]>
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> |
1576d986 | 74 | #include <linux/skb_array.h> |
1da177e4 | 75 | |
1da177e4 LT |
76 | #include <asm/uaccess.h> |
77 | ||
14daa021 RR |
78 | /* Uncomment to enable debugging */ |
79 | /* #define TUN_DEBUG 1 */ | |
80 | ||
1da177e4 LT |
81 | #ifdef TUN_DEBUG |
82 | static int debug; | |
14daa021 | 83 | |
6b8a66ee JP |
84 | #define tun_debug(level, tun, fmt, args...) \ |
85 | do { \ | |
86 | if (tun->debug) \ | |
87 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
88 | } while (0) | |
89 | #define DBG1(level, fmt, args...) \ | |
90 | do { \ | |
91 | if (debug == 2) \ | |
92 | printk(level fmt, ##args); \ | |
93 | } while (0) | |
14daa021 | 94 | #else |
6b8a66ee JP |
95 | #define tun_debug(level, tun, fmt, args...) \ |
96 | do { \ | |
97 | if (0) \ | |
98 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
99 | } while (0) | |
100 | #define DBG1(level, fmt, args...) \ | |
101 | do { \ | |
102 | if (0) \ | |
103 | printk(level fmt, ##args); \ | |
104 | } while (0) | |
14daa021 RR |
105 | #endif |
106 | ||
031f5e03 MT |
107 | /* TUN device flags */ |
108 | ||
109 | /* IFF_ATTACH_QUEUE is never stored in device flags, | |
110 | * overload it to mean fasync when stored there. | |
111 | */ | |
112 | #define TUN_FASYNC IFF_ATTACH_QUEUE | |
1cf8e410 MT |
113 | /* High bits in flags field are unused. */ |
114 | #define TUN_VNET_LE 0x80000000 | |
8b8e658b | 115 | #define TUN_VNET_BE 0x40000000 |
031f5e03 MT |
116 | |
117 | #define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \ | |
1cf8e410 | 118 | IFF_MULTI_QUEUE) |
0690899b MT |
119 | #define GOODCOPY_LEN 128 |
120 | ||
f271b2cc MK |
121 | #define FLT_EXACT_COUNT 8 |
122 | struct tap_filter { | |
123 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
124 | u32 mask[2]; /* Mask of the hashed addrs */ | |
125 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
126 | }; | |
127 | ||
baf71c5c PG |
128 | /* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal |
129 | * to max number of VCPUs in guest. */ | |
130 | #define MAX_TAP_QUEUES 256 | |
b8732fb7 | 131 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 132 | |
96442e42 JW |
133 | #define TUN_FLOW_EXPIRE (3 * HZ) |
134 | ||
608b9977 PA |
135 | struct tun_pcpu_stats { |
136 | u64 rx_packets; | |
137 | u64 rx_bytes; | |
138 | u64 tx_packets; | |
139 | u64 tx_bytes; | |
140 | struct u64_stats_sync syncp; | |
141 | u32 rx_dropped; | |
142 | u32 tx_dropped; | |
143 | u32 rx_frame_errors; | |
144 | }; | |
145 | ||
54f968d6 | 146 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 147 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
148 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
149 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 150 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 151 | * this). |
6e914fc7 JW |
152 | * |
153 | * RCU usage: | |
36fe8c09 | 154 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 155 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 156 | */ |
631ab46b | 157 | struct tun_file { |
54f968d6 JW |
158 | struct sock sk; |
159 | struct socket socket; | |
160 | struct socket_wq wq; | |
6e914fc7 | 161 | struct tun_struct __rcu *tun; |
54f968d6 JW |
162 | struct fasync_struct *fasync; |
163 | /* only used for fasnyc */ | |
164 | unsigned int flags; | |
fb7589a1 PE |
165 | union { |
166 | u16 queue_index; | |
167 | unsigned int ifindex; | |
168 | }; | |
4008e97f JW |
169 | struct list_head next; |
170 | struct tun_struct *detached; | |
1576d986 | 171 | struct skb_array tx_array; |
631ab46b EB |
172 | }; |
173 | ||
96442e42 JW |
174 | struct tun_flow_entry { |
175 | struct hlist_node hash_link; | |
176 | struct rcu_head rcu; | |
177 | struct tun_struct *tun; | |
178 | ||
179 | u32 rxhash; | |
9bc88939 | 180 | u32 rps_rxhash; |
96442e42 JW |
181 | int queue_index; |
182 | unsigned long updated; | |
183 | }; | |
184 | ||
185 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
186 | ||
54f968d6 | 187 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 188 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
189 | * file were attached to a persist device. |
190 | */ | |
14daa021 | 191 | struct tun_struct { |
c8d68e6b JW |
192 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
193 | unsigned int numqueues; | |
f271b2cc | 194 | unsigned int flags; |
0625c883 EB |
195 | kuid_t owner; |
196 | kgid_t group; | |
14daa021 | 197 | |
14daa021 | 198 | struct net_device *dev; |
c8f44aff | 199 | netdev_features_t set_features; |
88255375 | 200 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
e3e3c423 | 201 | NETIF_F_TSO6|NETIF_F_UFO) |
d9d52b51 | 202 | |
eaea34b2 | 203 | int align; |
d9d52b51 | 204 | int vnet_hdr_sz; |
54f968d6 JW |
205 | int sndbuf; |
206 | struct tap_filter txflt; | |
207 | struct sock_fprog fprog; | |
208 | /* protected by rtnl lock */ | |
209 | bool filter_attached; | |
14daa021 RR |
210 | #ifdef TUN_DEBUG |
211 | int debug; | |
1da177e4 | 212 | #endif |
96442e42 | 213 | spinlock_t lock; |
96442e42 JW |
214 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
215 | struct timer_list flow_gc_timer; | |
216 | unsigned long ageing_time; | |
4008e97f JW |
217 | unsigned int numdisabled; |
218 | struct list_head disabled; | |
5dbbaf2d | 219 | void *security; |
b8732fb7 | 220 | u32 flow_count; |
608b9977 | 221 | struct tun_pcpu_stats __percpu *pcpu_stats; |
14daa021 | 222 | }; |
1da177e4 | 223 | |
8b8e658b GK |
224 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
225 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
226 | { | |
227 | return tun->flags & TUN_VNET_BE ? false : | |
228 | virtio_legacy_is_little_endian(); | |
229 | } | |
230 | ||
231 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
232 | { | |
233 | int be = !!(tun->flags & TUN_VNET_BE); | |
234 | ||
235 | if (put_user(be, argp)) | |
236 | return -EFAULT; | |
237 | ||
238 | return 0; | |
239 | } | |
240 | ||
241 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
242 | { | |
243 | int be; | |
244 | ||
245 | if (get_user(be, argp)) | |
246 | return -EFAULT; | |
247 | ||
248 | if (be) | |
249 | tun->flags |= TUN_VNET_BE; | |
250 | else | |
251 | tun->flags &= ~TUN_VNET_BE; | |
252 | ||
253 | return 0; | |
254 | } | |
255 | #else | |
256 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
257 | { | |
258 | return virtio_legacy_is_little_endian(); | |
259 | } | |
260 | ||
261 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
262 | { | |
263 | return -EINVAL; | |
264 | } | |
265 | ||
266 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
267 | { | |
268 | return -EINVAL; | |
269 | } | |
270 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
271 | ||
25bd55bb GK |
272 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
273 | { | |
7d824109 | 274 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 275 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
276 | } |
277 | ||
56f0dcc5 MT |
278 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
279 | { | |
25bd55bb | 280 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
281 | } |
282 | ||
283 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
284 | { | |
25bd55bb | 285 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
286 | } |
287 | ||
96442e42 JW |
288 | static inline u32 tun_hashfn(u32 rxhash) |
289 | { | |
290 | return rxhash & 0x3ff; | |
291 | } | |
292 | ||
293 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
294 | { | |
295 | struct tun_flow_entry *e; | |
96442e42 | 296 | |
b67bfe0d | 297 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
298 | if (e->rxhash == rxhash) |
299 | return e; | |
300 | } | |
301 | return NULL; | |
302 | } | |
303 | ||
304 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
305 | struct hlist_head *head, | |
306 | u32 rxhash, u16 queue_index) | |
307 | { | |
9fdc6bef ED |
308 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
309 | ||
96442e42 JW |
310 | if (e) { |
311 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
312 | rxhash, queue_index); | |
313 | e->updated = jiffies; | |
314 | e->rxhash = rxhash; | |
9bc88939 | 315 | e->rps_rxhash = 0; |
96442e42 JW |
316 | e->queue_index = queue_index; |
317 | e->tun = tun; | |
318 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 319 | ++tun->flow_count; |
96442e42 JW |
320 | } |
321 | return e; | |
322 | } | |
323 | ||
96442e42 JW |
324 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
325 | { | |
326 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
327 | e->rxhash, e->queue_index); | |
328 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 329 | kfree_rcu(e, rcu); |
b8732fb7 | 330 | --tun->flow_count; |
96442e42 JW |
331 | } |
332 | ||
333 | static void tun_flow_flush(struct tun_struct *tun) | |
334 | { | |
335 | int i; | |
336 | ||
337 | spin_lock_bh(&tun->lock); | |
338 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
339 | struct tun_flow_entry *e; | |
b67bfe0d | 340 | struct hlist_node *n; |
96442e42 | 341 | |
b67bfe0d | 342 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
343 | tun_flow_delete(tun, e); |
344 | } | |
345 | spin_unlock_bh(&tun->lock); | |
346 | } | |
347 | ||
348 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
349 | { | |
350 | int i; | |
351 | ||
352 | spin_lock_bh(&tun->lock); | |
353 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
354 | struct tun_flow_entry *e; | |
b67bfe0d | 355 | struct hlist_node *n; |
96442e42 | 356 | |
b67bfe0d | 357 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
358 | if (e->queue_index == queue_index) |
359 | tun_flow_delete(tun, e); | |
360 | } | |
361 | } | |
362 | spin_unlock_bh(&tun->lock); | |
363 | } | |
364 | ||
365 | static void tun_flow_cleanup(unsigned long data) | |
366 | { | |
367 | struct tun_struct *tun = (struct tun_struct *)data; | |
368 | unsigned long delay = tun->ageing_time; | |
369 | unsigned long next_timer = jiffies + delay; | |
370 | unsigned long count = 0; | |
371 | int i; | |
372 | ||
373 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
374 | ||
375 | spin_lock_bh(&tun->lock); | |
376 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
377 | struct tun_flow_entry *e; | |
b67bfe0d | 378 | struct hlist_node *n; |
96442e42 | 379 | |
b67bfe0d | 380 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
381 | unsigned long this_timer; |
382 | count++; | |
383 | this_timer = e->updated + delay; | |
384 | if (time_before_eq(this_timer, jiffies)) | |
385 | tun_flow_delete(tun, e); | |
386 | else if (time_before(this_timer, next_timer)) | |
387 | next_timer = this_timer; | |
388 | } | |
389 | } | |
390 | ||
391 | if (count) | |
392 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
393 | spin_unlock_bh(&tun->lock); | |
394 | } | |
395 | ||
49974420 | 396 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 397 | struct tun_file *tfile) |
96442e42 JW |
398 | { |
399 | struct hlist_head *head; | |
