blob: 9c8b5bc2b9d8e851c53cc4d311c245654ebe75e2 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
16 */
17
18/*
19 * Changes:
20 *
Mike Kershawff4cc3a2005-09-01 17:40:05 -070021 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
23 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070024 * Mark Smith <markzzzsmith@yahoo.com.au>
Joe Perches344dc8e2012-07-12 19:33:09 +000025 * Use eth_random_addr() for tap MAC address.
Linus Torvalds1da177e2005-04-16 15:20:36 -070026 *
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
30 * Added ethtool API.
31 * Minor cleanups
32 *
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
35 */
36
Joe Perches6b8a66e2011-03-02 07:18:10 +000037#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
38
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#define DRV_NAME "tun"
40#define DRV_VERSION "1.6"
41#define DRV_DESCRIPTION "Universal TUN/TAP device driver"
42#define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
43
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/module.h>
45#include <linux/errno.h>
46#include <linux/kernel.h>
47#include <linux/major.h>
48#include <linux/slab.h>
49#include <linux/poll.h>
50#include <linux/fcntl.h>
51#include <linux/init.h>
52#include <linux/skbuff.h>
53#include <linux/netdevice.h>
54#include <linux/etherdevice.h>
55#include <linux/miscdevice.h>
56#include <linux/ethtool.h>
57#include <linux/rtnetlink.h>
Arnd Bergmann50857e22009-11-06 22:52:32 -080058#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/if.h>
60#include <linux/if_arp.h>
61#include <linux/if_ether.h>
62#include <linux/if_tun.h>
Jason Wang6680ec62013-07-25 13:00:33 +080063#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064#include <linux/crc32.h>
Pavel Emelyanovd647a592008-04-16 00:41:16 -070065#include <linux/nsproxy.h>
Rusty Russellf43798c2008-07-03 03:48:02 -070066#include <linux/virtio_net.h>
Michael S. Tsirkin99405162010-02-14 01:01:10 +000067#include <linux/rcupdate.h>
Eric W. Biederman881d9662007-09-17 11:56:21 -070068#include <net/net_namespace.h>
Pavel Emelyanov79d17602008-04-16 00:40:46 -070069#include <net/netns/generic.h>
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -080070#include <net/rtnetlink.h>
Herbert Xu33dccbb2009-02-05 21:25:32 -080071#include <net/sock.h>
Masatake YAMATO93e14b62014-01-29 16:43:31 +090072#include <linux/seq_file.h>
Herbert Xue0b46d02014-11-07 21:22:23 +080073#include <linux/uio.h>
Jason Wang1576d982016-06-30 14:45:36 +080074#include <linux/skb_array.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <asm/uaccess.h>
77
Rusty Russell14daa022008-04-12 18:48:58 -070078/* Uncomment to enable debugging */
79/* #define TUN_DEBUG 1 */
80
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#ifdef TUN_DEBUG
82static int debug;
Rusty Russell14daa022008-04-12 18:48:58 -070083
Joe Perches6b8a66e2011-03-02 07:18:10 +000084#define tun_debug(level, tun, fmt, args...) \
85do { \
86 if (tun->debug) \
87 netdev_printk(level, tun->dev, fmt, ##args); \
88} while (0)
89#define DBG1(level, fmt, args...) \
90do { \
91 if (debug == 2) \
92 printk(level fmt, ##args); \
93} while (0)
Rusty Russell14daa022008-04-12 18:48:58 -070094#else
Joe Perches6b8a66e2011-03-02 07:18:10 +000095#define tun_debug(level, tun, fmt, args...) \
96do { \
97 if (0) \
98 netdev_printk(level, tun->dev, fmt, ##args); \
99} while (0)
100#define DBG1(level, fmt, args...) \
101do { \
102 if (0) \
103 printk(level fmt, ##args); \
104} while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#endif
106
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +0200107/* TUN device flags */
108
109/* IFF_ATTACH_QUEUE is never stored in device flags,
110 * overload it to mean fasync when stored there.
111 */
112#define TUN_FASYNC IFF_ATTACH_QUEUE
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200113/* High bits in flags field are unused. */
114#define TUN_VNET_LE 0x80000000
Greg Kurz8b8e6582015-04-24 14:50:36 +0200115#define TUN_VNET_BE 0x40000000
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +0200116
117#define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +0200118 IFF_MULTI_QUEUE)
Michael S. Tsirkin06908992012-07-20 09:23:23 +0000119#define GOODCOPY_LEN 128
120
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700121#define FLT_EXACT_COUNT 8
122struct tap_filter {
123 unsigned int count; /* Number of addrs. Zero means disabled */
124 u32 mask[2]; /* Mask of the hashed addrs */
125 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
126};
127
Pankaj Guptabaf71c52015-01-12 11:41:29 +0530128/* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal
129 * to max number of VCPUs in guest. */
130#define MAX_TAP_QUEUES 256
Jason Wangb8732fb2013-01-23 03:59:13 +0000131#define MAX_TAP_FLOWS 4096
Jason Wangc8d68e62012-10-31 19:46:00 +0000132
Jason Wang96442e422012-10-31 19:46:02 +0000133#define TUN_FLOW_EXPIRE (3 * HZ)
134
Paolo Abeni608b9972016-04-13 10:52:20 +0200135struct tun_pcpu_stats {
136 u64 rx_packets;
137 u64 rx_bytes;
138 u64 tx_packets;
139 u64 tx_bytes;
140 struct u64_stats_sync syncp;
141 u32 rx_dropped;
142 u32 tx_dropped;
143 u32 rx_frame_errors;
144};
145
Jason Wang54f968d2012-10-31 19:45:57 +0000146/* A tun_file connects an open character device to a tuntap netdevice. It
stephen hemminger92d4ea62013-12-05 20:42:58 -0800147 * also contains all socket related structures (except sock_fprog and tap_filter)
Jason Wang54f968d2012-10-31 19:45:57 +0000148 * to serve as one transmit queue for tuntap device. The sock_fprog and
149 * tap_filter were kept in tun_struct since they were used for filtering for the
Rami Rosen36fe8c092012-11-25 22:07:40 +0000150 * netdevice not for a specific queue (at least I didn't see the requirement for
Jason Wang54f968d2012-10-31 19:45:57 +0000151 * this).
Jason Wang6e914fc2012-10-31 19:45:58 +0000152 *
153 * RCU usage:
Rami Rosen36fe8c092012-11-25 22:07:40 +0000154 * The tun_file and tun_struct are loosely coupled, the pointer from one to the
Jason Wang6e914fc2012-10-31 19:45:58 +0000155 * other can only be read while rcu_read_lock or rtnl_lock is held.
Jason Wang54f968d2012-10-31 19:45:57 +0000156 */
Eric W. Biederman631ab462009-01-20 11:00:40 +0000157struct tun_file {
Jason Wang54f968d2012-10-31 19:45:57 +0000158 struct sock sk;
159 struct socket socket;
160 struct socket_wq wq;
Jason Wang6e914fc2012-10-31 19:45:58 +0000161 struct tun_struct __rcu *tun;
Jason Wang54f968d2012-10-31 19:45:57 +0000162 struct fasync_struct *fasync;
163 /* only used for fasnyc */
164 unsigned int flags;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +0400165 union {
166 u16 queue_index;
167 unsigned int ifindex;
168 };
Jason Wang4008e972012-12-13 23:53:30 +0000169 struct list_head next;
170 struct tun_struct *detached;
Jason Wang1576d982016-06-30 14:45:36 +0800171 struct skb_array tx_array;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000172};
173
Jason Wang96442e422012-10-31 19:46:02 +0000174struct tun_flow_entry {
175 struct hlist_node hash_link;
176 struct rcu_head rcu;
177 struct tun_struct *tun;
178
179 u32 rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800180 u32 rps_rxhash;
Jason Wang96442e422012-10-31 19:46:02 +0000181 int queue_index;
182 unsigned long updated;
183};
184
185#define TUN_NUM_FLOW_ENTRIES 1024
186
Jason Wang54f968d2012-10-31 19:45:57 +0000187/* Since the socket were moved to tun_file, to preserve the behavior of persist
Rami Rosen36fe8c092012-11-25 22:07:40 +0000188 * device, socket filter, sndbuf and vnet header size were restore when the
Jason Wang54f968d2012-10-31 19:45:57 +0000189 * file were attached to a persist device.
190 */
Rusty Russell14daa022008-04-12 18:48:58 -0700191struct tun_struct {
Jason Wangc8d68e62012-10-31 19:46:00 +0000192 struct tun_file __rcu *tfiles[MAX_TAP_QUEUES];
193 unsigned int numqueues;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700194 unsigned int flags;
Eric W. Biederman0625c882012-02-07 16:48:55 -0800195 kuid_t owner;
196 kgid_t group;
Rusty Russell14daa022008-04-12 18:48:58 -0700197
Rusty Russell14daa022008-04-12 18:48:58 -0700198 struct net_device *dev;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000199 netdev_features_t set_features;
Michał Mirosław88255372011-04-19 06:13:10 +0000200#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
Vlad Yaseviche3e3c422015-02-03 16:36:17 -0500201 NETIF_F_TSO6|NETIF_F_UFO)
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200202
Paolo Abenieaea34b2016-02-26 10:45:40 +0100203 int align;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +0200204 int vnet_hdr_sz;
Jason Wang54f968d2012-10-31 19:45:57 +0000205 int sndbuf;
206 struct tap_filter txflt;
207 struct sock_fprog fprog;
208 /* protected by rtnl lock */
209 bool filter_attached;
Rusty Russell14daa022008-04-12 18:48:58 -0700210#ifdef TUN_DEBUG
211 int debug;
212#endif
Jason Wang96442e422012-10-31 19:46:02 +0000213 spinlock_t lock;
Jason Wang96442e422012-10-31 19:46:02 +0000214 struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
215 struct timer_list flow_gc_timer;
216 unsigned long ageing_time;
Jason Wang4008e972012-12-13 23:53:30 +0000217 unsigned int numdisabled;
218 struct list_head disabled;
Paul Moore5dbbaf22013-01-14 07:12:19 +0000219 void *security;
Jason Wangb8732fb2013-01-23 03:59:13 +0000220 u32 flow_count;
Paolo Abeni608b9972016-04-13 10:52:20 +0200221 struct tun_pcpu_stats __percpu *pcpu_stats;
Rusty Russell14daa022008-04-12 18:48:58 -0700222};
223
Greg Kurz8b8e6582015-04-24 14:50:36 +0200224#ifdef CONFIG_TUN_VNET_CROSS_LE
225static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
226{
227 return tun->flags & TUN_VNET_BE ? false :
228 virtio_legacy_is_little_endian();
229}
230
231static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
232{
233 int be = !!(tun->flags & TUN_VNET_BE);
234
235 if (put_user(be, argp))
236 return -EFAULT;
237
238 return 0;
239}
240
241static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
242{
243 int be;
244
245 if (get_user(be, argp))
246 return -EFAULT;
247
248 if (be)
249 tun->flags |= TUN_VNET_BE;
250 else
251 tun->flags &= ~TUN_VNET_BE;
252
253 return 0;
254}
255#else
256static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
257{
258 return virtio_legacy_is_little_endian();
259}
260
261static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
262{
263 return -EINVAL;
264}
265
266static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
267{
268 return -EINVAL;
269}
270#endif /* CONFIG_TUN_VNET_CROSS_LE */
271
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200272static inline bool tun_is_little_endian(struct tun_struct *tun)
273{
Greg Kurz7d824102015-04-24 14:26:24 +0200274 return tun->flags & TUN_VNET_LE ||
Greg Kurz8b8e6582015-04-24 14:50:36 +0200275 tun_legacy_is_little_endian(tun);
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200276}
277
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300278static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val)
279{
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200280 return __virtio16_to_cpu(tun_is_little_endian(tun), val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300281}
282
283static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val)
284{
Greg Kurz25bd55bb2015-04-24 14:24:38 +0200285 return __cpu_to_virtio16(tun_is_little_endian(tun), val);
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +0300286}
287
Jason Wang96442e422012-10-31 19:46:02 +0000288static inline u32 tun_hashfn(u32 rxhash)
289{
290 return rxhash & 0x3ff;
291}
292
293static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash)
294{
295 struct tun_flow_entry *e;
Jason Wang96442e422012-10-31 19:46:02 +0000296
Sasha Levinb67bfe02013-02-27 17:06:00 -0800297 hlist_for_each_entry_rcu(e, head, hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000298 if (e->rxhash == rxhash)
299 return e;
300 }
301 return NULL;
302}
303
304static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun,
305 struct hlist_head *head,
306 u32 rxhash, u16 queue_index)
307{
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000308 struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC);
309
Jason Wang96442e422012-10-31 19:46:02 +0000310 if (e) {
311 tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n",
312 rxhash, queue_index);
313 e->updated = jiffies;
314 e->rxhash = rxhash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800315 e->rps_rxhash = 0;
Jason Wang96442e422012-10-31 19:46:02 +0000316 e->queue_index = queue_index;
317 e->tun = tun;
318 hlist_add_head_rcu(&e->hash_link, head);
Jason Wangb8732fb2013-01-23 03:59:13 +0000319 ++tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000320 }
321 return e;
322}
323
Jason Wang96442e422012-10-31 19:46:02 +0000324static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e)
325{
326 tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n",
327 e->rxhash, e->queue_index);
328 hlist_del_rcu(&e->hash_link);
Eric Dumazet9fdc6be2012-12-21 07:17:21 +0000329 kfree_rcu(e, rcu);
Jason Wangb8732fb2013-01-23 03:59:13 +0000330 --tun->flow_count;
Jason Wang96442e422012-10-31 19:46:02 +0000331}
332
333static void tun_flow_flush(struct tun_struct *tun)
334{
335 int i;
336
337 spin_lock_bh(&tun->lock);
338 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
339 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800340 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000341
Sasha Levinb67bfe02013-02-27 17:06:00 -0800342 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
Jason Wang96442e422012-10-31 19:46:02 +0000343 tun_flow_delete(tun, e);
344 }
345 spin_unlock_bh(&tun->lock);
346}
347
348static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
349{
350 int i;
351
352 spin_lock_bh(&tun->lock);
353 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
354 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800355 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000356
Sasha Levinb67bfe02013-02-27 17:06:00 -0800357 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000358 if (e->queue_index == queue_index)
359 tun_flow_delete(tun, e);
360 }
361 }
362 spin_unlock_bh(&tun->lock);
363}
364
365static void tun_flow_cleanup(unsigned long data)
366{
367 struct tun_struct *tun = (struct tun_struct *)data;
368 unsigned long delay = tun->ageing_time;
369 unsigned long next_timer = jiffies + delay;
370 unsigned long count = 0;
371 int i;
372
373 tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n");
374
375 spin_lock_bh(&tun->lock);
376 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
377 struct tun_flow_entry *e;
Sasha Levinb67bfe02013-02-27 17:06:00 -0800378 struct hlist_node *n;
Jason Wang96442e422012-10-31 19:46:02 +0000379
Sasha Levinb67bfe02013-02-27 17:06:00 -0800380 hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
Jason Wang96442e422012-10-31 19:46:02 +0000381 unsigned long this_timer;
382 count++;
383 this_timer = e->updated + delay;
384 if (time_before_eq(this_timer, jiffies))
385 tun_flow_delete(tun, e);
386 else if (time_before(this_timer, next_timer))
387 next_timer = this_timer;
388 }
389 }
390
391 if (count)
392 mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer));
393 spin_unlock_bh(&tun->lock);
394}
395
Eric Dumazet49974422012-12-12 19:22:57 +0000396static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
Jason Wang9e857222013-01-28 01:05:19 +0000397 struct tun_file *tfile)
Jason Wang96442e422012-10-31 19:46:02 +0000398{
399 struct hlist_head *head;
400 struct tun_flow_entry *e;
401 unsigned long delay = tun->ageing_time;
Jason Wang9e857222013-01-28 01:05:19 +0000402 u16 queue_index = tfile->queue_index;
Jason Wang96442e422012-10-31 19:46:02 +0000403
404 if (!rxhash)
405 return;
406 else
407 head = &tun->flows[tun_hashfn(rxhash)];
408
409 rcu_read_lock();
410
Jason Wang9e857222013-01-28 01:05:19 +0000411 /* We may get a very small possibility of OOO during switching, not
412 * worth to optimize.*/
413 if (tun->numqueues == 1 || tfile->detached)
Jason Wang96442e422012-10-31 19:46:02 +0000414 goto unlock;
415
416 e = tun_flow_find(head, rxhash);
417 if (likely(e)) {
418 /* TODO: keep queueing to old queue until it's empty? */
419 e->queue_index = queue_index;
420 e->updated = jiffies;
Tom Herbert9bc88932013-12-22 18:54:32 +0800421 sock_rps_record_flow_hash(e->rps_rxhash);
Jason Wang96442e422012-10-31 19:46:02 +0000422 } else {
423 spin_lock_bh(&tun->lock);
Jason Wangb8732fb2013-01-23 03:59:13 +0000424 if (!tun_flow_find(head, rxhash) &&
425 tun->flow_count < MAX_TAP_FLOWS)
Jason Wang96442e422012-10-31 19:46:02 +0000426 tun_flow_create(tun, head, rxhash, queue_index);
427
428 if (!timer_pending(&tun->flow_gc_timer))
429 mod_timer(&tun->flow_gc_timer,
430 round_jiffies_up(jiffies + delay));
431 spin_unlock_bh(&tun->lock);
432 }
433
434unlock:
435 rcu_read_unlock();
436}
437
Tom Herbert9bc88932013-12-22 18:54:32 +0800438/**
439 * Save the hash received in the stack receive path and update the
440 * flow_hash table accordingly.
