Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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 | * | |
ff4cc3ac MK |
21 | * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14 |
22 | * Add TUNSETLINK ioctl to set the link encapsulation | |
23 | * | |
1da177e4 | 24 | * Mark Smith <markzzzsmith@yahoo.com.au> |
344dc8ed | 25 | * Use eth_random_addr() for tap MAC address. |
1da177e4 LT |
26 | * |
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 | ||
6b8a66ee JP |
37 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
38 | ||
1da177e4 LT |
39 | #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 | ||
1da177e4 LT |
44 | #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> | |
50857e2a | 58 | #include <linux/compat.h> |
1da177e4 LT |
59 | #include <linux/if.h> |
60 | #include <linux/if_arp.h> | |
61 | #include <linux/if_ether.h> | |
62 | #include <linux/if_tun.h> | |
6680ec68 | 63 | #include <linux/if_vlan.h> |
1da177e4 | 64 | #include <linux/crc32.h> |
d647a591 | 65 | #include <linux/nsproxy.h> |
f43798c2 | 66 | #include <linux/virtio_net.h> |
99405162 | 67 | #include <linux/rcupdate.h> |
881d966b | 68 | #include <net/net_namespace.h> |
79d17604 | 69 | #include <net/netns/generic.h> |
f019a7a5 | 70 | #include <net/rtnetlink.h> |
33dccbb0 | 71 | #include <net/sock.h> |
1da177e4 | 72 | |
1da177e4 LT |
73 | #include <asm/uaccess.h> |
74 | ||
14daa021 RR |
75 | /* Uncomment to enable debugging */ |
76 | /* #define TUN_DEBUG 1 */ | |
77 | ||
1da177e4 LT |
78 | #ifdef TUN_DEBUG |
79 | static int debug; | |
14daa021 | 80 | |
6b8a66ee JP |
81 | #define tun_debug(level, tun, fmt, args...) \ |
82 | do { \ | |
83 | if (tun->debug) \ | |
84 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
85 | } while (0) | |
86 | #define DBG1(level, fmt, args...) \ | |
87 | do { \ | |
88 | if (debug == 2) \ | |
89 | printk(level fmt, ##args); \ | |
90 | } while (0) | |
14daa021 | 91 | #else |
6b8a66ee JP |
92 | #define tun_debug(level, tun, fmt, args...) \ |
93 | do { \ | |
94 | if (0) \ | |
95 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
96 | } while (0) | |
97 | #define DBG1(level, fmt, args...) \ | |
98 | do { \ | |
99 | if (0) \ | |
100 | printk(level fmt, ##args); \ | |
101 | } while (0) | |
14daa021 RR |
102 | #endif |
103 | ||
0690899b MT |
104 | #define GOODCOPY_LEN 128 |
105 | ||
f271b2cc MK |
106 | #define FLT_EXACT_COUNT 8 |
107 | struct tap_filter { | |
108 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
109 | u32 mask[2]; /* Mask of the hashed addrs */ | |
110 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
111 | }; | |
112 | ||
edfb6a14 JW |
113 | /* DEFAULT_MAX_NUM_RSS_QUEUES were choosed to let the rx/tx queues allocated for |
114 | * the netdevice to be fit in one page. So we can make sure the success of | |
115 | * memory allocation. TODO: increase the limit. */ | |
116 | #define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES | |
b8732fb7 | 117 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 118 | |
96442e42 JW |
119 | #define TUN_FLOW_EXPIRE (3 * HZ) |
120 | ||
54f968d6 JW |
121 | /* A tun_file connects an open character device to a tuntap netdevice. It |
122 | * also contains all socket related strctures (except sock_fprog and tap_filter) | |
123 | * to serve as one transmit queue for tuntap device. The sock_fprog and | |
124 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 125 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 126 | * this). |
6e914fc7 JW |
127 | * |
128 | * RCU usage: | |
36fe8c09 | 129 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 130 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 131 | */ |
631ab46b | 132 | struct tun_file { |
54f968d6 JW |
133 | struct sock sk; |
134 | struct socket socket; | |
135 | struct socket_wq wq; | |
6e914fc7 | 136 | struct tun_struct __rcu *tun; |
36b50bab | 137 | struct net *net; |
54f968d6 JW |
138 | struct fasync_struct *fasync; |
139 | /* only used for fasnyc */ | |
140 | unsigned int flags; | |
c8d68e6b | 141 | u16 queue_index; |
4008e97f JW |
142 | struct list_head next; |
143 | struct tun_struct *detached; | |
631ab46b EB |
144 | }; |
145 | ||
96442e42 JW |
146 | struct tun_flow_entry { |
147 | struct hlist_node hash_link; | |
148 | struct rcu_head rcu; | |
149 | struct tun_struct *tun; | |
150 | ||
151 | u32 rxhash; | |
152 | int queue_index; | |
153 | unsigned long updated; | |
154 | }; | |
155 | ||
156 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
157 | ||
54f968d6 | 158 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 159 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
160 | * file were attached to a persist device. |
161 | */ | |
14daa021 | 162 | struct tun_struct { |
c8d68e6b JW |
163 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
164 | unsigned int numqueues; | |
f271b2cc | 165 | unsigned int flags; |
0625c883 EB |
166 | kuid_t owner; |
167 | kgid_t group; | |
14daa021 | 168 | |
14daa021 | 169 | struct net_device *dev; |
c8f44aff | 170 | netdev_features_t set_features; |
88255375 MM |
171 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
172 | NETIF_F_TSO6|NETIF_F_UFO) | |
d9d52b51 MT |
173 | |
174 | int vnet_hdr_sz; | |
54f968d6 JW |
175 | int sndbuf; |
176 | struct tap_filter txflt; | |
177 | struct sock_fprog fprog; | |
178 | /* protected by rtnl lock */ | |
179 | bool filter_attached; | |
14daa021 RR |
180 | #ifdef TUN_DEBUG |
181 | int debug; | |
1da177e4 | 182 | #endif |
96442e42 | 183 | spinlock_t lock; |
96442e42 JW |
184 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
185 | struct timer_list flow_gc_timer; | |
186 | unsigned long ageing_time; | |
4008e97f JW |
187 | unsigned int numdisabled; |
188 | struct list_head disabled; | |
5dbbaf2d | 189 | void *security; |
b8732fb7 | 190 | u32 flow_count; |
14daa021 | 191 | }; |
1da177e4 | 192 | |
96442e42 JW |
193 | static inline u32 tun_hashfn(u32 rxhash) |
194 | { | |
195 | return rxhash & 0x3ff; | |
196 | } | |
197 | ||
198 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
199 | { | |
200 | struct tun_flow_entry *e; | |
96442e42 | 201 | |
b67bfe0d | 202 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
203 | if (e->rxhash == rxhash) |
204 | return e; | |
205 | } | |
206 | return NULL; | |
207 | } | |
208 | ||
209 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
210 | struct hlist_head *head, | |
211 | u32 rxhash, u16 queue_index) | |
212 | { | |
9fdc6bef ED |
213 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
214 | ||
96442e42 JW |
215 | if (e) { |
216 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
217 | rxhash, queue_index); | |
218 | e->updated = jiffies; | |
219 | e->rxhash = rxhash; | |
220 | e->queue_index = queue_index; | |
221 | e->tun = tun; | |
222 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 223 | ++tun->flow_count; |
96442e42 JW |
224 | } |
225 | return e; | |
226 | } | |
227 | ||
96442e42 JW |
228 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
229 | { | |
230 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
231 | e->rxhash, e->queue_index); | |
232 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 233 | kfree_rcu(e, rcu); |
b8732fb7 | 234 | --tun->flow_count; |
96442e42 JW |
235 | } |
236 | ||
237 | static void tun_flow_flush(struct tun_struct *tun) | |
238 | { | |
239 | int i; | |
240 | ||
241 | spin_lock_bh(&tun->lock); | |
242 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
243 | struct tun_flow_entry *e; | |
b67bfe0d | 244 | struct hlist_node *n; |
96442e42 | 245 | |
b67bfe0d | 246 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
247 | tun_flow_delete(tun, e); |
248 | } | |
249 | spin_unlock_bh(&tun->lock); | |
250 | } | |
251 | ||
252 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
253 | { | |
254 | int i; | |
255 | ||
256 | spin_lock_bh(&tun->lock); | |
257 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
258 | struct tun_flow_entry *e; | |
b67bfe0d | 259 | struct hlist_node *n; |
96442e42 | 260 | |
b67bfe0d | 261 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
262 | if (e->queue_index == queue_index) |
263 | tun_flow_delete(tun, e); | |
264 | } | |
265 | } | |
266 | spin_unlock_bh(&tun->lock); | |
267 | } | |
268 | ||
269 | static void tun_flow_cleanup(unsigned long data) | |
270 | { | |
271 | struct tun_struct *tun = (struct tun_struct *)data; | |
272 | unsigned long delay = tun->ageing_time; | |
273 | unsigned long next_timer = jiffies + delay; | |
274 | unsigned long count = 0; | |
275 | int i; | |
276 | ||
277 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
278 | ||
279 | spin_lock_bh(&tun->lock); | |
280 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
281 | struct tun_flow_entry *e; | |
b67bfe0d | 282 | struct hlist_node *n; |
96442e42 | 283 | |
b67bfe0d | 284 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
285 | unsigned long this_timer; |
286 | count++; | |
287 | this_timer = e->updated + delay; | |
288 | if (time_before_eq(this_timer, jiffies)) | |
289 | tun_flow_delete(tun, e); | |
290 | else if (time_before(this_timer, next_timer)) | |
291 | next_timer = this_timer; | |
292 | } | |
293 | } | |
294 | ||
295 | if (count) | |
296 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
297 | spin_unlock_bh(&tun->lock); | |
298 | } | |
299 | ||
49974420 | 300 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 301 | struct tun_file *tfile) |
96442e42 JW |
302 | { |
303 | struct hlist_head *head; | |
304 | struct tun_flow_entry *e; | |
305 | unsigned long delay = tun->ageing_time; | |
9e85722d | 306 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
307 | |
308 | if (!rxhash) | |
309 | return; | |
310 | else | |
311 | head = &tun->flows[tun_hashfn(rxhash)]; | |
312 | ||
313 | rcu_read_lock(); | |
314 | ||
9e85722d JW |
315 | /* We may get a very small possibility of OOO during switching, not |
316 | * worth to optimize.*/ | |
317 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
318 | goto unlock; |
319 | ||
320 | e = tun_flow_find(head, rxhash); | |
321 | if (likely(e)) { | |
322 | /* TODO: keep queueing to old queue until it's empty? */ | |
323 | e->queue_index = queue_index; | |
324 | e->updated = jiffies; | |
325 | } else { | |
326 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
327 | if (!tun_flow_find(head, rxhash) && |
328 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