400 | struct tun_flow_entry *e; | |
401 | unsigned long delay = tun->ageing_time; | |
9e85722d | 402 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
403 | |
404 | if (!rxhash) | |
405 | return; | |
406 | else | |
407 | head = &tun->flows[tun_hashfn(rxhash)]; | |
408 | ||
409 | rcu_read_lock(); | |
410 | ||
9e85722d JW |
411 | /* We may get a very small possibility of OOO during switching, not |
412 | * worth to optimize.*/ | |
413 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
414 | goto unlock; |
415 | ||
416 | e = tun_flow_find(head, rxhash); | |
417 | if (likely(e)) { | |
418 | /* TODO: keep queueing to old queue until it's empty? */ | |
419 | e->queue_index = queue_index; | |
420 | e->updated = jiffies; | |
9bc88939 | 421 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
422 | } else { |
423 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
424 | if (!tun_flow_find(head, rxhash) && |
425 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
426 | tun_flow_create(tun, head, rxhash, queue_index); |
427 | ||
428 | if (!timer_pending(&tun->flow_gc_timer)) | |
429 | mod_timer(&tun->flow_gc_timer, | |
430 | round_jiffies_up(jiffies + delay)); | |
431 | spin_unlock_bh(&tun->lock); | |
432 | } | |
433 | ||
434 | unlock: | |
435 | rcu_read_unlock(); | |
436 | } | |
437 | ||
9bc88939 TH |
438 | /** |
439 | * Save the hash received in the stack receive path and update the | |
440 | * flow_hash table accordingly. | |
441 | */ | |
442 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
443 | { | |
567e4b79 | 444 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 445 | e->rps_rxhash = hash; |
9bc88939 TH |
446 | } |
447 | ||
c8d68e6b | 448 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 449 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
450 | * the rxq based on the txq where the last packet of the flow comes. As |
451 | * the userspace application move between processors, we may get a | |
452 | * different rxq no. here. If we could not get rxhash, then we would | |
453 | * hope the rxq no. may help here. | |
454 | */ | |
f663dd9a | 455 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 456 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
457 | { |
458 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 459 | struct tun_flow_entry *e; |
c8d68e6b JW |
460 | u32 txq = 0; |
461 | u32 numqueues = 0; | |
462 | ||
463 | rcu_read_lock(); | |
92bb73ea | 464 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 465 | |
3958afa1 | 466 | txq = skb_get_hash(skb); |
c8d68e6b | 467 | if (txq) { |
96442e42 | 468 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 469 | if (e) { |
9bc88939 | 470 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 471 | txq = e->queue_index; |
9bc88939 | 472 | } else |
96442e42 JW |
473 | /* use multiply and shift instead of expensive divide */ |
474 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
475 | } else if (likely(skb_rx_queue_recorded(skb))) { |
476 | txq = skb_get_rx_queue(skb); | |
477 | while (unlikely(txq >= numqueues)) | |
478 | txq -= numqueues; | |
479 | } | |
480 | ||
481 | rcu_read_unlock(); | |
482 | return txq; | |
483 | } | |
484 | ||
cde8b15f JW |
485 | static inline bool tun_not_capable(struct tun_struct *tun) |
486 | { | |
487 | const struct cred *cred = current_cred(); | |
c260b772 | 488 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
489 | |
490 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
491 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 492 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
493 | } |
494 | ||
c8d68e6b JW |
495 | static void tun_set_real_num_queues(struct tun_struct *tun) |
496 | { | |
497 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
498 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
499 | } | |
500 | ||
4008e97f JW |
501 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
502 | { | |
503 | tfile->detached = tun; | |
504 | list_add_tail(&tfile->next, &tun->disabled); | |
505 | ++tun->numdisabled; | |
506 | } | |
507 | ||
d32649d1 | 508 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
509 | { |
510 | struct tun_struct *tun = tfile->detached; | |
511 | ||
512 | tfile->detached = NULL; | |
513 | list_del_init(&tfile->next); | |
514 | --tun->numdisabled; | |
515 | return tun; | |
516 | } | |
517 | ||
4bfb0513 JW |
518 | static void tun_queue_purge(struct tun_file *tfile) |
519 | { | |
1576d986 JW |
520 | struct sk_buff *skb; |
521 | ||
522 | while ((skb = skb_array_consume(&tfile->tx_array)) != NULL) | |
523 | kfree_skb(skb); | |
524 | ||
4bfb0513 JW |
525 | skb_queue_purge(&tfile->sk.sk_error_queue); |
526 | } | |
527 | ||
c8d68e6b JW |
528 | static void __tun_detach(struct tun_file *tfile, bool clean) |
529 | { | |
530 | struct tun_file *ntfile; | |
531 | struct tun_struct *tun; | |
c8d68e6b | 532 | |
b8deabd3 JW |
533 | tun = rtnl_dereference(tfile->tun); |
534 | ||
9e85722d | 535 | if (tun && !tfile->detached) { |
c8d68e6b JW |
536 | u16 index = tfile->queue_index; |
537 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
538 | |
539 | rcu_assign_pointer(tun->tfiles[index], | |
540 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 541 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
542 | ntfile->queue_index = index; |
543 | ||
544 | --tun->numqueues; | |
9e85722d | 545 | if (clean) { |
c956674b | 546 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 547 | sock_put(&tfile->sk); |
9e85722d | 548 | } else |
4008e97f | 549 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
550 | |
551 | synchronize_net(); | |
96442e42 | 552 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 553 | /* Drop read queue */ |
4bfb0513 | 554 | tun_queue_purge(tfile); |
c8d68e6b | 555 | tun_set_real_num_queues(tun); |
dd38bd85 | 556 | } else if (tfile->detached && clean) { |
4008e97f | 557 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
558 | sock_put(&tfile->sk); |
559 | } | |
c8d68e6b JW |
560 | |
561 | if (clean) { | |
af668b3c MT |
562 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
563 | netif_carrier_off(tun->dev); | |
564 | ||
40630b82 | 565 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 566 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 567 | unregister_netdevice(tun->dev); |
af668b3c | 568 | } |
1576d986 JW |
569 | if (tun) |
570 | skb_array_cleanup(&tfile->tx_array); | |
140e807d | 571 | sock_put(&tfile->sk); |
c8d68e6b JW |
572 | } |
573 | } | |
574 | ||
575 | static void tun_detach(struct tun_file *tfile, bool clean) | |
576 | { | |
577 | rtnl_lock(); | |
578 | __tun_detach(tfile, clean); | |
579 | rtnl_unlock(); | |
580 | } | |
581 | ||
582 | static void tun_detach_all(struct net_device *dev) | |
583 | { | |
584 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 585 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
586 | int i, n = tun->numqueues; |
587 | ||
588 | for (i = 0; i < n; i++) { | |
b8deabd3 | 589 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 590 | BUG_ON(!tfile); |
addf8fc4 | 591 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 592 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 593 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
594 | --tun->numqueues; |
595 | } | |
9e85722d | 596 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 597 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 598 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 599 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 600 | } |
c8d68e6b JW |
601 | BUG_ON(tun->numqueues != 0); |
602 | ||
603 | synchronize_net(); | |
604 | for (i = 0; i < n; i++) { | |
b8deabd3 | 605 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 606 | /* Drop read queue */ |
4bfb0513 | 607 | tun_queue_purge(tfile); |
c8d68e6b JW |
608 | sock_put(&tfile->sk); |
609 | } | |
4008e97f JW |
610 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
611 | tun_enable_queue(tfile); | |
4bfb0513 | 612 | tun_queue_purge(tfile); |
4008e97f JW |
613 | sock_put(&tfile->sk); |
614 | } | |
615 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 616 | |
40630b82 | 617 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 618 | module_put(THIS_MODULE); |
c8d68e6b JW |
619 | } |
620 | ||
849c9b6f | 621 | static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter) |
a7385ba2 | 622 | { |
631ab46b | 623 | struct tun_file *tfile = file->private_data; |
1576d986 | 624 | struct net_device *dev = tun->dev; |
38231b7a | 625 | int err; |
a7385ba2 | 626 | |
5dbbaf2d PM |
627 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
628 | if (err < 0) | |
629 | goto out; | |
630 | ||
38231b7a | 631 | err = -EINVAL; |
9e85722d | 632 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
633 | goto out; |
634 | ||
635 | err = -EBUSY; | |
40630b82 | 636 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
637 | goto out; |
638 | ||
639 | err = -E2BIG; | |
4008e97f JW |
640 | if (!tfile->detached && |
641 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
642 | goto out; |
643 | ||
644 | err = 0; | |
54f968d6 | 645 | |
92d4ea6e | 646 | /* Re-attach the filter to persist device */ |
849c9b6f | 647 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
648 | lock_sock(tfile->socket.sk); |
649 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
650 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
651 | if (!err) |
652 | goto out; | |
653 | } | |
1576d986 JW |
654 | |
655 | if (!tfile->detached && | |
656 | skb_array_init(&tfile->tx_array, dev->tx_queue_len, GFP_KERNEL)) { | |
657 | err = -ENOMEM; | |
658 | goto out; | |
659 | } | |
660 | ||
c8d68e6b | 661 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 662 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
6e914fc7 | 663 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 664 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 665 | tun->numqueues++; |
a7385ba2 | 666 | |
4008e97f JW |
667 | if (tfile->detached) |
668 | tun_enable_queue(tfile); | |
669 | else | |
670 | sock_hold(&tfile->sk); | |
671 | ||
c8d68e6b | 672 | tun_set_real_num_queues(tun); |
a7385ba2 | 673 | |
c8d68e6b JW |
674 | /* device is allowed to go away first, so no need to hold extra |
675 | * refcnt. | |
676 | */ | |
677 | ||
678 | out: | |
679 | return err; | |
631ab46b EB |
680 | } |
681 | ||
682 | static struct tun_struct *__tun_get(struct tun_file *tfile) | |
683 | { | |
6e914fc7 | 684 | struct tun_struct *tun; |
c70f1829 | 685 | |
6e914fc7 JW |
686 | rcu_read_lock(); |
687 | tun = rcu_dereference(tfile->tun); | |
688 | if (tun) | |
689 | dev_hold(tun->dev); | |
690 | rcu_read_unlock(); | |
c70f1829 EB |
691 | |
692 | return tun; | |
631ab46b EB |
693 | } |
694 | ||
695 | static struct tun_struct *tun_get(struct file *file) | |
696 | { | |
697 | return __tun_get(file->private_data); | |
698 | } | |
699 | ||
700 | static void tun_put(struct tun_struct *tun) | |
701 | { | |
6e914fc7 | 702 | dev_put(tun->dev); |
