blob: 186ce541c65762f8ee1720aae7f573f145982406 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
Linus Torvalds1da177e2005-04-16 15:20:36 -070074#include <asm/uaccess.h>
75
Rusty Russell14daa022008-04-12 18:48:58 -070076/* Uncomment to enable debugging */
77/* #define TUN_DEBUG 1 */
78
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#ifdef TUN_DEBUG
80static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070081
Joe Perches6b8a66e2011-03-02 07:18:10 +000082#define tun_debug(level, tun, fmt, args...) \
83do { \
84 if (tun->debug) \
85 netdev_printk(level, tun->dev, fmt, ##args); \
86} while (0)
87#define DBG1(level, fmt, args...) \
88do { \
89 if (debug == 2) \
90 printk(level fmt, ##args); \
91} while (0)
Rusty Russell14daa022008-04-12 18:48:58 -070092#else
Joe Perches6b8a66e2011-03-02 07:18:10 +000093#define tun_debug(level, tun, fmt, args...) \
94do { \
95 if (0) \
96 netdev_printk(level, tun->dev, fmt, ##args); \
97} while (0)
98#define DBG1(level, fmt, args...) \
99do { \
100 if (0) \
101 printk(level fmt, ##args); \
102} while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103#endif
104
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000105#define GOODCOPY_LEN 128
106
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700107#define FLT_EXACT_COUNT 8
108struct tap_filter {
109 unsigned int count; /* Number of addrs. Zero means disabled */
110 u32 mask[2]; /* Mask of the hashed addrs */
111 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
112};
113
stephen hemminger92d4ea62013-12-05 20:42:58 -0800114/* DEFAULT_MAX_NUM_RSS_QUEUES were chosen to let the rx/tx queues allocated for
Jason Wangedfb6a12013-01-23 03:59:12 +0000115 * the netdevice to be fit in one page. So we can make sure the success of
116 * memory allocation. TODO: increase the limit. */
117#define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES
Jason Wangb8732fb2013-01-23 03:59:13 +0000118#define MAX_TAP_FLOWS 4096
Jason Wangc8d68e62012-10-31 19:46:00 +0000119
Jason Wang96442e422012-10-31 19:46:02 +0000120#define TUN_FLOW_EXPIRE (3 * HZ)
121
Jason Wang54f968d2012-10-31 19:45:57 +0000122/* A tun_file connects an open character device to a tuntap netdevice. It
stephen hemminger92d4ea62013-12-05 20:42:58 -0800123 * also contains all socket related structures (except sock_fprog and tap_filter)
Jason Wang54f968d2012-10-31 19:45:57 +0000124 * to serve as one transmit queue for tuntap device. The sock_fprog and
125 * tap_filter were kept in tun_struct since they were used for filtering for the
Rami Rosen36fe8c092012-11-25 22:07:40 +0000126 * netdevice not for a specific queue (at least I didn't see the requirement for
Jason Wang54f968d2012-10-31 19:45:57 +0000127 * this).
Jason Wang6e914fc2012-10-31 19:45:58 +0000128 *
129 * RCU usage:
Rami Rosen36fe8c092012-11-25 22:07:40 +0000130 * The tun_file and tun_struct are loosely coupled, the pointer from one to the
Jason Wang6e914fc2012-10-31 19:45:58 +0000131 * other can only be read while rcu_read_lock or rtnl_lock is held.
Jason Wang54f968d2012-10-31 19:45:57 +0000132 */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000133struct tun_file {
Jason Wang54f968d2012-10-31 19:45:57 +0000134 struct sock sk;
135 struct socket socket;
136 struct socket_wq wq;
Jason Wang6e914fc2012-10-31 19:45:58 +0000137 struct tun_struct __rcu *tun;
Eric W. Biederman36b50ba2009-01-20 11:01:48 +0000138 struct net *net;
Jason Wang54f968d2012-10-31 19:45:57 +0000139 struct fasync_struct *fasync;
140 /* only used for fasnyc */
141 unsigned int flags;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +0400142 union {
143 u16 queue_index;
144 unsigned int ifindex;
145 };
Jason Wang4008e972012-12-13 23:53:30 +0000146 struct list_head next;
147 struct tun_struct *detached;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000148};
149
Jason Wang96442e422012-10-31 19:46:02 +0000150struct tun_flow_entry {
151 struct hlist_node hash_link;
152 struct rcu_head rcu;
153 struct tun_struct *tun;
154
155 u32 rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800156 u32 rps_rxhash;
Jason Wang96442e422012-10-31 19:46:02 +0000157 int queue_index;
158 unsigned long updated;
159};
160
161#define TUN_NUM_FLOW_ENTRIES 1024
162
Jason Wang54f968d2012-10-31 19:45:57 +0000163/* Since the socket were moved to tun_file, to preserve the behavior of persist
Rami Rosen36fe8c092012-11-25 22:07:40 +0000164 * device, socket filter, sndbuf and vnet header size were restore when the
Jason Wang54f968d2012-10-31 19:45:57 +0000165 * file were attached to a persist device.
166 */
Rusty Russell14daa022008-04-12 18:48:58 -0700167struct tun_struct {
Jason Wangc8d68e62012-10-31 19:46:00 +0000168 struct tun_file __rcu *tfiles[MAX_TAP_QUEUES];
169 unsigned int numqueues;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700170 unsigned int flags;
Eric W. Biederman0625c882012-02-07 16:48:55 -0800171 kuid_t owner;
172 kgid_t group;
Rusty Russell14daa022008-04-12 18:48:58 -0700173
Rusty Russell14daa022008-04-12 18:48:58 -0700174 struct net_device *dev;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000175 netdev_features_t set_features;
Michał Mirosław88255372011-04-19 06:13:10 +0000176#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
177 NETIF_F_TSO6|NETIF_F_UFO)
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200178
179 int vnet_hdr_sz;
Jason Wang54f968d2012-10-31 19:45:57 +0000180 int sndbuf;
181 struct tap_filter txflt;
182 struct sock_fprog fprog;
183 /* protected by rtnl lock */
184 bool filter_attached;
Rusty Russell14daa022008-04-12 18:48:58 -0700185#ifdef TUN_DEBUG
186 int debug;
187#endif
Jason Wang96442e422012-10-31 19:46:02 +0000188 spinlock_t lock;
Jason Wang96442e422012-10-31 19:46:02 +0000189 struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
190 struct timer_list flow_gc_timer;
191 unsigned long ageing_time;
Jason Wang4008e972012-12-13 23:53:30 +0000192 unsigned int numdisabled;
193 struct list_head disabled;
Paul Moore5dbbaf22013-01-14 07:12:19 +0000194 void *security;
Jason Wangb8732fb2013-01-23 03:59:13 +0000195 u32 flow_count;
Rusty Russell14daa022008-04-12 18:48:58 -0700196};
197
Jason Wang96442e422012-10-31 19:46:02 +0000198static inline u32 tun_hashfn(u32 rxhash)
199{
200 return rxhash & 0x3ff;
201}
202
203static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash)
204{
205 struct tun_flow_entry *e;
Jason Wang96442e422012-10-31 19:46:02 +0000206
Sasha Levinb67bfe02013-02-27 17:06:00 -0800207 hlist_for_each_entry_rcu(e, head, hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000208 if (e->rxhash == rxhash)
209 return e;
210 }
211 return NULL;
212}
213
214static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun,
215 struct hlist_head *head,
216 u32 rxhash, u16 queue_index)
217{
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000218 struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC);
219
Jason Wang96442e422012-10-31 19:46:02 +0000220 if (e) {
221 tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n",
222 rxhash, queue_index);
223 e->updated = jiffies;
224 e->rxhash = rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800225 e->rps_rxhash = 0;
Jason Wang96442e422012-10-31 19:46:02 +0000226 e->queue_index = queue_index;
227 e->tun = tun;
228 hlist_add_head_rcu(&e->hash_link, head);
Jason Wangb8732fb2013-01-23 03:59:13 +0000229 ++tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000230 }
231 return e;
232}
233
Jason Wang96442e422012-10-31 19:46:02 +0000234static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e)
235{
236 tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n",
237 e->rxhash, e->queue_index);
Tom Herbert9bc88932013-12-22 18:54:32 +0800238 sock_rps_reset_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000239 hlist_del_rcu(&e->hash_link);
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000240 kfree_rcu(e, rcu);
Jason Wangb8732fb2013-01-23 03:59:13 +0000241 --tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000242}
243
244static void tun_flow_flush(struct tun_struct *tun)
245{
246 int i;
247
248 spin_lock_bh(&tun->lock);
249 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
250 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800251 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000252
Sasha Levinb67bfe02013-02-27 17:06:00 -0800253 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
Jason Wang96442e422012-10-31 19:46:02 +0000254 tun_flow_delete(tun, e);
255 }
256 spin_unlock_bh(&tun->lock);
257}
258
259static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
260{
261 int i;
262
263 spin_lock_bh(&tun->lock);
264 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
265 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800266 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000267
Sasha Levinb67bfe02013-02-27 17:06:00 -0800268 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000269 if (e->queue_index == queue_index)
270 tun_flow_delete(tun, e);
271 }
272 }
273 spin_unlock_bh(&tun->lock);
274}
275
276static void tun_flow_cleanup(unsigned long data)
277{
278 struct tun_struct *tun = (struct tun_struct *)data;
279 unsigned long delay = tun->ageing_time;
280 unsigned long next_timer = jiffies + delay;
281 unsigned long count = 0;
282 int i;
283
284 tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n");
285
286 spin_lock_bh(&tun->lock);
287 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
288 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800289 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000290
Sasha Levinb67bfe02013-02-27 17:06:00 -0800291 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000292 unsigned long this_timer;
293 count++;
294 this_timer = e->updated + delay;
295 if (time_before_eq(this_timer, jiffies))
296 tun_flow_delete(tun, e);
297 else if (time_before(this_timer, next_timer))
298 next_timer = this_timer;
299 }
300 }
301
302 if (count)
303 mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer));
304 spin_unlock_bh(&tun->lock);
305}
306
Eric Dumazet49974422012-12-12 19:22:57 +0000307static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
Jason Wang9e857222013-01-28 01:05:19 +0000308 struct tun_file *tfile)
Jason Wang96442e422012-10-31 19:46:02 +0000309{
310 struct hlist_head *head;
311 struct tun_flow_entry *e;
312 unsigned long delay = tun->ageing_time;
Jason Wang9e857222013-01-28 01:05:19 +0000313 u16 queue_index = tfile->queue_index;
Jason Wang96442e422012-10-31 19:46:02 +0000314
315 if (!rxhash)
316 return;
317 else
318 head = &tun->flows[tun_hashfn(rxhash)];
319
320 rcu_read_lock();
321
Jason Wang9e857222013-01-28 01:05:19 +0000322 /* We may get a very small possibility of OOO during switching, not
323 * worth to optimize.*/
324 if (tun->numqueues == 1 || tfile->detached)
Jason Wang96442e422012-10-31 19:46:02 +0000325 goto unlock;
326
327 e = tun_flow_find(head, rxhash);
328 if (likely(e)) {
329 /* TODO: keep queueing to old queue until it's empty? */
330 e->queue_index = queue_index;
331 e->updated = jiffies;
Tom Herbert9bc88932013-12-22 18:54:32 +0800332 sock_rps_record_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000333 } else {
334 spin_lock_bh(&tun->lock);
Jason Wangb8732fb2013-01-23 03:59:13 +0000335 if (!tun_flow_find(head, rxhash) &&
336 tun->flow_count < MAX_TAP_FLOWS)
Jason Wang96442e422012-10-31 19:46:02 +0000337 tun_flow_create(tun, head, rxhash, queue_index);
338
339 if (!timer_pending(&tun->flow_gc_timer))
340 mod_timer(&tun->flow_gc_timer,
341 round_jiffies_up(jiffies + delay));
342 spin_unlock_bh(&tun->lock);
343 }
344
345unlock:
346 rcu_read_unlock();
347}
348
Tom Herbert9bc88932013-12-22 18:54:32 +0800349/**
350 * Save the hash received in the stack receive path and update the
351 * flow_hash table accordingly.
