]>
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 | |
7c0f6ba6 | 76 | #include <linux/uaccess.h> |
1da177e4 | 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; |
5503fcec | 221 | u32 rx_batched; |
608b9977 | 222 | struct tun_pcpu_stats __percpu *pcpu_stats; |
14daa021 | 223 | }; |
1da177e4 | 224 | |
8b8e658b GK |
225 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
226 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
227 | { | |
228 | return tun->flags & TUN_VNET_BE ? false : | |
229 | virtio_legacy_is_little_endian(); | |
230 | } | |
231 | ||
232 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
233 | { | |
234 | int be = !!(tun->flags & TUN_VNET_BE); | |
235 | ||
236 | if (put_user(be, argp)) | |
237 | return -EFAULT; | |
238 | ||
239 | return 0; | |
240 | } | |
241 | ||
242 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
243 | { | |
244 | int be; | |
245 | ||
246 | if (get_user(be, argp)) | |
247 | return -EFAULT; | |
248 | ||
249 | if (be) | |
250 | tun->flags |= TUN_VNET_BE; | |
251 | else | |
252 | tun->flags &= ~TUN_VNET_BE; | |
253 | ||
254 | return 0; | |
255 | } | |
256 | #else | |
257 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
258 | { | |
259 | return virtio_legacy_is_little_endian(); | |
260 | } | |
261 | ||
262 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
263 | { | |
264 | return -EINVAL; | |
265 | } | |
266 | ||
267 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
268 | { | |
269 | return -EINVAL; | |
270 | } | |
271 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
272 | ||
25bd55bb GK |
273 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
274 | { | |
7d824109 | 275 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 276 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
277 | } |
278 | ||
56f0dcc5 MT |
279 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
280 | { | |
25bd55bb | 281 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
282 | } |
283 | ||
284 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
285 | { | |
25bd55bb | 286 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
287 | } |
288 | ||
96442e42 JW |
289 | static inline u32 tun_hashfn(u32 rxhash) |
290 | { | |
291 | return rxhash & 0x3ff; | |
292 | } | |
293 | ||
294 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
295 | { | |
296 | struct tun_flow_entry *e; | |
96442e42 | 297 | |
b67bfe0d | 298 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
299 | if (e->rxhash == rxhash) |
300 | return e; | |
301 | } | |
302 | return NULL; | |
303 | } | |
304 | ||
305 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
306 | struct hlist_head *head, | |
307 | u32 rxhash, u16 queue_index) | |
308 | { | |
9fdc6bef ED |
309 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
310 | ||
96442e42 JW |
311 | if (e) { |
312 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
313 | rxhash, queue_index); | |
314 | e->updated = jiffies; | |
315 | e->rxhash = rxhash; | |
9bc88939 | 316 | e->rps_rxhash = 0; |
96442e42 JW |
317 | e->queue_index = queue_index; |
318 | e->tun = tun; | |
319 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 320 | ++tun->flow_count; |
96442e42 JW |
321 | } |
322 | return e; | |
323 | } | |
324 | ||
96442e42 JW |
325 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
326 | { | |
327 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
328 | e->rxhash, e->queue_index); | |
329 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 330 | kfree_rcu(e, rcu); |
b8732fb7 | 331 | --tun->flow_count; |
96442e42 JW |
332 | } |
333 | ||
334 | static void tun_flow_flush(struct tun_struct *tun) | |
335 | { | |
336 | int i; | |
337 | ||
338 | spin_lock_bh(&tun->lock); | |
339 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
340 | struct tun_flow_entry *e; | |
b67bfe0d | 341 | struct hlist_node *n; |
96442e42 | 342 | |
b67bfe0d | 343 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
344 | tun_flow_delete(tun, e); |
345 | } | |
346 | spin_unlock_bh(&tun->lock); | |
347 | } | |
348 | ||
349 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
350 | { | |
351 | int i; | |
352 | ||
353 | spin_lock_bh(&tun->lock); | |
354 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
355 | struct tun_flow_entry *e; | |
b67bfe0d | 356 | struct hlist_node *n; |
96442e42 | 357 | |
b67bfe0d | 358 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
359 | if (e->queue_index == queue_index) |
360 | tun_flow_delete(tun, e); | |
361 | } | |
362 | } | |
363 | spin_unlock_bh(&tun->lock); | |
364 | } | |
365 | ||
366 | static void tun_flow_cleanup(unsigned long data) | |
367 | { | |
368 | struct tun_struct *tun = (struct tun_struct *)data; | |
369 | unsigned long delay = tun->ageing_time; | |
370 | unsigned long next_timer = jiffies + delay; | |
371 | unsigned long count = 0; | |
372 | int i; | |
373 | ||
374 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
375 | ||
376 | spin_lock_bh(&tun->lock); | |
377 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
378 | struct tun_flow_entry *e; | |
b67bfe0d | 379 | struct hlist_node *n; |
96442e42 | 380 | |
b67bfe0d | 381 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
382 | unsigned long this_timer; |
383 | count++; | |
384 | this_timer = e->updated + delay; | |
385 | if (time_before_eq(this_timer, jiffies)) | |
386 | tun_flow_delete(tun, e); | |
387 | else if (time_before(this_timer, next_timer)) | |
388 | next_timer = this_timer; | |
389 | } | |
390 | } | |
391 | ||
392 | if (count) | |
393 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
394 | spin_unlock_bh(&tun->lock); | |
395 | } | |
396 | ||
49974420 | 397 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 398 | struct tun_file *tfile) |
96442e42 JW |
399 | { |
400 | struct hlist_head *head; | |
401 | struct tun_flow_entry *e; | |
402 | unsigned long delay = tun->ageing_time; | |
9e85722d | 403 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
404 | |
405 | if (!rxhash) | |
406 | return; | |