631ab46b EB |
703 | } |
704 | ||
6b8a66ee | 705 | /* TAP filtering */ |
f271b2cc MK |
706 | static void addr_hash_set(u32 *mask, const u8 *addr) |
707 | { | |
708 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
709 | mask[n >> 5] |= (1 << (n & 31)); | |
710 | } | |
711 | ||
712 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
713 | { | |
714 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
715 | return mask[n >> 5] & (1 << (n & 31)); | |
716 | } | |
717 | ||
718 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
719 | { | |
720 | struct { u8 u[ETH_ALEN]; } *addr; | |
721 | struct tun_filter uf; | |
722 | int err, alen, n, nexact; | |
723 | ||
724 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
725 | return -EFAULT; | |
726 | ||
727 | if (!uf.count) { | |
728 | /* Disabled */ | |
729 | filter->count = 0; | |
730 | return 0; | |
731 | } | |
732 | ||
733 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
734 | addr = memdup_user(arg + sizeof(uf), alen); |
735 | if (IS_ERR(addr)) | |
736 | return PTR_ERR(addr); | |
f271b2cc MK |
737 | |
738 | /* The filter is updated without holding any locks. Which is | |
739 | * perfectly safe. We disable it first and in the worst | |
740 | * case we'll accept a few undesired packets. */ | |
741 | filter->count = 0; | |
742 | wmb(); | |
743 | ||
744 | /* Use first set of addresses as an exact filter */ | |
745 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
746 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
747 | ||
748 | nexact = n; | |
749 | ||
cfbf84fc AW |
750 | /* Remaining multicast addresses are hashed, |
751 | * unicast will leave the filter disabled. */ | |
f271b2cc | 752 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
753 | for (; n < uf.count; n++) { |
754 | if (!is_multicast_ether_addr(addr[n].u)) { | |
755 | err = 0; /* no filter */ | |
3b8d2a69 | 756 | goto free_addr; |
cfbf84fc | 757 | } |
f271b2cc | 758 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 759 | } |
f271b2cc MK |
760 | |
761 | /* For ALLMULTI just set the mask to all ones. | |
762 | * This overrides the mask populated above. */ | |
763 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
764 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
765 | ||
766 | /* Now enable the filter */ | |
767 | wmb(); | |
768 | filter->count = nexact; | |
769 | ||
770 | /* Return the number of exact filters */ | |
771 | err = nexact; | |
3b8d2a69 | 772 | free_addr: |
f271b2cc MK |
773 | kfree(addr); |
774 | return err; | |
775 | } | |
776 | ||
777 | /* Returns: 0 - drop, !=0 - accept */ | |
778 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
779 | { | |
780 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
781 | * at this point. */ | |
782 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
783 | int i; | |
784 | ||
785 | /* Exact match */ | |
786 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 787 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
788 | return 1; |
789 | ||
790 | /* Inexact match (multicast only) */ | |
791 | if (is_multicast_ether_addr(eh->h_dest)) | |
792 | return addr_hash_test(filter->mask, eh->h_dest); | |
793 | ||
794 | return 0; | |
795 | } | |
796 | ||
797 | /* | |
798 | * Checks whether the packet is accepted or not. | |
799 | * Returns: 0 - drop, !=0 - accept | |
800 | */ | |
801 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
802 | { | |
803 | if (!filter->count) | |
804 | return 1; | |
805 | ||
806 | return run_filter(filter, skb); | |
807 | } | |
808 | ||
1da177e4 LT |
809 | /* Network device part of the driver */ |
810 | ||
7282d491 | 811 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 812 | |
c70f1829 EB |
813 | /* Net device detach from fd. */ |
814 | static void tun_net_uninit(struct net_device *dev) | |
815 | { | |
c8d68e6b | 816 | tun_detach_all(dev); |
c70f1829 EB |
817 | } |
818 | ||
1da177e4 LT |
819 | /* Net device open. */ |
820 | static int tun_net_open(struct net_device *dev) | |
821 | { | |
c8d68e6b | 822 | netif_tx_start_all_queues(dev); |
1da177e4 LT |
823 | return 0; |
824 | } | |
825 | ||
826 | /* Net device close. */ | |
827 | static int tun_net_close(struct net_device *dev) | |
828 | { | |
c8d68e6b | 829 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
830 | return 0; |
831 | } | |
832 | ||
833 | /* Net device start xmit */ | |
424efe9c | 834 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
835 | { |
836 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 837 | int txq = skb->queue_mapping; |
6e914fc7 | 838 | struct tun_file *tfile; |
fa35864e | 839 | u32 numqueues = 0; |
1da177e4 | 840 | |
6e914fc7 | 841 | rcu_read_lock(); |
c8d68e6b | 842 | tfile = rcu_dereference(tun->tfiles[txq]); |
fa35864e | 843 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 844 | |
1da177e4 | 845 | /* Drop packet if interface is not attached */ |
fa35864e | 846 | if (txq >= numqueues) |
1da177e4 LT |
847 | goto drop; |
848 | ||
3df97ba8 JW |
849 | #ifdef CONFIG_RPS |
850 | if (numqueues == 1 && static_key_false(&rps_needed)) { | |
9bc88939 TH |
851 | /* Select queue was not called for the skbuff, so we extract the |
852 | * RPS hash and save it into the flow_table here. | |
853 | */ | |
854 | __u32 rxhash; | |
855 | ||
856 | rxhash = skb_get_hash(skb); | |
857 | if (rxhash) { | |
858 | struct tun_flow_entry *e; | |
859 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
860 | rxhash); | |
861 | if (e) | |
862 | tun_flow_save_rps_rxhash(e, rxhash); | |
863 | } | |
864 | } | |
3df97ba8 | 865 | #endif |
9bc88939 | 866 | |
6e914fc7 JW |
867 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
868 | ||
c8d68e6b JW |
869 | BUG_ON(!tfile); |
870 | ||
f271b2cc MK |
871 | /* Drop if the filter does not like it. |
872 | * This is a noop if the filter is disabled. | |
873 | * Filter can be enabled only for the TAP devices. */ | |
874 | if (!check_filter(&tun->txflt, skb)) | |
875 | goto drop; | |
876 | ||
54f968d6 JW |
877 | if (tfile->socket.sk->sk_filter && |
878 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
879 | goto drop; |
880 | ||
36fe8c09 | 881 | /* Limit the number of packets queued by dividing txq length with the |
c8d68e6b JW |
882 | * number of queues. |
883 | */ | |
fa35864e DC |
884 | if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues |
885 | >= dev->tx_queue_len) | |
5d097109 | 886 | goto drop; |
1da177e4 | 887 | |
7bf66305 JW |
888 | if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC))) |
889 | goto drop; | |
890 | ||
7b996243 | 891 | skb_tx_timestamp(skb); |
eda29772 | 892 | |
0110d6f2 | 893 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
894 | * for indefinite time. |
895 | */ | |
0110d6f2 MT |
896 | skb_orphan(skb); |
897 | ||
f8af75f3 ED |
898 | nf_reset(skb); |
899 | ||
1576d986 JW |
900 | if (skb_array_produce(&tfile->tx_array, skb)) |
901 | goto drop; | |
1da177e4 LT |
902 | |
903 | /* Notify and wake up reader process */ | |
54f968d6 JW |
904 | if (tfile->flags & TUN_FASYNC) |
905 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 906 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
907 | |
908 | rcu_read_unlock(); | |
6ed10654 | 909 | return NETDEV_TX_OK; |
1da177e4 LT |
910 | |
911 | drop: | |
608b9977 | 912 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
149d36f7 | 913 | skb_tx_error(skb); |
1da177e4 | 914 | kfree_skb(skb); |
6e914fc7 | 915 | rcu_read_unlock(); |
baeababb | 916 | return NET_XMIT_DROP; |
1da177e4 LT |
917 | } |
918 | ||
f271b2cc | 919 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 920 | { |
f271b2cc MK |
921 | /* |
922 | * This callback is supposed to deal with mc filter in | |
923 | * _rx_ path and has nothing to do with the _tx_ path. | |
924 | * In rx path we always accept everything userspace gives us. | |
925 | */ | |
1da177e4 LT |
926 | } |
927 | ||
c8f44aff MM |
928 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
929 | netdev_features_t features) | |
88255375 MM |
930 | { |
931 | struct tun_struct *tun = netdev_priv(dev); | |
932 | ||
933 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
934 | } | |
bebd097a NH |
935 | #ifdef CONFIG_NET_POLL_CONTROLLER |
936 | static void tun_poll_controller(struct net_device *dev) | |
937 | { | |
938 | /* | |
939 | * Tun only receives frames when: | |
940 | * 1) the char device endpoint gets data from user space | |
941 | * 2) the tun socket gets a sendmsg call from user space | |
92d4ea6e | 942 | * Since both of those are synchronous operations, we are guaranteed |
bebd097a | 943 | * never to have pending data when we poll for it |
92d4ea6e | 944 | * so there is nothing to do here but return. |
bebd097a NH |
945 | * We need this though so netpoll recognizes us as an interface that |
946 | * supports polling, which enables bridge devices in virt setups to | |
947 | * still use netconsole | |
948 | */ | |
949 | return; | |
950 | } | |
951 | #endif | |
eaea34b2 PA |
952 | |
953 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
954 | { | |
955 | struct tun_struct *tun = netdev_priv(dev); | |
956 | ||
957 | if (new_hr < NET_SKB_PAD) | |
958 | new_hr = NET_SKB_PAD; | |
959 | ||
960 | tun->align = new_hr; | |
961 | } | |
962 | ||
608b9977 PA |
963 | static struct rtnl_link_stats64 * |
964 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) | |
965 | { | |
966 | u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; | |
967 | struct tun_struct *tun = netdev_priv(dev); | |
968 | struct tun_pcpu_stats *p; | |
969 | int i; | |
970 | ||
971 | for_each_possible_cpu(i) { | |
972 | u64 rxpackets, rxbytes, txpackets, txbytes; | |
973 | unsigned int start; | |
974 | ||
975 | p = per_cpu_ptr(tun->pcpu_stats, i); | |
976 | do { | |
977 | start = u64_stats_fetch_begin(&p->syncp); | |
978 | rxpackets = p->rx_packets; | |
979 | rxbytes = p->rx_bytes; | |
980 | txpackets = p->tx_packets; | |
981 | txbytes = p->tx_bytes; | |
982 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
983 | ||
984 | stats->rx_packets += rxpackets; | |
985 | stats->rx_bytes += rxbytes; | |
986 | stats->tx_packets += txpackets; | |
987 | stats->tx_bytes += txbytes; | |
988 | ||
989 | /* u32 counters */ | |
990 | rx_dropped += p->rx_dropped; | |
991 | rx_frame_errors += p->rx_frame_errors; | |
992 | tx_dropped += p->tx_dropped; | |
993 | } | |
994 | stats->rx_dropped = rx_dropped; | |
995 | stats->rx_frame_errors = rx_frame_errors; | |
996 | stats->tx_dropped = tx_dropped; | |
997 | return stats; | |
998 | } | |
999 | ||
758e43b7 | 1000 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 1001 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1002 | .ndo_open = tun_net_open, |
1003 | .ndo_stop = tun_net_close, | |
00829823 | 1004 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1005 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1006 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1007 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1008 | .ndo_poll_controller = tun_poll_controller, | |
