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> |
1da177e4 | 74 | |
1da177e4 LT |
75 | #include <asm/uaccess.h> |
76 | ||
14daa021 RR |
77 | /* Uncomment to enable debugging */ |
78 | /* #define TUN_DEBUG 1 */ | |
79 | ||
1da177e4 LT |
80 | #ifdef TUN_DEBUG |
81 | static int debug; | |
14daa021 | 82 | |
6b8a66ee JP |
83 | #define tun_debug(level, tun, fmt, args...) \ |
84 | do { \ | |
85 | if (tun->debug) \ | |
86 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
87 | } while (0) | |
88 | #define DBG1(level, fmt, args...) \ | |
89 | do { \ | |
90 | if (debug == 2) \ | |
91 | printk(level fmt, ##args); \ | |
92 | } while (0) | |
14daa021 | 93 | #else |
6b8a66ee JP |
94 | #define tun_debug(level, tun, fmt, args...) \ |
95 | do { \ | |
96 | if (0) \ | |
97 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
98 | } while (0) | |
99 | #define DBG1(level, fmt, args...) \ | |
100 | do { \ | |
101 | if (0) \ | |
102 | printk(level fmt, ##args); \ | |
103 | } while (0) | |
14daa021 RR |
104 | #endif |
105 | ||
031f5e03 MT |
106 | /* TUN device flags */ |
107 | ||
108 | /* IFF_ATTACH_QUEUE is never stored in device flags, | |
109 | * overload it to mean fasync when stored there. | |
110 | */ | |
111 | #define TUN_FASYNC IFF_ATTACH_QUEUE | |
112 | #define TUN_NO_PI IFF_NO_PI | |
113 | /* This flag has no real effect */ | |
114 | #define TUN_ONE_QUEUE IFF_ONE_QUEUE | |
115 | #define TUN_PERSIST IFF_PERSIST | |
116 | #define TUN_VNET_HDR IFF_VNET_HDR | |
117 | #define TUN_TAP_MQ IFF_MULTI_QUEUE | |
118 | ||
119 | #define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \ | |
120 | IFF_MULTI_QUEUE) | |
0690899b MT |
121 | #define GOODCOPY_LEN 128 |
122 | ||
f271b2cc MK |
123 | #define FLT_EXACT_COUNT 8 |
124 | struct tap_filter { | |
125 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
126 | u32 mask[2]; /* Mask of the hashed addrs */ | |
127 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
128 | }; | |
129 | ||
92d4ea6e | 130 | /* DEFAULT_MAX_NUM_RSS_QUEUES were chosen to let the rx/tx queues allocated for |
edfb6a14 JW |
131 | * the netdevice to be fit in one page. So we can make sure the success of |
132 | * memory allocation. TODO: increase the limit. */ | |
133 | #define MAX_TAP_QUEUES DEFAULT_MAX_NUM_RSS_QUEUES | |
b8732fb7 | 134 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 135 | |
96442e42 JW |
136 | #define TUN_FLOW_EXPIRE (3 * HZ) |
137 | ||
54f968d6 | 138 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 139 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
140 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
141 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 142 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 143 | * this). |
6e914fc7 JW |
144 | * |
145 | * RCU usage: | |
36fe8c09 | 146 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 147 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 148 | */ |
631ab46b | 149 | struct tun_file { |
54f968d6 JW |
150 | struct sock sk; |
151 | struct socket socket; | |
152 | struct socket_wq wq; | |
6e914fc7 | 153 | struct tun_struct __rcu *tun; |
36b50bab | 154 | struct net *net; |
54f968d6 JW |
155 | struct fasync_struct *fasync; |
156 | /* only used for fasnyc */ | |
157 | unsigned int flags; | |
fb7589a1 PE |
158 | union { |
159 | u16 queue_index; | |
160 | unsigned int ifindex; | |
161 | }; | |
4008e97f JW |
162 | struct list_head next; |
163 | struct tun_struct *detached; | |
631ab46b EB |
164 | }; |
165 | ||
96442e42 JW |
166 | struct tun_flow_entry { |
167 | struct hlist_node hash_link; | |
168 | struct rcu_head rcu; | |
169 | struct tun_struct *tun; | |
170 | ||
171 | u32 rxhash; | |
9bc88939 | 172 | u32 rps_rxhash; |
96442e42 JW |
173 | int queue_index; |
174 | unsigned long updated; | |
175 | }; | |
176 | ||
177 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
178 | ||
54f968d6 | 179 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 180 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
181 | * file were attached to a persist device. |
182 | */ | |
14daa021 | 183 | struct tun_struct { |
c8d68e6b JW |
184 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
185 | unsigned int numqueues; | |
f271b2cc | 186 | unsigned int flags; |
0625c883 EB |
187 | kuid_t owner; |
188 | kgid_t group; | |
14daa021 | 189 | |
14daa021 | 190 | struct net_device *dev; |
c8f44aff | 191 | netdev_features_t set_features; |
88255375 | 192 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
3d0ad094 | 193 | NETIF_F_TSO6) |
d9d52b51 MT |
194 | |
195 | int vnet_hdr_sz; | |
54f968d6 JW |
196 | int sndbuf; |
197 | struct tap_filter txflt; | |
198 | struct sock_fprog fprog; | |
199 | /* protected by rtnl lock */ | |
200 | bool filter_attached; | |
14daa021 RR |
201 | #ifdef TUN_DEBUG |
202 | int debug; | |
1da177e4 | 203 | #endif |
96442e42 | 204 | spinlock_t lock; |
96442e42 JW |
205 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
206 | struct timer_list flow_gc_timer; | |
207 | unsigned long ageing_time; | |
4008e97f JW |
208 | unsigned int numdisabled; |
209 | struct list_head disabled; | |
5dbbaf2d | 210 | void *security; |
b8732fb7 | 211 | u32 flow_count; |
14daa021 | 212 | }; |
1da177e4 | 213 | |
96442e42 JW |
214 | static inline u32 tun_hashfn(u32 rxhash) |
215 | { | |
216 | return rxhash & 0x3ff; | |
217 | } | |
218 | ||
219 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
220 | { | |
221 | struct tun_flow_entry *e; | |
96442e42 | 222 | |
b67bfe0d | 223 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
224 | if (e->rxhash == rxhash) |
225 | return e; | |
226 | } | |
227 | return NULL; | |
228 | } | |
229 | ||
230 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
231 | struct hlist_head *head, | |
232 | u32 rxhash, u16 queue_index) | |
233 | { | |
9fdc6bef ED |
234 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
235 | ||
96442e42 JW |
236 | if (e) { |
237 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
238 | rxhash, queue_index); | |
239 | e->updated = jiffies; | |
240 | e->rxhash = rxhash; | |
9bc88939 | 241 | e->rps_rxhash = 0; |
96442e42 JW |
242 | e->queue_index = queue_index; |
243 | e->tun = tun; | |
244 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 245 | ++tun->flow_count; |
96442e42 JW |
246 | } |
247 | return e; | |
248 | } | |
249 | ||
96442e42 JW |
250 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
251 | { | |
252 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
253 | e->rxhash, e->queue_index); | |
9bc88939 | 254 | sock_rps_reset_flow_hash(e->rps_rxhash); |
96442e42 | 255 | hlist_del_rcu(&e->hash_link); |
9fdc6bef | 256 | kfree_rcu(e, rcu); |
b8732fb7 | 257 | --tun->flow_count; |
96442e42 JW |
258 | } |
259 | ||
260 | static void tun_flow_flush(struct tun_struct *tun) | |
261 | { | |
262 | int i; | |
263 | ||
264 | spin_lock_bh(&tun->lock); | |
265 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
266 | struct tun_flow_entry *e; | |
b67bfe0d | 267 | struct hlist_node *n; |
96442e42 | 268 | |
b67bfe0d | 269 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
270 | tun_flow_delete(tun, e); |
271 | } | |
272 | spin_unlock_bh(&tun->lock); | |
273 | } | |
274 | ||
275 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
276 | { | |
277 | int i; | |
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 | if (e->queue_index == queue_index) |
286 | tun_flow_delete(tun, e); | |
287 | } | |
288 | } | |
289 | spin_unlock_bh(&tun->lock); | |
290 | } | |
291 | ||
292 | static void tun_flow_cleanup(unsigned long data) | |
293 | { | |
294 | struct tun_struct *tun = (struct tun_struct *)data; | |
295 | unsigned long delay = tun->ageing_time; | |
296 | unsigned long next_timer = jiffies + delay; | |
297 | unsigned long count = 0; | |
298 | int i; | |
299 | ||
300 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
301 | ||
302 | spin_lock_bh(&tun->lock); | |
303 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
304 | struct tun_flow_entry *e; | |
b67bfe0d | 305 | struct hlist_node *n; |
96442e42 | 306 | |
b67bfe0d | 307 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
308 | unsigned long this_timer; |
309 | count++; | |
310 | this_timer = e->updated + delay; | |
311 | if (time_before_eq(this_timer, jiffies)) | |
312 | tun_flow_delete(tun, e); | |
313 | else if (time_before(this_timer, next_timer)) | |
314 | next_timer = this_timer; | |
315 | } | |
316 | } | |
317 | ||
318 | if (count) | |
319 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
320 | spin_unlock_bh(&tun->lock); | |
321 | } | |
322 | ||
49974420 | 323 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 324 | struct tun_file *tfile) |
96442e42 JW |
325 | { |
326 | struct hlist_head *head; | |
327 | struct tun_flow_entry *e; | |
328 | unsigned long delay = tun->ageing_time; | |
9e85722d | 329 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
330 | |
331 | if (!rxhash) | |
332 | return; | |
333 | else | |
334 | head = &tun->flows[tun_hashfn(rxhash)]; | |
335 | ||
336 | rcu_read_lock(); | |
337 | ||
9e85722d JW |
338 | /* We may get a very small possibility of OOO during switching, not |
339 | * worth to optimize.*/ | |
340 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
341 | goto unlock; |
342 | ||
343 | e = tun_flow_find(head, rxhash); | |
344 | if (likely(e)) { | |
345 | /* TODO: keep queueing to old queue until it's empty? */ | |
346 | e->queue_index = queue_index; | |
347 | e->updated = jiffies; | |
9bc88939 | 348 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
349 | } else { |
350 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
351 | if (!tun_flow_find(head, rxhash) && |
352 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
353 | tun_flow_create(tun, head, rxhash, queue_index); |
354 | ||
355 | if (!timer_pending(&tun->flow_gc_timer)) | |
356 | mod_timer(&tun->flow_gc_timer, | |
357 | round_jiffies_up(jiffies + delay)); | |
358 | spin_unlock_bh(&tun->lock); | |
359 | } | |
360 | ||
361 | unlock: | |
362 | rcu_read_unlock(); | |
363 | } | |
364 | ||
9bc88939 TH |
365 | /** |
366 | * Save the hash received in the stack receive path and update the | |
367 | * flow_hash table accordingly. | |
368 | */ | |
369 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
370 | { | |
