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> |
5188cd44 | 68 | #include <net/ipv6.h> |
881d966b | 69 | #include <net/net_namespace.h> |
79d17604 | 70 | #include <net/netns/generic.h> |
f019a7a5 | 71 | #include <net/rtnetlink.h> |
33dccbb0 | 72 | #include <net/sock.h> |
93e14b6d | 73 | #include <linux/seq_file.h> |
e0b46d0e | 74 | #include <linux/uio.h> |
1da177e4 | 75 | |
1da177e4 LT |
76 | #include <asm/uaccess.h> |
77 | ||
14daa021 RR |
78 | /* Uncomment to enable debugging */ |
79 | /* #define TUN_DEBUG 1 */ | |
80 | ||
1da177e4 LT |
81 | #ifdef TUN_DEBUG |
82 | static int debug; | |
14daa021 | 83 | |
6b8a66ee JP |
84 | #define tun_debug(level, tun, fmt, args...) \ |
85 | do { \ | |
86 | if (tun->debug) \ | |
87 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
88 | } while (0) | |
89 | #define DBG1(level, fmt, args...) \ | |
90 | do { \ | |
91 | if (debug == 2) \ | |
92 | printk(level fmt, ##args); \ | |
93 | } while (0) | |
14daa021 | 94 | #else |
6b8a66ee JP |
95 | #define tun_debug(level, tun, fmt, args...) \ |
96 | do { \ | |
97 | if (0) \ | |
98 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
99 | } while (0) | |
100 | #define DBG1(level, fmt, args...) \ | |
101 | do { \ | |
102 | if (0) \ | |
103 | printk(level fmt, ##args); \ | |
104 | } while (0) | |
14daa021 RR |
105 | #endif |
106 | ||
0690899b MT |
107 | #define GOODCOPY_LEN 128 |
108 | ||
f271b2cc MK |
109 | #define FLT_EXACT_COUNT 8 |
110 | struct tap_filter { | |
111 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
112 | u32 mask[2]; /* Mask of the hashed addrs */ | |
113 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
114 | }; | |
115 | ||
92d4ea6e | 116 | /* DEFAULT_MAX_NUM_RSS_QUEUES were chosen to let the rx/tx queues allocated for |
edfb6a14 JW |
117 | * the netdevice to be fit in one page. So we can make sure the success of |
118 | * memory allocation. TODO: increase the limit. */ | |
119 | #define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES | |
b8732fb7 | 120 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 121 | |
96442e42 JW |
122 | #define TUN_FLOW_EXPIRE (3 * HZ) |
123 | ||
54f968d6 | 124 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 125 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
126 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
127 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 128 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 129 | * this). |
6e914fc7 JW |
130 | * |
131 | * RCU usage: | |
36fe8c09 | 132 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 133 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 134 | */ |
631ab46b | 135 | struct tun_file { |
54f968d6 JW |
136 | struct sock sk; |
137 | struct socket socket; | |
138 | struct socket_wq wq; | |
6e914fc7 | 139 | struct tun_struct __rcu *tun; |
36b50bab | 140 | struct net *net; |
54f968d6 JW |
141 | struct fasync_struct *fasync; |
142 | /* only used for fasnyc */ | |
143 | unsigned int flags; | |
fb7589a1 PE |
144 | union { |
145 | u16 queue_index; | |
146 | unsigned int ifindex; | |
147 | }; | |
4008e97f JW |
148 | struct list_head next; |
149 | struct tun_struct *detached; | |
631ab46b EB |
150 | }; |
151 | ||
96442e42 JW |
152 | struct tun_flow_entry { |
153 | struct hlist_node hash_link; | |
154 | struct rcu_head rcu; | |
155 | struct tun_struct *tun; | |
156 | ||
157 | u32 rxhash; | |
9bc88939 | 158 | u32 rps_rxhash; |
96442e42 JW |
159 | int queue_index; |
160 | unsigned long updated; | |
161 | }; | |
162 | ||
163 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
164 | ||
54f968d6 | 165 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 166 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
167 | * file were attached to a persist device. |
168 | */ | |
14daa021 | 169 | struct tun_struct { |
c8d68e6b JW |
170 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
171 | unsigned int numqueues; | |
f271b2cc | 172 | unsigned int flags; |
0625c883 EB |
173 | kuid_t owner; |
174 | kgid_t group; | |
14daa021 | 175 | |
14daa021 | 176 | struct net_device *dev; |
c8f44aff | 177 | netdev_features_t set_features; |
88255375 | 178 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
3d0ad094 | 179 | NETIF_F_TSO6) |
d9d52b51 MT |
180 | |
181 | int vnet_hdr_sz; | |
54f968d6 JW |
182 | int sndbuf; |
183 | struct tap_filter txflt; | |
184 | struct sock_fprog fprog; | |
185 | /* protected by rtnl lock */ | |
186 | bool filter_attached; | |
14daa021 RR |
187 | #ifdef TUN_DEBUG |
188 | int debug; | |
1da177e4 | 189 | #endif |
96442e42 | 190 | spinlock_t lock; |
96442e42 JW |
191 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
192 | struct timer_list flow_gc_timer; | |
193 | unsigned long ageing_time; | |
4008e97f JW |
194 | unsigned int numdisabled; |
195 | struct list_head disabled; | |
5dbbaf2d | 196 | void *security; |
b8732fb7 | 197 | u32 flow_count; |
14daa021 | 198 | }; |
1da177e4 | 199 | |
96442e42 JW |
200 | static inline u32 tun_hashfn(u32 rxhash) |
201 | { | |
202 | return rxhash & 0x3ff; | |
203 | } | |
204 | ||
205 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
206 | { | |
207 | struct tun_flow_entry *e; | |
96442e42 | 208 | |
b67bfe0d | 209 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
210 | if (e->rxhash == rxhash) |
211 | return e; | |
212 | } | |
213 | return NULL; | |
214 | } | |
215 | ||
216 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
217 | struct hlist_head *head, | |
218 | u32 rxhash, u16 queue_index) | |
219 | { | |
9fdc6bef ED |
220 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
221 | ||
96442e42 JW |
222 | if (e) { |
223 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
224 | rxhash, queue_index); | |
225 | e->updated = jiffies; | |
226 | e->rxhash = rxhash; | |
9bc88939 | 227 | e->rps_rxhash = 0; |
96442e42 JW |
228 | e->queue_index = queue_index; |
229 | e->tun = tun; | |
230 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 231 | ++tun->flow_count; |
96442e42 JW |
232 | } |
233 | return e; | |
234 | } | |
235 | ||
96442e42 JW |
236 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
237 | { | |
238 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
239 | e->rxhash, e->queue_index); | |
9bc88939 | 240 | sock_rps_reset_flow_hash(e->rps_rxhash); |
96442e42 | 241 | hlist_del_rcu(&e->hash_link); |
9fdc6bef | 242 | kfree_rcu(e, rcu); |
b8732fb7 | 243 | --tun->flow_count; |
96442e42 JW |
244 | } |
245 | ||
246 | static void tun_flow_flush(struct tun_struct *tun) | |
247 | { | |
248 | int i; | |
249 | ||
250 | spin_lock_bh(&tun->lock); | |
251 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
252 | struct tun_flow_entry *e; | |
b67bfe0d | 253 | struct hlist_node *n; |
96442e42 | 254 | |
b67bfe0d | 255 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
256 | tun_flow_delete(tun, e); |
257 | } | |
258 | spin_unlock_bh(&tun->lock); | |
259 | } | |
260 | ||
261 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
262 | { | |
263 | int i; | |
264 | ||
265 | spin_lock_bh(&tun->lock); | |
266 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
267 | struct tun_flow_entry *e; | |
b67bfe0d | 268 | struct hlist_node *n; |
96442e42 | 269 | |
b67bfe0d | 270 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
271 | if (e->queue_index == queue_index) |
272 | tun_flow_delete(tun, e); | |
273 | } | |
274 | } | |
275 | spin_unlock_bh(&tun->lock); | |
276 | } | |
277 | ||
278 | static void tun_flow_cleanup(unsigned long data) | |
279 | { | |
280 | struct tun_struct *tun = (struct tun_struct *)data; | |
281 | unsigned long delay = tun->ageing_time; | |
282 | unsigned long next_timer = jiffies + delay; | |
283 | unsigned long count = 0; | |
284 | int i; | |
285 | ||
286 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
287 | ||
288 | spin_lock_bh(&tun->lock); | |
289 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
290 | struct tun_flow_entry *e; | |
b67bfe0d | 291 | struct hlist_node *n; |
96442e42 | 292 | |
b67bfe0d | 293 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
294 | unsigned long this_timer; |
295 | count++; | |
296 | this_timer = e->updated + delay; | |
297 | if (time_before_eq(this_timer, jiffies)) | |
298 | tun_flow_delete(tun, e); | |
299 | else if (time_before(this_timer, next_timer)) | |
300 | next_timer = this_timer; | |
301 | } | |
302 | } | |
303 | ||
304 | if (count) | |
305 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
306 | spin_unlock_bh(&tun->lock); | |
307 | } | |
308 | ||
49974420 | 309 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 310 | struct tun_file *tfile) |
96442e42 JW |
311 | { |
312 | struct hlist_head *head; | |
313 | struct tun_flow_entry *e; | |
314 | unsigned long delay = tun->ageing_time; | |
9e85722d | 315 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
316 | |
317 | if (!rxhash) | |
318 | return; | |
319 | else | |
320 | head = &tun->flows[tun_hashfn(rxhash)]; | |
321 | ||
322 | rcu_read_lock(); | |
323 | ||
9e85722d JW |
324 | /* We may get a very small possibility of OOO during switching, not |
325 | * worth to optimize.*/ | |
326 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
327 | goto unlock; |
328 | ||
329 | e = tun_flow_find(head, rxhash); | |
330 | if (likely(e)) { | |
331 | /* TODO: keep queueing to old queue until it's empty? */ | |
332 | e->queue_index = queue_index; | |
333 | e->updated = jiffies; | |
9bc88939 | 334 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
335 | } else { |
336 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
337 | if (!tun_flow_find(head, rxhash) && |
338 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
339 | tun_flow_create(tun, head, rxhash, queue_index); |
340 | ||
341 | if (!timer_pending(&tun->flow_gc_timer)) | |
342 | mod_timer(&tun->flow_gc_timer, | |
343 | round_jiffies_up(jiffies + delay)); | |
344 | spin_unlock_bh(&tun->lock); | |
345 | } | |
346 | ||
347 | unlock: | |
348 | rcu_read_unlock(); | |
349 | } | |
350 | ||
9bc88939 TH |
351 | /** |
352 | * Save the hash received in the stack receive path and update the | |
353 | * flow_hash table accordingly. | |
354 | */ | |
355 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
356 | { | |
357 | if (unlikely(e->rps_rxhash != hash)) { | |
358 | sock_rps_reset_flow_hash(e->rps_rxhash); | |
359 | e->rps_rxhash = hash; | |
360 | } | |
361 | } | |
362 | ||