352 */
353static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash)
354{
355 if (unlikely(e->rps_rxhash != hash)) {
356 sock_rps_reset_flow_hash(e->rps_rxhash);
357 e->rps_rxhash = hash;
358 }
359}
360
Jason Wangc8d68e62012-10-31 19:46:00 +0000361/* We try to identify a flow through its rxhash first. The reason that
stephen hemminger92d4ea62013-12-05 20:42:58 -0800362 * we do not check rxq no. is because some cards(e.g 82599), chooses
Jason Wangc8d68e62012-10-31 19:46:00 +0000363 * the rxq based on the txq where the last packet of the flow comes. As
364 * the userspace application move between processors, we may get a
365 * different rxq no. here. If we could not get rxhash, then we would
366 * hope the rxq no. may help here.
367 */
Jason Wangf663dd92014-01-10 16:18:26 +0800368static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100369 void *accel_priv, select_queue_fallback_t fallback)
Jason Wangc8d68e62012-10-31 19:46:00 +0000370{
371 struct tun_struct *tun = netdev_priv(dev);
Jason Wang96442e422012-10-31 19:46:02 +0000372 struct tun_flow_entry *e;
Jason Wangc8d68e62012-10-31 19:46:00 +0000373 u32 txq = 0;
374 u32 numqueues = 0;
375
376 rcu_read_lock();
Jason Wang92bb73e2013-06-05 16:44:57 +0800377 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000378
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800379 txq = skb_get_hash(skb);
Jason Wangc8d68e62012-10-31 19:46:00 +0000380 if (txq) {
Jason Wang96442e422012-10-31 19:46:02 +0000381 e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
Tom Herbert9bc88932013-12-22 18:54:32 +0800382 if (e) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800383 tun_flow_save_rps_rxhash(e, txq);
Zhi Yong Wufbe4d452014-01-02 13:24:28 +0800384 txq = e->queue_index;
Tom Herbert9bc88932013-12-22 18:54:32 +0800385 } else
Jason Wang96442e422012-10-31 19:46:02 +0000386 /* use multiply and shift instead of expensive divide */
387 txq = ((u64)txq * numqueues) >> 32;
Jason Wangc8d68e62012-10-31 19:46:00 +0000388 } else if (likely(skb_rx_queue_recorded(skb))) {
389 txq = skb_get_rx_queue(skb);
390 while (unlikely(txq >= numqueues))
391 txq -= numqueues;
392 }
393
394 rcu_read_unlock();
395 return txq;
396}
397
Jason Wangcde8b152012-10-31 19:46:01 +0000398static inline bool tun_not_capable(struct tun_struct *tun)
399{
400 const struct cred *cred = current_cred();
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000401 struct net *net = dev_net(tun->dev);
Jason Wangcde8b152012-10-31 19:46:01 +0000402
403 return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
404 (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000405 !ns_capable(net->user_ns, CAP_NET_ADMIN);
Jason Wangcde8b152012-10-31 19:46:01 +0000406}
407
Jason Wangc8d68e62012-10-31 19:46:00 +0000408static void tun_set_real_num_queues(struct tun_struct *tun)
409{
410 netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
411 netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
412}
413
Jason Wang4008e972012-12-13 23:53:30 +0000414static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
415{
416 tfile->detached = tun;
417 list_add_tail(&tfile->next, &tun->disabled);
418 ++tun->numdisabled;
419}
420
Jason Wangd32649d2012-12-18 11:00:27 +0800421static struct tun_struct *tun_enable_queue(struct tun_file *tfile)
Jason Wang4008e972012-12-13 23:53:30 +0000422{
423 struct tun_struct *tun = tfile->detached;
424
425 tfile->detached = NULL;
426 list_del_init(&tfile->next);
427 --tun->numdisabled;
428 return tun;
429}
430
Jason Wang4bfb0512013-09-05 17:53:59 +0800431static void tun_queue_purge(struct tun_file *tfile)
432{
433 skb_queue_purge(&tfile->sk.sk_receive_queue);
434 skb_queue_purge(&tfile->sk.sk_error_queue);
435}
436
Jason Wangc8d68e62012-10-31 19:46:00 +0000437static void __tun_detach(struct tun_file *tfile, bool clean)
438{
439 struct tun_file *ntfile;
440 struct tun_struct *tun;
Jason Wangc8d68e62012-10-31 19:46:00 +0000441
Jason Wangb8deabd2013-01-11 16:59:32 +0000442 tun = rtnl_dereference(tfile->tun);
443
Jason Wang9e857222013-01-28 01:05:19 +0000444 if (tun && !tfile->detached) {
Jason Wangc8d68e62012-10-31 19:46:00 +0000445 u16 index = tfile->queue_index;
446 BUG_ON(index >= tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000447
448 rcu_assign_pointer(tun->tfiles[index],
449 tun->tfiles[tun->numqueues - 1]);
Jason Wangb8deabd2013-01-11 16:59:32 +0000450 ntfile = rtnl_dereference(tun->tfiles[index]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000451 ntfile->queue_index = index;
452
453 --tun->numqueues;
Jason Wang9e857222013-01-28 01:05:19 +0000454 if (clean) {
Monam Agarwalc9566742014-03-24 00:02:32 +0530455 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang4008e972012-12-13 23:53:30 +0000456 sock_put(&tfile->sk);
Jason Wang9e857222013-01-28 01:05:19 +0000457 } else
Jason Wang4008e972012-12-13 23:53:30 +0000458 tun_disable_queue(tun, tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000459
460 synchronize_net();
Jason Wang96442e422012-10-31 19:46:02 +0000461 tun_flow_delete_by_queue(tun, tun->numqueues + 1);
Jason Wangc8d68e62012-10-31 19:46:00 +0000462 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800463 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000464 tun_set_real_num_queues(tun);
Jason Wangdd38bd82013-01-11 16:59:34 +0000465 } else if (tfile->detached && clean) {
Jason Wang4008e972012-12-13 23:53:30 +0000466 tun = tun_enable_queue(tfile);
Jason Wangdd38bd82013-01-11 16:59:34 +0000467 sock_put(&tfile->sk);
468 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000469
470 if (clean) {
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000471 if (tun && tun->numqueues == 0 && tun->numdisabled == 0) {
472 netif_carrier_off(tun->dev);
473
474 if (!(tun->flags & TUN_PERSIST) &&
475 tun->dev->reg_state == NETREG_REGISTERED)
Jason Wang4008e972012-12-13 23:53:30 +0000476 unregister_netdevice(tun->dev);
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000477 }
Jason Wang4008e972012-12-13 23:53:30 +0000478
Jason Wangc8d68e62012-10-31 19:46:00 +0000479 BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED,
480 &tfile->socket.flags));
481 sk_release_kernel(&tfile->sk);
482 }
483}
484
485static void tun_detach(struct tun_file *tfile, bool clean)
486{
487 rtnl_lock();
488 __tun_detach(tfile, clean);
489 rtnl_unlock();
490}
491
492static void tun_detach_all(struct net_device *dev)
493{
494 struct tun_struct *tun = netdev_priv(dev);
Jason Wang4008e972012-12-13 23:53:30 +0000495 struct tun_file *tfile, *tmp;
Jason Wangc8d68e62012-10-31 19:46:00 +0000496 int i, n = tun->numqueues;
497
498 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000499 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000500 BUG_ON(!tfile);
Xi Wang9e641bd2014-05-16 15:11:48 -0700501 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530502 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wangc8d68e62012-10-31 19:46:00 +0000503 --tun->numqueues;
504 }
Jason Wang9e857222013-01-28 01:05:19 +0000505 list_for_each_entry(tfile, &tun->disabled, next) {
Xi Wang9e641bd2014-05-16 15:11:48 -0700506 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530507 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang9e857222013-01-28 01:05:19 +0000508 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000509 BUG_ON(tun->numqueues != 0);
510
511 synchronize_net();
512 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000513 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000514 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800515 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000516 sock_put(&tfile->sk);
517 }
Jason Wang4008e972012-12-13 23:53:30 +0000518 list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
519 tun_enable_queue(tfile);
Jason Wang4bfb0512013-09-05 17:53:59 +0800520 tun_queue_purge(tfile);
Jason Wang4008e972012-12-13 23:53:30 +0000521 sock_put(&tfile->sk);
522 }
523 BUG_ON(tun->numdisabled != 0);
Jason Wangdd38bd82013-01-11 16:59:34 +0000524
525 if (tun->flags & TUN_PERSIST)
526 module_put(THIS_MODULE);
Jason Wangc8d68e62012-10-31 19:46:00 +0000527}
528
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400529static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter)
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000530{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000531 struct tun_file *tfile = file->private_data;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000532 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000533
Paul Moore5dbbaf22013-01-14 07:12:19 +0000534 err = security_tun_dev_attach(tfile->socket.sk, tun->security);
535 if (err < 0)
536 goto out;
537
Eric W. Biederman38231b72009-01-20 11:02:28 +0000538 err = -EINVAL;
Jason Wang9e857222013-01-28 01:05:19 +0000539 if (rtnl_dereference(tfile->tun) && !tfile->detached)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000540 goto out;
541
542 err = -EBUSY;
Jason Wangc8d68e62012-10-31 19:46:00 +0000543 if (!(tun->flags & TUN_TAP_MQ) && tun->numqueues == 1)
544 goto out;
545
546 err = -E2BIG;
Jason Wang4008e972012-12-13 23:53:30 +0000547 if (!tfile->detached &&
548 tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000549 goto out;
550
551 err = 0;
Jason Wang54f968d2012-10-31 19:45:57 +0000552
stephen hemminger92d4ea62013-12-05 20:42:58 -0800553 /* Re-attach the filter to persist device */
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400554 if (!skip_filter && (tun->filter_attached == true)) {
Jason Wang54f968d2012-10-31 19:45:57 +0000555 err = sk_attach_filter(&tun->fprog, tfile->socket.sk);
556 if (!err)
557 goto out;
558 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000559 tfile->queue_index = tun->numqueues;
Jason Wang6e914fc2012-10-31 19:45:58 +0000560 rcu_assign_pointer(tfile->tun, tun);
Jason Wangc8d68e62012-10-31 19:46:00 +0000561 rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000562 tun->numqueues++;
563
Jason Wang4008e972012-12-13 23:53:30 +0000564 if (tfile->detached)
565 tun_enable_queue(tfile);
566 else
567 sock_hold(&tfile->sk);
568
Jason Wangc8d68e62012-10-31 19:46:00 +0000569 tun_set_real_num_queues(tun);
570
Jason Wangc8d68e62012-10-31 19:46:00 +0000571 /* device is allowed to go away first, so no need to hold extra
572 * refcnt.