371 | if (unlikely(e->rps_rxhash != hash)) { | |
372 | sock_rps_reset_flow_hash(e->rps_rxhash); | |
373 | e->rps_rxhash = hash; | |
374 | } | |
375 | } | |
376 | ||
c8d68e6b | 377 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 378 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
379 | * the rxq based on the txq where the last packet of the flow comes. As |
380 | * the userspace application move between processors, we may get a | |
381 | * different rxq no. here. If we could not get rxhash, then we would | |
382 | * hope the rxq no. may help here. | |
383 | */ | |
f663dd9a | 384 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 385 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
386 | { |
387 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 388 | struct tun_flow_entry *e; |
c8d68e6b JW |
389 | u32 txq = 0; |
390 | u32 numqueues = 0; | |
391 | ||
392 | rcu_read_lock(); | |
92bb73ea | 393 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 394 | |
3958afa1 | 395 | txq = skb_get_hash(skb); |
c8d68e6b | 396 | if (txq) { |
96442e42 | 397 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 398 | if (e) { |
9bc88939 | 399 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 400 | txq = e->queue_index; |
9bc88939 | 401 | } else |
96442e42 JW |
402 | /* use multiply and shift instead of expensive divide */ |
403 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
404 | } else if (likely(skb_rx_queue_recorded(skb))) { |
405 | txq = skb_get_rx_queue(skb); | |
406 | while (unlikely(txq >= numqueues)) | |
407 | txq -= numqueues; | |
408 | } | |
409 | ||
410 | rcu_read_unlock(); | |
411 | return txq; | |
412 | } | |
413 | ||
cde8b15f JW |
414 | static inline bool tun_not_capable(struct tun_struct *tun) |
415 | { | |
416 | const struct cred *cred = current_cred(); | |
c260b772 | 417 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
418 | |
419 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
420 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 421 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
422 | } |
423 | ||
c8d68e6b JW |
424 | static void tun_set_real_num_queues(struct tun_struct *tun) |
425 | { | |
426 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
427 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
428 | } | |
429 | ||
4008e97f JW |
430 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
431 | { | |
432 | tfile->detached = tun; | |
433 | list_add_tail(&tfile->next, &tun->disabled); | |
434 | ++tun->numdisabled; | |
435 | } | |
436 | ||
d32649d1 | 437 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
438 | { |
439 | struct tun_struct *tun = tfile->detached; | |
440 | ||
441 | tfile->detached = NULL; | |
442 | list_del_init(&tfile->next); | |
443 | --tun->numdisabled; | |
444 | return tun; | |
445 | } | |
446 | ||
4bfb0513 JW |
447 | static void tun_queue_purge(struct tun_file *tfile) |
448 | { | |
449 | skb_queue_purge(&tfile->sk.sk_receive_queue); | |
450 | skb_queue_purge(&tfile->sk.sk_error_queue); | |
451 | } | |
452 | ||
c8d68e6b JW |
453 | static void __tun_detach(struct tun_file *tfile, bool clean) |
454 | { | |
455 | struct tun_file *ntfile; | |
456 | struct tun_struct *tun; | |
c8d68e6b | 457 | |
b8deabd3 JW |
458 | tun = rtnl_dereference(tfile->tun); |
459 | ||
9e85722d | 460 | if (tun && !tfile->detached) { |
c8d68e6b JW |
461 | u16 index = tfile->queue_index; |
462 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
463 | |
464 | rcu_assign_pointer(tun->tfiles[index], | |
465 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 466 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
467 | ntfile->queue_index = index; |
468 | ||
469 | --tun->numqueues; | |
9e85722d | 470 | if (clean) { |
c956674b | 471 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 472 | sock_put(&tfile->sk); |
9e85722d | 473 | } else |
4008e97f | 474 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
475 | |
476 | synchronize_net(); | |
96442e42 | 477 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 478 | /* Drop read queue */ |
4bfb0513 | 479 | tun_queue_purge(tfile); |
c8d68e6b | 480 | tun_set_real_num_queues(tun); |
dd38bd85 | 481 | } else if (tfile->detached && clean) { |
4008e97f | 482 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
483 | sock_put(&tfile->sk); |
484 | } | |
c8d68e6b JW |
485 | |
486 | if (clean) { | |
af668b3c MT |
487 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
488 | netif_carrier_off(tun->dev); | |
489 | ||
490 | if (!(tun->flags & TUN_PERSIST) && | |
491 | tun->dev->reg_state == NETREG_REGISTERED) | |
4008e97f | 492 | unregister_netdevice(tun->dev); |
af668b3c | 493 | } |
4008e97f | 494 | |
c8d68e6b JW |
495 | BUG_ON(!test_bit(SOCK_EXTERNALLY_ALLOCATED, |
496 | &tfile->socket.flags)); | |
497 | sk_release_kernel(&tfile->sk); | |
498 | } | |
499 | } | |
500 | ||
501 | static void tun_detach(struct tun_file *tfile, bool clean) | |
502 | { | |
503 | rtnl_lock(); | |
504 | __tun_detach(tfile, clean); | |
505 | rtnl_unlock(); | |
506 | } | |
507 | ||
508 | static void tun_detach_all(struct net_device *dev) | |
509 | { | |
510 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 511 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
512 | int i, n = tun->numqueues; |
513 | ||
514 | for (i = 0; i < n; i++) { | |
b8deabd3 | 515 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 516 | BUG_ON(!tfile); |
9e641bdc | 517 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 518 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
519 | --tun->numqueues; |
520 | } | |
9e85722d | 521 | list_for_each_entry(tfile, &tun->disabled, next) { |
9e641bdc | 522 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 523 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 524 | } |
c8d68e6b JW |
525 | BUG_ON(tun->numqueues != 0); |
526 | ||
527 | synchronize_net(); | |
528 | for (i = 0; i < n; i++) { | |
b8deabd3 | 529 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 530 | /* Drop read queue */ |
4bfb0513 | 531 | tun_queue_purge(tfile); |
c8d68e6b JW |
532 | sock_put(&tfile->sk); |
533 | } | |
4008e97f JW |
534 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
535 | tun_enable_queue(tfile); | |
4bfb0513 | 536 | tun_queue_purge(tfile); |
4008e97f JW |
537 | sock_put(&tfile->sk); |
538 | } | |
539 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 JW |
540 | |
541 | if (tun->flags & TUN_PERSIST) | |
542 | module_put(THIS_MODULE); | |
c8d68e6b JW |
543 | } |
544 | ||
849c9b6f | 545 | static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter) |
a7385ba2 | 546 | { |
631ab46b | 547 | struct tun_file *tfile = file->private_data; |
38231b7a | 548 | int err; |
a7385ba2 | 549 | |
5dbbaf2d PM |
550 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
551 | if (err < 0) | |
552 | goto out; | |
553 | ||
38231b7a | 554 | err = -EINVAL; |
9e85722d | 555 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
556 | goto out; |
557 | ||
558 | err = -EBUSY; | |
c8d68e6b JW |
559 | if (!(tun->flags & TUN_TAP_MQ) && tun->numqueues == 1) |
560 | goto out; | |
561 | ||
562 | err = -E2BIG; | |
4008e97f JW |
563 | if (!tfile->detached && |
564 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
565 | goto out; |
566 | ||
567 | err = 0; | |
54f968d6 | 568 | |
92d4ea6e | 569 | /* Re-attach the filter to persist device */ |
849c9b6f | 570 | if (!skip_filter && (tun->filter_attached == true)) { |
54f968d6 JW |
571 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); |
572 | if (!err) | |
573 | goto out; | |
574 | } | |
c8d68e6b | 575 | tfile->queue_index = tun->numqueues; |
6e914fc7 | 576 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 577 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 578 | tun->numqueues++; |
a7385ba2 | 579 | |
4008e97f JW |
580 | if (tfile->detached) |
581 | tun_enable_queue(tfile); | |
582 | else | |
583 | sock_hold(&tfile->sk); | |
584 | ||
c8d68e6b | 585 | tun_set_real_num_queues(tun); |
a7385ba2 | 586 | |
c8d68e6b JW |
587 | /* device is allowed to go away first, so no need to hold extra |
588 | * refcnt. | |
589 | */ | |
590 | ||
591 | out: | |
592 | return err; | |
631ab46b EB |
593 | } |
594 | ||
595 | static struct tun_struct *__tun_get(struct tun_file *tfile) | |
596 | { | |
6e914fc7 | 597 | struct tun_struct *tun; |
c70f1829 | 598 | |
6e914fc7 JW |
599 | rcu_read_lock(); |
600 | tun = rcu_dereference(tfile->tun); | |
601 | if (tun) | |
602 | dev_hold(tun->dev); | |
603 | rcu_read_unlock(); | |
c70f1829 EB |
604 | |
605 | return tun; | |
631ab46b EB |
606 | } |
607 | ||
608 | static struct tun_struct *tun_get(struct file *file) | |
609 | { | |
610 | return __tun_get(file->private_data); | |
611 | } | |
612 | ||
613 | static void tun_put(struct tun_struct *tun) | |
614 | { | |
6e914fc7 | 615 | dev_put(tun->dev); |
631ab46b EB |
616 | } |
617 | ||
6b8a66ee | 618 | /* TAP filtering */ |
f271b2cc MK |
619 | static void addr_hash_set(u32 *mask, const u8 *addr) |
620 | { | |
621 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
622 | mask[n >> 5] |= (1 << (n & 31)); | |
623 | } | |
624 | ||
625 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
626 | { | |
627 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
628 | return mask[n >> 5] & (1 << (n & 31)); | |
629 | } | |
630 | ||
631 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
632 | { | |
633 | struct { u8 u[ETH_ALEN]; } *addr; | |
634 | struct tun_filter uf; | |
635 | int err, alen, n, nexact; | |
636 | ||
637 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
638 | return -EFAULT; | |
639 | ||
640 | if (!uf.count) { | |
641 | /* Disabled */ | |
642 | filter->count = 0; | |
643 | return 0; | |
644 | } | |
645 | ||
646 | alen = ETH_ALEN * uf.count; | |
647 | addr = kmalloc(alen, GFP_KERNEL); | |
648 | if (!addr) | |
649 | return -ENOMEM; | |
650 | ||
651 | if (copy_from_user(addr, arg + sizeof(uf), alen)) { | |
652 | err = -EFAULT; | |
653 | goto done; | |
654 | } | |
655 | ||
656 | /* The filter is updated without holding any locks. Which is | |
657 | * perfectly safe. We disable it first and in the worst | |
658 | * case we'll accept a few undesired packets. */ | |
659 | filter->count = 0; | |
660 | wmb(); | |
661 | ||
662 | /* Use first set of addresses as an exact filter */ | |
