Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * TUN - Universal TUN/TAP device driver. | |
3 | * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com> | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify | |
6 | * it under the terms of the GNU General Public License as published by | |
7 | * the Free Software Foundation; either version 2 of the License, or | |
8 | * (at your option) any later version. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, | |
11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
13 | * GNU General Public License for more details. | |
14 | * | |
15 | * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $ | |
16 | */ | |
17 | ||
18 | /* | |
19 | * Changes: | |
20 | * | |
ff4cc3ac MK |
21 | * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14 |
22 | * Add TUNSETLINK ioctl to set the link encapsulation | |
23 | * | |
1da177e4 | 24 | * Mark Smith <markzzzsmith@yahoo.com.au> |
344dc8ed | 25 | * Use eth_random_addr() for tap MAC address. |
1da177e4 LT |
26 | * |
27 | * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20 | |
28 | * Fixes in packet dropping, queue length setting and queue wakeup. | |
29 | * Increased default tx queue length. | |
30 | * Added ethtool API. | |
31 | * Minor cleanups | |
32 | * | |
33 | * Daniel Podlejski <underley@underley.eu.org> | |
34 | * Modifications for 2.3.99-pre5 kernel. | |
35 | */ | |
36 | ||
6b8a66ee JP |
37 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
38 | ||
1da177e4 LT |
39 | #define DRV_NAME "tun" |
40 | #define DRV_VERSION "1.6" | |
41 | #define DRV_DESCRIPTION "Universal TUN/TAP device driver" | |
42 | #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>" | |
43 | ||
1da177e4 LT |
44 | #include <linux/module.h> |
45 | #include <linux/errno.h> | |
46 | #include <linux/kernel.h> | |
47 | #include <linux/major.h> | |
48 | #include <linux/slab.h> | |
49 | #include <linux/poll.h> | |
50 | #include <linux/fcntl.h> | |
51 | #include <linux/init.h> | |
52 | #include <linux/skbuff.h> | |
53 | #include <linux/netdevice.h> | |
54 | #include <linux/etherdevice.h> | |
55 | #include <linux/miscdevice.h> | |
56 | #include <linux/ethtool.h> | |
57 | #include <linux/rtnetlink.h> | |
50857e2a | 58 | #include <linux/compat.h> |
1da177e4 LT |
59 | #include <linux/if.h> |
60 | #include <linux/if_arp.h> | |
61 | #include <linux/if_ether.h> | |
62 | #include <linux/if_tun.h> | |
6680ec68 | 63 | #include <linux/if_vlan.h> |
1da177e4 | 64 | #include <linux/crc32.h> |
d647a591 | 65 | #include <linux/nsproxy.h> |
f43798c2 | 66 | #include <linux/virtio_net.h> |
99405162 | 67 | #include <linux/rcupdate.h> |
881d966b | 68 | #include <net/net_namespace.h> |
79d17604 | 69 | #include <net/netns/generic.h> |
f019a7a5 | 70 | #include <net/rtnetlink.h> |
33dccbb0 | 71 | #include <net/sock.h> |
93e14b6d | 72 | #include <linux/seq_file.h> |
e0b46d0e | 73 | #include <linux/uio.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 | |
1cf8e410 MT |
112 | /* High bits in flags field are unused. */ |
113 | #define TUN_VNET_LE 0x80000000 | |
8b8e658b | 114 | #define TUN_VNET_BE 0x40000000 |
031f5e03 MT |
115 | |
116 | #define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \ | |
1cf8e410 | 117 | IFF_MULTI_QUEUE) |
0690899b MT |
118 | #define GOODCOPY_LEN 128 |
119 | ||
f271b2cc MK |
120 | #define FLT_EXACT_COUNT 8 |
121 | struct tap_filter { | |
122 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
123 | u32 mask[2]; /* Mask of the hashed addrs */ | |
124 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
125 | }; | |
126 | ||
baf71c5c PG |
127 | /* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal |
128 | * to max number of VCPUs in guest. */ | |
129 | #define MAX_TAP_QUEUES 256 | |
b8732fb7 | 130 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 131 | |
96442e42 JW |
132 | #define TUN_FLOW_EXPIRE (3 * HZ) |
133 | ||
608b9977 PA |
134 | struct tun_pcpu_stats { |
135 | u64 rx_packets; | |
136 | u64 rx_bytes; | |
137 | u64 tx_packets; | |
138 | u64 tx_bytes; | |
139 | struct u64_stats_sync syncp; | |
140 | u32 rx_dropped; | |
141 | u32 tx_dropped; | |
142 | u32 rx_frame_errors; | |
143 | }; | |
144 | ||
54f968d6 | 145 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 146 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
147 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
148 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 149 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 150 | * this). |
6e914fc7 JW |
151 | * |
152 | * RCU usage: | |
36fe8c09 | 153 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 154 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 155 | */ |
631ab46b | 156 | struct tun_file { |
54f968d6 JW |
157 | struct sock sk; |
158 | struct socket socket; | |
159 | struct socket_wq wq; | |
6e914fc7 | 160 | struct tun_struct __rcu *tun; |
54f968d6 JW |
161 | struct fasync_struct *fasync; |
162 | /* only used for fasnyc */ | |
163 | unsigned int flags; | |
fb7589a1 PE |
164 | union { |
165 | u16 queue_index; | |
166 | unsigned int ifindex; | |
167 | }; | |
4008e97f JW |
168 | struct list_head next; |
169 | struct tun_struct *detached; | |
631ab46b EB |
170 | }; |
171 | ||
96442e42 JW |
172 | struct tun_flow_entry { |
173 | struct hlist_node hash_link; | |
174 | struct rcu_head rcu; | |
175 | struct tun_struct *tun; | |
176 | ||
177 | u32 rxhash; | |
9bc88939 | 178 | u32 rps_rxhash; |
96442e42 JW |
179 | int queue_index; |
180 | unsigned long updated; | |
181 | }; | |
182 | ||
183 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
184 | ||
54f968d6 | 185 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 186 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
187 | * file were attached to a persist device. |
188 | */ | |
14daa021 | 189 | struct tun_struct { |
c8d68e6b JW |
190 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
191 | unsigned int numqueues; | |
f271b2cc | 192 | unsigned int flags; |
0625c883 EB |
193 | kuid_t owner; |
194 | kgid_t group; | |
14daa021 | 195 | |
14daa021 | 196 | struct net_device *dev; |
c8f44aff | 197 | netdev_features_t set_features; |
88255375 | 198 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
e3e3c423 | 199 | NETIF_F_TSO6|NETIF_F_UFO) |
d9d52b51 | 200 | |
eaea34b2 | 201 | int align; |
d9d52b51 | 202 | int vnet_hdr_sz; |
54f968d6 JW |
203 | int sndbuf; |
204 | struct tap_filter txflt; | |
205 | struct sock_fprog fprog; | |
206 | /* protected by rtnl lock */ | |
207 | bool filter_attached; | |
14daa021 RR |
208 | #ifdef TUN_DEBUG |
209 | int debug; | |
1da177e4 | 210 | #endif |
96442e42 | 211 | spinlock_t lock; |
96442e42 JW |
212 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
213 | struct timer_list flow_gc_timer; | |
214 | unsigned long ageing_time; | |
4008e97f JW |
215 | unsigned int numdisabled; |
216 | struct list_head disabled; | |
5dbbaf2d | 217 | void *security; |
b8732fb7 | 218 | u32 flow_count; |
608b9977 | 219 | struct tun_pcpu_stats __percpu *pcpu_stats; |
14daa021 | 220 | }; |
1da177e4 | 221 | |
8b8e658b GK |
222 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
223 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
224 | { | |
225 | return tun->flags & TUN_VNET_BE ? false : | |
226 | virtio_legacy_is_little_endian(); | |
227 | } | |
228 | ||
229 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
230 | { | |
231 | int be = !!(tun->flags & TUN_VNET_BE); | |
232 | ||
233 | if (put_user(be, argp)) | |
234 | return -EFAULT; | |
235 | ||
236 | return 0; | |
237 | } | |
238 | ||
239 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
240 | { | |
241 | int be; | |
242 | ||
243 | if (get_user(be, argp)) | |
244 | return -EFAULT; | |
245 | ||
246 | if (be) | |
247 | tun->flags |= TUN_VNET_BE; | |
248 | else | |
249 | tun->flags &= ~TUN_VNET_BE; | |
250 | ||
251 | return 0; | |
252 | } | |
253 | #else | |
254 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
255 | { | |
256 | return virtio_legacy_is_little_endian(); | |
257 | } | |
258 | ||
259 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
260 | { | |
261 | return -EINVAL; | |
262 | } | |
263 | ||
264 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
265 | { | |
266 | return -EINVAL; | |
267 | } | |
268 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
269 | ||
25bd55bb GK |
270 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
271 | { | |
7d824109 | 272 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 273 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
274 | } |
275 | ||
56f0dcc5 MT |
276 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
277 | { | |
25bd55bb | 278 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
279 | } |
280 | ||
281 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
282 | { | |
25bd55bb | 283 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
284 | } |
285 | ||
96442e42 JW |
286 | static inline u32 tun_hashfn(u32 rxhash) |
287 | { | |
288 | return rxhash & 0x3ff; | |
289 | } | |
290 | ||
291 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
292 | { | |
293 | struct tun_flow_entry *e; | |
96442e42 | 294 | |
b67bfe0d | 295 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
296 | if (e->rxhash == rxhash) |
297 | return e; | |
298 | } | |
299 | return NULL; | |
300 | } | |
301 | ||
302 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
303 | struct hlist_head *head, | |
304 | u32 rxhash, u16 queue_index) | |
305 | { | |
9fdc6bef ED |
306 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
307 | ||
96442e42 JW |
308 | if (e) { |
309 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
310 | rxhash, queue_index); | |
311 | e->updated = jiffies; | |
312 | e->rxhash = rxhash; | |
9bc88939 | 313 | e->rps_rxhash = 0; |
96442e42 JW |
314 | e->queue_index = queue_index; |
315 | e->tun = tun; | |
316 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 317 | ++tun->flow_count; |
96442e42 JW |
318 | } |
319 | return e; | |
320 | } | |
321 | ||
96442e42 JW |
322 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
323 | { | |
324 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
325 | e->rxhash, e->queue_index); | |
326 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 327 | kfree_rcu(e, rcu); |
b8732fb7 | 328 | --tun->flow_count; |
96442e42 JW |
329 | } |
330 | ||
331 | static void tun_flow_flush(struct tun_struct *tun) | |
332 | { | |
333 | int i; | |
334 | ||
335 | spin_lock_bh(&tun->lock); | |
336 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
337 | struct tun_flow_entry *e; | |
b67bfe0d | 338 | struct hlist_node *n; |
96442e42 | 339 | |
b67bfe0d | 340 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
341 | tun_flow_delete(tun, e); |
342 | } | |
343 | spin_unlock_bh(&tun->lock); | |
344 | } | |
345 | ||
346 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
347 | { | |
348 | int i; | |
349 | ||
350 | spin_lock_bh(&tun->lock); | |
351 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
352 | struct tun_flow_entry *e; | |
b67bfe0d | 353 | struct hlist_node *n; |
96442e42 | 354 | |
b67bfe0d | 355 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
356 | if (e->queue_index == queue_index) |
