arg);
2008 if (!uid_valid(owner)) {
2009 ret = -EINVAL;
2010 break;
2011 }
2012 tun->owner = owner;
Jason Wang1e588332012-10-31 19:45:56 +00002013 tun_debug(KERN_INFO, tun, "owner set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08002014 from_kuid(&init_user_ns, tun->owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 break;
2016
Guido Guenther8c644622007-07-02 22:50:25 -07002017 case TUNSETGROUP:
2018 /* Set group of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08002019 group = make_kgid(current_user_ns(), arg);
2020 if (!gid_valid(group)) {
2021 ret = -EINVAL;
2022 break;
2023 }
2024 tun->group = group;
Jason Wang1e588332012-10-31 19:45:56 +00002025 tun_debug(KERN_INFO, tun, "group set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08002026 from_kgid(&init_user_ns, tun->group));
Guido Guenther8c644622007-07-02 22:50:25 -07002027 break;
2028
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002029 case TUNSETLINK:
2030 /* Only allow setting the type when the interface is down */
2031 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002032 tun_debug(KERN_INFO, tun,
2033 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07002034 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002035 } else {
2036 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00002037 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
2038 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07002039 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002040 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00002041 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002042
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043#ifdef TUN_DEBUG
2044 case TUNSETDEBUG:
2045 tun->debug = arg;
2046 break;
2047#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07002048 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00002049 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002050 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07002051
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002052 case TUNSETTXFILTER:
2053 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00002054 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002055 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Eric W. Biederman631ab462009-01-20 11:00:40 +00002056 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07002057 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002058 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059
2060 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04002061 /* Get hw address */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002062 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
2063 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08002064 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00002065 ret = -EFAULT;
2066 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067
2068 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002069 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00002070 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
2071 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08002072
Kim B. Heino40102372008-02-29 12:26:21 -08002073 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002074 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002075
2076 case TUNGETSNDBUF:
Jason Wang54f968d2012-10-31 19:45:57 +00002077 sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002078 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
2079 ret = -EFAULT;
2080 break;
2081
2082 case TUNSETSNDBUF:
2083 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
2084 ret = -EFAULT;
2085 break;
2086 }
2087
Jason Wangc8d68e62012-10-31 19:46:00 +00002088 tun->sndbuf = sndbuf;
2089 tun_set_sndbuf(tun);
Herbert Xu33dccbb2009-02-05 21:25:32 -08002090 break;
2091
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02002092 case TUNGETVNETHDRSZ:
2093 vnet_hdr_sz = tun->vnet_hdr_sz;
2094 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
2095 ret = -EFAULT;
2096 break;
2097
2098 case TUNSETVNETHDRSZ:
2099 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
2100 ret = -EFAULT;
2101 break;
2102 }
2103 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
2104 ret = -EINVAL;
2105 break;
2106 }
2107
2108 tun->vnet_hdr_sz = vnet_hdr_sz;
2109 break;
2110
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02002111 case TUNGETVNETLE:
2112 le = !!(tun->flags & TUN_VNET_LE);
2113 if (put_user(le, (int __user *)argp))
2114 ret = -EFAULT;
2115 break;
2116
2117 case TUNSETVNETLE:
2118 if (get_user(le, (int __user *)argp)) {
2119 ret = -EFAULT;
2120 break;
2121 }
2122 if (le)
2123 tun->flags |= TUN_VNET_LE;
2124 else
2125 tun->flags &= ~TUN_VNET_LE;
2126 break;
2127
Greg Kurz8b8e6582015-04-24 14:50:36 +02002128 case TUNGETVNETBE:
2129 ret = tun_get_vnet_be(tun, argp);
2130 break;
2131
2132 case TUNSETVNETBE:
2133 ret = tun_set_vnet_be(tun, argp);
2134 break;
2135
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002136 case TUNATTACHFILTER:
2137 /* Can be set only for TAPs */
2138 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002139 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002140 break;
2141 ret = -EFAULT;
Jason Wang54f968d2012-10-31 19:45:57 +00002142 if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog)))
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002143 break;
2144
Jason Wangc8d68e62012-10-31 19:46:00 +00002145 ret = tun_attach_filter(tun);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002146 break;
2147
2148 case TUNDETACHFILTER:
2149 /* Can be set only for TAPs */
2150 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002151 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002152 break;
Jason Wangc8d68e62012-10-31 19:46:00 +00002153 ret = 0;
2154 tun_detach_filter(tun, tun->numqueues);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002155 break;
2156
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002157 case TUNGETFILTER:
2158 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002159 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002160 break;
2161 ret = -EFAULT;
2162 if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog)))
2163 break;
2164 ret = 0;
2165 break;
2166
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00002168 ret = -EINVAL;
2169 break;
Joe Perchesee289b62010-05-17 22:47:34 -07002170 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171
Herbert Xu876bfd42009-08-06 14:22:44 +00002172unlock:
2173 rtnl_unlock();
2174 if (tun)
2175 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002176 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177}
2178
Arnd Bergmann50857e22009-11-06 22:52:32 -08002179static long tun_chr_ioctl(struct file *file,
2180 unsigned int cmd, unsigned long arg)
2181{
2182 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
2183}
2184
2185#ifdef CONFIG_COMPAT
2186static long tun_chr_compat_ioctl(struct file *file,
2187 unsigned int cmd, unsigned long arg)
2188{
2189 switch (cmd) {
2190 case TUNSETIFF:
2191 case TUNGETIFF:
2192 case TUNSETTXFILTER:
2193 case TUNGETSNDBUF:
2194 case TUNSETSNDBUF:
2195 case SIOCGIFHWADDR:
2196 case SIOCSIFHWADDR:
2197 arg = (unsigned long)compat_ptr(arg);
2198 break;
2199 default:
2200 arg = (compat_ulong_t)arg;
2201 break;
2202 }
2203
2204 /*
2205 * compat_ifreq is shorter than ifreq, so we must not access beyond
2206 * the end of that structure. All fields that are used in this
2207 * driver are compatible though, we don't need to convert the
2208 * contents.