663 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
664 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
665 | ||
666 | nexact = n; | |
667 | ||
cfbf84fc AW |
668 | /* Remaining multicast addresses are hashed, |
669 | * unicast will leave the filter disabled. */ | |
f271b2cc | 670 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
671 | for (; n < uf.count; n++) { |
672 | if (!is_multicast_ether_addr(addr[n].u)) { | |
673 | err = 0; /* no filter */ | |
674 | goto done; | |
675 | } | |
f271b2cc | 676 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 677 | } |
f271b2cc MK |
678 | |
679 | /* For ALLMULTI just set the mask to all ones. | |
680 | * This overrides the mask populated above. */ | |
681 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
682 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
683 | ||
684 | /* Now enable the filter */ | |
685 | wmb(); | |
686 | filter->count = nexact; | |
687 | ||
688 | /* Return the number of exact filters */ | |
689 | err = nexact; | |
690 | ||
691 | done: | |
692 | kfree(addr); | |
693 | return err; | |
694 | } | |
695 | ||
696 | /* Returns: 0 - drop, !=0 - accept */ | |
697 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
698 | { | |
699 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
700 | * at this point. */ | |
701 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
702 | int i; | |
703 | ||
704 | /* Exact match */ | |
705 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 706 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
707 | return 1; |
708 | ||
709 | /* Inexact match (multicast only) */ | |
710 | if (is_multicast_ether_addr(eh->h_dest)) | |
711 | return addr_hash_test(filter->mask, eh->h_dest); | |
712 | ||
713 | return 0; | |
714 | } | |
715 | ||
716 | /* | |
717 | * Checks whether the packet is accepted or not. | |
718 | * Returns: 0 - drop, !=0 - accept | |
719 | */ | |
720 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
721 | { | |
722 | if (!filter->count) | |
723 | return 1; | |
724 | ||
725 | return run_filter(filter, skb); | |
726 | } | |
727 | ||
1da177e4 LT |
728 | /* Network device part of the driver */ |
729 | ||
7282d491 | 730 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 731 | |
c70f1829 EB |
732 | /* Net device detach from fd. */ |
733 | static void tun_net_uninit(struct net_device *dev) | |
734 | { | |
c8d68e6b | 735 | tun_detach_all(dev); |
c70f1829 EB |
736 | } |
737 | ||
1da177e4 LT |
738 | /* Net device open. */ |
739 | static int tun_net_open(struct net_device *dev) | |
740 | { | |
c8d68e6b | 741 | netif_tx_start_all_queues(dev); |
1da177e4 LT |
742 | return 0; |
743 | } | |
744 | ||
745 | /* Net device close. */ | |
746 | static int tun_net_close(struct net_device *dev) | |
747 | { | |
c8d68e6b | 748 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
749 | return 0; |
750 | } | |
751 | ||
752 | /* Net device start xmit */ | |
424efe9c | 753 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
754 | { |
755 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 756 | int txq = skb->queue_mapping; |
6e914fc7 | 757 | struct tun_file *tfile; |
fa35864e | 758 | u32 numqueues = 0; |
1da177e4 | 759 | |
6e914fc7 | 760 | rcu_read_lock(); |
c8d68e6b | 761 | tfile = rcu_dereference(tun->tfiles[txq]); |
fa35864e | 762 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 763 | |
1da177e4 | 764 | /* Drop packet if interface is not attached */ |
fa35864e | 765 | if (txq >= numqueues) |
1da177e4 LT |
766 | goto drop; |
767 | ||
fa35864e | 768 | if (numqueues == 1) { |
9bc88939 TH |
769 | /* Select queue was not called for the skbuff, so we extract the |
770 | * RPS hash and save it into the flow_table here. | |
771 | */ | |
772 | __u32 rxhash; | |
773 | ||
774 | rxhash = skb_get_hash(skb); | |
775 | if (rxhash) { | |
776 | struct tun_flow_entry *e; | |
777 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
778 | rxhash); | |
779 | if (e) | |
780 | tun_flow_save_rps_rxhash(e, rxhash); | |
781 | } | |
782 | } | |
783 | ||
6e914fc7 JW |
784 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
785 | ||
c8d68e6b JW |
786 | BUG_ON(!tfile); |
787 | ||
f271b2cc MK |
788 | /* Drop if the filter does not like it. |
789 | * This is a noop if the filter is disabled. | |
790 | * Filter can be enabled only for the TAP devices. */ | |
791 | if (!check_filter(&tun->txflt, skb)) | |
792 | goto drop; | |
793 | ||
54f968d6 JW |
794 | if (tfile->socket.sk->sk_filter && |
795 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
796 | goto drop; |
797 | ||
36fe8c09 | 798 | /* Limit the number of packets queued by dividing txq length with the |
c8d68e6b JW |
799 | * number of queues. |
800 | */ | |
fa35864e DC |
801 | if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues |
802 | >= dev->tx_queue_len) | |
5d097109 | 803 | goto drop; |
1da177e4 | 804 | |
7bf66305 JW |
805 | if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC))) |
806 | goto drop; | |
807 | ||
eda29772 RC |
808 | if (skb->sk) { |
809 | sock_tx_timestamp(skb->sk, &skb_shinfo(skb)->tx_flags); | |
810 | sw_tx_timestamp(skb); | |
811 | } | |
812 | ||
0110d6f2 | 813 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
814 | * for indefinite time. |
815 | */ | |
0110d6f2 MT |
816 | skb_orphan(skb); |
817 | ||
f8af75f3 ED |
818 | nf_reset(skb); |
819 | ||
f271b2cc | 820 | /* Enqueue packet */ |
54f968d6 | 821 | skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb); |
1da177e4 LT |
822 | |
823 | /* Notify and wake up reader process */ | |
54f968d6 JW |
824 | if (tfile->flags & TUN_FASYNC) |
825 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 826 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
827 | |
828 | rcu_read_unlock(); | |
6ed10654 | 829 | return NETDEV_TX_OK; |
1da177e4 LT |
830 | |
831 | drop: | |
09f75cd7 | 832 | dev->stats.tx_dropped++; |
149d36f7 | 833 | skb_tx_error(skb); |
1da177e4 | 834 | kfree_skb(skb); |
6e914fc7 | 835 | rcu_read_unlock(); |
6ed10654 | 836 | return NETDEV_TX_OK; |
1da177e4 LT |
837 | } |
838 | ||
f271b2cc | 839 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 840 | { |
f271b2cc MK |
841 | /* |
842 | * This callback is supposed to deal with mc filter in | |
843 | * _rx_ path and has nothing to do with the _tx_ path. | |
844 | * In rx path we always accept everything userspace gives us. | |
845 | */ | |
1da177e4 LT |
846 | } |
847 | ||
4885a504 ES |
848 | #define MIN_MTU 68 |
849 | #define MAX_MTU 65535 | |
850 | ||
851 | static int | |
852 | tun_net_change_mtu(struct net_device *dev, int new_mtu) | |
853 | { | |
854 | if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU) | |
855 | return -EINVAL; | |
856 | dev->mtu = new_mtu; | |
857 | return 0; | |
858 | } | |
859 | ||
c8f44aff MM |
860 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
861 | netdev_features_t features) | |
88255375 MM |
862 | { |
863 | struct tun_struct *tun = netdev_priv(dev); | |
864 | ||
865 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
866 | } | |
bebd097a NH |
867 | #ifdef CONFIG_NET_POLL_CONTROLLER |
868 | static void tun_poll_controller(struct net_device *dev) | |
869 | { | |
870 | /* | |
871 | * Tun only receives frames when: | |
872 | * 1) the char device endpoint gets data from user space | |
873 | * 2) the tun socket gets a sendmsg call from user space | |
92d4ea6e | 874 | * Since both of those are synchronous operations, we are guaranteed |
bebd097a | 875 | * never to have pending data when we poll for it |
92d4ea6e | 876 | * so there is nothing to do here but return. |
bebd097a NH |
877 | * We need this though so netpoll recognizes us as an interface that |
878 | * supports polling, which enables bridge devices in virt setups to | |
879 | * still use netconsole | |
880 | */ | |
881 | return; | |
882 | } | |
883 | #endif | |
758e43b7 | 884 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 885 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
886 | .ndo_open = tun_net_open, |
887 | .ndo_stop = tun_net_close, | |
00829823 | 888 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 889 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 890 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 891 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
892 | #ifdef CONFIG_NET_POLL_CONTROLLER |
893 | .ndo_poll_controller = tun_poll_controller, | |
894 | #endif | |
758e43b7 SH |
895 | }; |
896 | ||
897 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 898 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
899 | .ndo_open = tun_net_open, |
900 | .ndo_stop = tun_net_close, | |
00829823 | 901 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 902 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 903 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 904 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
905 | .ndo_set_mac_address = eth_mac_addr, |
906 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 907 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
908 | #ifdef CONFIG_NET_POLL_CONTROLLER |
909 | .ndo_poll_controller = tun_poll_controller, | |
910 | #endif | |
758e43b7 SH |
911 | }; |
912 | ||
944a1376 | 913 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
914 | { |
915 | int i; | |
916 | ||
96442e42 JW |
917 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
918 | INIT_HLIST_HEAD(&tun->flows[i]); | |
919 | ||
920 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
921 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
922 | mod_timer(&tun->flow_gc_timer, | |
923 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
924 | } |
925 | ||
926 | static void tun_flow_uninit(struct tun_struct *tun) | |
927 | { | |
928 | del_timer_sync(&tun->flow_gc_timer); | |
929 | tun_flow_flush(tun); | |
96442e42 JW |
930 | } |
931 | ||
1da177e4 LT |
932 | /* Initialize net device. */ |
933 | static void tun_net_init(struct net_device *dev) | |
934 | { | |
935 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 936 | |
1da177e4 LT |
937 | switch (tun->flags & TUN_TYPE_MASK) { |
938 | case TUN_TUN_DEV: | |
758e43b7 SH |
939 | dev->netdev_ops = &tun_netdev_ops; |
940 | ||
1da177e4 LT |
941 | /* Point-to-Point TUN Device */ |
942 | dev->hard_header_len = 0; | |
943 | dev->addr_len = 0; | |