329 | tun_flow_create(tun, head, rxhash, queue_index); |
330 | ||
331 | if (!timer_pending(&tun->flow_gc_timer)) | |
332 | mod_timer(&tun->flow_gc_timer, | |
333 | round_jiffies_up(jiffies + delay)); | |
334 | spin_unlock_bh(&tun->lock); | |
335 | } | |
336 | ||
337 | unlock: | |
338 | rcu_read_unlock(); | |
339 | } | |
340 | ||
c8d68e6b JW |
341 | /* We try to identify a flow through its rxhash first. The reason that |
342 | * we do not check rxq no. is becuase some cards(e.g 82599), chooses | |
343 | * the rxq based on the txq where the last packet of the flow comes. As | |
344 | * the userspace application move between processors, we may get a | |
345 | * different rxq no. here. If we could not get rxhash, then we would | |
346 | * hope the rxq no. may help here. | |
347 | */ | |
348 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb) | |
349 | { | |
350 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 351 | struct tun_flow_entry *e; |
c8d68e6b JW |
352 | u32 txq = 0; |
353 | u32 numqueues = 0; | |
354 | ||
355 | rcu_read_lock(); | |
92bb73ea | 356 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b JW |
357 | |
358 | txq = skb_get_rxhash(skb); | |
359 | if (txq) { | |
96442e42 JW |
360 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
361 | if (e) | |
362 | txq = e->queue_index; | |
363 | else | |
364 | /* use multiply and shift instead of expensive divide */ | |
365 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
366 | } else if (likely(skb_rx_queue_recorded(skb))) { |
367 | txq = skb_get_rx_queue(skb); | |
368 | while (unlikely(txq >= numqueues)) | |
369 | txq -= numqueues; | |
370 | } | |
371 | ||
372 | rcu_read_unlock(); | |
373 | return txq; | |
374 | } | |
375 | ||
cde8b15f JW |
376 | static inline bool tun_not_capable(struct tun_struct *tun) |
377 | { | |
378 | const struct cred *cred = current_cred(); | |
c260b772 | 379 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
380 | |
381 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
382 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 383 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
384 | } |
385 | ||
c8d68e6b JW |
386 | static void tun_set_real_num_queues(struct tun_struct *tun) |
387 | { | |
388 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
389 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
390 | } | |
391 | ||
4008e97f JW |
392 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
393 | { | |
394 | tfile->detached = tun; | |
395 | list_add_tail(&tfile->next, &tun->disabled); | |
396 | ++tun->numdisabled; | |
397 | } | |
398 | ||
d32649d1 | 399 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
400 | { |
401 | struct tun_struct *tun = tfile->detached; | |
402 | ||
403 | tfile->detached = NULL; | |
404 | list_del_init(&tfile->next); | |
405 | --tun->numdisabled; | |
406 | return tun; | |
407 | } | |
408 | ||
c8d68e6b JW |
409 | static void __tun_detach(struct tun_file *tfile, bool clean) |
410 | { | |
411 | struct tun_file *ntfile; | |
412 | struct tun_struct *tun; | |
c8d68e6b | 413 | |
b8deabd3 JW |
414 | tun = rtnl_dereference(tfile->tun); |
415 | ||
9e85722d | 416 | if (tun && !tfile->detached) { |
c8d68e6b JW |
417 | u16 index = tfile->queue_index; |
418 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
419 | |
420 | rcu_assign_pointer(tun->tfiles[index], | |
421 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 422 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
423 | ntfile->queue_index = index; |
424 | ||
425 | --tun->numqueues; | |
9e85722d JW |
426 | if (clean) { |
427 | rcu_assign_pointer(tfile->tun, NULL); | |
4008e97f | 428 | sock_put(&tfile->sk); |
9e85722d | 429 | } else |
4008e97f | 430 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
431 | |
432 | synchronize_net(); | |
96442e42 | 433 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b JW |
434 | /* Drop read queue */ |
435 | skb_queue_purge(&tfile->sk.sk_receive_queue); | |
436 | tun_set_real_num_queues(tun); | |
dd38bd85 | 437 | } else if (tfile->detached && clean) { |
4008e97f | 438 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
439 | sock_put(&tfile->sk); |
440 | } | |
c8d68e6b JW |
441 | |
442 | if (clean) { | |
af668b3c MT |
443 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
444 | netif_carrier_off(tun->dev); | |
445 | ||
446 | if (!(tun->flags & TUN_PERSIST) && | |
447 | tun->dev->reg_state == NETREG_REGISTERED) | |
4008e97f | 448 | unregister_netdevice(tun->dev); |
af668b3c | 449 | } |
4008e97f | 450 | |
c8d68e6b JW |
451 | BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED, |
452 | &tfile->socket.flags)); | |
453 | sk_release_kernel(&tfile->sk); | |
454 | } | |
455 | } | |
456 | ||
457 | static void tun_detach(struct tun_file *tfile, bool clean) | |
458 | { | |
459 | rtnl_lock(); | |
460 | __tun_detach(tfile, clean); | |
461 | rtnl_unlock(); | |
462 | } | |
463 | ||
464 | static void tun_detach_all(struct net_device *dev) | |
465 | { | |
466 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 467 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
468 | int i, n = tun->numqueues; |
469 | ||
470 | for (i = 0; i < n; i++) { | |
b8deabd3 | 471 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
472 | BUG_ON(!tfile); |
473 | wake_up_all(&tfile->wq.wait); | |
474 | rcu_assign_pointer(tfile->tun, NULL); | |
475 | --tun->numqueues; | |
476 | } | |
9e85722d JW |
477 | list_for_each_entry(tfile, &tun->disabled, next) { |
478 | wake_up_all(&tfile->wq.wait); | |
479 | rcu_assign_pointer(tfile->tun, NULL); | |
480 | } | |
c8d68e6b JW |
481 | BUG_ON(tun->numqueues != 0); |
482 | ||
483 | synchronize_net(); | |
484 | for (i = 0; i < n; i++) { | |
b8deabd3 | 485 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
486 | /* Drop read queue */ |
487 | skb_queue_purge(&tfile->sk.sk_receive_queue); | |
488 | sock_put(&tfile->sk); | |
489 | } | |
4008e97f JW |
490 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
491 | tun_enable_queue(tfile); | |
492 | skb_queue_purge(&tfile->sk.sk_receive_queue); | |
493 | sock_put(&tfile->sk); | |
494 | } | |
495 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 JW |
496 | |
497 | if (tun->flags & TUN_PERSIST) | |
498 | module_put(THIS_MODULE); | |
c8d68e6b JW |
499 | } |
500 | ||
a7385ba2 EB |
501 | static int tun_attach(struct tun_struct *tun, struct file *file) |
502 | { | |
631ab46b | 503 | struct tun_file *tfile = file->private_data; |
38231b7a | 504 | int err; |
a7385ba2 | 505 | |
5dbbaf2d PM |
506 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
507 | if (err < 0) | |
508 | goto out; | |
509 | ||
38231b7a | 510 | err = -EINVAL; |
9e85722d | 511 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
512 | goto out; |
513 | ||
514 | err = -EBUSY; | |
c8d68e6b JW |
515 | if (!(tun->flags & TUN_TAP_MQ) && tun->numqueues == 1) |
516 | goto out; | |
517 | ||
518 | err = -E2BIG; | |
4008e97f JW |
519 | if (!tfile->detached && |
520 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
521 | goto out; |
522 | ||
523 | err = 0; | |
54f968d6 | 524 | |
c8d68e6b | 525 | /* Re-attach the filter to presist device */ |
54f968d6 JW |
526 | if (tun->filter_attached == true) { |
527 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
528 | if (!err) | |
529 | goto out; | |
530 | } | |
c8d68e6b | 531 | tfile->queue_index = tun->numqueues; |
6e914fc7 | 532 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 533 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 534 | tun->numqueues++; |
a7385ba2 | 535 | |
4008e97f JW |
536 | if (tfile->detached) |
537 | tun_enable_queue(tfile); | |
538 | else | |
539 | sock_hold(&tfile->sk); | |
540 | ||
c8d68e6b | 541 | tun_set_real_num_queues(tun); |
a7385ba2 | 542 | |
c8d68e6b JW |
543 | /* device is allowed to go away first, so no need to hold extra |
544 | * refcnt. | |
545 | */ | |
546 | ||
547 | out: | |
548 | return err; | |
631ab46b EB |
549 | } |
550 | ||
551 | static struct tun_struct *__tun_get(struct tun_file *tfile) | |
552 | { | |
6e914fc7 | 553 | struct tun_struct *tun; |
c70f1829 | 554 | |
6e914fc7 JW |
555 | rcu_read_lock(); |
556 | tun = rcu_dereference(tfile->tun); | |
557 | if (tun) | |
558 | dev_hold(tun->dev); | |
559 | rcu_read_unlock(); | |
c70f1829 EB |
560 | |
561 | return tun; | |
631ab46b EB |
562 | } |
563 | ||
564 | static struct tun_struct *tun_get(struct file *file) | |
565 | { | |
566 | return __tun_get(file->private_data); | |
567 | } | |
568 | ||
569 | static void tun_put(struct tun_struct *tun) | |
570 | { | |
6e914fc7 | 571 | dev_put(tun->dev); |
631ab46b EB |
572 | } |
573 | ||
6b8a66ee | 574 | /* TAP filtering */ |
f271b2cc MK |
575 | static void addr_hash_set(u32 *mask, const u8 *addr) |
576 | { | |
577 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
578 | mask[n >> 5] |= (1 << (n & 31)); | |
579 | } | |
580 | ||
581 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
582 | { | |
583 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
584 | return mask[n >> 5] & (1 << (n & 31)); | |
585 | } | |
586 | ||
587 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
588 | { | |
589 | struct { u8 u[ETH_ALEN]; } *addr; | |
590 | struct tun_filter uf; | |
591 | int err, alen, n, nexact; | |
592 | ||
593 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
594 | return -EFAULT; | |
595 | ||
596 | if (!uf.count) { | |
597 | /* Disabled */ | |
598 | filter->count = 0; | |
599 | return 0; | |
600 | } | |
601 | ||
602 | alen = ETH_ALEN * uf.count; | |
603 | addr = kmalloc(alen, GFP_KERNEL); | |
604 | if (!addr) | |
605 | return -ENOMEM; | |
606 | ||
607 | if (copy_from_user(addr, arg + sizeof(uf), alen)) { | |
608 | err = -EFAULT; | |
609 | goto done; | |
610 | } | |
611 | ||
612 | /* The filter is updated without holding any locks. Which is | |
613 | * perfectly safe. We disable it first and in the worst | |
614 | * case we'll accept a few undesired packets. */ | |
615 | filter->count = 0; | |
616 | wmb(); | |
617 | ||
618 | /* Use first set of addresses as an exact filter */ | |
619 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
620 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
621 | ||
622 | nexact = n; | |
623 | ||
cfbf84fc AW |
624 | /* Remaining multicast addresses are hashed, |
625 | * unicast will leave the filter disabled. */ | |
f271b2cc | 626 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