441 */
442static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash)
443{
Eric Dumazet567e4b72015-02-06 12:59:01 -0800444 if (unlikely(e->rps_rxhash != hash))
Tom Herbert9bc88932013-12-22 18:54:32 +0800445 e->rps_rxhash = hash;
Tom Herbert9bc88932013-12-22 18:54:32 +0800446}
447
Jason Wangc8d68e62012-10-31 19:46:00 +0000448/* We try to identify a flow through its rxhash first. The reason that
stephen hemminger92d4ea62013-12-05 20:42:58 -0800449 * we do not check rxq no. is because some cards(e.g 82599), chooses
Jason Wangc8d68e62012-10-31 19:46:00 +0000450 * the rxq based on the txq where the last packet of the flow comes. As
451 * the userspace application move between processors, we may get a
452 * different rxq no. here. If we could not get rxhash, then we would
453 * hope the rxq no. may help here.
454 */
Jason Wangf663dd92014-01-10 16:18:26 +0800455static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100456 void *accel_priv, select_queue_fallback_t fallback)
Jason Wangc8d68e62012-10-31 19:46:00 +0000457{
458 struct tun_struct *tun = netdev_priv(dev);
Jason Wang96442e422012-10-31 19:46:02 +0000459 struct tun_flow_entry *e;
Jason Wangc8d68e62012-10-31 19:46:00 +0000460 u32 txq = 0;
461 u32 numqueues = 0;
462
463 rcu_read_lock();
Jason Wang92bb73e2013-06-05 16:44:57 +0800464 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000465
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800466 txq = skb_get_hash(skb);
Jason Wangc8d68e62012-10-31 19:46:00 +0000467 if (txq) {
Jason Wang96442e422012-10-31 19:46:02 +0000468 e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
Tom Herbert9bc88932013-12-22 18:54:32 +0800469 if (e) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800470 tun_flow_save_rps_rxhash(e, txq);
Zhi Yong Wufbe4d452014-01-02 13:24:28 +0800471 txq = e->queue_index;
Tom Herbert9bc88932013-12-22 18:54:32 +0800472 } else
Jason Wang96442e422012-10-31 19:46:02 +0000473 /* use multiply and shift instead of expensive divide */
474 txq = ((u64)txq * numqueues) >> 32;
Jason Wangc8d68e62012-10-31 19:46:00 +0000475 } else if (likely(skb_rx_queue_recorded(skb))) {
476 txq = skb_get_rx_queue(skb);
477 while (unlikely(txq >= numqueues))
478 txq -= numqueues;
479 }
480
481 rcu_read_unlock();
482 return txq;
483}
484
Jason Wangcde8b152012-10-31 19:46:01 +0000485static inline bool tun_not_capable(struct tun_struct *tun)
486{
487 const struct cred *cred = current_cred();
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000488 struct net *net = dev_net(tun->dev);
Jason Wangcde8b152012-10-31 19:46:01 +0000489
490 return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
491 (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
Eric W. Biedermanc260b772012-11-18 21:34:11 +0000492 !ns_capable(net->user_ns, CAP_NET_ADMIN);
Jason Wangcde8b152012-10-31 19:46:01 +0000493}
494
Jason Wangc8d68e62012-10-31 19:46:00 +0000495static void tun_set_real_num_queues(struct tun_struct *tun)
496{
497 netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
498 netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
499}
500
Jason Wang4008e972012-12-13 23:53:30 +0000501static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
502{
503 tfile->detached = tun;
504 list_add_tail(&tfile->next, &tun->disabled);
505 ++tun->numdisabled;
506}
507
Jason Wangd32649d2012-12-18 11:00:27 +0800508static struct tun_struct *tun_enable_queue(struct tun_file *tfile)
Jason Wang4008e972012-12-13 23:53:30 +0000509{
510 struct tun_struct *tun = tfile->detached;
511
512 tfile->detached = NULL;
513 list_del_init(&tfile->next);
514 --tun->numdisabled;
515 return tun;
516}
517
Jason Wang4bfb0512013-09-05 17:53:59 +0800518static void tun_queue_purge(struct tun_file *tfile)
519{
Jason Wang1576d982016-06-30 14:45:36 +0800520 struct sk_buff *skb;
521
522 while ((skb = skb_array_consume(&tfile->tx_array)) != NULL)
523 kfree_skb(skb);
524
Jason Wang4bfb0512013-09-05 17:53:59 +0800525 skb_queue_purge(&tfile->sk.sk_error_queue);
526}
527
Jason Wangc8d68e62012-10-31 19:46:00 +0000528static void __tun_detach(struct tun_file *tfile, bool clean)
529{
530 struct tun_file *ntfile;
531 struct tun_struct *tun;
Jason Wangc8d68e62012-10-31 19:46:00 +0000532
Jason Wangb8deabd2013-01-11 16:59:32 +0000533 tun = rtnl_dereference(tfile->tun);
534
Jason Wang9e857222013-01-28 01:05:19 +0000535 if (tun && !tfile->detached) {
Jason Wangc8d68e62012-10-31 19:46:00 +0000536 u16 index = tfile->queue_index;
537 BUG_ON(index >= tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000538
539 rcu_assign_pointer(tun->tfiles[index],
540 tun->tfiles[tun->numqueues - 1]);
Jason Wangb8deabd2013-01-11 16:59:32 +0000541 ntfile = rtnl_dereference(tun->tfiles[index]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000542 ntfile->queue_index = index;
543
544 --tun->numqueues;
Jason Wang9e857222013-01-28 01:05:19 +0000545 if (clean) {
Monam Agarwalc9566742014-03-24 00:02:32 +0530546 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang4008e972012-12-13 23:53:30 +0000547 sock_put(&tfile->sk);
Jason Wang9e857222013-01-28 01:05:19 +0000548 } else
Jason Wang4008e972012-12-13 23:53:30 +0000549 tun_disable_queue(tun, tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000550
551 synchronize_net();
Jason Wang96442e422012-10-31 19:46:02 +0000552 tun_flow_delete_by_queue(tun, tun->numqueues + 1);
Jason Wangc8d68e62012-10-31 19:46:00 +0000553 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800554 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000555 tun_set_real_num_queues(tun);
Jason Wangdd38bd82013-01-11 16:59:34 +0000556 } else if (tfile->detached && clean) {
Jason Wang4008e972012-12-13 23:53:30 +0000557 tun = tun_enable_queue(tfile);
Jason Wangdd38bd82013-01-11 16:59:34 +0000558 sock_put(&tfile->sk);
559 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000560
561 if (clean) {
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000562 if (tun && tun->numqueues == 0 && tun->numdisabled == 0) {
563 netif_carrier_off(tun->dev);
564
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200565 if (!(tun->flags & IFF_PERSIST) &&
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000566 tun->dev->reg_state == NETREG_REGISTERED)
Jason Wang4008e972012-12-13 23:53:30 +0000567 unregister_netdevice(tun->dev);
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +0000568 }
Jason Wang1576d982016-06-30 14:45:36 +0800569 if (tun)
570 skb_array_cleanup(&tfile->tx_array);
Eric W. Biederman140e807da2015-05-08 21:07:08 -0500571 sock_put(&tfile->sk);
Jason Wangc8d68e62012-10-31 19:46:00 +0000572 }
573}
574
575static void tun_detach(struct tun_file *tfile, bool clean)
576{
577 rtnl_lock();
578 __tun_detach(tfile, clean);
579 rtnl_unlock();
580}
581
582static void tun_detach_all(struct net_device *dev)
583{
584 struct tun_struct *tun = netdev_priv(dev);
Jason Wang4008e972012-12-13 23:53:30 +0000585 struct tun_file *tfile, *tmp;
Jason Wangc8d68e62012-10-31 19:46:00 +0000586 int i, n = tun->numqueues;
587
588 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000589 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000590 BUG_ON(!tfile);
Jason Wangaddf8fc2016-05-19 13:36:51 +0800591 tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN;
Xi Wang9e641bd2014-05-16 15:11:48 -0700592 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530593 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wangc8d68e62012-10-31 19:46:00 +0000594 --tun->numqueues;
595 }
Jason Wang9e857222013-01-28 01:05:19 +0000596 list_for_each_entry(tfile, &tun->disabled, next) {
Jason Wangaddf8fc2016-05-19 13:36:51 +0800597 tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN;
Xi Wang9e641bd2014-05-16 15:11:48 -0700598 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Monam Agarwalc9566742014-03-24 00:02:32 +0530599 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang9e857222013-01-28 01:05:19 +0000600 }
Jason Wangc8d68e62012-10-31 19:46:00 +0000601 BUG_ON(tun->numqueues != 0);
602
603 synchronize_net();
604 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +0000605 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +0000606 /* Drop read queue */
Jason Wang4bfb0512013-09-05 17:53:59 +0800607 tun_queue_purge(tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000608 sock_put(&tfile->sk);
609 }
Jason Wang4008e972012-12-13 23:53:30 +0000610 list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
611 tun_enable_queue(tfile);
Jason Wang4bfb0512013-09-05 17:53:59 +0800612 tun_queue_purge(tfile);
Jason Wang4008e972012-12-13 23:53:30 +0000613 sock_put(&tfile->sk);
614 }
615 BUG_ON(tun->numdisabled != 0);
Jason Wangdd38bd82013-01-11 16:59:34 +0000616
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200617 if (tun->flags & IFF_PERSIST)
Jason Wangdd38bd82013-01-11 16:59:34 +0000618 module_put(THIS_MODULE);
Jason Wangc8d68e62012-10-31 19:46:00 +0000619}
620
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400621static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter)
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000622{
Eric W. Biederman631ab462009-01-20 11:00:40 +0000623 struct tun_file *tfile = file->private_data;
Jason Wang1576d982016-06-30 14:45:36 +0800624 struct net_device *dev = tun->dev;
Eric W. Biederman38231b72009-01-20 11:02:28 +0000625 int err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000626
Paul Moore5dbbaf22013-01-14 07:12:19 +0000627 err = security_tun_dev_attach(tfile->socket.sk, tun->security);
628 if (err < 0)
629 goto out;
630
Eric W. Biederman38231b72009-01-20 11:02:28 +0000631 err = -EINVAL;
Jason Wang9e857222013-01-28 01:05:19 +0000632 if (rtnl_dereference(tfile->tun) && !tfile->detached)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000633 goto out;
634
635 err = -EBUSY;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +0200636 if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1)
Jason Wangc8d68e62012-10-31 19:46:00 +0000637 goto out;
638
639 err = -E2BIG;
Jason Wang4008e972012-12-13 23:53:30 +0000640 if (!tfile->detached &&
641 tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
Eric W. Biederman38231b72009-01-20 11:02:28 +0000642 goto out;
643
644 err = 0;
Jason Wang54f968d2012-10-31 19:45:57 +0000645
stephen hemminger92d4ea62013-12-05 20:42:58 -0800646 /* Re-attach the filter to persist device */
Pavel Emelyanov849c9b62013-08-21 14:32:21 +0400647 if (!skip_filter && (tun->filter_attached == true)) {
Hannes Frederic Sowa8ced4252016-04-05 17:10:16 +0200648 lock_sock(tfile->socket.sk);
649 err = sk_attach_filter(&tun->fprog, tfile->socket.sk);
650 release_sock(tfile->socket.sk);
Jason Wang54f968d2012-10-31 19:45:57 +0000651 if (!err)
652 goto out;
653 }
Jason Wang1576d982016-06-30 14:45:36 +0800654
655 if (!tfile->detached &&
656 skb_array_init(&tfile->tx_array, dev->tx_queue_len, GFP_KERNEL)) {
657 err = -ENOMEM;
658 goto out;
659 }
660
Jason Wangc8d68e62012-10-31 19:46:00 +0000661 tfile->queue_index = tun->numqueues;
Jason Wangaddf8fc2016-05-19 13:36:51 +0800662 tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN;
Jason Wang6e914fc2012-10-31 19:45:58 +0000663 rcu_assign_pointer(tfile->tun, tun);
Jason Wangc8d68e62012-10-31 19:46:00 +0000664 rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
Jason Wangc8d68e62012-10-31 19:46:00 +0000665 tun->numqueues++;
666
Jason Wang4008e972012-12-13 23:53:30 +0000667 if (tfile->detached)
668 tun_enable_queue(tfile);
669 else
670 sock_hold(&tfile->sk);
671
Jason Wangc8d68e62012-10-31 19:46:00 +0000672 tun_set_real_num_queues(tun);
673
Jason Wangc8d68e62012-10-31 19:46:00 +0000674 /* device is allowed to go away first, so no need to hold extra
675 * refcnt.