407 | else | |
408 | head = &tun->flows[tun_hashfn(rxhash)]; | |
409 | ||
410 | rcu_read_lock(); | |
411 | ||
9e85722d JW |
412 | /* We may get a very small possibility of OOO during switching, not |
413 | * worth to optimize.*/ | |
414 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
415 | goto unlock; |
416 | ||
417 | e = tun_flow_find(head, rxhash); | |
418 | if (likely(e)) { | |
419 | /* TODO: keep queueing to old queue until it's empty? */ | |
420 | e->queue_index = queue_index; | |
421 | e->updated = jiffies; | |
9bc88939 | 422 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
423 | } else { |
424 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
425 | if (!tun_flow_find(head, rxhash) && |
426 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
427 | tun_flow_create(tun, head, rxhash, queue_index); |
428 | ||
429 | if (!timer_pending(&tun->flow_gc_timer)) | |
430 | mod_timer(&tun->flow_gc_timer, | |
431 | round_jiffies_up(jiffies + delay)); | |
432 | spin_unlock_bh(&tun->lock); | |
433 | } | |
434 | ||
435 | unlock: | |
436 | rcu_read_unlock(); | |
437 | } | |
438 | ||
9bc88939 TH |
439 | /** |
440 | * Save the hash received in the stack receive path and update the | |
441 | * flow_hash table accordingly. | |
442 | */ | |
443 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
444 | { | |
567e4b79 | 445 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 446 | e->rps_rxhash = hash; |
9bc88939 TH |
447 | } |
448 | ||
c8d68e6b | 449 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 450 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
451 | * the rxq based on the txq where the last packet of the flow comes. As |
452 | * the userspace application move between processors, we may get a | |
453 | * different rxq no. here. If we could not get rxhash, then we would | |
454 | * hope the rxq no. may help here. | |
455 | */ | |
f663dd9a | 456 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 457 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
458 | { |
459 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 460 | struct tun_flow_entry *e; |
c8d68e6b JW |
461 | u32 txq = 0; |
462 | u32 numqueues = 0; | |
463 | ||
464 | rcu_read_lock(); | |
92bb73ea | 465 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 466 | |
3958afa1 | 467 | txq = skb_get_hash(skb); |
c8d68e6b | 468 | if (txq) { |
96442e42 | 469 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 470 | if (e) { |
9bc88939 | 471 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 472 | txq = e->queue_index; |
9bc88939 | 473 | } else |
96442e42 JW |
474 | /* use multiply and shift instead of expensive divide */ |
475 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
476 | } else if (likely(skb_rx_queue_recorded(skb))) { |
477 | txq = skb_get_rx_queue(skb); | |
478 | while (unlikely(txq >= numqueues)) | |
479 | txq -= numqueues; | |
480 | } | |
481 | ||
482 | rcu_read_unlock(); | |
483 | return txq; | |
484 | } | |
485 | ||
cde8b15f JW |
486 | static inline bool tun_not_capable(struct tun_struct *tun) |
487 | { | |
488 | const struct cred *cred = current_cred(); | |
c260b772 | 489 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
490 | |
491 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
492 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 493 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
494 | } |
495 | ||
c8d68e6b JW |
496 | static void tun_set_real_num_queues(struct tun_struct *tun) |
497 | { | |
498 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
499 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
500 | } | |
501 | ||
4008e97f JW |
502 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
503 | { | |
504 | tfile->detached = tun; | |
505 | list_add_tail(&tfile->next, &tun->disabled); | |
506 | ++tun->numdisabled; | |
507 | } | |
508 | ||
d32649d1 | 509 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
510 | { |
511 | struct tun_struct *tun = tfile->detached; | |
512 | ||
513 | tfile->detached = NULL; | |
514 | list_del_init(&tfile->next); | |
515 | --tun->numdisabled; | |
516 | return tun; | |
517 | } | |
518 | ||
4bfb0513 JW |
519 | static void tun_queue_purge(struct tun_file *tfile) |
520 | { | |
1576d986 JW |
521 | struct sk_buff *skb; |
522 | ||
523 | while ((skb = skb_array_consume(&tfile->tx_array)) != NULL) | |
524 | kfree_skb(skb); | |
525 | ||
5503fcec | 526 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
527 | skb_queue_purge(&tfile->sk.sk_error_queue); |
528 | } | |
529 | ||
c8d68e6b JW |
530 | static void __tun_detach(struct tun_file *tfile, bool clean) |
531 | { | |
532 | struct tun_file *ntfile; | |
533 | struct tun_struct *tun; | |
c8d68e6b | 534 | |
b8deabd3 JW |
535 | tun = rtnl_dereference(tfile->tun); |
536 | ||
9e85722d | 537 | if (tun && !tfile->detached) { |
c8d68e6b JW |
538 | u16 index = tfile->queue_index; |
539 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
540 | |
541 | rcu_assign_pointer(tun->tfiles[index], | |
542 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 543 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
544 | ntfile->queue_index = index; |
545 | ||
546 | --tun->numqueues; | |
9e85722d | 547 | if (clean) { |
c956674b | 548 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 549 | sock_put(&tfile->sk); |
9e85722d | 550 | } else |
4008e97f | 551 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
552 | |
553 | synchronize_net(); | |
96442e42 | 554 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 555 | /* Drop read queue */ |
4bfb0513 | 556 | tun_queue_purge(tfile); |
c8d68e6b | 557 | tun_set_real_num_queues(tun); |
dd38bd85 | 558 | } else if (tfile->detached && clean) { |
4008e97f | 559 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
560 | sock_put(&tfile->sk); |
561 | } | |
c8d68e6b JW |
562 | |
563 | if (clean) { | |
af668b3c MT |
564 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
565 | netif_carrier_off(tun->dev); | |
566 | ||