627 | for (; n < uf.count; n++) { |
628 | if (!is_multicast_ether_addr(addr[n].u)) { | |
629 | err = 0; /* no filter */ | |
630 | goto done; | |
631 | } | |
f271b2cc | 632 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 633 | } |
f271b2cc MK |
634 | |
635 | /* For ALLMULTI just set the mask to all ones. | |
636 | * This overrides the mask populated above. */ | |
637 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
638 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
639 | ||
640 | /* Now enable the filter */ | |
641 | wmb(); | |
642 | filter->count = nexact; | |
643 | ||
644 | /* Return the number of exact filters */ | |
645 | err = nexact; | |
646 | ||
647 | done: | |
648 | kfree(addr); | |
649 | return err; | |
650 | } | |
651 | ||
652 | /* Returns: 0 - drop, !=0 - accept */ | |
653 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
654 | { | |
655 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
656 | * at this point. */ | |
657 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
658 | int i; | |
659 | ||
660 | /* Exact match */ | |
661 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 662 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
663 | return 1; |
664 | ||
665 | /* Inexact match (multicast only) */ | |
666 | if (is_multicast_ether_addr(eh->h_dest)) | |
667 | return addr_hash_test(filter->mask, eh->h_dest); | |
668 | ||
669 | return 0; | |
670 | } | |
671 | ||
672 | /* | |
673 | * Checks whether the packet is accepted or not. | |
674 | * Returns: 0 - drop, !=0 - accept | |
675 | */ | |
676 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
677 | { | |
678 | if (!filter->count) | |
679 | return 1; | |
680 | ||
681 | return run_filter(filter, skb); | |
682 | } | |
683 | ||
1da177e4 LT |
684 | /* Network device part of the driver */ |
685 | ||
7282d491 | 686 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 687 | |
c70f1829 EB |
688 | /* Net device detach from fd. */ |
689 | static void tun_net_uninit(struct net_device *dev) | |
690 | { | |
c8d68e6b | 691 | tun_detach_all(dev); |
c70f1829 EB |
692 | } |
693 | ||
1da177e4 LT |
694 | /* Net device open. */ |
695 | static int tun_net_open(struct net_device *dev) | |
696 | { | |
c8d68e6b | 697 | netif_tx_start_all_queues(dev); |
1da177e4 LT |
698 | return 0; |
699 | } | |
700 | ||
701 | /* Net device close. */ | |
702 | static int tun_net_close(struct net_device *dev) | |
703 | { | |
c8d68e6b | 704 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
705 | return 0; |
706 | } | |
707 | ||
708 | /* Net device start xmit */ | |
424efe9c | 709 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
710 | { |
711 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 712 | int txq = skb->queue_mapping; |
6e914fc7 | 713 | struct tun_file *tfile; |
1da177e4 | 714 | |
6e914fc7 | 715 | rcu_read_lock(); |
c8d68e6b JW |
716 | tfile = rcu_dereference(tun->tfiles[txq]); |
717 | ||
1da177e4 | 718 | /* Drop packet if interface is not attached */ |
c8d68e6b | 719 | if (txq >= tun->numqueues) |
1da177e4 LT |
720 | goto drop; |
721 | ||
6e914fc7 JW |
722 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
723 | ||
c8d68e6b JW |
724 | BUG_ON(!tfile); |
725 | ||
f271b2cc MK |
726 | /* Drop if the filter does not like it. |
727 | * This is a noop if the filter is disabled. | |
728 | * Filter can be enabled only for the TAP devices. */ | |
729 | if (!check_filter(&tun->txflt, skb)) | |
730 | goto drop; | |
731 | ||
54f968d6 JW |
732 | if (tfile->socket.sk->sk_filter && |
733 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
734 | goto drop; |
735 | ||
36fe8c09 | 736 | /* Limit the number of packets queued by dividing txq length with the |
c8d68e6b JW |
737 | * number of queues. |
738 | */ | |
54f968d6 | 739 | if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) |
5d097109 MT |
740 | >= dev->tx_queue_len / tun->numqueues) |
741 | goto drop; | |
1da177e4 | 742 | |
eda29772 RC |
743 | if (skb->sk) { |
744 | sock_tx_timestamp(skb->sk, &skb_shinfo(skb)->tx_flags); | |
745 | sw_tx_timestamp(skb); | |
746 | } | |
747 | ||
0110d6f2 MT |
748 | /* Orphan the skb - required as we might hang on to it |
749 | * for indefinite time. */ | |
868eefeb MT |
750 | if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC))) |
751 | goto drop; | |
0110d6f2 MT |
752 | skb_orphan(skb); |
753 | ||
f8af75f3 ED |
754 | nf_reset(skb); |
755 | ||
f271b2cc | 756 | /* Enqueue packet */ |
54f968d6 | 757 | skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb); |
1da177e4 LT |
758 | |
759 | /* Notify and wake up reader process */ | |
54f968d6 JW |
760 | if (tfile->flags & TUN_FASYNC) |
761 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
762 | wake_up_interruptible_poll(&tfile->wq.wait, POLLIN | | |
05c2828c | 763 | POLLRDNORM | POLLRDBAND); |
6e914fc7 JW |
764 | |
765 | rcu_read_unlock(); | |
6ed10654 | 766 | return NETDEV_TX_OK; |
1da177e4 LT |
767 | |
768 | drop: | |
09f75cd7 | 769 | dev->stats.tx_dropped++; |
149d36f7 | 770 | skb_tx_error(skb); |
1da177e4 | 771 | kfree_skb(skb); |
6e914fc7 | 772 | rcu_read_unlock(); |
6ed10654 | 773 | return NETDEV_TX_OK; |
1da177e4 LT |
774 | } |
775 | ||
f271b2cc | 776 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 777 | { |
f271b2cc MK |
778 | /* |
779 | * This callback is supposed to deal with mc filter in | |
780 | * _rx_ path and has nothing to do with the _tx_ path. | |
781 | * In rx path we always accept everything userspace gives us. | |
782 | */ | |
1da177e4 LT |
783 | } |
784 | ||
4885a504 ES |
785 | #define MIN_MTU 68 |
786 | #define MAX_MTU 65535 | |
787 | ||
788 | static int | |
789 | tun_net_change_mtu(struct net_device *dev, int new_mtu) | |
790 | { | |
791 | if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU) | |
792 | return -EINVAL; | |
793 | dev->mtu = new_mtu; | |
794 | return 0; | |
795 | } | |
796 | ||
c8f44aff MM |
797 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
798 | netdev_features_t features) | |
88255375 MM |
799 | { |
800 | struct tun_struct *tun = netdev_priv(dev); | |
801 | ||
802 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
803 | } | |
bebd097a NH |
804 | #ifdef CONFIG_NET_POLL_CONTROLLER |
805 | static void tun_poll_controller(struct net_device *dev) | |
806 | { | |
807 | /* | |
808 | * Tun only receives frames when: | |
809 | * 1) the char device endpoint gets data from user space | |
810 | * 2) the tun socket gets a sendmsg call from user space | |
811 | * Since both of those are syncronous operations, we are guaranteed | |
812 | * never to have pending data when we poll for it | |
813 | * so theres nothing to do here but return. | |
814 | * We need this though so netpoll recognizes us as an interface that | |
815 | * supports polling, which enables bridge devices in virt setups to | |
816 | * still use netconsole | |
817 | */ | |
818 | return; | |
819 | } | |
820 | #endif | |
758e43b7 | 821 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 822 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
823 | .ndo_open = tun_net_open, |
824 | .ndo_stop = tun_net_close, | |
00829823 | 825 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 826 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 827 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 828 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
829 | #ifdef CONFIG_NET_POLL_CONTROLLER |
830 | .ndo_poll_controller = tun_poll_controller, | |
831 | #endif | |
758e43b7 SH |
832 | }; |
833 | ||
834 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 835 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
836 | .ndo_open = tun_net_open, |
837 | .ndo_stop = tun_net_close, | |
00829823 | 838 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 839 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 840 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 841 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
842 | .ndo_set_mac_address = eth_mac_addr, |
843 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 844 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
845 | #ifdef CONFIG_NET_POLL_CONTROLLER |
846 | .ndo_poll_controller = tun_poll_controller, | |
847 | #endif | |
758e43b7 SH |
848 | }; |
849 | ||
944a1376 | 850 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
851 | { |
852 | int i; | |
853 | ||
96442e42 JW |
854 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
855 | INIT_HLIST_HEAD(&tun->flows[i]); | |
856 | ||
857 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
858 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
859 | mod_timer(&tun->flow_gc_timer, | |
860 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
861 | } |
862 | ||
863 | static void tun_flow_uninit(struct tun_struct *tun) | |
864 | { | |
865 | del_timer_sync(&tun->flow_gc_timer); | |
866 | tun_flow_flush(tun); | |
96442e42 JW |
867 | } |
868 | ||
1da177e4 LT |
869 | /* Initialize net device. */ |
870 | static void tun_net_init(struct net_device *dev) | |
871 | { | |
872 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 873 | |
1da177e4 LT |
874 | switch (tun->flags & TUN_TYPE_MASK) { |
875 | case TUN_TUN_DEV: | |
758e43b7 SH |
876 | dev->netdev_ops = &tun_netdev_ops; |
877 | ||
1da177e4 LT |
878 | /* Point-to-Point TUN Device */ |
879 | dev->hard_header_len = 0; | |
880 | dev->addr_len = 0; | |
881 | dev->mtu = 1500; | |
882 | ||
883 | /* Zero header length */ | |
6aa20a22 | 884 | dev->type = ARPHRD_NONE; |
1da177e4 LT |
885 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
886 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ | |
887 | break; | |
888 | ||
889 | case TUN_TAP_DEV: | |
7a0a9608 | 890 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 891 | /* Ethernet TAP Device */ |
1da177e4 | 892 | ether_setup(dev); |
550fd08c | 893 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 894 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 895 | |
f2cedb63 | 896 | eth_hw_addr_random(dev); |
36226a8d | 897 | |
1da177e4 LT |
898 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ |
899 | break; | |
900 | } | |
901 | } | |
902 | ||
903 | /* Character device part */ | |
904 | ||
905 | /* Poll */ | |
c8d68e6b | 906 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 907 | { |
b2430de3 EB |
908 | struct tun_file *tfile = file->private_data; |
909 | struct tun_struct *tun = __tun_get(tfile); | |