676 */
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000677
Eric W. Biederman38231b72009-01-20 11:02:28 +0000678out:
Eric W. Biederman38231b72009-01-20 11:02:28 +0000679 return err;
Eric W. Biedermana7385ba2009-01-20 10:57:48 +0000680}
681
Eric W. Biederman631ab462009-01-20 11:00:40 +0000682static struct tun_struct *__tun_get(struct tun_file *tfile)
683{
Jason Wang6e914fc2012-10-31 19:45:58 +0000684 struct tun_struct *tun;
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000685
Jason Wang6e914fc2012-10-31 19:45:58 +0000686 rcu_read_lock();
687 tun = rcu_dereference(tfile->tun);
688 if (tun)
689 dev_hold(tun->dev);
690 rcu_read_unlock();
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000691
692 return tun;
Eric W. Biederman631ab462009-01-20 11:00:40 +0000693}
694
695static struct tun_struct *tun_get(struct file *file)
696{
697 return __tun_get(file->private_data);
698}
699
700static void tun_put(struct tun_struct *tun)
701{
Jason Wang6e914fc2012-10-31 19:45:58 +0000702 dev_put(tun->dev);
Eric W. Biederman631ab462009-01-20 11:00:40 +0000703}
704
Joe Perches6b8a66e2011-03-02 07:18:10 +0000705/* TAP filtering */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700706static void addr_hash_set(u32 *mask, const u8 *addr)
707{
708 int n = ether_crc(ETH_ALEN, addr) >> 26;
709 mask[n >> 5] |= (1 << (n & 31));
710}
711
712static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
713{
714 int n = ether_crc(ETH_ALEN, addr) >> 26;
715 return mask[n >> 5] & (1 << (n & 31));
716}
717
718static int update_filter(struct tap_filter *filter, void __user *arg)
719{
720 struct { u8 u[ETH_ALEN]; } *addr;
721 struct tun_filter uf;
722 int err, alen, n, nexact;
723
724 if (copy_from_user(&uf, arg, sizeof(uf)))
725 return -EFAULT;
726
727 if (!uf.count) {
728 /* Disabled */
729 filter->count = 0;
730 return 0;
731 }
732
733 alen = ETH_ALEN * uf.count;
734 addr = kmalloc(alen, GFP_KERNEL);
735 if (!addr)
736 return -ENOMEM;
737
738 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
739 err = -EFAULT;
740 goto done;
741 }
742
743 /* The filter is updated without holding any locks. Which is
744 * perfectly safe. We disable it first and in the worst
745 * case we'll accept a few undesired packets. */
746 filter->count = 0;
747 wmb();
748
749 /* Use first set of addresses as an exact filter */
750 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
751 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
752
753 nexact = n;
754
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800755 /* Remaining multicast addresses are hashed,
756 * unicast will leave the filter disabled. */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700757 memset(filter->mask, 0, sizeof(filter->mask));
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800758 for (; n < uf.count; n++) {
759 if (!is_multicast_ether_addr(addr[n].u)) {
760 err = 0; /* no filter */
761 goto done;
762 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700763 addr_hash_set(filter->mask, addr[n].u);
Alex Williamsoncfbf84f2009-02-08 17:49:17 -0800764 }
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700765
766 /* For ALLMULTI just set the mask to all ones.
767 * This overrides the mask populated above. */
768 if ((uf.flags & TUN_FLT_ALLMULTI))
769 memset(filter->mask, ~0, sizeof(filter->mask));
770
771 /* Now enable the filter */
772 wmb();
773 filter->count = nexact;
774
775 /* Return the number of exact filters */
776 err = nexact;
777
778done:
779 kfree(addr);
780 return err;
781}
782
783/* Returns: 0 - drop, !=0 - accept */
784static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
785{
786 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
787 * at this point. */
788 struct ethhdr *eh = (struct ethhdr *) skb->data;
789 int i;
790
791 /* Exact match */
792 for (i = 0; i < filter->count; i++)
Joe Perches2e42e472012-05-09 17:17:46 +0000793 if (ether_addr_equal(eh->h_dest, filter->addr[i]))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700794 return 1;
795
796 /* Inexact match (multicast only) */
797 if (is_multicast_ether_addr(eh->h_dest))
798 return addr_hash_test(filter->mask, eh->h_dest);
799
800 return 0;
801}
802
803/*
804 * Checks whether the packet is accepted or not.
805 * Returns: 0 - drop, !=0 - accept
806 */
807static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
808{
809 if (!filter->count)
810 return 1;
811
812 return run_filter(filter, skb);
813}
814
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815/* Network device part of the driver */
816
Jeff Garzik7282d492006-09-13 14:30:00 -0400817static const struct ethtool_ops tun_ethtool_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000819/* Net device detach from fd. */
820static void tun_net_uninit(struct net_device *dev)
821{
Jason Wangc8d68e62012-10-31 19:46:00 +0000822 tun_detach_all(dev);
Eric W. Biedermanc70f1822009-01-20 11:07:17 +0000823}
824
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825/* Net device open. */
826static int tun_net_open(struct net_device *dev)
827{
Jason Wangc8d68e62012-10-31 19:46:00 +0000828 netif_tx_start_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 return 0;
830}
831
832/* Net device close. */
833static int tun_net_close(struct net_device *dev)
834{
Jason Wangc8d68e62012-10-31 19:46:00 +0000835 netif_tx_stop_all_queues(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 return 0;
837}
838
839/* Net device start xmit */
Stephen Hemminger424efe92009-08-31 19:50:51 +0000840static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
842 struct tun_struct *tun = netdev_priv(dev);
Jason Wangc8d68e62012-10-31 19:46:00 +0000843 int txq = skb->queue_mapping;
Jason Wang6e914fc2012-10-31 19:45:58 +0000844 struct tun_file *tfile;
Dominic Curranfa358642014-01-22 03:03:23 +0000845 u32 numqueues = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
Jason Wang6e914fc2012-10-31 19:45:58 +0000847 rcu_read_lock();
Jason Wangc8d68e62012-10-31 19:46:00 +0000848 tfile = rcu_dereference(tun->tfiles[txq]);
Dominic Curranfa358642014-01-22 03:03:23 +0000849 numqueues = ACCESS_ONCE(tun->numqueues);
Jason Wangc8d68e62012-10-31 19:46:00 +0000850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 /* Drop packet if interface is not attached */
Dominic Curranfa358642014-01-22 03:03:23 +0000852 if (txq >= numqueues)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 goto drop;
854
Jason Wang3df97ba2016-04-25 23:13:42 -0400855#ifdef CONFIG_RPS
856 if (numqueues == 1 && static_key_false(&rps_needed)) {
Tom Herbert9bc88932013-12-22 18:54:32 +0800857 /* Select queue was not called for the skbuff, so we extract the
858 * RPS hash and save it into the flow_table here.
859 */
860 __u32 rxhash;
861
862 rxhash = skb_get_hash(skb);
863 if (rxhash) {
864 struct tun_flow_entry *e;
865 e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)],
866 rxhash);
867 if (e)
868 tun_flow_save_rps_rxhash(e, rxhash);
869 }
870 }
Jason Wang3df97ba2016-04-25 23:13:42 -0400871#endif
Tom Herbert9bc88932013-12-22 18:54:32 +0800872
Jason Wang6e914fc2012-10-31 19:45:58 +0000873 tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len);
874
Jason Wangc8d68e62012-10-31 19:46:00 +0000875 BUG_ON(!tfile);
876
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700877 /* Drop if the filter does not like it.
878 * This is a noop if the filter is disabled.
879 * Filter can be enabled only for the TAP devices. */
880 if (!check_filter(&tun->txflt, skb))
881 goto drop;
882
Jason Wang54f968d2012-10-31 19:45:57 +0000883 if (tfile->socket.sk->sk_filter &&
884 sk_filter(tfile->socket.sk, skb))
Michael S. Tsirkin99405162010-02-14 01:01:10 +0000885 goto drop;
886
Rami Rosen36fe8c092012-11-25 22:07:40 +0000887 /* Limit the number of packets queued by dividing txq length with the
Jason Wangc8d68e62012-10-31 19:46:00 +0000888 * number of queues.
889 */
Dominic Curranfa358642014-01-22 03:03:23 +0000890 if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues
891 >= dev->tx_queue_len)
Michael S. Tsirkin5d097102012-12-03 10:07:14 +0000892 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
Jason Wang7bf66302013-09-05 17:54:00 +0800894 if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
895 goto drop;
896
Eric Dumazet5fcd2d82015-10-09 15:42:21 -0700897 if (skb->sk && sk_fullsock(skb->sk)) {
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400898 sock_tx_timestamp(skb->sk, skb->sk->sk_tsflags,
899 &skb_shinfo(skb)->tx_flags);
Richard Cochraneda29772013-07-19 19:40:10 +0200900 sw_tx_timestamp(skb);
901 }
902
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000903 /* Orphan the skb - required as we might hang on to it
Jason Wang7bf66302013-09-05 17:54:00 +0800904 * for indefinite time.
905 */
Michael S. Tsirkin0110d6f2010-04-13 04:59:44 +0000906 skb_orphan(skb);
907
Eric Dumazetf8af75f2013-03-06 11:02:37 +0000908 nf_reset(skb);
909
Jason Wang1576d982016-06-30 14:45:36 +0800910 if (skb_array_produce(&tfile->tx_array, skb))
911 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
913 /* Notify and wake up reader process */
Jason Wang54f968d2012-10-31 19:45:57 +0000914 if (tfile->flags & TUN_FASYNC)
915 kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
Xi Wang9e641bd2014-05-16 15:11:48 -0700916 tfile->socket.sk->sk_data_ready(tfile->socket.sk);
Jason Wang6e914fc2012-10-31 19:45:58 +0000917
918 rcu_read_unlock();
Patrick McHardy6ed10652009-06-23 06:03:08 +0000919 return NETDEV_TX_OK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
921drop:
Paolo Abeni608b9972016-04-13 10:52:20 +0200922 this_cpu_inc(tun->pcpu_stats->tx_dropped);
Michael S. Tsirkin149d36f2012-11-01 09:16:32 +0000923 skb_tx_error(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 kfree_skb(skb);
Jason Wang6e914fc2012-10-31 19:45:58 +0000925 rcu_read_unlock();
Jason Wangbaeabab2014-11-18 13:20:41 +0800926 return NET_XMIT_DROP;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927}
928
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700929static void tun_net_mclist(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930{
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -0700931 /*
932 * This callback is supposed to deal with mc filter in
933 * _rx_ path and has nothing to do with the _tx_ path.
934 * In rx path we always accept everything userspace gives us.
935 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936}
937
Ed Swierk4885a502007-09-16 12:21:38 -0700938#define MIN_MTU 68
939#define MAX_MTU 65535
940
941static int
942tun_net_change_mtu(struct net_device *dev, int new_mtu)
943{
944 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
945 return -EINVAL;
946 dev->mtu = new_mtu;
947 return 0;
948}
949
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000950static netdev_features_t tun_net_fix_features(struct net_device *dev,
951 netdev_features_t features)
Michał Mirosław88255372011-04-19 06:13:10 +0000952{
953 struct tun_struct *tun = netdev_priv(dev);
954
955 return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
956}
Neil Hormanbebd0972011-06-15 05:25:01 +0000957#ifdef CONFIG_NET_POLL_CONTROLLER
958static void tun_poll_controller(struct net_device *dev)
959{
960 /*
961 * Tun only receives frames when:
962 * 1) the char device endpoint gets data from user space
963 * 2) the tun socket gets a sendmsg call from user space
stephen hemminger92d4ea62013-12-05 20:42:58 -0800964 * Since both of those are synchronous operations, we are guaranteed
Neil Hormanbebd0972011-06-15 05:25:01 +0000965 * never to have pending data when we poll for it
stephen hemminger92d4ea62013-12-05 20:42:58 -0800966 * so there is nothing to do here but return.