944 | dev->mtu = 1500; | |
945 | ||
946 | /* Zero header length */ | |
6aa20a22 | 947 | dev->type = ARPHRD_NONE; |
1da177e4 LT |
948 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
949 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ | |
950 | break; | |
951 | ||
952 | case TUN_TAP_DEV: | |
7a0a9608 | 953 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 954 | /* Ethernet TAP Device */ |
1da177e4 | 955 | ether_setup(dev); |
550fd08c | 956 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 957 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 958 | |
f2cedb63 | 959 | eth_hw_addr_random(dev); |
36226a8d | 960 | |
1da177e4 LT |
961 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ |
962 | break; | |
963 | } | |
964 | } | |
965 | ||
966 | /* Character device part */ | |
967 | ||
968 | /* Poll */ | |
c8d68e6b | 969 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 970 | { |
b2430de3 EB |
971 | struct tun_file *tfile = file->private_data; |
972 | struct tun_struct *tun = __tun_get(tfile); | |
3c8a9c63 | 973 | struct sock *sk; |
33dccbb0 | 974 | unsigned int mask = 0; |
1da177e4 LT |
975 | |
976 | if (!tun) | |
eac9e902 | 977 | return POLLERR; |
1da177e4 | 978 | |
54f968d6 | 979 | sk = tfile->socket.sk; |
3c8a9c63 | 980 | |
6b8a66ee | 981 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 982 | |
9e641bdc | 983 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 984 | |
89f56d1e | 985 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
986 | mask |= POLLIN | POLLRDNORM; |
987 | ||
33dccbb0 HX |
988 | if (sock_writeable(sk) || |
989 | (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
990 | sock_writeable(sk))) | |
991 | mask |= POLLOUT | POLLWRNORM; | |
992 | ||
c70f1829 EB |
993 | if (tun->dev->reg_state != NETREG_REGISTERED) |
994 | mask = POLLERR; | |
995 | ||
631ab46b | 996 | tun_put(tun); |
1da177e4 LT |
997 | return mask; |
998 | } | |
999 | ||
f42157cb RR |
1000 | /* prepad is the amount to reserve at front. len is length after that. |
1001 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1002 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1003 | size_t prepad, size_t len, |
1004 | size_t linear, int noblock) | |
f42157cb | 1005 | { |
54f968d6 | 1006 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1007 | struct sk_buff *skb; |
33dccbb0 | 1008 | int err; |
f42157cb RR |
1009 | |
1010 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1011 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1012 | linear = len; |
f42157cb | 1013 | |
33dccbb0 | 1014 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1015 | &err, 0); |
f42157cb | 1016 | if (!skb) |
33dccbb0 | 1017 | return ERR_PTR(err); |
f42157cb RR |
1018 | |
1019 | skb_reserve(skb, prepad); | |
1020 | skb_put(skb, linear); | |
33dccbb0 HX |
1021 | skb->data_len = len - linear; |
1022 | skb->len += len - linear; | |
f42157cb RR |
1023 | |
1024 | return skb; | |
1025 | } | |
1026 | ||
1da177e4 | 1027 | /* Get packet from user space buffer */ |
54f968d6 JW |
1028 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
1029 | void *msg_control, const struct iovec *iv, | |
1030 | size_t total_len, size_t count, int noblock) | |
1da177e4 | 1031 | { |
09640e63 | 1032 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1033 | struct sk_buff *skb; |
3dd5c330 | 1034 | size_t len = total_len, align = NET_SKB_PAD, linear; |
f43798c2 | 1035 | struct virtio_net_hdr gso = { 0 }; |
96f8d9ec | 1036 | int good_linear; |
6f26c9a7 | 1037 | int offset = 0; |
0690899b MT |
1038 | int copylen; |
1039 | bool zerocopy = false; | |
1040 | int err; | |
49974420 | 1041 | u32 rxhash; |
1da177e4 LT |
1042 | |
1043 | if (!(tun->flags & TUN_NO_PI)) { | |
15718ea0 | 1044 | if (len < sizeof(pi)) |
1da177e4 | 1045 | return -EINVAL; |
15718ea0 | 1046 | len -= sizeof(pi); |
1da177e4 | 1047 | |
6f26c9a7 | 1048 | if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi))) |
1da177e4 | 1049 | return -EFAULT; |
6f26c9a7 | 1050 | offset += sizeof(pi); |
1da177e4 LT |
1051 | } |
1052 | ||
f43798c2 | 1053 | if (tun->flags & TUN_VNET_HDR) { |
15718ea0 | 1054 | if (len < tun->vnet_hdr_sz) |
f43798c2 | 1055 | return -EINVAL; |
15718ea0 | 1056 | len -= tun->vnet_hdr_sz; |
f43798c2 | 1057 | |
6f26c9a7 | 1058 | if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso))) |
f43798c2 RR |
1059 | return -EFAULT; |
1060 | ||
4909122f HX |
1061 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
1062 | gso.csum_start + gso.csum_offset + 2 > gso.hdr_len) | |
1063 | gso.hdr_len = gso.csum_start + gso.csum_offset + 2; | |
1064 | ||
f43798c2 RR |
1065 | if (gso.hdr_len > len) |
1066 | return -EINVAL; | |
d9d52b51 | 1067 | offset += tun->vnet_hdr_sz; |
f43798c2 RR |
1068 | } |
1069 | ||
e01bf1c8 | 1070 | if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) { |
a504b86e | 1071 | align += NET_IP_ALIGN; |
0eca93bc HX |
1072 | if (unlikely(len < ETH_HLEN || |
1073 | (gso.hdr_len && gso.hdr_len < ETH_HLEN))) | |
e01bf1c8 RR |
1074 | return -EINVAL; |
1075 | } | |
6aa20a22 | 1076 | |
96f8d9ec JW |
1077 | good_linear = SKB_MAX_HEAD(align); |
1078 | ||
88529176 JW |
1079 | if (msg_control) { |
1080 | /* There are 256 bytes to be copied in skb, so there is | |
1081 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1082 | * The rest of the buffer is mapped from userspace. |
1083 | */ | |
88529176 | 1084 | copylen = gso.hdr_len ? gso.hdr_len : GOODCOPY_LEN; |
96f8d9ec JW |
1085 | if (copylen > good_linear) |
1086 | copylen = good_linear; | |
3dd5c330 | 1087 | linear = copylen; |
88529176 JW |
1088 | if (iov_pages(iv, offset + copylen, count) <= MAX_SKB_FRAGS) |
1089 | zerocopy = true; | |
1090 | } | |
1091 | ||
1092 | if (!zerocopy) { | |
0690899b | 1093 | copylen = len; |
96f8d9ec JW |
1094 | if (gso.hdr_len > good_linear) |
1095 | linear = good_linear; | |
1096 | else | |
1097 | linear = gso.hdr_len; | |
3dd5c330 | 1098 | } |
0690899b | 1099 | |
3dd5c330 | 1100 | skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); |
33dccbb0 HX |
1101 | if (IS_ERR(skb)) { |
1102 | if (PTR_ERR(skb) != -EAGAIN) | |
1103 | tun->dev->stats.rx_dropped++; | |
1104 | return PTR_ERR(skb); | |
1da177e4 LT |
1105 | } |
1106 | ||
0690899b MT |
1107 | if (zerocopy) |
1108 | err = zerocopy_sg_from_iovec(skb, iv, offset, count); | |
88529176 | 1109 | else { |
0690899b | 1110 | err = skb_copy_datagram_from_iovec(skb, 0, iv, offset, len); |
88529176 JW |
1111 | if (!err && msg_control) { |
1112 | struct ubuf_info *uarg = msg_control; | |
1113 | uarg->callback(uarg, false); | |
1114 | } | |
1115 | } | |
0690899b MT |
1116 | |
1117 | if (err) { | |
09f75cd7 | 1118 | tun->dev->stats.rx_dropped++; |
8f22757e | 1119 | kfree_skb(skb); |
1da177e4 | 1120 | return -EFAULT; |
8f22757e | 1121 | } |
1da177e4 | 1122 | |
f43798c2 RR |
1123 | if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { |
1124 | if (!skb_partial_csum_set(skb, gso.csum_start, | |
1125 | gso.csum_offset)) { | |
1126 | tun->dev->stats.rx_frame_errors++; | |
1127 | kfree_skb(skb); | |
1128 | return -EINVAL; | |
1129 | } | |
88255375 | 1130 | } |
f43798c2 | 1131 | |
1da177e4 LT |
1132 | switch (tun->flags & TUN_TYPE_MASK) { |
1133 | case TUN_TUN_DEV: | |
f09f7ee2 AWC |
1134 | if (tun->flags & TUN_NO_PI) { |
1135 | switch (skb->data[0] & 0xf0) { | |
1136 | case 0x40: | |
1137 | pi.proto = htons(ETH_P_IP); | |
1138 | break; | |
1139 | case 0x60: | |
1140 | pi.proto = htons(ETH_P_IPV6); | |
1141 | break; | |
1142 | default: | |
1143 | tun->dev->stats.rx_dropped++; | |
1144 | kfree_skb(skb); | |
1145 | return -EINVAL; | |
1146 | } | |
1147 | } | |
1148 | ||
459a98ed | 1149 | skb_reset_mac_header(skb); |
1da177e4 | 1150 | skb->protocol = pi.proto; |
4c13eb66 | 1151 | skb->dev = tun->dev; |
1da177e4 LT |
1152 | break; |
1153 | case TUN_TAP_DEV: | |
1154 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1155 | break; | |
6403eab1 | 1156 | } |
1da177e4 | 1157 | |
5188cd44 BH |
1158 | skb_reset_network_header(skb); |
1159 | ||
f43798c2 RR |
1160 | if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) { |
1161 | pr_debug("GSO!\n"); | |
1162 | switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { | |
1163 | case VIRTIO_NET_HDR_GSO_TCPV4: | |
c9af6db4 | 1164 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; |
f43798c2 RR |
1165 | break; |
1166 | case VIRTIO_NET_HDR_GSO_TCPV6: | |
c9af6db4 | 1167 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; |
f43798c2 | 1168 | break; |
e36aa25a | 1169 | case VIRTIO_NET_HDR_GSO_UDP: |
3d0ad094 BH |
1170 | { |
1171 | static bool warned; | |
1172 | ||
1173 | if (!warned) { | |
1174 | warned = true; | |
1175 | netdev_warn(tun->dev, | |
1176 | "%s: using disabled UFO feature; please fix this program\n", | |
1177 | current->comm); | |
1178 | } | |
c9af6db4 | 1179 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
5188cd44 BH |
1180 | if (skb->protocol == htons(ETH_P_IPV6)) |
1181 | ipv6_proxy_select_ident(skb); | |
e36aa25a | 1182 | break; |
3d0ad094 | 1183 | } |
f43798c2 RR |
1184 | default: |
1185 | tun->dev->stats.rx_frame_errors++; | |
1186 | kfree_skb(skb); | |
1187 | return -EINVAL; | |
1188 | } | |
1189 | ||
1190 | if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN) | |
c9af6db4 | 1191 | skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; |
f43798c2 RR |
1192 | |
1193 | skb_shinfo(skb)->gso_size = gso.gso_size; | |
1194 | if (skb_shinfo(skb)->gso_size == 0) { | |
1195 | tun->dev->stats.rx_frame_errors++; | |
1196 | kfree_skb(skb); | |
1197 | return -EINVAL; | |
1198 | } | |
1199 | ||
1200 | /* Header must be checked, and gso_segs computed. */ | |
1201 | skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY; | |
1202 | skb_shinfo(skb)->gso_segs = 0; | |
1203 | } | |
6aa20a22 | 1204 | |
0690899b MT |
1205 | /* copy skb_ubuf_info for callback when skb has no error */ |
1206 | if (zerocopy) { | |
1207 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1208 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1209 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
0690899b MT |
1210 | } |
1211 | ||
40893fd0 | 1212 | skb_probe_transport_header(skb, 0); |
38502af7 | 1213 | |
3958afa1 | 1214 | rxhash = skb_get_hash(skb); |
1da177e4 | 1215 | netif_rx_ni(skb); |
6aa20a22 | 1216 | |