573 */
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000574
Eric W. Biederman38231b72009-01-20 11:02:28 +0000575out:
Eric W. Biederman38231b72009-01-20 11:02:28 +0000576 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000577}
578
Eric W. Biederman631ab462009-01-20 11:00:40 +0000579static struct tun_struct *__tun_get(struct tun_file *tfile)
580{
Jason Wang6e914fc2012-10-31 19:45:58 +0000581 struct tun_struct *tun;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000582
Jason Wang6e914fc2012-10-31 19:45:58 +0000583 rcu_read_lock();
584 tun = rcu_dereference(tfile->tun);
585 if (tun)
586 dev_hold(tun->dev);
587 rcu_read_unlock();
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000588
589 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000590}
591
592static struct tun_struct *tun_get(struct file *file)
593{
594 return __tun_get(file->private_data);
595}
596
597static void tun_put(struct tun_struct *tun)
598{
Jason Wang6e914fc2012-10-31 19:45:58 +0000599 dev_put(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000600}
601
Joe Perches6b8a66e2011-03-02 07:18:10 +0000602/* TAP filtering */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700603static void addr_hash_set(u32 *mask, const u8 *addr)
604{
605 int n = ether_crc(ETH_ALEN, addr) >> 26;
606 mask[n >> 5] |= (1 << (n & 31));
607}
608
609static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
610{
611 int n = ether_crc(ETH_ALEN, addr) >> 26;
612 return mask[n >> 5] & (1 << (n & 31));
613}
614
615static int update_filter(struct tap_filter *filter, void __user *arg)
616{
617 struct { u8 u[ETH_ALEN]; } *addr;
618 struct tun_filter uf;
619 int err, alen, n, nexact;
620
621 if (copy_from_user(&uf, arg, sizeof(uf)))
622 return -EFAULT;
623
624 if (!uf.count) {
625 /* Disabled */
626 filter->count = 0;
627 return 0;
628 }
629
630 alen = ETH_ALEN * uf.count;
631 addr = kmalloc(alen, GFP_KERNEL);
632 if (!addr)
633 return -ENOMEM;
634
635 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
636 err = -EFAULT;
637 goto done;
638 }
639
640 /* The filter is updated without holding any locks. Which is
641 * perfectly safe. We disable it first and in the worst
642 * case we'll accept a few undesired packets. */
643 filter->count = 0;
644 wmb();
645
646 /* Use first set of addresses as an exact filter */
647 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
648 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
649
650 nexact = n;
651
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800652 /* Remaining multicast addresses are hashed,
653 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700654 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800655 for (; n < uf.count; n++) {
656 if (!is_multicast_ether_addr(addr[n].u)) {
657 err = 0; /* no filter */
658 goto done;
659 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700660 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800661 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700662
663 /* For ALLMULTI just set the mask to all ones.
664 * This overrides the mask populated above. */
665 if ((uf.flags & TUN_FLT_ALLMULTI))
666 memset(filter->mask, ~0, sizeof(filter->mask));
667
668 /* Now enable the filter */
669 wmb();
670 filter->count = nexact;
671
672 /* Return the number of exact filters */
673 err = nexact;
674
675done:
676 kfree(addr);
677 return err;
678}
679
680/* Returns: 0 - drop, !=0 - accept */
681static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
682{
683 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
684 * at this point. */
685 struct ethhdr *eh = (struct ethhdr *) skb->data;
686 int i;
687
688 /* Exact match */
689 for (i = 0; i < filter->count; i++)
Joe Perches2e42e472012-05-09 17:17:46 +0000690 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700691 return 1;
692
693 /* Inexact match (multicast only) */
694 if (is_multicast_ether_addr(eh->h_dest))
695 return addr_hash_test(filter->mask, eh->h_dest);
696
697 return 0;
698}
699
700/*
701 * Checks whether the packet is accepted or not.
702 * Returns: 0 - drop, !=0 - accept
703 */
704static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
705{
706 if (!filter->count)
707 return 1;
708
709 return run_filter(filter, skb);
710}
711
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712/* Network device part of the driver */
713
Jeff Garzik7282d492006-09-13 14:30:00 -0400714static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000716/* Net device detach from fd. */
717static void tun_net_uninit(struct net_device *dev)
718{
Jason Wangc8d68e62012-10-31 19:46:00 +0000719 tun_detach_all(dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000720}
721
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722/* Net device open. */
723static int tun_net_open(struct net_device *dev)
724{
Jason Wangc8d68e62012-10-31 19:46:00 +0000725 netif_tx_start_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 return 0;
727}
728
729/* Net device close. */
730static int tun_net_close(struct net_device *dev)
731{
Jason Wangc8d68e62012-10-31 19:46:00 +0000732 netif_tx_stop_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 return 0;
734}
735
736/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000737static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738{
739 struct tun_struct *tun = netdev_priv(dev);
Jason Wangc8d68e62012-10-31 19:46:00 +0000740 int txq = skb->queue_mapping;
Jason Wang6e914fc2012-10-31 19:45:58 +0000741 struct tun_file *tfile;
Dominic Curranfa358642014-01-22 03:03:23 +0000742 u32 numqueues = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Jason Wang6e914fc2012-10-31 19:45:58 +0000744 rcu_read_lock();
Jason Wangc8d68e62012-10-31 19:46:00 +0000745 tfile = rcu_dereference(tun->tfiles[txq]);
Dominic Curranfa358642014-01-22 03:03:23 +0000746 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 /* Drop packet if interface is not attached */
Dominic Curranfa358642014-01-22 03:03:23 +0000749 if (txq >= numqueues)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 goto drop;
751
Dominic Curranfa358642014-01-22 03:03:23 +0000752 if (numqueues == 1) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800753 /* Select queue was not called for the skbuff, so we extract the
754 * RPS hash and save it into the flow_table here.
755 */
756 __u32 rxhash;
757
758 rxhash = skb_get_hash(skb);
759 if (rxhash) {
760 struct tun_flow_entry *e;
761 e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)],
762 rxhash);
763 if (e)
764 tun_flow_save_rps_rxhash(e, rxhash);
765 }
766 }
767
Jason Wang6e914fc2012-10-31 19:45:58 +0000768 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
769
Jason Wangc8d68e62012-10-31 19:46:00 +0000770 BUG_ON(!tfile);
771
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700772 /* Drop if the filter does not like it.
773 * This is a noop if the filter is disabled.
774 * Filter can be enabled only for the TAP devices. */
775 if (!check_filter(&tun->txflt, skb))
776 goto drop;
777
Jason Wang54f968d2012-10-31 19:45:57 +0000778 if (tfile->socket.sk->sk_filter &&
779 sk_filter(tfile->socket.sk, skb))
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000780 goto drop;
781
Rami Rosen36fe8c092012-11-25 22:07:40 +0000782 /* Limit the number of packets queued by dividing txq length with the
Jason Wangc8d68e62012-10-31 19:46:00 +0000783 * number of queues.
784 */
Dominic Curranfa358642014-01-22 03:03:23 +0000785 if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues
786 >= dev->tx_queue_len)
Michael S. Tsirkin5d097102012-12-03 10:07:14 +0000787 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
Jason Wang7bf66302013-09-05 17:54:00 +0800789 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
790 goto drop;
791
Richard Cochraneda29772013-07-19 19:40:10 +0200792 if (skb->sk) {
793 sock_tx_timestamp(skb->sk, &skb_shinfo(skb)->tx_flags);
794 sw_tx_timestamp(skb);
795 }
796
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000797 /* Orphan the skb - required as we might hang on to it
Jason Wang7bf66302013-09-05 17:54:00 +0800798 * for indefinite time.
799 */
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000800 skb_orphan(skb);
801
Eric Dumazetf8af75f2013-03-06 11:02:37 +0000802 nf_reset(skb);
803
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700804 /* Enqueue packet */
Jason Wang54f968d2012-10-31 19:45:57 +0000805 skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806
807 /* Notify and wake up reader process */
Jason Wang54f968d2012-10-31 19:45:57 +0000808 if (tfile->flags & TUN_FASYNC)
809 kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
Xi Wang9e641bd2014-05-16 15:11:48 -0700810 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Jason Wang6e914fc2012-10-31 19:45:58 +0000811
812 rcu_read_unlock();
Patrick McHardy6ed10652009-06-23 06:03:08 +0000813 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
815drop:
Jeff Garzik09f75cd2007-10-03 17:41:50 -0700816 dev->stats.tx_dropped++;
Michael S. Tsirkin149d36f2012-11-01 09:16:32 +0000817 skb_tx_error(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 kfree_skb(skb);
Jason Wang6e914fc2012-10-31 19:45:58 +0000819 rcu_read_unlock();
Patrick McHardy6ed10652009-06-23 06:03:08 +0000820 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821}
822
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700823static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700825 /*
826 * This callback is supposed to deal with mc filter in
827 * _rx_ path and has nothing to do with the _tx_ path.
828 * In rx path we always accept everything userspace gives us.
829 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830}
831
Ed Swierk4885a502007-09-16 12:21:38 -0700832#define MIN_MTU 68
833#define MAX_MTU 65535
834
835static int
836tun_net_change_mtu(struct net_device *dev, int new_mtu)
837{
838 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
839 return -EINVAL;
840 dev->mtu = new_mtu;
841 return 0;
842}
843
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000844static netdev_features_t tun_net_fix_features(struct net_device *dev,
845 netdev_features_t features)
Michał Mirosław88255372011-04-19 06:13:10 +0000846{
847 struct tun_struct *tun = netdev_priv(dev);
848
849 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
850}
Neil Hormanbebd0972011-06-15 05:25:01 +0000851#ifdef CONFIG_NET_POLL_CONTROLLER
852static void tun_poll_controller(struct net_device *dev)
853{
854 /*
855 * Tun only receives frames when:
856 * 1) the char device endpoint gets data from user space
857 * 2) the tun socket gets a sendmsg call from user space
stephen hemminger92d4ea62013-12-05 20:42:58 -0800858 * Since both of those are synchronous operations, we are guaranteed
Neil Hormanbebd0972011-06-15 05:25:01 +0000859 * never to have pending data when we poll for it
stephen hemminger92d4ea62013-12-05 20:42:58 -0800860 * so there is nothing to do here but return.