357 | tun_flow_delete(tun, e); | |
358 | } | |
359 | } | |
360 | spin_unlock_bh(&tun->lock); | |
361 | } | |
362 | ||
363 | static void tun_flow_cleanup(unsigned long data) | |
364 | { | |
365 | struct tun_struct *tun = (struct tun_struct *)data; | |
366 | unsigned long delay = tun->ageing_time; | |
367 | unsigned long next_timer = jiffies + delay; | |
368 | unsigned long count = 0; | |
369 | int i; | |
370 | ||
371 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
372 | ||
373 | spin_lock_bh(&tun->lock); | |
374 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
375 | struct tun_flow_entry *e; | |
b67bfe0d | 376 | struct hlist_node *n; |
96442e42 | 377 | |
b67bfe0d | 378 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
379 | unsigned long this_timer; |
380 | count++; | |
381 | this_timer = e->updated + delay; | |
382 | if (time_before_eq(this_timer, jiffies)) | |
383 | tun_flow_delete(tun, e); | |
384 | else if (time_before(this_timer, next_timer)) | |
385 | next_timer = this_timer; | |
386 | } | |
387 | } | |
388 | ||
389 | if (count) | |
390 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
391 | spin_unlock_bh(&tun->lock); | |
392 | } | |
393 | ||
49974420 | 394 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 395 | struct tun_file *tfile) |
96442e42 JW |
396 | { |
397 | struct hlist_head *head; | |
398 | struct tun_flow_entry *e; | |
399 | unsigned long delay = tun->ageing_time; | |
9e85722d | 400 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
401 | |
402 | if (!rxhash) | |
403 | return; | |
404 | else | |
405 | head = &tun->flows[tun_hashfn(rxhash)]; | |
406 | ||
407 | rcu_read_lock(); | |
408 | ||
9e85722d JW |
409 | /* We may get a very small possibility of OOO during switching, not |
410 | * worth to optimize.*/ | |
411 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
412 | goto unlock; |
413 | ||
414 | e = tun_flow_find(head, rxhash); | |
415 | if (likely(e)) { | |
416 | /* TODO: keep queueing to old queue until it's empty? */ | |
417 | e->queue_index = queue_index; | |
418 | e->updated = jiffies; | |
9bc88939 | 419 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
420 | } else { |
421 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
422 | if (!tun_flow_find(head, rxhash) && |
423 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
424 | tun_flow_create(tun, head, rxhash, queue_index); |
425 | ||
426 | if (!timer_pending(&tun->flow_gc_timer)) | |
427 | mod_timer(&tun->flow_gc_timer, | |
428 | round_jiffies_up(jiffies + delay)); | |
429 | spin_unlock_bh(&tun->lock); | |
430 | } | |
431 | ||
432 | unlock: | |
433 | rcu_read_unlock(); | |
434 | } | |
435 | ||
9bc88939 TH |
436 | /** |
437 | * Save the hash received in the stack receive path and update the | |
438 | * flow_hash table accordingly. | |
439 | */ | |
440 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
441 | { | |
567e4b79 | 442 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 443 | e->rps_rxhash = hash; |
9bc88939 TH |
444 | } |
445 | ||
c8d68e6b | 446 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 447 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
448 | * the rxq based on the txq where the last packet of the flow comes. As |
449 | * the userspace application move between processors, we may get a | |
450 | * different rxq no. here. If we could not get rxhash, then we would | |
451 | * hope the rxq no. may help here. | |
452 | */ | |
f663dd9a | 453 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 454 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
455 | { |
456 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 457 | struct tun_flow_entry *e; |
c8d68e6b JW |
458 | u32 txq = 0; |
459 | u32 numqueues = 0; | |
460 | ||
461 | rcu_read_lock(); | |
92bb73ea | 462 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 463 | |
3958afa1 | 464 | txq = skb_get_hash(skb); |
c8d68e6b | 465 | if (txq) { |
96442e42 | 466 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 467 | if (e) { |
9bc88939 | 468 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 469 | txq = e->queue_index; |
9bc88939 | 470 | } else |
96442e42 JW |
471 | /* use multiply and shift instead of expensive divide */ |
472 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
473 | } else if (likely(skb_rx_queue_recorded(skb))) { |
474 | txq = skb_get_rx_queue(skb); | |
475 | while (unlikely(txq >= numqueues)) | |
476 | txq -= numqueues; | |
477 | } | |
478 | ||
479 | rcu_read_unlock(); | |
480 | return txq; | |
481 | } | |
482 | ||
cde8b15f JW |
483 | static inline bool tun_not_capable(struct tun_struct *tun) |
484 | { | |
485 | const struct cred *cred = current_cred(); | |
c260b772 | 486 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
487 | |
488 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
489 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 490 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
491 | } |
492 | ||
c8d68e6b JW |
493 | static void tun_set_real_num_queues(struct tun_struct *tun) |
494 | { | |
495 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
496 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
497 | } | |
498 | ||
4008e97f JW |
499 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
500 | { | |
501 | tfile->detached = tun; | |
502 | list_add_tail(&tfile->next, &tun->disabled); | |
503 | ++tun->numdisabled; | |
504 | } | |
505 | ||
d32649d1 | 506 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
507 | { |
508 | struct tun_struct *tun = tfile->detached; | |
509 | ||
510 | tfile->detached = NULL; | |
511 | list_del_init(&tfile->next); | |
512 | --tun->numdisabled; | |
513 | return tun; | |
514 | } | |
515 | ||
4bfb0513 JW |
516 | static void tun_queue_purge(struct tun_file *tfile) |
517 | { | |
518 | skb_queue_purge(&tfile->sk.sk_receive_queue); | |
519 | skb_queue_purge(&tfile->sk.sk_error_queue); | |
520 | } | |
521 | ||
c8d68e6b JW |
522 | static void __tun_detach(struct tun_file *tfile, bool clean) |
523 | { | |
524 | struct tun_file *ntfile; | |
525 | struct tun_struct *tun; | |
c8d68e6b | 526 | |
b8deabd3 JW |
527 | tun = rtnl_dereference(tfile->tun); |
528 | ||
9e85722d | 529 | if (tun && !tfile->detached) { |
c8d68e6b JW |
530 | u16 index = tfile->queue_index; |
531 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
532 | |
533 | rcu_assign_pointer(tun->tfiles[index], | |
534 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 535 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
536 | ntfile->queue_index = index; |
537 | ||
538 | --tun->numqueues; | |
9e85722d | 539 | if (clean) { |
c956674b | 540 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 541 | sock_put(&tfile->sk); |
9e85722d | 542 | } else |
4008e97f | 543 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
544 | |
545 | synchronize_net(); | |
96442e42 | 546 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 547 | /* Drop read queue */ |
4bfb0513 | 548 | tun_queue_purge(tfile); |
c8d68e6b | 549 | tun_set_real_num_queues(tun); |
dd38bd85 | 550 | } else if (tfile->detached && clean) { |
4008e97f | 551 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
552 | sock_put(&tfile->sk); |
553 | } | |
c8d68e6b JW |
554 | |
555 | if (clean) { | |
af668b3c MT |
556 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
557 | netif_carrier_off(tun->dev); | |
558 | ||
40630b82 | 559 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 560 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 561 | unregister_netdevice(tun->dev); |
af668b3c | 562 | } |
140e807d | 563 | sock_put(&tfile->sk); |
c8d68e6b JW |
564 | } |
565 | } | |
566 | ||
567 | static void tun_detach(struct tun_file *tfile, bool clean) | |
568 | { | |
569 | rtnl_lock(); | |
570 | __tun_detach(tfile, clean); | |
571 | rtnl_unlock(); | |
572 | } | |
573 | ||
574 | static void tun_detach_all(struct net_device *dev) | |
575 | { | |
576 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 577 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
578 | int i, n = tun->numqueues; |
579 | ||
580 | for (i = 0; i < n; i++) { | |
b8deabd3 | 581 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 582 | BUG_ON(!tfile); |
9e641bdc | 583 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 584 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
585 | --tun->numqueues; |
586 | } | |
9e85722d | 587 | list_for_each_entry(tfile, &tun->disabled, next) { |
9e641bdc | 588 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 589 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 590 | } |
c8d68e6b JW |
591 | BUG_ON(tun->numqueues != 0); |
592 | ||
593 | synchronize_net(); | |
594 | for (i = 0; i < n; i++) { | |
b8deabd3 | 595 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 596 | /* Drop read queue */ |
4bfb0513 | 597 | tun_queue_purge(tfile); |
c8d68e6b JW |
598 | sock_put(&tfile->sk); |
599 | } | |
4008e97f JW |
600 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
601 | tun_enable_queue(tfile); | |
4bfb0513 | 602 | tun_queue_purge(tfile); |
4008e97f JW |
603 | sock_put(&tfile->sk); |
604 | } | |
605 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 606 | |
40630b82 | 607 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 608 | module_put(THIS_MODULE); |
c8d68e6b JW |
609 | } |
610 | ||
849c9b6f | 611 | static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter) |
a7385ba2 | 612 | { |
631ab46b | 613 | struct tun_file *tfile = file->private_data; |
38231b7a | 614 | int err; |
a7385ba2 | 615 | |
5dbbaf2d PM |
616 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
617 | if (err < 0) | |
618 | goto out; | |
619 | ||
38231b7a | 620 | err = -EINVAL; |
9e85722d | 621 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
622 | goto out; |
623 | ||
624 | err = -EBUSY; | |
40630b82 | 625 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
626 | goto out; |
627 | ||
628 | err = -E2BIG; | |
4008e97f JW |
629 | if (!tfile->detached && |
630 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
631 | goto out; |
632 | ||
633 | err = 0; | |
54f968d6 | 634 | |
92d4ea6e | 635 | /* Re-attach the filter to persist device */ |
849c9b6f | 636 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
637 | lock_sock(tfile->socket.sk); |
638 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
639 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
640 | if (!err) |
641 | goto out; | |
642 | } | |
c8d68e6b | 643 | tfile->queue_index = tun->numqueues; |
6e914fc7 | 644 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 645 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 646 | tun->numqueues++; |
a7385ba2 | 647 | |
4008e97f JW |
648 | if (tfile->detached) |
649 | tun_enable_queue(tfile); | |
650 | else | |
651 | sock_hold(&tfile->sk); | |
652 | ||
c8d68e6b | 653 | tun_set_real_num_queues(tun); |
a7385ba2 | 654 | |
c8d68e6b JW |
655 | /* device is allowed to go away first, so no need to hold extra |
656 | * refcnt. | |
657 | */ | |
658 | ||
659 | out: | |
660 | return err; | |
631ab46b EB |
661 | } |
662 | ||
663 | static struct tun_struct *__tun_get(struct tun_file *tfile) | |
664 | { | |
6e914fc7 | 665 | struct tun_struct *tun; |