09f75cd7 JG |
1217 | tun->dev->stats.rx_packets++; |
1218 | tun->dev->stats.rx_bytes += len; | |
1da177e4 | 1219 | |
9e85722d | 1220 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1221 | return total_len; |
6aa20a22 | 1222 | } |
1da177e4 | 1223 | |
ee0b3e67 BP |
1224 | static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv, |
1225 | unsigned long count, loff_t pos) | |
1da177e4 | 1226 | { |
33dccbb0 | 1227 | struct file *file = iocb->ki_filp; |
ab46d779 | 1228 | struct tun_struct *tun = tun_get(file); |
54f968d6 | 1229 | struct tun_file *tfile = file->private_data; |
631ab46b | 1230 | ssize_t result; |
1da177e4 LT |
1231 | |
1232 | if (!tun) | |
1233 | return -EBADFD; | |
1234 | ||
6b8a66ee | 1235 | tun_debug(KERN_INFO, tun, "tun_chr_write %ld\n", count); |
1da177e4 | 1236 | |
54f968d6 JW |
1237 | result = tun_get_user(tun, tfile, NULL, iv, iov_length(iv, count), |
1238 | count, file->f_flags & O_NONBLOCK); | |
631ab46b EB |
1239 | |
1240 | tun_put(tun); | |
1241 | return result; | |
1da177e4 LT |
1242 | } |
1243 | ||
1da177e4 | 1244 | /* Put packet to the user space buffer */ |
6f7c156c | 1245 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1246 | struct tun_file *tfile, |
6f7c156c | 1247 | struct sk_buff *skb, |
1248 | const struct iovec *iv, int len) | |
1da177e4 LT |
1249 | { |
1250 | struct tun_pi pi = { 0, skb->protocol }; | |
1251 | ssize_t total = 0; | |
e6fd07c8 | 1252 | int vlan_offset = 0, copied; |
a8f9bfdf | 1253 | int vlan_hlen = 0; |
2eb783c4 | 1254 | int vnet_hdr_sz = 0; |
a8f9bfdf HX |
1255 | |
1256 | if (vlan_tx_tag_present(skb)) | |
1257 | vlan_hlen = VLAN_HLEN; | |
1da177e4 | 1258 | |
2eb783c4 HX |
1259 | if (tun->flags & TUN_VNET_HDR) |
1260 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
1261 | ||
1da177e4 LT |
1262 | if (!(tun->flags & TUN_NO_PI)) { |
1263 | if ((len -= sizeof(pi)) < 0) | |
1264 | return -EINVAL; | |
1265 | ||
2eb783c4 | 1266 | if (len < skb->len + vlan_hlen + vnet_hdr_sz) { |
1da177e4 LT |
1267 | /* Packet will be striped */ |
1268 | pi.flags |= TUN_PKT_STRIP; | |
1269 | } | |
6aa20a22 | 1270 | |
43b39dcd | 1271 | if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi))) |
1da177e4 LT |
1272 | return -EFAULT; |
1273 | total += sizeof(pi); | |
6aa20a22 | 1274 | } |
1da177e4 | 1275 | |
2eb783c4 | 1276 | if (vnet_hdr_sz) { |
f43798c2 | 1277 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ |
2eb783c4 | 1278 | if ((len -= vnet_hdr_sz) < 0) |
f43798c2 RR |
1279 | return -EINVAL; |
1280 | ||
1281 | if (skb_is_gso(skb)) { | |
1282 | struct skb_shared_info *sinfo = skb_shinfo(skb); | |
1283 | ||
1284 | /* This is a hint as to how much should be linear. */ | |
1285 | gso.hdr_len = skb_headlen(skb); | |
1286 | gso.gso_size = sinfo->gso_size; | |
1287 | if (sinfo->gso_type & SKB_GSO_TCPV4) | |
1288 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4; | |
1289 | else if (sinfo->gso_type & SKB_GSO_TCPV6) | |
1290 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6; | |
ef3db4a5 | 1291 | else { |
6b8a66ee | 1292 | pr_err("unexpected GSO type: " |
ef3db4a5 MT |
1293 | "0x%x, gso_size %d, hdr_len %d\n", |
1294 | sinfo->gso_type, gso.gso_size, | |
1295 | gso.hdr_len); | |
1296 | print_hex_dump(KERN_ERR, "tun: ", | |
1297 | DUMP_PREFIX_NONE, | |
1298 | 16, 1, skb->head, | |
1299 | min((int)gso.hdr_len, 64), true); | |
1300 | WARN_ON_ONCE(1); | |
1301 | return -EINVAL; | |
1302 | } | |
f43798c2 RR |
1303 | if (sinfo->gso_type & SKB_GSO_TCP_ECN) |
1304 | gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN; | |
1305 | } else | |
1306 | gso.gso_type = VIRTIO_NET_HDR_GSO_NONE; | |
1307 | ||
1308 | if (skb->ip_summed == CHECKSUM_PARTIAL) { | |
1309 | gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; | |
a8f9bfdf HX |
1310 | gso.csum_start = skb_checksum_start_offset(skb) + |
1311 | vlan_hlen; | |
f43798c2 | 1312 | gso.csum_offset = skb->csum_offset; |
10a8d94a JW |
1313 | } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) { |
1314 | gso.flags = VIRTIO_NET_HDR_F_DATA_VALID; | |
f43798c2 RR |
1315 | } /* else everything is zero */ |
1316 | ||
43b39dcd MT |
1317 | if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total, |
1318 | sizeof(gso)))) | |
f43798c2 | 1319 | return -EFAULT; |
2eb783c4 | 1320 | total += vnet_hdr_sz; |
f43798c2 RR |
1321 | } |
1322 | ||
e6fd07c8 | 1323 | copied = total; |
a8f9bfdf HX |
1324 | len = min_t(int, skb->len + vlan_hlen, len); |
1325 | total += skb->len + vlan_hlen; | |
1326 | if (vlan_hlen) { | |
6680ec68 JW |
1327 | int copy, ret; |
1328 | struct { | |
1329 | __be16 h_vlan_proto; | |
1330 | __be16 h_vlan_TCI; | |
1331 | } veth; | |
1332 | ||
1333 | veth.h_vlan_proto = skb->vlan_proto; | |
1334 | veth.h_vlan_TCI = htons(vlan_tx_tag_get(skb)); | |
1335 | ||
1336 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 JW |
1337 | |
1338 | copy = min_t(int, vlan_offset, len); | |
e6fd07c8 | 1339 | ret = skb_copy_datagram_const_iovec(skb, 0, iv, copied, copy); |
6680ec68 | 1340 | len -= copy; |
e6fd07c8 | 1341 | copied += copy; |
6680ec68 JW |
1342 | if (ret || !len) |
1343 | goto done; | |
1344 | ||
1345 | copy = min_t(int, sizeof(veth), len); | |
e6fd07c8 | 1346 | ret = memcpy_toiovecend(iv, (void *)&veth, copied, copy); |
6680ec68 | 1347 | len -= copy; |
e6fd07c8 | 1348 | copied += copy; |
6680ec68 JW |
1349 | if (ret || !len) |
1350 | goto done; | |
1351 | } | |
1da177e4 | 1352 | |
e6fd07c8 | 1353 | skb_copy_datagram_const_iovec(skb, vlan_offset, iv, copied, len); |
1da177e4 | 1354 | |
6680ec68 | 1355 | done: |
09f75cd7 JG |
1356 | tun->dev->stats.tx_packets++; |
1357 | tun->dev->stats.tx_bytes += len; | |
1da177e4 LT |
1358 | |
1359 | return total; | |
1360 | } | |
1361 | ||
54f968d6 | 1362 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
f96eb74c | 1363 | const struct iovec *iv, ssize_t len, int noblock) |
1da177e4 | 1364 | { |
1da177e4 | 1365 | struct sk_buff *skb; |
05c2828c | 1366 | ssize_t ret = 0; |
9e641bdc | 1367 | int peeked, err, off = 0; |
1da177e4 | 1368 | |
3872baf6 | 1369 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1370 | |
9e641bdc XW |
1371 | if (!len) |
1372 | return ret; | |
1da177e4 | 1373 | |
9e641bdc XW |
1374 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1375 | return -EIO; | |
1da177e4 | 1376 | |
9e641bdc XW |
1377 | /* Read frames from queue */ |
1378 | skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0, | |
1379 | &peeked, &off, &err); | |
1380 | if (skb) { | |
54f968d6 | 1381 | ret = tun_put_user(tun, tfile, skb, iv, len); |
f271b2cc | 1382 | kfree_skb(skb); |
9e641bdc XW |
1383 | } else |
1384 | ret = err; | |
1da177e4 | 1385 | |
05c2828c MT |
1386 | return ret; |
1387 | } | |
1388 | ||
1389 | static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv, | |
1390 | unsigned long count, loff_t pos) | |
1391 | { | |
1392 | struct file *file = iocb->ki_filp; | |
1393 | struct tun_file *tfile = file->private_data; | |
1394 | struct tun_struct *tun = __tun_get(tfile); | |
1395 | ssize_t len, ret; | |
1396 | ||
1397 | if (!tun) | |
1398 | return -EBADFD; | |
1399 | len = iov_length(iv, count); | |
1400 | if (len < 0) { | |
1401 | ret = -EINVAL; | |
1402 | goto out; | |
1403 | } | |
1404 | ||
f96eb74c | 1405 | ret = tun_do_read(tun, tfile, iv, len, |
54f968d6 | 1406 | file->f_flags & O_NONBLOCK); |
42404c09 | 1407 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
1408 | if (ret > 0) |
1409 | iocb->ki_pos = ret; | |
631ab46b EB |
1410 | out: |
1411 | tun_put(tun); | |
1da177e4 LT |
1412 | return ret; |
1413 | } | |
1414 | ||
96442e42 JW |
1415 | static void tun_free_netdev(struct net_device *dev) |
1416 | { | |
1417 | struct tun_struct *tun = netdev_priv(dev); | |
1418 | ||
4008e97f | 1419 | BUG_ON(!(list_empty(&tun->disabled))); |
96442e42 | 1420 | tun_flow_uninit(tun); |
5dbbaf2d | 1421 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1422 | free_netdev(dev); |
1423 | } | |
1424 | ||
1da177e4 LT |
1425 | static void tun_setup(struct net_device *dev) |
1426 | { | |
1427 | struct tun_struct *tun = netdev_priv(dev); | |
1428 | ||
0625c883 EB |
1429 | tun->owner = INVALID_UID; |
1430 | tun->group = INVALID_GID; | |
1da177e4 | 1431 | |
1da177e4 | 1432 | dev->ethtool_ops = &tun_ethtool_ops; |
96442e42 | 1433 | dev->destructor = tun_free_netdev; |
1da177e4 LT |
1434 | } |
1435 | ||
f019a7a5 EB |
1436 | /* Trivial set of netlink ops to allow deleting tun or tap |
1437 | * device with netlink. | |
1438 | */ | |
1439 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1440 | { | |
1441 | return -EINVAL; | |
1442 | } | |
1443 | ||
1444 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
1445 | .kind = DRV_NAME, | |
1446 | .priv_size = sizeof(struct tun_struct), | |
1447 | .setup = tun_setup, | |
1448 | .validate = tun_validate, | |
1449 | }; | |
1450 | ||
33dccbb0 HX |
1451 | static void tun_sock_write_space(struct sock *sk) |
1452 | { | |
54f968d6 | 1453 | struct tun_file *tfile; |
43815482 | 1454 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
1455 | |
1456 | if (!sock_writeable(sk)) | |
1457 | return; | |
1458 | ||
33dccbb0 HX |
1459 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
1460 | return; | |
1461 | ||
43815482 ED |
1462 | wqueue = sk_sleep(sk); |
1463 | if (wqueue && waitqueue_active(wqueue)) | |
1464 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 1465 | POLLWRNORM | POLLWRBAND); |
c722c625 | 1466 | |
54f968d6 JW |
1467 | tfile = container_of(sk, struct tun_file, sk); |
1468 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
1469 | } |
1470 | ||
05c2828c MT |
1471 | static int tun_sendmsg(struct kiocb *iocb, struct socket *sock, |
1472 | struct msghdr *m, size_t total_len) | |
1473 | { | |
54f968d6 JW |
1474 | int ret; |
1475 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1476 | struct tun_struct *tun = __tun_get(tfile); | |
1477 | ||
1478 | if (!tun) | |
1479 | return -EBADFD; | |
54f968d6 JW |
1480 | ret = tun_get_user(tun, tfile, m->msg_control, m->msg_iov, total_len, |
1481 | m->msg_iovlen, m->msg_flags & MSG_DONTWAIT); | |
1482 | tun_put(tun); | |
1483 | return ret; | |
05c2828c MT |
1484 | } |
1485 | ||
1486 | static int tun_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1487 | struct msghdr *m, size_t total_len, | |
1488 | int flags) | |
1489 | { | |