c8d68e6b | 363 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 364 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
365 | * the rxq based on the txq where the last packet of the flow comes. As |
366 | * the userspace application move between processors, we may get a | |
367 | * different rxq no. here. If we could not get rxhash, then we would | |
368 | * hope the rxq no. may help here. | |
369 | */ | |
f663dd9a | 370 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 371 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
372 | { |
373 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 374 | struct tun_flow_entry *e; |
c8d68e6b JW |
375 | u32 txq = 0; |
376 | u32 numqueues = 0; | |
377 | ||
378 | rcu_read_lock(); | |
92bb73ea | 379 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 380 | |
3958afa1 | 381 | txq = skb_get_hash(skb); |
c8d68e6b | 382 | if (txq) { |
96442e42 | 383 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 384 | if (e) { |
9bc88939 | 385 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 386 | txq = e->queue_index; |
9bc88939 | 387 | } else |
96442e42 JW |
388 | /* use multiply and shift instead of expensive divide */ |
389 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
390 | } else if (likely(skb_rx_queue_recorded(skb))) { |
391 | txq = skb_get_rx_queue(skb); | |
392 | while (unlikely(txq >= numqueues)) | |
393 | txq -= numqueues; | |
394 | } | |
395 | ||
396 | rcu_read_unlock(); | |
397 | return txq; | |
398 | } | |
399 | ||
cde8b15f JW |
400 | static inline bool tun_not_capable(struct tun_struct *tun) |
401 | { | |
402 | const struct cred *cred = current_cred(); | |
c260b772 | 403 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
404 | |
405 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
406 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 407 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
408 | } |
409 | ||
c8d68e6b JW |
410 | static void tun_set_real_num_queues(struct tun_struct *tun) |
411 | { | |
412 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
413 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
414 | } | |
415 | ||
4008e97f JW |
416 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
417 | { | |
418 | tfile->detached = tun; | |
419 | list_add_tail(&tfile->next, &tun->disabled); | |
420 | ++tun->numdisabled; | |
421 | } | |
422 | ||
d32649d1 | 423 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
424 | { |
425 | struct tun_struct *tun = tfile->detached; | |
426 | ||
427 | tfile->detached = NULL; | |
428 | list_del_init(&tfile->next); | |
429 | --tun->numdisabled; | |
430 | return tun; | |
431 | } | |
432 | ||
4bfb0513 JW |
433 | static void tun_queue_purge(struct tun_file *tfile) |
434 | { | |
435 | skb_queue_purge(&tfile->sk.sk_receive_queue); | |
436 | skb_queue_purge(&tfile->sk.sk_error_queue); | |
437 | } | |
438 | ||
c8d68e6b JW |
439 | static void __tun_detach(struct tun_file *tfile, bool clean) |
440 | { | |
441 | struct tun_file *ntfile; | |
442 | struct tun_struct *tun; | |
c8d68e6b | 443 | |
b8deabd3 JW |
444 | tun = rtnl_dereference(tfile->tun); |
445 | ||
9e85722d | 446 | if (tun && !tfile->detached) { |
c8d68e6b JW |
447 | u16 index = tfile->queue_index; |
448 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
449 | |
450 | rcu_assign_pointer(tun->tfiles[index], | |
451 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 452 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
453 | ntfile->queue_index = index; |
454 | ||
455 | --tun->numqueues; | |
9e85722d | 456 | if (clean) { |
c956674b | 457 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 458 | sock_put(&tfile->sk); |
9e85722d | 459 | } else |
4008e97f | 460 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
461 | |
462 | synchronize_net(); | |
96442e42 | 463 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 464 | /* Drop read queue */ |
4bfb0513 | 465 | tun_queue_purge(tfile); |
c8d68e6b | 466 | tun_set_real_num_queues(tun); |
dd38bd85 | 467 | } else if (tfile->detached && clean) { |
4008e97f | 468 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
469 | sock_put(&tfile->sk); |
470 | } | |
c8d68e6b JW |
471 | |
472 | if (clean) { | |
af668b3c MT |
473 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
474 | netif_carrier_off(tun->dev); | |
475 | ||
476 | if (!(tun->flags & TUN_PERSIST) && | |
477 | tun->dev->reg_state == NETREG_REGISTERED) | |
4008e97f | 478 | unregister_netdevice(tun->dev); |
af668b3c | 479 | } |
4008e97f | 480 | |
c8d68e6b JW |
481 | BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED, |
482 | &tfile->socket.flags)); | |
483 | sk_release_kernel(&tfile->sk); | |
484 | } | |
485 | } | |
486 | ||
487 | static void tun_detach(struct tun_file *tfile, bool clean) | |
488 | { | |
489 | rtnl_lock(); | |
490 | __tun_detach(tfile, clean); | |
491 | rtnl_unlock(); | |
492 | } | |
493 | ||
494 | static void tun_detach_all(struct net_device *dev) | |
495 | { | |
496 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 497 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
498 | int i, n = tun->numqueues; |
499 | ||
500 | for (i = 0; i < n; i++) { | |
b8deabd3 | 501 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 502 | BUG_ON(!tfile); |
9e641bdc | 503 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 504 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
505 | --tun->numqueues; |
506 | } | |
9e85722d | 507 | list_for_each_entry(tfile, &tun->disabled, next) { |
9e641bdc | 508 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 509 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 510 | } |
c8d68e6b JW |
511 | BUG_ON(tun->numqueues != 0); |
512 | ||
513 | synchronize_net(); | |
514 | for (i = 0; i < n; i++) { | |
b8deabd3 | 515 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 516 | /* Drop read queue */ |
4bfb0513 | 517 | tun_queue_purge(tfile); |
c8d68e6b JW |
518 | sock_put(&tfile->sk); |
519 | } | |
4008e97f JW |
520 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
521 | tun_enable_queue(tfile); | |
4bfb0513 | 522 | tun_queue_purge(tfile); |
4008e97f JW |
523 | sock_put(&tfile->sk); |
524 | } | |
525 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 JW |
526 | |
527 | if (tun->flags & TUN_PERSIST) | |
528 | module_put(THIS_MODULE); | |
c8d68e6b JW |
529 | } |
530 | ||
849c9b6f | 531 | static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter) |
a7385ba2 | 532 | { |
631ab46b | 533 | struct tun_file *tfile = file->private_data; |
38231b7a | 534 | int err; |
a7385ba2 | 535 | |
5dbbaf2d PM |
536 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
537 | if (err < 0) | |
538 | goto out; | |
539 | ||
38231b7a | 540 | err = -EINVAL; |
9e85722d | 541 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
542 | goto out; |
543 | ||
544 | err = -EBUSY; | |
c8d68e6b JW |
545 | if (!(tun->flags & TUN_TAP_MQ) && tun->numqueues == 1) |
546 | goto out; | |
547 | ||
548 | err = -E2BIG; | |
4008e97f JW |
549 | if (!tfile->detached && |
550 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
551 | goto out; |
552 | ||
553 | err = 0; | |
54f968d6 | 554 | |
92d4ea6e | 555 | /* Re-attach the filter to persist device */ |
849c9b6f | 556 | if (!skip_filter && (tun->filter_attached == true)) { |
54f968d6 JW |
557 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); |
558 | if (!err) | |
559 | goto out; | |
560 | } | |
c8d68e6b | 561 | tfile->queue_index = tun->numqueues; |
6e914fc7 | 562 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 563 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 564 | tun->numqueues++; |
a7385ba2 | 565 | |
4008e97f JW |
566 | if (tfile->detached) |
567 | tun_enable_queue(tfile); | |
568 | else | |
569 | sock_hold(&tfile->sk); | |
570 | ||
c8d68e6b | 571 | tun_set_real_num_queues(tun); |
a7385ba2 | 572 | |
c8d68e6b JW |
573 | /* device is allowed to go away first, so no need to hold extra |
574 | * refcnt. | |
575 | */ | |
576 | ||
577 | out: | |
578 | return err; | |
631ab46b EB |
579 | } |
580 | ||
581 | static struct tun_struct *__tun_get(struct tun_file *tfile) | |
582 | { | |
6e914fc7 | 583 | struct tun_struct *tun; |
c70f1829 | 584 | |
6e914fc7 JW |
585 | rcu_read_lock(); |
586 | tun = rcu_dereference(tfile->tun); | |
587 | if (tun) | |
588 | dev_hold(tun->dev); | |
589 | rcu_read_unlock(); | |
c70f1829 EB |
590 | |
591 | return tun; | |
631ab46b EB |
592 | } |
593 | ||
594 | static struct tun_struct *tun_get(struct file *file) | |
595 | { | |
596 | return __tun_get(file->private_data); | |
597 | } | |
598 | ||
599 | static void tun_put(struct tun_struct *tun) | |
600 | { | |
6e914fc7 | 601 | dev_put(tun->dev); |
631ab46b EB |
602 | } |
603 | ||
6b8a66ee | 604 | /* TAP filtering */ |
f271b2cc MK |
605 | static void addr_hash_set(u32 *mask, const u8 *addr) |
606 | { | |
607 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
608 | mask[n >> 5] |= (1 << (n & 31)); | |
609 | } | |
610 | ||
611 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
612 | { | |
613 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
614 | return mask[n >> 5] & (1 << (n & 31)); | |
615 | } | |
616 | ||
617 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
618 | { | |
619 | struct { u8 u[ETH_ALEN]; } *addr; | |
620 | struct tun_filter uf; | |
621 | int err, alen, n, nexact; | |
622 | ||
623 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
624 | return -EFAULT; | |
625 | ||
626 | if (!uf.count) { | |
627 | /* Disabled */ | |
628 | filter->count = 0; | |
629 | return 0; | |
630 | } | |
631 | ||
632 | alen = ETH_ALEN * uf.count; | |
633 | addr = kmalloc(alen, GFP_KERNEL); | |
634 | if (!addr) | |
635 | return -ENOMEM; | |
636 | ||
637 | if (copy_from_user(addr, arg + sizeof(uf), alen)) { | |
638 | err = -EFAULT; | |
639 | goto done; | |
640 | } | |
641 | ||
642 | /* The filter is updated without holding any locks. Which is | |
643 | * perfectly safe. We disable it first and in the worst | |
644 | * case we'll accept a few undesired packets. */ | |
645 | filter->count = 0; | |
646 | wmb(); | |
647 | ||
648 | /* Use first set of addresses as an exact filter */ | |
649 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
650 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
651 | ||
652 | nexact = n; | |
653 | ||
cfbf84fc AW |
654 | /* Remaining multicast addresses are hashed, |
655 | * unicast will leave the filter disabled. */ | |