40630b82 | 567 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 568 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 569 | unregister_netdevice(tun->dev); |
af668b3c | 570 | } |
1576d986 JW |
571 | if (tun) |
572 | skb_array_cleanup(&tfile->tx_array); | |
140e807d | 573 | sock_put(&tfile->sk); |
c8d68e6b JW |
574 | } |
575 | } | |
576 | ||
577 | static void tun_detach(struct tun_file *tfile, bool clean) | |
578 | { | |
579 | rtnl_lock(); | |
580 | __tun_detach(tfile, clean); | |
581 | rtnl_unlock(); | |
582 | } | |
583 | ||
584 | static void tun_detach_all(struct net_device *dev) | |
585 | { | |
586 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 587 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
588 | int i, n = tun->numqueues; |
589 | ||
590 | for (i = 0; i < n; i++) { | |
b8deabd3 | 591 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 592 | BUG_ON(!tfile); |
addf8fc4 | 593 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 594 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 595 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
596 | --tun->numqueues; |
597 | } | |
9e85722d | 598 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 599 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 600 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 601 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 602 | } |
c8d68e6b JW |
603 | BUG_ON(tun->numqueues != 0); |
604 | ||
605 | synchronize_net(); | |
606 | for (i = 0; i < n; i++) { | |
b8deabd3 | 607 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 608 | /* Drop read queue */ |
4bfb0513 | 609 | tun_queue_purge(tfile); |
c8d68e6b JW |
610 | sock_put(&tfile->sk); |
611 | } | |
4008e97f JW |
612 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
613 | tun_enable_queue(tfile); | |
4bfb0513 | 614 | tun_queue_purge(tfile); |
4008e97f JW |
615 | sock_put(&tfile->sk); |
616 | } | |
617 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 618 | |
40630b82 | 619 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 620 | module_put(THIS_MODULE); |
c8d68e6b JW |
621 | } |
622 | ||
849c9b6f | 623 | static int tun_attach(struct tun_struct *tun, struct file *file, bool skip_filter) |
a7385ba2 | 624 | { |
631ab46b | 625 | struct tun_file *tfile = file->private_data; |
1576d986 | 626 | struct net_device *dev = tun->dev; |
38231b7a | 627 | int err; |
a7385ba2 | 628 | |
5dbbaf2d PM |
629 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
630 | if (err < 0) | |
631 | goto out; | |
632 | ||
38231b7a | 633 | err = -EINVAL; |
9e85722d | 634 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
635 | goto out; |
636 | ||
637 | err = -EBUSY; | |
40630b82 | 638 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
639 | goto out; |
640 | ||
641 | err = -E2BIG; | |
4008e97f JW |
642 | if (!tfile->detached && |
643 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
644 | goto out; |
645 | ||
646 | err = 0; | |
54f968d6 | 647 | |
92d4ea6e | 648 | /* Re-attach the filter to persist device */ |
849c9b6f | 649 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
650 | lock_sock(tfile->socket.sk); |
651 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
652 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
653 | if (!err) |
654 | goto out; | |
655 | } | |
1576d986 JW |
656 | |
657 | if (!tfile->detached && | |
658 | skb_array_init(&tfile->tx_array, dev->tx_queue_len, GFP_KERNEL)) { | |
659 | err = -ENOMEM; | |
660 | goto out; | |
661 | } | |
662 | ||
c8d68e6b | 663 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 664 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
6e914fc7 | 665 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 666 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 667 | tun->numqueues++; |
a7385ba2 | 668 | |
4008e97f JW |
669 | if (tfile->detached) |
670 | tun_enable_queue(tfile); | |
671 | else | |
672 | sock_hold(&tfile->sk); | |
673 | ||
c8d68e6b | 674 | tun_set_real_num_queues(tun); |
a7385ba2 | 675 | |
c8d68e6b JW |
676 | /* device is allowed to go away first, so no need to hold extra |
677 | * refcnt. | |
678 | */ | |
679 | ||
680 | out: | |
681 | return err; | |
631ab46b EB |
682 | } |
683 | ||
684 | static struct tun_struct *__tun_get(struct tun_file *tfile) | |
685 | { | |
6e914fc7 | 686 | struct tun_struct *tun; |
c70f1829 | 687 | |
6e914fc7 JW |
688 | rcu_read_lock(); |
689 | tun = rcu_dereference(tfile->tun); | |
690 | if (tun) | |
691 | dev_hold(tun->dev); | |
692 | rcu_read_unlock(); | |
c70f1829 EB |
693 | |
694 | return tun; | |
631ab46b EB |
695 | } |
696 | ||
697 | static struct tun_struct *tun_get(struct file *file) | |
698 | { | |
699 | return __tun_get(file->private_data); | |
700 | } | |
701 | ||
702 | static void tun_put(struct tun_struct *tun) | |
703 | { | |
6e914fc7 | 704 | dev_put(tun->dev); |
631ab46b EB |
705 | } |
706 | ||
6b8a66ee | 707 | /* TAP filtering */ |
f271b2cc MK |
708 | static void addr_hash_set(u32 *mask, const u8 *addr) |
709 | { | |
710 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
711 | mask[n >> 5] |= (1 << (n & 31)); | |
712 | } | |
713 | ||
714 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
715 | { | |
716 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
717 | return mask[n >> 5] & (1 << (n & 31)); | |
718 | } | |
719 | ||
720 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
721 | { | |
722 | struct { u8 u[ETH_ALEN]; } *addr; | |
723 | struct tun_filter uf; | |
724 | int err, alen, n, nexact; | |
725 | ||
726 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
727 | return -EFAULT; | |
728 | ||
729 | if (!uf.count) { | |
730 | /* Disabled */ | |
731 | filter->count = 0; | |
732 | return 0; | |
733 | } | |
734 | ||
735 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
736 | addr = memdup_user(arg + sizeof(uf), alen); |
737 | if (IS_ERR(addr)) | |
738 | return PTR_ERR(addr); | |
f271b2cc MK |
739 | |
740 | /* The filter is updated without holding any locks. Which is | |
741 | * perfectly safe. We disable it first and in the worst | |