Neil Hormanbebd0972011-06-15 05:25:01 +0000861 * We need this though so netpoll recognizes us as an interface that
862 * supports polling, which enables bridge devices in virt setups to
863 * still use netconsole
864 */
865 return;
866}
867#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800868static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000869 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800870 .ndo_open = tun_net_open,
871 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800872 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800873 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000874 .ndo_fix_features = tun_net_fix_features,
Jason Wangc8d68e62012-10-31 19:46:00 +0000875 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000876#ifdef CONFIG_NET_POLL_CONTROLLER
877 .ndo_poll_controller = tun_poll_controller,
878#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800879};
880
881static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000882 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800883 .ndo_open = tun_net_open,
884 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -0800885 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800886 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +0000887 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +0000888 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800889 .ndo_set_mac_address = eth_mac_addr,
890 .ndo_validate_addr = eth_validate_addr,
Jason Wangc8d68e62012-10-31 19:46:00 +0000891 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +0000892#ifdef CONFIG_NET_POLL_CONTROLLER
893 .ndo_poll_controller = tun_poll_controller,
894#endif
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800895};
896
Pavel Emelyanov944a1372013-06-11 17:01:08 +0400897static void tun_flow_init(struct tun_struct *tun)
Jason Wang96442e422012-10-31 19:46:02 +0000898{
899 int i;
900
Jason Wang96442e422012-10-31 19:46:02 +0000901 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
902 INIT_HLIST_HEAD(&tun->flows[i]);
903
904 tun->ageing_time = TUN_FLOW_EXPIRE;
905 setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun);
906 mod_timer(&tun->flow_gc_timer,
907 round_jiffies_up(jiffies + tun->ageing_time));
Jason Wang96442e422012-10-31 19:46:02 +0000908}
909
910static void tun_flow_uninit(struct tun_struct *tun)
911{
912 del_timer_sync(&tun->flow_gc_timer);
913 tun_flow_flush(tun);
Jason Wang96442e422012-10-31 19:46:02 +0000914}
915
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916/* Initialize net device. */
917static void tun_net_init(struct net_device *dev)
918{
919 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400920
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 switch (tun->flags & TUN_TYPE_MASK) {
922 case TUN_TUN_DEV:
Stephen Hemminger758e43b2008-11-19 22:10:37 -0800923 dev->netdev_ops = &tun_netdev_ops;
924
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 /* Point-to-Point TUN Device */
926 dev->hard_header_len = 0;
927 dev->addr_len = 0;
928 dev->mtu = 1500;
929
930 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400931 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
933 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
934 break;
935
936 case TUN_TAP_DEV:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -0800937 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700939 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +0000940 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
stephen hemmingera6768472012-12-10 15:16:00 +0000941 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
Danny Kukawkaf2cedb62012-02-15 06:45:39 +0000943 eth_hw_addr_random(dev);
Brian Braunstein36226a82007-04-26 01:00:55 -0700944
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
946 break;
947 }
948}
949
950/* Character device part */
951
952/* Poll */
Jason Wangc8d68e62012-10-31 19:46:00 +0000953static unsigned int tun_chr_poll(struct file *file, poll_table *wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400954{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +0000955 struct tun_file *tfile = file->private_data;
956 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000957 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800958 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +0000961 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962
Jason Wang54f968d2012-10-31 19:45:57 +0000963 sk = tfile->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +0000964
Joe Perches6b8a66e2011-03-02 07:18:10 +0000965 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
Xi Wang9e641bd2014-05-16 15:11:48 -0700967 poll_wait(file, sk_sleep(sk), wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -0400968
Michael S. Tsirkin89f56d12009-08-30 07:04:42 +0000969 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 mask |= POLLIN | POLLRDNORM;
971
Herbert Xu33dccbb2009-02-05 21:25:32 -0800972 if (sock_writeable(sk) ||
973 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
974 sock_writeable(sk)))
975 mask |= POLLOUT | POLLWRNORM;
976
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000977 if (tun->dev->reg_state != NETREG_REGISTERED)
978 mask = POLLERR;
979
Eric W. Biederman631ab462009-01-20 11:00:40 +0000980 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 return mask;
982}
983
Rusty Russellf42157c2008-08-15 15:15:10 -0700984/* prepad is the amount to reserve at front. len is length after that.
985 * linear is a hint as to how much to copy (usually headers). */
Jason Wang54f968d2012-10-31 19:45:57 +0000986static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +0000987 size_t prepad, size_t len,
988 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -0700989{
Jason Wang54f968d2012-10-31 19:45:57 +0000990 struct sock *sk = tfile->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -0700991 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -0800992 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -0700993
994 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -0700995 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -0800996 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -0700997
Herbert Xu33dccbb2009-02-05 21:25:32 -0800998 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
Eric Dumazet28d64272013-08-08 14:38:47 -0700999 &err, 0);
Rusty Russellf42157c2008-08-15 15:15:10 -07001000 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001001 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -07001002
1003 skb_reserve(skb, prepad);
1004 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001005 skb->data_len = len - linear;
1006 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -07001007
1008 return skb;
1009}
1010
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011/* Get packet from user space buffer */
Jason Wang54f968d2012-10-31 19:45:57 +00001012static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
1013 void *msg_control, const struct iovec *iv,
1014 size_t total_len, size_t count, int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015{
Harvey Harrison09640e62009-02-01 00:45:17 -08001016 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 struct sk_buff *skb;
Jason Wang3dd5c332013-07-10 13:43:27 +08001018 size_t len = total_len, align = NET_SKB_PAD, linear;
Rusty Russellf43798c2008-07-03 03:48:02 -07001019 struct virtio_net_hdr gso = { 0 };
Jason Wang96f8d9e2013-11-13 14:00:39 +08001020 int good_linear;
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001021 int offset = 0;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001022 int copylen;
1023 bool zerocopy = false;
1024 int err;
Eric Dumazet49974422012-12-12 19:22:57 +00001025 u32 rxhash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026
1027 if (!(tun->flags & TUN_NO_PI)) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001028 if (len < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001030 len -= sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001032 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 return -EFAULT;
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001034 offset += sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 }
1036
Rusty Russellf43798c2008-07-03 03:48:02 -07001037 if (tun->flags & TUN_VNET_HDR) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001038 if (len < tun->vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001039 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001040 len -= tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001041
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001042 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
Rusty Russellf43798c2008-07-03 03:48:02 -07001043 return -EFAULT;
1044
Herbert Xu49091222009-06-08 00:20:01 -07001045 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
1046 gso.csum_start + gso.csum_offset + 2 > gso.hdr_len)
1047 gso.hdr_len = gso.csum_start + gso.csum_offset + 2;
1048
Rusty Russellf43798c2008-07-03 03:48:02 -07001049 if (gso.hdr_len > len)
1050 return -EINVAL;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001051 offset += tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001052 }
1053
Rusty Russelle01bf1c2008-04-12 18:49:30 -07001054 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
stephen hemmingera504b862011-06-08 14:33:07 +00001055 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -07001056 if (unlikely(len < ETH_HLEN ||
1057 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -07001058 return -EINVAL;
1059 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001060
Jason Wang96f8d9e2013-11-13 14:00:39 +08001061 good_linear = SKB_MAX_HEAD(align);
1062
Jason Wang88529172013-07-18 10:55:15 +08001063 if (msg_control) {
1064 /* There are 256 bytes to be copied in skb, so there is
1065 * enough room for skb expand head in case it is used.
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001066 * The rest of the buffer is mapped from userspace.
1067 */
Jason Wang88529172013-07-18 10:55:15 +08001068 copylen = gso.hdr_len ? gso.hdr_len : GOODCOPY_LEN;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001069 if (copylen > good_linear)
1070 copylen = good_linear;
Jason Wang3dd5c332013-07-10 13:43:27 +08001071 linear = copylen;
Jason Wang88529172013-07-18 10:55:15 +08001072 if (iov_pages(iv, offset + copylen, count) <= MAX_SKB_FRAGS)
1073 zerocopy = true;
1074 }
1075
1076 if (!zerocopy) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001077 copylen = len;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001078 if (gso.hdr_len > good_linear)
1079 linear = good_linear;
1080 else
1081 linear = gso.hdr_len;
Jason Wang3dd5c332013-07-10 13:43:27 +08001082 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001083
Jason Wang3dd5c332013-07-10 13:43:27 +08001084 skb = tun_alloc_skb(tfile, align, copylen, linear, noblock);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001085 if (IS_ERR(skb)) {
1086 if (PTR_ERR(skb) != -EAGAIN)
1087 tun->dev->stats.rx_dropped++;
1088 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 }
1090
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001091 if (zerocopy)
1092 err = zerocopy_sg_from_iovec(skb, iv, offset, count);
Jason Wang88529172013-07-18 10:55:15 +08001093 else {
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001094 err = skb_copy_datagram_from_iovec(skb, 0, iv, offset, len);
Jason Wang88529172013-07-18 10:55:15 +08001095 if (!err && msg_control) {
1096 struct ubuf_info *uarg = msg_control;
1097 uarg->callback(uarg, false);
1098 }
1099 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001100
1101 if (err) {
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001102 tun->dev->stats.rx_dropped++;
Dave Jones8f227572006-03-11 18:49:13 -08001103 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -08001105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106
Rusty Russellf43798c2008-07-03 03:48:02 -07001107 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
1108 if (!skb_partial_csum_set(skb, gso.csum_start,
1109 gso.csum_offset)) {
1110 tun->dev->stats.rx_frame_errors++;
1111 kfree_skb(skb);
1112 return -EINVAL;
1113 }
Michał Mirosław88255372011-04-19 06:13:10 +00001114 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001115
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 switch (tun->flags & TUN_TYPE_MASK) {
1117 case TUN_TUN_DEV:
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -07001118 if (tun->flags & TUN_NO_PI) {
1119 switch (skb->data[0] & 0xf0) {
1120 case 0x40:
1121 pi.proto = htons(ETH_P_IP);
1122 break;
1123 case 0x60:
1124 pi.proto = htons(ETH_P_IPV6);
1125 break;
1126 default:
1127 tun->dev->stats.rx_dropped++;
1128 kfree_skb(skb);