f271b2cc | 656 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
657 | for (; n < uf.count; n++) { |
658 | if (!is_multicast_ether_addr(addr[n].u)) { | |
659 | err = 0; /* no filter */ | |
660 | goto done; | |
661 | } | |
f271b2cc | 662 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 663 | } |
f271b2cc MK |
664 | |
665 | /* For ALLMULTI just set the mask to all ones. | |
666 | * This overrides the mask populated above. */ | |
667 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
668 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
669 | ||
670 | /* Now enable the filter */ | |
671 | wmb(); | |
672 | filter->count = nexact; | |
673 | ||
674 | /* Return the number of exact filters */ | |
675 | err = nexact; | |
676 | ||
677 | done: | |
678 | kfree(addr); | |
679 | return err; | |
680 | } | |
681 | ||
682 | /* Returns: 0 - drop, !=0 - accept */ | |
683 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
684 | { | |
685 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
686 | * at this point. */ | |
687 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
688 | int i; | |
689 | ||
690 | /* Exact match */ | |
691 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 692 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
693 | return 1; |
694 | ||
695 | /* Inexact match (multicast only) */ | |
696 | if (is_multicast_ether_addr(eh->h_dest)) | |
697 | return addr_hash_test(filter->mask, eh->h_dest); | |
698 | ||
699 | return 0; | |
700 | } | |
701 | ||
702 | /* | |
703 | * Checks whether the packet is accepted or not. | |
704 | * Returns: 0 - drop, !=0 - accept | |
705 | */ | |
706 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
707 | { | |
708 | if (!filter->count) | |
709 | return 1; | |
710 | ||
711 | return run_filter(filter, skb); | |
712 | } | |
713 | ||
1da177e4 LT |
714 | /* Network device part of the driver */ |
715 | ||
7282d491 | 716 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 717 | |
c70f1829 EB |
718 | /* Net device detach from fd. */ |
719 | static void tun_net_uninit(struct net_device *dev) | |
720 | { | |
c8d68e6b | 721 | tun_detach_all(dev); |
c70f1829 EB |
722 | } |
723 | ||
1da177e4 LT |
724 | /* Net device open. */ |
725 | static int tun_net_open(struct net_device *dev) | |
726 | { | |
c8d68e6b | 727 | netif_tx_start_all_queues(dev); |
1da177e4 LT |
728 | return 0; |
729 | } | |
730 | ||
731 | /* Net device close. */ | |
732 | static int tun_net_close(struct net_device *dev) | |
733 | { | |
c8d68e6b | 734 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
735 | return 0; |
736 | } | |
737 | ||
738 | /* Net device start xmit */ | |
424efe9c | 739 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
740 | { |
741 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 742 | int txq = skb->queue_mapping; |
6e914fc7 | 743 | struct tun_file *tfile; |
fa35864e | 744 | u32 numqueues = 0; |
1da177e4 | 745 | |
6e914fc7 | 746 | rcu_read_lock(); |
c8d68e6b | 747 | tfile = rcu_dereference(tun->tfiles[txq]); |
fa35864e | 748 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 749 | |
1da177e4 | 750 | /* Drop packet if interface is not attached */ |
fa35864e | 751 | if (txq >= numqueues) |
1da177e4 LT |
752 | goto drop; |
753 | ||
fa35864e | 754 | if (numqueues == 1) { |
9bc88939 TH |
755 | /* Select queue was not called for the skbuff, so we extract the |
756 | * RPS hash and save it into the flow_table here. | |
757 | */ | |
758 | __u32 rxhash; | |
759 | ||
760 | rxhash = skb_get_hash(skb); | |
761 | if (rxhash) { | |
762 | struct tun_flow_entry *e; | |
763 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
764 | rxhash); | |
765 | if (e) | |
766 | tun_flow_save_rps_rxhash(e, rxhash); | |
767 | } | |
768 | } | |
769 | ||
6e914fc7 JW |
770 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
771 | ||
c8d68e6b JW |
772 | BUG_ON(!tfile); |
773 | ||
f271b2cc MK |
774 | /* Drop if the filter does not like it. |
775 | * This is a noop if the filter is disabled. | |
776 | * Filter can be enabled only for the TAP devices. */ | |
777 | if (!check_filter(&tun->txflt, skb)) | |
778 | goto drop; | |
779 | ||
54f968d6 JW |
780 | if (tfile->socket.sk->sk_filter && |
781 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
782 | goto drop; |
783 | ||
36fe8c09 | 784 | /* Limit the number of packets queued by dividing txq length with the |
c8d68e6b JW |
785 | * number of queues. |
786 | */ | |
fa35864e DC |
787 | if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues |
788 | >= dev->tx_queue_len) | |
5d097109 | 789 | goto drop; |
1da177e4 | 790 | |
7bf66305 JW |
791 | if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC))) |
792 | goto drop; | |
793 | ||
eda29772 RC |
794 | if (skb->sk) { |
795 | sock_tx_timestamp(skb->sk, &skb_shinfo(skb)->tx_flags); | |
796 | sw_tx_timestamp(skb); | |
797 | } | |
798 | ||
0110d6f2 | 799 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
800 | * for indefinite time. |
801 | */ | |
0110d6f2 MT |
802 | skb_orphan(skb); |
803 | ||
f8af75f3 ED |
804 | nf_reset(skb); |
805 | ||
f271b2cc | 806 | /* Enqueue packet */ |
54f968d6 | 807 | skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb); |
1da177e4 LT |
808 | |
809 | /* Notify and wake up reader process */ | |
54f968d6 JW |
810 | if (tfile->flags & TUN_FASYNC) |
811 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 812 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
813 | |
814 | rcu_read_unlock(); | |
6ed10654 | 815 | return NETDEV_TX_OK; |
1da177e4 LT |
816 | |
817 | drop: | |
09f75cd7 | 818 | dev->stats.tx_dropped++; |
149d36f7 | 819 | skb_tx_error(skb); |
1da177e4 | 820 | kfree_skb(skb); |
6e914fc7 | 821 | rcu_read_unlock(); |
baeababb | 822 | return NET_XMIT_DROP; |
1da177e4 LT |
823 | } |
824 | ||
f271b2cc | 825 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 826 | { |
f271b2cc MK |
827 | /* |
828 | * This callback is supposed to deal with mc filter in | |
829 | * _rx_ path and has nothing to do with the _tx_ path. | |
830 | * In rx path we always accept everything userspace gives us. | |
831 | */ | |
1da177e4 LT |
832 | } |
833 | ||
4885a504 ES |
834 | #define MIN_MTU 68 |
835 | #define MAX_MTU 65535 | |
836 | ||
837 | static int | |
838 | tun_net_change_mtu(struct net_device *dev, int new_mtu) | |
839 | { | |
840 | if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU) | |
841 | return -EINVAL; | |
842 | dev->mtu = new_mtu; | |
843 | return 0; | |
844 | } | |
845 | ||
c8f44aff MM |
846 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
847 | netdev_features_t features) | |
88255375 MM |
848 | { |
849 | struct tun_struct *tun = netdev_priv(dev); | |
850 | ||
851 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
852 | } | |
bebd097a NH |
853 | #ifdef CONFIG_NET_POLL_CONTROLLER |
854 | static void tun_poll_controller(struct net_device *dev) | |
855 | { | |
856 | /* | |
857 | * Tun only receives frames when: | |
858 | * 1) the char device endpoint gets data from user space | |
859 | * 2) the tun socket gets a sendmsg call from user space | |
92d4ea6e | 860 | * Since both of those are synchronous operations, we are guaranteed |
bebd097a | 861 | * never to have pending data when we poll for it |
92d4ea6e | 862 | * so there is nothing to do here but return. |
bebd097a NH |
863 | * We need this though so netpoll recognizes us as an interface that |
864 | * supports polling, which enables bridge devices in virt setups to | |
865 | * still use netconsole | |
866 | */ | |
867 | return; | |
868 | } | |
869 | #endif | |
758e43b7 | 870 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 871 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
872 | .ndo_open = tun_net_open, |
873 | .ndo_stop = tun_net_close, | |
00829823 | 874 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 875 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 876 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 877 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
878 | #ifdef CONFIG_NET_POLL_CONTROLLER |
879 | .ndo_poll_controller = tun_poll_controller, | |
880 | #endif | |
758e43b7 SH |
881 | }; |
882 | ||
883 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 884 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
885 | .ndo_open = tun_net_open, |
886 | .ndo_stop = tun_net_close, | |
00829823 | 887 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 888 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 889 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 890 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
891 | .ndo_set_mac_address = eth_mac_addr, |
892 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 893 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
894 | #ifdef CONFIG_NET_POLL_CONTROLLER |
895 | .ndo_poll_controller = tun_poll_controller, | |
896 | #endif | |
758e43b7 SH |
897 | }; |
898 | ||
944a1376 | 899 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
900 | { |
901 | int i; | |
902 | ||
96442e42 JW |
903 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
904 | INIT_HLIST_HEAD(&tun->flows[i]); | |
905 | ||
906 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
907 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
908 | mod_timer(&tun->flow_gc_timer, | |
909 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
910 | } |
911 | ||
912 | static void tun_flow_uninit(struct tun_struct *tun) | |
913 | { | |
914 | del_timer_sync(&tun->flow_gc_timer); | |
915 | tun_flow_flush(tun); | |
96442e42 JW |
916 | } |
917 | ||
1da177e4 LT |
918 | /* Initialize net device. */ |
919 | static void tun_net_init(struct net_device *dev) | |
920 | { | |
921 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 922 | |
1da177e4 LT |
923 | switch (tun->flags & TUN_TYPE_MASK) { |
924 | case TUN_TUN_DEV: | |
758e43b7 SH |
925 | dev->netdev_ops = &tun_netdev_ops; |
926 | ||
1da177e4 LT |
927 | /* Point-to-Point TUN Device */ |
928 | dev->hard_header_len = 0; | |
929 | dev->addr_len = 0; | |
930 | dev->mtu = 1500; | |
931 | ||
932 | /* Zero header length */ | |
6aa20a22 | 933 | dev->type = ARPHRD_NONE; |
1da177e4 LT |
934 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
935 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ | |
936 | break; | |
937 | ||
938 | case TUN_TAP_DEV: | |
7a0a9608 | 939 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 940 | /* Ethernet TAP Device */ |