1009 | #endif | |
eaea34b2 | 1010 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1011 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1012 | }; |
1013 | ||
1014 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 1015 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1016 | .ndo_open = tun_net_open, |
1017 | .ndo_stop = tun_net_close, | |
00829823 | 1018 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1019 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1020 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1021 | .ndo_set_mac_address = eth_mac_addr, |
1022 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1023 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1024 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1025 | .ndo_poll_controller = tun_poll_controller, | |
1026 | #endif | |
5e52796a | 1027 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1028 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1029 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1030 | }; |
1031 | ||
944a1376 | 1032 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1033 | { |
1034 | int i; | |
1035 | ||
96442e42 JW |
1036 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1037 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1038 | ||
1039 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
1040 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
1041 | mod_timer(&tun->flow_gc_timer, | |
1042 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1043 | } |
1044 | ||
1045 | static void tun_flow_uninit(struct tun_struct *tun) | |
1046 | { | |
1047 | del_timer_sync(&tun->flow_gc_timer); | |
1048 | tun_flow_flush(tun); | |
96442e42 JW |
1049 | } |
1050 | ||
91572088 JW |
1051 | #define MIN_MTU 68 |
1052 | #define MAX_MTU 65535 | |
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 | } | |
91572088 JW |
1084 | |
1085 | dev->min_mtu = MIN_MTU; | |
1086 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1087 | } |
1088 | ||
1089 | /* Character device part */ | |
1090 | ||
1091 | /* Poll */ | |
c8d68e6b | 1092 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1093 | { |
b2430de3 EB |
1094 | struct tun_file *tfile = file->private_data; |
1095 | struct tun_struct *tun = __tun_get(tfile); | |
3c8a9c63 | 1096 | struct sock *sk; |
33dccbb0 | 1097 | unsigned int mask = 0; |
1da177e4 LT |
1098 | |
1099 | if (!tun) | |
eac9e902 | 1100 | return POLLERR; |
1da177e4 | 1101 | |
54f968d6 | 1102 | sk = tfile->socket.sk; |
3c8a9c63 | 1103 | |
6b8a66ee | 1104 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 1105 | |
9e641bdc | 1106 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1107 | |
1576d986 | 1108 | if (!skb_array_empty(&tfile->tx_array)) |
1da177e4 LT |
1109 | mask |= POLLIN | POLLRDNORM; |
1110 | ||
33dccbb0 | 1111 | if (sock_writeable(sk) || |
9cd3e072 | 1112 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && |
33dccbb0 HX |
1113 | sock_writeable(sk))) |
1114 | mask |= POLLOUT | POLLWRNORM; | |
1115 | ||
c70f1829 EB |
1116 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1117 | mask = POLLERR; | |
1118 | ||
631ab46b | 1119 | tun_put(tun); |
1da177e4 LT |
1120 | return mask; |
1121 | } | |
1122 | ||
f42157cb RR |
1123 | /* prepad is the amount to reserve at front. len is length after that. |
1124 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1125 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1126 | size_t prepad, size_t len, |
1127 | size_t linear, int noblock) | |
f42157cb | 1128 | { |
54f968d6 | 1129 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1130 | struct sk_buff *skb; |
33dccbb0 | 1131 | int err; |
f42157cb RR |
1132 | |
1133 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1134 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1135 | linear = len; |
f42157cb | 1136 | |
33dccbb0 | 1137 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1138 | &err, 0); |
f42157cb | 1139 | if (!skb) |
33dccbb0 | 1140 | return ERR_PTR(err); |
f42157cb RR |
1141 | |
1142 | skb_reserve(skb, prepad); | |
1143 | skb_put(skb, linear); | |
33dccbb0 HX |
1144 | skb->data_len = len - linear; |
1145 | skb->len += len - linear; | |
f42157cb RR |
1146 | |
1147 | return skb; | |
1148 | } | |
1149 | ||
1da177e4 | 1150 | /* Get packet from user space buffer */ |
54f968d6 | 1151 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 AV |
1152 | void *msg_control, struct iov_iter *from, |
1153 | int noblock) | |
1da177e4 | 1154 | { |
09640e63 | 1155 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1156 | struct sk_buff *skb; |
f5ff53b4 | 1157 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1158 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1159 | struct virtio_net_hdr gso = { 0 }; |
608b9977 | 1160 | struct tun_pcpu_stats *stats; |
96f8d9ec | 1161 | int good_linear; |
0690899b MT |
1162 | int copylen; |
1163 | bool zerocopy = false; | |
1164 | int err; | |
49974420 | 1165 | u32 rxhash; |
f5ff53b4 | 1166 | ssize_t n; |
1da177e4 | 1167 | |
1bd4978a ED |
1168 | if (!(tun->dev->flags & IFF_UP)) |
1169 | return -EIO; | |
1170 | ||
40630b82 | 1171 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1172 | if (len < sizeof(pi)) |
1da177e4 | 1173 | return -EINVAL; |
15718ea0 | 1174 | len -= sizeof(pi); |
1da177e4 | 1175 | |
f5ff53b4 AV |
1176 | n = copy_from_iter(&pi, sizeof(pi), from); |
1177 | if (n != sizeof(pi)) | |
1da177e4 LT |
1178 | return -EFAULT; |
1179 | } | |
1180 | ||
40630b82 | 1181 | if (tun->flags & IFF_VNET_HDR) { |
15718ea0 | 1182 | if (len < tun->vnet_hdr_sz) |
f43798c2 | 1183 | return -EINVAL; |
15718ea0 | 1184 | len -= tun->vnet_hdr_sz; |
f43798c2 | 1185 | |
f5ff53b4 AV |
1186 | n = copy_from_iter(&gso, sizeof(gso), from); |
1187 | if (n != sizeof(gso)) | |
f43798c2 RR |
1188 | return -EFAULT; |
1189 | ||
4909122f | 1190 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1191 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1192 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1193 | |
56f0dcc5 | 1194 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1195 | return -EINVAL; |
d8febb77 | 1196 | iov_iter_advance(from, tun->vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1197 | } |
1198 | ||
40630b82 | 1199 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1200 | align += NET_IP_ALIGN; |
0eca93bc | 1201 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1202 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1203 | return -EINVAL; |
1204 | } | |
6aa20a22 | 1205 | |
96f8d9ec JW |
1206 | good_linear = SKB_MAX_HEAD(align); |
1207 | ||
88529176 | 1208 | if (msg_control) { |
f5ff53b4 AV |
1209 | struct iov_iter i = *from; |
1210 | ||
88529176 JW |
1211 | /* There are 256 bytes to be copied in skb, so there is |
1212 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1213 | * The rest of the buffer is mapped from userspace. |
1214 | */ | |
56f0dcc5 | 1215 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1216 | if (copylen > good_linear) |
1217 | copylen = good_linear; | |
3dd5c330 | 1218 | linear = copylen; |
f5ff53b4 AV |
1219 | iov_iter_advance(&i, copylen); |
1220 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1221 | zerocopy = true; |
1222 | } | |
1223 | ||
1224 | if (!zerocopy) { | |
0690899b | 1225 | copylen = len; |
56f0dcc5 | 1226 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) |
96f8d9ec JW |
1227 | linear = good_linear; |
1228 | else | |
56f0dcc5 | 1229 | linear = tun16_to_cpu(tun, gso.hdr_len); |
3dd5c330 | 1230 | } |
0690899b | 1231 | |
3dd5c330 | 1232 | skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); |
33dccbb0 HX |
1233 | if (IS_ERR(skb)) { |
1234 | if (PTR_ERR(skb) != -EAGAIN) | |
608b9977 | 1235 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
33dccbb0 | 1236 | return PTR_ERR(skb); |
1da177e4 LT |
1237 | } |
1238 | ||
0690899b | 1239 | if (zerocopy) |
f5ff53b4 | 1240 | err = zerocopy_sg_from_iter(skb, from); |
88529176 | 1241 | else { |
f5ff53b4 | 1242 | err = skb_copy_datagram_from_iter(skb, 0, from, len); |
88529176 JW |
1243 | if (!err && msg_control) { |
1244 | struct ubuf_info *uarg = msg_control; | |
1245 | uarg->callback(uarg, false); | |
1246 | } | |
1247 | } | |
0690899b MT |
1248 | |
1249 | if (err) { | |
608b9977 | 1250 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
8f22757e | 1251 | kfree_skb(skb); |
1da177e4 | 1252 | return -EFAULT; |
8f22757e | 1253 | } |
1da177e4 | 1254 | |
df10db98 PA |
1255 | err = virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun)); |
1256 | if (err) { | |
1257 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); | |
1258 | kfree_skb(skb); | |
1259 | return -EINVAL; | |
1260 | } | |
1261 | ||
1da177e4 | 1262 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1263 | case IFF_TUN: |
1264 | if (tun->flags & IFF_NO_PI) { | |
f09f7ee2 AWC |
1265 | switch (skb->data[0] & 0xf0) { |
1266 | case 0x40: | |
1267 | pi.proto = htons(ETH_P_IP); | |
1268 | break; | |
1269 | case 0x60: | |
1270 | pi.proto = htons(ETH_P_IPV6); | |
1271 | break; | |
1272 | default: | |
608b9977 | 1273 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
f09f7ee2 AWC |
1274 | kfree_skb(skb); |
1275 | return -EINVAL; | |
1276 | } | |
1277 | } | |
1278 | ||
459a98ed | 1279 | skb_reset_mac_header(skb); |
1da177e4 | 1280 | skb->protocol = pi.proto; |
4c13eb66 | 1281 | skb->dev = tun->dev; |
1da177e4 | 1282 | break; |
40630b82 | 1283 | case IFF_TAP: |
1da177e4 LT |
1284 | skb->protocol = eth_type_trans(skb, tun->dev); |
1285 | break; | |
6403eab1 | 1286 | } |
1da177e4 | 1287 | |
0690899b MT |
1288 | /* copy skb_ubuf_info for callback when skb has no error */ |
1289 | if (zerocopy) { | |
1290 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1291 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1292 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
0690899b MT |
1293 | } |
1294 | ||
72f65107 | 1295 | skb_reset_network_header(skb); |
40893fd0 | 1296 | skb_probe_transport_header(skb, 0); |
38502af7 | 1297 | |
3958afa1 | 1298 | rxhash = skb_get_hash(skb); |
1da177e4 | 1299 | netif_rx_ni(skb); |
6aa20a22 | 1300 | |
608b9977 PA |
1301 | stats = get_cpu_ptr(tun->pcpu_stats); |
1302 | u64_stats_update_begin(&stats->syncp); | |
1303 | stats->rx_packets++; | |
1304 | stats->rx_bytes += len; | |
1305 | u64_stats_update_end(&stats->syncp); | |
1306 | put_cpu_ptr(stats); | |
1da177e4 | 1307 | |
9e85722d | 1308 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1309 | return total_len; |
6aa20a22 | 1310 | } |
1da177e4 | 1311 | |
f5ff53b4 | 1312 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1313 | { |
33dccbb0 | 1314 | struct file *file = iocb->ki_filp; |
ab46d779 | 1315 | struct tun_struct *tun = tun_get(file); |
54f968d6 | 1316 | struct tun_file *tfile = file->private_data; |
631ab46b | 1317 | ssize_t result; |
1da177e4 LT |
1318 | |
1319 | if (!tun) | |
1320 | return -EBADFD; | |
1321 | ||