1129 return -EINVAL;
1130 }
1131 }
1132
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001133 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -07001135 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 break;
1137 case TUN_TAP_DEV:
1138 skb->protocol = eth_type_trans(skb, tun->dev);
1139 break;
Joe Perches6403eab2011-06-03 11:51:20 +00001140 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141
Rusty Russellf43798c2008-07-03 03:48:02 -07001142 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
1143 pr_debug("GSO!\n");
1144 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
1145 case VIRTIO_NET_HDR_GSO_TCPV4:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001146 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
Rusty Russellf43798c2008-07-03 03:48:02 -07001147 break;
1148 case VIRTIO_NET_HDR_GSO_TCPV6:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001149 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
Rusty Russellf43798c2008-07-03 03:48:02 -07001150 break;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001151 case VIRTIO_NET_HDR_GSO_UDP:
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001152 skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001153 break;
Rusty Russellf43798c2008-07-03 03:48:02 -07001154 default:
1155 tun->dev->stats.rx_frame_errors++;
1156 kfree_skb(skb);
1157 return -EINVAL;
1158 }
1159
1160 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001161 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
Rusty Russellf43798c2008-07-03 03:48:02 -07001162
1163 skb_shinfo(skb)->gso_size = gso.gso_size;
1164 if (skb_shinfo(skb)->gso_size == 0) {
1165 tun->dev->stats.rx_frame_errors++;
1166 kfree_skb(skb);
1167 return -EINVAL;
1168 }
1169
1170 /* Header must be checked, and gso_segs computed. */
1171 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
1172 skb_shinfo(skb)->gso_segs = 0;
1173 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001174
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001175 /* copy skb_ubuf_info for callback when skb has no error */
1176 if (zerocopy) {
1177 skb_shinfo(skb)->destructor_arg = msg_control;
1178 skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001179 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001180 }
1181
Eric Dumazet76fe4582012-12-17 04:39:20 +00001182 skb_reset_network_header(skb);
Jason Wang40893fd2013-03-26 23:11:22 +00001183 skb_probe_transport_header(skb, 0);
Jason Wang38502af2013-03-25 20:19:56 +00001184
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001185 rxhash = skb_get_hash(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001187
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001188 tun->dev->stats.rx_packets++;
1189 tun->dev->stats.rx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190
Jason Wang9e857222013-01-28 01:05:19 +00001191 tun_flow_update(tun, rxhash, tfile);
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001192 return total_len;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001193}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
Badari Pulavartyee0b3e62006-09-30 23:28:47 -07001195static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
1196 unsigned long count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001198 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -08001199 struct tun_struct *tun = tun_get(file);
Jason Wang54f968d2012-10-31 19:45:57 +00001200 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001201 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202
1203 if (!tun)
1204 return -EBADFD;
1205
Joe Perches6b8a66e2011-03-02 07:18:10 +00001206 tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
Jason Wang54f968d2012-10-31 19:45:57 +00001208 result = tun_get_user(tun, tfile, NULL, iv, iov_length(iv, count),
1209 count, file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001210
1211 tun_put(tun);
1212 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213}
1214
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +00001216static ssize_t tun_put_user(struct tun_struct *tun,
Jason Wang54f968d2012-10-31 19:45:57 +00001217 struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001218 struct sk_buff *skb,
1219 const struct iovec *iv, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220{
1221 struct tun_pi pi = { 0, skb->protocol };
1222 ssize_t total = 0;
Jason Wange6fd07c2013-12-11 13:08:33 +08001223 int vlan_offset = 0, copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224
1225 if (!(tun->flags & TUN_NO_PI)) {
1226 if ((len -= sizeof(pi)) < 0)
1227 return -EINVAL;
1228
1229 if (len < skb->len) {
1230 /* Packet will be striped */
1231 pi.flags |= TUN_PKT_STRIP;
1232 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001233
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +00001234 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 return -EFAULT;
1236 total += sizeof(pi);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001237 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238
Rusty Russellf43798c2008-07-03 03:48:02 -07001239 if (tun->flags & TUN_VNET_HDR) {
1240 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001241 if ((len -= tun->vnet_hdr_sz) < 0)
Rusty Russellf43798c2008-07-03 03:48:02 -07001242 return -EINVAL;
1243
1244 if (skb_is_gso(skb)) {
1245 struct skb_shared_info *sinfo = skb_shinfo(skb);
1246
1247 /* This is a hint as to how much should be linear. */
1248 gso.hdr_len = skb_headlen(skb);
1249 gso.gso_size = sinfo->gso_size;
1250 if (sinfo->gso_type & SKB_GSO_TCPV4)
1251 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
1252 else if (sinfo->gso_type & SKB_GSO_TCPV6)
1253 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001254 else if (sinfo->gso_type & SKB_GSO_UDP)
1255 gso.gso_type = VIRTIO_NET_HDR_GSO_UDP;
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001256 else {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001257 pr_err("unexpected GSO type: "
Michael S. Tsirkinef3db4a2010-07-21 04:32:45 +00001258 "0x%x, gso_size %d, hdr_len %d\n",
1259 sinfo->gso_type, gso.gso_size,
1260 gso.hdr_len);
1261 print_hex_dump(KERN_ERR, "tun: ",
1262 DUMP_PREFIX_NONE,
1263 16, 1, skb->head,
1264 min((int)gso.hdr_len, 64), true);
1265 WARN_ON_ONCE(1);
1266 return -EINVAL;
1267 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001268 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
1269 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
1270 } else
1271 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
1272
1273 if (skb->ip_summed == CHECKSUM_PARTIAL) {
1274 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
Michał Mirosław55508d62010-12-14 15:24:08 +00001275 gso.csum_start = skb_checksum_start_offset(skb);
Rusty Russellf43798c2008-07-03 03:48:02 -07001276 gso.csum_offset = skb->csum_offset;
Jason Wang10a8d942011-06-10 00:56:17 +00001277 } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
1278 gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
Rusty Russellf43798c2008-07-03 03:48:02 -07001279 } /* else everything is zero */
1280
Michael S. Tsirkin43b39dc2009-04-20 01:25:59 +00001281 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
1282 sizeof(gso))))
Rusty Russellf43798c2008-07-03 03:48:02 -07001283 return -EFAULT;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001284 total += tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001285 }
1286
Jason Wange6fd07c2013-12-11 13:08:33 +08001287 copied = total;
1288 total += skb->len;
Jason Wang6680ec62013-07-25 13:00:33 +08001289 if (!vlan_tx_tag_present(skb)) {
1290 len = min_t(int, skb->len, len);
1291 } else {
1292 int copy, ret;
1293 struct {
1294 __be16 h_vlan_proto;
1295 __be16 h_vlan_TCI;
1296 } veth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
Jason Wang6680ec62013-07-25 13:00:33 +08001298 veth.h_vlan_proto = skb->vlan_proto;
1299 veth.h_vlan_TCI = htons(vlan_tx_tag_get(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
Jason Wang6680ec62013-07-25 13:00:33 +08001301 vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto);
1302 len = min_t(int, skb->len + VLAN_HLEN, len);
Jason Wange6fd07c2013-12-11 13:08:33 +08001303 total += VLAN_HLEN;
Jason Wang6680ec62013-07-25 13:00:33 +08001304
1305 copy = min_t(int, vlan_offset, len);
Jason Wange6fd07c2013-12-11 13:08:33 +08001306 ret = skb_copy_datagram_const_iovec(skb, 0, iv, copied, copy);
Jason Wang6680ec62013-07-25 13:00:33 +08001307 len -= copy;
Jason Wange6fd07c2013-12-11 13:08:33 +08001308 copied += copy;
Jason Wang6680ec62013-07-25 13:00:33 +08001309 if (ret || !len)
1310 goto done;
1311
1312 copy = min_t(int, sizeof(veth), len);
Jason Wange6fd07c2013-12-11 13:08:33 +08001313 ret = memcpy_toiovecend(iv, (void *)&veth, copied, copy);
Jason Wang6680ec62013-07-25 13:00:33 +08001314 len -= copy;
Jason Wange6fd07c2013-12-11 13:08:33 +08001315 copied += copy;
Jason Wang6680ec62013-07-25 13:00:33 +08001316 if (ret || !len)
1317 goto done;
1318 }
1319
Jason Wange6fd07c2013-12-11 13:08:33 +08001320 skb_copy_datagram_const_iovec(skb, vlan_offset, iv, copied, len);
Jason Wang6680ec62013-07-25 13:00:33 +08001321
1322done:
Jeff Garzik09f75cd2007-10-03 17:41:50 -07001323 tun->dev->stats.tx_packets++;
1324 tun->dev->stats.tx_bytes += len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325
1326 return total;
1327}
1328
Jason Wang54f968d2012-10-31 19:45:57 +00001329static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
Zhi Yong Wuf96eb742013-12-07 04:13:06 +08001330 const struct iovec *iv, ssize_t len, int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 struct sk_buff *skb;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001333 ssize_t ret = 0;
Xi Wang9e641bd2014-05-16 15:11:48 -07001334 int peeked, err, off = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
Rami Rosen3872baf2012-11-25 22:07:41 +00001336 tun_debug(KERN_INFO, tun, "tun_do_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
Xi Wang9e641bd2014-05-16 15:11:48 -07001338 if (!len)
1339 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Xi Wang9e641bd2014-05-16 15:11:48 -07001341 if (tun->dev->reg_state != NETREG_REGISTERED)
1342 return -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343
Xi Wang9e641bd2014-05-16 15:11:48 -07001344 /* Read frames from queue */
1345 skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0,
1346 &peeked, &off, &err);
1347 if (skb) {
Jason Wang54f968d2012-10-31 19:45:57 +00001348 ret = tun_put_user(tun, tfile, skb, iv, len);
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001349 kfree_skb(skb);
Xi Wang9e641bd2014-05-16 15:11:48 -07001350 } else
1351 ret = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001353 return ret;
1354}
1355
1356static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
1357 unsigned long count, loff_t pos)
1358{
1359 struct file *file = iocb->ki_filp;
1360 struct tun_file *tfile = file->private_data;
1361 struct tun_struct *tun = __tun_get(tfile);
1362 ssize_t len, ret;
1363
1364 if (!tun)
1365 return -EBADFD;
1366 len = iov_length(iv, count);
1367 if (len < 0) {
1368 ret = -EINVAL;
1369 goto out;
1370 }
1371
Zhi Yong Wuf96eb742013-12-07 04:13:06 +08001372 ret = tun_do_read(tun, tfile, iv, len,
Jason Wang54f968d2012-10-31 19:45:57 +00001373 file->f_flags & O_NONBLOCK);
David S. Miller42404c02013-12-10 22:05:45 -05001374 ret = min_t(ssize_t, ret, len);
Zhi Yong Wud0b7da8a2013-12-06 14:16:51 +08001375 if (ret > 0)
1376 iocb->ki_pos = ret;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001377out:
1378 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 return ret;
1380}
1381
Jason Wang96442e422012-10-31 19:46:02 +00001382static void tun_free_netdev(struct net_device *dev)
1383{
1384 struct tun_struct *tun = netdev_priv(dev);
1385
Jason Wang4008e972012-12-13 23:53:30 +00001386 BUG_ON(!(list_empty(&tun->disabled)));
Jason Wang96442e422012-10-31 19:46:02 +00001387 tun_flow_uninit(tun);
Paul Moore5dbbaf22013-01-14 07:12:19 +00001388 security_tun_dev_free_security(tun->security);
Jason Wang96442e422012-10-31 19:46:02 +00001389 free_netdev(dev);
1390}
1391
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392static void tun_setup(struct net_device *dev)
1393{
1394 struct tun_struct *tun = netdev_priv(dev);
1395
Eric W. Biederman0625c882012-02-07 16:48:55 -08001396 tun->owner = INVALID_UID;
1397 tun->group = INVALID_GID;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 dev->ethtool_ops = &tun_ethtool_ops;
Jason Wang96442e422012-10-31 19:46:02 +00001400 dev->destructor = tun_free_netdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401}
1402
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001403/* Trivial set of netlink ops to allow deleting tun or tap
1404 * device with netlink.