1da177e4 | 941 | ether_setup(dev); |
550fd08c | 942 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 943 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 944 | |
f2cedb63 | 945 | eth_hw_addr_random(dev); |
36226a8d | 946 | |
1da177e4 LT |
947 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ |
948 | break; | |
949 | } | |
950 | } | |
951 | ||
952 | /* Character device part */ | |
953 | ||
954 | /* Poll */ | |
c8d68e6b | 955 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 956 | { |
b2430de3 EB |
957 | struct tun_file *tfile = file->private_data; |
958 | struct tun_struct *tun = __tun_get(tfile); | |
3c8a9c63 | 959 | struct sock *sk; |
33dccbb0 | 960 | unsigned int mask = 0; |
1da177e4 LT |
961 | |
962 | if (!tun) | |
eac9e902 | 963 | return POLLERR; |
1da177e4 | 964 | |
54f968d6 | 965 | sk = tfile->socket.sk; |
3c8a9c63 | 966 | |
6b8a66ee | 967 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 968 | |
9e641bdc | 969 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 970 | |
89f56d1e | 971 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
972 | mask |= POLLIN | POLLRDNORM; |
973 | ||
33dccbb0 HX |
974 | if (sock_writeable(sk) || |
975 | (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
976 | sock_writeable(sk))) | |
977 | mask |= POLLOUT | POLLWRNORM; | |
978 | ||
c70f1829 EB |
979 | if (tun->dev->reg_state != NETREG_REGISTERED) |
980 | mask = POLLERR; | |
981 | ||
631ab46b | 982 | tun_put(tun); |
1da177e4 LT |
983 | return mask; |
984 | } | |
985 | ||
f42157cb RR |
986 | /* prepad is the amount to reserve at front. len is length after that. |
987 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 988 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 989 | size_t prepad, size_t len, |
990 | size_t linear, int noblock) | |
f42157cb | 991 | { |
54f968d6 | 992 | struct sock *sk = tfile->socket.sk; |
f42157cb | 993 | struct sk_buff *skb; |
33dccbb0 | 994 | int err; |
f42157cb RR |
995 | |
996 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 997 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 998 | linear = len; |
f42157cb | 999 | |
33dccbb0 | 1000 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1001 | &err, 0); |
f42157cb | 1002 | if (!skb) |
33dccbb0 | 1003 | return ERR_PTR(err); |
f42157cb RR |
1004 | |
1005 | skb_reserve(skb, prepad); | |
1006 | skb_put(skb, linear); | |
33dccbb0 HX |
1007 | skb->data_len = len - linear; |
1008 | skb->len += len - linear; | |
f42157cb RR |
1009 | |
1010 | return skb; | |
1011 | } | |
1012 | ||
1da177e4 | 1013 | /* Get packet from user space buffer */ |
54f968d6 | 1014 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 AV |
1015 | void *msg_control, struct iov_iter *from, |
1016 | int noblock) | |
1da177e4 | 1017 | { |
09640e63 | 1018 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1019 | struct sk_buff *skb; |
f5ff53b4 | 1020 | size_t total_len = iov_iter_count(from); |
3dd5c330 | 1021 | size_t len = total_len, align = NET_SKB_PAD, linear; |
f43798c2 | 1022 | struct virtio_net_hdr gso = { 0 }; |
96f8d9ec | 1023 | int good_linear; |
0690899b MT |
1024 | int copylen; |
1025 | bool zerocopy = false; | |
1026 | int err; | |
49974420 | 1027 | u32 rxhash; |
f5ff53b4 | 1028 | ssize_t n; |
1da177e4 LT |
1029 | |
1030 | if (!(tun->flags & TUN_NO_PI)) { | |
15718ea0 | 1031 | if (len < sizeof(pi)) |
1da177e4 | 1032 | return -EINVAL; |
15718ea0 | 1033 | len -= sizeof(pi); |
1da177e4 | 1034 | |
f5ff53b4 AV |
1035 | n = copy_from_iter(&pi, sizeof(pi), from); |
1036 | if (n != sizeof(pi)) | |
1da177e4 LT |
1037 | return -EFAULT; |
1038 | } | |
1039 | ||
f43798c2 | 1040 | if (tun->flags & TUN_VNET_HDR) { |
15718ea0 | 1041 | if (len < tun->vnet_hdr_sz) |
f43798c2 | 1042 | return -EINVAL; |
15718ea0 | 1043 | len -= tun->vnet_hdr_sz; |
f43798c2 | 1044 | |
f5ff53b4 AV |
1045 | n = copy_from_iter(&gso, sizeof(gso), from); |
1046 | if (n != sizeof(gso)) | |
f43798c2 RR |
1047 | return -EFAULT; |
1048 | ||
4909122f HX |
1049 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
1050 | gso.csum_start + gso.csum_offset + 2 > gso.hdr_len) | |
1051 | gso.hdr_len = gso.csum_start + gso.csum_offset + 2; | |
1052 | ||
f43798c2 RR |
1053 | if (gso.hdr_len > len) |
1054 | return -EINVAL; | |
d8febb77 | 1055 | iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1056 | } |
1057 | ||
e01bf1c8 | 1058 | if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) { |
a504b86e | 1059 | align += NET_IP_ALIGN; |
0eca93bc HX |
1060 | if (unlikely(len < ETH_HLEN || |
1061 | (gso.hdr_len && gso.hdr_len < ETH_HLEN))) | |
e01bf1c8 RR |
1062 | return -EINVAL; |
1063 | } | |
6aa20a22 | 1064 | |
96f8d9ec JW |
1065 | good_linear = SKB_MAX_HEAD(align); |
1066 | ||
88529176 | 1067 | if (msg_control) { |
f5ff53b4 AV |
1068 | struct iov_iter i = *from; |
1069 | ||
88529176 JW |
1070 | /* There are 256 bytes to be copied in skb, so there is |
1071 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1072 | * The rest of the buffer is mapped from userspace. |
1073 | */ | |
88529176 | 1074 | copylen = gso.hdr_len ? gso.hdr_len : GOODCOPY_LEN; |
96f8d9ec JW |
1075 | if (copylen > good_linear) |
1076 | copylen = good_linear; | |
3dd5c330 | 1077 | linear = copylen; |
f5ff53b4 AV |
1078 | iov_iter_advance(&i, copylen); |
1079 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1080 | zerocopy = true; |
1081 | } | |
1082 | ||
1083 | if (!zerocopy) { | |
0690899b | 1084 | copylen = len; |
96f8d9ec JW |
1085 | if (gso.hdr_len > good_linear) |
1086 | linear = good_linear; | |
1087 | else | |
1088 | linear = gso.hdr_len; | |
3dd5c330 | 1089 | } |
0690899b | 1090 | |
3dd5c330 | 1091 | skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); |
33dccbb0 HX |
1092 | if (IS_ERR(skb)) { |
1093 | if (PTR_ERR(skb) != -EAGAIN) | |
1094 | tun->dev->stats.rx_dropped++; | |
1095 | return PTR_ERR(skb); | |
1da177e4 LT |
1096 | } |
1097 | ||
0690899b | 1098 | if (zerocopy) |
f5ff53b4 | 1099 | err = zerocopy_sg_from_iter(skb, from); |
88529176 | 1100 | else { |
f5ff53b4 | 1101 | err = skb_copy_datagram_from_iter(skb, 0, from, len); |
88529176 JW |
1102 | if (!err && msg_control) { |
1103 | struct ubuf_info *uarg = msg_control; | |
1104 | uarg->callback(uarg, false); | |
1105 | } | |
1106 | } | |
0690899b MT |
1107 | |
1108 | if (err) { | |
09f75cd7 | 1109 | tun->dev->stats.rx_dropped++; |
8f22757e | 1110 | kfree_skb(skb); |
1da177e4 | 1111 | return -EFAULT; |
8f22757e | 1112 | } |
1da177e4 | 1113 | |
f43798c2 RR |
1114 | if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { |
1115 | if (!skb_partial_csum_set(skb, gso.csum_start, | |
1116 | gso.csum_offset)) { | |
1117 | tun->dev->stats.rx_frame_errors++; | |
1118 | kfree_skb(skb); | |
1119 | return -EINVAL; | |
1120 | } | |
88255375 | 1121 | } |
f43798c2 | 1122 | |
1da177e4 LT |
1123 | switch (tun->flags & TUN_TYPE_MASK) { |
1124 | case TUN_TUN_DEV: | |
f09f7ee2 AWC |
1125 | if (tun->flags & TUN_NO_PI) { |
1126 | switch (skb->data[0] & 0xf0) { | |
1127 | case 0x40: | |
1128 | pi.proto = htons(ETH_P_IP); | |
1129 | break; | |
1130 | case 0x60: | |
1131 | pi.proto = htons(ETH_P_IPV6); | |
1132 | break; | |
1133 | default: | |
1134 | tun->dev->stats.rx_dropped++; | |
1135 | kfree_skb(skb); | |
1136 | return -EINVAL; | |
1137 | } | |
1138 | } | |
1139 | ||
459a98ed | 1140 | skb_reset_mac_header(skb); |
1da177e4 | 1141 | skb->protocol = pi.proto; |
4c13eb66 | 1142 | skb->dev = tun->dev; |
1da177e4 LT |
1143 | break; |
1144 | case TUN_TAP_DEV: | |
1145 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1146 | break; | |
6403eab1 | 1147 | } |
1da177e4 | 1148 | |
5188cd44 BH |
1149 | skb_reset_network_header(skb); |
1150 | ||
f43798c2 RR |
1151 | if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) { |
1152 | pr_debug("GSO!\n"); | |
1153 | switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { | |
1154 | case VIRTIO_NET_HDR_GSO_TCPV4: | |
c9af6db4 | 1155 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; |
f43798c2 RR |
1156 | break; |
1157 | case VIRTIO_NET_HDR_GSO_TCPV6: | |
c9af6db4 | 1158 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; |
f43798c2 | 1159 | break; |
e36aa25a | 1160 | case VIRTIO_NET_HDR_GSO_UDP: |
3d0ad094 BH |
1161 | { |
1162 | static bool warned; | |
1163 | ||
1164 | if (!warned) { | |
1165 | warned = true; | |
1166 | netdev_warn(tun->dev, | |
1167 | "%s: using disabled UFO feature; please fix this program\n", | |
1168 | current->comm); | |
1169 | } | |
c9af6db4 | 1170 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
5188cd44 BH |
1171 | if (skb->protocol == htons(ETH_P_IPV6)) |
1172 | ipv6_proxy_select_ident(skb); | |
e36aa25a | 1173 | break; |
3d0ad094 | 1174 | } |
f43798c2 RR |
1175 | default: |
1176 | tun->dev->stats.rx_frame_errors++; | |
1177 | kfree_skb(skb); | |
1178 | return -EINVAL; | |
1179 | } | |
1180 | ||
1181 | if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN) | |
c9af6db4 | 1182 | skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; |
f43798c2 RR |
1183 | |
1184 | skb_shinfo(skb)->gso_size = gso.gso_size; | |
1185 | if (skb_shinfo(skb)->gso_size == 0) { | |
1186 | tun->dev->stats.rx_frame_errors++; | |
1187 | kfree_skb(skb); | |
1188 | return -EINVAL; | |
1189 | } | |
1190 | ||
1191 | /* Header must be checked, and gso_segs computed. */ | |
1192 | skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY; | |
1193 | skb_shinfo(skb)->gso_segs = 0; | |
1194 | } | |
6aa20a22 | 1195 | |
0690899b MT |
1196 | /* copy skb_ubuf_info for callback when skb has no error */ |
1197 | if (zerocopy) { | |
1198 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1199 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1200 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
0690899b MT |
1201 | } |
1202 | ||
40893fd0 | 1203 | skb_probe_transport_header(skb, 0); |
38502af7 | 1204 | |
3958afa1 | 1205 | rxhash = skb_get_hash(skb); |
1da177e4 | 1206 | netif_rx_ni(skb); |
6aa20a22 | 1207 | |
09f75cd7 JG |
1208 | tun->dev->stats.rx_packets++; |
1209 | tun->dev->stats.rx_bytes += len; | |
1da177e4 | 1210 | |
9e85722d | 1211 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1212 | return total_len; |
6aa20a22 | 1213 | } |
1da177e4 | 1214 | |
f5ff53b4 | 1215 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1216 | { |
33dccbb0 | 1217 | struct file *file = iocb->ki_filp; |