Neil Hormanbebd0972011-06-15 05:25:01 +0000967 * We need this though so netpoll recognizes us as an interface that
968 * supports polling, which enables bridge devices in virt setups to
969 * still use netconsole
970 */
971 return;
972}
973#endif
Paolo Abenieaea34b2016-02-26 10:45:40 +0100974
975static void tun_set_headroom(struct net_device *dev, int new_hr)
976{
977 struct tun_struct *tun = netdev_priv(dev);
978
979 if (new_hr < NET_SKB_PAD)
980 new_hr = NET_SKB_PAD;
981
982 tun->align = new_hr;
983}
984
Paolo Abeni608b9972016-04-13 10:52:20 +0200985static struct rtnl_link_stats64 *
986tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
987{
988 u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0;
989 struct tun_struct *tun = netdev_priv(dev);
990 struct tun_pcpu_stats *p;
991 int i;
992
993 for_each_possible_cpu(i) {
994 u64 rxpackets, rxbytes, txpackets, txbytes;
995 unsigned int start;
996
997 p = per_cpu_ptr(tun->pcpu_stats, i);
998 do {
999 start = u64_stats_fetch_begin(&p->syncp);
1000 rxpackets = p->rx_packets;
1001 rxbytes = p->rx_bytes;
1002 txpackets = p->tx_packets;
1003 txbytes = p->tx_bytes;
1004 } while (u64_stats_fetch_retry(&p->syncp, start));
1005
1006 stats->rx_packets += rxpackets;
1007 stats->rx_bytes += rxbytes;
1008 stats->tx_packets += txpackets;
1009 stats->tx_bytes += txbytes;
1010
1011 /* u32 counters */
1012 rx_dropped += p->rx_dropped;
1013 rx_frame_errors += p->rx_frame_errors;
1014 tx_dropped += p->tx_dropped;
1015 }
1016 stats->rx_dropped = rx_dropped;
1017 stats->rx_frame_errors = rx_frame_errors;
1018 stats->tx_dropped = tx_dropped;
1019 return stats;
1020}
1021
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001022static const struct net_device_ops tun_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001023 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001024 .ndo_open = tun_net_open,
1025 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -08001026 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001027 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +00001028 .ndo_fix_features = tun_net_fix_features,
Jason Wangc8d68e62012-10-31 19:46:00 +00001029 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +00001030#ifdef CONFIG_NET_POLL_CONTROLLER
1031 .ndo_poll_controller = tun_poll_controller,
1032#endif
Paolo Abenieaea34b2016-02-26 10:45:40 +01001033 .ndo_set_rx_headroom = tun_set_headroom,
Paolo Abeni608b9972016-04-13 10:52:20 +02001034 .ndo_get_stats64 = tun_net_get_stats64,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001035};
1036
1037static const struct net_device_ops tap_netdev_ops = {
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001038 .ndo_uninit = tun_net_uninit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001039 .ndo_open = tun_net_open,
1040 .ndo_stop = tun_net_close,
Stephen Hemminger00829822008-11-20 20:14:53 -08001041 .ndo_start_xmit = tun_net_xmit,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001042 .ndo_change_mtu = tun_net_change_mtu,
Michał Mirosław88255372011-04-19 06:13:10 +00001043 .ndo_fix_features = tun_net_fix_features,
Jiri Pirkoafc4b132011-08-16 06:29:01 +00001044 .ndo_set_rx_mode = tun_net_mclist,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001045 .ndo_set_mac_address = eth_mac_addr,
1046 .ndo_validate_addr = eth_validate_addr,
Jason Wangc8d68e62012-10-31 19:46:00 +00001047 .ndo_select_queue = tun_select_queue,
Neil Hormanbebd0972011-06-15 05:25:01 +00001048#ifdef CONFIG_NET_POLL_CONTROLLER
1049 .ndo_poll_controller = tun_poll_controller,
1050#endif
Toshiaki Makita5e527962015-07-31 15:03:27 +09001051 .ndo_features_check = passthru_features_check,
Paolo Abenieaea34b2016-02-26 10:45:40 +01001052 .ndo_set_rx_headroom = tun_set_headroom,
Paolo Abeni608b9972016-04-13 10:52:20 +02001053 .ndo_get_stats64 = tun_net_get_stats64,
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001054};
1055
Pavel Emelyanov944a1372013-06-11 17:01:08 +04001056static void tun_flow_init(struct tun_struct *tun)
Jason Wang96442e422012-10-31 19:46:02 +00001057{
1058 int i;
1059
Jason Wang96442e422012-10-31 19:46:02 +00001060 for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
1061 INIT_HLIST_HEAD(&tun->flows[i]);
1062
1063 tun->ageing_time = TUN_FLOW_EXPIRE;
1064 setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun);
1065 mod_timer(&tun->flow_gc_timer,
1066 round_jiffies_up(jiffies + tun->ageing_time));
Jason Wang96442e422012-10-31 19:46:02 +00001067}
1068
1069static void tun_flow_uninit(struct tun_struct *tun)
1070{
1071 del_timer_sync(&tun->flow_gc_timer);
1072 tun_flow_flush(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001073}
1074
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075/* Initialize net device. */
1076static void tun_net_init(struct net_device *dev)
1077{
1078 struct tun_struct *tun = netdev_priv(dev);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001079
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001081 case IFF_TUN:
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001082 dev->netdev_ops = &tun_netdev_ops;
1083
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 /* Point-to-Point TUN Device */
1085 dev->hard_header_len = 0;
1086 dev->addr_len = 0;
1087 dev->mtu = 1500;
1088
1089 /* Zero header length */
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001090 dev->type = ARPHRD_NONE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 break;
1093
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001094 case IFF_TAP:
Kusanagi Kouichi7a0a9602008-12-29 18:23:28 -08001095 dev->netdev_ops = &tap_netdev_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 /* Ethernet TAP Device */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001097 ether_setup(dev);
Neil Horman550fd082011-07-26 06:05:38 +00001098 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
stephen hemmingera6768472012-12-10 15:16:00 +00001099 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100
Danny Kukawkaf2cedb62012-02-15 06:45:39 +00001101 eth_hw_addr_random(dev);
Brian Braunstein36226a82007-04-26 01:00:55 -07001102
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 break;
1104 }
1105}
1106
1107/* Character device part */
1108
1109/* Poll */
Jason Wangc8d68e62012-10-31 19:46:00 +00001110static unsigned int tun_chr_poll(struct file *file, poll_table *wait)
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001111{
Eric W. Biedermanb2430de2009-01-20 11:03:21 +00001112 struct tun_file *tfile = file->private_data;
1113 struct tun_struct *tun = __tun_get(tfile);
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +00001114 struct sock *sk;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001115 unsigned int mask = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
1117 if (!tun)
Eric W. Biedermaneac9e902009-01-20 10:59:05 +00001118 return POLLERR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
Jason Wang54f968d2012-10-31 19:45:57 +00001120 sk = tfile->socket.sk;
Mariusz Kozlowski3c8a9c62009-07-05 19:48:35 +00001121
Joe Perches6b8a66e2011-03-02 07:18:10 +00001122 tun_debug(KERN_INFO, tun, "tun_chr_poll\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
Xi Wang9e641bd2014-05-16 15:11:48 -07001124 poll_wait(file, sk_sleep(sk), wait);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001125
Jason Wang1576d982016-06-30 14:45:36 +08001126 if (!skb_array_empty(&tfile->tx_array))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 mask |= POLLIN | POLLRDNORM;
1128
Herbert Xu33dccbb2009-02-05 21:25:32 -08001129 if (sock_writeable(sk) ||
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001130 (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
Herbert Xu33dccbb2009-02-05 21:25:32 -08001131 sock_writeable(sk)))
1132 mask |= POLLOUT | POLLWRNORM;
1133
Eric W. Biedermanc70f1822009-01-20 11:07:17 +00001134 if (tun->dev->reg_state != NETREG_REGISTERED)
1135 mask = POLLERR;
1136
Eric W. Biederman631ab462009-01-20 11:00:40 +00001137 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 return mask;
1139}
1140
Rusty Russellf42157c2008-08-15 15:15:10 -07001141/* prepad is the amount to reserve at front. len is length after that.
1142 * linear is a hint as to how much to copy (usually headers). */
Jason Wang54f968d2012-10-31 19:45:57 +00001143static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001144 size_t prepad, size_t len,
1145 size_t linear, int noblock)
Rusty Russellf42157c2008-08-15 15:15:10 -07001146{
Jason Wang54f968d2012-10-31 19:45:57 +00001147 struct sock *sk = tfile->socket.sk;
Rusty Russellf42157c2008-08-15 15:15:10 -07001148 struct sk_buff *skb;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001149 int err;
Rusty Russellf42157c2008-08-15 15:15:10 -07001150
1151 /* Under a page? Don't bother with paged skb. */
Herbert Xu0eca93b2009-04-14 02:09:43 -07001152 if (prepad + len < PAGE_SIZE || !linear)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001153 linear = len;
Rusty Russellf42157c2008-08-15 15:15:10 -07001154
Herbert Xu33dccbb2009-02-05 21:25:32 -08001155 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
Eric Dumazet28d64272013-08-08 14:38:47 -07001156 &err, 0);
Rusty Russellf42157c2008-08-15 15:15:10 -07001157 if (!skb)
Herbert Xu33dccbb2009-02-05 21:25:32 -08001158 return ERR_PTR(err);
Rusty Russellf42157c2008-08-15 15:15:10 -07001159
1160 skb_reserve(skb, prepad);
1161 skb_put(skb, linear);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001162 skb->data_len = len - linear;
1163 skb->len += len - linear;
Rusty Russellf42157c2008-08-15 15:15:10 -07001164
1165 return skb;
1166}
1167
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168/* Get packet from user space buffer */
Jason Wang54f968d2012-10-31 19:45:57 +00001169static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
Al Virof5ff53b2014-06-19 15:36:49 -04001170 void *msg_control, struct iov_iter *from,
1171 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
Harvey Harrison09640e62009-02-01 00:45:17 -08001173 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 struct sk_buff *skb;
Al Virof5ff53b2014-06-19 15:36:49 -04001175 size_t total_len = iov_iter_count(from);
Paolo Abenieaea34b2016-02-26 10:45:40 +01001176 size_t len = total_len, align = tun->align, linear;
Rusty Russellf43798c2008-07-03 03:48:02 -07001177 struct virtio_net_hdr gso = { 0 };
Paolo Abeni608b9972016-04-13 10:52:20 +02001178 struct tun_pcpu_stats *stats;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001179 int good_linear;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001180 int copylen;
1181 bool zerocopy = false;
1182 int err;
Eric Dumazet49974422012-12-12 19:22:57 +00001183 u32 rxhash;
Al Virof5ff53b2014-06-19 15:36:49 -04001184 ssize_t n;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
Eric Dumazet1bd4978a2015-12-16 08:57:37 -08001186 if (!(tun->dev->flags & IFF_UP))
1187 return -EIO;
1188
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001189 if (!(tun->flags & IFF_NO_PI)) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001190 if (len < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001192 len -= sizeof(pi);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
Al Virof5ff53b2014-06-19 15:36:49 -04001194 n = copy_from_iter(&pi, sizeof(pi), from);
1195 if (n != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 return -EFAULT;
1197 }
1198
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001199 if (tun->flags & IFF_VNET_HDR) {
Dan Carpenter15718ea2013-08-15 15:52:57 +03001200 if (len < tun->vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001201 return -EINVAL;
Dan Carpenter15718ea2013-08-15 15:52:57 +03001202 len -= tun->vnet_hdr_sz;
Rusty Russellf43798c2008-07-03 03:48:02 -07001203
Al Virof5ff53b2014-06-19 15:36:49 -04001204 n = copy_from_iter(&gso, sizeof(gso), from);
1205 if (n != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001206 return -EFAULT;
1207
Herbert Xu49091222009-06-08 00:20:01 -07001208 if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001209 tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len))
1210 gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2);
Herbert Xu49091222009-06-08 00:20:01 -07001211
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001212 if (tun16_to_cpu(tun, gso.hdr_len) > len)
Rusty Russellf43798c2008-07-03 03:48:02 -07001213 return -EINVAL;
Herbert Xud8febb72014-11-30 18:03:31 +08001214 iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001215 }
1216
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001217 if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) {
stephen hemmingera504b862011-06-08 14:33:07 +00001218 align += NET_IP_ALIGN;
Herbert Xu0eca93b2009-04-14 02:09:43 -07001219 if (unlikely(len < ETH_HLEN ||
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001220 (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN)))
Rusty Russelle01bf1c2008-04-12 18:49:30 -07001221 return -EINVAL;
1222 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001223
Jason Wang96f8d9e2013-11-13 14:00:39 +08001224 good_linear = SKB_MAX_HEAD(align);
1225
Jason Wang88529172013-07-18 10:55:15 +08001226 if (msg_control) {
Al Virof5ff53b2014-06-19 15:36:49 -04001227 struct iov_iter i = *from;
1228
Jason Wang88529172013-07-18 10:55:15 +08001229 /* There are 256 bytes to be copied in skb, so there is
1230 * enough room for skb expand head in case it is used.
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001231 * The rest of the buffer is mapped from userspace.