1405 */
1406static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
1407{
1408 return -EINVAL;
1409}
1410
1411static struct rtnl_link_ops tun_link_ops __read_mostly = {
1412 .kind = DRV_NAME,
1413 .priv_size = sizeof(struct tun_struct),
1414 .setup = tun_setup,
1415 .validate = tun_validate,
1416};
1417
Herbert Xu33dccbb2009-02-05 21:25:32 -08001418static void tun_sock_write_space(struct sock *sk)
1419{
Jason Wang54f968d2012-10-31 19:45:57 +00001420 struct tun_file *tfile;
Eric Dumazet43815482010-04-29 11:01:49 +00001421 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001422
1423 if (!sock_writeable(sk))
1424 return;
1425
Herbert Xu33dccbb2009-02-05 21:25:32 -08001426 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
1427 return;
1428
Eric Dumazet43815482010-04-29 11:01:49 +00001429 wqueue = sk_sleep(sk);
1430 if (wqueue && waitqueue_active(wqueue))
1431 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001432 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -07001433
Jason Wang54f968d2012-10-31 19:45:57 +00001434 tfile = container_of(sk, struct tun_file, sk);
1435 kill_fasync(&tfile->fasync, SIGIO, POLL_OUT);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001436}
1437
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001438static int tun_sendmsg(struct kiocb *iocb, struct socket *sock,
1439 struct msghdr *m, size_t total_len)
1440{
Jason Wang54f968d2012-10-31 19:45:57 +00001441 int ret;
1442 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1443 struct tun_struct *tun = __tun_get(tfile);
1444
1445 if (!tun)
1446 return -EBADFD;
Jason Wang54f968d2012-10-31 19:45:57 +00001447 ret = tun_get_user(tun, tfile, m->msg_control, m->msg_iov, total_len,
1448 m->msg_iovlen, m->msg_flags & MSG_DONTWAIT);
1449 tun_put(tun);
1450 return ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001451}
1452
1453static int tun_recvmsg(struct kiocb *iocb, struct socket *sock,
1454 struct msghdr *m, size_t total_len,
1455 int flags)
1456{
Jason Wang54f968d2012-10-31 19:45:57 +00001457 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1458 struct tun_struct *tun = __tun_get(tfile);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001459 int ret;
Jason Wang54f968d2012-10-31 19:45:57 +00001460
1461 if (!tun)
1462 return -EBADFD;
1463
Richard Cochraneda29772013-07-19 19:40:10 +02001464 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) {
Gao feng3811ae72013-04-24 21:59:23 +00001465 ret = -EINVAL;
1466 goto out;
1467 }
Richard Cochraneda29772013-07-19 19:40:10 +02001468 if (flags & MSG_ERRQUEUE) {
1469 ret = sock_recv_errqueue(sock->sk, m, total_len,
1470 SOL_PACKET, TUN_TX_TIMESTAMP);
1471 goto out;
1472 }
Zhi Yong Wuf96eb742013-12-07 04:13:06 +08001473 ret = tun_do_read(tun, tfile, m->msg_iov, total_len,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001474 flags & MSG_DONTWAIT);
David S. Miller42404c02013-12-10 22:05:45 -05001475 if (ret > total_len) {
1476 m->msg_flags |= MSG_TRUNC;
1477 ret = flags & MSG_TRUNC ? ret : total_len;
1478 }
Gao feng3811ae72013-04-24 21:59:23 +00001479out:
Jason Wang54f968d2012-10-31 19:45:57 +00001480 tun_put(tun);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001481 return ret;
1482}
1483
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001484static int tun_release(struct socket *sock)
1485{
1486 if (sock->sk)
1487 sock_put(sock->sk);
1488 return 0;
1489}
1490
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001491/* Ops structure to mimic raw sockets with tun */
1492static const struct proto_ops tun_socket_ops = {
1493 .sendmsg = tun_sendmsg,
1494 .recvmsg = tun_recvmsg,
Stanislav Kinsbursky1ab5ecb2012-03-12 02:59:41 +00001495 .release = tun_release,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001496};
1497
Herbert Xu33dccbb2009-02-05 21:25:32 -08001498static struct proto tun_proto = {
1499 .name = "tun",
1500 .owner = THIS_MODULE,
Jason Wang54f968d2012-10-31 19:45:57 +00001501 .obj_size = sizeof(struct tun_file),
Herbert Xu33dccbb2009-02-05 21:25:32 -08001502};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001503
David Woodhouse980c9e82009-05-09 22:54:21 -07001504static int tun_flags(struct tun_struct *tun)
1505{
1506 int flags = 0;
1507
1508 if (tun->flags & TUN_TUN_DEV)
1509 flags |= IFF_TUN;
1510 else
1511 flags |= IFF_TAP;
1512
1513 if (tun->flags & TUN_NO_PI)
1514 flags |= IFF_NO_PI;
1515
Michael S. Tsirkin5d097102012-12-03 10:07:14 +00001516 /* This flag has no real effect. We track the value for backwards
1517 * compatibility.
1518 */
David Woodhouse980c9e82009-05-09 22:54:21 -07001519 if (tun->flags & TUN_ONE_QUEUE)
1520 flags |= IFF_ONE_QUEUE;
1521
1522 if (tun->flags & TUN_VNET_HDR)
1523 flags |= IFF_VNET_HDR;
1524
Jason Wangc8d68e62012-10-31 19:46:00 +00001525 if (tun->flags & TUN_TAP_MQ)
1526 flags |= IFF_MULTI_QUEUE;
1527
Pavel Emelyanov274038f2013-06-11 14:41:24 +04001528 if (tun->flags & TUN_PERSIST)
1529 flags |= IFF_PERSIST;
1530
David Woodhouse980c9e82009-05-09 22:54:21 -07001531 return flags;
1532}
1533
1534static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1535 char *buf)
1536{
1537 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1538 return sprintf(buf, "0x%x\n", tun_flags(tun));
1539}
1540
1541static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1542 char *buf)
1543{
1544 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001545 return uid_valid(tun->owner)?
1546 sprintf(buf, "%u\n",
1547 from_kuid_munged(current_user_ns(), tun->owner)):
1548 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001549}
1550
1551static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1552 char *buf)
1553{
1554 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001555 return gid_valid(tun->group) ?
1556 sprintf(buf, "%u\n",
1557 from_kgid_munged(current_user_ns(), tun->group)):
1558 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001559}
1560
1561static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1562static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1563static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1564
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001565static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566{
1567 struct tun_struct *tun;
Jason Wang54f968d2012-10-31 19:45:57 +00001568 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 struct net_device *dev;
1570 int err;
1571
Jason Wang7c0c3b12013-01-11 16:59:33 +00001572 if (tfile->detached)
1573 return -EINVAL;
1574
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001575 dev = __dev_get_by_name(net, ifr->ifr_name);
1576 if (dev) {
David Woodhousef85ba782009-04-27 03:23:54 -07001577 if (ifr->ifr_flags & IFF_TUN_EXCL)
1578 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001579 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1580 tun = netdev_priv(dev);
1581 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1582 tun = netdev_priv(dev);
1583 else
1584 return -EINVAL;
1585
Jason Wang8e6d91a2013-05-28 18:32:11 +00001586 if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) !=
1587 !!(tun->flags & TUN_TAP_MQ))
1588 return -EINVAL;
1589
Jason Wangcde8b152012-10-31 19:46:01 +00001590 if (tun_not_capable(tun))
Paul Moore2b980db2009-08-28 18:12:43 -04001591 return -EPERM;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001592 err = security_tun_dev_open(tun->security);
Paul Moore2b980db2009-08-28 18:12:43 -04001593 if (err < 0)
1594 return err;
1595
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001596 err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001597 if (err < 0)
1598 return err;
Jason Wang4008e972012-12-13 23:53:30 +00001599
1600 if (tun->flags & TUN_TAP_MQ &&
Jason Wange8dbad62013-04-22 20:40:39 +00001601 (tun->numqueues + tun->numdisabled > 1)) {
1602 /* One or more queue has already been attached, no need
1603 * to initialize the device again.
1604 */
1605 return 0;
1606 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001607 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 else {
1609 char *name;
1610 unsigned long flags = 0;
Jason Wangedfb6a12013-01-23 03:59:12 +00001611 int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
1612 MAX_TAP_QUEUES : 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
Eric W. Biedermanc260b772012-11-18 21:34:11 +00001614 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001615 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001616 err = security_tun_dev_create();
1617 if (err < 0)
1618 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001619
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 /* Set dev type */
1621 if (ifr->ifr_flags & IFF_TUN) {
1622 /* TUN device */
1623 flags |= TUN_TUN_DEV;
1624 name = "tun%d";
1625 } else if (ifr->ifr_flags & IFF_TAP) {
1626 /* TAP device */
1627 flags |= TUN_TAP_DEV;
1628 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001629 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001630 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001631
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 if (*ifr->ifr_name)
1633 name = ifr->ifr_name;
1634
Jason Wangc8d68e62012-10-31 19:46:00 +00001635 dev = alloc_netdev_mqs(sizeof(struct tun_struct), name,
Tom Gundersenc835a672014-07-14 16:37:24 +02001636 NET_NAME_UNKNOWN, tun_setup, queues,
1637 queues);
Jason Wangedfb6a12013-01-23 03:59:12 +00001638
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 if (!dev)
1640 return -ENOMEM;
1641
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001642 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001643 dev->rtnl_link_ops = &tun_link_ops;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001644 dev->ifindex = tfile->ifindex;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001645
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 tun = netdev_priv(dev);
1647 tun->dev = dev;
1648 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001649 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001650 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651
Jason Wang54f968d2012-10-31 19:45:57 +00001652 tun->filter_attached = false;
1653 tun->sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001654
Jason Wang96442e422012-10-31 19:46:02 +00001655 spin_lock_init(&tun->lock);
1656
Paul Moore5dbbaf22013-01-14 07:12:19 +00001657 err = security_tun_dev_alloc_security(&tun->security);
1658 if (err < 0)
1659 goto err_free_dev;
Paul Moore2b980db2009-08-28 18:12:43 -04001660
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 tun_net_init(dev);
Pavel Emelyanov944a1372013-06-11 17:01:08 +04001662 tun_flow_init(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001663
Michał Mirosław88255372011-04-19 06:13:10 +00001664 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
Jason Wang6680ec62013-07-25 13:00:33 +08001665 TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
1666 NETIF_F_HW_VLAN_STAG_TX;
Michał Mirosław88255372011-04-19 06:13:10 +00001667 dev->features = dev->hw_features;
Fernando Luis Vazquez Cao6671b222014-02-18 21:20:09 +09001668 dev->vlan_features = dev->features &
1669 ~(NETIF_F_HW_VLAN_CTAG_TX |
1670 NETIF_F_HW_VLAN_STAG_TX);
Michał Mirosław88255372011-04-19 06:13:10 +00001671
Jason Wang4008e972012-12-13 23:53:30 +00001672 INIT_LIST_HEAD(&tun->disabled);
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001673 err = tun_attach(tun, file, false);
Jason Wangeb0fb362012-12-02 17:19:45 +00001674 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001675 goto err_free_flow;
Jason Wangeb0fb362012-12-02 17:19:45 +00001676
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 err = register_netdevice(tun->dev);
1678 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001679 goto err_detach;
Herbert Xu9c3fea62009-04-18 14:15:52 +00001680
David Woodhouse980c9e82009-05-09 22:54:21 -07001681 if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) ||
1682 device_create_file(&tun->dev->dev, &dev_attr_owner) ||
1683 device_create_file(&tun->dev->dev, &dev_attr_group))
Joe Perches6b8a66e2011-03-02 07:18:10 +00001684 pr_err("Failed to create tun sysfs files\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 }
1686
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +00001687 netif_carrier_on(tun->dev);
1688
Joe Perches6b8a66e2011-03-02 07:18:10 +00001689 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690
1691 if (ifr->ifr_flags & IFF_NO_PI)
1692 tun->flags |= TUN_NO_PI;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001693 else
1694 tun->flags &= ~TUN_NO_PI;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695
Michael S. Tsirkin5d097102012-12-03 10:07:14 +00001696 /* This flag has no real effect. We track the value for backwards
1697 * compatibility.
1698 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 if (ifr->ifr_flags & IFF_ONE_QUEUE)
1700 tun->flags |= TUN_ONE_QUEUE;
Nathaniel Filardoa26af1e2008-02-05 03:05:07 -08001701 else
1702 tun->flags &= ~TUN_ONE_QUEUE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703
Rusty Russellf43798c2008-07-03 03:48:02 -07001704 if (ifr->ifr_flags & IFF_VNET_HDR)
1705 tun->flags |= TUN_VNET_HDR;
1706 else
1707 tun->flags &= ~TUN_VNET_HDR;
1708
Jason Wangc8d68e62012-10-31 19:46:00 +00001709 if (ifr->ifr_flags & IFF_MULTI_QUEUE)
1710 tun->flags |= TUN_TAP_MQ;
1711 else
1712 tun->flags &= ~TUN_TAP_MQ;
1713
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001714 /* Make sure persistent devices do not get stuck in
1715 * xoff state.