f5ff53b4 | 1322 | result = tun_get_user(tun, tfile, NULL, from, file->f_flags & O_NONBLOCK); |
631ab46b EB |
1323 | |
1324 | tun_put(tun); | |
1325 | return result; | |
1da177e4 LT |
1326 | } |
1327 | ||
1da177e4 | 1328 | /* Put packet to the user space buffer */ |
6f7c156c | 1329 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1330 | struct tun_file *tfile, |
6f7c156c | 1331 | struct sk_buff *skb, |
e0b46d0e | 1332 | struct iov_iter *iter) |
1da177e4 LT |
1333 | { |
1334 | struct tun_pi pi = { 0, skb->protocol }; | |
608b9977 | 1335 | struct tun_pcpu_stats *stats; |
e0b46d0e | 1336 | ssize_t total; |
8c847d25 | 1337 | int vlan_offset = 0; |
a8f9bfdf | 1338 | int vlan_hlen = 0; |
2eb783c4 | 1339 | int vnet_hdr_sz = 0; |
a8f9bfdf | 1340 | |
df8a39de | 1341 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 1342 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 1343 | |
40630b82 | 1344 | if (tun->flags & IFF_VNET_HDR) |
2eb783c4 | 1345 | vnet_hdr_sz = tun->vnet_hdr_sz; |
1da177e4 | 1346 | |
e0b46d0e HX |
1347 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
1348 | ||
40630b82 | 1349 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 1350 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
1351 | return -EINVAL; |
1352 | ||
e0b46d0e HX |
1353 | total += sizeof(pi); |
1354 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
1355 | /* Packet will be striped */ |
1356 | pi.flags |= TUN_PKT_STRIP; | |
1357 | } | |
6aa20a22 | 1358 | |
e0b46d0e | 1359 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 1360 | return -EFAULT; |
6aa20a22 | 1361 | } |
1da177e4 | 1362 | |
2eb783c4 | 1363 | if (vnet_hdr_sz) { |
f43798c2 | 1364 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ |
34166093 MR |
1365 | int ret; |
1366 | ||
e0b46d0e | 1367 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
1368 | return -EINVAL; |
1369 | ||
34166093 MR |
1370 | ret = virtio_net_hdr_from_skb(skb, &gso, |
1371 | tun_is_little_endian(tun)); | |
1372 | if (ret) { | |
f43798c2 | 1373 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
1374 | pr_err("unexpected GSO type: " |
1375 | "0x%x, gso_size %d, hdr_len %d\n", | |
1376 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
1377 | tun16_to_cpu(tun, gso.hdr_len)); | |
1378 | print_hex_dump(KERN_ERR, "tun: ", | |
1379 | DUMP_PREFIX_NONE, | |
1380 | 16, 1, skb->head, | |
1381 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
1382 | WARN_ON_ONCE(1); | |
1383 | return -EINVAL; | |
1384 | } | |
f43798c2 | 1385 | |
e0b46d0e | 1386 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 1387 | return -EFAULT; |
8c847d25 JW |
1388 | |
1389 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
1390 | } |
1391 | ||
a8f9bfdf | 1392 | if (vlan_hlen) { |
e0b46d0e | 1393 | int ret; |
6680ec68 JW |
1394 | struct { |
1395 | __be16 h_vlan_proto; | |
1396 | __be16 h_vlan_TCI; | |
1397 | } veth; | |
1398 | ||
1399 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 1400 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
1401 | |
1402 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 1403 | |
e0b46d0e HX |
1404 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
1405 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
1406 | goto done; |
1407 | ||
e0b46d0e HX |
1408 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
1409 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
1410 | goto done; |
1411 | } | |
1da177e4 | 1412 | |
e0b46d0e | 1413 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 1414 | |
6680ec68 | 1415 | done: |
608b9977 PA |
1416 | /* caller is in process context, */ |
1417 | stats = get_cpu_ptr(tun->pcpu_stats); | |
1418 | u64_stats_update_begin(&stats->syncp); | |
1419 | stats->tx_packets++; | |
1420 | stats->tx_bytes += skb->len + vlan_hlen; | |
1421 | u64_stats_update_end(&stats->syncp); | |
1422 | put_cpu_ptr(tun->pcpu_stats); | |
1da177e4 LT |
1423 | |
1424 | return total; | |
1425 | } | |
1426 | ||
1576d986 JW |
1427 | static struct sk_buff *tun_ring_recv(struct tun_file *tfile, int noblock, |
1428 | int *err) | |
1429 | { | |
1430 | DECLARE_WAITQUEUE(wait, current); | |
1431 | struct sk_buff *skb = NULL; | |
f48cc6b2 | 1432 | int error = 0; |
1576d986 JW |
1433 | |
1434 | skb = skb_array_consume(&tfile->tx_array); | |
1435 | if (skb) | |
1436 | goto out; | |
1437 | if (noblock) { | |
f48cc6b2 | 1438 | error = -EAGAIN; |
1576d986 JW |
1439 | goto out; |
1440 | } | |
1441 | ||
1442 | add_wait_queue(&tfile->wq.wait, &wait); | |
1443 | current->state = TASK_INTERRUPTIBLE; | |
1444 | ||
1445 | while (1) { | |
1446 | skb = skb_array_consume(&tfile->tx_array); | |
1447 | if (skb) | |
1448 | break; | |
1449 | if (signal_pending(current)) { | |
f48cc6b2 | 1450 | error = -ERESTARTSYS; |
1576d986 JW |
1451 | break; |
1452 | } | |
1453 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 1454 | error = -EFAULT; |
1576d986 JW |
1455 | break; |
1456 | } | |
1457 | ||
1458 | schedule(); | |
1459 | } | |
1460 | ||
1461 | current->state = TASK_RUNNING; | |
1462 | remove_wait_queue(&tfile->wq.wait, &wait); | |
1463 | ||
1464 | out: | |
f48cc6b2 | 1465 | *err = error; |
1576d986 JW |
1466 | return skb; |
1467 | } | |
1468 | ||
54f968d6 | 1469 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 AV |
1470 | struct iov_iter *to, |
1471 | int noblock) | |
1da177e4 | 1472 | { |
1da177e4 | 1473 | struct sk_buff *skb; |
9b067034 | 1474 | ssize_t ret; |
1576d986 | 1475 | int err; |
1da177e4 | 1476 | |
3872baf6 | 1477 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1478 | |
9b067034 AV |
1479 | if (!iov_iter_count(to)) |
1480 | return 0; | |
1da177e4 | 1481 | |
1576d986 JW |
1482 | /* Read frames from ring */ |
1483 | skb = tun_ring_recv(tfile, noblock, &err); | |
e0b46d0e | 1484 | if (!skb) |
957f094f | 1485 | return err; |
e0b46d0e | 1486 | |
9b067034 | 1487 | ret = tun_put_user(tun, tfile, skb, to); |
f51a5e82 | 1488 | if (unlikely(ret < 0)) |
f271b2cc | 1489 | kfree_skb(skb); |
f51a5e82 JW |
1490 | else |
1491 | consume_skb(skb); | |
1da177e4 | 1492 | |
05c2828c MT |
1493 | return ret; |
1494 | } | |
1495 | ||
9b067034 | 1496 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
1497 | { |
1498 | struct file *file = iocb->ki_filp; | |
1499 | struct tun_file *tfile = file->private_data; | |
1500 | struct tun_struct *tun = __tun_get(tfile); | |
9b067034 | 1501 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
1502 | |
1503 | if (!tun) | |
1504 | return -EBADFD; | |
9b067034 | 1505 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK); |
42404c09 | 1506 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
1507 | if (ret > 0) |
1508 | iocb->ki_pos = ret; | |
631ab46b | 1509 | tun_put(tun); |
1da177e4 LT |
1510 | return ret; |
1511 | } | |
1512 | ||
96442e42 JW |
1513 | static void tun_free_netdev(struct net_device *dev) |
1514 | { | |
1515 | struct tun_struct *tun = netdev_priv(dev); | |
1516 | ||
4008e97f | 1517 | BUG_ON(!(list_empty(&tun->disabled))); |
608b9977 | 1518 | free_percpu(tun->pcpu_stats); |
96442e42 | 1519 | tun_flow_uninit(tun); |
5dbbaf2d | 1520 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1521 | free_netdev(dev); |
1522 | } | |
1523 | ||
1da177e4 LT |
1524 | static void tun_setup(struct net_device *dev) |
1525 | { | |
1526 | struct tun_struct *tun = netdev_priv(dev); | |
1527 | ||
0625c883 EB |
1528 | tun->owner = INVALID_UID; |
1529 | tun->group = INVALID_GID; | |
1da177e4 | 1530 | |
1da177e4 | 1531 | dev->ethtool_ops = &tun_ethtool_ops; |
96442e42 | 1532 | dev->destructor = tun_free_netdev; |
016adb72 JW |
1533 | /* We prefer our own queue length */ |
1534 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
1535 | } |
1536 | ||
f019a7a5 EB |
1537 | /* Trivial set of netlink ops to allow deleting tun or tap |
1538 | * device with netlink. | |
1539 | */ | |
1540 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1541 | { | |
1542 | return -EINVAL; | |
1543 | } | |
1544 | ||
1545 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
1546 | .kind = DRV_NAME, | |
1547 | .priv_size = sizeof(struct tun_struct), | |
1548 | .setup = tun_setup, | |
1549 | .validate = tun_validate, | |
1550 | }; | |
1551 | ||
33dccbb0 HX |
1552 | static void tun_sock_write_space(struct sock *sk) |
1553 | { | |
54f968d6 | 1554 | struct tun_file *tfile; |
43815482 | 1555 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
1556 | |
1557 | if (!sock_writeable(sk)) | |
1558 | return; | |
1559 | ||
9cd3e072 | 1560 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
1561 | return; |
1562 | ||
43815482 ED |
1563 | wqueue = sk_sleep(sk); |
1564 | if (wqueue && waitqueue_active(wqueue)) | |
1565 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 1566 | POLLWRNORM | POLLWRBAND); |
c722c625 | 1567 | |
54f968d6 JW |
1568 | tfile = container_of(sk, struct tun_file, sk); |
1569 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
1570 | } |
1571 | ||
1b784140 | 1572 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 1573 | { |
54f968d6 JW |
1574 | int ret; |
1575 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1576 | struct tun_struct *tun = __tun_get(tfile); | |
1577 | ||
1578 | if (!tun) | |
1579 | return -EBADFD; | |
f5ff53b4 | 1580 | |
c0371da6 | 1581 | ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter, |
f5ff53b4 | 1582 | m->msg_flags & MSG_DONTWAIT); |
54f968d6 JW |
1583 | tun_put(tun); |
1584 | return ret; | |
05c2828c MT |
1585 | } |
1586 | ||
1b784140 | 1587 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
1588 | int flags) |
1589 | { | |
54f968d6 JW |
1590 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
1591 | struct tun_struct *tun = __tun_get(tfile); | |
05c2828c | 1592 | int ret; |
54f968d6 JW |
1593 | |
1594 | if (!tun) | |
1595 | return -EBADFD; | |
1596 | ||
eda29772 | 1597 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
1598 | ret = -EINVAL; |
1599 | goto out; | |
1600 | } | |
eda29772 RC |
1601 | if (flags & MSG_ERRQUEUE) { |
1602 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
1603 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
1604 | goto out; | |
1605 | } | |
c0371da6 | 1606 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT); |
87897931 | 1607 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
1608 | m->msg_flags |= MSG_TRUNC; |
1609 | ret = flags & MSG_TRUNC ? ret : total_len; | |
1610 | } | |
3811ae76 | 1611 | out: |
54f968d6 | 1612 | tun_put(tun); |
05c2828c MT |
1613 | return ret; |
1614 | } | |
1615 | ||
1576d986 JW |
1616 | static int tun_peek_len(struct socket *sock) |
1617 | { | |
1618 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1619 | struct tun_struct *tun; | |
1620 | int ret = 0; | |
1621 | ||
1622 | tun = __tun_get(tfile); | |
1623 | if (!tun) | |
1624 | return 0; | |
1625 | ||
1626 | ret = skb_array_peek_len(&tfile->tx_array); | |
1627 | tun_put(tun); | |
1628 | ||
1629 | return ret; | |
1630 | } | |
1631 | ||
05c2828c MT |
1632 | /* Ops structure to mimic raw sockets with tun */ |