742 | * case we'll accept a few undesired packets. */ | |
743 | filter->count = 0; | |
744 | wmb(); | |
745 | ||
746 | /* Use first set of addresses as an exact filter */ | |
747 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
748 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
749 | ||
750 | nexact = n; | |
751 | ||
cfbf84fc AW |
752 | /* Remaining multicast addresses are hashed, |
753 | * unicast will leave the filter disabled. */ | |
f271b2cc | 754 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
755 | for (; n < uf.count; n++) { |
756 | if (!is_multicast_ether_addr(addr[n].u)) { | |
757 | err = 0; /* no filter */ | |
3b8d2a69 | 758 | goto free_addr; |
cfbf84fc | 759 | } |
f271b2cc | 760 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 761 | } |
f271b2cc MK |
762 | |
763 | /* For ALLMULTI just set the mask to all ones. | |
764 | * This overrides the mask populated above. */ | |
765 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
766 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
767 | ||
768 | /* Now enable the filter */ | |
769 | wmb(); | |
770 | filter->count = nexact; | |
771 | ||
772 | /* Return the number of exact filters */ | |
773 | err = nexact; | |
3b8d2a69 | 774 | free_addr: |
f271b2cc MK |
775 | kfree(addr); |
776 | return err; | |
777 | } | |
778 | ||
779 | /* Returns: 0 - drop, !=0 - accept */ | |
780 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
781 | { | |
782 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
783 | * at this point. */ | |
784 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
785 | int i; | |
786 | ||
787 | /* Exact match */ | |
788 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 789 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
790 | return 1; |
791 | ||
792 | /* Inexact match (multicast only) */ | |
793 | if (is_multicast_ether_addr(eh->h_dest)) | |
794 | return addr_hash_test(filter->mask, eh->h_dest); | |
795 | ||
796 | return 0; | |
797 | } | |
798 | ||
799 | /* | |
800 | * Checks whether the packet is accepted or not. | |
801 | * Returns: 0 - drop, !=0 - accept | |
802 | */ | |
803 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
804 | { | |
805 | if (!filter->count) | |
806 | return 1; | |
807 | ||
808 | return run_filter(filter, skb); | |
809 | } | |
810 | ||
1da177e4 LT |
811 | /* Network device part of the driver */ |
812 | ||
7282d491 | 813 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 814 | |
c70f1829 EB |
815 | /* Net device detach from fd. */ |
816 | static void tun_net_uninit(struct net_device *dev) | |
817 | { | |
c8d68e6b | 818 | tun_detach_all(dev); |
c70f1829 EB |
819 | } |
820 | ||
1da177e4 LT |
821 | /* Net device open. */ |
822 | static int tun_net_open(struct net_device *dev) | |
823 | { | |
c8d68e6b | 824 | netif_tx_start_all_queues(dev); |
1da177e4 LT |
825 | return 0; |
826 | } | |
827 | ||
828 | /* Net device close. */ | |
829 | static int tun_net_close(struct net_device *dev) | |
830 | { | |
c8d68e6b | 831 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
832 | return 0; |
833 | } | |
834 | ||
835 | /* Net device start xmit */ | |
424efe9c | 836 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
837 | { |
838 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 839 | int txq = skb->queue_mapping; |
6e914fc7 | 840 | struct tun_file *tfile; |
fa35864e | 841 | u32 numqueues = 0; |
1da177e4 | 842 | |
6e914fc7 | 843 | rcu_read_lock(); |
c8d68e6b | 844 | tfile = rcu_dereference(tun->tfiles[txq]); |
fa35864e | 845 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 846 | |
1da177e4 | 847 | /* Drop packet if interface is not attached */ |
fa35864e | 848 | if (txq >= numqueues) |
1da177e4 LT |
849 | goto drop; |
850 | ||
3df97ba8 JW |
851 | #ifdef CONFIG_RPS |
852 | if (numqueues == 1 && static_key_false(&rps_needed)) { | |
9bc88939 TH |
853 | /* Select queue was not called for the skbuff, so we extract the |
854 | * RPS hash and save it into the flow_table here. | |
855 | */ | |
856 | __u32 rxhash; | |
857 | ||
858 | rxhash = skb_get_hash(skb); | |
859 | if (rxhash) { | |
860 | struct tun_flow_entry *e; | |
861 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
862 | rxhash); | |
863 | if (e) | |
864 | tun_flow_save_rps_rxhash(e, rxhash); | |
865 | } | |
866 | } | |
3df97ba8 | 867 | #endif |
9bc88939 | 868 | |
6e914fc7 JW |
869 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
870 | ||
c8d68e6b JW |
871 | BUG_ON(!tfile); |
872 | ||
f271b2cc MK |
873 | /* Drop if the filter does not like it. |
874 | * This is a noop if the filter is disabled. | |
875 | * Filter can be enabled only for the TAP devices. */ | |
876 | if (!check_filter(&tun->txflt, skb)) | |
877 | goto drop; | |
878 | ||
54f968d6 JW |
879 | if (tfile->socket.sk->sk_filter && |
880 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
881 | goto drop; |
882 | ||
7bf66305 JW |
883 | if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC))) |
884 | goto drop; | |
885 | ||
7b996243 | 886 | skb_tx_timestamp(skb); |
eda29772 | 887 | |
0110d6f2 | 888 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
889 | * for indefinite time. |
890 | */ | |
0110d6f2 MT |
891 | skb_orphan(skb); |
892 | ||
f8af75f3 ED |
893 | nf_reset(skb); |
894 | ||
1576d986 JW |
895 | if (skb_array_produce(&tfile->tx_array, skb)) |
896 | goto drop; | |
1da177e4 LT |
897 | |
898 | /* Notify and wake up reader process */ | |
54f968d6 JW |
899 | if (tfile->flags & TUN_FASYNC) |
900 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 901 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
902 | |
903 | rcu_read_unlock(); | |
6ed10654 | 904 | return NETDEV_TX_OK; |
1da177e4 LT |
905 | |
906 | drop: | |
608b9977 | 907 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
149d36f7 | 908 | skb_tx_error(skb); |
1da177e4 | 909 | kfree_skb(skb); |
6e914fc7 | 910 | rcu_read_unlock(); |
baeababb | 911 | return NET_XMIT_DROP; |
1da177e4 LT |
912 | } |
913 | ||
f271b2cc | 914 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 915 | { |