3c8a9c63 | 910 | struct sock *sk; |
33dccbb0 | 911 | unsigned int mask = 0; |
1da177e4 LT |
912 | |
913 | if (!tun) | |
eac9e902 | 914 | return POLLERR; |
1da177e4 | 915 | |
54f968d6 | 916 | sk = tfile->socket.sk; |
3c8a9c63 | 917 | |
6b8a66ee | 918 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 919 | |
54f968d6 | 920 | poll_wait(file, &tfile->wq.wait, wait); |
6aa20a22 | 921 | |
89f56d1e | 922 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
923 | mask |= POLLIN | POLLRDNORM; |
924 | ||
33dccbb0 HX |
925 | if (sock_writeable(sk) || |
926 | (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
927 | sock_writeable(sk))) | |
928 | mask |= POLLOUT | POLLWRNORM; | |
929 | ||
c70f1829 EB |
930 | if (tun->dev->reg_state != NETREG_REGISTERED) |
931 | mask = POLLERR; | |
932 | ||
631ab46b | 933 | tun_put(tun); |
1da177e4 LT |
934 | return mask; |
935 | } | |
936 | ||
f42157cb RR |
937 | /* prepad is the amount to reserve at front. len is length after that. |
938 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 939 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 940 | size_t prepad, size_t len, |
941 | size_t linear, int noblock) | |
f42157cb | 942 | { |
54f968d6 | 943 | struct sock *sk = tfile->socket.sk; |
f42157cb | 944 | struct sk_buff *skb; |
33dccbb0 | 945 | int err; |
f42157cb RR |
946 | |
947 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 948 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 949 | linear = len; |
f42157cb | 950 | |
33dccbb0 HX |
951 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
952 | &err); | |
f42157cb | 953 | if (!skb) |
33dccbb0 | 954 | return ERR_PTR(err); |
f42157cb RR |
955 | |
956 | skb_reserve(skb, prepad); | |
957 | skb_put(skb, linear); | |
33dccbb0 HX |
958 | skb->data_len = len - linear; |
959 | skb->len += len - linear; | |
f42157cb RR |
960 | |
961 | return skb; | |
962 | } | |
963 | ||
0690899b MT |
964 | /* set skb frags from iovec, this can move to core network code for reuse */ |
965 | static int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *from, | |
966 | int offset, size_t count) | |
967 | { | |
968 | int len = iov_length(from, count) - offset; | |
969 | int copy = skb_headlen(skb); | |
970 | int size, offset1 = 0; | |
971 | int i = 0; | |
972 | ||
973 | /* Skip over from offset */ | |
974 | while (count && (offset >= from->iov_len)) { | |
975 | offset -= from->iov_len; | |
976 | ++from; | |
977 | --count; | |
978 | } | |
979 | ||
980 | /* copy up to skb headlen */ | |
981 | while (count && (copy > 0)) { | |
982 | size = min_t(unsigned int, copy, from->iov_len - offset); | |
983 | if (copy_from_user(skb->data + offset1, from->iov_base + offset, | |
984 | size)) | |
985 | return -EFAULT; | |
986 | if (copy > size) { | |
987 | ++from; | |
988 | --count; | |
989 | offset = 0; | |
990 | } else | |
991 | offset += size; | |
992 | copy -= size; | |
993 | offset1 += size; | |
994 | } | |
995 | ||
996 | if (len == offset1) | |
997 | return 0; | |
998 | ||
999 | while (count--) { | |
1000 | struct page *page[MAX_SKB_FRAGS]; | |
1001 | int num_pages; | |
1002 | unsigned long base; | |
1003 | unsigned long truesize; | |
1004 | ||
1005 | len = from->iov_len - offset; | |
1006 | if (!len) { | |
1007 | offset = 0; | |
1008 | ++from; | |
1009 | continue; | |
1010 | } | |
1011 | base = (unsigned long)from->iov_base + offset; | |
1012 | size = ((base & ~PAGE_MASK) + len + ~PAGE_MASK) >> PAGE_SHIFT; | |
1013 | if (i + size > MAX_SKB_FRAGS) | |
1014 | return -EMSGSIZE; | |
1015 | num_pages = get_user_pages_fast(base, size, 0, &page[i]); | |
1016 | if (num_pages != size) { | |
7e24bfbe MT |
1017 | int j; |
1018 | ||
1019 | for (j = 0; j < num_pages; j++) | |
1020 | put_page(page[i + j]); | |
0690899b MT |
1021 | return -EFAULT; |
1022 | } | |
1023 | truesize = size * PAGE_SIZE; | |
1024 | skb->data_len += len; | |
1025 | skb->len += len; | |
1026 | skb->truesize += truesize; | |
1027 | atomic_add(truesize, &skb->sk->sk_wmem_alloc); | |
1028 | while (len) { | |
1029 | int off = base & ~PAGE_MASK; | |
1030 | int size = min_t(int, len, PAGE_SIZE - off); | |
1031 | __skb_fill_page_desc(skb, i, page[i], off, size); | |
1032 | skb_shinfo(skb)->nr_frags++; | |
1033 | /* increase sk_wmem_alloc */ | |
1034 | base += size; | |
1035 | len -= size; | |
1036 | i++; | |
1037 | } | |
1038 | offset = 0; | |
1039 | ++from; | |
1040 | } | |
1041 | return 0; | |
1042 | } | |
1043 | ||
1da177e4 | 1044 | /* Get packet from user space buffer */ |
54f968d6 JW |
1045 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
1046 | void *msg_control, const struct iovec *iv, | |
1047 | size_t total_len, size_t count, int noblock) | |
1da177e4 | 1048 | { |
09640e63 | 1049 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1050 | struct sk_buff *skb; |
3dd5c330 | 1051 | size_t len = total_len, align = NET_SKB_PAD, linear; |
f43798c2 | 1052 | struct virtio_net_hdr gso = { 0 }; |
6f26c9a7 | 1053 | int offset = 0; |
0690899b MT |
1054 | int copylen; |
1055 | bool zerocopy = false; | |
1056 | int err; | |
49974420 | 1057 | u32 rxhash; |
1da177e4 LT |
1058 | |
1059 | if (!(tun->flags & TUN_NO_PI)) { | |
0690899b | 1060 | if ((len -= sizeof(pi)) > total_len) |
1da177e4 LT |
1061 | return -EINVAL; |
1062 | ||
6f26c9a7 | 1063 | if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi))) |
1da177e4 | 1064 | return -EFAULT; |
6f26c9a7 | 1065 | offset += sizeof(pi); |
1da177e4 LT |
1066 | } |
1067 | ||
f43798c2 | 1068 | if (tun->flags & TUN_VNET_HDR) { |
0690899b | 1069 | if ((len -= tun->vnet_hdr_sz) > total_len) |
f43798c2 RR |
1070 | return -EINVAL; |
1071 | ||
6f26c9a7 | 1072 | if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso))) |
f43798c2 RR |
1073 | return -EFAULT; |
1074 | ||
4909122f HX |
1075 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
1076 | gso.csum_start + gso.csum_offset + 2 > gso.hdr_len) | |
1077 | gso.hdr_len = gso.csum_start + gso.csum_offset + 2; | |
1078 | ||
f43798c2 RR |
1079 | if (gso.hdr_len > len) |
1080 | return -EINVAL; | |
d9d52b51 | 1081 | offset += tun->vnet_hdr_sz; |
f43798c2 RR |
1082 | } |
1083 | ||
e01bf1c8 | 1084 | if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) { |
a504b86e | 1085 | align += NET_IP_ALIGN; |
0eca93bc HX |
1086 | if (unlikely(len < ETH_HLEN || |
1087 | (gso.hdr_len && gso.hdr_len < ETH_HLEN))) | |
e01bf1c8 RR |
1088 | return -EINVAL; |
1089 | } | |
6aa20a22 | 1090 | |
88529176 JW |
1091 | if (msg_control) { |
1092 | /* There are 256 bytes to be copied in skb, so there is | |
1093 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1094 | * The rest of the buffer is mapped from userspace. |
1095 | */ | |
88529176 | 1096 | copylen = gso.hdr_len ? gso.hdr_len : GOODCOPY_LEN; |
3dd5c330 | 1097 | linear = copylen; |
88529176 JW |
1098 | if (iov_pages(iv, offset + copylen, count) <= MAX_SKB_FRAGS) |
1099 | zerocopy = true; | |
1100 | } | |
1101 | ||
1102 | if (!zerocopy) { | |
0690899b | 1103 | copylen = len; |
3dd5c330 JW |
1104 | linear = gso.hdr_len; |
1105 | } | |
0690899b | 1106 | |
3dd5c330 | 1107 | skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); |
33dccbb0 HX |
1108 | if (IS_ERR(skb)) { |
1109 | if (PTR_ERR(skb) != -EAGAIN) | |
1110 | tun->dev->stats.rx_dropped++; | |
1111 | return PTR_ERR(skb); | |
1da177e4 LT |
1112 | } |
1113 | ||
0690899b MT |
1114 | if (zerocopy) |
1115 | err = zerocopy_sg_from_iovec(skb, iv, offset, count); | |
88529176 | 1116 | else { |
0690899b | 1117 | err = skb_copy_datagram_from_iovec(skb, 0, iv, offset, len); |
88529176 JW |
1118 | if (!err && msg_control) { |
1119 | struct ubuf_info *uarg = msg_control; | |
1120 | uarg->callback(uarg, false); | |
1121 | } | |
1122 | } | |
0690899b MT |
1123 | |
1124 | if (err) { | |
09f75cd7 | 1125 | tun->dev->stats.rx_dropped++; |
8f22757e | 1126 | kfree_skb(skb); |
1da177e4 | 1127 | return -EFAULT; |
8f22757e | 1128 | } |
1da177e4 | 1129 | |
f43798c2 RR |
1130 | if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { |
1131 | if (!skb_partial_csum_set(skb, gso.csum_start, | |
1132 | gso.csum_offset)) { | |
1133 | tun->dev->stats.rx_frame_errors++; | |
1134 | kfree_skb(skb); | |
1135 | return -EINVAL; | |
1136 | } | |
88255375 | 1137 | } |
f43798c2 | 1138 | |
1da177e4 LT |
1139 | switch (tun->flags & TUN_TYPE_MASK) { |
1140 | case TUN_TUN_DEV: | |
f09f7ee2 AWC |
1141 | if (tun->flags & TUN_NO_PI) { |
1142 | switch (skb->data[0] & 0xf0) { | |
1143 | case 0x40: | |
1144 | pi.proto = htons(ETH_P_IP); | |
1145 | break; | |
1146 | case 0x60: | |
1147 | pi.proto = htons(ETH_P_IPV6); | |
1148 | break; | |
1149 | default: | |
1150 | tun->dev->stats.rx_dropped++; | |
1151 | kfree_skb(skb); | |
1152 | return -EINVAL; | |
1153 | } | |
1154 | } | |
1155 | ||
459a98ed | 1156 | skb_reset_mac_header(skb); |
1da177e4 | 1157 | skb->protocol = pi.proto; |
4c13eb66 | 1158 | skb->dev = tun->dev; |
1da177e4 LT |
1159 | break; |
1160 | case TUN_TAP_DEV: | |
1161 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1162 | break; | |
6403eab1 | 1163 | } |
1da177e4 | 1164 | |
f43798c2 RR |
1165 | if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) { |
1166 | pr_debug("GSO!\n"); | |
1167 | switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { | |
1168 | case VIRTIO_NET_HDR_GSO_TCPV4: | |
c9af6db4 | 1169 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; |
f43798c2 RR |
1170 | break; |
1171 | case VIRTIO_NET_HDR_GSO_TCPV6: | |
c9af6db4 | 1172 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; |
f43798c2 | 1173 | break; |
e36aa25a | 1174 | case VIRTIO_NET_HDR_GSO_UDP: |
c9af6db4 | 1175 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
e36aa25a | 1176 | break; |
f43798c2 RR |
1177 | default: |
1178 | tun->dev->stats.rx_frame_errors++; | |
1179 | kfree_skb(skb); | |
1180 | return -EINVAL; | |
1181 | } | |
1182 | ||
1183 | if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN) | |
c9af6db4 | 1184 | skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; |
f43798c2 RR |
1185 | |
1186 | skb_shinfo(skb)->gso_size = gso.gso_size; | |
1187 | if (skb_shinfo(skb)->gso_size == 0) { | |
1188 | tun->dev->stats.rx_frame_errors++; | |
1189 | kfree_skb(skb); | |
1190 | return -EINVAL; | |
1191 | } | |
1192 | ||
1193 | /* Header must be checked, and gso_segs computed. */ | |
1194 | skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY; | |
1195 | skb_shinfo(skb)->gso_segs = 0; | |