1633 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 1634 | .peek_len = tun_peek_len, |
05c2828c MT |
1635 | .sendmsg = tun_sendmsg, |
1636 | .recvmsg = tun_recvmsg, | |
1637 | }; | |
1638 | ||
33dccbb0 HX |
1639 | static struct proto tun_proto = { |
1640 | .name = "tun", | |
1641 | .owner = THIS_MODULE, | |
54f968d6 | 1642 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 1643 | }; |
f019a7a5 | 1644 | |
980c9e8c DW |
1645 | static int tun_flags(struct tun_struct *tun) |
1646 | { | |
031f5e03 | 1647 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
1648 | } |
1649 | ||
1650 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
1651 | char *buf) | |
1652 | { | |
1653 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
1654 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
1655 | } | |
1656 | ||
1657 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
1658 | char *buf) | |
1659 | { | |
1660 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1661 | return uid_valid(tun->owner)? |
1662 | sprintf(buf, "%u\n", | |
1663 | from_kuid_munged(current_user_ns(), tun->owner)): | |
1664 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1665 | } |
1666 | ||
1667 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
1668 | char *buf) | |
1669 | { | |
1670 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1671 | return gid_valid(tun->group) ? |
1672 | sprintf(buf, "%u\n", | |
1673 | from_kgid_munged(current_user_ns(), tun->group)): | |
1674 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1675 | } |
1676 | ||
1677 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
1678 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
1679 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
1680 | ||
c4d33e24 TI |
1681 | static struct attribute *tun_dev_attrs[] = { |
1682 | &dev_attr_tun_flags.attr, | |
1683 | &dev_attr_owner.attr, | |
1684 | &dev_attr_group.attr, | |
1685 | NULL | |
1686 | }; | |
1687 | ||
1688 | static const struct attribute_group tun_attr_group = { | |
1689 | .attrs = tun_dev_attrs | |
1690 | }; | |
1691 | ||
d647a591 | 1692 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
1693 | { |
1694 | struct tun_struct *tun; | |
54f968d6 | 1695 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
1696 | struct net_device *dev; |
1697 | int err; | |
1698 | ||
7c0c3b1a JW |
1699 | if (tfile->detached) |
1700 | return -EINVAL; | |
1701 | ||
74a3e5a7 EB |
1702 | dev = __dev_get_by_name(net, ifr->ifr_name); |
1703 | if (dev) { | |
f85ba780 DW |
1704 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
1705 | return -EBUSY; | |
74a3e5a7 EB |
1706 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
1707 | tun = netdev_priv(dev); | |
1708 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
1709 | tun = netdev_priv(dev); | |
1710 | else | |
1711 | return -EINVAL; | |
1712 | ||
8e6d91ae | 1713 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 1714 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
1715 | return -EINVAL; |
1716 | ||
cde8b15f | 1717 | if (tun_not_capable(tun)) |
2b980dbd | 1718 | return -EPERM; |
5dbbaf2d | 1719 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
1720 | if (err < 0) |
1721 | return err; | |
1722 | ||
849c9b6f | 1723 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER); |
a7385ba2 EB |
1724 | if (err < 0) |
1725 | return err; | |
4008e97f | 1726 | |
40630b82 | 1727 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
1728 | (tun->numqueues + tun->numdisabled > 1)) { |
1729 | /* One or more queue has already been attached, no need | |
1730 | * to initialize the device again. | |
1731 | */ | |
1732 | return 0; | |
1733 | } | |
6aa20a22 | 1734 | } |
1da177e4 LT |
1735 | else { |
1736 | char *name; | |
1737 | unsigned long flags = 0; | |
edfb6a14 JW |
1738 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
1739 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 1740 | |
c260b772 | 1741 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 1742 | return -EPERM; |
2b980dbd PM |
1743 | err = security_tun_dev_create(); |
1744 | if (err < 0) | |
1745 | return err; | |
ca6bb5d7 | 1746 | |
1da177e4 LT |
1747 | /* Set dev type */ |
1748 | if (ifr->ifr_flags & IFF_TUN) { | |
1749 | /* TUN device */ | |
40630b82 | 1750 | flags |= IFF_TUN; |
1da177e4 LT |
1751 | name = "tun%d"; |
1752 | } else if (ifr->ifr_flags & IFF_TAP) { | |
1753 | /* TAP device */ | |
40630b82 | 1754 | flags |= IFF_TAP; |
1da177e4 | 1755 | name = "tap%d"; |
6aa20a22 | 1756 | } else |
36989b90 | 1757 | return -EINVAL; |
6aa20a22 | 1758 | |
1da177e4 LT |
1759 | if (*ifr->ifr_name) |
1760 | name = ifr->ifr_name; | |
1761 | ||
c8d68e6b | 1762 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
1763 | NET_NAME_UNKNOWN, tun_setup, queues, |
1764 | queues); | |
edfb6a14 | 1765 | |
1da177e4 LT |
1766 | if (!dev) |
1767 | return -ENOMEM; | |
1768 | ||
fc54c658 | 1769 | dev_net_set(dev, net); |
f019a7a5 | 1770 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 1771 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 1772 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 1773 | |
1da177e4 LT |
1774 | tun = netdev_priv(dev); |
1775 | tun->dev = dev; | |
1776 | tun->flags = flags; | |
f271b2cc | 1777 | tun->txflt.count = 0; |
d9d52b51 | 1778 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 1779 | |
eaea34b2 | 1780 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
1781 | tun->filter_attached = false; |
1782 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
33dccbb0 | 1783 | |
608b9977 PA |
1784 | tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); |
1785 | if (!tun->pcpu_stats) { | |
1786 | err = -ENOMEM; | |
1787 | goto err_free_dev; | |
1788 | } | |
1789 | ||
96442e42 JW |
1790 | spin_lock_init(&tun->lock); |
1791 | ||
5dbbaf2d PM |
1792 | err = security_tun_dev_alloc_security(&tun->security); |
1793 | if (err < 0) | |
608b9977 | 1794 | goto err_free_stat; |
2b980dbd | 1795 | |
1da177e4 | 1796 | tun_net_init(dev); |
944a1376 | 1797 | tun_flow_init(tun); |
96442e42 | 1798 | |
88255375 | 1799 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
1800 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
1801 | NETIF_F_HW_VLAN_STAG_TX; | |
2a2bbf17 | 1802 | dev->features = dev->hw_features | NETIF_F_LLTX; |
6671b224 FLVC |
1803 | dev->vlan_features = dev->features & |
1804 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
1805 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 1806 | |
4008e97f | 1807 | INIT_LIST_HEAD(&tun->disabled); |
849c9b6f | 1808 | err = tun_attach(tun, file, false); |
eb0fb363 | 1809 | if (err < 0) |
662ca437 | 1810 | goto err_free_flow; |
eb0fb363 | 1811 | |
1da177e4 LT |
1812 | err = register_netdevice(tun->dev); |
1813 | if (err < 0) | |
662ca437 | 1814 | goto err_detach; |
1da177e4 LT |
1815 | } |
1816 | ||
af668b3c MT |
1817 | netif_carrier_on(tun->dev); |
1818 | ||
6b8a66ee | 1819 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 1820 | |
031f5e03 MT |
1821 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
1822 | (ifr->ifr_flags & TUN_FEATURES); | |
c8d68e6b | 1823 | |
e35259a9 MK |
1824 | /* Make sure persistent devices do not get stuck in |
1825 | * xoff state. | |
1826 | */ | |
1827 | if (netif_running(tun->dev)) | |
c8d68e6b | 1828 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 1829 | |
1da177e4 LT |
1830 | strcpy(ifr->ifr_name, tun->dev->name); |
1831 | return 0; | |
1832 | ||
662ca437 JW |
1833 | err_detach: |
1834 | tun_detach_all(dev); | |
1835 | err_free_flow: | |
1836 | tun_flow_uninit(tun); | |
1837 | security_tun_dev_free_security(tun->security); | |
608b9977 PA |
1838 | err_free_stat: |
1839 | free_percpu(tun->pcpu_stats); | |
662ca437 | 1840 | err_free_dev: |
1da177e4 | 1841 | free_netdev(dev); |
1da177e4 LT |
1842 | return err; |
1843 | } | |
1844 | ||
9ce99cf6 | 1845 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 1846 | struct ifreq *ifr) |
e3b99556 | 1847 | { |
6b8a66ee | 1848 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
1849 | |
1850 | strcpy(ifr->ifr_name, tun->dev->name); | |
1851 | ||
980c9e8c | 1852 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 1853 | |
e3b99556 MM |
1854 | } |
1855 | ||
5228ddc9 RR |
1856 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
1857 | * privs required. */ | |
88255375 | 1858 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 1859 | { |
c8f44aff | 1860 | netdev_features_t features = 0; |
5228ddc9 RR |
1861 | |
1862 | if (arg & TUN_F_CSUM) { | |
88255375 | 1863 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
1864 | arg &= ~TUN_F_CSUM; |
1865 | ||
1866 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
1867 | if (arg & TUN_F_TSO_ECN) { | |
1868 | features |= NETIF_F_TSO_ECN; | |
1869 | arg &= ~TUN_F_TSO_ECN; | |
1870 | } | |
1871 | if (arg & TUN_F_TSO4) | |
1872 | features |= NETIF_F_TSO; | |
1873 | if (arg & TUN_F_TSO6) | |
1874 | features |= NETIF_F_TSO6; | |
1875 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
1876 | } | |
e3e3c423 VY |
1877 | |
1878 | if (arg & TUN_F_UFO) { | |
1879 | features |= NETIF_F_UFO; | |
1880 | arg &= ~TUN_F_UFO; | |
1881 | } | |
5228ddc9 RR |
1882 | } |
1883 | ||
1884 | /* This gives the user a way to test for new features in future by | |
1885 | * trying to set them. */ | |
1886 | if (arg) | |
1887 | return -EINVAL; | |
1888 | ||
88255375 MM |
1889 | tun->set_features = features; |
1890 | netdev_update_features(tun->dev); | |
5228ddc9 RR |
1891 | |
1892 | return 0; | |
1893 | } | |
1894 | ||
c8d68e6b JW |
1895 | static void tun_detach_filter(struct tun_struct *tun, int n) |
1896 | { | |
1897 | int i; | |
1898 | struct tun_file *tfile; | |
1899 | ||
1900 | for (i = 0; i < n; i++) { | |
b8deabd3 | 1901 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
1902 | lock_sock(tfile->socket.sk); |
1903 | sk_detach_filter(tfile->socket.sk); | |
1904 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
1905 | } |
1906 | ||
1907 | tun->filter_attached = false; | |
1908 | } | |
1909 | ||
1910 | static int tun_attach_filter(struct tun_struct *tun) | |
1911 | { | |
1912 | int i, ret = 0; | |
1913 | struct tun_file *tfile; | |
1914 | ||
1915 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1916 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
1917 | lock_sock(tfile->socket.sk); |
1918 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
1919 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
1920 | if (ret) { |
1921 | tun_detach_filter(tun, i); | |
1922 | return ret; | |
1923 | } | |
1924 | } | |
1925 | ||
1926 | tun->filter_attached = true; | |
1927 | return ret; | |
1928 | } | |
1929 | ||
1930 | static void tun_set_sndbuf(struct tun_struct *tun) | |
1931 | { | |
1932 | struct tun_file *tfile; | |
1933 | int i; | |
1934 | ||
1935 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1936 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1937 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