f271b2cc MK |
916 | /* |
917 | * This callback is supposed to deal with mc filter in | |
918 | * _rx_ path and has nothing to do with the _tx_ path. | |
919 | * In rx path we always accept everything userspace gives us. | |
920 | */ | |
1da177e4 LT |
921 | } |
922 | ||
c8f44aff MM |
923 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
924 | netdev_features_t features) | |
88255375 MM |
925 | { |
926 | struct tun_struct *tun = netdev_priv(dev); | |
927 | ||
928 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
929 | } | |
bebd097a NH |
930 | #ifdef CONFIG_NET_POLL_CONTROLLER |
931 | static void tun_poll_controller(struct net_device *dev) | |
932 | { | |
933 | /* | |
934 | * Tun only receives frames when: | |
935 | * 1) the char device endpoint gets data from user space | |
936 | * 2) the tun socket gets a sendmsg call from user space | |
92d4ea6e | 937 | * Since both of those are synchronous operations, we are guaranteed |
bebd097a | 938 | * never to have pending data when we poll for it |
92d4ea6e | 939 | * so there is nothing to do here but return. |
bebd097a NH |
940 | * We need this though so netpoll recognizes us as an interface that |
941 | * supports polling, which enables bridge devices in virt setups to | |
942 | * still use netconsole | |
943 | */ | |
944 | return; | |
945 | } | |
946 | #endif | |
eaea34b2 PA |
947 | |
948 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
949 | { | |
950 | struct tun_struct *tun = netdev_priv(dev); | |
951 | ||
952 | if (new_hr < NET_SKB_PAD) | |
953 | new_hr = NET_SKB_PAD; | |
954 | ||
955 | tun->align = new_hr; | |
956 | } | |
957 | ||
bc1f4470 | 958 | static void |
608b9977 PA |
959 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
960 | { | |
961 | u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; | |
962 | struct tun_struct *tun = netdev_priv(dev); | |
963 | struct tun_pcpu_stats *p; | |
964 | int i; | |
965 | ||
966 | for_each_possible_cpu(i) { | |
967 | u64 rxpackets, rxbytes, txpackets, txbytes; | |
968 | unsigned int start; | |
969 | ||
970 | p = per_cpu_ptr(tun->pcpu_stats, i); | |
971 | do { | |
972 | start = u64_stats_fetch_begin(&p->syncp); | |
973 | rxpackets = p->rx_packets; | |
974 | rxbytes = p->rx_bytes; | |
975 | txpackets = p->tx_packets; | |
976 | txbytes = p->tx_bytes; | |
977 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
978 | ||
979 | stats->rx_packets += rxpackets; | |
980 | stats->rx_bytes += rxbytes; | |
981 | stats->tx_packets += txpackets; | |
982 | stats->tx_bytes += txbytes; | |
983 | ||
984 | /* u32 counters */ | |
985 | rx_dropped += p->rx_dropped; | |
986 | rx_frame_errors += p->rx_frame_errors; | |
987 | tx_dropped += p->tx_dropped; | |
988 | } | |
989 | stats->rx_dropped = rx_dropped; | |
990 | stats->rx_frame_errors = rx_frame_errors; | |
991 | stats->tx_dropped = tx_dropped; | |
608b9977 PA |
992 | } |
993 | ||
758e43b7 | 994 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 995 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
996 | .ndo_open = tun_net_open, |
997 | .ndo_stop = tun_net_close, | |
00829823 | 998 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 999 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1000 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1001 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1002 | .ndo_poll_controller = tun_poll_controller, | |
1003 | #endif | |
eaea34b2 | 1004 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1005 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1006 | }; |
1007 | ||
1008 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 1009 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1010 | .ndo_open = tun_net_open, |
1011 | .ndo_stop = tun_net_close, | |
00829823 | 1012 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1013 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1014 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1015 | .ndo_set_mac_address = eth_mac_addr, |
1016 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1017 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1018 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1019 | .ndo_poll_controller = tun_poll_controller, | |
1020 | #endif | |
5e52796a | 1021 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1022 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1023 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1024 | }; |
1025 | ||
944a1376 | 1026 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1027 | { |
1028 | int i; | |
1029 | ||
96442e42 JW |
1030 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1031 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1032 | ||
1033 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
1034 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
1035 | mod_timer(&tun->flow_gc_timer, | |
1036 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1037 | } |
1038 | ||
1039 | static void tun_flow_uninit(struct tun_struct *tun) | |
1040 | { | |
1041 | del_timer_sync(&tun->flow_gc_timer); | |
1042 | tun_flow_flush(tun); | |
96442e42 JW |
1043 | } |
1044 | ||
91572088 JW |
1045 | #define MIN_MTU 68 |
1046 | #define MAX_MTU 65535 | |
1047 | ||
1da177e4 LT |
1048 | /* Initialize net device. */ |
1049 | static void tun_net_init(struct net_device *dev) | |
1050 | { | |
1051 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1052 | |
1da177e4 | 1053 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1054 | case IFF_TUN: |
758e43b7 SH |
1055 | dev->netdev_ops = &tun_netdev_ops; |
1056 | ||
1da177e4 LT |
1057 | /* Point-to-Point TUN Device */ |
1058 | dev->hard_header_len = 0; | |
1059 | dev->addr_len = 0; | |
1060 | dev->mtu = 1500; | |
1061 | ||
1062 | /* Zero header length */ | |
6aa20a22 | 1063 | dev->type = ARPHRD_NONE; |
1da177e4 | 1064 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1065 | break; |
1066 | ||
40630b82 | 1067 | case IFF_TAP: |
7a0a9608 | 1068 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1069 | /* Ethernet TAP Device */ |