1196 | } | |
6aa20a22 | 1197 | |
0690899b MT |
1198 | /* copy skb_ubuf_info for callback when skb has no error */ |
1199 | if (zerocopy) { | |
1200 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1201 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1202 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
0690899b MT |
1203 | } |
1204 | ||
76fe4581 | 1205 | skb_reset_network_header(skb); |
40893fd0 | 1206 | skb_probe_transport_header(skb, 0); |
38502af7 | 1207 | |
49974420 | 1208 | rxhash = skb_get_rxhash(skb); |
1da177e4 | 1209 | netif_rx_ni(skb); |
6aa20a22 | 1210 | |
09f75cd7 JG |
1211 | tun->dev->stats.rx_packets++; |
1212 | tun->dev->stats.rx_bytes += len; | |
1da177e4 | 1213 | |
9e85722d | 1214 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1215 | return total_len; |
6aa20a22 | 1216 | } |
1da177e4 | 1217 | |
ee0b3e67 BP |
1218 | static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv, |
1219 | unsigned long count, loff_t pos) | |
1da177e4 | 1220 | { |
33dccbb0 | 1221 | struct file *file = iocb->ki_filp; |
ab46d779 | 1222 | struct tun_struct *tun = tun_get(file); |
54f968d6 | 1223 | struct tun_file *tfile = file->private_data; |
631ab46b | 1224 | ssize_t result; |
1da177e4 LT |
1225 | |
1226 | if (!tun) | |
1227 | return -EBADFD; | |
1228 | ||
6b8a66ee | 1229 | tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count); |
1da177e4 | 1230 | |
54f968d6 JW |
1231 | result = tun_get_user(tun, tfile, NULL, iv, iov_length(iv, count), |
1232 | count, file->f_flags & O_NONBLOCK); | |
631ab46b EB |
1233 | |
1234 | tun_put(tun); | |
1235 | return result; | |
1da177e4 LT |
1236 | } |
1237 | ||
1da177e4 | 1238 | /* Put packet to the user space buffer */ |
6f7c156c | 1239 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1240 | struct tun_file *tfile, |
6f7c156c | 1241 | struct sk_buff *skb, |
1242 | const struct iovec *iv, int len) | |
1da177e4 LT |
1243 | { |
1244 | struct tun_pi pi = { 0, skb->protocol }; | |
1245 | ssize_t total = 0; | |
6680ec68 | 1246 | int vlan_offset = 0; |
1da177e4 LT |
1247 | |
1248 | if (!(tun->flags & TUN_NO_PI)) { | |
1249 | if ((len -= sizeof(pi)) < 0) | |
1250 | return -EINVAL; | |
1251 | ||
1252 | if (len < skb->len) { | |
1253 | /* Packet will be striped */ | |
1254 | pi.flags |= TUN_PKT_STRIP; | |
1255 | } | |
6aa20a22 | 1256 | |
43b39dcd | 1257 | if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi))) |
1da177e4 LT |
1258 | return -EFAULT; |
1259 | total += sizeof(pi); | |
6aa20a22 | 1260 | } |
1da177e4 | 1261 | |
f43798c2 RR |
1262 | if (tun->flags & TUN_VNET_HDR) { |
1263 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ | |
d9d52b51 | 1264 | if ((len -= tun->vnet_hdr_sz) < 0) |
f43798c2 RR |
1265 | return -EINVAL; |
1266 | ||
1267 | if (skb_is_gso(skb)) { | |
1268 | struct skb_shared_info *sinfo = skb_shinfo(skb); | |
1269 | ||
1270 | /* This is a hint as to how much should be linear. */ | |
1271 | gso.hdr_len = skb_headlen(skb); | |
1272 | gso.gso_size = sinfo->gso_size; | |
1273 | if (sinfo->gso_type & SKB_GSO_TCPV4) | |
1274 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4; | |
1275 | else if (sinfo->gso_type & SKB_GSO_TCPV6) | |
1276 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6; | |
e36aa25a SS |
1277 | else if (sinfo->gso_type & SKB_GSO_UDP) |
1278 | gso.gso_type = VIRTIO_NET_HDR_GSO_UDP; | |
ef3db4a5 | 1279 | else { |
6b8a66ee | 1280 | pr_err("unexpected GSO type: " |
ef3db4a5 MT |
1281 | "0x%x, gso_size %d, hdr_len %d\n", |
1282 | sinfo->gso_type, gso.gso_size, | |
1283 | gso.hdr_len); | |
1284 | print_hex_dump(KERN_ERR, "tun: ", | |
1285 | DUMP_PREFIX_NONE, | |
1286 | 16, 1, skb->head, | |
1287 | min((int)gso.hdr_len, 64), true); | |
1288 | WARN_ON_ONCE(1); | |
1289 | return -EINVAL; | |
1290 | } | |
f43798c2 RR |
1291 | if (sinfo->gso_type & SKB_GSO_TCP_ECN) |
1292 | gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN; | |
1293 | } else | |
1294 | gso.gso_type = VIRTIO_NET_HDR_GSO_NONE; | |
1295 | ||
1296 | if (skb->ip_summed == CHECKSUM_PARTIAL) { | |
1297 | gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; | |
55508d60 | 1298 | gso.csum_start = skb_checksum_start_offset(skb); |
f43798c2 | 1299 | gso.csum_offset = skb->csum_offset; |
10a8d94a JW |
1300 | } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) { |
1301 | gso.flags = VIRTIO_NET_HDR_F_DATA_VALID; | |
f43798c2 RR |
1302 | } /* else everything is zero */ |
1303 | ||
43b39dcd MT |
1304 | if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total, |
1305 | sizeof(gso)))) | |
f43798c2 | 1306 | return -EFAULT; |
d9d52b51 | 1307 | total += tun->vnet_hdr_sz; |
f43798c2 RR |
1308 | } |
1309 | ||
6680ec68 JW |
1310 | if (!vlan_tx_tag_present(skb)) { |
1311 | len = min_t(int, skb->len, len); | |
1312 | } else { | |
1313 | int copy, ret; | |
1314 | struct { | |
1315 | __be16 h_vlan_proto; | |
1316 | __be16 h_vlan_TCI; | |
1317 | } veth; | |
1318 | ||
1319 | veth.h_vlan_proto = skb->vlan_proto; | |
1320 | veth.h_vlan_TCI = htons(vlan_tx_tag_get(skb)); | |
1321 | ||
1322 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
1323 | len = min_t(int, skb->len + VLAN_HLEN, len); | |
1324 | ||
1325 | copy = min_t(int, vlan_offset, len); | |
1326 | ret = skb_copy_datagram_const_iovec(skb, 0, iv, total, copy); | |
1327 | len -= copy; | |
1328 | total += copy; | |
1329 | if (ret || !len) | |
1330 | goto done; | |
1331 | ||
1332 | copy = min_t(int, sizeof(veth), len); | |
1333 | ret = memcpy_toiovecend(iv, (void *)&veth, total, copy); | |
1334 | len -= copy; | |
1335 | total += copy; | |
1336 | if (ret || !len) | |
1337 | goto done; | |
1338 | } | |
1da177e4 | 1339 | |
6680ec68 JW |
1340 | skb_copy_datagram_const_iovec(skb, vlan_offset, iv, total, len); |
1341 | total += len; | |
1da177e4 | 1342 | |
6680ec68 | 1343 | done: |
09f75cd7 JG |
1344 | tun->dev->stats.tx_packets++; |
1345 | tun->dev->stats.tx_bytes += len; | |
1da177e4 LT |
1346 | |
1347 | return total; | |
1348 | } | |
1349 | ||
54f968d6 | 1350 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
05c2828c MT |
1351 | struct kiocb *iocb, const struct iovec *iv, |
1352 | ssize_t len, int noblock) | |
1da177e4 | 1353 | { |
1da177e4 LT |
1354 | DECLARE_WAITQUEUE(wait, current); |
1355 | struct sk_buff *skb; | |
05c2828c | 1356 | ssize_t ret = 0; |
1da177e4 | 1357 | |
3872baf6 | 1358 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1359 | |
61a5ff15 | 1360 | if (unlikely(!noblock)) |
54f968d6 | 1361 | add_wait_queue(&tfile->wq.wait, &wait); |
1da177e4 | 1362 | while (len) { |
1da177e4 LT |
1363 | current->state = TASK_INTERRUPTIBLE; |
1364 | ||
1365 | /* Read frames from the queue */ | |
54f968d6 | 1366 | if (!(skb = skb_dequeue(&tfile->socket.sk->sk_receive_queue))) { |
05c2828c | 1367 | if (noblock) { |
1da177e4 LT |
1368 | ret = -EAGAIN; |
1369 | break; | |
1370 | } | |
1371 | if (signal_pending(current)) { | |
1372 | ret = -ERESTARTSYS; | |
1373 | break; | |
1374 | } | |
c70f1829 EB |
1375 | if (tun->dev->reg_state != NETREG_REGISTERED) { |
1376 | ret = -EIO; | |
1377 | break; | |
1378 | } | |
1da177e4 LT |
1379 | |
1380 | /* Nothing to read, let's sleep */ | |
1381 | schedule(); | |
1382 | continue; | |
1383 | } | |
1da177e4 | 1384 | |
54f968d6 | 1385 | ret = tun_put_user(tun, tfile, skb, iv, len); |
f271b2cc MK |
1386 | kfree_skb(skb); |
1387 | break; | |
1da177e4 LT |
1388 | } |
1389 | ||
1390 | current->state = TASK_RUNNING; | |
61a5ff15 | 1391 | if (unlikely(!noblock)) |
54f968d6 | 1392 | remove_wait_queue(&tfile->wq.wait, &wait); |
1da177e4 | 1393 | |
05c2828c MT |
1394 | return ret; |
1395 | } | |
1396 | ||
1397 | static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv, | |
1398 | unsigned long count, loff_t pos) | |
1399 | { | |
1400 | struct file *file = iocb->ki_filp; | |
1401 | struct tun_file *tfile = file->private_data; | |
1402 | struct tun_struct *tun = __tun_get(tfile); | |
1403 | ssize_t len, ret; | |
1404 | ||
1405 | if (!tun) | |
1406 | return -EBADFD; | |
1407 | len = iov_length(iv, count); | |
1408 | if (len < 0) { | |
1409 | ret = -EINVAL; | |
1410 | goto out; | |
1411 | } | |
1412 | ||
54f968d6 JW |
1413 | ret = tun_do_read(tun, tfile, iocb, iv, len, |
1414 | file->f_flags & O_NONBLOCK); | |
05c2828c | 1415 | ret = min_t(ssize_t, ret, len); |
631ab46b EB |
1416 | out: |
1417 | tun_put(tun); | |
1da177e4 LT |
1418 | return ret; |
1419 | } | |
1420 | ||
96442e42 JW |
1421 | static void tun_free_netdev(struct net_device *dev) |
1422 | { | |
1423 | struct tun_struct *tun = netdev_priv(dev); | |
1424 | ||
4008e97f | 1425 | BUG_ON(!(list_empty(&tun->disabled))); |
96442e42 | 1426 | tun_flow_uninit(tun); |
5dbbaf2d | 1427 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1428 | free_netdev(dev); |
1429 | } | |
1430 | ||
1da177e4 LT |
1431 | static void tun_setup(struct net_device *dev) |
1432 | { | |
1433 | struct tun_struct *tun = netdev_priv(dev); | |
1434 | ||
0625c883 EB |
1435 | tun->owner = INVALID_UID; |
1436 | tun->group = INVALID_GID; | |
1da177e4 | 1437 | |
1da177e4 | 1438 | dev->ethtool_ops = &tun_ethtool_ops; |
96442e42 | 1439 | dev->destructor = tun_free_netdev; |
1da177e4 LT |
1440 | } |
1441 | ||
f019a7a5 EB |
1442 | /* Trivial set of netlink ops to allow deleting tun or tap |
1443 | * device with netlink. | |
1444 | */ | |
1445 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1446 | { | |
1447 | return -EINVAL; | |
1448 | } | |
1449 | ||
1450 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
1451 | .kind = DRV_NAME, | |
1452 | .priv_size = sizeof(struct tun_struct), | |
1453 | .setup = tun_setup, | |
1454 | .validate = tun_validate, | |
1455 | }; | |
1456 | ||
33dccbb0 HX |
1457 | static void tun_sock_write_space(struct sock *sk) |
1458 | { | |
54f968d6 | 1459 | struct tun_file *tfile; |
43815482 | 1460 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
1461 | |
1462 | if (!sock_writeable(sk)) | |
1463 | return; | |
1464 | ||
33dccbb0 HX |
1465 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
1466 | return; | |
1467 | ||
43815482 ED |
1468 | wqueue = sk_sleep(sk); |
1469 | if (wqueue && waitqueue_active(wqueue)) | |
1470 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 1471 | POLLWRNORM | POLLWRBAND); |
c722c625 | 1472 | |
54f968d6 JW |
1473 | tfile = container_of(sk, struct tun_file, sk); |
1474 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
1475 | } |
1476 | ||