c70f1829 | 666 | |
6e914fc7 JW |
667 | rcu_read_lock(); |
668 | tun = rcu_dereference(tfile->tun); | |
669 | if (tun) | |
670 | dev_hold(tun->dev); | |
671 | rcu_read_unlock(); | |
c70f1829 EB |
672 | |
673 | return tun; | |
631ab46b EB |
674 | } |
675 | ||
676 | static struct tun_struct *tun_get(struct file *file) | |
677 | { | |
678 | return __tun_get(file->private_data); | |
679 | } | |
680 | ||
681 | static void tun_put(struct tun_struct *tun) | |
682 | { | |
6e914fc7 | 683 | dev_put(tun->dev); |
631ab46b EB |
684 | } |
685 | ||
6b8a66ee | 686 | /* TAP filtering */ |
f271b2cc MK |
687 | static void addr_hash_set(u32 *mask, const u8 *addr) |
688 | { | |
689 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
690 | mask[n >> 5] |= (1 << (n & 31)); | |
691 | } | |
692 | ||
693 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
694 | { | |
695 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
696 | return mask[n >> 5] & (1 << (n & 31)); | |
697 | } | |
698 | ||
699 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
700 | { | |
701 | struct { u8 u[ETH_ALEN]; } *addr; | |
702 | struct tun_filter uf; | |
703 | int err, alen, n, nexact; | |
704 | ||
705 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
706 | return -EFAULT; | |
707 | ||
708 | if (!uf.count) { | |
709 | /* Disabled */ | |
710 | filter->count = 0; | |
711 | return 0; | |
712 | } | |
713 | ||
714 | alen = ETH_ALEN * uf.count; | |
715 | addr = kmalloc(alen, GFP_KERNEL); | |
716 | if (!addr) | |
717 | return -ENOMEM; | |
718 | ||
719 | if (copy_from_user(addr, arg + sizeof(uf), alen)) { | |
720 | err = -EFAULT; | |
721 | goto done; | |
722 | } | |
723 | ||
724 | /* The filter is updated without holding any locks. Which is | |
725 | * perfectly safe. We disable it first and in the worst | |
726 | * case we'll accept a few undesired packets. */ | |
727 | filter->count = 0; | |
728 | wmb(); | |
729 | ||
730 | /* Use first set of addresses as an exact filter */ | |
731 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
732 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
733 | ||
734 | nexact = n; | |
735 | ||
cfbf84fc AW |
736 | /* Remaining multicast addresses are hashed, |
737 | * unicast will leave the filter disabled. */ | |
f271b2cc | 738 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
739 | for (; n < uf.count; n++) { |
740 | if (!is_multicast_ether_addr(addr[n].u)) { | |
741 | err = 0; /* no filter */ | |
742 | goto done; | |
743 | } | |
f271b2cc | 744 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 745 | } |
f271b2cc MK |
746 | |
747 | /* For ALLMULTI just set the mask to all ones. | |
748 | * This overrides the mask populated above. */ | |
749 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
750 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
751 | ||
752 | /* Now enable the filter */ | |
753 | wmb(); | |
754 | filter->count = nexact; | |
755 | ||
756 | /* Return the number of exact filters */ | |
757 | err = nexact; | |
758 | ||
759 | done: | |
760 | kfree(addr); | |
761 | return err; | |
762 | } | |
763 | ||
764 | /* Returns: 0 - drop, !=0 - accept */ | |
765 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
766 | { | |
767 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
768 | * at this point. */ | |
769 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
770 | int i; | |
771 | ||
772 | /* Exact match */ | |
773 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 774 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
775 | return 1; |
776 | ||
777 | /* Inexact match (multicast only) */ | |
778 | if (is_multicast_ether_addr(eh->h_dest)) | |
779 | return addr_hash_test(filter->mask, eh->h_dest); | |
780 | ||
781 | return 0; | |
782 | } | |
783 | ||
784 | /* | |
785 | * Checks whether the packet is accepted or not. | |
786 | * Returns: 0 - drop, !=0 - accept | |
787 | */ | |
788 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
789 | { | |
790 | if (!filter->count) | |
791 | return 1; | |
792 | ||
793 | return run_filter(filter, skb); | |
794 | } | |
795 | ||
1da177e4 LT |
796 | /* Network device part of the driver */ |
797 | ||
7282d491 | 798 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 799 | |
c70f1829 EB |
800 | /* Net device detach from fd. */ |
801 | static void tun_net_uninit(struct net_device *dev) | |
802 | { | |
c8d68e6b | 803 | tun_detach_all(dev); |
c70f1829 EB |
804 | } |
805 | ||
1da177e4 LT |
806 | /* Net device open. */ |
807 | static int tun_net_open(struct net_device *dev) | |
808 | { | |
c8d68e6b | 809 | netif_tx_start_all_queues(dev); |
1da177e4 LT |
810 | return 0; |
811 | } | |
812 | ||
813 | /* Net device close. */ | |
814 | static int tun_net_close(struct net_device *dev) | |
815 | { | |
c8d68e6b | 816 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
817 | return 0; |
818 | } | |
819 | ||
820 | /* Net device start xmit */ | |
424efe9c | 821 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
822 | { |
823 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 824 | int txq = skb->queue_mapping; |
6e914fc7 | 825 | struct tun_file *tfile; |
fa35864e | 826 | u32 numqueues = 0; |
1da177e4 | 827 | |
6e914fc7 | 828 | rcu_read_lock(); |
c8d68e6b | 829 | tfile = rcu_dereference(tun->tfiles[txq]); |
fa35864e | 830 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 831 | |
1da177e4 | 832 | /* Drop packet if interface is not attached */ |
fa35864e | 833 | if (txq >= numqueues) |
1da177e4 LT |
834 | goto drop; |
835 | ||
fa35864e | 836 | if (numqueues == 1) { |
9bc88939 TH |
837 | /* Select queue was not called for the skbuff, so we extract the |
838 | * RPS hash and save it into the flow_table here. | |
839 | */ | |
840 | __u32 rxhash; | |
841 | ||
842 | rxhash = skb_get_hash(skb); | |
843 | if (rxhash) { | |
844 | struct tun_flow_entry *e; | |
845 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
846 | rxhash); | |
847 | if (e) | |
848 | tun_flow_save_rps_rxhash(e, rxhash); | |
849 | } | |
850 | } | |
851 | ||
6e914fc7 JW |
852 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
853 | ||
c8d68e6b JW |
854 | BUG_ON(!tfile); |
855 | ||
f271b2cc MK |
856 | /* Drop if the filter does not like it. |
857 | * This is a noop if the filter is disabled. | |
858 | * Filter can be enabled only for the TAP devices. */ | |
859 | if (!check_filter(&tun->txflt, skb)) | |
860 | goto drop; | |
861 | ||
54f968d6 JW |
862 | if (tfile->socket.sk->sk_filter && |
863 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
864 | goto drop; |
865 | ||
36fe8c09 | 866 | /* Limit the number of packets queued by dividing txq length with the |
c8d68e6b JW |
867 | * number of queues. |
868 | */ | |
fa35864e DC |
869 | if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues |
870 | >= dev->tx_queue_len) | |
5d097109 | 871 | goto drop; |
1da177e4 | 872 | |
7bf66305 JW |
873 | if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC))) |
874 | goto drop; | |
875 | ||
5fcd2d8b | 876 | if (skb->sk && sk_fullsock(skb->sk)) { |
c14ac945 SHY |
877 | sock_tx_timestamp(skb->sk, skb->sk->sk_tsflags, |
878 | &skb_shinfo(skb)->tx_flags); | |
eda29772 RC |
879 | sw_tx_timestamp(skb); |
880 | } | |
881 | ||
0110d6f2 | 882 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
883 | * for indefinite time. |
884 | */ | |
0110d6f2 MT |
885 | skb_orphan(skb); |
886 | ||
f8af75f3 ED |
887 | nf_reset(skb); |
888 | ||
f271b2cc | 889 | /* Enqueue packet */ |
54f968d6 | 890 | skb_queue_tail(&tfile->socket.sk->sk_receive_queue, skb); |
1da177e4 LT |
891 | |
892 | /* Notify and wake up reader process */ | |
54f968d6 JW |
893 | if (tfile->flags & TUN_FASYNC) |
894 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 895 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
896 | |
897 | rcu_read_unlock(); | |
6ed10654 | 898 | return NETDEV_TX_OK; |
1da177e4 LT |
899 | |
900 | drop: | |
608b9977 | 901 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
149d36f7 | 902 | skb_tx_error(skb); |
1da177e4 | 903 | kfree_skb(skb); |
6e914fc7 | 904 | rcu_read_unlock(); |
baeababb | 905 | return NET_XMIT_DROP; |
1da177e4 LT |
906 | } |
907 | ||
f271b2cc | 908 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 909 | { |
f271b2cc MK |
910 | /* |
911 | * This callback is supposed to deal with mc filter in | |
912 | * _rx_ path and has nothing to do with the _tx_ path. | |
913 | * In rx path we always accept everything userspace gives us. | |
914 | */ | |
1da177e4 LT |
915 | } |
916 | ||
4885a504 ES |
917 | #define MIN_MTU 68 |
918 | #define MAX_MTU 65535 | |
919 | ||
920 | static int | |
921 | tun_net_change_mtu(struct net_device *dev, int new_mtu) | |
922 | { | |
923 | if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU) | |
924 | return -EINVAL; | |
925 | dev->mtu = new_mtu; | |
926 | return 0; | |
927 | } | |
928 | ||
c8f44aff MM |
929 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
930 | netdev_features_t features) | |
88255375 MM |
931 | { |
932 | struct tun_struct *tun = netdev_priv(dev); | |
933 | ||
934 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
935 | } | |
bebd097a NH |
936 | #ifdef CONFIG_NET_POLL_CONTROLLER |
937 | static void tun_poll_controller(struct net_device *dev) | |
938 | { | |
939 | /* | |
940 | * Tun only receives frames when: | |
941 | * 1) the char device endpoint gets data from user space | |
942 | * 2) the tun socket gets a sendmsg call from user space | |
92d4ea6e | 943 | * Since both of those are synchronous operations, we are guaranteed |
bebd097a | 944 | * never to have pending data when we poll for it |
92d4ea6e | 945 | * so there is nothing to do here but return. |
bebd097a NH |
946 | * We need this though so netpoll recognizes us as an interface that |
947 | * supports polling, which enables bridge devices in virt setups to | |
948 | * still use netconsole | |
949 | */ | |
950 | return; | |
951 | } | |
952 | #endif | |
eaea34b2 PA |
953 | |
954 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
955 | { | |
956 | struct tun_struct *tun = netdev_priv(dev); | |
957 | ||
958 | if (new_hr < NET_SKB_PAD) | |
959 | new_hr = NET_SKB_PAD; | |
960 | ||
961 | tun->align = new_hr; | |
962 | } | |
963 | ||
608b9977 PA |
964 | static struct rtnl_link_stats64 * |
965 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) | |
966 | { | |
967 | u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; | |
968 | struct tun_struct *tun = netdev_priv(dev); | |
969 | struct tun_pcpu_stats *p; | |
970 | int i; | |
971 | ||
972 | for_each_possible_cpu(i) { | |
973 | u64 rxpackets, rxbytes, txpackets, txbytes; | |
974 | unsigned int start; | |
975 | ||
976 | p = per_cpu_ptr(tun->pcpu_stats, i); | |
977 | do { | |
978 | start = u64_stats_fetch_begin(&p->syncp); | |
979 | rxpackets = p->rx_packets; | |
980 | rxbytes = p->rx_bytes; | |
981 | txpackets = p->tx_packets; | |
982 | txbytes = p->tx_bytes; | |
983 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
984 | ||
985 | stats->rx_packets += rxpackets; | |
986 | stats->rx_bytes += rxbytes; | |
987 | stats->tx_packets += txpackets; | |
988 | stats->tx_bytes += txbytes; | |
989 | ||
990 | /* u32 counters */ | |