1da177e4 | 1070 | ether_setup(dev); |
550fd08c | 1071 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1072 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1073 | |
f2cedb63 | 1074 | eth_hw_addr_random(dev); |
36226a8d | 1075 | |
1da177e4 LT |
1076 | break; |
1077 | } | |
91572088 JW |
1078 | |
1079 | dev->min_mtu = MIN_MTU; | |
1080 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1081 | } |
1082 | ||
1083 | /* Character device part */ | |
1084 | ||
1085 | /* Poll */ | |
c8d68e6b | 1086 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1087 | { |
b2430de3 EB |
1088 | struct tun_file *tfile = file->private_data; |
1089 | struct tun_struct *tun = __tun_get(tfile); | |
3c8a9c63 | 1090 | struct sock *sk; |
33dccbb0 | 1091 | unsigned int mask = 0; |
1da177e4 LT |
1092 | |
1093 | if (!tun) | |
eac9e902 | 1094 | return POLLERR; |
1da177e4 | 1095 | |
54f968d6 | 1096 | sk = tfile->socket.sk; |
3c8a9c63 | 1097 | |
6b8a66ee | 1098 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 1099 | |
9e641bdc | 1100 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1101 | |
1576d986 | 1102 | if (!skb_array_empty(&tfile->tx_array)) |
1da177e4 LT |
1103 | mask |= POLLIN | POLLRDNORM; |
1104 | ||
33dccbb0 | 1105 | if (sock_writeable(sk) || |
9cd3e072 | 1106 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && |
33dccbb0 HX |
1107 | sock_writeable(sk))) |
1108 | mask |= POLLOUT | POLLWRNORM; | |
1109 | ||
c70f1829 EB |
1110 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1111 | mask = POLLERR; | |
1112 | ||
631ab46b | 1113 | tun_put(tun); |
1da177e4 LT |
1114 | return mask; |
1115 | } | |
1116 | ||
f42157cb RR |
1117 | /* prepad is the amount to reserve at front. len is length after that. |
1118 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1119 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1120 | size_t prepad, size_t len, |
1121 | size_t linear, int noblock) | |
f42157cb | 1122 | { |
54f968d6 | 1123 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1124 | struct sk_buff *skb; |
33dccbb0 | 1125 | int err; |
f42157cb RR |
1126 | |
1127 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1128 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1129 | linear = len; |
f42157cb | 1130 | |
33dccbb0 | 1131 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1132 | &err, 0); |
f42157cb | 1133 | if (!skb) |
33dccbb0 | 1134 | return ERR_PTR(err); |
f42157cb RR |
1135 | |
1136 | skb_reserve(skb, prepad); | |
1137 | skb_put(skb, linear); | |
33dccbb0 HX |
1138 | skb->data_len = len - linear; |
1139 | skb->len += len - linear; | |
f42157cb RR |
1140 | |
1141 | return skb; | |
1142 | } | |
1143 | ||
5503fcec JW |
1144 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1145 | struct sk_buff *skb, int more) | |
1146 | { | |
1147 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1148 | struct sk_buff_head process_queue; | |
1149 | u32 rx_batched = tun->rx_batched; | |
1150 | bool rcv = false; | |
1151 | ||
1152 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1153 | local_bh_disable(); | |
1154 | netif_receive_skb(skb); | |
1155 | local_bh_enable(); | |
1156 | return; | |
1157 | } | |
1158 | ||
1159 | spin_lock(&queue->lock); | |
1160 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1161 | __skb_queue_head_init(&process_queue); | |
1162 | skb_queue_splice_tail_init(queue, &process_queue); | |
1163 | rcv = true; | |
1164 | } else { | |
1165 | __skb_queue_tail(queue, skb); | |
1166 | } | |
1167 | spin_unlock(&queue->lock); | |
1168 | ||
1169 | if (rcv) { | |
1170 | struct sk_buff *nskb; | |
1171 | ||
1172 | local_bh_disable(); | |
1173 | while ((nskb = __skb_dequeue(&process_queue))) | |
1174 | netif_receive_skb(nskb); | |
1175 | netif_receive_skb(skb); | |
1176 | local_bh_enable(); | |
1177 | } | |
1178 | } | |
1179 | ||
1da177e4 | 1180 | /* Get packet from user space buffer */ |
54f968d6 | 1181 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1182 | void *msg_control, struct iov_iter *from, |
5503fcec | 1183 | int noblock, bool more) |
1da177e4 | 1184 | { |
09640e63 | 1185 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1186 | struct sk_buff *skb; |
f5ff53b4 | 1187 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1188 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1189 | struct virtio_net_hdr gso = { 0 }; |
608b9977 | 1190 | struct tun_pcpu_stats *stats; |
96f8d9ec | 1191 | int good_linear; |
0690899b MT |
1192 | int copylen; |
1193 | bool zerocopy = false; | |
1194 | int err; | |
49974420 | 1195 | u32 rxhash; |
1da177e4 | 1196 | |
1bd4978a ED |
1197 | if (!(tun->dev->flags & IFF_UP)) |
1198 | return -EIO; | |
1199 | ||
40630b82 | 1200 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1201 | if (len < sizeof(pi)) |
1da177e4 | 1202 | return -EINVAL; |
15718ea0 | 1203 | len -= sizeof(pi); |
1da177e4 | 1204 | |
cbbd26b8 | 1205 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1206 | return -EFAULT; |
1207 | } | |
1208 | ||
40630b82 | 1209 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1210 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1211 | ||
1212 | if (len < vnet_hdr_sz) | |
f43798c2 | 1213 | return -EINVAL; |
e1edab87 | 1214 | len -= vnet_hdr_sz; |
f43798c2 | 1215 | |
cbbd26b8 | 1216 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1217 | return -EFAULT; |
1218 | ||
4909122f | 1219 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1220 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1221 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1222 | |
56f0dcc5 | 1223 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1224 | return -EINVAL; |
e1edab87 | 1225 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1226 | } |
1227 | ||
40630b82 | 1228 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1229 | align += NET_IP_ALIGN; |
0eca93bc | 1230 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1231 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1232 | return -EINVAL; |