05c2828c MT |
1477 | static int tun_sendmsg(struct kiocb *iocb, struct socket *sock, |
1478 | struct msghdr *m, size_t total_len) | |
1479 | { | |
54f968d6 JW |
1480 | int ret; |
1481 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1482 | struct tun_struct *tun = __tun_get(tfile); | |
1483 | ||
1484 | if (!tun) | |
1485 | return -EBADFD; | |
54f968d6 JW |
1486 | ret = tun_get_user(tun, tfile, m->msg_control, m->msg_iov, total_len, |
1487 | m->msg_iovlen, m->msg_flags & MSG_DONTWAIT); | |
1488 | tun_put(tun); | |
1489 | return ret; | |
05c2828c MT |
1490 | } |
1491 | ||
1492 | static int tun_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1493 | struct msghdr *m, size_t total_len, | |
1494 | int flags) | |
1495 | { | |
54f968d6 JW |
1496 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
1497 | struct tun_struct *tun = __tun_get(tfile); | |
05c2828c | 1498 | int ret; |
54f968d6 JW |
1499 | |
1500 | if (!tun) | |
1501 | return -EBADFD; | |
1502 | ||
eda29772 | 1503 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
1504 | ret = -EINVAL; |
1505 | goto out; | |
1506 | } | |
eda29772 RC |
1507 | if (flags & MSG_ERRQUEUE) { |
1508 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
1509 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
1510 | goto out; | |
1511 | } | |
54f968d6 | 1512 | ret = tun_do_read(tun, tfile, iocb, m->msg_iov, total_len, |
05c2828c MT |
1513 | flags & MSG_DONTWAIT); |
1514 | if (ret > total_len) { | |
1515 | m->msg_flags |= MSG_TRUNC; | |
1516 | ret = flags & MSG_TRUNC ? ret : total_len; | |
1517 | } | |
3811ae76 | 1518 | out: |
54f968d6 | 1519 | tun_put(tun); |
05c2828c MT |
1520 | return ret; |
1521 | } | |
1522 | ||
1ab5ecb9 SK |
1523 | static int tun_release(struct socket *sock) |
1524 | { | |
1525 | if (sock->sk) | |
1526 | sock_put(sock->sk); | |
1527 | return 0; | |
1528 | } | |
1529 | ||
05c2828c MT |
1530 | /* Ops structure to mimic raw sockets with tun */ |
1531 | static const struct proto_ops tun_socket_ops = { | |
1532 | .sendmsg = tun_sendmsg, | |
1533 | .recvmsg = tun_recvmsg, | |
1ab5ecb9 | 1534 | .release = tun_release, |
05c2828c MT |
1535 | }; |
1536 | ||
33dccbb0 HX |
1537 | static struct proto tun_proto = { |
1538 | .name = "tun", | |
1539 | .owner = THIS_MODULE, | |
54f968d6 | 1540 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 1541 | }; |
f019a7a5 | 1542 | |
980c9e8c DW |
1543 | static int tun_flags(struct tun_struct *tun) |
1544 | { | |
1545 | int flags = 0; | |
1546 | ||
1547 | if (tun->flags & TUN_TUN_DEV) | |
1548 | flags |= IFF_TUN; | |
1549 | else | |
1550 | flags |= IFF_TAP; | |
1551 | ||
1552 | if (tun->flags & TUN_NO_PI) | |
1553 | flags |= IFF_NO_PI; | |
1554 | ||
5d097109 MT |
1555 | /* This flag has no real effect. We track the value for backwards |
1556 | * compatibility. | |
1557 | */ | |
980c9e8c DW |
1558 | if (tun->flags & TUN_ONE_QUEUE) |
1559 | flags |= IFF_ONE_QUEUE; | |
1560 | ||
1561 | if (tun->flags & TUN_VNET_HDR) | |
1562 | flags |= IFF_VNET_HDR; | |
1563 | ||
c8d68e6b JW |
1564 | if (tun->flags & TUN_TAP_MQ) |
1565 | flags |= IFF_MULTI_QUEUE; | |
1566 | ||
274038f8 PE |
1567 | if (tun->flags & TUN_PERSIST) |
1568 | flags |= IFF_PERSIST; | |
1569 | ||
980c9e8c DW |
1570 | return flags; |
1571 | } | |
1572 | ||
1573 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
1574 | char *buf) | |
1575 | { | |
1576 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
1577 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
1578 | } | |
1579 | ||
1580 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
1581 | char *buf) | |
1582 | { | |
1583 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1584 | return uid_valid(tun->owner)? |
1585 | sprintf(buf, "%u\n", | |
1586 | from_kuid_munged(current_user_ns(), tun->owner)): | |
1587 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1588 | } |
1589 | ||
1590 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
1591 | char *buf) | |
1592 | { | |
1593 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1594 | return gid_valid(tun->group) ? |
1595 | sprintf(buf, "%u\n", | |
1596 | from_kgid_munged(current_user_ns(), tun->group)): | |
1597 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1598 | } |
1599 | ||
1600 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
1601 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
1602 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
1603 | ||
d647a591 | 1604 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
1605 | { |
1606 | struct tun_struct *tun; | |
54f968d6 | 1607 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
1608 | struct net_device *dev; |
1609 | int err; | |
1610 | ||
7c0c3b1a JW |
1611 | if (tfile->detached) |
1612 | return -EINVAL; | |
1613 | ||
74a3e5a7 EB |
1614 | dev = __dev_get_by_name(net, ifr->ifr_name); |
1615 | if (dev) { | |
f85ba780 DW |
1616 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
1617 | return -EBUSY; | |
74a3e5a7 EB |
1618 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
1619 | tun = netdev_priv(dev); | |
1620 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
1621 | tun = netdev_priv(dev); | |
1622 | else | |
1623 | return -EINVAL; | |
1624 | ||
8e6d91ae JW |
1625 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
1626 | !!(tun->flags & TUN_TAP_MQ)) | |
1627 | return -EINVAL; | |
1628 | ||
cde8b15f | 1629 | if (tun_not_capable(tun)) |
2b980dbd | 1630 | return -EPERM; |
5dbbaf2d | 1631 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
1632 | if (err < 0) |
1633 | return err; | |
1634 | ||
a7385ba2 EB |
1635 | err = tun_attach(tun, file); |
1636 | if (err < 0) | |
1637 | return err; | |
4008e97f JW |
1638 | |
1639 | if (tun->flags & TUN_TAP_MQ && | |
e8dbad66 JW |
1640 | (tun->numqueues + tun->numdisabled > 1)) { |
1641 | /* One or more queue has already been attached, no need | |
1642 | * to initialize the device again. | |
1643 | */ | |
1644 | return 0; | |
1645 | } | |
6aa20a22 | 1646 | } |
1da177e4 LT |
1647 | else { |
1648 | char *name; | |
1649 | unsigned long flags = 0; | |
edfb6a14 JW |
1650 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
1651 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 1652 | |
c260b772 | 1653 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 1654 | return -EPERM; |
2b980dbd PM |
1655 | err = security_tun_dev_create(); |
1656 | if (err < 0) | |
1657 | return err; | |
ca6bb5d7 | 1658 | |
1da177e4 LT |
1659 | /* Set dev type */ |
1660 | if (ifr->ifr_flags & IFF_TUN) { | |
1661 | /* TUN device */ | |
1662 | flags |= TUN_TUN_DEV; | |
1663 | name = "tun%d"; | |
1664 | } else if (ifr->ifr_flags & IFF_TAP) { | |
1665 | /* TAP device */ | |
1666 | flags |= TUN_TAP_DEV; | |
1667 | name = "tap%d"; | |
6aa20a22 | 1668 | } else |
36989b90 | 1669 | return -EINVAL; |
6aa20a22 | 1670 | |
1da177e4 LT |
1671 | if (*ifr->ifr_name) |
1672 | name = ifr->ifr_name; | |
1673 | ||
c8d68e6b | 1674 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
edfb6a14 JW |
1675 | tun_setup, queues, queues); |
1676 | ||
1da177e4 LT |
1677 | if (!dev) |
1678 | return -ENOMEM; | |
1679 | ||
fc54c658 | 1680 | dev_net_set(dev, net); |
f019a7a5 | 1681 | dev->rtnl_link_ops = &tun_link_ops; |
758e43b7 | 1682 | |
1da177e4 LT |
1683 | tun = netdev_priv(dev); |
1684 | tun->dev = dev; | |
1685 | tun->flags = flags; | |
f271b2cc | 1686 | tun->txflt.count = 0; |
d9d52b51 | 1687 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 1688 | |
54f968d6 JW |
1689 | tun->filter_attached = false; |
1690 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
33dccbb0 | 1691 | |
96442e42 JW |
1692 | spin_lock_init(&tun->lock); |
1693 | ||
5dbbaf2d PM |
1694 | err = security_tun_dev_alloc_security(&tun->security); |
1695 | if (err < 0) | |
1696 | goto err_free_dev; | |
2b980dbd | 1697 | |
1da177e4 | 1698 | tun_net_init(dev); |
944a1376 | 1699 | tun_flow_init(tun); |
96442e42 | 1700 | |
88255375 | 1701 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
1702 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
1703 | NETIF_F_HW_VLAN_STAG_TX; | |
88255375 | 1704 | dev->features = dev->hw_features; |
c0317998 | 1705 | dev->vlan_features = dev->features; |
88255375 | 1706 | |
4008e97f | 1707 | INIT_LIST_HEAD(&tun->disabled); |
eb0fb363 JW |
1708 | err = tun_attach(tun, file); |
1709 | if (err < 0) | |
1710 | goto err_free_dev; | |
1711 | ||
1da177e4 LT |
1712 | err = register_netdevice(tun->dev); |
1713 | if (err < 0) | |
54f968d6 | 1714 | goto err_free_dev; |
9c3fea6a | 1715 | |
980c9e8c DW |
1716 | if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) || |
1717 | device_create_file(&tun->dev->dev, &dev_attr_owner) || | |
1718 | device_create_file(&tun->dev->dev, &dev_attr_group)) | |
6b8a66ee | 1719 | pr_err("Failed to create tun sysfs files\n"); |
1da177e4 LT |
1720 | } |
1721 | ||
af668b3c MT |
1722 | netif_carrier_on(tun->dev); |
1723 | ||
6b8a66ee | 1724 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 LT |
1725 | |
1726 | if (ifr->ifr_flags & IFF_NO_PI) | |
1727 | tun->flags |= TUN_NO_PI; | |
a26af1e0 NF |
1728 | else |
1729 | tun->flags &= ~TUN_NO_PI; | |
1da177e4 | 1730 | |
5d097109 MT |
1731 | /* This flag has no real effect. We track the value for backwards |
1732 | * compatibility. | |
1733 | */ | |
1da177e4 LT |
1734 | if (ifr->ifr_flags & IFF_ONE_QUEUE) |
1735 | tun->flags |= TUN_ONE_QUEUE; | |
a26af1e0 NF |
1736 | else |
1737 | tun->flags &= ~TUN_ONE_QUEUE; | |
1da177e4 | 1738 | |
f43798c2 RR |
1739 | if (ifr->ifr_flags & IFF_VNET_HDR) |
1740 | tun->flags |= TUN_VNET_HDR; | |
1741 | else | |
1742 | tun->flags &= ~TUN_VNET_HDR; | |
1743 | ||
c8d68e6b JW |
1744 | if (ifr->ifr_flags & IFF_MULTI_QUEUE) |
1745 | tun->flags |= TUN_TAP_MQ; | |
1746 | else | |
1747 | tun->flags &= ~TUN_TAP_MQ; | |
1748 | ||
e35259a9 MK |
1749 | /* Make sure persistent devices do not get stuck in |
1750 | * xoff state. | |
1751 | */ | |
1752 | if (netif_running(tun->dev)) | |
c8d68e6b | 1753 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 1754 | |
1da177e4 LT |
1755 | strcpy(ifr->ifr_name, tun->dev->name); |
1756 | return 0; | |
1757 | ||
1758 | err_free_dev: | |
1759 | free_netdev(dev); | |
1da177e4 LT |
1760 | return err; |
1761 | } | |
1762 | ||
9ce99cf6 | 1763 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 1764 | struct ifreq *ifr) |
e3b99556 | 1765 | { |
6b8a66ee | 1766 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