991 | rx_dropped += p->rx_dropped; | |
992 | rx_frame_errors += p->rx_frame_errors; | |
993 | tx_dropped += p->tx_dropped; | |
994 | } | |
995 | stats->rx_dropped = rx_dropped; | |
996 | stats->rx_frame_errors = rx_frame_errors; | |
997 | stats->tx_dropped = tx_dropped; | |
998 | return stats; | |
999 | } | |
1000 | ||
758e43b7 | 1001 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 1002 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1003 | .ndo_open = tun_net_open, |
1004 | .ndo_stop = tun_net_close, | |
00829823 | 1005 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 1006 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 1007 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1008 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1009 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1010 | .ndo_poll_controller = tun_poll_controller, | |
1011 | #endif | |
eaea34b2 | 1012 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1013 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1014 | }; |
1015 | ||
1016 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 1017 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1018 | .ndo_open = tun_net_open, |
1019 | .ndo_stop = tun_net_close, | |
00829823 | 1020 | .ndo_start_xmit = tun_net_xmit, |
758e43b7 | 1021 | .ndo_change_mtu = tun_net_change_mtu, |
88255375 | 1022 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1023 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1024 | .ndo_set_mac_address = eth_mac_addr, |
1025 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1026 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1027 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1028 | .ndo_poll_controller = tun_poll_controller, | |
1029 | #endif | |
5e52796a | 1030 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1031 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1032 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1033 | }; |
1034 | ||
944a1376 | 1035 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1036 | { |
1037 | int i; | |
1038 | ||
96442e42 JW |
1039 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1040 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1041 | ||
1042 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
1043 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
1044 | mod_timer(&tun->flow_gc_timer, | |
1045 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1046 | } |
1047 | ||
1048 | static void tun_flow_uninit(struct tun_struct *tun) | |
1049 | { | |
1050 | del_timer_sync(&tun->flow_gc_timer); | |
1051 | tun_flow_flush(tun); | |
96442e42 JW |
1052 | } |
1053 | ||
1da177e4 LT |
1054 | /* Initialize net device. */ |
1055 | static void tun_net_init(struct net_device *dev) | |
1056 | { | |
1057 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1058 | |
1da177e4 | 1059 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1060 | case IFF_TUN: |
758e43b7 SH |
1061 | dev->netdev_ops = &tun_netdev_ops; |
1062 | ||
1da177e4 LT |
1063 | /* Point-to-Point TUN Device */ |
1064 | dev->hard_header_len = 0; | |
1065 | dev->addr_len = 0; | |
1066 | dev->mtu = 1500; | |
1067 | ||
1068 | /* Zero header length */ | |
6aa20a22 | 1069 | dev->type = ARPHRD_NONE; |
1da177e4 | 1070 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1071 | break; |
1072 | ||
40630b82 | 1073 | case IFF_TAP: |
7a0a9608 | 1074 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1075 | /* Ethernet TAP Device */ |
1da177e4 | 1076 | ether_setup(dev); |
550fd08c | 1077 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1078 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1079 | |
f2cedb63 | 1080 | eth_hw_addr_random(dev); |
36226a8d | 1081 | |
1da177e4 LT |
1082 | break; |
1083 | } | |
1084 | } | |
1085 | ||
1086 | /* Character device part */ | |
1087 | ||
1088 | /* Poll */ | |
c8d68e6b | 1089 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1090 | { |
b2430de3 EB |
1091 | struct tun_file *tfile = file->private_data; |
1092 | struct tun_struct *tun = __tun_get(tfile); | |
3c8a9c63 | 1093 | struct sock *sk; |
33dccbb0 | 1094 | unsigned int mask = 0; |
1da177e4 LT |
1095 | |
1096 | if (!tun) | |
eac9e902 | 1097 | return POLLERR; |
1da177e4 | 1098 | |
54f968d6 | 1099 | sk = tfile->socket.sk; |
3c8a9c63 | 1100 | |
6b8a66ee | 1101 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 1102 | |
9e641bdc | 1103 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1104 | |
89f56d1e | 1105 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
1da177e4 LT |
1106 | mask |= POLLIN | POLLRDNORM; |
1107 | ||
33dccbb0 | 1108 | if (sock_writeable(sk) || |
9cd3e072 | 1109 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && |
33dccbb0 HX |
1110 | sock_writeable(sk))) |
1111 | mask |= POLLOUT | POLLWRNORM; | |
1112 | ||
c70f1829 EB |
1113 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1114 | mask = POLLERR; | |
1115 | ||
631ab46b | 1116 | tun_put(tun); |
1da177e4 LT |
1117 | return mask; |
1118 | } | |
1119 | ||
f42157cb RR |
1120 | /* prepad is the amount to reserve at front. len is length after that. |
1121 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1122 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1123 | size_t prepad, size_t len, |
1124 | size_t linear, int noblock) | |
f42157cb | 1125 | { |
54f968d6 | 1126 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1127 | struct sk_buff *skb; |
33dccbb0 | 1128 | int err; |
f42157cb RR |
1129 | |
1130 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1131 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1132 | linear = len; |
f42157cb | 1133 | |
33dccbb0 | 1134 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1135 | &err, 0); |
f42157cb | 1136 | if (!skb) |
33dccbb0 | 1137 | return ERR_PTR(err); |
f42157cb RR |
1138 | |
1139 | skb_reserve(skb, prepad); | |
1140 | skb_put(skb, linear); | |
33dccbb0 HX |
1141 | skb->data_len = len - linear; |
1142 | skb->len += len - linear; | |
f42157cb RR |
1143 | |
1144 | return skb; | |
1145 | } | |
1146 | ||
1da177e4 | 1147 | /* Get packet from user space buffer */ |
54f968d6 | 1148 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 AV |
1149 | void *msg_control, struct iov_iter *from, |
1150 | int noblock) | |
1da177e4 | 1151 | { |
09640e63 | 1152 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1153 | struct sk_buff *skb; |
f5ff53b4 | 1154 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1155 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1156 | struct virtio_net_hdr gso = { 0 }; |
608b9977 | 1157 | struct tun_pcpu_stats *stats; |
96f8d9ec | 1158 | int good_linear; |
0690899b MT |
1159 | int copylen; |
1160 | bool zerocopy = false; | |
1161 | int err; | |
49974420 | 1162 | u32 rxhash; |
f5ff53b4 | 1163 | ssize_t n; |
1da177e4 | 1164 | |
1bd4978a ED |
1165 | if (!(tun->dev->flags & IFF_UP)) |
1166 | return -EIO; | |
1167 | ||
40630b82 | 1168 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1169 | if (len < sizeof(pi)) |
1da177e4 | 1170 | return -EINVAL; |
15718ea0 | 1171 | len -= sizeof(pi); |
1da177e4 | 1172 | |
f5ff53b4 AV |
1173 | n = copy_from_iter(&pi, sizeof(pi), from); |
1174 | if (n != sizeof(pi)) | |
1da177e4 LT |
1175 | return -EFAULT; |
1176 | } | |
1177 | ||
40630b82 | 1178 | if (tun->flags & IFF_VNET_HDR) { |
15718ea0 | 1179 | if (len < tun->vnet_hdr_sz) |
f43798c2 | 1180 | return -EINVAL; |
15718ea0 | 1181 | len -= tun->vnet_hdr_sz; |
f43798c2 | 1182 | |
f5ff53b4 AV |
1183 | n = copy_from_iter(&gso, sizeof(gso), from); |
1184 | if (n != sizeof(gso)) | |
f43798c2 RR |
1185 | return -EFAULT; |
1186 | ||
4909122f | 1187 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1188 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1189 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1190 | |
56f0dcc5 | 1191 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1192 | return -EINVAL; |
d8febb77 | 1193 | iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1194 | } |
1195 | ||
40630b82 | 1196 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1197 | align += NET_IP_ALIGN; |
0eca93bc | 1198 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1199 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1200 | return -EINVAL; |
1201 | } | |
6aa20a22 | 1202 | |
96f8d9ec JW |
1203 | good_linear = SKB_MAX_HEAD(align); |
1204 | ||
88529176 | 1205 | if (msg_control) { |
f5ff53b4 AV |
1206 | struct iov_iter i = *from; |
1207 | ||
88529176 JW |
1208 | /* There are 256 bytes to be copied in skb, so there is |
1209 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1210 | * The rest of the buffer is mapped from userspace. |
1211 | */ | |
56f0dcc5 | 1212 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1213 | if (copylen > good_linear) |
1214 | copylen = good_linear; | |
3dd5c330 | 1215 | linear = copylen; |
f5ff53b4 AV |
1216 | iov_iter_advance(&i, copylen); |
1217 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1218 | zerocopy = true; |
1219 | } | |
1220 | ||
1221 | if (!zerocopy) { | |
0690899b | 1222 | copylen = len; |
56f0dcc5 | 1223 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) |
96f8d9ec JW |
1224 | linear = good_linear; |
1225 | else | |
56f0dcc5 | 1226 | linear = tun16_to_cpu(tun, gso.hdr_len); |
3dd5c330 | 1227 | } |
0690899b | 1228 | |
3dd5c330 | 1229 | skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); |
33dccbb0 HX |
1230 | if (IS_ERR(skb)) { |
1231 | if (PTR_ERR(skb) != -EAGAIN) | |
608b9977 | 1232 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
33dccbb0 | 1233 | return PTR_ERR(skb); |
1da177e4 LT |
1234 | } |
1235 | ||
0690899b | 1236 | if (zerocopy) |
f5ff53b4 | 1237 | err = zerocopy_sg_from_iter(skb, from); |
88529176 | 1238 | else { |
f5ff53b4 | 1239 | err = skb_copy_datagram_from_iter(skb, 0, from, len); |
88529176 JW |
1240 | if (!err && msg_control) { |
1241 | struct ubuf_info *uarg = msg_control; | |
1242 | uarg->callback(uarg, false); | |
1243 | } | |
1244 | } | |
0690899b MT |
1245 | |
1246 | if (err) { | |
608b9977 | 1247 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
8f22757e | 1248 | kfree_skb(skb); |
1da177e4 | 1249 | return -EFAULT; |
8f22757e | 1250 | } |
1da177e4 | 1251 | |
f43798c2 | 1252 | if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { |
56f0dcc5 MT |
1253 | if (!skb_partial_csum_set(skb, tun16_to_cpu(tun, gso.csum_start), |
1254 | tun16_to_cpu(tun, gso.csum_offset))) { | |
608b9977 | 1255 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
f43798c2 RR |
1256 | kfree_skb(skb); |
1257 | return -EINVAL; | |
1258 | } | |
88255375 | 1259 | } |
f43798c2 | 1260 | |
1da177e4 | 1261 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1262 | case IFF_TUN: |
1263 | if (tun->flags & IFF_NO_PI) { | |
f09f7ee2 AWC |
1264 | switch (skb->data[0] & 0xf0) { |
1265 | case 0x40: | |
1266 | pi.proto = htons(ETH_P_IP); | |
1267 | break; | |
1268 | case 0x60: | |
1269 | pi.proto = htons(ETH_P_IPV6); | |
1270 | break; | |
1271 | default: | |
608b9977 | 1272 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
f09f7ee2 AWC |
1273 | kfree_skb(skb); |
1274 | return -EINVAL; | |
1275 | } | |
1276 | } | |
1277 | ||