ab46d779 | 1218 | struct tun_struct *tun = tun_get(file); |
54f968d6 | 1219 | struct tun_file *tfile = file->private_data; |
631ab46b | 1220 | ssize_t result; |
1da177e4 LT |
1221 | |
1222 | if (!tun) | |
1223 | return -EBADFD; | |
1224 | ||
f5ff53b4 | 1225 | result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK); |
631ab46b EB |
1226 | |
1227 | tun_put(tun); | |
1228 | return result; | |
1da177e4 LT |
1229 | } |
1230 | ||
1da177e4 | 1231 | /* Put packet to the user space buffer */ |
6f7c156c | 1232 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1233 | struct tun_file *tfile, |
6f7c156c | 1234 | struct sk_buff *skb, |
e0b46d0e | 1235 | struct iov_iter *iter) |
1da177e4 LT |
1236 | { |
1237 | struct tun_pi pi = { 0, skb->protocol }; | |
e0b46d0e | 1238 | ssize_t total; |
8c847d25 | 1239 | int vlan_offset = 0; |
a8f9bfdf | 1240 | int vlan_hlen = 0; |
2eb783c4 | 1241 | int vnet_hdr_sz = 0; |
a8f9bfdf HX |
1242 | |
1243 | if (vlan_tx_tag_present(skb)) | |
1244 | vlan_hlen = VLAN_HLEN; | |
1da177e4 | 1245 | |
2eb783c4 HX |
1246 | if (tun->flags & TUN_VNET_HDR) |
1247 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
1248 | ||
e0b46d0e HX |
1249 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
1250 | ||
1da177e4 | 1251 | if (!(tun->flags & TUN_NO_PI)) { |
e0b46d0e | 1252 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
1253 | return -EINVAL; |
1254 | ||
e0b46d0e HX |
1255 | total += sizeof(pi); |
1256 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
1257 | /* Packet will be striped */ |
1258 | pi.flags |= TUN_PKT_STRIP; | |
1259 | } | |
6aa20a22 | 1260 | |
e0b46d0e | 1261 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 1262 | return -EFAULT; |
6aa20a22 | 1263 | } |
1da177e4 | 1264 | |
2eb783c4 | 1265 | if (vnet_hdr_sz) { |
f43798c2 | 1266 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ |
e0b46d0e | 1267 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
1268 | return -EINVAL; |
1269 | ||
1270 | if (skb_is_gso(skb)) { | |
1271 | struct skb_shared_info *sinfo = skb_shinfo(skb); | |
1272 | ||
1273 | /* This is a hint as to how much should be linear. */ | |
1274 | gso.hdr_len = skb_headlen(skb); | |
1275 | gso.gso_size = sinfo->gso_size; | |
1276 | if (sinfo->gso_type & SKB_GSO_TCPV4) | |
1277 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4; | |
1278 | else if (sinfo->gso_type & SKB_GSO_TCPV6) | |
1279 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6; | |
ef3db4a5 | 1280 | else { |
6b8a66ee | 1281 | pr_err("unexpected GSO type: " |
ef3db4a5 MT |
1282 | "0x%x, gso_size %d, hdr_len %d\n", |
1283 | sinfo->gso_type, gso.gso_size, | |
1284 | gso.hdr_len); | |
1285 | print_hex_dump(KERN_ERR, "tun: ", | |
1286 | DUMP_PREFIX_NONE, | |
1287 | 16, 1, skb->head, | |
1288 | min((int)gso.hdr_len, 64), true); | |
1289 | WARN_ON_ONCE(1); | |
1290 | return -EINVAL; | |
1291 | } | |
f43798c2 RR |
1292 | if (sinfo->gso_type & SKB_GSO_TCP_ECN) |
1293 | gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN; | |
1294 | } else | |
1295 | gso.gso_type = VIRTIO_NET_HDR_GSO_NONE; | |
1296 | ||
1297 | if (skb->ip_summed == CHECKSUM_PARTIAL) { | |
1298 | gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; | |
a8f9bfdf HX |
1299 | gso.csum_start = skb_checksum_start_offset(skb) + |
1300 | vlan_hlen; | |
f43798c2 | 1301 | gso.csum_offset = skb->csum_offset; |
10a8d94a JW |
1302 | } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) { |
1303 | gso.flags = VIRTIO_NET_HDR_F_DATA_VALID; | |
f43798c2 RR |
1304 | } /* else everything is zero */ |
1305 | ||
e0b46d0e | 1306 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 1307 | return -EFAULT; |
8c847d25 JW |
1308 | |
1309 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
1310 | } |
1311 | ||
a8f9bfdf | 1312 | if (vlan_hlen) { |
e0b46d0e | 1313 | int ret; |
6680ec68 JW |
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); | |
6680ec68 | 1323 | |
e0b46d0e HX |
1324 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
1325 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
1326 | goto done; |
1327 | ||
e0b46d0e HX |
1328 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
1329 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
1330 | goto done; |
1331 | } | |
1da177e4 | 1332 | |
e0b46d0e | 1333 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 1334 | |
6680ec68 | 1335 | done: |
09f75cd7 | 1336 | tun->dev->stats.tx_packets++; |
e0b46d0e | 1337 | tun->dev->stats.tx_bytes += skb->len + vlan_hlen; |
1da177e4 LT |
1338 | |
1339 | return total; | |
1340 | } | |
1341 | ||
54f968d6 | 1342 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 AV |
1343 | struct iov_iter *to, |
1344 | int noblock) | |
1da177e4 | 1345 | { |
1da177e4 | 1346 | struct sk_buff *skb; |
9b067034 | 1347 | ssize_t ret; |
9e641bdc | 1348 | int peeked, err, off = 0; |
1da177e4 | 1349 | |
3872baf6 | 1350 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1351 | |
9b067034 AV |
1352 | if (!iov_iter_count(to)) |
1353 | return 0; | |
1da177e4 | 1354 | |
9e641bdc XW |
1355 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1356 | return -EIO; | |
1da177e4 | 1357 | |
9e641bdc XW |
1358 | /* Read frames from queue */ |
1359 | skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0, | |
1360 | &peeked, &off, &err); | |
e0b46d0e | 1361 | if (!skb) |
9b067034 | 1362 | return 0; |
e0b46d0e | 1363 | |
9b067034 | 1364 | ret = tun_put_user(tun, tfile, skb, to); |
f51a5e82 JW |
1365 | if (unlikely(ret < 0)) |
1366 | kfree_skb(skb); | |
1367 | else | |
1368 | consume_skb(skb); | |
1da177e4 | 1369 | |
05c2828c MT |
1370 | return ret; |
1371 | } | |
1372 | ||
9b067034 | 1373 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
1374 | { |
1375 | struct file *file = iocb->ki_filp; | |
1376 | struct tun_file *tfile = file->private_data; | |
1377 | struct tun_struct *tun = __tun_get(tfile); | |
9b067034 | 1378 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
1379 | |
1380 | if (!tun) | |
1381 | return -EBADFD; | |
9b067034 | 1382 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK); |
42404c09 | 1383 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
1384 | if (ret > 0) |
1385 | iocb->ki_pos = ret; | |
631ab46b | 1386 | tun_put(tun); |
1da177e4 LT |
1387 | return ret; |
1388 | } | |
1389 | ||
96442e42 JW |
1390 | static void tun_free_netdev(struct net_device *dev) |
1391 | { | |
1392 | struct tun_struct *tun = netdev_priv(dev); | |
1393 | ||
4008e97f | 1394 | BUG_ON(!(list_empty(&tun->disabled))); |
96442e42 | 1395 | tun_flow_uninit(tun); |
5dbbaf2d | 1396 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1397 | free_netdev(dev); |
1398 | } | |
1399 | ||
1da177e4 LT |
1400 | static void tun_setup(struct net_device *dev) |
1401 | { | |
1402 | struct tun_struct *tun = netdev_priv(dev); | |
1403 | ||
0625c883 EB |
1404 | tun->owner = INVALID_UID; |
1405 | tun->group = INVALID_GID; | |
1da177e4 | 1406 | |
1da177e4 | 1407 | dev->ethtool_ops = &tun_ethtool_ops; |
96442e42 | 1408 | dev->destructor = tun_free_netdev; |
1da177e4 LT |
1409 | } |
1410 | ||
f019a7a5 EB |
1411 | /* Trivial set of netlink ops to allow deleting tun or tap |
1412 | * device with netlink. | |
1413 | */ | |
1414 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1415 | { | |
1416 | return -EINVAL; | |
1417 | } | |
1418 | ||
1419 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
1420 | .kind = DRV_NAME, | |
1421 | .priv_size = sizeof(struct tun_struct), | |
1422 | .setup = tun_setup, | |
1423 | .validate = tun_validate, | |
1424 | }; | |
1425 | ||
33dccbb0 HX |
1426 | static void tun_sock_write_space(struct sock *sk) |
1427 | { | |
54f968d6 | 1428 | struct tun_file *tfile; |
43815482 | 1429 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
1430 | |
1431 | if (!sock_writeable(sk)) | |
1432 | return; | |
1433 | ||
33dccbb0 HX |
1434 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
1435 | return; | |
1436 | ||
43815482 ED |
1437 | wqueue = sk_sleep(sk); |
1438 | if (wqueue && waitqueue_active(wqueue)) | |
1439 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 1440 | POLLWRNORM | POLLWRBAND); |
c722c625 | 1441 | |
54f968d6 JW |
1442 | tfile = container_of(sk, struct tun_file, sk); |
1443 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
1444 | } |
1445 | ||
05c2828c MT |
1446 | static int tun_sendmsg(struct kiocb *iocb, struct socket *sock, |
1447 | struct msghdr *m, size_t total_len) | |
1448 | { | |
54f968d6 JW |
1449 | int ret; |
1450 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1451 | struct tun_struct *tun = __tun_get(tfile); | |
f5ff53b4 | 1452 | struct iov_iter from; |
54f968d6 JW |
1453 | |
1454 | if (!tun) | |
1455 | return -EBADFD; | |
f5ff53b4 AV |
1456 | |
1457 | iov_iter_init(&from, WRITE, m->msg_iov, m->msg_iovlen, total_len); | |
1458 | ret = tun_get_user(tun, tfile, m->msg_control, &from, | |
1459 | m->msg_flags & MSG_DONTWAIT); | |
54f968d6 JW |
1460 | tun_put(tun); |
1461 | return ret; | |
05c2828c MT |
1462 | } |
1463 | ||
1464 | static int tun_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1465 | struct msghdr *m, size_t total_len, | |
1466 | int flags) | |
1467 | { | |
54f968d6 JW |
1468 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
1469 | struct tun_struct *tun = __tun_get(tfile); | |
9b067034 | 1470 | struct iov_iter to; |
05c2828c | 1471 | int ret; |
54f968d6 JW |
1472 | |
1473 | if (!tun) | |
1474 | return -EBADFD; | |
1475 | ||
eda29772 | 1476 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
1477 | ret = -EINVAL; |
1478 | goto out; | |
1479 | } | |
eda29772 RC |
1480 | if (flags & MSG_ERRQUEUE) { |
1481 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
1482 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
1483 | goto out; | |
1484 | } | |
9b067034 AV |
1485 | iov_iter_init(&to, READ, m->msg_iov, m->msg_iovlen, total_len); |
1486 | ret = tun_do_read(tun, tfile, &to, flags & MSG_DONTWAIT); | |
42404c09 DM |
1487 | if (ret > total_len) { |
1488 | m->msg_flags |= MSG_TRUNC; | |
1489 | ret = flags & MSG_TRUNC ? ret : total_len; | |
1490 | } | |
3811ae76 | 1491 | out: |
54f968d6 | 1492 | tun_put(tun); |
05c2828c MT |
1493 | return ret; |
1494 | } | |
1495 | ||
1ab5ecb9 SK |
1496 | static int tun_release(struct socket *sock) |
1497 | { | |
1498 | if (sock->sk) | |
1499 | sock_put(sock->sk); | |
1500 | return 0; | |
1501 | } | |
1502 | ||
05c2828c MT |
1503 | /* Ops structure to mimic raw sockets with tun */ |