1716 */
1717 if (netif_running(tun->dev))
Jason Wangc8d68e62012-10-31 19:46:00 +00001718 netif_tx_wake_all_queues(tun->dev);
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001719
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 strcpy(ifr->ifr_name, tun->dev->name);
1721 return 0;
1722
Jason Wang662ca432013-09-11 18:09:48 +08001723err_detach:
1724 tun_detach_all(dev);
1725err_free_flow:
1726 tun_flow_uninit(tun);
1727 security_tun_dev_free_security(tun->security);
1728err_free_dev:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 return err;
1731}
1732
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001733static void tun_get_iff(struct net *net, struct tun_struct *tun,
Herbert Xu876bfd42009-08-06 14:22:44 +00001734 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001735{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001736 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001737
1738 strcpy(ifr->ifr_name, tun->dev->name);
1739
David Woodhouse980c9e82009-05-09 22:54:21 -07001740 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001741
Mark McLoughline3b99552008-08-15 15:09:56 -07001742}
1743
Rusty Russell5228ddc2008-07-03 03:46:16 -07001744/* This is like a cut-down ethtool ops, except done via tun fd so no
1745 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001746static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001747{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001748 netdev_features_t features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001749
1750 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001751 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001752 arg &= ~TUN_F_CSUM;
1753
1754 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1755 if (arg & TUN_F_TSO_ECN) {
1756 features |= NETIF_F_TSO_ECN;
1757 arg &= ~TUN_F_TSO_ECN;
1758 }
1759 if (arg & TUN_F_TSO4)
1760 features |= NETIF_F_TSO;
1761 if (arg & TUN_F_TSO6)
1762 features |= NETIF_F_TSO6;
1763 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1764 }
Sridhar Samudralae36aa252009-07-14 14:21:04 +00001765
1766 if (arg & TUN_F_UFO) {
1767 features |= NETIF_F_UFO;
1768 arg &= ~TUN_F_UFO;
1769 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001770 }
1771
1772 /* This gives the user a way to test for new features in future by
1773 * trying to set them. */
1774 if (arg)
1775 return -EINVAL;
1776
Michał Mirosław88255372011-04-19 06:13:10 +00001777 tun->set_features = features;
1778 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001779
1780 return 0;
1781}
1782
Jason Wangc8d68e62012-10-31 19:46:00 +00001783static void tun_detach_filter(struct tun_struct *tun, int n)
1784{
1785 int i;
1786 struct tun_file *tfile;
1787
1788 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001789 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001790 sk_detach_filter(tfile->socket.sk);
1791 }
1792
1793 tun->filter_attached = false;
1794}
1795
1796static int tun_attach_filter(struct tun_struct *tun)
1797{
1798 int i, ret = 0;
1799 struct tun_file *tfile;
1800
1801 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001802 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001803 ret = sk_attach_filter(&tun->fprog, tfile->socket.sk);
1804 if (ret) {
1805 tun_detach_filter(tun, i);
1806 return ret;
1807 }
1808 }
1809
1810 tun->filter_attached = true;
1811 return ret;
1812}
1813
1814static void tun_set_sndbuf(struct tun_struct *tun)
1815{
1816 struct tun_file *tfile;
1817 int i;
1818
1819 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001820 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001821 tfile->socket.sk->sk_sndbuf = tun->sndbuf;
1822 }
1823}
1824
Jason Wangcde8b152012-10-31 19:46:01 +00001825static int tun_set_queue(struct file *file, struct ifreq *ifr)
1826{
1827 struct tun_file *tfile = file->private_data;
1828 struct tun_struct *tun;
Jason Wangcde8b152012-10-31 19:46:01 +00001829 int ret = 0;
1830
1831 rtnl_lock();
1832
1833 if (ifr->ifr_flags & IFF_ATTACH_QUEUE) {
Jason Wang4008e972012-12-13 23:53:30 +00001834 tun = tfile->detached;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001835 if (!tun) {
Jason Wangcde8b152012-10-31 19:46:01 +00001836 ret = -EINVAL;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001837 goto unlock;
1838 }
1839 ret = security_tun_dev_attach_queue(tun->security);
1840 if (ret < 0)
1841 goto unlock;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001842 ret = tun_attach(tun, file, false);
Jason Wang4008e972012-12-13 23:53:30 +00001843 } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001844 tun = rtnl_dereference(tfile->tun);
Jason Wang9e857222013-01-28 01:05:19 +00001845 if (!tun || !(tun->flags & TUN_TAP_MQ) || tfile->detached)
Jason Wang4008e972012-12-13 23:53:30 +00001846 ret = -EINVAL;
1847 else
1848 __tun_detach(tfile, false);
1849 } else
Jason Wangcde8b152012-10-31 19:46:01 +00001850 ret = -EINVAL;
1851
Paul Moore5dbbaf22013-01-14 07:12:19 +00001852unlock:
Jason Wangcde8b152012-10-31 19:46:01 +00001853 rtnl_unlock();
1854 return ret;
1855}
1856
Arnd Bergmann50857e22009-11-06 22:52:32 -08001857static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1858 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001860 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001861 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 void __user* argp = (void __user*)arg;
1863 struct ifreq ifr;
Eric W. Biederman0625c882012-02-07 16:48:55 -08001864 kuid_t owner;
1865 kgid_t group;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001866 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001867 int vnet_hdr_sz;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001868 unsigned int ifindex;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001869 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870
Jason Wangcde8b152012-10-31 19:46:01 +00001871 if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) {
Arnd Bergmann50857e22009-11-06 22:52:32 -08001872 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 return -EFAULT;
David S. Miller8bbb1812012-07-30 14:52:48 -07001874 } else {
Mathias Krausea117dac2012-07-29 19:45:14 +00001875 memset(&ifr, 0, sizeof(ifr));
David S. Miller8bbb1812012-07-30 14:52:48 -07001876 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001877 if (cmd == TUNGETFEATURES) {
1878 /* Currently this just means: "what IFF flags are valid?".
1879 * This is needed because we never checked for invalid flags on
1880 * TUNSETIFF. */
1881 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
Jason Wangcde8b152012-10-31 19:46:01 +00001882 IFF_VNET_HDR | IFF_MULTI_QUEUE,
Eric W. Biederman631ab462009-01-20 11:00:40 +00001883 (unsigned int __user*)argp);
Jason Wangcde8b152012-10-31 19:46:01 +00001884 } else if (cmd == TUNSETQUEUE)
1885 return tun_set_queue(file, &ifr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001886
Jason Wangc8d68e62012-10-31 19:46:00 +00001887 ret = 0;
Herbert Xu876bfd42009-08-06 14:22:44 +00001888 rtnl_lock();
1889
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001890 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1893
Herbert Xu876bfd42009-08-06 14:22:44 +00001894 ret = tun_set_iff(tfile->net, file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895
Herbert Xu876bfd42009-08-06 14:22:44 +00001896 if (ret)
1897 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898
Arnd Bergmann50857e22009-11-06 22:52:32 -08001899 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00001900 ret = -EFAULT;
1901 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902 }
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001903 if (cmd == TUNSETIFINDEX) {
1904 ret = -EPERM;
1905 if (tun)
1906 goto unlock;
1907
1908 ret = -EFAULT;
1909 if (copy_from_user(&ifindex, argp, sizeof(ifindex)))
1910 goto unlock;
1911
1912 ret = 0;
1913 tfile->ifindex = ifindex;
1914 goto unlock;
1915 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916
Herbert Xu876bfd42009-08-06 14:22:44 +00001917 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00001919 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920
Jason Wang1e588332012-10-31 19:45:56 +00001921 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922
Eric W. Biederman631ab462009-01-20 11:00:40 +00001923 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07001925 case TUNGETIFF:
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001926 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07001927
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001928 if (tfile->detached)
1929 ifr.ifr_flags |= IFF_DETACH_QUEUE;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001930 if (!tfile->socket.sk->sk_filter)
1931 ifr.ifr_flags |= IFF_NOFILTER;
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04001932
Arnd Bergmann50857e22009-11-06 22:52:32 -08001933 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00001934 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07001935 break;
1936
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 case TUNSETNOCSUM:
1938 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939
Michał Mirosław88255372011-04-19 06:13:10 +00001940 /* [unimplemented] */
1941 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00001942 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 break;
1944
1945 case TUNSETPERSIST:
Jason Wang54f968d2012-10-31 19:45:57 +00001946 /* Disable/Enable persist mode. Keep an extra reference to the
1947 * module to prevent the module being unprobed.
1948 */
Jason Wangdd38bd82013-01-11 16:59:34 +00001949 if (arg && !(tun->flags & TUN_PERSIST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 tun->flags |= TUN_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001951 __module_get(THIS_MODULE);
Jason Wangdd38bd82013-01-11 16:59:34 +00001952 }
1953 if (!arg && (tun->flags & TUN_PERSIST)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 tun->flags &= ~TUN_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00001955 module_put(THIS_MODULE);
1956 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
Joe Perches6b8a66e2011-03-02 07:18:10 +00001958 tun_debug(KERN_INFO, tun, "persist %s\n",
1959 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 break;
1961
1962 case TUNSETOWNER:
1963 /* Set owner of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001964 owner = make_kuid(current_user_ns(), arg);
1965 if (!uid_valid(owner)) {
1966 ret = -EINVAL;
1967 break;
1968 }
1969 tun->owner = owner;
Jason Wang1e588332012-10-31 19:45:56 +00001970 tun_debug(KERN_INFO, tun, "owner set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08001971 from_kuid(&init_user_ns, tun->owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 break;
1973
Guido Guenther8c644622007-07-02 22:50:25 -07001974 case TUNSETGROUP:
1975 /* Set group of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08001976 group = make_kgid(current_user_ns(), arg);
1977 if (!gid_valid(group)) {
1978 ret = -EINVAL;
1979 break;
1980 }
1981 tun->group = group;
Jason Wang1e588332012-10-31 19:45:56 +00001982 tun_debug(KERN_INFO, tun, "group set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08001983 from_kgid(&init_user_ns, tun->group));
Guido Guenther8c644622007-07-02 22:50:25 -07001984 break;
1985
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001986 case TUNSETLINK:
1987 /* Only allow setting the type when the interface is down */
1988 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00001989 tun_debug(KERN_INFO, tun,
1990 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07001991 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001992 } else {
1993 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00001994 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
1995 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07001996 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001997 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00001998 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07001999
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000#ifdef TUN_DEBUG
2001 case TUNSETDEBUG:
2002 tun->debug = arg;
2003 break;
2004#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07002005 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00002006 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002007 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07002008
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002009 case TUNSETTXFILTER:
2010 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00002011 ret = -EINVAL;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002012 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
Eric W. Biederman631ab462009-01-20 11:00:40 +00002013 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07002014 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002015 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
2017 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04002018 /* Get hw address */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002019 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
2020 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08002021 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00002022 ret = -EFAULT;
2023 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024
2025 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002026 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00002027 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
2028 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08002029
Kim B. Heino40102372008-02-29 12:26:21 -08002030 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002031 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002032
2033 case TUNGETSNDBUF:
Jason Wang54f968d2012-10-31 19:45:57 +00002034 sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002035 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
2036 ret = -EFAULT;
2037 break;
2038
2039 case TUNSETSNDBUF:
2040 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
2041 ret = -EFAULT;
2042 break;
2043 }
2044
Jason Wangc8d68e62012-10-31 19:46:00 +00002045 tun->sndbuf = sndbuf;
2046 tun_set_sndbuf(tun);
Herbert Xu33dccbb2009-02-05 21:25:32 -08002047 break;
2048
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02002049 case TUNGETVNETHDRSZ:
2050 vnet_hdr_sz = tun->vnet_hdr_sz;
2051 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
2052 ret = -EFAULT;
2053 break;
2054
2055 case TUNSETVNETHDRSZ:
2056 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
2057 ret = -EFAULT;
2058 break;
2059 }
2060 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
2061 ret = -EINVAL;
2062 break;
2063 }
2064
2065 tun->vnet_hdr_sz = vnet_hdr_sz;
2066 break;
2067
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002068 case TUNATTACHFILTER:
2069 /* Can be set only for TAPs */
2070 ret = -EINVAL;
2071 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
2072 break;
2073 ret = -EFAULT;
Jason Wang54f968d2012-10-31 19:45:57 +00002074 if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog)))
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002075 break;
2076
Jason Wangc8d68e62012-10-31 19:46:00 +00002077 ret = tun_attach_filter(tun);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002078 break;
2079
2080 case TUNDETACHFILTER:
2081 /* Can be set only for TAPs */
2082 ret = -EINVAL;
2083 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
2084 break;
Jason Wangc8d68e62012-10-31 19:46:00 +00002085 ret = 0;
2086 tun_detach_filter(tun, tun->numqueues);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002087 break;
2088
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002089 case TUNGETFILTER:
2090 ret = -EINVAL;
2091 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
2092 break;
2093 ret = -EFAULT;
2094 if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog)))
2095 break;
2096 ret = 0;
2097 break;
2098
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00002100 ret = -EINVAL;
2101 break;
Joe Perchesee289b62010-05-17 22:47:34 -07002102 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103
Herbert Xu876bfd42009-08-06 14:22:44 +00002104unlock:
2105 rtnl_unlock();
2106 if (tun)
2107 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002108 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109}
2110
Arnd Bergmann50857e22009-11-06 22:52:32 -08002111static long tun_chr_ioctl(struct file *file,
2112 unsigned int cmd, unsigned long arg)
2113{
2114 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
2115}
2116
2117#ifdef CONFIG_COMPAT
2118static long tun_chr_compat_ioctl(struct file *file,
2119 unsigned int cmd, unsigned long arg)
2120{
2121 switch (cmd) {
2122 case TUNSETIFF:
2123 case TUNGETIFF:
2124 case TUNSETTXFILTER:
2125 case TUNGETSNDBUF:
2126 case TUNSETSNDBUF:
2127 case SIOCGIFHWADDR:
2128 case SIOCSIFHWADDR:
2129 arg = (unsigned long)compat_ptr(arg);
2130 break;
2131 default:
2132 arg = (compat_ulong_t)arg;
2133 break;
2134 }
2135
2136 /*
2137 * compat_ifreq is shorter than ifreq, so we must not access beyond
2138 * the end of that structure. All fields that are used in this
2139 * driver are compatible though, we don't need to convert the
2140 * contents.