1767 | |
1768 | strcpy(ifr->ifr_name, tun->dev->name); | |
1769 | ||
980c9e8c | 1770 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 1771 | |
e3b99556 MM |
1772 | } |
1773 | ||
5228ddc9 RR |
1774 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
1775 | * privs required. */ | |
88255375 | 1776 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 1777 | { |
c8f44aff | 1778 | netdev_features_t features = 0; |
5228ddc9 RR |
1779 | |
1780 | if (arg & TUN_F_CSUM) { | |
88255375 | 1781 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
1782 | arg &= ~TUN_F_CSUM; |
1783 | ||
1784 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
1785 | if (arg & TUN_F_TSO_ECN) { | |
1786 | features |= NETIF_F_TSO_ECN; | |
1787 | arg &= ~TUN_F_TSO_ECN; | |
1788 | } | |
1789 | if (arg & TUN_F_TSO4) | |
1790 | features |= NETIF_F_TSO; | |
1791 | if (arg & TUN_F_TSO6) | |
1792 | features |= NETIF_F_TSO6; | |
1793 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
1794 | } | |
e36aa25a SS |
1795 | |
1796 | if (arg & TUN_F_UFO) { | |
1797 | features |= NETIF_F_UFO; | |
1798 | arg &= ~TUN_F_UFO; | |
1799 | } | |
5228ddc9 RR |
1800 | } |
1801 | ||
1802 | /* This gives the user a way to test for new features in future by | |
1803 | * trying to set them. */ | |
1804 | if (arg) | |
1805 | return -EINVAL; | |
1806 | ||
88255375 MM |
1807 | tun->set_features = features; |
1808 | netdev_update_features(tun->dev); | |
5228ddc9 RR |
1809 | |
1810 | return 0; | |
1811 | } | |
1812 | ||
c8d68e6b JW |
1813 | static void tun_detach_filter(struct tun_struct *tun, int n) |
1814 | { | |
1815 | int i; | |
1816 | struct tun_file *tfile; | |
1817 | ||
1818 | for (i = 0; i < n; i++) { | |
b8deabd3 | 1819 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1820 | sk_detach_filter(tfile->socket.sk); |
1821 | } | |
1822 | ||
1823 | tun->filter_attached = false; | |
1824 | } | |
1825 | ||
1826 | static int tun_attach_filter(struct tun_struct *tun) | |
1827 | { | |
1828 | int i, ret = 0; | |
1829 | struct tun_file *tfile; | |
1830 | ||
1831 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1832 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1833 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); |
1834 | if (ret) { | |
1835 | tun_detach_filter(tun, i); | |
1836 | return ret; | |
1837 | } | |
1838 | } | |
1839 | ||
1840 | tun->filter_attached = true; | |
1841 | return ret; | |
1842 | } | |
1843 | ||
1844 | static void tun_set_sndbuf(struct tun_struct *tun) | |
1845 | { | |
1846 | struct tun_file *tfile; | |
1847 | int i; | |
1848 | ||
1849 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1850 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1851 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
1852 | } | |
1853 | } | |
1854 | ||
cde8b15f JW |
1855 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
1856 | { | |
1857 | struct tun_file *tfile = file->private_data; | |
1858 | struct tun_struct *tun; | |
cde8b15f JW |
1859 | int ret = 0; |
1860 | ||
1861 | rtnl_lock(); | |
1862 | ||
1863 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 1864 | tun = tfile->detached; |
5dbbaf2d | 1865 | if (!tun) { |
cde8b15f | 1866 | ret = -EINVAL; |
5dbbaf2d PM |
1867 | goto unlock; |
1868 | } | |
1869 | ret = security_tun_dev_attach_queue(tun->security); | |
1870 | if (ret < 0) | |
1871 | goto unlock; | |
1872 | ret = tun_attach(tun, file); | |
4008e97f | 1873 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 1874 | tun = rtnl_dereference(tfile->tun); |
9e85722d | 1875 | if (!tun || !(tun->flags & TUN_TAP_MQ) || tfile->detached) |
4008e97f JW |
1876 | ret = -EINVAL; |
1877 | else | |
1878 | __tun_detach(tfile, false); | |
1879 | } else | |
cde8b15f JW |
1880 | ret = -EINVAL; |
1881 | ||
5dbbaf2d | 1882 | unlock: |
cde8b15f JW |
1883 | rtnl_unlock(); |
1884 | return ret; | |
1885 | } | |
1886 | ||
50857e2a AB |
1887 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
1888 | unsigned long arg, int ifreq_len) | |
1da177e4 | 1889 | { |
36b50bab | 1890 | struct tun_file *tfile = file->private_data; |
631ab46b | 1891 | struct tun_struct *tun; |
1da177e4 LT |
1892 | void __user* argp = (void __user*)arg; |
1893 | struct ifreq ifr; | |
0625c883 EB |
1894 | kuid_t owner; |
1895 | kgid_t group; | |
33dccbb0 | 1896 | int sndbuf; |
d9d52b51 | 1897 | int vnet_hdr_sz; |
f271b2cc | 1898 | int ret; |
1da177e4 | 1899 | |
cde8b15f | 1900 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) { |
50857e2a | 1901 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 1902 | return -EFAULT; |
8bbb1813 | 1903 | } else { |
a117dacd | 1904 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 1905 | } |
631ab46b EB |
1906 | if (cmd == TUNGETFEATURES) { |
1907 | /* Currently this just means: "what IFF flags are valid?". | |
1908 | * This is needed because we never checked for invalid flags on | |
1909 | * TUNSETIFF. */ | |
1910 | return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE | | |
cde8b15f | 1911 | IFF_VNET_HDR | IFF_MULTI_QUEUE, |
631ab46b | 1912 | (unsigned int __user*)argp); |
cde8b15f JW |
1913 | } else if (cmd == TUNSETQUEUE) |
1914 | return tun_set_queue(file, &ifr); | |
631ab46b | 1915 | |
c8d68e6b | 1916 | ret = 0; |
876bfd4d HX |
1917 | rtnl_lock(); |
1918 | ||
36b50bab | 1919 | tun = __tun_get(tfile); |
1da177e4 | 1920 | if (cmd == TUNSETIFF && !tun) { |
1da177e4 LT |
1921 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
1922 | ||
876bfd4d | 1923 | ret = tun_set_iff(tfile->net, file, &ifr); |
1da177e4 | 1924 | |
876bfd4d HX |
1925 | if (ret) |
1926 | goto unlock; | |
1da177e4 | 1927 | |
50857e2a | 1928 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
1929 | ret = -EFAULT; |
1930 | goto unlock; | |
1da177e4 LT |
1931 | } |
1932 | ||
876bfd4d | 1933 | ret = -EBADFD; |
1da177e4 | 1934 | if (!tun) |
876bfd4d | 1935 | goto unlock; |
1da177e4 | 1936 | |
1e588338 | 1937 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 1938 | |
631ab46b | 1939 | ret = 0; |
1da177e4 | 1940 | switch (cmd) { |
e3b99556 | 1941 | case TUNGETIFF: |
9ce99cf6 | 1942 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 1943 | |
50857e2a | 1944 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 1945 | ret = -EFAULT; |
e3b99556 MM |
1946 | break; |
1947 | ||
1da177e4 LT |
1948 | case TUNSETNOCSUM: |
1949 | /* Disable/Enable checksum */ | |
1da177e4 | 1950 | |
88255375 MM |
1951 | /* [unimplemented] */ |
1952 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 1953 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
1954 | break; |
1955 | ||
1956 | case TUNSETPERSIST: | |
54f968d6 JW |
1957 | /* Disable/Enable persist mode. Keep an extra reference to the |
1958 | * module to prevent the module being unprobed. | |
1959 | */ | |
dd38bd85 | 1960 | if (arg && !(tun->flags & TUN_PERSIST)) { |
1da177e4 | 1961 | tun->flags |= TUN_PERSIST; |
54f968d6 | 1962 | __module_get(THIS_MODULE); |
dd38bd85 JW |
1963 | } |
1964 | if (!arg && (tun->flags & TUN_PERSIST)) { | |
1da177e4 | 1965 | tun->flags &= ~TUN_PERSIST; |
54f968d6 JW |
1966 | module_put(THIS_MODULE); |
1967 | } | |
1da177e4 | 1968 | |
6b8a66ee JP |
1969 | tun_debug(KERN_INFO, tun, "persist %s\n", |
1970 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
1971 | break; |
1972 | ||
1973 | case TUNSETOWNER: | |
1974 | /* Set owner of the device */ | |
0625c883 EB |
1975 | owner = make_kuid(current_user_ns(), arg); |
1976 | if (!uid_valid(owner)) { | |
1977 | ret = -EINVAL; | |
1978 | break; | |
1979 | } | |
1980 | tun->owner = owner; | |
1e588338 | 1981 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 1982 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
1983 | break; |
1984 | ||
8c644623 GG |
1985 | case TUNSETGROUP: |
1986 | /* Set group of the device */ | |
0625c883 EB |
1987 | group = make_kgid(current_user_ns(), arg); |
1988 | if (!gid_valid(group)) { | |
1989 | ret = -EINVAL; | |
1990 | break; | |
1991 | } | |
1992 | tun->group = group; | |
1e588338 | 1993 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 1994 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
1995 | break; |
1996 | ||
ff4cc3ac MK |
1997 | case TUNSETLINK: |
1998 | /* Only allow setting the type when the interface is down */ | |
1999 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
2000 | tun_debug(KERN_INFO, tun, |
2001 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 2002 | ret = -EBUSY; |
ff4cc3ac MK |
2003 | } else { |
2004 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2005 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2006 | tun->dev->type); | |
48abfe05 | 2007 | ret = 0; |
ff4cc3ac | 2008 | } |
631ab46b | 2009 | break; |
ff4cc3ac | 2010 | |
1da177e4 LT |
2011 | #ifdef TUN_DEBUG |
2012 | case TUNSETDEBUG: | |
2013 | tun->debug = arg; | |
2014 | break; | |
2015 | #endif | |
5228ddc9 | 2016 | case TUNSETOFFLOAD: |
88255375 | 2017 | ret = set_offload(tun, arg); |
631ab46b | 2018 | break; |
5228ddc9 | 2019 | |
f271b2cc MK |
2020 | case TUNSETTXFILTER: |
2021 | /* Can be set only for TAPs */ | |
631ab46b | 2022 | ret = -EINVAL; |
f271b2cc | 2023 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) |
631ab46b | 2024 | break; |
c0e5a8c2 | 2025 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2026 | break; |
1da177e4 LT |
2027 | |
2028 | case SIOCGIFHWADDR: | |
b595076a | 2029 | /* Get hw address */ |
f271b2cc MK |
2030 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2031 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2032 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2033 | ret = -EFAULT; |
2034 | break; | |
1da177e4 LT |
2035 | |
2036 | case SIOCSIFHWADDR: | |
f271b2cc | 2037 | /* Set hw address */ |
6b8a66ee JP |
2038 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2039 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2040 | |
40102371 | 2041 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2042 | break; |
33dccbb0 HX |
2043 | |
2044 | case TUNGETSNDBUF: | |
54f968d6 | 2045 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2046 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2047 | ret = -EFAULT; | |
2048 | break; | |
2049 | ||
2050 | case TUNSETSNDBUF: | |
2051 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2052 | ret = -EFAULT; | |
2053 | break; | |
2054 | } | |
2055 | ||
c8d68e6b JW |
2056 | tun->sndbuf = sndbuf; |