1232 */
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001233 copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN;
Jason Wang96f8d9e2013-11-13 14:00:39 +08001234 if (copylen > good_linear)
1235 copylen = good_linear;
Jason Wang3dd5c332013-07-10 13:43:27 +08001236 linear = copylen;
Al Virof5ff53b2014-06-19 15:36:49 -04001237 iov_iter_advance(&i, copylen);
1238 if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS)
Jason Wang88529172013-07-18 10:55:15 +08001239 zerocopy = true;
1240 }
1241
1242 if (!zerocopy) {
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001243 copylen = len;
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001244 if (tun16_to_cpu(tun, gso.hdr_len) > good_linear)
Jason Wang96f8d9e2013-11-13 14:00:39 +08001245 linear = good_linear;
1246 else
Michael S. Tsirkin56f0dcc2014-10-23 22:59:31 +03001247 linear = tun16_to_cpu(tun, gso.hdr_len);
Jason Wang3dd5c332013-07-10 13:43:27 +08001248 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001249
Jason Wang3dd5c332013-07-10 13:43:27 +08001250 skb = tun_alloc_skb(tfile, align, copylen, linear, noblock);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001251 if (IS_ERR(skb)) {
1252 if (PTR_ERR(skb) != -EAGAIN)
Paolo Abeni608b9972016-04-13 10:52:20 +02001253 this_cpu_inc(tun->pcpu_stats->rx_dropped);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001254 return PTR_ERR(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 }
1256
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001257 if (zerocopy)
Al Virof5ff53b2014-06-19 15:36:49 -04001258 err = zerocopy_sg_from_iter(skb, from);
Jason Wang88529172013-07-18 10:55:15 +08001259 else {
Al Virof5ff53b2014-06-19 15:36:49 -04001260 err = skb_copy_datagram_from_iter(skb, 0, from, len);
Jason Wang88529172013-07-18 10:55:15 +08001261 if (!err && msg_control) {
1262 struct ubuf_info *uarg = msg_control;
1263 uarg->callback(uarg, false);
1264 }
1265 }
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001266
1267 if (err) {
Paolo Abeni608b9972016-04-13 10:52:20 +02001268 this_cpu_inc(tun->pcpu_stats->rx_dropped);
Dave Jones8f227572006-03-11 18:49:13 -08001269 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 return -EFAULT;
Dave Jones8f227572006-03-11 18:49:13 -08001271 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
Paolo Abenidf10db92016-06-14 00:00:04 +02001273 err = virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun));
1274 if (err) {
1275 this_cpu_inc(tun->pcpu_stats->rx_frame_errors);
1276 kfree_skb(skb);
1277 return -EINVAL;
1278 }
1279
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001281 case IFF_TUN:
1282 if (tun->flags & IFF_NO_PI) {
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -07001283 switch (skb->data[0] & 0xf0) {
1284 case 0x40:
1285 pi.proto = htons(ETH_P_IP);
1286 break;
1287 case 0x60:
1288 pi.proto = htons(ETH_P_IPV6);
1289 break;
1290 default:
Paolo Abeni608b9972016-04-13 10:52:20 +02001291 this_cpu_inc(tun->pcpu_stats->rx_dropped);
Ang Way Chuangf09f7ee2008-06-17 21:10:33 -07001292 kfree_skb(skb);
1293 return -EINVAL;
1294 }
1295 }
1296
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001297 skb_reset_mac_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 skb->protocol = pi.proto;
Arnaldo Carvalho de Melo4c13eb62007-04-25 17:40:23 -07001299 skb->dev = tun->dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001301 case IFF_TAP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 skb->protocol = eth_type_trans(skb, tun->dev);
1303 break;
Joe Perches6403eab2011-06-03 11:51:20 +00001304 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001306 /* copy skb_ubuf_info for callback when skb has no error */
1307 if (zerocopy) {
1308 skb_shinfo(skb)->destructor_arg = msg_control;
1309 skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00001310 skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001311 }
1312
Vlad Yasevich72f65102015-02-03 16:36:16 -05001313 skb_reset_network_header(skb);
Jason Wang40893fd2013-03-26 23:11:22 +00001314 skb_probe_transport_header(skb, 0);
Jason Wang38502af2013-03-25 20:19:56 +00001315
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001316 rxhash = skb_get_hash(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 netif_rx_ni(skb);
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001318
Paolo Abeni608b9972016-04-13 10:52:20 +02001319 stats = get_cpu_ptr(tun->pcpu_stats);
1320 u64_stats_update_begin(&stats->syncp);
1321 stats->rx_packets++;
1322 stats->rx_bytes += len;
1323 u64_stats_update_end(&stats->syncp);
1324 put_cpu_ptr(stats);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325
Jason Wang9e857222013-01-28 01:05:19 +00001326 tun_flow_update(tun, rxhash, tfile);
Michael S. Tsirkin06908992012-07-20 09:23:23 +00001327 return total_len;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001328}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329
Al Virof5ff53b2014-06-19 15:36:49 -04001330static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331{
Herbert Xu33dccbb2009-02-05 21:25:32 -08001332 struct file *file = iocb->ki_filp;
Herbert Xuab46d772009-02-14 20:46:39 -08001333 struct tun_struct *tun = tun_get(file);
Jason Wang54f968d2012-10-31 19:45:57 +00001334 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001335 ssize_t result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
1337 if (!tun)
1338 return -EBADFD;
1339
Al Virof5ff53b2014-06-19 15:36:49 -04001340 result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK);
Eric W. Biederman631ab462009-01-20 11:00:40 +00001341
1342 tun_put(tun);
1343 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344}
1345
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346/* Put packet to the user space buffer */
stephen hemminger6f7c1562011-06-08 14:33:08 +00001347static ssize_t tun_put_user(struct tun_struct *tun,
Jason Wang54f968d2012-10-31 19:45:57 +00001348 struct tun_file *tfile,
stephen hemminger6f7c1562011-06-08 14:33:08 +00001349 struct sk_buff *skb,
Herbert Xue0b46d02014-11-07 21:22:23 +08001350 struct iov_iter *iter)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351{
1352 struct tun_pi pi = { 0, skb->protocol };
Paolo Abeni608b9972016-04-13 10:52:20 +02001353 struct tun_pcpu_stats *stats;
Herbert Xue0b46d02014-11-07 21:22:23 +08001354 ssize_t total;
Jason Wang8c847d22014-11-13 16:54:14 +08001355 int vlan_offset = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001356 int vlan_hlen = 0;
Herbert Xu2eb783c2014-11-03 04:30:14 +08001357 int vnet_hdr_sz = 0;
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001358
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001359 if (skb_vlan_tag_present(skb))
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001360 vlan_hlen = VLAN_HLEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001362 if (tun->flags & IFF_VNET_HDR)
Herbert Xu2eb783c2014-11-03 04:30:14 +08001363 vnet_hdr_sz = tun->vnet_hdr_sz;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
Herbert Xue0b46d02014-11-07 21:22:23 +08001365 total = skb->len + vlan_hlen + vnet_hdr_sz;
1366
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001367 if (!(tun->flags & IFF_NO_PI)) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001368 if (iov_iter_count(iter) < sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 return -EINVAL;
1370
Herbert Xue0b46d02014-11-07 21:22:23 +08001371 total += sizeof(pi);
1372 if (iov_iter_count(iter) < total) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 /* Packet will be striped */
1374 pi.flags |= TUN_PKT_STRIP;
1375 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001376
Herbert Xue0b46d02014-11-07 21:22:23 +08001377 if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 return -EFAULT;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001379 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
Herbert Xu2eb783c2014-11-03 04:30:14 +08001381 if (vnet_hdr_sz) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001382 struct virtio_net_hdr gso = { 0 }; /* no info leak */
Mike Rapoport34166092016-06-08 16:09:20 +03001383 int ret;
1384
Herbert Xue0b46d02014-11-07 21:22:23 +08001385 if (iov_iter_count(iter) < vnet_hdr_sz)
Rusty Russellf43798c2008-07-03 03:48:02 -07001386 return -EINVAL;
1387
Mike Rapoport34166092016-06-08 16:09:20 +03001388 ret = virtio_net_hdr_from_skb(skb, &gso,
1389 tun_is_little_endian(tun));
1390 if (ret) {
Rusty Russellf43798c2008-07-03 03:48:02 -07001391 struct skb_shared_info *sinfo = skb_shinfo(skb);
Mike Rapoport34166092016-06-08 16:09:20 +03001392 pr_err("unexpected GSO type: "
1393 "0x%x, gso_size %d, hdr_len %d\n",
1394 sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size),
1395 tun16_to_cpu(tun, gso.hdr_len));
1396 print_hex_dump(KERN_ERR, "tun: ",
1397 DUMP_PREFIX_NONE,
1398 16, 1, skb->head,
1399 min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true);
1400 WARN_ON_ONCE(1);
1401 return -EINVAL;
1402 }
Rusty Russellf43798c2008-07-03 03:48:02 -07001403
Herbert Xue0b46d02014-11-07 21:22:23 +08001404 if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso))
Rusty Russellf43798c2008-07-03 03:48:02 -07001405 return -EFAULT;
Jason Wang8c847d22014-11-13 16:54:14 +08001406
1407 iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso));
Rusty Russellf43798c2008-07-03 03:48:02 -07001408 }
1409
Herbert Xua8f9bfd2014-11-03 04:30:13 +08001410 if (vlan_hlen) {
Herbert Xue0b46d02014-11-07 21:22:23 +08001411 int ret;
Jason Wang6680ec62013-07-25 13:00:33 +08001412 struct {
1413 __be16 h_vlan_proto;
1414 __be16 h_vlan_TCI;
1415 } veth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416
Jason Wang6680ec62013-07-25 13:00:33 +08001417 veth.h_vlan_proto = skb->vlan_proto;
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001418 veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419
Jason Wang6680ec62013-07-25 13:00:33 +08001420 vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto);
Jason Wang6680ec62013-07-25 13:00:33 +08001421
Herbert Xue0b46d02014-11-07 21:22:23 +08001422 ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset);
1423 if (ret || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001424 goto done;
1425
Herbert Xue0b46d02014-11-07 21:22:23 +08001426 ret = copy_to_iter(&veth, sizeof(veth), iter);
1427 if (ret != sizeof(veth) || !iov_iter_count(iter))
Jason Wang6680ec62013-07-25 13:00:33 +08001428 goto done;
1429 }
1430
Herbert Xue0b46d02014-11-07 21:22:23 +08001431 skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset);
Jason Wang6680ec62013-07-25 13:00:33 +08001432
1433done:
Paolo Abeni608b9972016-04-13 10:52:20 +02001434 /* caller is in process context, */
1435 stats = get_cpu_ptr(tun->pcpu_stats);
1436 u64_stats_update_begin(&stats->syncp);
1437 stats->tx_packets++;
1438 stats->tx_bytes += skb->len + vlan_hlen;
1439 u64_stats_update_end(&stats->syncp);
1440 put_cpu_ptr(tun->pcpu_stats);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
1442 return total;
1443}
1444
Jason Wang1576d982016-06-30 14:45:36 +08001445static struct sk_buff *tun_ring_recv(struct tun_file *tfile, int noblock,
1446 int *err)
1447{
1448 DECLARE_WAITQUEUE(wait, current);
1449 struct sk_buff *skb = NULL;
Jason Wangf48cc6b2016-07-04 13:53:38 +08001450 int error = 0;
Jason Wang1576d982016-06-30 14:45:36 +08001451
1452 skb = skb_array_consume(&tfile->tx_array);
1453 if (skb)
1454 goto out;
1455 if (noblock) {
Jason Wangf48cc6b2016-07-04 13:53:38 +08001456 error = -EAGAIN;
Jason Wang1576d982016-06-30 14:45:36 +08001457 goto out;
1458 }
1459
1460 add_wait_queue(&tfile->wq.wait, &wait);
1461 current->state = TASK_INTERRUPTIBLE;
1462
1463 while (1) {
1464 skb = skb_array_consume(&tfile->tx_array);
1465 if (skb)
1466 break;
1467 if (signal_pending(current)) {
Jason Wangf48cc6b2016-07-04 13:53:38 +08001468 error = -ERESTARTSYS;
Jason Wang1576d982016-06-30 14:45:36 +08001469 break;
1470 }
1471 if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) {
Jason Wangf48cc6b2016-07-04 13:53:38 +08001472 error = -EFAULT;
Jason Wang1576d982016-06-30 14:45:36 +08001473 break;
1474 }
1475
1476 schedule();
1477 }
1478
1479 current->state = TASK_RUNNING;
1480 remove_wait_queue(&tfile->wq.wait, &wait);
1481
1482out:
Jason Wangf48cc6b2016-07-04 13:53:38 +08001483 *err = error;
Jason Wang1576d982016-06-30 14:45:36 +08001484 return skb;
1485}
1486
Jason Wang54f968d2012-10-31 19:45:57 +00001487static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
Al Viro9b067032014-11-07 13:52:07 -05001488 struct iov_iter *to,
1489 int noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 struct sk_buff *skb;
Al Viro9b067032014-11-07 13:52:07 -05001492 ssize_t ret;
Jason Wang1576d982016-06-30 14:45:36 +08001493 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494
Rami Rosen3872baf2012-11-25 22:07:41 +00001495 tun_debug(KERN_INFO, tun, "tun_do_read\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496
Al Viro9b067032014-11-07 13:52:07 -05001497 if (!iov_iter_count(to))
1498 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499
Jason Wang1576d982016-06-30 14:45:36 +08001500 /* Read frames from ring */
1501 skb = tun_ring_recv(tfile, noblock, &err);
Herbert Xue0b46d02014-11-07 21:22:23 +08001502 if (!skb)
Alex Gartrell957f0942014-12-25 23:22:49 -08001503 return err;
Herbert Xue0b46d02014-11-07 21:22:23 +08001504
Al Viro9b067032014-11-07 13:52:07 -05001505 ret = tun_put_user(tun, tfile, skb, to);
Jason Wangf51a5e82014-12-01 16:53:15 +08001506 if (unlikely(ret < 0))
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001507 kfree_skb(skb);
Jason Wangf51a5e82014-12-01 16:53:15 +08001508 else
1509 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001511 return ret;
1512}
1513
Al Viro9b067032014-11-07 13:52:07 -05001514static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001515{
1516 struct file *file = iocb->ki_filp;
1517 struct tun_file *tfile = file->private_data;
1518 struct tun_struct *tun = __tun_get(tfile);
Al Viro9b067032014-11-07 13:52:07 -05001519 ssize_t len = iov_iter_count(to), ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001520
1521 if (!tun)
1522 return -EBADFD;
Al Viro9b067032014-11-07 13:52:07 -05001523 ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK);
David S. Miller42404c02013-12-10 22:05:45 -05001524 ret = min_t(ssize_t, ret, len);
Zhi Yong Wud0b7da8a2013-12-06 14:16:51 +08001525 if (ret > 0)
1526 iocb->ki_pos = ret;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001527 tun_put(tun);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 return ret;
1529}
1530
Jason Wang96442e422012-10-31 19:46:02 +00001531static void tun_free_netdev(struct net_device *dev)
1532{
1533 struct tun_struct *tun = netdev_priv(dev);
1534
Jason Wang4008e972012-12-13 23:53:30 +00001535 BUG_ON(!(list_empty(&tun->disabled)));
Paolo Abeni608b9972016-04-13 10:52:20 +02001536 free_percpu(tun->pcpu_stats);
Jason Wang96442e422012-10-31 19:46:02 +00001537 tun_flow_uninit(tun);
Paul Moore5dbbaf22013-01-14 07:12:19 +00001538 security_tun_dev_free_security(tun->security);
Jason Wang96442e422012-10-31 19:46:02 +00001539 free_netdev(dev);
1540}
1541
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542static void tun_setup(struct net_device *dev)
1543{
1544 struct tun_struct *tun = netdev_priv(dev);
1545
Eric W. Biederman0625c882012-02-07 16:48:55 -08001546 tun->owner = INVALID_UID;
1547 tun->group = INVALID_GID;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 dev->ethtool_ops = &tun_ethtool_ops;
Jason Wang96442e422012-10-31 19:46:02 +00001550 dev->destructor = tun_free_netdev;
Jason Wang016adb72016-04-08 13:26:48 +08001551 /* We prefer our own queue length */
1552 dev->tx_queue_len = TUN_READQ_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553}
1554
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001555/* Trivial set of netlink ops to allow deleting tun or tap
1556 * device with netlink.