1938 | } | |
1939 | } | |
1940 | ||
cde8b15f JW |
1941 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
1942 | { | |
1943 | struct tun_file *tfile = file->private_data; | |
1944 | struct tun_struct *tun; | |
cde8b15f JW |
1945 | int ret = 0; |
1946 | ||
1947 | rtnl_lock(); | |
1948 | ||
1949 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 1950 | tun = tfile->detached; |
5dbbaf2d | 1951 | if (!tun) { |
cde8b15f | 1952 | ret = -EINVAL; |
5dbbaf2d PM |
1953 | goto unlock; |
1954 | } | |
1955 | ret = security_tun_dev_attach_queue(tun->security); | |
1956 | if (ret < 0) | |
1957 | goto unlock; | |
849c9b6f | 1958 | ret = tun_attach(tun, file, false); |
4008e97f | 1959 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 1960 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 1961 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
1962 | ret = -EINVAL; |
1963 | else | |
1964 | __tun_detach(tfile, false); | |
1965 | } else | |
cde8b15f JW |
1966 | ret = -EINVAL; |
1967 | ||
5dbbaf2d | 1968 | unlock: |
cde8b15f JW |
1969 | rtnl_unlock(); |
1970 | return ret; | |
1971 | } | |
1972 | ||
50857e2a AB |
1973 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
1974 | unsigned long arg, int ifreq_len) | |
1da177e4 | 1975 | { |
36b50bab | 1976 | struct tun_file *tfile = file->private_data; |
631ab46b | 1977 | struct tun_struct *tun; |
1da177e4 LT |
1978 | void __user* argp = (void __user*)arg; |
1979 | struct ifreq ifr; | |
0625c883 EB |
1980 | kuid_t owner; |
1981 | kgid_t group; | |
33dccbb0 | 1982 | int sndbuf; |
d9d52b51 | 1983 | int vnet_hdr_sz; |
fb7589a1 | 1984 | unsigned int ifindex; |
1cf8e410 | 1985 | int le; |
f271b2cc | 1986 | int ret; |
1da177e4 | 1987 | |
20861f26 | 1988 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == SOCK_IOC_TYPE) { |
50857e2a | 1989 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 1990 | return -EFAULT; |
8bbb1813 | 1991 | } else { |
a117dacd | 1992 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 1993 | } |
631ab46b EB |
1994 | if (cmd == TUNGETFEATURES) { |
1995 | /* Currently this just means: "what IFF flags are valid?". | |
1996 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
1997 | * TUNSETIFF. |
1998 | */ | |
1999 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 2000 | (unsigned int __user*)argp); |
cde8b15f JW |
2001 | } else if (cmd == TUNSETQUEUE) |
2002 | return tun_set_queue(file, &ifr); | |
631ab46b | 2003 | |
c8d68e6b | 2004 | ret = 0; |
876bfd4d HX |
2005 | rtnl_lock(); |
2006 | ||
36b50bab | 2007 | tun = __tun_get(tfile); |
0f16bc13 GF |
2008 | if (cmd == TUNSETIFF) { |
2009 | ret = -EEXIST; | |
2010 | if (tun) | |
2011 | goto unlock; | |
2012 | ||
1da177e4 LT |
2013 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
2014 | ||
140e807d | 2015 | ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr); |
1da177e4 | 2016 | |
876bfd4d HX |
2017 | if (ret) |
2018 | goto unlock; | |
1da177e4 | 2019 | |
50857e2a | 2020 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
2021 | ret = -EFAULT; |
2022 | goto unlock; | |
1da177e4 | 2023 | } |
fb7589a1 PE |
2024 | if (cmd == TUNSETIFINDEX) { |
2025 | ret = -EPERM; | |
2026 | if (tun) | |
2027 | goto unlock; | |
2028 | ||
2029 | ret = -EFAULT; | |
2030 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
2031 | goto unlock; | |
2032 | ||
2033 | ret = 0; | |
2034 | tfile->ifindex = ifindex; | |
2035 | goto unlock; | |
2036 | } | |
1da177e4 | 2037 | |
876bfd4d | 2038 | ret = -EBADFD; |
1da177e4 | 2039 | if (!tun) |
876bfd4d | 2040 | goto unlock; |
1da177e4 | 2041 | |
1e588338 | 2042 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 2043 | |
631ab46b | 2044 | ret = 0; |
1da177e4 | 2045 | switch (cmd) { |
e3b99556 | 2046 | case TUNGETIFF: |
9ce99cf6 | 2047 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 2048 | |
3d407a80 PE |
2049 | if (tfile->detached) |
2050 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
2051 | if (!tfile->socket.sk->sk_filter) |
2052 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 2053 | |
50857e2a | 2054 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 2055 | ret = -EFAULT; |
e3b99556 MM |
2056 | break; |
2057 | ||
1da177e4 LT |
2058 | case TUNSETNOCSUM: |
2059 | /* Disable/Enable checksum */ | |
1da177e4 | 2060 | |
88255375 MM |
2061 | /* [unimplemented] */ |
2062 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 2063 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
2064 | break; |
2065 | ||
2066 | case TUNSETPERSIST: | |
54f968d6 JW |
2067 | /* Disable/Enable persist mode. Keep an extra reference to the |
2068 | * module to prevent the module being unprobed. | |
2069 | */ | |
40630b82 MT |
2070 | if (arg && !(tun->flags & IFF_PERSIST)) { |
2071 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 2072 | __module_get(THIS_MODULE); |
dd38bd85 | 2073 | } |
40630b82 MT |
2074 | if (!arg && (tun->flags & IFF_PERSIST)) { |
2075 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 JW |
2076 | module_put(THIS_MODULE); |
2077 | } | |
1da177e4 | 2078 | |
6b8a66ee JP |
2079 | tun_debug(KERN_INFO, tun, "persist %s\n", |
2080 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
2081 | break; |
2082 | ||
2083 | case TUNSETOWNER: | |
2084 | /* Set owner of the device */ | |
0625c883 EB |
2085 | owner = make_kuid(current_user_ns(), arg); |
2086 | if (!uid_valid(owner)) { | |
2087 | ret = -EINVAL; | |
2088 | break; | |
2089 | } | |
2090 | tun->owner = owner; | |
1e588338 | 2091 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 2092 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
2093 | break; |
2094 | ||
8c644623 GG |
2095 | case TUNSETGROUP: |
2096 | /* Set group of the device */ | |
0625c883 EB |
2097 | group = make_kgid(current_user_ns(), arg); |
2098 | if (!gid_valid(group)) { | |
2099 | ret = -EINVAL; | |
2100 | break; | |
2101 | } | |
2102 | tun->group = group; | |
1e588338 | 2103 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 2104 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
2105 | break; |
2106 | ||
ff4cc3ac MK |
2107 | case TUNSETLINK: |
2108 | /* Only allow setting the type when the interface is down */ | |
2109 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
2110 | tun_debug(KERN_INFO, tun, |
2111 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 2112 | ret = -EBUSY; |
ff4cc3ac MK |
2113 | } else { |
2114 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2115 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2116 | tun->dev->type); | |
48abfe05 | 2117 | ret = 0; |
ff4cc3ac | 2118 | } |
631ab46b | 2119 | break; |
ff4cc3ac | 2120 | |
1da177e4 LT |
2121 | #ifdef TUN_DEBUG |
2122 | case TUNSETDEBUG: | |
2123 | tun->debug = arg; | |
2124 | break; | |
2125 | #endif | |
5228ddc9 | 2126 | case TUNSETOFFLOAD: |
88255375 | 2127 | ret = set_offload(tun, arg); |
631ab46b | 2128 | break; |
5228ddc9 | 2129 | |
f271b2cc MK |
2130 | case TUNSETTXFILTER: |
2131 | /* Can be set only for TAPs */ | |
631ab46b | 2132 | ret = -EINVAL; |
40630b82 | 2133 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 2134 | break; |
c0e5a8c2 | 2135 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2136 | break; |
1da177e4 LT |
2137 | |
2138 | case SIOCGIFHWADDR: | |
b595076a | 2139 | /* Get hw address */ |
f271b2cc MK |
2140 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2141 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2142 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2143 | ret = -EFAULT; |
2144 | break; | |
1da177e4 LT |
2145 | |
2146 | case SIOCSIFHWADDR: | |
f271b2cc | 2147 | /* Set hw address */ |
6b8a66ee JP |
2148 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2149 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2150 | |
40102371 | 2151 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2152 | break; |
33dccbb0 HX |
2153 | |
2154 | case TUNGETSNDBUF: | |
54f968d6 | 2155 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2156 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2157 | ret = -EFAULT; | |
2158 | break; | |
2159 | ||
2160 | case TUNSETSNDBUF: | |
2161 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2162 | ret = -EFAULT; | |
2163 | break; | |
2164 | } | |
2165 | ||
c8d68e6b JW |
2166 | tun->sndbuf = sndbuf; |
2167 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2168 | break; |
2169 | ||
d9d52b51 MT |
2170 | case TUNGETVNETHDRSZ: |
2171 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2172 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2173 | ret = -EFAULT; | |
2174 | break; | |
2175 | ||
2176 | case TUNSETVNETHDRSZ: | |
2177 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2178 | ret = -EFAULT; | |
2179 | break; | |
2180 | } | |
2181 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2182 | ret = -EINVAL; | |
2183 | break; | |
2184 | } | |
2185 | ||
2186 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2187 | break; | |
2188 | ||
1cf8e410 MT |
2189 | case TUNGETVNETLE: |
2190 | le = !!(tun->flags & TUN_VNET_LE); | |
2191 | if (put_user(le, (int __user *)argp)) | |
2192 | ret = -EFAULT; | |
2193 | break; | |
2194 | ||
2195 | case TUNSETVNETLE: | |
2196 | if (get_user(le, (int __user *)argp)) { | |
2197 | ret = -EFAULT; | |
2198 | break; | |
2199 | } | |
2200 | if (le) | |
2201 | tun->flags |= TUN_VNET_LE; | |
2202 | else | |
2203 | tun->flags &= ~TUN_VNET_LE; | |
2204 | break; | |
2205 | ||
8b8e658b GK |
2206 | case TUNGETVNETBE: |
2207 | ret = tun_get_vnet_be(tun, argp); | |
2208 | break; | |
2209 | ||
2210 | case TUNSETVNETBE: | |
2211 | ret = tun_set_vnet_be(tun, argp); | |
2212 | break; | |
2213 | ||
99405162 MT |
2214 | case TUNATTACHFILTER: |
2215 | /* Can be set only for TAPs */ | |
2216 | ret = -EINVAL; | |
40630b82 | 2217 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
2218 | break; |
2219 | ret = -EFAULT; | |
54f968d6 | 2220 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2221 | break; |
2222 | ||
c8d68e6b | 2223 | ret = tun_attach_filter(tun); |
99405162 MT |
2224 | break; |
2225 | ||
2226 | case TUNDETACHFILTER: | |
2227 | /* Can be set only for TAPs */ | |
2228 | ret = -EINVAL; | |
40630b82 | 2229 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 2230 | break; |
c8d68e6b JW |
2231 | ret = 0; |
2232 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2233 | break; |
2234 | ||
76975e9c PE |
2235 | case TUNGETFILTER: |
2236 | ret = -EINVAL; | |
40630b82 | 2237 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
2238 | break; |
2239 | ret = -EFAULT; | |
2240 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2241 | break; | |
2242 | ret = 0; | |
2243 | break; | |
2244 | ||
1da177e4 | 2245 | default: |
631ab46b EB |
2246 | ret = -EINVAL; |
2247 | break; | |
ee289b64 | 2248 | } |
1da177e4 | 2249 | |
876bfd4d HX |
2250 | unlock: |
2251 | rtnl_unlock(); | |
2252 | if (tun) | |
2253 | tun_put(tun); | |
631ab46b | 2254 | return ret; |
1da177e4 LT |
2255 | } |
2256 | ||