54f968d6 JW |
1490 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
1491 | struct tun_struct *tun = __tun_get(tfile); | |
05c2828c | 1492 | int ret; |
54f968d6 JW |
1493 | |
1494 | if (!tun) | |
1495 | return -EBADFD; | |
1496 | ||
eda29772 | 1497 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
1498 | ret = -EINVAL; |
1499 | goto out; | |
1500 | } | |
eda29772 RC |
1501 | if (flags & MSG_ERRQUEUE) { |
1502 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
1503 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
1504 | goto out; | |
1505 | } | |
f96eb74c | 1506 | ret = tun_do_read(tun, tfile, m->msg_iov, total_len, |
05c2828c | 1507 | flags & MSG_DONTWAIT); |
42404c09 DM |
1508 | if (ret > total_len) { |
1509 | m->msg_flags |= MSG_TRUNC; | |
1510 | ret = flags & MSG_TRUNC ? ret : total_len; | |
1511 | } | |
3811ae76 | 1512 | out: |
54f968d6 | 1513 | tun_put(tun); |
05c2828c MT |
1514 | return ret; |
1515 | } | |
1516 | ||
1ab5ecb9 SK |
1517 | static int tun_release(struct socket *sock) |
1518 | { | |
1519 | if (sock->sk) | |
1520 | sock_put(sock->sk); | |
1521 | return 0; | |
1522 | } | |
1523 | ||
05c2828c MT |
1524 | /* Ops structure to mimic raw sockets with tun */ |
1525 | static const struct proto_ops tun_socket_ops = { | |
1526 | .sendmsg = tun_sendmsg, | |
1527 | .recvmsg = tun_recvmsg, | |
1ab5ecb9 | 1528 | .release = tun_release, |
05c2828c MT |
1529 | }; |
1530 | ||
33dccbb0 HX |
1531 | static struct proto tun_proto = { |
1532 | .name = "tun", | |
1533 | .owner = THIS_MODULE, | |
54f968d6 | 1534 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 1535 | }; |
f019a7a5 | 1536 | |
980c9e8c DW |
1537 | static int tun_flags(struct tun_struct *tun) |
1538 | { | |
031f5e03 | 1539 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
1540 | } |
1541 | ||
1542 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
1543 | char *buf) | |
1544 | { | |
1545 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
1546 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
1547 | } | |
1548 | ||
1549 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
1550 | char *buf) | |
1551 | { | |
1552 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1553 | return uid_valid(tun->owner)? |
1554 | sprintf(buf, "%u\n", | |
1555 | from_kuid_munged(current_user_ns(), tun->owner)): | |
1556 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1557 | } |
1558 | ||
1559 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
1560 | char *buf) | |
1561 | { | |
1562 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1563 | return gid_valid(tun->group) ? |
1564 | sprintf(buf, "%u\n", | |
1565 | from_kgid_munged(current_user_ns(), tun->group)): | |
1566 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1567 | } |
1568 | ||
1569 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
1570 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
1571 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
1572 | ||
d647a591 | 1573 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
1574 | { |
1575 | struct tun_struct *tun; | |
54f968d6 | 1576 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
1577 | struct net_device *dev; |
1578 | int err; | |
1579 | ||
7c0c3b1a JW |
1580 | if (tfile->detached) |
1581 | return -EINVAL; | |
1582 | ||
74a3e5a7 EB |
1583 | dev = __dev_get_by_name(net, ifr->ifr_name); |
1584 | if (dev) { | |
f85ba780 DW |
1585 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
1586 | return -EBUSY; | |
74a3e5a7 EB |
1587 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
1588 | tun = netdev_priv(dev); | |
1589 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
1590 | tun = netdev_priv(dev); | |
1591 | else | |
1592 | return -EINVAL; | |
1593 | ||
8e6d91ae JW |
1594 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
1595 | !!(tun->flags & TUN_TAP_MQ)) | |
1596 | return -EINVAL; | |
1597 | ||
cde8b15f | 1598 | if (tun_not_capable(tun)) |
2b980dbd | 1599 | return -EPERM; |
5dbbaf2d | 1600 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
1601 | if (err < 0) |
1602 | return err; | |
1603 | ||
849c9b6f | 1604 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER); |
a7385ba2 EB |
1605 | if (err < 0) |
1606 | return err; | |
4008e97f JW |
1607 | |
1608 | if (tun->flags & TUN_TAP_MQ && | |
e8dbad66 JW |
1609 | (tun->numqueues + tun->numdisabled > 1)) { |
1610 | /* One or more queue has already been attached, no need | |
1611 | * to initialize the device again. | |
1612 | */ | |
1613 | return 0; | |
1614 | } | |
6aa20a22 | 1615 | } |
1da177e4 LT |
1616 | else { |
1617 | char *name; | |
1618 | unsigned long flags = 0; | |
edfb6a14 JW |
1619 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
1620 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 1621 | |
c260b772 | 1622 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 1623 | return -EPERM; |
2b980dbd PM |
1624 | err = security_tun_dev_create(); |
1625 | if (err < 0) | |
1626 | return err; | |
ca6bb5d7 | 1627 | |
1da177e4 LT |
1628 | /* Set dev type */ |
1629 | if (ifr->ifr_flags & IFF_TUN) { | |
1630 | /* TUN device */ | |
1631 | flags |= TUN_TUN_DEV; | |
1632 | name = "tun%d"; | |
1633 | } else if (ifr->ifr_flags & IFF_TAP) { | |
1634 | /* TAP device */ | |
1635 | flags |= TUN_TAP_DEV; | |
1636 | name = "tap%d"; | |
6aa20a22 | 1637 | } else |
36989b90 | 1638 | return -EINVAL; |
6aa20a22 | 1639 | |
1da177e4 LT |
1640 | if (*ifr->ifr_name) |
1641 | name = ifr->ifr_name; | |
1642 | ||
c8d68e6b | 1643 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
1644 | NET_NAME_UNKNOWN, tun_setup, queues, |
1645 | queues); | |
edfb6a14 | 1646 | |
1da177e4 LT |
1647 | if (!dev) |
1648 | return -ENOMEM; | |
1649 | ||
fc54c658 | 1650 | dev_net_set(dev, net); |
f019a7a5 | 1651 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 1652 | dev->ifindex = tfile->ifindex; |
758e43b7 | 1653 | |
1da177e4 LT |
1654 | tun = netdev_priv(dev); |
1655 | tun->dev = dev; | |
1656 | tun->flags = flags; | |
f271b2cc | 1657 | tun->txflt.count = 0; |
d9d52b51 | 1658 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 1659 | |
54f968d6 JW |
1660 | tun->filter_attached = false; |
1661 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
33dccbb0 | 1662 | |
96442e42 JW |
1663 | spin_lock_init(&tun->lock); |
1664 | ||
5dbbaf2d PM |
1665 | err = security_tun_dev_alloc_security(&tun->security); |
1666 | if (err < 0) | |
1667 | goto err_free_dev; | |
2b980dbd | 1668 | |
1da177e4 | 1669 | tun_net_init(dev); |
944a1376 | 1670 | tun_flow_init(tun); |
96442e42 | 1671 | |
88255375 | 1672 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
1673 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
1674 | NETIF_F_HW_VLAN_STAG_TX; | |
88255375 | 1675 | dev->features = dev->hw_features; |
6671b224 FLVC |
1676 | dev->vlan_features = dev->features & |
1677 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
1678 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 1679 | |
4008e97f | 1680 | INIT_LIST_HEAD(&tun->disabled); |
849c9b6f | 1681 | err = tun_attach(tun, file, false); |
eb0fb363 | 1682 | if (err < 0) |
662ca437 | 1683 | goto err_free_flow; |
eb0fb363 | 1684 | |
1da177e4 LT |
1685 | err = register_netdevice(tun->dev); |
1686 | if (err < 0) | |
662ca437 | 1687 | goto err_detach; |
9c3fea6a | 1688 | |
980c9e8c DW |
1689 | if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) || |
1690 | device_create_file(&tun->dev->dev, &dev_attr_owner) || | |
1691 | device_create_file(&tun->dev->dev, &dev_attr_group)) | |
6b8a66ee | 1692 | pr_err("Failed to create tun sysfs files\n"); |
1da177e4 LT |
1693 | } |
1694 | ||
af668b3c MT |
1695 | netif_carrier_on(tun->dev); |
1696 | ||
6b8a66ee | 1697 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 1698 | |
031f5e03 MT |
1699 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
1700 | (ifr->ifr_flags & TUN_FEATURES); | |
c8d68e6b | 1701 | |
e35259a9 MK |
1702 | /* Make sure persistent devices do not get stuck in |
1703 | * xoff state. | |
1704 | */ | |
1705 | if (netif_running(tun->dev)) | |
c8d68e6b | 1706 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 1707 | |
1da177e4 LT |
1708 | strcpy(ifr->ifr_name, tun->dev->name); |
1709 | return 0; | |
1710 | ||
662ca437 JW |
1711 | err_detach: |
1712 | tun_detach_all(dev); | |
1713 | err_free_flow: | |
1714 | tun_flow_uninit(tun); | |
1715 | security_tun_dev_free_security(tun->security); | |
1716 | err_free_dev: | |
1da177e4 | 1717 | free_netdev(dev); |
1da177e4 LT |
1718 | return err; |
1719 | } | |
1720 | ||
9ce99cf6 | 1721 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 1722 | struct ifreq *ifr) |
e3b99556 | 1723 | { |
6b8a66ee | 1724 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
1725 | |
1726 | strcpy(ifr->ifr_name, tun->dev->name); | |
1727 | ||
980c9e8c | 1728 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 1729 | |
e3b99556 MM |
1730 | } |
1731 | ||
5228ddc9 RR |
1732 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
1733 | * privs required. */ | |
88255375 | 1734 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 1735 | { |
c8f44aff | 1736 | netdev_features_t features = 0; |
5228ddc9 RR |
1737 | |
1738 | if (arg & TUN_F_CSUM) { | |
88255375 | 1739 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
1740 | arg &= ~TUN_F_CSUM; |
1741 | ||
1742 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
1743 | if (arg & TUN_F_TSO_ECN) { | |
1744 | features |= NETIF_F_TSO_ECN; | |
1745 | arg &= ~TUN_F_TSO_ECN; | |
1746 | } | |
1747 | if (arg & TUN_F_TSO4) | |
1748 | features |= NETIF_F_TSO; | |
1749 | if (arg & TUN_F_TSO6) | |
1750 | features |= NETIF_F_TSO6; | |
1751 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
1752 | } | |
1753 | } | |
1754 | ||
1755 | /* This gives the user a way to test for new features in future by | |
1756 | * trying to set them. */ | |
1757 | if (arg) | |
1758 | return -EINVAL; | |
1759 | ||
88255375 MM |
1760 | tun->set_features = features; |
1761 | netdev_update_features(tun->dev); | |
5228ddc9 RR |
1762 | |
1763 | return 0; | |
1764 | } | |
1765 | ||
c8d68e6b JW |
1766 | static void tun_detach_filter(struct tun_struct *tun, int n) |
1767 | { | |
1768 | int i; | |
1769 | struct tun_file *tfile; | |
1770 | ||
1771 | for (i = 0; i < n; i++) { | |