1557 */
1558static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
1559{
1560 return -EINVAL;
1561}
1562
1563static struct rtnl_link_ops tun_link_ops __read_mostly = {
1564 .kind = DRV_NAME,
1565 .priv_size = sizeof(struct tun_struct),
1566 .setup = tun_setup,
1567 .validate = tun_validate,
1568};
1569
Herbert Xu33dccbb2009-02-05 21:25:32 -08001570static void tun_sock_write_space(struct sock *sk)
1571{
Jason Wang54f968d2012-10-31 19:45:57 +00001572 struct tun_file *tfile;
Eric Dumazet43815482010-04-29 11:01:49 +00001573 wait_queue_head_t *wqueue;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001574
1575 if (!sock_writeable(sk))
1576 return;
1577
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001578 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags))
Herbert Xu33dccbb2009-02-05 21:25:32 -08001579 return;
1580
Eric Dumazet43815482010-04-29 11:01:49 +00001581 wqueue = sk_sleep(sk);
1582 if (wqueue && waitqueue_active(wqueue))
1583 wake_up_interruptible_sync_poll(wqueue, POLLOUT |
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001584 POLLWRNORM | POLLWRBAND);
Herbert Xuc722c622009-06-03 21:45:55 -07001585
Jason Wang54f968d2012-10-31 19:45:57 +00001586 tfile = container_of(sk, struct tun_file, sk);
1587 kill_fasync(&tfile->fasync, SIGIO, POLL_OUT);
Herbert Xu33dccbb2009-02-05 21:25:32 -08001588}
1589
Ying Xue1b784142015-03-02 15:37:48 +08001590static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001591{
Jason Wang54f968d2012-10-31 19:45:57 +00001592 int ret;
1593 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1594 struct tun_struct *tun = __tun_get(tfile);
1595
1596 if (!tun)
1597 return -EBADFD;
Al Virof5ff53b2014-06-19 15:36:49 -04001598
Al Viroc0371da2014-11-24 10:42:55 -05001599 ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04001600 m->msg_flags & MSG_DONTWAIT);
Jason Wang54f968d2012-10-31 19:45:57 +00001601 tun_put(tun);
1602 return ret;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001603}
1604
Ying Xue1b784142015-03-02 15:37:48 +08001605static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001606 int flags)
1607{
Jason Wang54f968d2012-10-31 19:45:57 +00001608 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1609 struct tun_struct *tun = __tun_get(tfile);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001610 int ret;
Jason Wang54f968d2012-10-31 19:45:57 +00001611
1612 if (!tun)
1613 return -EBADFD;
1614
Richard Cochraneda29772013-07-19 19:40:10 +02001615 if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) {
Gao feng3811ae72013-04-24 21:59:23 +00001616 ret = -EINVAL;
1617 goto out;
1618 }
Richard Cochraneda29772013-07-19 19:40:10 +02001619 if (flags & MSG_ERRQUEUE) {
1620 ret = sock_recv_errqueue(sock->sk, m, total_len,
1621 SOL_PACKET, TUN_TX_TIMESTAMP);
1622 goto out;
1623 }
Al Viroc0371da2014-11-24 10:42:55 -05001624 ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT);
Alex Gartrell87897932014-12-25 23:05:03 -08001625 if (ret > (ssize_t)total_len) {
David S. Miller42404c02013-12-10 22:05:45 -05001626 m->msg_flags |= MSG_TRUNC;
1627 ret = flags & MSG_TRUNC ? ret : total_len;
1628 }
Gao feng3811ae72013-04-24 21:59:23 +00001629out:
Jason Wang54f968d2012-10-31 19:45:57 +00001630 tun_put(tun);
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001631 return ret;
1632}
1633
Jason Wang1576d982016-06-30 14:45:36 +08001634static int tun_peek_len(struct socket *sock)
1635{
1636 struct tun_file *tfile = container_of(sock, struct tun_file, socket);
1637 struct tun_struct *tun;
1638 int ret = 0;
1639
1640 tun = __tun_get(tfile);
1641 if (!tun)
1642 return 0;
1643
1644 ret = skb_array_peek_len(&tfile->tx_array);
1645 tun_put(tun);
1646
1647 return ret;
1648}
1649
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001650/* Ops structure to mimic raw sockets with tun */
1651static const struct proto_ops tun_socket_ops = {
Jason Wang1576d982016-06-30 14:45:36 +08001652 .peek_len = tun_peek_len,
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00001653 .sendmsg = tun_sendmsg,
1654 .recvmsg = tun_recvmsg,
1655};
1656
Herbert Xu33dccbb2009-02-05 21:25:32 -08001657static struct proto tun_proto = {
1658 .name = "tun",
1659 .owner = THIS_MODULE,
Jason Wang54f968d2012-10-31 19:45:57 +00001660 .obj_size = sizeof(struct tun_file),
Herbert Xu33dccbb2009-02-05 21:25:32 -08001661};
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001662
David Woodhouse980c9e82009-05-09 22:54:21 -07001663static int tun_flags(struct tun_struct *tun)
1664{
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001665 return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP);
David Woodhouse980c9e82009-05-09 22:54:21 -07001666}
1667
1668static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
1669 char *buf)
1670{
1671 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
1672 return sprintf(buf, "0x%x\n", tun_flags(tun));
1673}
1674
1675static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
1676 char *buf)
1677{
1678 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001679 return uid_valid(tun->owner)?
1680 sprintf(buf, "%u\n",
1681 from_kuid_munged(current_user_ns(), tun->owner)):
1682 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001683}
1684
1685static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
1686 char *buf)
1687{
1688 struct tun_struct *tun = netdev_priv(to_net_dev(dev));
Eric W. Biederman0625c882012-02-07 16:48:55 -08001689 return gid_valid(tun->group) ?
1690 sprintf(buf, "%u\n",
1691 from_kgid_munged(current_user_ns(), tun->group)):
1692 sprintf(buf, "-1\n");
David Woodhouse980c9e82009-05-09 22:54:21 -07001693}
1694
1695static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
1696static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
1697static DEVICE_ATTR(group, 0444, tun_show_group, NULL);
1698
Takashi Iwaic4d33e22015-02-04 14:37:34 +01001699static struct attribute *tun_dev_attrs[] = {
1700 &dev_attr_tun_flags.attr,
1701 &dev_attr_owner.attr,
1702 &dev_attr_group.attr,
1703 NULL
1704};
1705
1706static const struct attribute_group tun_attr_group = {
1707 .attrs = tun_dev_attrs
1708};
1709
Pavel Emelyanovd647a592008-04-16 00:41:16 -07001710static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711{
1712 struct tun_struct *tun;
Jason Wang54f968d2012-10-31 19:45:57 +00001713 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 struct net_device *dev;
1715 int err;
1716
Jason Wang7c0c3b12013-01-11 16:59:33 +00001717 if (tfile->detached)
1718 return -EINVAL;
1719
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001720 dev = __dev_get_by_name(net, ifr->ifr_name);
1721 if (dev) {
David Woodhousef85ba782009-04-27 03:23:54 -07001722 if (ifr->ifr_flags & IFF_TUN_EXCL)
1723 return -EBUSY;
Eric W. Biederman74a3e5a2009-01-20 10:56:20 +00001724 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
1725 tun = netdev_priv(dev);
1726 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
1727 tun = netdev_priv(dev);
1728 else
1729 return -EINVAL;
1730
Jason Wang8e6d91a2013-05-28 18:32:11 +00001731 if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) !=
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001732 !!(tun->flags & IFF_MULTI_QUEUE))
Jason Wang8e6d91a2013-05-28 18:32:11 +00001733 return -EINVAL;
1734
Jason Wangcde8b152012-10-31 19:46:01 +00001735 if (tun_not_capable(tun))
Paul Moore2b980db2009-08-28 18:12:43 -04001736 return -EPERM;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001737 err = security_tun_dev_open(tun->security);
Paul Moore2b980db2009-08-28 18:12:43 -04001738 if (err < 0)
1739 return err;
1740
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001741 err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER);
Eric W. Biedermana7385ba2009-01-20 10:57:48 +00001742 if (err < 0)
1743 return err;
Jason Wang4008e972012-12-13 23:53:30 +00001744
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001745 if (tun->flags & IFF_MULTI_QUEUE &&
Jason Wange8dbad62013-04-22 20:40:39 +00001746 (tun->numqueues + tun->numdisabled > 1)) {
1747 /* One or more queue has already been attached, no need
1748 * to initialize the device again.
1749 */
1750 return 0;
1751 }
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001752 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 else {
1754 char *name;
1755 unsigned long flags = 0;
Jason Wangedfb6a12013-01-23 03:59:12 +00001756 int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
1757 MAX_TAP_QUEUES : 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
Eric W. Biedermanc260b772012-11-18 21:34:11 +00001759 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001760 return -EPERM;
Paul Moore2b980db2009-08-28 18:12:43 -04001761 err = security_tun_dev_create();
1762 if (err < 0)
1763 return err;
David Woodhouseca6bb5d2006-06-22 16:07:52 -07001764
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 /* Set dev type */
1766 if (ifr->ifr_flags & IFF_TUN) {
1767 /* TUN device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001768 flags |= IFF_TUN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 name = "tun%d";
1770 } else if (ifr->ifr_flags & IFF_TAP) {
1771 /* TAP device */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001772 flags |= IFF_TAP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 name = "tap%d";
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001774 } else
Kusanagi Kouichi36989b92009-09-16 21:36:13 +00001775 return -EINVAL;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04001776
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 if (*ifr->ifr_name)
1778 name = ifr->ifr_name;
1779
Jason Wangc8d68e62012-10-31 19:46:00 +00001780 dev = alloc_netdev_mqs(sizeof(struct tun_struct), name,
Tom Gundersenc835a672014-07-14 16:37:24 +02001781 NET_NAME_UNKNOWN, tun_setup, queues,
1782 queues);
Jason Wangedfb6a12013-01-23 03:59:12 +00001783
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 if (!dev)
1785 return -ENOMEM;
1786
Pavel Emelyanovfc54c652008-04-16 00:41:53 -07001787 dev_net_set(dev, net);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08001788 dev->rtnl_link_ops = &tun_link_ops;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04001789 dev->ifindex = tfile->ifindex;
Takashi Iwaic4d33e22015-02-04 14:37:34 +01001790 dev->sysfs_groups[0] = &tun_attr_group;
Stephen Hemminger758e43b2008-11-19 22:10:37 -08001791
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 tun = netdev_priv(dev);
1793 tun->dev = dev;
1794 tun->flags = flags;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07001795 tun->txflt.count = 0;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02001796 tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797
Paolo Abenieaea34b2016-02-26 10:45:40 +01001798 tun->align = NET_SKB_PAD;
Jason Wang54f968d2012-10-31 19:45:57 +00001799 tun->filter_attached = false;
1800 tun->sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08001801
Paolo Abeni608b9972016-04-13 10:52:20 +02001802 tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats);
1803 if (!tun->pcpu_stats) {
1804 err = -ENOMEM;
1805 goto err_free_dev;
1806 }
1807
Jason Wang96442e422012-10-31 19:46:02 +00001808 spin_lock_init(&tun->lock);
1809
Paul Moore5dbbaf22013-01-14 07:12:19 +00001810 err = security_tun_dev_alloc_security(&tun->security);
1811 if (err < 0)
Paolo Abeni608b9972016-04-13 10:52:20 +02001812 goto err_free_stat;
Paul Moore2b980db2009-08-28 18:12:43 -04001813
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 tun_net_init(dev);
Pavel Emelyanov944a1372013-06-11 17:01:08 +04001815 tun_flow_init(tun);
Jason Wang96442e422012-10-31 19:46:02 +00001816
Michał Mirosław88255372011-04-19 06:13:10 +00001817 dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
Jason Wang6680ec62013-07-25 13:00:33 +08001818 TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
1819 NETIF_F_HW_VLAN_STAG_TX;
Paolo Abeni2a2bbf12016-04-14 18:39:39 +02001820 dev->features = dev->hw_features | NETIF_F_LLTX;
Fernando Luis Vazquez Cao6671b222014-02-18 21:20:09 +09001821 dev->vlan_features = dev->features &
1822 ~(NETIF_F_HW_VLAN_CTAG_TX |
1823 NETIF_F_HW_VLAN_STAG_TX);
Michał Mirosław88255372011-04-19 06:13:10 +00001824
Jason Wang4008e972012-12-13 23:53:30 +00001825 INIT_LIST_HEAD(&tun->disabled);
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001826 err = tun_attach(tun, file, false);
Jason Wangeb0fb362012-12-02 17:19:45 +00001827 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001828 goto err_free_flow;
Jason Wangeb0fb362012-12-02 17:19:45 +00001829
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 err = register_netdevice(tun->dev);
1831 if (err < 0)
Jason Wang662ca432013-09-11 18:09:48 +08001832 goto err_detach;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 }
1834
Michael S. Tsirkinaf668b32013-01-28 00:38:02 +00001835 netif_carrier_on(tun->dev);
1836
Joe Perches6b8a66e2011-03-02 07:18:10 +00001837 tun_debug(KERN_INFO, tun, "tun_set_iff\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02001839 tun->flags = (tun->flags & ~TUN_FEATURES) |
1840 (ifr->ifr_flags & TUN_FEATURES);
Jason Wangc8d68e62012-10-31 19:46:00 +00001841
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001842 /* Make sure persistent devices do not get stuck in
1843 * xoff state.