1504 | static const struct proto_ops tun_socket_ops = { | |
1505 | .sendmsg = tun_sendmsg, | |
1506 | .recvmsg = tun_recvmsg, | |
1ab5ecb9 | 1507 | .release = tun_release, |
05c2828c MT |
1508 | }; |
1509 | ||
33dccbb0 HX |
1510 | static struct proto tun_proto = { |
1511 | .name = "tun", | |
1512 | .owner = THIS_MODULE, | |
54f968d6 | 1513 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 1514 | }; |
f019a7a5 | 1515 | |
980c9e8c DW |
1516 | static int tun_flags(struct tun_struct *tun) |
1517 | { | |
1518 | int flags = 0; | |
1519 | ||
1520 | if (tun->flags & TUN_TUN_DEV) | |
1521 | flags |= IFF_TUN; | |
1522 | else | |
1523 | flags |= IFF_TAP; | |
1524 | ||
1525 | if (tun->flags & TUN_NO_PI) | |
1526 | flags |= IFF_NO_PI; | |
1527 | ||
5d097109 MT |
1528 | /* This flag has no real effect. We track the value for backwards |
1529 | * compatibility. | |
1530 | */ | |
980c9e8c DW |
1531 | if (tun->flags & TUN_ONE_QUEUE) |
1532 | flags |= IFF_ONE_QUEUE; | |
1533 | ||
1534 | if (tun->flags & TUN_VNET_HDR) | |
1535 | flags |= IFF_VNET_HDR; | |
1536 | ||
c8d68e6b JW |
1537 | if (tun->flags & TUN_TAP_MQ) |
1538 | flags |= IFF_MULTI_QUEUE; | |
1539 | ||
274038f8 PE |
1540 | if (tun->flags & TUN_PERSIST) |
1541 | flags |= IFF_PERSIST; | |
1542 | ||
980c9e8c DW |
1543 | return flags; |
1544 | } | |
1545 | ||
1546 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
1547 | char *buf) | |
1548 | { | |
1549 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
1550 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
1551 | } | |
1552 | ||
1553 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
1554 | char *buf) | |
1555 | { | |
1556 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1557 | return uid_valid(tun->owner)? |
1558 | sprintf(buf, "%u\n", | |
1559 | from_kuid_munged(current_user_ns(), tun->owner)): | |
1560 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1561 | } |
1562 | ||
1563 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
1564 | char *buf) | |
1565 | { | |
1566 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1567 | return gid_valid(tun->group) ? |
1568 | sprintf(buf, "%u\n", | |
1569 | from_kgid_munged(current_user_ns(), tun->group)): | |
1570 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1571 | } |
1572 | ||
1573 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
1574 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
1575 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
1576 | ||
d647a591 | 1577 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
1578 | { |
1579 | struct tun_struct *tun; | |
54f968d6 | 1580 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
1581 | struct net_device *dev; |
1582 | int err; | |
1583 | ||
7c0c3b1a JW |
1584 | if (tfile->detached) |
1585 | return -EINVAL; | |
1586 | ||
74a3e5a7 EB |
1587 | dev = __dev_get_by_name(net, ifr->ifr_name); |
1588 | if (dev) { | |
f85ba780 DW |
1589 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
1590 | return -EBUSY; | |
74a3e5a7 EB |
1591 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
1592 | tun = netdev_priv(dev); | |
1593 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
1594 | tun = netdev_priv(dev); | |
1595 | else | |
1596 | return -EINVAL; | |
1597 | ||
8e6d91ae JW |
1598 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
1599 | !!(tun->flags & TUN_TAP_MQ)) | |
1600 | return -EINVAL; | |
1601 | ||
cde8b15f | 1602 | if (tun_not_capable(tun)) |
2b980dbd | 1603 | return -EPERM; |
5dbbaf2d | 1604 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
1605 | if (err < 0) |
1606 | return err; | |
1607 | ||
849c9b6f | 1608 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER); |
a7385ba2 EB |
1609 | if (err < 0) |
1610 | return err; | |
4008e97f JW |
1611 | |
1612 | if (tun->flags & TUN_TAP_MQ && | |
e8dbad66 JW |
1613 | (tun->numqueues + tun->numdisabled > 1)) { |
1614 | /* One or more queue has already been attached, no need | |
1615 | * to initialize the device again. | |
1616 | */ | |
1617 | return 0; | |
1618 | } | |
6aa20a22 | 1619 | } |
1da177e4 LT |
1620 | else { |
1621 | char *name; | |
1622 | unsigned long flags = 0; | |
edfb6a14 JW |
1623 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
1624 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 1625 | |
c260b772 | 1626 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 1627 | return -EPERM; |
2b980dbd PM |
1628 | err = security_tun_dev_create(); |
1629 | if (err < 0) | |
1630 | return err; | |
ca6bb5d7 | 1631 | |
1da177e4 LT |
1632 | /* Set dev type */ |
1633 | if (ifr->ifr_flags & IFF_TUN) { | |
1634 | /* TUN device */ | |
1635 | flags |= TUN_TUN_DEV; | |
1636 | name = "tun%d"; | |
1637 | } else if (ifr->ifr_flags & IFF_TAP) { | |
1638 | /* TAP device */ | |
1639 | flags |= TUN_TAP_DEV; | |
1640 | name = "tap%d"; | |
6aa20a22 | 1641 | } else |
36989b90 | 1642 | return -EINVAL; |
6aa20a22 | 1643 | |
1da177e4 LT |
1644 | if (*ifr->ifr_name) |
1645 | name = ifr->ifr_name; | |
1646 | ||
c8d68e6b | 1647 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
1648 | NET_NAME_UNKNOWN, tun_setup, queues, |
1649 | queues); | |
edfb6a14 | 1650 | |
1da177e4 LT |
1651 | if (!dev) |
1652 | return -ENOMEM; | |
1653 | ||
fc54c658 | 1654 | dev_net_set(dev, net); |
f019a7a5 | 1655 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 1656 | dev->ifindex = tfile->ifindex; |
758e43b7 | 1657 | |
1da177e4 LT |
1658 | tun = netdev_priv(dev); |
1659 | tun->dev = dev; | |
1660 | tun->flags = flags; | |
f271b2cc | 1661 | tun->txflt.count = 0; |
d9d52b51 | 1662 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 1663 | |
54f968d6 JW |
1664 | tun->filter_attached = false; |
1665 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
33dccbb0 | 1666 | |
96442e42 JW |
1667 | spin_lock_init(&tun->lock); |
1668 | ||
5dbbaf2d PM |
1669 | err = security_tun_dev_alloc_security(&tun->security); |
1670 | if (err < 0) | |
1671 | goto err_free_dev; | |
2b980dbd | 1672 | |
1da177e4 | 1673 | tun_net_init(dev); |
944a1376 | 1674 | tun_flow_init(tun); |
96442e42 | 1675 | |
88255375 | 1676 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
1677 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
1678 | NETIF_F_HW_VLAN_STAG_TX; | |
88255375 | 1679 | dev->features = dev->hw_features; |
6671b224 FLVC |
1680 | dev->vlan_features = dev->features & |
1681 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
1682 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 1683 | |
4008e97f | 1684 | INIT_LIST_HEAD(&tun->disabled); |
849c9b6f | 1685 | err = tun_attach(tun, file, false); |
eb0fb363 | 1686 | if (err < 0) |
662ca437 | 1687 | goto err_free_flow; |
eb0fb363 | 1688 | |
1da177e4 LT |
1689 | err = register_netdevice(tun->dev); |
1690 | if (err < 0) | |
662ca437 | 1691 | goto err_detach; |
9c3fea6a | 1692 | |
980c9e8c DW |
1693 | if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) || |
1694 | device_create_file(&tun->dev->dev, &dev_attr_owner) || | |
1695 | device_create_file(&tun->dev->dev, &dev_attr_group)) | |
6b8a66ee | 1696 | pr_err("Failed to create tun sysfs files\n"); |
1da177e4 LT |
1697 | } |
1698 | ||
af668b3c MT |
1699 | netif_carrier_on(tun->dev); |
1700 | ||
6b8a66ee | 1701 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 LT |
1702 | |
1703 | if (ifr->ifr_flags & IFF_NO_PI) | |
1704 | tun->flags |= TUN_NO_PI; | |
a26af1e0 NF |
1705 | else |
1706 | tun->flags &= ~TUN_NO_PI; | |
1da177e4 | 1707 | |
5d097109 MT |
1708 | /* This flag has no real effect. We track the value for backwards |
1709 | * compatibility. | |
1710 | */ | |
1da177e4 LT |
1711 | if (ifr->ifr_flags & IFF_ONE_QUEUE) |
1712 | tun->flags |= TUN_ONE_QUEUE; | |
a26af1e0 NF |
1713 | else |
1714 | tun->flags &= ~TUN_ONE_QUEUE; | |
1da177e4 | 1715 | |
f43798c2 RR |
1716 | if (ifr->ifr_flags & IFF_VNET_HDR) |
1717 | tun->flags |= TUN_VNET_HDR; | |
1718 | else | |
1719 | tun->flags &= ~TUN_VNET_HDR; | |
1720 | ||
c8d68e6b JW |
1721 | if (ifr->ifr_flags & IFF_MULTI_QUEUE) |
1722 | tun->flags |= TUN_TAP_MQ; | |
1723 | else | |
1724 | tun->flags &= ~TUN_TAP_MQ; | |
1725 | ||
e35259a9 MK |
1726 | /* Make sure persistent devices do not get stuck in |
1727 | * xoff state. | |
1728 | */ | |
1729 | if (netif_running(tun->dev)) | |
c8d68e6b | 1730 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 1731 | |
1da177e4 LT |
1732 | strcpy(ifr->ifr_name, tun->dev->name); |
1733 | return 0; | |
1734 | ||
662ca437 JW |
1735 | err_detach: |
1736 | tun_detach_all(dev); | |
1737 | err_free_flow: | |
1738 | tun_flow_uninit(tun); | |
1739 | security_tun_dev_free_security(tun->security); | |
1740 | err_free_dev: | |
1da177e4 | 1741 | free_netdev(dev); |
1da177e4 LT |
1742 | return err; |
1743 | } | |
1744 | ||
9ce99cf6 | 1745 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 1746 | struct ifreq *ifr) |
e3b99556 | 1747 | { |
6b8a66ee | 1748 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
1749 | |
1750 | strcpy(ifr->ifr_name, tun->dev->name); | |
1751 | ||
980c9e8c | 1752 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 1753 | |
e3b99556 MM |
1754 | } |
1755 | ||
5228ddc9 RR |
1756 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
1757 | * privs required. */ | |
88255375 | 1758 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 1759 | { |
c8f44aff | 1760 | netdev_features_t features = 0; |
5228ddc9 RR |
1761 | |
1762 | if (arg & TUN_F_CSUM) { | |
88255375 | 1763 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
1764 | arg &= ~TUN_F_CSUM; |
1765 | ||
1766 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
1767 | if (arg & TUN_F_TSO_ECN) { | |
1768 | features |= NETIF_F_TSO_ECN; | |
1769 | arg &= ~TUN_F_TSO_ECN; | |
1770 | } | |
1771 | if (arg & TUN_F_TSO4) | |
1772 | features |= NETIF_F_TSO; | |
1773 | if (arg & TUN_F_TSO6) | |
1774 | features |= NETIF_F_TSO6; | |
1775 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
1776 | } | |
1777 | } | |
1778 | ||
1779 | /* This gives the user a way to test for new features in future by | |
1780 | * trying to set them. */ | |
1781 | if (arg) | |
1782 | return -EINVAL; | |
1783 | ||
88255375 MM |
1784 | tun->set_features = features; |
1785 | netdev_update_features(tun->dev); | |
5228ddc9 RR |
1786 | |
1787 | return 0; | |
1788 | } | |
1789 | ||
c8d68e6b JW |
1790 | static void tun_detach_filter(struct tun_struct *tun, int n) |
1791 | { | |