459a98ed | 1278 | skb_reset_mac_header(skb); |
1da177e4 | 1279 | skb->protocol = pi.proto; |
4c13eb66 | 1280 | skb->dev = tun->dev; |
1da177e4 | 1281 | break; |
40630b82 | 1282 | case IFF_TAP: |
1da177e4 LT |
1283 | skb->protocol = eth_type_trans(skb, tun->dev); |
1284 | break; | |
6403eab1 | 1285 | } |
1da177e4 | 1286 | |
f43798c2 RR |
1287 | if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) { |
1288 | pr_debug("GSO!\n"); | |
1289 | switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { | |
1290 | case VIRTIO_NET_HDR_GSO_TCPV4: | |
c9af6db4 | 1291 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; |
f43798c2 RR |
1292 | break; |
1293 | case VIRTIO_NET_HDR_GSO_TCPV6: | |
c9af6db4 | 1294 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; |
f43798c2 | 1295 | break; |
e36aa25a | 1296 | case VIRTIO_NET_HDR_GSO_UDP: |
c9af6db4 | 1297 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
e36aa25a | 1298 | break; |
f43798c2 | 1299 | default: |
608b9977 | 1300 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
f43798c2 RR |
1301 | kfree_skb(skb); |
1302 | return -EINVAL; | |
1303 | } | |
1304 | ||
1305 | if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN) | |
c9af6db4 | 1306 | skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; |
f43798c2 | 1307 | |
56f0dcc5 | 1308 | skb_shinfo(skb)->gso_size = tun16_to_cpu(tun, gso.gso_size); |
f43798c2 | 1309 | if (skb_shinfo(skb)->gso_size == 0) { |
608b9977 | 1310 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
f43798c2 RR |
1311 | kfree_skb(skb); |
1312 | return -EINVAL; | |
1313 | } | |
1314 | ||
1315 | /* Header must be checked, and gso_segs computed. */ | |
1316 | skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY; | |
1317 | skb_shinfo(skb)->gso_segs = 0; | |
1318 | } | |
6aa20a22 | 1319 | |
0690899b MT |
1320 | /* copy skb_ubuf_info for callback when skb has no error */ |
1321 | if (zerocopy) { | |
1322 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1323 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1324 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
0690899b MT |
1325 | } |
1326 | ||
72f65107 | 1327 | skb_reset_network_header(skb); |
40893fd0 | 1328 | skb_probe_transport_header(skb, 0); |
38502af7 | 1329 | |
3958afa1 | 1330 | rxhash = skb_get_hash(skb); |
1da177e4 | 1331 | netif_rx_ni(skb); |
6aa20a22 | 1332 | |
608b9977 PA |
1333 | stats = get_cpu_ptr(tun->pcpu_stats); |
1334 | u64_stats_update_begin(&stats->syncp); | |
1335 | stats->rx_packets++; | |
1336 | stats->rx_bytes += len; | |
1337 | u64_stats_update_end(&stats->syncp); | |
1338 | put_cpu_ptr(stats); | |
1da177e4 | 1339 | |
9e85722d | 1340 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1341 | return total_len; |
6aa20a22 | 1342 | } |
1da177e4 | 1343 | |
f5ff53b4 | 1344 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1345 | { |
33dccbb0 | 1346 | struct file *file = iocb->ki_filp; |
ab46d779 | 1347 | struct tun_struct *tun = tun_get(file); |
54f968d6 | 1348 | struct tun_file *tfile = file->private_data; |
631ab46b | 1349 | ssize_t result; |
1da177e4 LT |
1350 | |
1351 | if (!tun) | |
1352 | return -EBADFD; | |
1353 | ||
f5ff53b4 | 1354 | result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK); |
631ab46b EB |
1355 | |
1356 | tun_put(tun); | |
1357 | return result; | |
1da177e4 LT |
1358 | } |
1359 | ||
1da177e4 | 1360 | /* Put packet to the user space buffer */ |
6f7c156c | 1361 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1362 | struct tun_file *tfile, |
6f7c156c | 1363 | struct sk_buff *skb, |
e0b46d0e | 1364 | struct iov_iter *iter) |
1da177e4 LT |
1365 | { |
1366 | struct tun_pi pi = { 0, skb->protocol }; | |
608b9977 | 1367 | struct tun_pcpu_stats *stats; |
e0b46d0e | 1368 | ssize_t total; |
8c847d25 | 1369 | int vlan_offset = 0; |
a8f9bfdf | 1370 | int vlan_hlen = 0; |
2eb783c4 | 1371 | int vnet_hdr_sz = 0; |
a8f9bfdf | 1372 | |
df8a39de | 1373 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 1374 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 1375 | |
40630b82 | 1376 | if (tun->flags & IFF_VNET_HDR) |
2eb783c4 | 1377 | vnet_hdr_sz = tun->vnet_hdr_sz; |
1da177e4 | 1378 | |
e0b46d0e HX |
1379 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
1380 | ||
40630b82 | 1381 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 1382 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
1383 | return -EINVAL; |
1384 | ||
e0b46d0e HX |
1385 | total += sizeof(pi); |
1386 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
1387 | /* Packet will be striped */ |
1388 | pi.flags |= TUN_PKT_STRIP; | |
1389 | } | |
6aa20a22 | 1390 | |
e0b46d0e | 1391 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 1392 | return -EFAULT; |
6aa20a22 | 1393 | } |
1da177e4 | 1394 | |
2eb783c4 | 1395 | if (vnet_hdr_sz) { |
f43798c2 | 1396 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ |
e0b46d0e | 1397 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
1398 | return -EINVAL; |
1399 | ||
1400 | if (skb_is_gso(skb)) { | |
1401 | struct skb_shared_info *sinfo = skb_shinfo(skb); | |
1402 | ||
1403 | /* This is a hint as to how much should be linear. */ | |
56f0dcc5 MT |
1404 | gso.hdr_len = cpu_to_tun16(tun, skb_headlen(skb)); |
1405 | gso.gso_size = cpu_to_tun16(tun, sinfo->gso_size); | |
f43798c2 RR |
1406 | if (sinfo->gso_type & SKB_GSO_TCPV4) |
1407 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4; | |
1408 | else if (sinfo->gso_type & SKB_GSO_TCPV6) | |
1409 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6; | |
e3e3c423 VY |
1410 | else if (sinfo->gso_type & SKB_GSO_UDP) |
1411 | gso.gso_type = VIRTIO_NET_HDR_GSO_UDP; | |
ef3db4a5 | 1412 | else { |
6b8a66ee | 1413 | pr_err("unexpected GSO type: " |
ef3db4a5 | 1414 | "0x%x, gso_size %d, hdr_len %d\n", |
56f0dcc5 MT |
1415 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), |
1416 | tun16_to_cpu(tun, gso.hdr_len)); | |
ef3db4a5 MT |
1417 | print_hex_dump(KERN_ERR, "tun: ", |
1418 | DUMP_PREFIX_NONE, | |
1419 | 16, 1, skb->head, | |
56f0dcc5 | 1420 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); |
ef3db4a5 MT |
1421 | WARN_ON_ONCE(1); |
1422 | return -EINVAL; | |
1423 | } | |
f43798c2 RR |
1424 | if (sinfo->gso_type & SKB_GSO_TCP_ECN) |
1425 | gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN; | |
1426 | } else | |
1427 | gso.gso_type = VIRTIO_NET_HDR_GSO_NONE; | |
1428 | ||
1429 | if (skb->ip_summed == CHECKSUM_PARTIAL) { | |
1430 | gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; | |
56f0dcc5 MT |
1431 | gso.csum_start = cpu_to_tun16(tun, skb_checksum_start_offset(skb) + |
1432 | vlan_hlen); | |
1433 | gso.csum_offset = cpu_to_tun16(tun, skb->csum_offset); | |
10a8d94a JW |
1434 | } else if (skb->ip_summed == CHECKSUM_UNNECESSARY) { |
1435 | gso.flags = VIRTIO_NET_HDR_F_DATA_VALID; | |
f43798c2 RR |
1436 | } /* else everything is zero */ |
1437 | ||
e0b46d0e | 1438 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 1439 | return -EFAULT; |
8c847d25 JW |
1440 | |
1441 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
1442 | } |
1443 | ||
a8f9bfdf | 1444 | if (vlan_hlen) { |
e0b46d0e | 1445 | int ret; |
6680ec68 JW |
1446 | struct { |
1447 | __be16 h_vlan_proto; | |
1448 | __be16 h_vlan_TCI; | |
1449 | } veth; | |
1450 | ||
1451 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 1452 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
1453 | |
1454 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 1455 | |
e0b46d0e HX |
1456 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
1457 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
1458 | goto done; |
1459 | ||
e0b46d0e HX |
1460 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
1461 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
1462 | goto done; |
1463 | } | |
1da177e4 | 1464 | |
e0b46d0e | 1465 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 1466 | |
6680ec68 | 1467 | done: |
608b9977 PA |
1468 | /* caller is in process context, */ |
1469 | stats = get_cpu_ptr(tun->pcpu_stats); | |
1470 | u64_stats_update_begin(&stats->syncp); | |
1471 | stats->tx_packets++; | |
1472 | stats->tx_bytes += skb->len + vlan_hlen; | |
1473 | u64_stats_update_end(&stats->syncp); | |
1474 | put_cpu_ptr(tun->pcpu_stats); | |
1da177e4 LT |
1475 | |
1476 | return total; | |
1477 | } | |
1478 | ||
54f968d6 | 1479 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 AV |
1480 | struct iov_iter *to, |
1481 | int noblock) | |
1da177e4 | 1482 | { |
1da177e4 | 1483 | struct sk_buff *skb; |
9b067034 | 1484 | ssize_t ret; |
9e641bdc | 1485 | int peeked, err, off = 0; |
1da177e4 | 1486 | |
3872baf6 | 1487 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1488 | |
9b067034 AV |
1489 | if (!iov_iter_count(to)) |
1490 | return 0; | |
1da177e4 | 1491 | |
9e641bdc XW |
1492 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1493 | return -EIO; | |
1da177e4 | 1494 | |
9e641bdc XW |
1495 | /* Read frames from queue */ |
1496 | skb = __skb_recv_datagram(tfile->socket.sk, noblock ? MSG_DONTWAIT : 0, | |
1497 | &peeked, &off, &err); | |
e0b46d0e | 1498 | if (!skb) |
957f094f | 1499 | return err; |
e0b46d0e | 1500 | |
9b067034 | 1501 | ret = tun_put_user(tun, tfile, skb, to); |
f51a5e82 | 1502 | if (unlikely(ret < 0)) |
f271b2cc | 1503 | kfree_skb(skb); |
f51a5e82 JW |
1504 | else |
1505 | consume_skb(skb); | |
1da177e4 | 1506 | |
05c2828c MT |
1507 | return ret; |
1508 | } | |
1509 | ||
9b067034 | 1510 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
1511 | { |
1512 | struct file *file = iocb->ki_filp; | |
1513 | struct tun_file *tfile = file->private_data; | |
1514 | struct tun_struct *tun = __tun_get(tfile); | |
9b067034 | 1515 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
1516 | |
1517 | if (!tun) | |
1518 | return -EBADFD; | |
9b067034 | 1519 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK); |
42404c09 | 1520 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
1521 | if (ret > 0) |
1522 | iocb->ki_pos = ret; | |
631ab46b | 1523 | tun_put(tun); |
1da177e4 LT |
1524 | return ret; |
1525 | } | |
1526 | ||
96442e42 JW |
1527 | static void tun_free_netdev(struct net_device *dev) |
1528 | { | |
1529 | struct tun_struct *tun = netdev_priv(dev); | |
1530 | ||
4008e97f | 1531 | BUG_ON(!(list_empty(&tun->disabled))); |
608b9977 | 1532 | free_percpu(tun->pcpu_stats); |
96442e42 | 1533 | tun_flow_uninit(tun); |
5dbbaf2d | 1534 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1535 | free_netdev(dev); |
1536 | } | |
1537 | ||
1da177e4 LT |
1538 | static void tun_setup(struct net_device *dev) |
1539 | { | |
1540 | struct tun_struct *tun = netdev_priv(dev); | |
1541 | ||
0625c883 EB |
1542 | tun->owner = INVALID_UID; |
1543 | tun->group = INVALID_GID; | |
1da177e4 | 1544 | |
1da177e4 | 1545 | dev->ethtool_ops = &tun_ethtool_ops; |
96442e42 | 1546 | dev->destructor = tun_free_netdev; |
016adb72 JW |
1547 | /* We prefer our own queue length */ |
1548 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
1549 | } |
1550 | ||
f019a7a5 EB |
1551 | /* Trivial set of netlink ops to allow deleting tun or tap |
1552 | * device with netlink. | |
1553 | */ | |
1554 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1555 | { | |