1844 */
1845 if (netif_running(tun->dev))
Jason Wangc8d68e62012-10-31 19:46:00 +00001846 netif_tx_wake_all_queues(tun->dev);
Max Krasnyanskye35259a2008-07-10 16:59:11 -07001847
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 strcpy(ifr->ifr_name, tun->dev->name);
1849 return 0;
1850
Jason Wang662ca432013-09-11 18:09:48 +08001851err_detach:
1852 tun_detach_all(dev);
1853err_free_flow:
1854 tun_flow_uninit(tun);
1855 security_tun_dev_free_security(tun->security);
Paolo Abeni608b9972016-04-13 10:52:20 +02001856err_free_stat:
1857 free_percpu(tun->pcpu_stats);
Jason Wang662ca432013-09-11 18:09:48 +08001858err_free_dev:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 free_netdev(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 return err;
1861}
1862
Rami Rosen9ce99cf2012-11-23 03:58:10 +00001863static void tun_get_iff(struct net *net, struct tun_struct *tun,
Herbert Xu876bfd42009-08-06 14:22:44 +00001864 struct ifreq *ifr)
Mark McLoughline3b99552008-08-15 15:09:56 -07001865{
Joe Perches6b8a66e2011-03-02 07:18:10 +00001866 tun_debug(KERN_INFO, tun, "tun_get_iff\n");
Mark McLoughline3b99552008-08-15 15:09:56 -07001867
1868 strcpy(ifr->ifr_name, tun->dev->name);
1869
David Woodhouse980c9e82009-05-09 22:54:21 -07001870 ifr->ifr_flags = tun_flags(tun);
Mark McLoughline3b99552008-08-15 15:09:56 -07001871
Mark McLoughline3b99552008-08-15 15:09:56 -07001872}
1873
Rusty Russell5228ddc2008-07-03 03:46:16 -07001874/* This is like a cut-down ethtool ops, except done via tun fd so no
1875 * privs required. */
Michał Mirosław88255372011-04-19 06:13:10 +00001876static int set_offload(struct tun_struct *tun, unsigned long arg)
Rusty Russell5228ddc2008-07-03 03:46:16 -07001877{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001878 netdev_features_t features = 0;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001879
1880 if (arg & TUN_F_CSUM) {
Michał Mirosław88255372011-04-19 06:13:10 +00001881 features |= NETIF_F_HW_CSUM;
Rusty Russell5228ddc2008-07-03 03:46:16 -07001882 arg &= ~TUN_F_CSUM;
1883
1884 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1885 if (arg & TUN_F_TSO_ECN) {
1886 features |= NETIF_F_TSO_ECN;
1887 arg &= ~TUN_F_TSO_ECN;
1888 }
1889 if (arg & TUN_F_TSO4)
1890 features |= NETIF_F_TSO;
1891 if (arg & TUN_F_TSO6)
1892 features |= NETIF_F_TSO6;
1893 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1894 }
Vlad Yaseviche3e3c422015-02-03 16:36:17 -05001895
1896 if (arg & TUN_F_UFO) {
1897 features |= NETIF_F_UFO;
1898 arg &= ~TUN_F_UFO;
1899 }
Rusty Russell5228ddc2008-07-03 03:46:16 -07001900 }
1901
1902 /* This gives the user a way to test for new features in future by
1903 * trying to set them. */
1904 if (arg)
1905 return -EINVAL;
1906
Michał Mirosław88255372011-04-19 06:13:10 +00001907 tun->set_features = features;
1908 netdev_update_features(tun->dev);
Rusty Russell5228ddc2008-07-03 03:46:16 -07001909
1910 return 0;
1911}
1912
Jason Wangc8d68e62012-10-31 19:46:00 +00001913static void tun_detach_filter(struct tun_struct *tun, int n)
1914{
1915 int i;
1916 struct tun_file *tfile;
1917
1918 for (i = 0; i < n; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001919 tfile = rtnl_dereference(tun->tfiles[i]);
Hannes Frederic Sowa8ced4252016-04-05 17:10:16 +02001920 lock_sock(tfile->socket.sk);
1921 sk_detach_filter(tfile->socket.sk);
1922 release_sock(tfile->socket.sk);
Jason Wangc8d68e62012-10-31 19:46:00 +00001923 }
1924
1925 tun->filter_attached = false;
1926}
1927
1928static int tun_attach_filter(struct tun_struct *tun)
1929{
1930 int i, ret = 0;
1931 struct tun_file *tfile;
1932
1933 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001934 tfile = rtnl_dereference(tun->tfiles[i]);
Hannes Frederic Sowa8ced4252016-04-05 17:10:16 +02001935 lock_sock(tfile->socket.sk);
1936 ret = sk_attach_filter(&tun->fprog, tfile->socket.sk);
1937 release_sock(tfile->socket.sk);
Jason Wangc8d68e62012-10-31 19:46:00 +00001938 if (ret) {
1939 tun_detach_filter(tun, i);
1940 return ret;
1941 }
1942 }
1943
1944 tun->filter_attached = true;
1945 return ret;
1946}
1947
1948static void tun_set_sndbuf(struct tun_struct *tun)
1949{
1950 struct tun_file *tfile;
1951 int i;
1952
1953 for (i = 0; i < tun->numqueues; i++) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001954 tfile = rtnl_dereference(tun->tfiles[i]);
Jason Wangc8d68e62012-10-31 19:46:00 +00001955 tfile->socket.sk->sk_sndbuf = tun->sndbuf;
1956 }
1957}
1958
Jason Wangcde8b152012-10-31 19:46:01 +00001959static int tun_set_queue(struct file *file, struct ifreq *ifr)
1960{
1961 struct tun_file *tfile = file->private_data;
1962 struct tun_struct *tun;
Jason Wangcde8b152012-10-31 19:46:01 +00001963 int ret = 0;
1964
1965 rtnl_lock();
1966
1967 if (ifr->ifr_flags & IFF_ATTACH_QUEUE) {
Jason Wang4008e972012-12-13 23:53:30 +00001968 tun = tfile->detached;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001969 if (!tun) {
Jason Wangcde8b152012-10-31 19:46:01 +00001970 ret = -EINVAL;
Paul Moore5dbbaf22013-01-14 07:12:19 +00001971 goto unlock;
1972 }
1973 ret = security_tun_dev_attach_queue(tun->security);
1974 if (ret < 0)
1975 goto unlock;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04001976 ret = tun_attach(tun, file, false);
Jason Wang4008e972012-12-13 23:53:30 +00001977 } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) {
Jason Wangb8deabd2013-01-11 16:59:32 +00001978 tun = rtnl_dereference(tfile->tun);
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02001979 if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached)
Jason Wang4008e972012-12-13 23:53:30 +00001980 ret = -EINVAL;
1981 else
1982 __tun_detach(tfile, false);
1983 } else
Jason Wangcde8b152012-10-31 19:46:01 +00001984 ret = -EINVAL;
1985
Paul Moore5dbbaf22013-01-14 07:12:19 +00001986unlock:
Jason Wangcde8b152012-10-31 19:46:01 +00001987 rtnl_unlock();
1988 return ret;
1989}
1990
Arnd Bergmann50857e22009-11-06 22:52:32 -08001991static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
1992 unsigned long arg, int ifreq_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993{
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00001994 struct tun_file *tfile = file->private_data;
Eric W. Biederman631ab462009-01-20 11:00:40 +00001995 struct tun_struct *tun;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 void __user* argp = (void __user*)arg;
1997 struct ifreq ifr;
Eric W. Biederman0625c882012-02-07 16:48:55 -08001998 kuid_t owner;
1999 kgid_t group;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002000 int sndbuf;
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02002001 int vnet_hdr_sz;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002002 unsigned int ifindex;
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02002003 int le;
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002004 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005
Jason Wangcde8b152012-10-31 19:46:01 +00002006 if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) {
Arnd Bergmann50857e22009-11-06 22:52:32 -08002007 if (copy_from_user(&ifr, argp, ifreq_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 return -EFAULT;
David S. Miller8bbb1812012-07-30 14:52:48 -07002009 } else {
Mathias Krausea117dac2012-07-29 19:45:14 +00002010 memset(&ifr, 0, sizeof(ifr));
David S. Miller8bbb1812012-07-30 14:52:48 -07002011 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00002012 if (cmd == TUNGETFEATURES) {
2013 /* Currently this just means: "what IFF flags are valid?".
2014 * This is needed because we never checked for invalid flags on
Michael S. Tsirkin031f5e02014-11-19 14:44:40 +02002015 * TUNSETIFF.
2016 */
2017 return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES,
Eric W. Biederman631ab462009-01-20 11:00:40 +00002018 (unsigned int __user*)argp);
Jason Wangcde8b152012-10-31 19:46:01 +00002019 } else if (cmd == TUNSETQUEUE)
2020 return tun_set_queue(file, &ifr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002021
Jason Wangc8d68e62012-10-31 19:46:00 +00002022 ret = 0;
Herbert Xu876bfd42009-08-06 14:22:44 +00002023 rtnl_lock();
2024
Eric W. Biederman36b50ba2009-01-20 11:01:48 +00002025 tun = __tun_get(tfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 if (cmd == TUNSETIFF && !tun) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 ifr.ifr_name[IFNAMSIZ-1] = '\0';
2028
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002029 ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030
Herbert Xu876bfd42009-08-06 14:22:44 +00002031 if (ret)
2032 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Arnd Bergmann50857e22009-11-06 22:52:32 -08002034 if (copy_to_user(argp, &ifr, ifreq_len))
Herbert Xu876bfd42009-08-06 14:22:44 +00002035 ret = -EFAULT;
2036 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 }
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002038 if (cmd == TUNSETIFINDEX) {
2039 ret = -EPERM;
2040 if (tun)
2041 goto unlock;
2042
2043 ret = -EFAULT;
2044 if (copy_from_user(&ifindex, argp, sizeof(ifindex)))
2045 goto unlock;
2046
2047 ret = 0;
2048 tfile->ifindex = ifindex;
2049 goto unlock;
2050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051
Herbert Xu876bfd42009-08-06 14:22:44 +00002052 ret = -EBADFD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 if (!tun)
Herbert Xu876bfd42009-08-06 14:22:44 +00002054 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055
Jason Wang1e588332012-10-31 19:45:56 +00002056 tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057
Eric W. Biederman631ab462009-01-20 11:00:40 +00002058 ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 switch (cmd) {
Mark McLoughline3b99552008-08-15 15:09:56 -07002060 case TUNGETIFF:
Rami Rosen9ce99cf2012-11-23 03:58:10 +00002061 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
Mark McLoughline3b99552008-08-15 15:09:56 -07002062
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04002063 if (tfile->detached)
2064 ifr.ifr_flags |= IFF_DETACH_QUEUE;
Pavel Emelyanov849c9b62013-08-21 14:32:21 +04002065 if (!tfile->socket.sk->sk_filter)
2066 ifr.ifr_flags |= IFF_NOFILTER;
Pavel Emelyanov3d407a82013-08-21 14:32:00 +04002067
Arnd Bergmann50857e22009-11-06 22:52:32 -08002068 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00002069 ret = -EFAULT;
Mark McLoughline3b99552008-08-15 15:09:56 -07002070 break;
2071
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 case TUNSETNOCSUM:
2073 /* Disable/Enable checksum */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074
Michał Mirosław88255372011-04-19 06:13:10 +00002075 /* [unimplemented] */
2076 tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n",
Joe Perches6b8a66e2011-03-02 07:18:10 +00002077 arg ? "disabled" : "enabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 break;
2079
2080 case TUNSETPERSIST:
Jason Wang54f968d2012-10-31 19:45:57 +00002081 /* Disable/Enable persist mode. Keep an extra reference to the
2082 * module to prevent the module being unprobed.
2083 */
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002084 if (arg && !(tun->flags & IFF_PERSIST)) {
2085 tun->flags |= IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00002086 __module_get(THIS_MODULE);
Jason Wangdd38bd82013-01-11 16:59:34 +00002087 }
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002088 if (!arg && (tun->flags & IFF_PERSIST)) {
2089 tun->flags &= ~IFF_PERSIST;
Jason Wang54f968d2012-10-31 19:45:57 +00002090 module_put(THIS_MODULE);
2091 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
Joe Perches6b8a66e2011-03-02 07:18:10 +00002093 tun_debug(KERN_INFO, tun, "persist %s\n",
2094 arg ? "enabled" : "disabled");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 break;
2096
2097 case TUNSETOWNER:
2098 /* Set owner of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08002099 owner = make_kuid(current_user_ns(), arg);
2100 if (!uid_valid(owner)) {
2101 ret = -EINVAL;
2102 break;
2103 }
2104 tun->owner = owner;
Jason Wang1e588332012-10-31 19:45:56 +00002105 tun_debug(KERN_INFO, tun, "owner set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08002106 from_kuid(&init_user_ns, tun->owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 break;
2108
Guido Guenther8c644622007-07-02 22:50:25 -07002109 case TUNSETGROUP:
2110 /* Set group of the device */
Eric W. Biederman0625c882012-02-07 16:48:55 -08002111 group = make_kgid(current_user_ns(), arg);
2112 if (!gid_valid(group)) {
2113 ret = -EINVAL;
2114 break;
2115 }
2116 tun->group = group;
Jason Wang1e588332012-10-31 19:45:56 +00002117 tun_debug(KERN_INFO, tun, "group set to %u\n",
Eric W. Biederman0625c882012-02-07 16:48:55 -08002118 from_kgid(&init_user_ns, tun->group));
Guido Guenther8c644622007-07-02 22:50:25 -07002119 break;
2120
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002121 case TUNSETLINK:
2122 /* Only allow setting the type when the interface is down */
2123 if (tun->dev->flags & IFF_UP) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002124 tun_debug(KERN_INFO, tun,
2125 "Linktype set failed because interface is up\n");
David S. Miller48abfe02008-04-23 19:37:58 -07002126 ret = -EBUSY;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002127 } else {
2128 tun->dev->type = (int) arg;
Joe Perches6b8a66e2011-03-02 07:18:10 +00002129 tun_debug(KERN_INFO, tun, "linktype set to %d\n",
2130 tun->dev->type);
David S. Miller48abfe02008-04-23 19:37:58 -07002131 ret = 0;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002132 }
Eric W. Biederman631ab462009-01-20 11:00:40 +00002133 break;
Mike Kershawff4cc3a2005-09-01 17:40:05 -07002134
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135#ifdef TUN_DEBUG
2136 case TUNSETDEBUG:
2137 tun->debug = arg;
2138 break;
2139#endif
Rusty Russell5228ddc2008-07-03 03:46:16 -07002140 case TUNSETOFFLOAD:
Michał Mirosław88255372011-04-19 06:13:10 +00002141 ret = set_offload(tun, arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002142 break;
Rusty Russell5228ddc2008-07-03 03:46:16 -07002143
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002144 case TUNSETTXFILTER:
2145 /* Can be set only for TAPs */
Eric W. Biederman631ab462009-01-20 11:00:40 +00002146 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002147 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Eric W. Biederman631ab462009-01-20 11:00:40 +00002148 break;
Harvey Harrisonc0e5a8c2008-07-16 12:45:34 -07002149 ret = update_filter(&tun->txflt, (void __user *)arg);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002150 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151
2152 case SIOCGIFHWADDR:
Uwe Kleine-Königb5950762010-11-01 15:38:34 -04002153 /* Get hw address */
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002154 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
2155 ifr.ifr_hwaddr.sa_family = tun->dev->type;
Arnd Bergmann50857e22009-11-06 22:52:32 -08002156 if (copy_to_user(argp, &ifr, ifreq_len))
Eric W. Biederman631ab462009-01-20 11:00:40 +00002157 ret = -EFAULT;
2158 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159
2160 case SIOCSIFHWADDR:
Max Krasnyanskyf271b2c2008-07-14 22:18:19 -07002161 /* Set hw address */
Joe Perches6b8a66e2011-03-02 07:18:10 +00002162 tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n",
2163 ifr.ifr_hwaddr.sa_data);
Kim B. Heino40102372008-02-29 12:26:21 -08002164
Kim B. Heino40102372008-02-29 12:26:21 -08002165 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002166 break;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002167
2168 case TUNGETSNDBUF:
Jason Wang54f968d2012-10-31 19:45:57 +00002169 sndbuf = tfile->socket.sk->sk_sndbuf;
Herbert Xu33dccbb2009-02-05 21:25:32 -08002170 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
2171 ret = -EFAULT;
2172 break;
2173
2174 case TUNSETSNDBUF:
2175 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
2176 ret = -EFAULT;
2177 break;
2178 }
2179
Jason Wangc8d68e62012-10-31 19:46:00 +00002180 tun->sndbuf = sndbuf;
2181 tun_set_sndbuf(tun);
Herbert Xu33dccbb2009-02-05 21:25:32 -08002182 break;
2183
Michael S. Tsirkind9d52b52010-03-17 17:45:01 +02002184 case TUNGETVNETHDRSZ:
2185 vnet_hdr_sz = tun->vnet_hdr_sz;
2186 if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
2187 ret = -EFAULT;
2188 break;
2189
2190 case TUNSETVNETHDRSZ:
2191 if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
2192 ret = -EFAULT;
2193 break;
2194 }
2195 if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
2196 ret = -EINVAL;
2197 break;
2198 }
2199
2200 tun->vnet_hdr_sz = vnet_hdr_sz;
2201 break;
2202
Michael S. Tsirkin1cf8e412014-12-16 15:05:06 +02002203 case TUNGETVNETLE:
2204 le = !!(tun->flags & TUN_VNET_LE);
2205 if (put_user(le, (int __user *)argp))
2206 ret = -EFAULT;
2207 break;
2208
2209 case TUNSETVNETLE:
2210 if (get_user(le, (int __user *)argp)) {
2211 ret = -EFAULT;
2212 break;
2213 }
2214 if (le)
2215 tun->flags |= TUN_VNET_LE;
2216 else
2217 tun->flags &= ~TUN_VNET_LE;
2218 break;
2219
Greg Kurz8b8e6582015-04-24 14:50:36 +02002220 case TUNGETVNETBE:
2221 ret = tun_get_vnet_be(tun, argp);
2222 break;
2223
2224 case TUNSETVNETBE:
2225 ret = tun_set_vnet_be(tun, argp);
2226 break;
2227
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002228 case TUNATTACHFILTER:
2229 /* Can be set only for TAPs */
2230 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002231 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002232 break;
2233 ret = -EFAULT;
Jason Wang54f968d2012-10-31 19:45:57 +00002234 if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog)))
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002235 break;
2236
Jason Wangc8d68e62012-10-31 19:46:00 +00002237 ret = tun_attach_filter(tun);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002238 break;
2239
2240 case TUNDETACHFILTER:
2241 /* Can be set only for TAPs */
2242 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002243 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002244 break;
Jason Wangc8d68e62012-10-31 19:46:00 +00002245 ret = 0;
2246 tun_detach_filter(tun, tun->numqueues);
Michael S. Tsirkin99405162010-02-14 01:01:10 +00002247 break;
2248
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002249 case TUNGETFILTER:
2250 ret = -EINVAL;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002251 if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
Pavel Emelyanov76975e92013-08-21 14:32:39 +04002252 break;
2253 ret = -EFAULT;
2254 if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog)))
2255 break;
2256 ret = 0;
2257 break;
2258
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 default:
Eric W. Biederman631ab462009-01-20 11:00:40 +00002260 ret = -EINVAL;
2261 break;
Joe Perchesee289b62010-05-17 22:47:34 -07002262 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263
Herbert Xu876bfd42009-08-06 14:22:44 +00002264unlock:
2265 rtnl_unlock();
2266 if (tun)
2267 tun_put(tun);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002268 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269}
2270
Arnd Bergmann50857e22009-11-06 22:52:32 -08002271static long tun_chr_ioctl(struct file *file,
2272 unsigned int cmd, unsigned long arg)
2273{
2274 return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
2275}
2276
2277#ifdef CONFIG_COMPAT
2278static long tun_chr_compat_ioctl(struct file *file,
2279 unsigned int cmd, unsigned long arg)
2280{
2281 switch (cmd) {
2282 case TUNSETIFF:
2283 case TUNGETIFF:
2284 case TUNSETTXFILTER:
2285 case TUNGETSNDBUF:
2286 case TUNSETSNDBUF:
2287 case SIOCGIFHWADDR:
2288 case SIOCSIFHWADDR:
2289 arg = (unsigned long)compat_ptr(arg);
2290 break;
2291 default:
2292 arg = (compat_ulong_t)arg;
2293 break;
2294 }
2295
2296 /*
2297 * compat_ifreq is shorter than ifreq, so we must not access beyond
2298 * the end of that structure. All fields that are used in this
2299 * driver are compatible though, we don't need to convert the
2300 * contents.