2141 */
2142 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
2143}
2144#endif /* CONFIG_COMPAT */
2145
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146static int tun_chr_fasync(int fd, struct file *file, int on)
2147{
Jason Wang54f968d2012-10-31 19:45:57 +00002148 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 int ret;
2150
Jason Wang54f968d2012-10-31 19:45:57 +00002151 if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06002152 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002153
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 if (on) {
Jeff Laytone0b93ed2014-08-22 11:27:32 -04002155 __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Jason Wang54f968d2012-10-31 19:45:57 +00002156 tfile->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002157 } else
Jason Wang54f968d2012-10-31 19:45:57 +00002158 tfile->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06002159 ret = 0;
2160out:
Jonathan Corbet9d319522008-06-19 15:50:37 -06002161 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162}
2163
2164static int tun_chr_open(struct inode *inode, struct file * file)
2165{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002166 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07002167
Joe Perches6b8a66e2011-03-02 07:18:10 +00002168 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00002169
Jason Wang54f968d2012-10-31 19:45:57 +00002170 tfile = (struct tun_file *)sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL,
2171 &tun_proto);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002172 if (!tfile)
2173 return -ENOMEM;
Monam Agarwalc9566742014-03-24 00:02:32 +05302174 RCU_INIT_POINTER(tfile->tun, NULL);
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00002175 tfile->net = get_net(current->nsproxy->net_ns);
Jason Wang54f968d2012-10-31 19:45:57 +00002176 tfile->flags = 0;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002177 tfile->ifindex = 0;
Jason Wang54f968d2012-10-31 19:45:57 +00002178
Jason Wang54f968d2012-10-31 19:45:57 +00002179 init_waitqueue_head(&tfile->wq.wait);
Xi Wang9e641bd2014-05-16 15:11:48 -07002180 RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq);
Jason Wang54f968d2012-10-31 19:45:57 +00002181
2182 tfile->socket.file = file;
2183 tfile->socket.ops = &tun_socket_ops;
2184
2185 sock_init_data(&tfile->socket, &tfile->sk);
2186 sk_change_net(&tfile->sk, tfile->net);
2187
2188 tfile->sk.sk_write_space = tun_sock_write_space;
2189 tfile->sk.sk_sndbuf = INT_MAX;
2190
Eric W. Biederman631ab462009-01-20 11:00:40 +00002191 file->private_data = tfile;
Jason Wang54f968d2012-10-31 19:45:57 +00002192 set_bit(SOCK_EXTERNALLY_ALLOCATED, &tfile->socket.flags);
Jason Wang4008e972012-12-13 23:53:30 +00002193 INIT_LIST_HEAD(&tfile->next);
Jason Wang54f968d2012-10-31 19:45:57 +00002194
Jason Wang19a6afb2013-06-08 14:17:41 +08002195 sock_set_flag(&tfile->sk, SOCK_ZEROCOPY);
2196
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197 return 0;
2198}
2199
2200static int tun_chr_close(struct inode *inode, struct file *file)
2201{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002202 struct tun_file *tfile = file->private_data;
Jason Wang54f968d2012-10-31 19:45:57 +00002203 struct net *net = tfile->net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204
Jason Wangc8d68e62012-10-31 19:46:00 +00002205 tun_detach(tfile, true);
Jason Wang54f968d2012-10-31 19:45:57 +00002206 put_net(net);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207
2208 return 0;
2209}
2210
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002211#ifdef CONFIG_PROC_FS
2212static int tun_chr_show_fdinfo(struct seq_file *m, struct file *f)
2213{
2214 struct tun_struct *tun;
2215 struct ifreq ifr;
2216
2217 memset(&ifr, 0, sizeof(ifr));
2218
2219 rtnl_lock();
2220 tun = tun_get(f);
2221 if (tun)
2222 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
2223 rtnl_unlock();
2224
2225 if (tun)
2226 tun_put(tun);
2227
2228 return seq_printf(m, "iff:\t%s\n", ifr.ifr_name);
2229}
2230#endif
2231
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08002232static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002233 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 .llseek = no_llseek,
Badari Pulavartyee0b3e62006-09-30 23:28:47 -07002235 .read = do_sync_read,
2236 .aio_read = tun_chr_aio_read,
2237 .write = do_sync_write,
2238 .aio_write = tun_chr_aio_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08002240 .unlocked_ioctl = tun_chr_ioctl,
2241#ifdef CONFIG_COMPAT
2242 .compat_ioctl = tun_chr_compat_ioctl,
2243#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244 .open = tun_chr_open,
2245 .release = tun_chr_close,
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002246 .fasync = tun_chr_fasync,
2247#ifdef CONFIG_PROC_FS
2248 .show_fdinfo = tun_chr_show_fdinfo,
2249#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250};
2251
2252static struct miscdevice tun_miscdev = {
2253 .minor = TUN_MINOR,
2254 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02002255 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257};
2258
2259/* ethtool interface */
2260
2261static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
2262{
2263 cmd->supported = 0;
2264 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00002265 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266 cmd->duplex = DUPLEX_FULL;
2267 cmd->port = PORT_TP;
2268 cmd->phy_address = 0;
2269 cmd->transceiver = XCVR_INTERNAL;
2270 cmd->autoneg = AUTONEG_DISABLE;
2271 cmd->maxtxpkt = 0;
2272 cmd->maxrxpkt = 0;
2273 return 0;
2274}
2275
2276static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
2277{
2278 struct tun_struct *tun = netdev_priv(dev);
2279
Rick Jones33a5ba12011-11-15 14:59:53 +00002280 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
2281 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282
2283 switch (tun->flags & TUN_TYPE_MASK) {
2284 case TUN_TUN_DEV:
Rick Jones33a5ba12011-11-15 14:59:53 +00002285 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 break;
2287 case TUN_TAP_DEV:
Rick Jones33a5ba12011-11-15 14:59:53 +00002288 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 break;
2290 }
2291}
2292
2293static u32 tun_get_msglevel(struct net_device *dev)
2294{
2295#ifdef TUN_DEBUG
2296 struct tun_struct *tun = netdev_priv(dev);
2297 return tun->debug;
2298#else
2299 return -EOPNOTSUPP;
2300#endif
2301}
2302
2303static void tun_set_msglevel(struct net_device *dev, u32 value)
2304{
2305#ifdef TUN_DEBUG
2306 struct tun_struct *tun = netdev_priv(dev);
2307 tun->debug = value;
2308#endif
2309}
2310
Jeff Garzik7282d492006-09-13 14:30:00 -04002311static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002312 .get_settings = tun_get_settings,
2313 .get_drvinfo = tun_get_drvinfo,
2314 .get_msglevel = tun_get_msglevel,
2315 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00002316 .get_link = ethtool_op_get_link,
Richard Cochraneda29772013-07-19 19:40:10 +02002317 .get_ts_info = ethtool_op_get_ts_info,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318};
2319
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002320
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321static int __init tun_init(void)
2322{
2323 int ret = 0;
2324
Joe Perches6b8a66e2011-03-02 07:18:10 +00002325 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
2326 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002328 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002329 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002330 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002331 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002332 }
2333
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002335 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002336 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002337 goto err_misc;
2338 }
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002339 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002340err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002341 rtnl_link_unregister(&tun_link_ops);
2342err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 return ret;
2344}
2345
2346static void tun_cleanup(void)
2347{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002348 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002349 rtnl_link_unregister(&tun_link_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350}
2351
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002352/* Get an underlying socket object from tun file. Returns error unless file is
2353 * attached to a device. The returned object works like a packet socket, it
2354 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
2355 * holding a reference to the file for as long as the socket is in use. */
2356struct socket *tun_get_socket(struct file *file)
2357{
Jason Wang6e914fc2012-10-31 19:45:58 +00002358 struct tun_file *tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002359 if (file->f_op != &tun_fops)
2360 return ERR_PTR(-EINVAL);
Jason Wang6e914fc2012-10-31 19:45:58 +00002361 tfile = file->private_data;
2362 if (!tfile)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002363 return ERR_PTR(-EBADFD);
Jason Wang54f968d2012-10-31 19:45:57 +00002364 return &tfile->socket;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002365}
2366EXPORT_SYMBOL_GPL(tun_get_socket);
2367
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368module_init(tun_init);
2369module_exit(tun_cleanup);
2370MODULE_DESCRIPTION(DRV_DESCRIPTION);
2371MODULE_AUTHOR(DRV_COPYRIGHT);
2372MODULE_LICENSE("GPL");
2373MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02002374MODULE_ALIAS("devname:net/tun");