b8deabd3 | 1772 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1773 | sk_detach_filter(tfile->socket.sk); |
1774 | } | |
1775 | ||
1776 | tun->filter_attached = false; | |
1777 | } | |
1778 | ||
1779 | static int tun_attach_filter(struct tun_struct *tun) | |
1780 | { | |
1781 | int i, ret = 0; | |
1782 | struct tun_file *tfile; | |
1783 | ||
1784 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1785 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1786 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); |
1787 | if (ret) { | |
1788 | tun_detach_filter(tun, i); | |
1789 | return ret; | |
1790 | } | |
1791 | } | |
1792 | ||
1793 | tun->filter_attached = true; | |
1794 | return ret; | |
1795 | } | |
1796 | ||
1797 | static void tun_set_sndbuf(struct tun_struct *tun) | |
1798 | { | |
1799 | struct tun_file *tfile; | |
1800 | int i; | |
1801 | ||
1802 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1803 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1804 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
1805 | } | |
1806 | } | |
1807 | ||
cde8b15f JW |
1808 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
1809 | { | |
1810 | struct tun_file *tfile = file->private_data; | |
1811 | struct tun_struct *tun; | |
cde8b15f JW |
1812 | int ret = 0; |
1813 | ||
1814 | rtnl_lock(); | |
1815 | ||
1816 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 1817 | tun = tfile->detached; |
5dbbaf2d | 1818 | if (!tun) { |
cde8b15f | 1819 | ret = -EINVAL; |
5dbbaf2d PM |
1820 | goto unlock; |
1821 | } | |
1822 | ret = security_tun_dev_attach_queue(tun->security); | |
1823 | if (ret < 0) | |
1824 | goto unlock; | |
849c9b6f | 1825 | ret = tun_attach(tun, file, false); |
4008e97f | 1826 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 1827 | tun = rtnl_dereference(tfile->tun); |
9e85722d | 1828 | if (!tun || !(tun->flags & TUN_TAP_MQ) || tfile->detached) |
4008e97f JW |
1829 | ret = -EINVAL; |
1830 | else | |
1831 | __tun_detach(tfile, false); | |
1832 | } else | |
cde8b15f JW |
1833 | ret = -EINVAL; |
1834 | ||
5dbbaf2d | 1835 | unlock: |
cde8b15f JW |
1836 | rtnl_unlock(); |
1837 | return ret; | |
1838 | } | |
1839 | ||
50857e2a AB |
1840 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
1841 | unsigned long arg, int ifreq_len) | |
1da177e4 | 1842 | { |
36b50bab | 1843 | struct tun_file *tfile = file->private_data; |
631ab46b | 1844 | struct tun_struct *tun; |
1da177e4 LT |
1845 | void __user* argp = (void __user*)arg; |
1846 | struct ifreq ifr; | |
0625c883 EB |
1847 | kuid_t owner; |
1848 | kgid_t group; | |
33dccbb0 | 1849 | int sndbuf; |
d9d52b51 | 1850 | int vnet_hdr_sz; |
fb7589a1 | 1851 | unsigned int ifindex; |
f271b2cc | 1852 | int ret; |
1da177e4 | 1853 | |
cde8b15f | 1854 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) { |
50857e2a | 1855 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 1856 | return -EFAULT; |
8bbb1813 | 1857 | } else { |
a117dacd | 1858 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 1859 | } |
631ab46b EB |
1860 | if (cmd == TUNGETFEATURES) { |
1861 | /* Currently this just means: "what IFF flags are valid?". | |
1862 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
1863 | * TUNSETIFF. |
1864 | */ | |
1865 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 1866 | (unsigned int __user*)argp); |
cde8b15f JW |
1867 | } else if (cmd == TUNSETQUEUE) |
1868 | return tun_set_queue(file, &ifr); | |
631ab46b | 1869 | |
c8d68e6b | 1870 | ret = 0; |
876bfd4d HX |
1871 | rtnl_lock(); |
1872 | ||
36b50bab | 1873 | tun = __tun_get(tfile); |
1da177e4 | 1874 | if (cmd == TUNSETIFF && !tun) { |
1da177e4 LT |
1875 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
1876 | ||
876bfd4d | 1877 | ret = tun_set_iff(tfile->net, file, &ifr); |
1da177e4 | 1878 | |
876bfd4d HX |
1879 | if (ret) |
1880 | goto unlock; | |
1da177e4 | 1881 | |
50857e2a | 1882 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
1883 | ret = -EFAULT; |
1884 | goto unlock; | |
1da177e4 | 1885 | } |
fb7589a1 PE |
1886 | if (cmd == TUNSETIFINDEX) { |
1887 | ret = -EPERM; | |
1888 | if (tun) | |
1889 | goto unlock; | |
1890 | ||
1891 | ret = -EFAULT; | |
1892 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
1893 | goto unlock; | |
1894 | ||
1895 | ret = 0; | |
1896 | tfile->ifindex = ifindex; | |
1897 | goto unlock; | |
1898 | } | |
1da177e4 | 1899 | |
876bfd4d | 1900 | ret = -EBADFD; |
1da177e4 | 1901 | if (!tun) |
876bfd4d | 1902 | goto unlock; |
1da177e4 | 1903 | |
1e588338 | 1904 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 1905 | |
631ab46b | 1906 | ret = 0; |
1da177e4 | 1907 | switch (cmd) { |
e3b99556 | 1908 | case TUNGETIFF: |
9ce99cf6 | 1909 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 1910 | |
3d407a80 PE |
1911 | if (tfile->detached) |
1912 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
1913 | if (!tfile->socket.sk->sk_filter) |
1914 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 1915 | |
50857e2a | 1916 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 1917 | ret = -EFAULT; |
e3b99556 MM |
1918 | break; |
1919 | ||
1da177e4 LT |
1920 | case TUNSETNOCSUM: |
1921 | /* Disable/Enable checksum */ | |
1da177e4 | 1922 | |
88255375 MM |
1923 | /* [unimplemented] */ |
1924 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 1925 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
1926 | break; |
1927 | ||
1928 | case TUNSETPERSIST: | |
54f968d6 JW |
1929 | /* Disable/Enable persist mode. Keep an extra reference to the |
1930 | * module to prevent the module being unprobed. | |
1931 | */ | |
dd38bd85 | 1932 | if (arg && !(tun->flags & TUN_PERSIST)) { |
1da177e4 | 1933 | tun->flags |= TUN_PERSIST; |
54f968d6 | 1934 | __module_get(THIS_MODULE); |
dd38bd85 JW |
1935 | } |
1936 | if (!arg && (tun->flags & TUN_PERSIST)) { | |
1da177e4 | 1937 | tun->flags &= ~TUN_PERSIST; |
54f968d6 JW |
1938 | module_put(THIS_MODULE); |
1939 | } | |
1da177e4 | 1940 | |
6b8a66ee JP |
1941 | tun_debug(KERN_INFO, tun, "persist %s\n", |
1942 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
1943 | break; |
1944 | ||
1945 | case TUNSETOWNER: | |
1946 | /* Set owner of the device */ | |
0625c883 EB |
1947 | owner = make_kuid(current_user_ns(), arg); |
1948 | if (!uid_valid(owner)) { | |
1949 | ret = -EINVAL; | |
1950 | break; | |
1951 | } | |
1952 | tun->owner = owner; | |
1e588338 | 1953 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 1954 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
1955 | break; |
1956 | ||
8c644623 GG |
1957 | case TUNSETGROUP: |
1958 | /* Set group of the device */ | |
0625c883 EB |
1959 | group = make_kgid(current_user_ns(), arg); |
1960 | if (!gid_valid(group)) { | |
1961 | ret = -EINVAL; | |
1962 | break; | |
1963 | } | |
1964 | tun->group = group; | |
1e588338 | 1965 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 1966 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
1967 | break; |
1968 | ||
ff4cc3ac MK |
1969 | case TUNSETLINK: |
1970 | /* Only allow setting the type when the interface is down */ | |
1971 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
1972 | tun_debug(KERN_INFO, tun, |
1973 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 1974 | ret = -EBUSY; |
ff4cc3ac MK |
1975 | } else { |
1976 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
1977 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
1978 | tun->dev->type); | |
48abfe05 | 1979 | ret = 0; |
ff4cc3ac | 1980 | } |
631ab46b | 1981 | break; |
ff4cc3ac | 1982 | |
1da177e4 LT |
1983 | #ifdef TUN_DEBUG |
1984 | case TUNSETDEBUG: | |
1985 | tun->debug = arg; | |
1986 | break; | |
1987 | #endif | |
5228ddc9 | 1988 | case TUNSETOFFLOAD: |
88255375 | 1989 | ret = set_offload(tun, arg); |
631ab46b | 1990 | break; |
5228ddc9 | 1991 | |
f271b2cc MK |
1992 | case TUNSETTXFILTER: |
1993 | /* Can be set only for TAPs */ | |
631ab46b | 1994 | ret = -EINVAL; |
f271b2cc | 1995 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) |
631ab46b | 1996 | break; |
c0e5a8c2 | 1997 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 1998 | break; |
1da177e4 LT |
1999 | |
2000 | case SIOCGIFHWADDR: | |
b595076a | 2001 | /* Get hw address */ |
f271b2cc MK |
2002 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2003 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2004 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2005 | ret = -EFAULT; |
2006 | break; | |
1da177e4 LT |
2007 | |
2008 | case SIOCSIFHWADDR: | |
f271b2cc | 2009 | /* Set hw address */ |
6b8a66ee JP |
2010 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2011 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2012 | |
40102371 | 2013 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2014 | break; |
33dccbb0 HX |
2015 | |
2016 | case TUNGETSNDBUF: | |
54f968d6 | 2017 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2018 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2019 | ret = -EFAULT; | |
2020 | break; | |
2021 | ||
2022 | case TUNSETSNDBUF: | |
2023 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2024 | ret = -EFAULT; | |
2025 | break; | |
2026 | } | |
2027 | ||
c8d68e6b JW |
2028 | tun->sndbuf = sndbuf; |
2029 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2030 | break; |
2031 | ||
d9d52b51 MT |
2032 | case TUNGETVNETHDRSZ: |
2033 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2034 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2035 | ret = -EFAULT; | |
2036 | break; | |
2037 | ||
2038 | case TUNSETVNETHDRSZ: | |
2039 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2040 | ret = -EFAULT; | |
2041 | break; | |
2042 | } | |
2043 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2044 | ret = -EINVAL; | |
2045 | break; | |
2046 | } | |
2047 | ||
2048 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2049 | break; | |
2050 | ||
99405162 MT |
2051 | case TUNATTACHFILTER: |
2052 | /* Can be set only for TAPs */ | |
2053 | ret = -EINVAL; | |
2054 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2055 | break; | |
2056 | ret = -EFAULT; | |
54f968d6 | 2057 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2058 | break; |
2059 | ||
c8d68e6b | 2060 | ret = tun_attach_filter(tun); |
99405162 MT |
2061 | break; |
2062 | ||
2063 | case TUNDETACHFILTER: | |