1792 | int i; | |
1793 | struct tun_file *tfile; | |
1794 | ||
1795 | for (i = 0; i < n; i++) { | |
b8deabd3 | 1796 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1797 | sk_detach_filter(tfile->socket.sk); |
1798 | } | |
1799 | ||
1800 | tun->filter_attached = false; | |
1801 | } | |
1802 | ||
1803 | static int tun_attach_filter(struct tun_struct *tun) | |
1804 | { | |
1805 | int i, ret = 0; | |
1806 | struct tun_file *tfile; | |
1807 | ||
1808 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1809 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1810 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); |
1811 | if (ret) { | |
1812 | tun_detach_filter(tun, i); | |
1813 | return ret; | |
1814 | } | |
1815 | } | |
1816 | ||
1817 | tun->filter_attached = true; | |
1818 | return ret; | |
1819 | } | |
1820 | ||
1821 | static void tun_set_sndbuf(struct tun_struct *tun) | |
1822 | { | |
1823 | struct tun_file *tfile; | |
1824 | int i; | |
1825 | ||
1826 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1827 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1828 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
1829 | } | |
1830 | } | |
1831 | ||
cde8b15f JW |
1832 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
1833 | { | |
1834 | struct tun_file *tfile = file->private_data; | |
1835 | struct tun_struct *tun; | |
cde8b15f JW |
1836 | int ret = 0; |
1837 | ||
1838 | rtnl_lock(); | |
1839 | ||
1840 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 1841 | tun = tfile->detached; |
5dbbaf2d | 1842 | if (!tun) { |
cde8b15f | 1843 | ret = -EINVAL; |
5dbbaf2d PM |
1844 | goto unlock; |
1845 | } | |
1846 | ret = security_tun_dev_attach_queue(tun->security); | |
1847 | if (ret < 0) | |
1848 | goto unlock; | |
849c9b6f | 1849 | ret = tun_attach(tun, file, false); |
4008e97f | 1850 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 1851 | tun = rtnl_dereference(tfile->tun); |
9e85722d | 1852 | if (!tun || !(tun->flags & TUN_TAP_MQ) || tfile->detached) |
4008e97f JW |
1853 | ret = -EINVAL; |
1854 | else | |
1855 | __tun_detach(tfile, false); | |
1856 | } else | |
cde8b15f JW |
1857 | ret = -EINVAL; |
1858 | ||
5dbbaf2d | 1859 | unlock: |
cde8b15f JW |
1860 | rtnl_unlock(); |
1861 | return ret; | |
1862 | } | |
1863 | ||
50857e2a AB |
1864 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
1865 | unsigned long arg, int ifreq_len) | |
1da177e4 | 1866 | { |
36b50bab | 1867 | struct tun_file *tfile = file->private_data; |
631ab46b | 1868 | struct tun_struct *tun; |
1da177e4 LT |
1869 | void __user* argp = (void __user*)arg; |
1870 | struct ifreq ifr; | |
0625c883 EB |
1871 | kuid_t owner; |
1872 | kgid_t group; | |
33dccbb0 | 1873 | int sndbuf; |
d9d52b51 | 1874 | int vnet_hdr_sz; |
fb7589a1 | 1875 | unsigned int ifindex; |
f271b2cc | 1876 | int ret; |
1da177e4 | 1877 | |
cde8b15f | 1878 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) { |
50857e2a | 1879 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 1880 | return -EFAULT; |
8bbb1813 | 1881 | } else { |
a117dacd | 1882 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 1883 | } |
631ab46b EB |
1884 | if (cmd == TUNGETFEATURES) { |
1885 | /* Currently this just means: "what IFF flags are valid?". | |
1886 | * This is needed because we never checked for invalid flags on | |
1887 | * TUNSETIFF. */ | |
1888 | return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE | | |
cde8b15f | 1889 | IFF_VNET_HDR | IFF_MULTI_QUEUE, |
631ab46b | 1890 | (unsigned int __user*)argp); |
cde8b15f JW |
1891 | } else if (cmd == TUNSETQUEUE) |
1892 | return tun_set_queue(file, &ifr); | |
631ab46b | 1893 | |
c8d68e6b | 1894 | ret = 0; |
876bfd4d HX |
1895 | rtnl_lock(); |
1896 | ||
36b50bab | 1897 | tun = __tun_get(tfile); |
1da177e4 | 1898 | if (cmd == TUNSETIFF && !tun) { |
1da177e4 LT |
1899 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
1900 | ||
876bfd4d | 1901 | ret = tun_set_iff(tfile->net, file, &ifr); |
1da177e4 | 1902 | |
876bfd4d HX |
1903 | if (ret) |
1904 | goto unlock; | |
1da177e4 | 1905 | |
50857e2a | 1906 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
1907 | ret = -EFAULT; |
1908 | goto unlock; | |
1da177e4 | 1909 | } |
fb7589a1 PE |
1910 | if (cmd == TUNSETIFINDEX) { |
1911 | ret = -EPERM; | |
1912 | if (tun) | |
1913 | goto unlock; | |
1914 | ||
1915 | ret = -EFAULT; | |
1916 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
1917 | goto unlock; | |
1918 | ||
1919 | ret = 0; | |
1920 | tfile->ifindex = ifindex; | |
1921 | goto unlock; | |
1922 | } | |
1da177e4 | 1923 | |
876bfd4d | 1924 | ret = -EBADFD; |
1da177e4 | 1925 | if (!tun) |
876bfd4d | 1926 | goto unlock; |
1da177e4 | 1927 | |
1e588338 | 1928 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 1929 | |
631ab46b | 1930 | ret = 0; |
1da177e4 | 1931 | switch (cmd) { |
e3b99556 | 1932 | case TUNGETIFF: |
9ce99cf6 | 1933 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 1934 | |
3d407a80 PE |
1935 | if (tfile->detached) |
1936 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
1937 | if (!tfile->socket.sk->sk_filter) |
1938 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 1939 | |
50857e2a | 1940 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 1941 | ret = -EFAULT; |
e3b99556 MM |
1942 | break; |
1943 | ||
1da177e4 LT |
1944 | case TUNSETNOCSUM: |
1945 | /* Disable/Enable checksum */ | |
1da177e4 | 1946 | |
88255375 MM |
1947 | /* [unimplemented] */ |
1948 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 1949 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
1950 | break; |
1951 | ||
1952 | case TUNSETPERSIST: | |
54f968d6 JW |
1953 | /* Disable/Enable persist mode. Keep an extra reference to the |
1954 | * module to prevent the module being unprobed. | |
1955 | */ | |
dd38bd85 | 1956 | if (arg && !(tun->flags & TUN_PERSIST)) { |
1da177e4 | 1957 | tun->flags |= TUN_PERSIST; |
54f968d6 | 1958 | __module_get(THIS_MODULE); |
dd38bd85 JW |
1959 | } |
1960 | if (!arg && (tun->flags & TUN_PERSIST)) { | |
1da177e4 | 1961 | tun->flags &= ~TUN_PERSIST; |
54f968d6 JW |
1962 | module_put(THIS_MODULE); |
1963 | } | |
1da177e4 | 1964 | |
6b8a66ee JP |
1965 | tun_debug(KERN_INFO, tun, "persist %s\n", |
1966 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
1967 | break; |
1968 | ||
1969 | case TUNSETOWNER: | |
1970 | /* Set owner of the device */ | |
0625c883 EB |
1971 | owner = make_kuid(current_user_ns(), arg); |
1972 | if (!uid_valid(owner)) { | |
1973 | ret = -EINVAL; | |
1974 | break; | |
1975 | } | |
1976 | tun->owner = owner; | |
1e588338 | 1977 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 1978 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
1979 | break; |
1980 | ||
8c644623 GG |
1981 | case TUNSETGROUP: |
1982 | /* Set group of the device */ | |
0625c883 EB |
1983 | group = make_kgid(current_user_ns(), arg); |
1984 | if (!gid_valid(group)) { | |
1985 | ret = -EINVAL; | |
1986 | break; | |
1987 | } | |
1988 | tun->group = group; | |
1e588338 | 1989 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 1990 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
1991 | break; |
1992 | ||
ff4cc3ac MK |
1993 | case TUNSETLINK: |
1994 | /* Only allow setting the type when the interface is down */ | |
1995 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
1996 | tun_debug(KERN_INFO, tun, |
1997 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 1998 | ret = -EBUSY; |
ff4cc3ac MK |
1999 | } else { |
2000 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2001 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2002 | tun->dev->type); | |
48abfe05 | 2003 | ret = 0; |
ff4cc3ac | 2004 | } |
631ab46b | 2005 | break; |
ff4cc3ac | 2006 | |
1da177e4 LT |
2007 | #ifdef TUN_DEBUG |
2008 | case TUNSETDEBUG: | |
2009 | tun->debug = arg; | |
2010 | break; | |
2011 | #endif | |
5228ddc9 | 2012 | case TUNSETOFFLOAD: |
88255375 | 2013 | ret = set_offload(tun, arg); |
631ab46b | 2014 | break; |
5228ddc9 | 2015 | |
f271b2cc MK |
2016 | case TUNSETTXFILTER: |
2017 | /* Can be set only for TAPs */ | |
631ab46b | 2018 | ret = -EINVAL; |
f271b2cc | 2019 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) |
631ab46b | 2020 | break; |
c0e5a8c2 | 2021 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2022 | break; |
1da177e4 LT |
2023 | |
2024 | case SIOCGIFHWADDR: | |
b595076a | 2025 | /* Get hw address */ |
f271b2cc MK |
2026 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2027 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2028 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2029 | ret = -EFAULT; |
2030 | break; | |
1da177e4 LT |
2031 | |
2032 | case SIOCSIFHWADDR: | |
f271b2cc | 2033 | /* Set hw address */ |
6b8a66ee JP |
2034 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2035 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2036 | |
40102371 | 2037 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2038 | break; |
33dccbb0 HX |
2039 | |
2040 | case TUNGETSNDBUF: | |
54f968d6 | 2041 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2042 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2043 | ret = -EFAULT; | |
2044 | break; | |
2045 | ||
2046 | case TUNSETSNDBUF: | |
2047 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2048 | ret = -EFAULT; | |
2049 | break; | |
2050 | } | |
2051 | ||
c8d68e6b JW |
2052 | tun->sndbuf = sndbuf; |
2053 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2054 | break; |
2055 | ||
d9d52b51 MT |
2056 | case TUNGETVNETHDRSZ: |
2057 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2058 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2059 | ret = -EFAULT; | |
2060 | break; | |
2061 | ||
2062 | case TUNSETVNETHDRSZ: | |
2063 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2064 | ret = -EFAULT; | |
2065 | break; | |
2066 | } | |
2067 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2068 | ret = -EINVAL; | |
2069 | break; | |
2070 | } | |
2071 | ||
2072 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2073 | break; | |
2074 | ||
99405162 MT |
2075 | case TUNATTACHFILTER: |
2076 | /* Can be set only for TAPs */ | |
2077 | ret = -EINVAL; | |
2078 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2079 | break; | |
2080 | ret = -EFAULT; | |
54f968d6 | 2081 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2082 | break; |
2083 | ||
c8d68e6b | 2084 | ret = tun_attach_filter(tun); |