2301 */
2302 return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
2303}
2304#endif /* CONFIG_COMPAT */
2305
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306static int tun_chr_fasync(int fd, struct file *file, int on)
2307{
Jason Wang54f968d2012-10-31 19:45:57 +00002308 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 int ret;
2310
Jason Wang54f968d2012-10-31 19:45:57 +00002311 if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0)
Jonathan Corbet9d319522008-06-19 15:50:37 -06002312 goto out;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002313
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314 if (on) {
Jeff Laytone0b93ed2014-08-22 11:27:32 -04002315 __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
Jason Wang54f968d2012-10-31 19:45:57 +00002316 tfile->flags |= TUN_FASYNC;
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002317 } else
Jason Wang54f968d2012-10-31 19:45:57 +00002318 tfile->flags &= ~TUN_FASYNC;
Jonathan Corbet9d319522008-06-19 15:50:37 -06002319 ret = 0;
2320out:
Jonathan Corbet9d319522008-06-19 15:50:37 -06002321 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002322}
2323
2324static int tun_chr_open(struct inode *inode, struct file * file)
2325{
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002326 struct net *net = current->nsproxy->net_ns;
Eric W. Biederman631ab462009-01-20 11:00:40 +00002327 struct tun_file *tfile;
Thomas Gleixnerdeed49f2009-10-14 01:19:46 -07002328
Joe Perches6b8a66e2011-03-02 07:18:10 +00002329 DBG1(KERN_INFO, "tunX: tun_chr_open\n");
Eric W. Biederman631ab462009-01-20 11:00:40 +00002330
Eric W. Biederman140e807da2015-05-08 21:07:08 -05002331 tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05002332 &tun_proto, 0);
Eric W. Biederman631ab462009-01-20 11:00:40 +00002333 if (!tfile)
2334 return -ENOMEM;
Monam Agarwalc9566742014-03-24 00:02:32 +05302335 RCU_INIT_POINTER(tfile->tun, NULL);
Jason Wang54f968d2012-10-31 19:45:57 +00002336 tfile->flags = 0;
Pavel Emelyanovfb7589a2013-08-21 14:31:38 +04002337 tfile->ifindex = 0;
Jason Wang54f968d2012-10-31 19:45:57 +00002338
Jason Wang54f968d2012-10-31 19:45:57 +00002339 init_waitqueue_head(&tfile->wq.wait);
Xi Wang9e641bd2014-05-16 15:11:48 -07002340 RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq);
Jason Wang54f968d2012-10-31 19:45:57 +00002341
2342 tfile->socket.file = file;
2343 tfile->socket.ops = &tun_socket_ops;
2344
2345 sock_init_data(&tfile->socket, &tfile->sk);
Jason Wang54f968d2012-10-31 19:45:57 +00002346
2347 tfile->sk.sk_write_space = tun_sock_write_space;
2348 tfile->sk.sk_sndbuf = INT_MAX;
2349
Eric W. Biederman631ab462009-01-20 11:00:40 +00002350 file->private_data = tfile;
Jason Wang4008e972012-12-13 23:53:30 +00002351 INIT_LIST_HEAD(&tfile->next);
Jason Wang54f968d2012-10-31 19:45:57 +00002352
Jason Wang19a6afb2013-06-08 14:17:41 +08002353 sock_set_flag(&tfile->sk, SOCK_ZEROCOPY);
2354
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355 return 0;
2356}
2357
2358static int tun_chr_close(struct inode *inode, struct file *file)
2359{
Eric W. Biederman631ab462009-01-20 11:00:40 +00002360 struct tun_file *tfile = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361
Jason Wangc8d68e62012-10-31 19:46:00 +00002362 tun_detach(tfile, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363
2364 return 0;
2365}
2366
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002367#ifdef CONFIG_PROC_FS
Joe Perchesa3816ab2014-09-29 16:08:25 -07002368static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f)
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002369{
2370 struct tun_struct *tun;
2371 struct ifreq ifr;
2372
2373 memset(&ifr, 0, sizeof(ifr));
2374
2375 rtnl_lock();
2376 tun = tun_get(f);
2377 if (tun)
2378 tun_get_iff(current->nsproxy->net_ns, tun, &ifr);
2379 rtnl_unlock();
2380
2381 if (tun)
2382 tun_put(tun);
2383
Joe Perchesa3816ab2014-09-29 16:08:25 -07002384 seq_printf(m, "iff:\t%s\n", ifr.ifr_name);
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002385}
2386#endif
2387
Arjan van de Vend54b1fd2007-02-12 00:55:34 -08002388static const struct file_operations tun_fops = {
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002389 .owner = THIS_MODULE,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 .llseek = no_llseek,
Al Viro9b067032014-11-07 13:52:07 -05002391 .read_iter = tun_chr_read_iter,
Al Virof5ff53b2014-06-19 15:36:49 -04002392 .write_iter = tun_chr_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393 .poll = tun_chr_poll,
Arnd Bergmann50857e22009-11-06 22:52:32 -08002394 .unlocked_ioctl = tun_chr_ioctl,
2395#ifdef CONFIG_COMPAT
2396 .compat_ioctl = tun_chr_compat_ioctl,
2397#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398 .open = tun_chr_open,
2399 .release = tun_chr_close,
Masatake YAMATO93e14b62014-01-29 16:43:31 +09002400 .fasync = tun_chr_fasync,
2401#ifdef CONFIG_PROC_FS
2402 .show_fdinfo = tun_chr_show_fdinfo,
2403#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404};
2405
2406static struct miscdevice tun_miscdev = {
2407 .minor = TUN_MINOR,
2408 .name = "tun",
Kay Sieverse454cea2009-09-18 23:01:12 +02002409 .nodename = "net/tun",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002410 .fops = &tun_fops,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411};
2412
2413/* ethtool interface */
2414
2415static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
2416{
2417 cmd->supported = 0;
2418 cmd->advertising = 0;
David Decotigny70739492011-04-27 18:32:40 +00002419 ethtool_cmd_speed_set(cmd, SPEED_10);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002420 cmd->duplex = DUPLEX_FULL;
2421 cmd->port = PORT_TP;
2422 cmd->phy_address = 0;
2423 cmd->transceiver = XCVR_INTERNAL;
2424 cmd->autoneg = AUTONEG_DISABLE;
2425 cmd->maxtxpkt = 0;
2426 cmd->maxrxpkt = 0;
2427 return 0;
2428}
2429
2430static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
2431{
2432 struct tun_struct *tun = netdev_priv(dev);
2433
Rick Jones33a5ba12011-11-15 14:59:53 +00002434 strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
2435 strlcpy(info->version, DRV_VERSION, sizeof(info->version));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002436
2437 switch (tun->flags & TUN_TYPE_MASK) {
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002438 case IFF_TUN:
Rick Jones33a5ba12011-11-15 14:59:53 +00002439 strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440 break;
Michael S. Tsirkin40630b82014-11-19 15:17:31 +02002441 case IFF_TAP:
Rick Jones33a5ba12011-11-15 14:59:53 +00002442 strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443 break;
2444 }
2445}
2446
2447static u32 tun_get_msglevel(struct net_device *dev)
2448{
2449#ifdef TUN_DEBUG
2450 struct tun_struct *tun = netdev_priv(dev);
2451 return tun->debug;
2452#else
2453 return -EOPNOTSUPP;
2454#endif
2455}
2456
2457static void tun_set_msglevel(struct net_device *dev, u32 value)
2458{
2459#ifdef TUN_DEBUG
2460 struct tun_struct *tun = netdev_priv(dev);
2461 tun->debug = value;
2462#endif
2463}
2464
Jeff Garzik7282d492006-09-13 14:30:00 -04002465static const struct ethtool_ops tun_ethtool_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466 .get_settings = tun_get_settings,
2467 .get_drvinfo = tun_get_drvinfo,
2468 .get_msglevel = tun_get_msglevel,
2469 .set_msglevel = tun_set_msglevel,
Nolan Leakebee31362010-07-27 13:53:43 +00002470 .get_link = ethtool_op_get_link,
Richard Cochraneda29772013-07-19 19:40:10 +02002471 .get_ts_info = ethtool_op_get_ts_info,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472};
2473
Jason Wang1576d982016-06-30 14:45:36 +08002474static int tun_queue_resize(struct tun_struct *tun)
2475{
2476 struct net_device *dev = tun->dev;
2477 struct tun_file *tfile;
2478 struct skb_array **arrays;
2479 int n = tun->numqueues + tun->numdisabled;
2480 int ret, i;
2481
2482 arrays = kmalloc(sizeof *arrays * n, GFP_KERNEL);
2483 if (!arrays)
2484 return -ENOMEM;
2485
2486 for (i = 0; i < tun->numqueues; i++) {
2487 tfile = rtnl_dereference(tun->tfiles[i]);
2488 arrays[i] = &tfile->tx_array;
2489 }
2490 list_for_each_entry(tfile, &tun->disabled, next)
2491 arrays[i++] = &tfile->tx_array;
2492
2493 ret = skb_array_resize_multiple(arrays, n,
2494 dev->tx_queue_len, GFP_KERNEL);
2495
2496 kfree(arrays);
2497 return ret;
2498}
2499
2500static int tun_device_event(struct notifier_block *unused,
2501 unsigned long event, void *ptr)
2502{
2503 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2504 struct tun_struct *tun = netdev_priv(dev);
2505
Craig Gallek86dfb4ac2016-07-06 18:44:20 -04002506 if (dev->rtnl_link_ops != &tun_link_ops)
2507 return NOTIFY_DONE;
2508
Jason Wang1576d982016-06-30 14:45:36 +08002509 switch (event) {
2510 case NETDEV_CHANGE_TX_QUEUE_LEN:
2511 if (tun_queue_resize(tun))
2512 return NOTIFY_BAD;
2513 break;
2514 default:
2515 break;
2516 }
2517
2518 return NOTIFY_DONE;
2519}
2520
2521static struct notifier_block tun_notifier_block __read_mostly = {
2522 .notifier_call = tun_device_event,
2523};
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002524
Linus Torvalds1da177e2005-04-16 15:20:36 -07002525static int __init tun_init(void)
2526{
2527 int ret = 0;
2528
Joe Perches6b8a66e2011-03-02 07:18:10 +00002529 pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
2530 pr_info("%s\n", DRV_COPYRIGHT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002531
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002532 ret = rtnl_link_register(&tun_link_ops);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002533 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002534 pr_err("Can't register link_ops\n");
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002535 goto err_linkops;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002536 }
2537
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538 ret = misc_register(&tun_miscdev);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002539 if (ret) {
Joe Perches6b8a66e2011-03-02 07:18:10 +00002540 pr_err("Can't register misc device %d\n", TUN_MINOR);
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002541 goto err_misc;
2542 }
Jason Wang1576d982016-06-30 14:45:36 +08002543
2544 register_netdevice_notifier(&tun_notifier_block);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002545 return 0;
Pavel Emelyanov79d17602008-04-16 00:40:46 -07002546err_misc:
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002547 rtnl_link_unregister(&tun_link_ops);
2548err_linkops:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549 return ret;
2550}
2551
2552static void tun_cleanup(void)
2553{
Jeff Garzik6aa20a22006-09-13 13:24:59 -04002554 misc_deregister(&tun_miscdev);
Eric W. Biedermanf019a7a2009-01-21 16:02:16 -08002555 rtnl_link_unregister(&tun_link_ops);
Jason Wang1576d982016-06-30 14:45:36 +08002556 unregister_netdevice_notifier(&tun_notifier_block);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557}
2558
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002559/* Get an underlying socket object from tun file. Returns error unless file is
2560 * attached to a device. The returned object works like a packet socket, it
2561 * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for
2562 * holding a reference to the file for as long as the socket is in use. */
2563struct socket *tun_get_socket(struct file *file)
2564{
Jason Wang6e914fc2012-10-31 19:45:58 +00002565 struct tun_file *tfile;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002566 if (file->f_op != &tun_fops)
2567 return ERR_PTR(-EINVAL);
Jason Wang6e914fc2012-10-31 19:45:58 +00002568 tfile = file->private_data;
2569 if (!tfile)
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002570 return ERR_PTR(-EBADFD);
Jason Wang54f968d2012-10-31 19:45:57 +00002571 return &tfile->socket;
Michael S. Tsirkin05c28282010-01-14 06:17:09 +00002572}
2573EXPORT_SYMBOL_GPL(tun_get_socket);
2574
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575module_init(tun_init);
2576module_exit(tun_cleanup);
2577MODULE_DESCRIPTION(DRV_DESCRIPTION);
2578MODULE_AUTHOR(DRV_COPYRIGHT);
2579MODULE_LICENSE("GPL");
2580MODULE_ALIAS_MISCDEV(TUN_MINOR);
Kay Sievers578454f2010-05-20 18:07:20 +02002581MODULE_ALIAS("devname:net/tun");