2064 | /* Can be set only for TAPs */ | |
2065 | ret = -EINVAL; | |
2066 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2067 | break; | |
c8d68e6b JW |
2068 | ret = 0; |
2069 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2070 | break; |
2071 | ||
76975e9c PE |
2072 | case TUNGETFILTER: |
2073 | ret = -EINVAL; | |
2074 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) | |
2075 | break; | |
2076 | ret = -EFAULT; | |
2077 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2078 | break; | |
2079 | ret = 0; | |
2080 | break; | |
2081 | ||
1da177e4 | 2082 | default: |
631ab46b EB |
2083 | ret = -EINVAL; |
2084 | break; | |
ee289b64 | 2085 | } |
1da177e4 | 2086 | |
876bfd4d HX |
2087 | unlock: |
2088 | rtnl_unlock(); | |
2089 | if (tun) | |
2090 | tun_put(tun); | |
631ab46b | 2091 | return ret; |
1da177e4 LT |
2092 | } |
2093 | ||
50857e2a AB |
2094 | static long tun_chr_ioctl(struct file *file, |
2095 | unsigned int cmd, unsigned long arg) | |
2096 | { | |
2097 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2098 | } | |
2099 | ||
2100 | #ifdef CONFIG_COMPAT | |
2101 | static long tun_chr_compat_ioctl(struct file *file, | |
2102 | unsigned int cmd, unsigned long arg) | |
2103 | { | |
2104 | switch (cmd) { | |
2105 | case TUNSETIFF: | |
2106 | case TUNGETIFF: | |
2107 | case TUNSETTXFILTER: | |
2108 | case TUNGETSNDBUF: | |
2109 | case TUNSETSNDBUF: | |
2110 | case SIOCGIFHWADDR: | |
2111 | case SIOCSIFHWADDR: | |
2112 | arg = (unsigned long)compat_ptr(arg); | |
2113 | break; | |
2114 | default: | |
2115 | arg = (compat_ulong_t)arg; | |
2116 | break; | |
2117 | } | |
2118 | ||
2119 | /* | |
2120 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2121 | * the end of that structure. All fields that are used in this | |
2122 | * driver are compatible though, we don't need to convert the | |
2123 | * contents. | |
2124 | */ | |
2125 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2126 | } | |
2127 | #endif /* CONFIG_COMPAT */ | |
2128 | ||
1da177e4 LT |
2129 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2130 | { | |
54f968d6 | 2131 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2132 | int ret; |
2133 | ||
54f968d6 | 2134 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2135 | goto out; |
6aa20a22 | 2136 | |
1da177e4 | 2137 | if (on) { |
e0b93edd | 2138 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2139 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2140 | } else |
54f968d6 | 2141 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2142 | ret = 0; |
2143 | out: | |
9d319522 | 2144 | return ret; |
1da177e4 LT |
2145 | } |
2146 | ||
2147 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2148 | { | |
631ab46b | 2149 | struct tun_file *tfile; |
deed49fb | 2150 | |
6b8a66ee | 2151 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2152 | |
54f968d6 JW |
2153 | tfile = (struct tun_file *)sk_alloc(&init_net, AF_UNSPEC, GFP_KERNEL, |
2154 | &tun_proto); | |
631ab46b EB |
2155 | if (!tfile) |
2156 | return -ENOMEM; | |
c956674b | 2157 | RCU_INIT_POINTER(tfile->tun, NULL); |
36b50bab | 2158 | tfile->net = get_net(current->nsproxy->net_ns); |
54f968d6 | 2159 | tfile->flags = 0; |
fb7589a1 | 2160 | tfile->ifindex = 0; |
54f968d6 | 2161 | |
54f968d6 | 2162 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2163 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2164 | |
2165 | tfile->socket.file = file; | |
2166 | tfile->socket.ops = &tun_socket_ops; | |
2167 | ||
2168 | sock_init_data(&tfile->socket, &tfile->sk); | |
2169 | sk_change_net(&tfile->sk, tfile->net); | |
2170 | ||
2171 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2172 | tfile->sk.sk_sndbuf = INT_MAX; | |
2173 | ||
631ab46b | 2174 | file->private_data = tfile; |
54f968d6 | 2175 | set_bit(SOCK_EXTERNALLY_ALLOCATED, &tfile->socket.flags); |
4008e97f | 2176 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2177 | |
19a6afb2 JW |
2178 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2179 | ||
1da177e4 LT |
2180 | return 0; |
2181 | } | |
2182 | ||
2183 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2184 | { | |
631ab46b | 2185 | struct tun_file *tfile = file->private_data; |
54f968d6 | 2186 | struct net *net = tfile->net; |
1da177e4 | 2187 | |
c8d68e6b | 2188 | tun_detach(tfile, true); |
54f968d6 | 2189 | put_net(net); |
1da177e4 LT |
2190 | |
2191 | return 0; | |
2192 | } | |
2193 | ||
93e14b6d MY |
2194 | #ifdef CONFIG_PROC_FS |
2195 | static int tun_chr_show_fdinfo(struct seq_file *m, struct file *f) | |
2196 | { | |
2197 | struct tun_struct *tun; | |
2198 | struct ifreq ifr; | |
2199 | ||
2200 | memset(&ifr, 0, sizeof(ifr)); | |
2201 | ||
2202 | rtnl_lock(); | |
2203 | tun = tun_get(f); | |
2204 | if (tun) | |
2205 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2206 | rtnl_unlock(); | |
2207 | ||
2208 | if (tun) | |
2209 | tun_put(tun); | |
2210 | ||
2211 | return seq_printf(m, "iff:\t%s\n", ifr.ifr_name); | |
2212 | } | |
2213 | #endif | |
2214 | ||
d54b1fdb | 2215 | static const struct file_operations tun_fops = { |
6aa20a22 | 2216 | .owner = THIS_MODULE, |
1da177e4 | 2217 | .llseek = no_llseek, |
ee0b3e67 BP |
2218 | .read = do_sync_read, |
2219 | .aio_read = tun_chr_aio_read, | |
2220 | .write = do_sync_write, | |
2221 | .aio_write = tun_chr_aio_write, | |
1da177e4 | 2222 | .poll = tun_chr_poll, |
50857e2a AB |
2223 | .unlocked_ioctl = tun_chr_ioctl, |
2224 | #ifdef CONFIG_COMPAT | |
2225 | .compat_ioctl = tun_chr_compat_ioctl, | |
2226 | #endif | |
1da177e4 LT |
2227 | .open = tun_chr_open, |
2228 | .release = tun_chr_close, | |
93e14b6d MY |
2229 | .fasync = tun_chr_fasync, |
2230 | #ifdef CONFIG_PROC_FS | |
2231 | .show_fdinfo = tun_chr_show_fdinfo, | |
2232 | #endif | |
1da177e4 LT |
2233 | }; |
2234 | ||
2235 | static struct miscdevice tun_miscdev = { | |
2236 | .minor = TUN_MINOR, | |
2237 | .name = "tun", | |
e454cea2 | 2238 | .nodename = "net/tun", |
1da177e4 | 2239 | .fops = &tun_fops, |
1da177e4 LT |
2240 | }; |
2241 | ||
2242 | /* ethtool interface */ | |
2243 | ||
2244 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) | |
2245 | { | |
2246 | cmd->supported = 0; | |
2247 | cmd->advertising = 0; | |
70739497 | 2248 | ethtool_cmd_speed_set(cmd, SPEED_10); |
1da177e4 LT |
2249 | cmd->duplex = DUPLEX_FULL; |
2250 | cmd->port = PORT_TP; | |
2251 | cmd->phy_address = 0; | |
2252 | cmd->transceiver = XCVR_INTERNAL; | |
2253 | cmd->autoneg = AUTONEG_DISABLE; | |
2254 | cmd->maxtxpkt = 0; | |
2255 | cmd->maxrxpkt = 0; | |
2256 | return 0; | |
2257 | } | |
2258 | ||
2259 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2260 | { | |
2261 | struct tun_struct *tun = netdev_priv(dev); | |
2262 | ||
33a5ba14 RJ |
2263 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2264 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2265 | |
2266 | switch (tun->flags & TUN_TYPE_MASK) { | |
2267 | case TUN_TUN_DEV: | |
33a5ba14 | 2268 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 LT |
2269 | break; |
2270 | case TUN_TAP_DEV: | |
33a5ba14 | 2271 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2272 | break; |
2273 | } | |
2274 | } | |
2275 | ||
2276 | static u32 tun_get_msglevel(struct net_device *dev) | |
2277 | { | |
2278 | #ifdef TUN_DEBUG | |
2279 | struct tun_struct *tun = netdev_priv(dev); | |
2280 | return tun->debug; | |
2281 | #else | |
2282 | return -EOPNOTSUPP; | |
2283 | #endif | |
2284 | } | |
2285 | ||
2286 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2287 | { | |
2288 | #ifdef TUN_DEBUG | |
2289 | struct tun_struct *tun = netdev_priv(dev); | |
2290 | tun->debug = value; | |
2291 | #endif | |
2292 | } | |
2293 | ||
7282d491 | 2294 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2295 | .get_settings = tun_get_settings, |
2296 | .get_drvinfo = tun_get_drvinfo, | |
2297 | .get_msglevel = tun_get_msglevel, | |
2298 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2299 | .get_link = ethtool_op_get_link, |
eda29772 | 2300 | .get_ts_info = ethtool_op_get_ts_info, |
1da177e4 LT |
2301 | }; |
2302 | ||
79d17604 | 2303 | |
1da177e4 LT |
2304 | static int __init tun_init(void) |
2305 | { | |
2306 | int ret = 0; | |
2307 | ||
6b8a66ee JP |
2308 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
2309 | pr_info("%s\n", DRV_COPYRIGHT); | |
1da177e4 | 2310 | |
f019a7a5 | 2311 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 2312 | if (ret) { |
6b8a66ee | 2313 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 2314 | goto err_linkops; |
79d17604 PE |
2315 | } |
2316 | ||
1da177e4 | 2317 | ret = misc_register(&tun_miscdev); |
79d17604 | 2318 | if (ret) { |
6b8a66ee | 2319 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
2320 | goto err_misc; |
2321 | } | |
f019a7a5 | 2322 | return 0; |
79d17604 | 2323 | err_misc: |
f019a7a5 EB |
2324 | rtnl_link_unregister(&tun_link_ops); |
2325 | err_linkops: | |
1da177e4 LT |
2326 | return ret; |
2327 | } | |
2328 | ||
2329 | static void tun_cleanup(void) | |
2330 | { | |
6aa20a22 | 2331 | misc_deregister(&tun_miscdev); |
f019a7a5 | 2332 | rtnl_link_unregister(&tun_link_ops); |
1da177e4 LT |
2333 | } |
2334 | ||
05c2828c MT |
2335 | /* Get an underlying socket object from tun file. Returns error unless file is |
2336 | * attached to a device. The returned object works like a packet socket, it | |
2337 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
2338 | * holding a reference to the file for as long as the socket is in use. */ | |
2339 | struct socket *tun_get_socket(struct file *file) | |
2340 | { | |
6e914fc7 | 2341 | struct tun_file *tfile; |
05c2828c MT |
2342 | if (file->f_op != &tun_fops) |
2343 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
2344 | tfile = file->private_data; |
2345 | if (!tfile) | |
05c2828c | 2346 | return ERR_PTR(-EBADFD); |
54f968d6 | 2347 | return &tfile->socket; |
05c2828c MT |
2348 | } |
2349 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
2350 | ||
1da177e4 LT |
2351 | module_init(tun_init); |
2352 | module_exit(tun_cleanup); | |
2353 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
2354 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
2355 | MODULE_LICENSE("GPL"); | |
2356 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 2357 | MODULE_ALIAS("devname:net/tun"); |