50857e2a AB |
2257 | static long tun_chr_ioctl(struct file *file, |
2258 | unsigned int cmd, unsigned long arg) | |
2259 | { | |
2260 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2261 | } | |
2262 | ||
2263 | #ifdef CONFIG_COMPAT | |
2264 | static long tun_chr_compat_ioctl(struct file *file, | |
2265 | unsigned int cmd, unsigned long arg) | |
2266 | { | |
2267 | switch (cmd) { | |
2268 | case TUNSETIFF: | |
2269 | case TUNGETIFF: | |
2270 | case TUNSETTXFILTER: | |
2271 | case TUNGETSNDBUF: | |
2272 | case TUNSETSNDBUF: | |
2273 | case SIOCGIFHWADDR: | |
2274 | case SIOCSIFHWADDR: | |
2275 | arg = (unsigned long)compat_ptr(arg); | |
2276 | break; | |
2277 | default: | |
2278 | arg = (compat_ulong_t)arg; | |
2279 | break; | |
2280 | } | |
2281 | ||
2282 | /* | |
2283 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2284 | * the end of that structure. All fields that are used in this | |
2285 | * driver are compatible though, we don't need to convert the | |
2286 | * contents. | |
2287 | */ | |
2288 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2289 | } | |
2290 | #endif /* CONFIG_COMPAT */ | |
2291 | ||
1da177e4 LT |
2292 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2293 | { | |
54f968d6 | 2294 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2295 | int ret; |
2296 | ||
54f968d6 | 2297 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2298 | goto out; |
6aa20a22 | 2299 | |
1da177e4 | 2300 | if (on) { |
e0b93edd | 2301 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2302 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2303 | } else |
54f968d6 | 2304 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2305 | ret = 0; |
2306 | out: | |
9d319522 | 2307 | return ret; |
1da177e4 LT |
2308 | } |
2309 | ||
2310 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2311 | { | |
140e807d | 2312 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 2313 | struct tun_file *tfile; |
deed49fb | 2314 | |
6b8a66ee | 2315 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2316 | |
140e807d | 2317 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 2318 | &tun_proto, 0); |
631ab46b EB |
2319 | if (!tfile) |
2320 | return -ENOMEM; | |
c956674b | 2321 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 2322 | tfile->flags = 0; |
fb7589a1 | 2323 | tfile->ifindex = 0; |
54f968d6 | 2324 | |
54f968d6 | 2325 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2326 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2327 | |
2328 | tfile->socket.file = file; | |
2329 | tfile->socket.ops = &tun_socket_ops; | |
2330 | ||
2331 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
2332 | |
2333 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2334 | tfile->sk.sk_sndbuf = INT_MAX; | |
2335 | ||
631ab46b | 2336 | file->private_data = tfile; |
4008e97f | 2337 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2338 | |
19a6afb2 JW |
2339 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2340 | ||
1da177e4 LT |
2341 | return 0; |
2342 | } | |
2343 | ||
2344 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2345 | { | |
631ab46b | 2346 | struct tun_file *tfile = file->private_data; |
1da177e4 | 2347 | |
c8d68e6b | 2348 | tun_detach(tfile, true); |
1da177e4 LT |
2349 | |
2350 | return 0; | |
2351 | } | |
2352 | ||
93e14b6d | 2353 | #ifdef CONFIG_PROC_FS |
a3816ab0 | 2354 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f) |
93e14b6d MY |
2355 | { |
2356 | struct tun_struct *tun; | |
2357 | struct ifreq ifr; | |
2358 | ||
2359 | memset(&ifr, 0, sizeof(ifr)); | |
2360 | ||
2361 | rtnl_lock(); | |
2362 | tun = tun_get(f); | |
2363 | if (tun) | |
2364 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2365 | rtnl_unlock(); | |
2366 | ||
2367 | if (tun) | |
2368 | tun_put(tun); | |
2369 | ||
a3816ab0 | 2370 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
2371 | } |
2372 | #endif | |
2373 | ||
d54b1fdb | 2374 | static const struct file_operations tun_fops = { |
6aa20a22 | 2375 | .owner = THIS_MODULE, |
1da177e4 | 2376 | .llseek = no_llseek, |
9b067034 | 2377 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 2378 | .write_iter = tun_chr_write_iter, |
1da177e4 | 2379 | .poll = tun_chr_poll, |
50857e2a AB |
2380 | .unlocked_ioctl = tun_chr_ioctl, |
2381 | #ifdef CONFIG_COMPAT | |
2382 | .compat_ioctl = tun_chr_compat_ioctl, | |
2383 | #endif | |
1da177e4 LT |
2384 | .open = tun_chr_open, |
2385 | .release = tun_chr_close, | |
93e14b6d MY |
2386 | .fasync = tun_chr_fasync, |
2387 | #ifdef CONFIG_PROC_FS | |
2388 | .show_fdinfo = tun_chr_show_fdinfo, | |
2389 | #endif | |
1da177e4 LT |
2390 | }; |
2391 | ||
2392 | static struct miscdevice tun_miscdev = { | |
2393 | .minor = TUN_MINOR, | |
2394 | .name = "tun", | |
e454cea2 | 2395 | .nodename = "net/tun", |
1da177e4 | 2396 | .fops = &tun_fops, |
1da177e4 LT |
2397 | }; |
2398 | ||
2399 | /* ethtool interface */ | |
2400 | ||
2401 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) | |
2402 | { | |
2403 | cmd->supported = 0; | |
2404 | cmd->advertising = 0; | |
70739497 | 2405 | ethtool_cmd_speed_set(cmd, SPEED_10); |
1da177e4 LT |
2406 | cmd->duplex = DUPLEX_FULL; |
2407 | cmd->port = PORT_TP; | |
2408 | cmd->phy_address = 0; | |
2409 | cmd->transceiver = XCVR_INTERNAL; | |
2410 | cmd->autoneg = AUTONEG_DISABLE; | |
2411 | cmd->maxtxpkt = 0; | |
2412 | cmd->maxrxpkt = 0; | |
2413 | return 0; | |
2414 | } | |
2415 | ||
2416 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2417 | { | |
2418 | struct tun_struct *tun = netdev_priv(dev); | |
2419 | ||
33a5ba14 RJ |
2420 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2421 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2422 | |
2423 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 2424 | case IFF_TUN: |
33a5ba14 | 2425 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 2426 | break; |
40630b82 | 2427 | case IFF_TAP: |
33a5ba14 | 2428 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2429 | break; |
2430 | } | |
2431 | } | |
2432 | ||
2433 | static u32 tun_get_msglevel(struct net_device *dev) | |
2434 | { | |
2435 | #ifdef TUN_DEBUG | |
2436 | struct tun_struct *tun = netdev_priv(dev); | |
2437 | return tun->debug; | |
2438 | #else | |
2439 | return -EOPNOTSUPP; | |
2440 | #endif | |
2441 | } | |
2442 | ||
2443 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2444 | { | |
2445 | #ifdef TUN_DEBUG | |
2446 | struct tun_struct *tun = netdev_priv(dev); | |
2447 | tun->debug = value; | |
2448 | #endif | |
2449 | } | |
2450 | ||
7282d491 | 2451 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2452 | .get_settings = tun_get_settings, |
2453 | .get_drvinfo = tun_get_drvinfo, | |
2454 | .get_msglevel = tun_get_msglevel, | |
2455 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2456 | .get_link = ethtool_op_get_link, |
eda29772 | 2457 | .get_ts_info = ethtool_op_get_ts_info, |
1da177e4 LT |
2458 | }; |
2459 | ||
1576d986 JW |
2460 | static int tun_queue_resize(struct tun_struct *tun) |
2461 | { | |
2462 | struct net_device *dev = tun->dev; | |
2463 | struct tun_file *tfile; | |
2464 | struct skb_array **arrays; | |
2465 | int n = tun->numqueues + tun->numdisabled; | |
2466 | int ret, i; | |
2467 | ||
2468 | arrays = kmalloc(sizeof *arrays * n, GFP_KERNEL); | |
2469 | if (!arrays) | |
2470 | return -ENOMEM; | |
2471 | ||
2472 | for (i = 0; i < tun->numqueues; i++) { | |
2473 | tfile = rtnl_dereference(tun->tfiles[i]); | |
2474 | arrays[i] = &tfile->tx_array; | |
2475 | } | |
2476 | list_for_each_entry(tfile, &tun->disabled, next) | |
2477 | arrays[i++] = &tfile->tx_array; | |
2478 | ||
2479 | ret = skb_array_resize_multiple(arrays, n, | |
2480 | dev->tx_queue_len, GFP_KERNEL); | |
2481 | ||
2482 | kfree(arrays); | |
2483 | return ret; | |
2484 | } | |
2485 | ||
2486 | static int tun_device_event(struct notifier_block *unused, | |
2487 | unsigned long event, void *ptr) | |
2488 | { | |
2489 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
2490 | struct tun_struct *tun = netdev_priv(dev); | |
2491 | ||
86dfb4ac CG |
2492 | if (dev->rtnl_link_ops != &tun_link_ops) |
2493 | return NOTIFY_DONE; | |
2494 | ||
1576d986 JW |
2495 | switch (event) { |
2496 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
2497 | if (tun_queue_resize(tun)) | |
2498 | return NOTIFY_BAD; | |
2499 | break; | |
2500 | default: | |
2501 | break; | |
2502 | } | |
2503 | ||
2504 | return NOTIFY_DONE; | |
2505 | } | |
2506 | ||
2507 | static struct notifier_block tun_notifier_block __read_mostly = { | |
2508 | .notifier_call = tun_device_event, | |
2509 | }; | |
79d17604 | 2510 | |
1da177e4 LT |
2511 | static int __init tun_init(void) |
2512 | { | |
2513 | int ret = 0; | |
2514 | ||
6b8a66ee JP |
2515 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
2516 | pr_info("%s\n", DRV_COPYRIGHT); | |
1da177e4 | 2517 | |
f019a7a5 | 2518 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 2519 | if (ret) { |
6b8a66ee | 2520 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 2521 | goto err_linkops; |
79d17604 PE |
2522 | } |
2523 | ||
1da177e4 | 2524 | ret = misc_register(&tun_miscdev); |
79d17604 | 2525 | if (ret) { |
6b8a66ee | 2526 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
2527 | goto err_misc; |
2528 | } | |
1576d986 JW |
2529 | |
2530 | register_netdevice_notifier(&tun_notifier_block); | |
f019a7a5 | 2531 | return 0; |
79d17604 | 2532 | err_misc: |
f019a7a5 EB |
2533 | rtnl_link_unregister(&tun_link_ops); |
2534 | err_linkops: | |
1da177e4 LT |
2535 | return ret; |
2536 | } | |
2537 | ||
2538 | static void tun_cleanup(void) | |
2539 | { | |
6aa20a22 | 2540 | misc_deregister(&tun_miscdev); |
f019a7a5 | 2541 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 2542 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
2543 | } |
2544 | ||
05c2828c MT |
2545 | /* Get an underlying socket object from tun file. Returns error unless file is |
2546 | * attached to a device. The returned object works like a packet socket, it | |
2547 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
2548 | * holding a reference to the file for as long as the socket is in use. */ | |
2549 | struct socket *tun_get_socket(struct file *file) | |
2550 | { | |
6e914fc7 | 2551 | struct tun_file *tfile; |
05c2828c MT |
2552 | if (file->f_op != &tun_fops) |
2553 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
2554 | tfile = file->private_data; |
2555 | if (!tfile) | |
05c2828c | 2556 | return ERR_PTR(-EBADFD); |
54f968d6 | 2557 | return &tfile->socket; |
05c2828c MT |
2558 | } |
2559 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
2560 | ||
1da177e4 LT |
2561 | module_init(tun_init); |
2562 | module_exit(tun_cleanup); | |
2563 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
2564 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
2565 | MODULE_LICENSE("GPL"); | |
2566 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 2567 | MODULE_ALIAS("devname:net/tun"); |