2209 */
2210 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
2211}
2212#endif /* CONFIG_COMPAT */
2213
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214static int tun_chr_fasync(int fd, struct file *file, int on)
2215{
Jason Wang54f968d2012-10-31 19:45:57 +00002216 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 int ret;
2218
Jason Wang54f968d2012-10-31 19:45:57 +00002219 if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06002220 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002221
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222 if (on) {
Jeff Laytone0b93ed2014-08-22 11:27:32 -04002223 __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Jason Wang54f968d2012-10-31 19:45:57 +00002224 tfile->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002225 } else
Jason Wang54f968d2012-10-31 19:45:57 +00002226 tfile->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06002227 ret = 0;
2228out:
Jonathan Corbet9d319522008-06-19 15:50:37 -06002229 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230}
2231
2232static int tun_chr_open(struct inode *inode, struct file * file)
2233{
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002234 struct net *net = current->nsproxy->net_ns;
Eric W. Biederman631ab462009-01-20 11:00:40 +00002235 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07002236
Joe Perches6b8a66e2011-03-02 07:18:10 +00002237 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00002238
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002239 tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05002240 &tun_proto, 0);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002241 if (!tfile)
2242 return -ENOMEM;
Monam Agarwalc9566742014-03-24 00:02:32 +05302243 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang54f968d2012-10-31 19:45:57 +00002244 tfile->flags = 0;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002245 tfile->ifindex = 0;
Jason Wang54f968d2012-10-31 19:45:57 +00002246
Jason Wang54f968d2012-10-31 19:45:57 +00002247 init_waitqueue_head(&tfile->wq.wait);
Xi Wang9e641bd2014-05-16 15:11:48 -07002248 RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq);
Jason Wang54f968d2012-10-31 19:45:57 +00002249
2250 tfile->socket.file = file;
2251 tfile->socket.ops = &tun_socket_ops;
2252
2253 sock_init_data(&tfile->socket, &tfile->sk);
Jason Wang54f968d2012-10-31 19:45:57 +00002254
2255 tfile->sk.sk_write_space = tun_sock_write_space;
2256 tfile->sk.sk_sndbuf = INT_MAX;
2257
Eric W. Biederman631ab462009-01-20 11:00:40 +00002258 file->private_data = tfile;
Jason Wang4008e972012-12-13 23:53:30 +00002259 INIT_LIST_HEAD(&tfile->next);
Jason Wang54f968d2012-10-31 19:45:57 +00002260
Jason Wang19a6afb2013-06-08 14:17:41 +08002261 sock_set_flag(&tfile->sk, SOCK_ZEROCOPY);
2262
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263 return 0;
2264}
2265
2266static int tun_chr_close(struct inode *inode, struct file *file)
2267{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002268 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269
Jason Wangc8d68e62012-10-31 19:46:00 +00002270 tun_detach(tfile, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271
2272 return 0;
2273}
2274
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002275#ifdef CONFIG_PROC_FS
Joe Perchesa3816ab2014-09-29 16:08:25 -07002276static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f)
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002277{
2278 struct tun_struct *tun;
2279 struct ifreq ifr;
2280
2281 memset(&ifr, 0, sizeof(ifr));
2282
2283 rtnl_lock();
2284 tun = tun_get(f);
2285 if (tun)
2286 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
2287 rtnl_unlock();
2288
2289 if (tun)
2290 tun_put(tun);
2291
Joe Perchesa3816ab2014-09-29 16:08:25 -07002292 seq_printf(m, "iff:\t%s\n", ifr.ifr_name);
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002293}
2294#endif
2295
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08002296static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002297 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298 .llseek = no_llseek,
Al Viro9b067032014-11-07 13:52:07 -05002299 .read_iter = tun_chr_read_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04002300 .write_iter = tun_chr_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08002302 .unlocked_ioctl = tun_chr_ioctl,
2303#ifdef CONFIG_COMPAT
2304 .compat_ioctl = tun_chr_compat_ioctl,
2305#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 .open = tun_chr_open,
2307 .release = tun_chr_close,
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002308 .fasync = tun_chr_fasync,
2309#ifdef CONFIG_PROC_FS
2310 .show_fdinfo = tun_chr_show_fdinfo,
2311#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002312};
2313