1233 | } | |
6aa20a22 | 1234 | |
96f8d9ec JW |
1235 | good_linear = SKB_MAX_HEAD(align); |
1236 | ||
88529176 | 1237 | if (msg_control) { |
f5ff53b4 AV |
1238 | struct iov_iter i = *from; |
1239 | ||
88529176 JW |
1240 | /* There are 256 bytes to be copied in skb, so there is |
1241 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1242 | * The rest of the buffer is mapped from userspace. |
1243 | */ | |
56f0dcc5 | 1244 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1245 | if (copylen > good_linear) |
1246 | copylen = good_linear; | |
3dd5c330 | 1247 | linear = copylen; |
f5ff53b4 AV |
1248 | iov_iter_advance(&i, copylen); |
1249 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1250 | zerocopy = true; |
1251 | } | |
1252 | ||
1253 | if (!zerocopy) { | |
0690899b | 1254 | copylen = len; |
56f0dcc5 | 1255 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) |
96f8d9ec JW |
1256 | linear = good_linear; |
1257 | else | |
56f0dcc5 | 1258 | linear = tun16_to_cpu(tun, gso.hdr_len); |
3dd5c330 | 1259 | } |
0690899b | 1260 | |
3dd5c330 | 1261 | skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); |
33dccbb0 HX |
1262 | if (IS_ERR(skb)) { |
1263 | if (PTR_ERR(skb) != -EAGAIN) | |
608b9977 | 1264 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
33dccbb0 | 1265 | return PTR_ERR(skb); |
1da177e4 LT |
1266 | } |
1267 | ||
0690899b | 1268 | if (zerocopy) |
f5ff53b4 | 1269 | err = zerocopy_sg_from_iter(skb, from); |
af1cc7a2 | 1270 | else |
f5ff53b4 | 1271 | err = skb_copy_datagram_from_iter(skb, 0, from, len); |
0690899b MT |
1272 | |
1273 | if (err) { | |
608b9977 | 1274 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
8f22757e | 1275 | kfree_skb(skb); |
1da177e4 | 1276 | return -EFAULT; |
8f22757e | 1277 | } |
1da177e4 | 1278 | |
3e9e40e7 | 1279 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
df10db98 PA |
1280 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
1281 | kfree_skb(skb); | |
1282 | return -EINVAL; | |
1283 | } | |
1284 | ||
1da177e4 | 1285 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1286 | case IFF_TUN: |
1287 | if (tun->flags & IFF_NO_PI) { | |
f09f7ee2 AWC |
1288 | switch (skb->data[0] & 0xf0) { |
1289 | case 0x40: | |
1290 | pi.proto = htons(ETH_P_IP); | |
1291 | break; | |
1292 | case 0x60: | |
1293 | pi.proto = htons(ETH_P_IPV6); | |
1294 | break; | |
1295 | default: | |
608b9977 | 1296 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
f09f7ee2 AWC |
1297 | kfree_skb(skb); |
1298 | return -EINVAL; | |
1299 | } | |
1300 | } | |
1301 | ||
459a98ed | 1302 | skb_reset_mac_header(skb); |
1da177e4 | 1303 | skb->protocol = pi.proto; |
4c13eb66 | 1304 | skb->dev = tun->dev; |
1da177e4 | 1305 | break; |
40630b82 | 1306 | case IFF_TAP: |
1da177e4 LT |
1307 | skb->protocol = eth_type_trans(skb, tun->dev); |
1308 | break; | |
6403eab1 | 1309 | } |
1da177e4 | 1310 | |
0690899b MT |
1311 | /* copy skb_ubuf_info for callback when skb has no error */ |
1312 | if (zerocopy) { | |
1313 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1314 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1315 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
af1cc7a2 JW |
1316 | } else if (msg_control) { |
1317 | struct ubuf_info *uarg = msg_control; | |
1318 | uarg->callback(uarg, false); | |
0690899b MT |
1319 | } |
1320 | ||
72f65107 | 1321 | skb_reset_network_header(skb); |
40893fd0 | 1322 | skb_probe_transport_header(skb, 0); |
38502af7 | 1323 | |
3958afa1 | 1324 | rxhash = skb_get_hash(skb); |
d4aea20d | 1325 | #ifndef CONFIG_4KSTACKS |
5503fcec | 1326 | tun_rx_batched(tun, tfile, skb, more); |
d4aea20d | 1327 | #else |
1da177e4 | 1328 | netif_rx_ni(skb); |
d4aea20d | 1329 | #endif |
6aa20a22 | 1330 | |
608b9977 PA |
1331 | stats = get_cpu_ptr(tun->pcpu_stats); |
1332 | u64_stats_update_begin(&stats->syncp); | |
1333 | stats->rx_packets++; | |
1334 | stats->rx_bytes += len; | |
1335 | u64_stats_update_end(&stats->syncp); | |
1336 | put_cpu_ptr(stats); | |
1da177e4 | 1337 | |
9e85722d | 1338 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1339 | return total_len; |
6aa20a22 | 1340 | } |
1da177e4 | 1341 | |
f5ff53b4 | 1342 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1343 | { |
33dccbb0 | 1344 | struct file *file = iocb->ki_filp; |
ab46d779 | 1345 | struct tun_struct *tun = tun_get(file); |
54f968d6 | 1346 | struct tun_file *tfile = file->private_data; |
631ab46b | 1347 | ssize_t result; |
1da177e4 LT |
1348 | |
1349 | if (!tun) | |
1350 | return -EBADFD; | |
1351 | ||
5503fcec JW |
1352 | result = tun_get_user(tun, tfile, NULL, from, |
1353 | file->f_flags & O_NONBLOCK, false); | |
631ab46b EB |
1354 | |
1355 | tun_put(tun); | |
1356 | return result; | |
1da177e4 LT |
1357 | } |
1358 | ||
1da177e4 | 1359 | /* Put packet to the user space buffer */ |
6f7c156c | 1360 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1361 | struct tun_file *tfile, |
6f7c156c | 1362 | struct sk_buff *skb, |
e0b46d0e | 1363 | struct iov_iter *iter) |
1da177e4 LT |
1364 | { |
1365 | struct tun_pi pi = { 0, skb->protocol }; | |
608b9977 | 1366 | struct tun_pcpu_stats *stats; |
e0b46d0e | 1367 | ssize_t total; |
8c847d25 | 1368 | int vlan_offset = 0; |
a8f9bfdf | 1369 | int vlan_hlen = 0; |
2eb783c4 | 1370 | int vnet_hdr_sz = 0; |
a8f9bfdf | 1371 | |
df8a39de | 1372 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 1373 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 1374 | |
40630b82 | 1375 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 1376 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 1377 | |
e0b46d0e HX |
1378 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
1379 | ||
40630b82 | 1380 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 1381 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
1382 | return -EINVAL; |
1383 | ||
e0b46d0e HX |
1384 | total += sizeof(pi); |
1385 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
1386 | /* Packet will be striped */ |
1387 | pi.flags |= TUN_PKT_STRIP; | |