1556 | return -EINVAL; | |
1557 | } | |
1558 | ||
1559 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
1560 | .kind = DRV_NAME, | |
1561 | .priv_size = sizeof(struct tun_struct), | |
1562 | .setup = tun_setup, | |
1563 | .validate = tun_validate, | |
1564 | }; | |
1565 | ||
33dccbb0 HX |
1566 | static void tun_sock_write_space(struct sock *sk) |
1567 | { | |
54f968d6 | 1568 | struct tun_file *tfile; |
43815482 | 1569 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
1570 | |
1571 | if (!sock_writeable(sk)) | |
1572 | return; | |
1573 | ||
9cd3e072 | 1574 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
1575 | return; |
1576 | ||
43815482 ED |
1577 | wqueue = sk_sleep(sk); |
1578 | if (wqueue && waitqueue_active(wqueue)) | |
1579 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 1580 | POLLWRNORM | POLLWRBAND); |
c722c625 | 1581 | |
54f968d6 JW |
1582 | tfile = container_of(sk, struct tun_file, sk); |
1583 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
1584 | } |
1585 | ||
1b784140 | 1586 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 1587 | { |
54f968d6 JW |
1588 | int ret; |
1589 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1590 | struct tun_struct *tun = __tun_get(tfile); | |
1591 | ||
1592 | if (!tun) | |
1593 | return -EBADFD; | |
f5ff53b4 | 1594 | |
c0371da6 | 1595 | ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter, |
f5ff53b4 | 1596 | m->msg_flags & MSG_DONTWAIT); |
54f968d6 JW |
1597 | tun_put(tun); |
1598 | return ret; | |
05c2828c MT |
1599 | } |
1600 | ||
1b784140 | 1601 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
1602 | int flags) |
1603 | { | |
54f968d6 JW |
1604 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
1605 | struct tun_struct *tun = __tun_get(tfile); | |
05c2828c | 1606 | int ret; |
54f968d6 JW |
1607 | |
1608 | if (!tun) | |
1609 | return -EBADFD; | |
1610 | ||
eda29772 | 1611 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
1612 | ret = -EINVAL; |
1613 | goto out; | |
1614 | } | |
eda29772 RC |
1615 | if (flags & MSG_ERRQUEUE) { |
1616 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
1617 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
1618 | goto out; | |
1619 | } | |
c0371da6 | 1620 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT); |
87897931 | 1621 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
1622 | m->msg_flags |= MSG_TRUNC; |
1623 | ret = flags & MSG_TRUNC ? ret : total_len; | |
1624 | } | |
3811ae76 | 1625 | out: |
54f968d6 | 1626 | tun_put(tun); |
05c2828c MT |
1627 | return ret; |
1628 | } | |
1629 | ||
1630 | /* Ops structure to mimic raw sockets with tun */ | |
1631 | static const struct proto_ops tun_socket_ops = { | |
1632 | .sendmsg = tun_sendmsg, | |
1633 | .recvmsg = tun_recvmsg, | |
1634 | }; | |
1635 | ||
33dccbb0 HX |
1636 | static struct proto tun_proto = { |
1637 | .name = "tun", | |
1638 | .owner = THIS_MODULE, | |
54f968d6 | 1639 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 1640 | }; |
f019a7a5 | 1641 | |
980c9e8c DW |
1642 | static int tun_flags(struct tun_struct *tun) |
1643 | { | |
031f5e03 | 1644 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
1645 | } |
1646 | ||
1647 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
1648 | char *buf) | |
1649 | { | |
1650 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
1651 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
1652 | } | |
1653 | ||
1654 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
1655 | char *buf) | |
1656 | { | |
1657 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1658 | return uid_valid(tun->owner)? |
1659 | sprintf(buf, "%u\n", | |
1660 | from_kuid_munged(current_user_ns(), tun->owner)): | |
1661 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1662 | } |
1663 | ||
1664 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
1665 | char *buf) | |
1666 | { | |
1667 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1668 | return gid_valid(tun->group) ? |
1669 | sprintf(buf, "%u\n", | |
1670 | from_kgid_munged(current_user_ns(), tun->group)): | |
1671 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1672 | } |
1673 | ||
1674 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
1675 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
1676 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
1677 | ||
c4d33e24 TI |
1678 | static struct attribute *tun_dev_attrs[] = { |
1679 | &dev_attr_tun_flags.attr, | |
1680 | &dev_attr_owner.attr, | |
1681 | &dev_attr_group.attr, | |
1682 | NULL | |
1683 | }; | |
1684 | ||
1685 | static const struct attribute_group tun_attr_group = { | |
1686 | .attrs = tun_dev_attrs | |
1687 | }; | |
1688 | ||
d647a591 | 1689 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
1690 | { |
1691 | struct tun_struct *tun; | |
54f968d6 | 1692 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
1693 | struct net_device *dev; |
1694 | int err; | |
1695 | ||
7c0c3b1a JW |
1696 | if (tfile->detached) |
1697 | return -EINVAL; | |
1698 | ||
74a3e5a7 EB |
1699 | dev = __dev_get_by_name(net, ifr->ifr_name); |
1700 | if (dev) { | |
f85ba780 DW |
1701 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
1702 | return -EBUSY; | |
74a3e5a7 EB |
1703 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
1704 | tun = netdev_priv(dev); | |
1705 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
1706 | tun = netdev_priv(dev); | |
1707 | else | |
1708 | return -EINVAL; | |
1709 | ||
8e6d91ae | 1710 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 1711 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
1712 | return -EINVAL; |
1713 | ||
cde8b15f | 1714 | if (tun_not_capable(tun)) |
2b980dbd | 1715 | return -EPERM; |
5dbbaf2d | 1716 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
1717 | if (err < 0) |
1718 | return err; | |
1719 | ||
849c9b6f | 1720 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER); |
a7385ba2 EB |
1721 | if (err < 0) |
1722 | return err; | |
4008e97f | 1723 | |
40630b82 | 1724 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
1725 | (tun->numqueues + tun->numdisabled > 1)) { |
1726 | /* One or more queue has already been attached, no need | |
1727 | * to initialize the device again. | |
1728 | */ | |
1729 | return 0; | |
1730 | } | |
6aa20a22 | 1731 | } |
1da177e4 LT |
1732 | else { |
1733 | char *name; | |
1734 | unsigned long flags = 0; | |
edfb6a14 JW |
1735 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
1736 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 1737 | |
c260b772 | 1738 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 1739 | return -EPERM; |
2b980dbd PM |
1740 | err = security_tun_dev_create(); |
1741 | if (err < 0) | |
1742 | return err; | |
ca6bb5d7 | 1743 | |
1da177e4 LT |
1744 | /* Set dev type */ |
1745 | if (ifr->ifr_flags & IFF_TUN) { | |
1746 | /* TUN device */ | |
40630b82 | 1747 | flags |= IFF_TUN; |
1da177e4 LT |
1748 | name = "tun%d"; |
1749 | } else if (ifr->ifr_flags & IFF_TAP) { | |
1750 | /* TAP device */ | |
40630b82 | 1751 | flags |= IFF_TAP; |
1da177e4 | 1752 | name = "tap%d"; |
6aa20a22 | 1753 | } else |
36989b90 | 1754 | return -EINVAL; |
6aa20a22 | 1755 | |
1da177e4 LT |
1756 | if (*ifr->ifr_name) |
1757 | name = ifr->ifr_name; | |
1758 | ||
c8d68e6b | 1759 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
1760 | NET_NAME_UNKNOWN, tun_setup, queues, |
1761 | queues); | |
edfb6a14 | 1762 | |
1da177e4 LT |
1763 | if (!dev) |
1764 | return -ENOMEM; | |
1765 | ||
fc54c658 | 1766 | dev_net_set(dev, net); |
f019a7a5 | 1767 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 1768 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 1769 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 1770 | |
1da177e4 LT |
1771 | tun = netdev_priv(dev); |
1772 | tun->dev = dev; | |
1773 | tun->flags = flags; | |
f271b2cc | 1774 | tun->txflt.count = 0; |
d9d52b51 | 1775 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 1776 | |
eaea34b2 | 1777 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
1778 | tun->filter_attached = false; |
1779 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
33dccbb0 | 1780 | |
608b9977 PA |
1781 | tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); |
1782 | if (!tun->pcpu_stats) { | |
1783 | err = -ENOMEM; | |
1784 | goto err_free_dev; | |
1785 | } | |
1786 | ||
96442e42 JW |
1787 | spin_lock_init(&tun->lock); |
1788 | ||
5dbbaf2d PM |
1789 | err = security_tun_dev_alloc_security(&tun->security); |
1790 | if (err < 0) | |
608b9977 | 1791 | goto err_free_stat; |
2b980dbd | 1792 | |
1da177e4 | 1793 | tun_net_init(dev); |
944a1376 | 1794 | tun_flow_init(tun); |
96442e42 | 1795 | |
88255375 | 1796 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
1797 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
1798 | NETIF_F_HW_VLAN_STAG_TX; | |
88255375 | 1799 | dev->features = dev->hw_features; |
6671b224 FLVC |
1800 | dev->vlan_features = dev->features & |
1801 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
1802 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 1803 | |
4008e97f | 1804 | INIT_LIST_HEAD(&tun->disabled); |
849c9b6f | 1805 | err = tun_attach(tun, file, false); |
eb0fb363 | 1806 | if (err < 0) |
662ca437 | 1807 | goto err_free_flow; |
eb0fb363 | 1808 | |
1da177e4 LT |
1809 | err = register_netdevice(tun->dev); |
1810 | if (err < 0) | |
662ca437 | 1811 | goto err_detach; |
1da177e4 LT |
1812 | } |
1813 | ||
af668b3c MT |
1814 | netif_carrier_on(tun->dev); |
1815 | ||
6b8a66ee | 1816 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 1817 | |
031f5e03 MT |
1818 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
1819 | (ifr->ifr_flags & TUN_FEATURES); | |
c8d68e6b | 1820 | |
e35259a9 MK |
1821 | /* Make sure persistent devices do not get stuck in |
1822 | * xoff state. | |
1823 | */ | |
1824 | if (netif_running(tun->dev)) | |
c8d68e6b | 1825 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 1826 | |
1da177e4 LT |
1827 | strcpy(ifr->ifr_name, tun->dev->name); |
1828 | return 0; | |
1829 | ||
662ca437 JW |
1830 | err_detach: |
1831 | tun_detach_all(dev); | |
1832 | err_free_flow: | |
1833 | tun_flow_uninit(tun); | |
1834 | security_tun_dev_free_security(tun->security); | |
608b9977 PA |
1835 | err_free_stat: |
1836 | free_percpu(tun->pcpu_stats); | |
662ca437 | 1837 | err_free_dev: |
1da177e4 | 1838 | free_netdev(dev); |
1da177e4 LT |
1839 | return err; |
1840 | } | |
1841 | ||
9ce99cf6 | 1842 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 1843 | struct ifreq *ifr) |
e3b99556 | 1844 | { |
6b8a66ee | 1845 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
1846 | |
1847 | strcpy(ifr->ifr_name, tun->dev->name); | |
1848 | ||
980c9e8c | 1849 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 1850 | |
e3b99556 MM |
1851 | } |
1852 | ||
5228ddc9 RR |
1853 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
1854 | * privs required. */ | |
88255375 | 1855 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 1856 | { |
c8f44aff | 1857 | netdev_features_t features = 0; |
5228ddc9 RR |
1858 | |
1859 | if (arg & TUN_F_CSUM) { | |
88255375 | 1860 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