2314static struct miscdevice tun_miscdev = {
2315 .minor = TUN_MINOR,
2316 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02002317 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319};
2320
2321/* ethtool interface */
2322
2323static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
2324{
2325 cmd->supported = 0;
2326 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00002327 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328 cmd->duplex = DUPLEX_FULL;
2329 cmd->port = PORT_TP;
2330 cmd->phy_address = 0;
2331 cmd->transceiver = XCVR_INTERNAL;
2332 cmd->autoneg = AUTONEG_DISABLE;
2333 cmd->maxtxpkt = 0;
2334 cmd->maxrxpkt = 0;
2335 return 0;
2336}
2337
2338static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
2339{
2340 struct tun_struct *tun = netdev_priv(dev);
2341
Rick Jones33a5ba12011-11-15 14:59:53 +00002342 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
2343 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344
2345 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002346 case IFF_TUN:
Rick Jones33a5ba12011-11-15 14:59:53 +00002347 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002349 case IFF_TAP:
Rick Jones33a5ba12011-11-15 14:59:53 +00002350 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 break;
2352 }
2353}
2354
2355static u32 tun_get_msglevel(struct net_device *dev)
2356{
2357#ifdef TUN_DEBUG
2358 struct tun_struct *tun = netdev_priv(dev);
2359 return tun->debug;
2360#else
2361 return -EOPNOTSUPP;
2362#endif
2363}
2364
2365static void tun_set_msglevel(struct net_device *dev, u32 value)
2366{
2367#ifdef TUN_DEBUG
2368 struct tun_struct *tun = netdev_priv(dev);
2369 tun->debug = value;
2370#endif
2371}
2372
Jeff Garzik7282d492006-09-13 14:30:00 -04002373static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 .get_settings = tun_get_settings,
2375 .get_drvinfo = tun_get_drvinfo,
2376 .get_msglevel = tun_get_msglevel,
2377 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00002378 .get_link = ethtool_op_get_link,
Richard Cochraneda29772013-07-19 19:40:10 +02002379 .get_ts_info = ethtool_op_get_ts_info,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380};
2381
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002382
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383static int __init tun_init(void)
2384{
2385 int ret = 0;
2386
Joe Perches6b8a66e2011-03-02 07:18:10 +00002387 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
2388 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002390 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002391 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002392 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002393 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002394 }
2395
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002397 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002398 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002399 goto err_misc;
2400 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002401 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002402err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002403 rtnl_link_unregister(&tun_link_ops);
2404err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002405 return ret;
2406}
2407
2408static void tun_cleanup(void)
2409{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002410 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002411 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412}
2413
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002414/* Get an underlying socket object from tun file. Returns error unless file is
2415 * attached to a device. The returned object works like a packet socket, it
2416 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
2417 * holding a reference to the file for as long as the socket is in use. */
2418struct socket *tun_get_socket(struct file *file)
2419{
Jason Wang6e914fc2012-10-31 19:45:58 +00002420 struct tun_file *tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002421 if (file->f_op != &tun_fops)
2422 return ERR_PTR(-EINVAL);
Jason Wang6e914fc2012-10-31 19:45:58 +00002423 tfile = file->private_data;
2424 if (!tfile)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002425 return ERR_PTR(-EBADFD);
Jason Wang54f968d2012-10-31 19:45:57 +00002426 return &tfile->socket;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002427}
2428EXPORT_SYMBOL_GPL(tun_get_socket);
2429
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430module_init(tun_init);
2431module_exit(tun_cleanup);
2432MODULE_DESCRIPTION(DRV_DESCRIPTION);
2433MODULE_AUTHOR(DRV_COPYRIGHT);
2434MODULE_LICENSE("GPL");
2435MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02002436MODULE_ALIAS("devname:net/tun");