99405162 MT |
2085 | break; |
2086 | ||
2087 | case TUNDETACHFILTER: | |
2088 | /* Can be set only for TAPs */ | |
2089 | ret = -EINVAL; | |
2090 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2091 | break; | |
c8d68e6b JW |
2092 | ret = 0; |
2093 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2094 | break; |
2095 | ||
76975e9c PE |
2096 | case TUNGETFILTER: |
2097 | ret = -EINVAL; | |
2098 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2099 | break; | |
2100 | ret = -EFAULT; | |
2101 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2102 | break; | |
2103 | ret = 0; | |
2104 | break; | |
2105 | ||
1da177e4 | 2106 | default: |
631ab46b EB |
2107 | ret = -EINVAL; |
2108 | break; | |
ee289b64 | 2109 | } |
1da177e4 | 2110 | |
876bfd4d HX |
2111 | unlock: |
2112 | rtnl_unlock(); | |
2113 | if (tun) | |
2114 | tun_put(tun); | |
631ab46b | 2115 | return ret; |
1da177e4 LT |
2116 | } |
2117 | ||
50857e2a AB |
2118 | static long tun_chr_ioctl(struct file *file, |
2119 | unsigned int cmd, unsigned long arg) | |
2120 | { | |
2121 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2122 | } | |
2123 | ||
2124 | #ifdef CONFIG_COMPAT | |
2125 | static long tun_chr_compat_ioctl(struct file *file, | |
2126 | unsigned int cmd, unsigned long arg) | |
2127 | { | |
2128 | switch (cmd) { | |
2129 | case TUNSETIFF: | |
2130 | case TUNGETIFF: | |
2131 | case TUNSETTXFILTER: | |
2132 | case TUNGETSNDBUF: | |
2133 | case TUNSETSNDBUF: | |
2134 | case SIOCGIFHWADDR: | |
2135 | case SIOCSIFHWADDR: | |
2136 | arg = (unsigned long)compat_ptr(arg); | |
2137 | break; | |
2138 | default: | |
2139 | arg = (compat_ulong_t)arg; | |
2140 | break; | |
2141 | } | |
2142 | ||
2143 | /* | |
2144 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2145 | * the end of that structure. All fields that are used in this | |
2146 | * driver are compatible though, we don't need to convert the | |
2147 | * contents. | |
2148 | */ | |
2149 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2150 | } | |
2151 | #endif /* CONFIG_COMPAT */ | |
2152 | ||
1da177e4 LT |
2153 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2154 | { | |
54f968d6 | 2155 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2156 | int ret; |
2157 | ||
54f968d6 | 2158 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2159 | goto out; |
6aa20a22 | 2160 | |
1da177e4 | 2161 | if (on) { |
e0b93edd | 2162 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2163 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2164 | } else |
54f968d6 | 2165 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2166 | ret = 0; |
2167 | out: | |
9d319522 | 2168 | return ret; |
1da177e4 LT |
2169 | } |
2170 | ||
2171 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2172 | { | |
631ab46b | 2173 | struct tun_file *tfile; |
deed49fb | 2174 | |
6b8a66ee | 2175 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2176 | |
54f968d6 JW |
2177 | tfile = (struct tun_file *)sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL, |
2178 | &tun_proto); | |
631ab46b EB |
2179 | if (!tfile) |
2180 | return -ENOMEM; | |
c956674b | 2181 | RCU_INIT_POINTER(tfile->tun, NULL); |
36b50bab | 2182 | tfile->net = get_net(current->nsproxy->net_ns); |
54f968d6 | 2183 | tfile->flags = 0; |
fb7589a1 | 2184 | tfile->ifindex = 0; |
54f968d6 | 2185 | |
54f968d6 | 2186 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2187 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2188 | |
2189 | tfile->socket.file = file; | |
2190 | tfile->socket.ops = &tun_socket_ops; | |
2191 | ||
2192 | sock_init_data(&tfile->socket, &tfile->sk); | |
2193 | sk_change_net(&tfile->sk, tfile->net); | |
2194 | ||
2195 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2196 | tfile->sk.sk_sndbuf = INT_MAX; | |
2197 | ||
631ab46b | 2198 | file->private_data = tfile; |
54f968d6 | 2199 | set_bit(SOCK_EXTERNALLY_ALLOCATED, &tfile->socket.flags); |
4008e97f | 2200 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2201 | |
19a6afb2 JW |
2202 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2203 | ||
1da177e4 LT |
2204 | return 0; |
2205 | } | |
2206 | ||
2207 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2208 | { | |
631ab46b | 2209 | struct tun_file *tfile = file->private_data; |
54f968d6 | 2210 | struct net *net = tfile->net; |
1da177e4 | 2211 | |
c8d68e6b | 2212 | tun_detach(tfile, true); |
54f968d6 | 2213 | put_net(net); |
1da177e4 LT |
2214 | |
2215 | return 0; | |
2216 | } | |
2217 | ||
93e14b6d MY |
2218 | #ifdef CONFIG_PROC_FS |
2219 | static int tun_chr_show_fdinfo(struct seq_file *m, struct file *f) | |
2220 | { | |
2221 | struct tun_struct *tun; | |
2222 | struct ifreq ifr; | |
2223 | ||
2224 | memset(&ifr, 0, sizeof(ifr)); | |
2225 | ||
2226 | rtnl_lock(); | |
2227 | tun = tun_get(f); | |
2228 | if (tun) | |
2229 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2230 | rtnl_unlock(); | |
2231 | ||
2232 | if (tun) | |
2233 | tun_put(tun); | |
2234 | ||
2235 | return seq_printf(m, "iff:\t%s\n", ifr.ifr_name); | |
2236 | } | |
2237 | #endif | |
2238 | ||
d54b1fdb | 2239 | static const struct file_operations tun_fops = { |
6aa20a22 | 2240 | .owner = THIS_MODULE, |
1da177e4 | 2241 | .llseek = no_llseek, |
9b067034 | 2242 | .read = new_sync_read, |
f5ff53b4 | 2243 | .write = new_sync_write, |
9b067034 | 2244 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 2245 | .write_iter = tun_chr_write_iter, |
1da177e4 | 2246 | .poll = tun_chr_poll, |
50857e2a AB |
2247 | .unlocked_ioctl = tun_chr_ioctl, |
2248 | #ifdef CONFIG_COMPAT | |
2249 | .compat_ioctl = tun_chr_compat_ioctl, | |
2250 | #endif | |
1da177e4 LT |
2251 | .open = tun_chr_open, |
2252 | .release = tun_chr_close, | |
93e14b6d MY |
2253 | .fasync = tun_chr_fasync, |
2254 | #ifdef CONFIG_PROC_FS | |
2255 | .show_fdinfo = tun_chr_show_fdinfo, | |
2256 | #endif | |
1da177e4 LT |
2257 | }; |
2258 | ||
2259 | static struct miscdevice tun_miscdev = { | |
2260 | .minor = TUN_MINOR, | |
2261 | .name = "tun", | |
e454cea2 | 2262 | .nodename = "net/tun", |
1da177e4 | 2263 | .fops = &tun_fops, |
1da177e4 LT |
2264 | }; |
2265 | ||
2266 | /* ethtool interface */ | |
2267 | ||
2268 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) | |
2269 | { | |
2270 | cmd->supported = 0; | |
2271 | cmd->advertising = 0; | |
70739497 | 2272 | ethtool_cmd_speed_set(cmd, SPEED_10); |
1da177e4 LT |
2273 | cmd->duplex = DUPLEX_FULL; |
2274 | cmd->port = PORT_TP; | |
2275 | cmd->phy_address = 0; | |
2276 | cmd->transceiver = XCVR_INTERNAL; | |
2277 | cmd->autoneg = AUTONEG_DISABLE; | |
2278 | cmd->maxtxpkt = 0; | |
2279 | cmd->maxrxpkt = 0; | |
2280 | return 0; | |
2281 | } | |
2282 | ||
2283 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2284 | { | |
2285 | struct tun_struct *tun = netdev_priv(dev); | |
2286 | ||
33a5ba14 RJ |
2287 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2288 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2289 | |
2290 | switch (tun->flags & TUN_TYPE_MASK) { | |
2291 | case TUN_TUN_DEV: | |
33a5ba14 | 2292 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 LT |
2293 | break; |
2294 | case TUN_TAP_DEV: | |
33a5ba14 | 2295 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2296 | break; |
2297 | } | |
2298 | } | |
2299 | ||
2300 | static u32 tun_get_msglevel(struct net_device *dev) | |
2301 | { | |
2302 | #ifdef TUN_DEBUG | |
2303 | struct tun_struct *tun = netdev_priv(dev); | |
2304 | return tun->debug; | |
2305 | #else | |
2306 | return -EOPNOTSUPP; | |
2307 | #endif | |
2308 | } | |
2309 | ||
2310 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2311 | { | |
2312 | #ifdef TUN_DEBUG | |
2313 | struct tun_struct *tun = netdev_priv(dev); | |
2314 | tun->debug = value; | |
2315 | #endif | |
2316 | } | |
2317 | ||
7282d491 | 2318 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2319 | .get_settings = tun_get_settings, |
2320 | .get_drvinfo = tun_get_drvinfo, | |
2321 | .get_msglevel = tun_get_msglevel, | |
2322 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2323 | .get_link = ethtool_op_get_link, |
eda29772 | 2324 | .get_ts_info = ethtool_op_get_ts_info, |
1da177e4 LT |
2325 | }; |
2326 | ||
79d17604 | 2327 | |
1da177e4 LT |
2328 | static int __init tun_init(void) |
2329 | { | |
2330 | int ret = 0; | |
2331 | ||
6b8a66ee JP |
2332 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
2333 | pr_info("%s\n", DRV_COPYRIGHT); | |
1da177e4 | 2334 | |
f019a7a5 | 2335 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 2336 | if (ret) { |
6b8a66ee | 2337 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 2338 | goto err_linkops; |
79d17604 PE |
2339 | } |
2340 | ||
1da177e4 | 2341 | ret = misc_register(&tun_miscdev); |
79d17604 | 2342 | if (ret) { |
6b8a66ee | 2343 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
2344 | goto err_misc; |
2345 | } | |
f019a7a5 | 2346 | return 0; |
79d17604 | 2347 | err_misc: |
f019a7a5 EB |
2348 | rtnl_link_unregister(&tun_link_ops); |
2349 | err_linkops: | |
1da177e4 LT |
2350 | return ret; |
2351 | } | |
2352 | ||
2353 | static void tun_cleanup(void) | |
2354 | { | |
6aa20a22 | 2355 | misc_deregister(&tun_miscdev); |
f019a7a5 | 2356 | rtnl_link_unregister(&tun_link_ops); |
1da177e4 LT |
2357 | } |
2358 | ||
05c2828c MT |
2359 | /* Get an underlying socket object from tun file. Returns error unless file is |
2360 | * attached to a device. The returned object works like a packet socket, it | |
2361 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
2362 | * holding a reference to the file for as long as the socket is in use. */ | |
2363 | struct socket *tun_get_socket(struct file *file) | |
2364 | { | |
6e914fc7 | 2365 | struct tun_file *tfile; |
05c2828c MT |
2366 | if (file->f_op != &tun_fops) |
2367 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
2368 | tfile = file->private_data; |
2369 | if (!tfile) | |
05c2828c | 2370 | return ERR_PTR(-EBADFD); |
54f968d6 | 2371 | return &tfile->socket; |
05c2828c MT |
2372 | } |
2373 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
2374 | ||
1da177e4 LT |
2375 | module_init(tun_init); |
2376 | module_exit(tun_cleanup); | |
2377 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
2378 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
2379 | MODULE_LICENSE("GPL"); | |
2380 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 2381 | MODULE_ALIAS("devname:net/tun"); |