1861 | arg &= ~TUN_F_CSUM; |
1862 | ||
1863 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
1864 | if (arg & TUN_F_TSO_ECN) { | |
1865 | features |= NETIF_F_TSO_ECN; | |
1866 | arg &= ~TUN_F_TSO_ECN; | |
1867 | } | |
1868 | if (arg & TUN_F_TSO4) | |
1869 | features |= NETIF_F_TSO; | |
1870 | if (arg & TUN_F_TSO6) | |
1871 | features |= NETIF_F_TSO6; | |
1872 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
1873 | } | |
e3e3c423 VY |
1874 | |
1875 | if (arg & TUN_F_UFO) { | |
1876 | features |= NETIF_F_UFO; | |
1877 | arg &= ~TUN_F_UFO; | |
1878 | } | |
5228ddc9 RR |
1879 | } |
1880 | ||
1881 | /* This gives the user a way to test for new features in future by | |
1882 | * trying to set them. */ | |
1883 | if (arg) | |
1884 | return -EINVAL; | |
1885 | ||
88255375 MM |
1886 | tun->set_features = features; |
1887 | netdev_update_features(tun->dev); | |
5228ddc9 RR |
1888 | |
1889 | return 0; | |
1890 | } | |
1891 | ||
c8d68e6b JW |
1892 | static void tun_detach_filter(struct tun_struct *tun, int n) |
1893 | { | |
1894 | int i; | |
1895 | struct tun_file *tfile; | |
1896 | ||
1897 | for (i = 0; i < n; i++) { | |
b8deabd3 | 1898 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
1899 | lock_sock(tfile->socket.sk); |
1900 | sk_detach_filter(tfile->socket.sk); | |
1901 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
1902 | } |
1903 | ||
1904 | tun->filter_attached = false; | |
1905 | } | |
1906 | ||
1907 | static int tun_attach_filter(struct tun_struct *tun) | |
1908 | { | |
1909 | int i, ret = 0; | |
1910 | struct tun_file *tfile; | |
1911 | ||
1912 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1913 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
1914 | lock_sock(tfile->socket.sk); |
1915 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
1916 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
1917 | if (ret) { |
1918 | tun_detach_filter(tun, i); | |
1919 | return ret; | |
1920 | } | |
1921 | } | |
1922 | ||
1923 | tun->filter_attached = true; | |
1924 | return ret; | |
1925 | } | |
1926 | ||
1927 | static void tun_set_sndbuf(struct tun_struct *tun) | |
1928 | { | |
1929 | struct tun_file *tfile; | |
1930 | int i; | |
1931 | ||
1932 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1933 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1934 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
1935 | } | |
1936 | } | |
1937 | ||
cde8b15f JW |
1938 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
1939 | { | |
1940 | struct tun_file *tfile = file->private_data; | |
1941 | struct tun_struct *tun; | |
cde8b15f JW |
1942 | int ret = 0; |
1943 | ||
1944 | rtnl_lock(); | |
1945 | ||
1946 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 1947 | tun = tfile->detached; |
5dbbaf2d | 1948 | if (!tun) { |
cde8b15f | 1949 | ret = -EINVAL; |
5dbbaf2d PM |
1950 | goto unlock; |
1951 | } | |
1952 | ret = security_tun_dev_attach_queue(tun->security); | |
1953 | if (ret < 0) | |
1954 | goto unlock; | |
849c9b6f | 1955 | ret = tun_attach(tun, file, false); |
4008e97f | 1956 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 1957 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 1958 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
1959 | ret = -EINVAL; |
1960 | else | |
1961 | __tun_detach(tfile, false); | |
1962 | } else | |
cde8b15f JW |
1963 | ret = -EINVAL; |
1964 | ||
5dbbaf2d | 1965 | unlock: |
cde8b15f JW |
1966 | rtnl_unlock(); |
1967 | return ret; | |
1968 | } | |
1969 | ||
50857e2a AB |
1970 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
1971 | unsigned long arg, int ifreq_len) | |
1da177e4 | 1972 | { |
36b50bab | 1973 | struct tun_file *tfile = file->private_data; |
631ab46b | 1974 | struct tun_struct *tun; |
1da177e4 LT |
1975 | void __user* argp = (void __user*)arg; |
1976 | struct ifreq ifr; | |
0625c883 EB |
1977 | kuid_t owner; |
1978 | kgid_t group; | |
33dccbb0 | 1979 | int sndbuf; |
d9d52b51 | 1980 | int vnet_hdr_sz; |
fb7589a1 | 1981 | unsigned int ifindex; |
1cf8e410 | 1982 | int le; |
f271b2cc | 1983 | int ret; |
1da177e4 | 1984 | |
cde8b15f | 1985 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == 0x89) { |
50857e2a | 1986 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 1987 | return -EFAULT; |
8bbb1813 | 1988 | } else { |
a117dacd | 1989 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 1990 | } |
631ab46b EB |
1991 | if (cmd == TUNGETFEATURES) { |
1992 | /* Currently this just means: "what IFF flags are valid?". | |
1993 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
1994 | * TUNSETIFF. |
1995 | */ | |
1996 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 1997 | (unsigned int __user*)argp); |
cde8b15f JW |
1998 | } else if (cmd == TUNSETQUEUE) |
1999 | return tun_set_queue(file, &ifr); | |
631ab46b | 2000 | |
c8d68e6b | 2001 | ret = 0; |
876bfd4d HX |
2002 | rtnl_lock(); |
2003 | ||
36b50bab | 2004 | tun = __tun_get(tfile); |
1da177e4 | 2005 | if (cmd == TUNSETIFF && !tun) { |
1da177e4 LT |
2006 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
2007 | ||
140e807d | 2008 | ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr); |
1da177e4 | 2009 | |
876bfd4d HX |
2010 | if (ret) |
2011 | goto unlock; | |
1da177e4 | 2012 | |
50857e2a | 2013 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
2014 | ret = -EFAULT; |
2015 | goto unlock; | |
1da177e4 | 2016 | } |
fb7589a1 PE |
2017 | if (cmd == TUNSETIFINDEX) { |
2018 | ret = -EPERM; | |
2019 | if (tun) | |
2020 | goto unlock; | |
2021 | ||
2022 | ret = -EFAULT; | |
2023 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
2024 | goto unlock; | |
2025 | ||
2026 | ret = 0; | |
2027 | tfile->ifindex = ifindex; | |
2028 | goto unlock; | |
2029 | } | |
1da177e4 | 2030 | |
876bfd4d | 2031 | ret = -EBADFD; |
1da177e4 | 2032 | if (!tun) |
876bfd4d | 2033 | goto unlock; |
1da177e4 | 2034 | |
1e588338 | 2035 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 2036 | |
631ab46b | 2037 | ret = 0; |
1da177e4 | 2038 | switch (cmd) { |
e3b99556 | 2039 | case TUNGETIFF: |
9ce99cf6 | 2040 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 2041 | |
3d407a80 PE |
2042 | if (tfile->detached) |
2043 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
2044 | if (!tfile->socket.sk->sk_filter) |
2045 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 2046 | |
50857e2a | 2047 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 2048 | ret = -EFAULT; |
e3b99556 MM |
2049 | break; |
2050 | ||
1da177e4 LT |
2051 | case TUNSETNOCSUM: |
2052 | /* Disable/Enable checksum */ | |
1da177e4 | 2053 | |
88255375 MM |
2054 | /* [unimplemented] */ |
2055 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 2056 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
2057 | break; |
2058 | ||
2059 | case TUNSETPERSIST: | |
54f968d6 JW |
2060 | /* Disable/Enable persist mode. Keep an extra reference to the |
2061 | * module to prevent the module being unprobed. | |
2062 | */ | |
40630b82 MT |
2063 | if (arg && !(tun->flags & IFF_PERSIST)) { |
2064 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 2065 | __module_get(THIS_MODULE); |
dd38bd85 | 2066 | } |
40630b82 MT |
2067 | if (!arg && (tun->flags & IFF_PERSIST)) { |
2068 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 JW |
2069 | module_put(THIS_MODULE); |
2070 | } | |
1da177e4 | 2071 | |
6b8a66ee JP |
2072 | tun_debug(KERN_INFO, tun, "persist %s\n", |
2073 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
2074 | break; |
2075 | ||
2076 | case TUNSETOWNER: | |
2077 | /* Set owner of the device */ | |
0625c883 EB |
2078 | owner = make_kuid(current_user_ns(), arg); |
2079 | if (!uid_valid(owner)) { | |
2080 | ret = -EINVAL; | |
2081 | break; | |
2082 | } | |
2083 | tun->owner = owner; | |
1e588338 | 2084 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 2085 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
2086 | break; |
2087 | ||
8c644623 GG |
2088 | case TUNSETGROUP: |
2089 | /* Set group of the device */ | |
0625c883 EB |
2090 | group = make_kgid(current_user_ns(), arg); |
2091 | if (!gid_valid(group)) { | |
2092 | ret = -EINVAL; | |
2093 | break; | |
2094 | } | |
2095 | tun->group = group; | |
1e588338 | 2096 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 2097 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
2098 | break; |
2099 | ||
ff4cc3ac MK |
2100 | case TUNSETLINK: |
2101 | /* Only allow setting the type when the interface is down */ | |
2102 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
2103 | tun_debug(KERN_INFO, tun, |
2104 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 2105 | ret = -EBUSY; |
ff4cc3ac MK |
2106 | } else { |
2107 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2108 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2109 | tun->dev->type); | |
48abfe05 | 2110 | ret = 0; |
ff4cc3ac | 2111 | } |
631ab46b | 2112 | break; |
ff4cc3ac | 2113 | |
1da177e4 LT |
2114 | #ifdef TUN_DEBUG |
2115 | case TUNSETDEBUG: | |
2116 | tun->debug = arg; | |
2117 | break; | |
2118 | #endif | |
5228ddc9 | 2119 | case TUNSETOFFLOAD: |
88255375 | 2120 | ret = set_offload(tun, arg); |
631ab46b | 2121 | break; |
5228ddc9 | 2122 | |
f271b2cc MK |
2123 | case TUNSETTXFILTER: |
2124 | /* Can be set only for TAPs */ | |
631ab46b | 2125 | ret = -EINVAL; |
40630b82 | 2126 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 2127 | break; |
c0e5a8c2 | 2128 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2129 | break; |
1da177e4 LT |
2130 | |
2131 | case SIOCGIFHWADDR: | |
b595076a | 2132 | /* Get hw address */ |
f271b2cc MK |
2133 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2134 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2135 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2136 | ret = -EFAULT; |
2137 | break; | |
1da177e4 LT |
2138 | |
2139 | case SIOCSIFHWADDR: | |
f271b2cc | 2140 | /* Set hw address */ |
6b8a66ee JP |
2141 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2142 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2143 | |
40102371 | 2144 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2145 | break; |
33dccbb0 HX |
2146 | |
2147 | case TUNGETSNDBUF: | |
54f968d6 | 2148 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2149 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2150 | ret = -EFAULT; | |
2151 | break; | |
2152 | ||
2153 | case TUNSETSNDBUF: | |
2154 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2155 | ret = -EFAULT; | |
2156 | break; | |
2157 | } | |
2158 | ||
c8d68e6b JW |
2159 | tun->sndbuf = sndbuf; |
2160 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2161 | break; |
2162 | ||
d9d52b51 MT |
2163 | case TUNGETVNETHDRSZ: |
2164 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2165 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2166 | ret = -EFAULT; | |
2167 | break; | |
2168 | ||
2169 | case TUNSETVNETHDRSZ: | |
2170 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2171 | ret = -EFAULT; | |
2172 | break; | |
2173 | } | |
2174 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2175 | ret = -EINVAL; | |
2176 | break; | |
2177 | } | |
2178 | ||
2179 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2180 | break; | |