2057 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2058 | break; |
2059 | ||
d9d52b51 MT |
2060 | case TUNGETVNETHDRSZ: |
2061 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2062 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2063 | ret = -EFAULT; | |
2064 | break; | |
2065 | ||
2066 | case TUNSETVNETHDRSZ: | |
2067 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2068 | ret = -EFAULT; | |
2069 | break; | |
2070 | } | |
2071 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2072 | ret = -EINVAL; | |
2073 | break; | |
2074 | } | |
2075 | ||
2076 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2077 | break; | |
2078 | ||
99405162 MT |
2079 | case TUNATTACHFILTER: |
2080 | /* Can be set only for TAPs */ | |
2081 | ret = -EINVAL; | |
2082 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2083 | break; | |
2084 | ret = -EFAULT; | |
54f968d6 | 2085 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2086 | break; |
2087 | ||
c8d68e6b | 2088 | ret = tun_attach_filter(tun); |
99405162 MT |
2089 | break; |
2090 | ||
2091 | case TUNDETACHFILTER: | |
2092 | /* Can be set only for TAPs */ | |
2093 | ret = -EINVAL; | |
2094 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2095 | break; | |
c8d68e6b JW |
2096 | ret = 0; |
2097 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2098 | break; |
2099 | ||
1da177e4 | 2100 | default: |
631ab46b EB |
2101 | ret = -EINVAL; |
2102 | break; | |
ee289b64 | 2103 | } |
1da177e4 | 2104 | |
876bfd4d HX |
2105 | unlock: |
2106 | rtnl_unlock(); | |
2107 | if (tun) | |
2108 | tun_put(tun); | |
631ab46b | 2109 | return ret; |
1da177e4 LT |
2110 | } |
2111 | ||
50857e2a AB |
2112 | static long tun_chr_ioctl(struct file *file, |
2113 | unsigned int cmd, unsigned long arg) | |
2114 | { | |
2115 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2116 | } | |
2117 | ||
2118 | #ifdef CONFIG_COMPAT | |
2119 | static long tun_chr_compat_ioctl(struct file *file, | |
2120 | unsigned int cmd, unsigned long arg) | |
2121 | { | |
2122 | switch (cmd) { | |
2123 | case TUNSETIFF: | |
2124 | case TUNGETIFF: | |
2125 | case TUNSETTXFILTER: | |
2126 | case TUNGETSNDBUF: | |
2127 | case TUNSETSNDBUF: | |
2128 | case SIOCGIFHWADDR: | |
2129 | case SIOCSIFHWADDR: | |
2130 | arg = (unsigned long)compat_ptr(arg); | |
2131 | break; | |
2132 | default: | |
2133 | arg = (compat_ulong_t)arg; | |
2134 | break; | |
2135 | } | |
2136 | ||
2137 | /* | |
2138 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2139 | * the end of that structure. All fields that are used in this | |
2140 | * driver are compatible though, we don't need to convert the | |
2141 | * contents. | |
2142 | */ | |
2143 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2144 | } | |
2145 | #endif /* CONFIG_COMPAT */ | |
2146 | ||
1da177e4 LT |
2147 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2148 | { | |
54f968d6 | 2149 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2150 | int ret; |
2151 | ||
54f968d6 | 2152 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2153 | goto out; |
6aa20a22 | 2154 | |
1da177e4 | 2155 | if (on) { |
609d7fa9 | 2156 | ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
1da177e4 | 2157 | if (ret) |
9d319522 | 2158 | goto out; |
54f968d6 | 2159 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2160 | } else |
54f968d6 | 2161 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2162 | ret = 0; |
2163 | out: | |
9d319522 | 2164 | return ret; |
1da177e4 LT |
2165 | } |
2166 | ||
2167 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2168 | { | |
631ab46b | 2169 | struct tun_file *tfile; |
deed49fb | 2170 | |
6b8a66ee | 2171 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2172 | |
54f968d6 JW |
2173 | tfile = (struct tun_file *)sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL, |
2174 | &tun_proto); | |
631ab46b EB |
2175 | if (!tfile) |
2176 | return -ENOMEM; | |
6e914fc7 | 2177 | rcu_assign_pointer(tfile->tun, NULL); |
36b50bab | 2178 | tfile->net = get_net(current->nsproxy->net_ns); |
54f968d6 JW |
2179 | tfile->flags = 0; |
2180 | ||
2181 | rcu_assign_pointer(tfile->socket.wq, &tfile->wq); | |
2182 | init_waitqueue_head(&tfile->wq.wait); | |
2183 | ||
2184 | tfile->socket.file = file; | |
2185 | tfile->socket.ops = &tun_socket_ops; | |
2186 | ||
2187 | sock_init_data(&tfile->socket, &tfile->sk); | |
2188 | sk_change_net(&tfile->sk, tfile->net); | |
2189 | ||
2190 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2191 | tfile->sk.sk_sndbuf = INT_MAX; | |
2192 | ||
631ab46b | 2193 | file->private_data = tfile; |
54f968d6 | 2194 | set_bit(SOCK_EXTERNALLY_ALLOCATED, &tfile->socket.flags); |
4008e97f | 2195 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2196 | |
19a6afb2 JW |
2197 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2198 | ||
1da177e4 LT |
2199 | return 0; |
2200 | } | |
2201 | ||
2202 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2203 | { | |
631ab46b | 2204 | struct tun_file *tfile = file->private_data; |
54f968d6 | 2205 | struct net *net = tfile->net; |
1da177e4 | 2206 | |
c8d68e6b | 2207 | tun_detach(tfile, true); |
54f968d6 | 2208 | put_net(net); |
1da177e4 LT |
2209 | |
2210 | return 0; | |
2211 | } | |
2212 | ||
d54b1fdb | 2213 | static const struct file_operations tun_fops = { |
6aa20a22 | 2214 | .owner = THIS_MODULE, |
1da177e4 | 2215 | .llseek = no_llseek, |
ee0b3e67 BP |
2216 | .read = do_sync_read, |
2217 | .aio_read = tun_chr_aio_read, | |
2218 | .write = do_sync_write, | |
2219 | .aio_write = tun_chr_aio_write, | |
1da177e4 | 2220 | .poll = tun_chr_poll, |
50857e2a AB |
2221 | .unlocked_ioctl = tun_chr_ioctl, |
2222 | #ifdef CONFIG_COMPAT | |
2223 | .compat_ioctl = tun_chr_compat_ioctl, | |
2224 | #endif | |
1da177e4 LT |
2225 | .open = tun_chr_open, |
2226 | .release = tun_chr_close, | |
6aa20a22 | 2227 | .fasync = tun_chr_fasync |
1da177e4 LT |
2228 | }; |
2229 | ||
2230 | static struct miscdevice tun_miscdev = { | |
2231 | .minor = TUN_MINOR, | |
2232 | .name = "tun", | |
e454cea2 | 2233 | .nodename = "net/tun", |
1da177e4 | 2234 | .fops = &tun_fops, |
1da177e4 LT |
2235 | }; |
2236 | ||
2237 | /* ethtool interface */ | |
2238 | ||
2239 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) | |
2240 | { | |
2241 | cmd->supported = 0; | |
2242 | cmd->advertising = 0; | |
70739497 | 2243 | ethtool_cmd_speed_set(cmd, SPEED_10); |
1da177e4 LT |
2244 | cmd->duplex = DUPLEX_FULL; |
2245 | cmd->port = PORT_TP; | |
2246 | cmd->phy_address = 0; | |
2247 | cmd->transceiver = XCVR_INTERNAL; | |
2248 | cmd->autoneg = AUTONEG_DISABLE; | |
2249 | cmd->maxtxpkt = 0; | |
2250 | cmd->maxrxpkt = 0; | |
2251 | return 0; | |
2252 | } | |
2253 | ||
2254 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2255 | { | |
2256 | struct tun_struct *tun = netdev_priv(dev); | |
2257 | ||
33a5ba14 RJ |
2258 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2259 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2260 | |
2261 | switch (tun->flags & TUN_TYPE_MASK) { | |
2262 | case TUN_TUN_DEV: | |
33a5ba14 | 2263 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 LT |
2264 | break; |
2265 | case TUN_TAP_DEV: | |
33a5ba14 | 2266 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2267 | break; |
2268 | } | |
2269 | } | |
2270 | ||
2271 | static u32 tun_get_msglevel(struct net_device *dev) | |
2272 | { | |
2273 | #ifdef TUN_DEBUG | |
2274 | struct tun_struct *tun = netdev_priv(dev); | |
2275 | return tun->debug; | |
2276 | #else | |
2277 | return -EOPNOTSUPP; | |
2278 | #endif | |
2279 | } | |
2280 | ||
2281 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2282 | { | |
2283 | #ifdef TUN_DEBUG | |
2284 | struct tun_struct *tun = netdev_priv(dev); | |
2285 | tun->debug = value; | |
2286 | #endif | |
2287 | } | |
2288 | ||
7282d491 | 2289 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2290 | .get_settings = tun_get_settings, |
2291 | .get_drvinfo = tun_get_drvinfo, | |
2292 | .get_msglevel = tun_get_msglevel, | |
2293 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2294 | .get_link = ethtool_op_get_link, |
eda29772 | 2295 | .get_ts_info = ethtool_op_get_ts_info, |
1da177e4 LT |
2296 | }; |
2297 | ||
79d17604 | 2298 | |
1da177e4 LT |
2299 | static int __init tun_init(void) |
2300 | { | |
2301 | int ret = 0; | |
2302 | ||
6b8a66ee JP |
2303 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
2304 | pr_info("%s\n", DRV_COPYRIGHT); | |
1da177e4 | 2305 | |
f019a7a5 | 2306 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 2307 | if (ret) { |
6b8a66ee | 2308 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 2309 | goto err_linkops; |
79d17604 PE |
2310 | } |
2311 | ||
1da177e4 | 2312 | ret = misc_register(&tun_miscdev); |
79d17604 | 2313 | if (ret) { |
6b8a66ee | 2314 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
2315 | goto err_misc; |
2316 | } | |
f019a7a5 | 2317 | return 0; |
79d17604 | 2318 | err_misc: |
f019a7a5 EB |
2319 | rtnl_link_unregister(&tun_link_ops); |
2320 | err_linkops: | |
1da177e4 LT |
2321 | return ret; |
2322 | } | |
2323 | ||
2324 | static void tun_cleanup(void) | |
2325 | { | |
6aa20a22 | 2326 | misc_deregister(&tun_miscdev); |
f019a7a5 | 2327 | rtnl_link_unregister(&tun_link_ops); |
1da177e4 LT |
2328 | } |
2329 | ||
05c2828c MT |
2330 | /* Get an underlying socket object from tun file. Returns error unless file is |
2331 | * attached to a device. The returned object works like a packet socket, it | |
2332 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
2333 | * holding a reference to the file for as long as the socket is in use. */ | |
2334 | struct socket *tun_get_socket(struct file *file) | |
2335 | { | |
6e914fc7 | 2336 | struct tun_file *tfile; |
05c2828c MT |
2337 | if (file->f_op != &tun_fops) |
2338 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
2339 | tfile = file->private_data; |
2340 | if (!tfile) | |
05c2828c | 2341 | return ERR_PTR(-EBADFD); |
54f968d6 | 2342 | return &tfile->socket; |
05c2828c MT |
2343 | } |
2344 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
2345 | ||
1da177e4 LT |
2346 | module_init(tun_init); |
2347 | module_exit(tun_cleanup); | |
2348 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
2349 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
2350 | MODULE_LICENSE("GPL"); | |
2351 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 2352 | MODULE_ALIAS("devname:net/tun"); |