2181 | ||
1cf8e410 MT |
2182 | case TUNGETVNETLE: |
2183 | le = !!(tun->flags & TUN_VNET_LE); | |
2184 | if (put_user(le, (int __user *)argp)) | |
2185 | ret = -EFAULT; | |
2186 | break; | |
2187 | ||
2188 | case TUNSETVNETLE: | |
2189 | if (get_user(le, (int __user *)argp)) { | |
2190 | ret = -EFAULT; | |
2191 | break; | |
2192 | } | |
2193 | if (le) | |
2194 | tun->flags |= TUN_VNET_LE; | |
2195 | else | |
2196 | tun->flags &= ~TUN_VNET_LE; | |
2197 | break; | |
2198 | ||
8b8e658b GK |
2199 | case TUNGETVNETBE: |
2200 | ret = tun_get_vnet_be(tun, argp); | |
2201 | break; | |
2202 | ||
2203 | case TUNSETVNETBE: | |
2204 | ret = tun_set_vnet_be(tun, argp); | |
2205 | break; | |
2206 | ||
99405162 MT |
2207 | case TUNATTACHFILTER: |
2208 | /* Can be set only for TAPs */ | |
2209 | ret = -EINVAL; | |
40630b82 | 2210 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
2211 | break; |
2212 | ret = -EFAULT; | |
54f968d6 | 2213 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2214 | break; |
2215 | ||
c8d68e6b | 2216 | ret = tun_attach_filter(tun); |
99405162 MT |
2217 | break; |
2218 | ||
2219 | case TUNDETACHFILTER: | |
2220 | /* Can be set only for TAPs */ | |
2221 | ret = -EINVAL; | |
40630b82 | 2222 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 2223 | break; |
c8d68e6b JW |
2224 | ret = 0; |
2225 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2226 | break; |
2227 | ||
76975e9c PE |
2228 | case TUNGETFILTER: |
2229 | ret = -EINVAL; | |
40630b82 | 2230 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
2231 | break; |
2232 | ret = -EFAULT; | |
2233 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2234 | break; | |
2235 | ret = 0; | |
2236 | break; | |
2237 | ||
1da177e4 | 2238 | default: |
631ab46b EB |
2239 | ret = -EINVAL; |
2240 | break; | |
ee289b64 | 2241 | } |
1da177e4 | 2242 | |
876bfd4d HX |
2243 | unlock: |
2244 | rtnl_unlock(); | |
2245 | if (tun) | |
2246 | tun_put(tun); | |
631ab46b | 2247 | return ret; |
1da177e4 LT |
2248 | } |
2249 | ||
50857e2a AB |
2250 | static long tun_chr_ioctl(struct file *file, |
2251 | unsigned int cmd, unsigned long arg) | |
2252 | { | |
2253 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2254 | } | |
2255 | ||
2256 | #ifdef CONFIG_COMPAT | |
2257 | static long tun_chr_compat_ioctl(struct file *file, | |
2258 | unsigned int cmd, unsigned long arg) | |
2259 | { | |
2260 | switch (cmd) { | |
2261 | case TUNSETIFF: | |
2262 | case TUNGETIFF: | |
2263 | case TUNSETTXFILTER: | |
2264 | case TUNGETSNDBUF: | |
2265 | case TUNSETSNDBUF: | |
2266 | case SIOCGIFHWADDR: | |
2267 | case SIOCSIFHWADDR: | |
2268 | arg = (unsigned long)compat_ptr(arg); | |
2269 | break; | |
2270 | default: | |
2271 | arg = (compat_ulong_t)arg; | |
2272 | break; | |
2273 | } | |
2274 | ||
2275 | /* | |
2276 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2277 | * the end of that structure. All fields that are used in this | |
2278 | * driver are compatible though, we don't need to convert the | |
2279 | * contents. | |
2280 | */ | |
2281 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2282 | } | |
2283 | #endif /* CONFIG_COMPAT */ | |
2284 | ||
1da177e4 LT |
2285 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2286 | { | |
54f968d6 | 2287 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2288 | int ret; |
2289 | ||
54f968d6 | 2290 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2291 | goto out; |
6aa20a22 | 2292 | |
1da177e4 | 2293 | if (on) { |
e0b93edd | 2294 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2295 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2296 | } else |
54f968d6 | 2297 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2298 | ret = 0; |
2299 | out: | |
9d319522 | 2300 | return ret; |
1da177e4 LT |
2301 | } |
2302 | ||
2303 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2304 | { | |
140e807d | 2305 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 2306 | struct tun_file *tfile; |
deed49fb | 2307 | |
6b8a66ee | 2308 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2309 | |
140e807d | 2310 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 2311 | &tun_proto, 0); |
631ab46b EB |
2312 | if (!tfile) |
2313 | return -ENOMEM; | |
c956674b | 2314 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 2315 | tfile->flags = 0; |
fb7589a1 | 2316 | tfile->ifindex = 0; |
54f968d6 | 2317 | |
54f968d6 | 2318 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2319 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2320 | |
2321 | tfile->socket.file = file; | |
2322 | tfile->socket.ops = &tun_socket_ops; | |
2323 | ||
2324 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
2325 | |
2326 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2327 | tfile->sk.sk_sndbuf = INT_MAX; | |
2328 | ||
631ab46b | 2329 | file->private_data = tfile; |
4008e97f | 2330 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2331 | |
19a6afb2 JW |
2332 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2333 | ||
1da177e4 LT |
2334 | return 0; |
2335 | } | |
2336 | ||
2337 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2338 | { | |
631ab46b | 2339 | struct tun_file *tfile = file->private_data; |
1da177e4 | 2340 | |
c8d68e6b | 2341 | tun_detach(tfile, true); |
1da177e4 LT |
2342 | |
2343 | return 0; | |
2344 | } | |
2345 | ||
93e14b6d | 2346 | #ifdef CONFIG_PROC_FS |
a3816ab0 | 2347 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f) |
93e14b6d MY |
2348 | { |
2349 | struct tun_struct *tun; | |
2350 | struct ifreq ifr; | |
2351 | ||
2352 | memset(&ifr, 0, sizeof(ifr)); | |
2353 | ||
2354 | rtnl_lock(); | |
2355 | tun = tun_get(f); | |
2356 | if (tun) | |
2357 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2358 | rtnl_unlock(); | |
2359 | ||
2360 | if (tun) | |
2361 | tun_put(tun); | |
2362 | ||
a3816ab0 | 2363 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
2364 | } |
2365 | #endif | |
2366 | ||
d54b1fdb | 2367 | static const struct file_operations tun_fops = { |
6aa20a22 | 2368 | .owner = THIS_MODULE, |
1da177e4 | 2369 | .llseek = no_llseek, |
9b067034 | 2370 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 2371 | .write_iter = tun_chr_write_iter, |
1da177e4 | 2372 | .poll = tun_chr_poll, |
50857e2a AB |
2373 | .unlocked_ioctl = tun_chr_ioctl, |
2374 | #ifdef CONFIG_COMPAT | |
2375 | .compat_ioctl = tun_chr_compat_ioctl, | |
2376 | #endif | |
1da177e4 LT |
2377 | .open = tun_chr_open, |
2378 | .release = tun_chr_close, | |
93e14b6d MY |
2379 | .fasync = tun_chr_fasync, |
2380 | #ifdef CONFIG_PROC_FS | |
2381 | .show_fdinfo = tun_chr_show_fdinfo, | |
2382 | #endif | |
1da177e4 LT |
2383 | }; |
2384 | ||
2385 | static struct miscdevice tun_miscdev = { | |
2386 | .minor = TUN_MINOR, | |
2387 | .name = "tun", | |
e454cea2 | 2388 | .nodename = "net/tun", |
1da177e4 | 2389 | .fops = &tun_fops, |
1da177e4 LT |
2390 | }; |
2391 | ||
2392 | /* ethtool interface */ | |
2393 | ||
2394 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) | |
2395 | { | |
2396 | cmd->supported = 0; | |
2397 | cmd->advertising = 0; | |
70739497 | 2398 | ethtool_cmd_speed_set(cmd, SPEED_10); |
1da177e4 LT |
2399 | cmd->duplex = DUPLEX_FULL; |
2400 | cmd->port = PORT_TP; | |
2401 | cmd->phy_address = 0; | |
2402 | cmd->transceiver = XCVR_INTERNAL; | |
2403 | cmd->autoneg = AUTONEG_DISABLE; | |
2404 | cmd->maxtxpkt = 0; | |
2405 | cmd->maxrxpkt = 0; | |
2406 | return 0; | |
2407 | } | |
2408 | ||
2409 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2410 | { | |
2411 | struct tun_struct *tun = netdev_priv(dev); | |
2412 | ||
33a5ba14 RJ |
2413 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2414 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2415 | |
2416 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 2417 | case IFF_TUN: |
33a5ba14 | 2418 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 2419 | break; |
40630b82 | 2420 | case IFF_TAP: |
33a5ba14 | 2421 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2422 | break; |
2423 | } | |
2424 | } | |
2425 | ||
2426 | static u32 tun_get_msglevel(struct net_device *dev) | |
2427 | { | |
2428 | #ifdef TUN_DEBUG | |
2429 | struct tun_struct *tun = netdev_priv(dev); | |
2430 | return tun->debug; | |
2431 | #else | |
2432 | return -EOPNOTSUPP; | |
2433 | #endif | |
2434 | } | |
2435 | ||
2436 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2437 | { | |
2438 | #ifdef TUN_DEBUG | |
2439 | struct tun_struct *tun = netdev_priv(dev); | |
2440 | tun->debug = value; | |
2441 | #endif | |
2442 | } | |
2443 | ||
7282d491 | 2444 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2445 | .get_settings = tun_get_settings, |
2446 | .get_drvinfo = tun_get_drvinfo, | |
2447 | .get_msglevel = tun_get_msglevel, | |
2448 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2449 | .get_link = ethtool_op_get_link, |
eda29772 | 2450 | .get_ts_info = ethtool_op_get_ts_info, |
1da177e4 LT |
2451 | }; |
2452 | ||
79d17604 | 2453 | |
1da177e4 LT |
2454 | static int __init tun_init(void) |
2455 | { | |
2456 | int ret = 0; | |
2457 | ||
6b8a66ee JP |
2458 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
2459 | pr_info("%s\n", DRV_COPYRIGHT); | |
1da177e4 | 2460 | |
f019a7a5 | 2461 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 2462 | if (ret) { |
6b8a66ee | 2463 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 2464 | goto err_linkops; |
79d17604 PE |
2465 | } |
2466 | ||
1da177e4 | 2467 | ret = misc_register(&tun_miscdev); |
79d17604 | 2468 | if (ret) { |
6b8a66ee | 2469 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
2470 | goto err_misc; |
2471 | } | |
f019a7a5 | 2472 | return 0; |
79d17604 | 2473 | err_misc: |
f019a7a5 EB |
2474 | rtnl_link_unregister(&tun_link_ops); |
2475 | err_linkops: | |
1da177e4 LT |
2476 | return ret; |
2477 | } | |
2478 | ||
2479 | static void tun_cleanup(void) | |
2480 | { | |
6aa20a22 | 2481 | misc_deregister(&tun_miscdev); |
f019a7a5 | 2482 | rtnl_link_unregister(&tun_link_ops); |
1da177e4 LT |
2483 | } |
2484 | ||
05c2828c MT |
2485 | /* Get an underlying socket object from tun file. Returns error unless file is |
2486 | * attached to a device. The returned object works like a packet socket, it | |
2487 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
2488 | * holding a reference to the file for as long as the socket is in use. */ | |
2489 | struct socket *tun_get_socket(struct file *file) | |
2490 | { | |
6e914fc7 | 2491 | struct tun_file *tfile; |
05c2828c MT |
2492 | if (file->f_op != &tun_fops) |
2493 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
2494 | tfile = file->private_data; |
2495 | if (!tfile) | |
05c2828c | 2496 | return ERR_PTR(-EBADFD); |
54f968d6 | 2497 | return &tfile->socket; |
05c2828c MT |
2498 | } |
2499 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
2500 | ||
1da177e4 LT |
2501 | module_init(tun_init); |
2502 | module_exit(tun_cleanup); | |
2503 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
2504 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
2505 | MODULE_LICENSE("GPL"); | |
2506 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 2507 | MODULE_ALIAS("devname:net/tun"); |