1388 | } | |
6aa20a22 | 1389 | |
e0b46d0e | 1390 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 1391 | return -EFAULT; |
6aa20a22 | 1392 | } |
1da177e4 | 1393 | |
2eb783c4 | 1394 | if (vnet_hdr_sz) { |
9403cd7c | 1395 | struct virtio_net_hdr gso; |
34166093 | 1396 | |
e0b46d0e | 1397 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
1398 | return -EINVAL; |
1399 | ||
3e9e40e7 | 1400 | if (virtio_net_hdr_from_skb(skb, &gso, |
6391a448 | 1401 | tun_is_little_endian(tun), true)) { |
f43798c2 | 1402 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
1403 | pr_err("unexpected GSO type: " |
1404 | "0x%x, gso_size %d, hdr_len %d\n", | |
1405 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
1406 | tun16_to_cpu(tun, gso.hdr_len)); | |
1407 | print_hex_dump(KERN_ERR, "tun: ", | |
1408 | DUMP_PREFIX_NONE, | |
1409 | 16, 1, skb->head, | |
1410 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
1411 | WARN_ON_ONCE(1); | |
1412 | return -EINVAL; | |
1413 | } | |
f43798c2 | 1414 | |
e0b46d0e | 1415 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 1416 | return -EFAULT; |
8c847d25 JW |
1417 | |
1418 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
1419 | } |
1420 | ||
a8f9bfdf | 1421 | if (vlan_hlen) { |
e0b46d0e | 1422 | int ret; |
6680ec68 JW |
1423 | struct { |
1424 | __be16 h_vlan_proto; | |
1425 | __be16 h_vlan_TCI; | |
1426 | } veth; | |
1427 | ||
1428 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 1429 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
1430 | |
1431 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 1432 | |
e0b46d0e HX |
1433 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
1434 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
1435 | goto done; |
1436 | ||
e0b46d0e HX |
1437 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
1438 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
1439 | goto done; |
1440 | } | |
1da177e4 | 1441 | |
e0b46d0e | 1442 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 1443 | |
6680ec68 | 1444 | done: |
608b9977 PA |
1445 | /* caller is in process context, */ |
1446 | stats = get_cpu_ptr(tun->pcpu_stats); | |
1447 | u64_stats_update_begin(&stats->syncp); | |
1448 | stats->tx_packets++; | |
1449 | stats->tx_bytes += skb->len + vlan_hlen; | |
1450 | u64_stats_update_end(&stats->syncp); | |
1451 | put_cpu_ptr(tun->pcpu_stats); | |
1da177e4 LT |
1452 | |
1453 | return total; | |
1454 | } | |
1455 | ||
1576d986 JW |
1456 | static struct sk_buff *tun_ring_recv(struct tun_file *tfile, int noblock, |
1457 | int *err) | |
1458 | { | |
1459 | DECLARE_WAITQUEUE(wait, current); | |
1460 | struct sk_buff *skb = NULL; | |
f48cc6b2 | 1461 | int error = 0; |
1576d986 JW |
1462 | |
1463 | skb = skb_array_consume(&tfile->tx_array); | |
1464 | if (skb) | |
1465 | goto out; | |
1466 | if (noblock) { | |
f48cc6b2 | 1467 | error = -EAGAIN; |
1576d986 JW |
1468 | goto out; |
1469 | } | |
1470 | ||
1471 | add_wait_queue(&tfile->wq.wait, &wait); | |
1472 | current->state = TASK_INTERRUPTIBLE; | |
1473 | ||
1474 | while (1) { | |
1475 | skb = skb_array_consume(&tfile->tx_array); | |
1476 | if (skb) | |
1477 | break; | |
1478 | if (signal_pending(current)) { | |
f48cc6b2 | 1479 | error = -ERESTARTSYS; |
1576d986 JW |
1480 | break; |
1481 | } | |
1482 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 1483 | error = -EFAULT; |
1576d986 JW |
1484 | break; |
1485 | } | |
1486 | ||
1487 | schedule(); | |
1488 | } | |
1489 | ||
1490 | current->state = TASK_RUNNING; | |
1491 | remove_wait_queue(&tfile->wq.wait, &wait); | |
1492 | ||
1493 | out: | |
f48cc6b2 | 1494 | *err = error; |
1576d986 JW |
1495 | return skb; |
1496 | } | |
1497 | ||
54f968d6 | 1498 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 AV |
1499 | struct iov_iter *to, |
1500 | int noblock) | |
1da177e4 | 1501 | { |
1da177e4 | 1502 | struct sk_buff *skb; |
9b067034 | 1503 | ssize_t ret; |
1576d986 | 1504 | int err; |
1da177e4 | 1505 | |
3872baf6 | 1506 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1507 | |
9b067034 AV |
1508 | if (!iov_iter_count(to)) |
1509 | return 0; | |
1da177e4 | 1510 | |
1576d986 JW |
1511 | /* Read frames from ring */ |
1512 | skb = tun_ring_recv(tfile, noblock, &err); | |
e0b46d0e | 1513 | if (!skb) |
957f094f | 1514 | return err; |
e0b46d0e | 1515 | |
9b067034 | 1516 | ret = tun_put_user(tun, tfile, skb, to); |
f51a5e82 | 1517 | if (unlikely(ret < 0)) |
f271b2cc | 1518 | kfree_skb(skb); |
f51a5e82 JW |
1519 | else |
1520 | consume_skb(skb); | |
1da177e4 | 1521 | |
05c2828c MT |
1522 | return ret; |
1523 | } | |
1524 | ||
9b067034 | 1525 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
1526 | { |
1527 | struct file *file = iocb->ki_filp; | |
1528 | struct tun_file *tfile = file->private_data; | |
1529 | struct tun_struct *tun = __tun_get(tfile); | |
9b067034 | 1530 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
1531 | |
1532 | if (!tun) | |
1533 | return -EBADFD; | |
9b067034 | 1534 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK); |
42404c09 | 1535 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
1536 | if (ret > 0) |
1537 | iocb->ki_pos = ret; | |
631ab46b | 1538 | tun_put(tun); |
1da177e4 LT |
1539 | return ret; |
1540 | } | |
1541 | ||
96442e42 JW |
1542 | static void tun_free_netdev(struct net_device *dev) |
1543 | { | |
1544 | struct tun_struct *tun = netdev_priv(dev); | |
1545 | ||
4008e97f | 1546 | BUG_ON(!(list_empty(&tun->disabled))); |
608b9977 | 1547 | free_percpu(tun->pcpu_stats); |
96442e42 | 1548 | tun_flow_uninit(tun); |
5dbbaf2d | 1549 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1550 | free_netdev(dev); |
1551 | } | |
1552 | ||
1da177e4 LT |
1553 | static void tun_setup(struct net_device *dev) |
1554 | { | |
1555 | struct tun_struct *tun = netdev_priv(dev); | |
1556 | ||
0625c883 EB |
1557 | tun->owner = INVALID_UID; |
1558 | tun->group = INVALID_GID; | |
1da177e4 | 1559 | |
1da177e4 | 1560 | dev->ethtool_ops = &tun_ethtool_ops; |
96442e42 | 1561 | dev->destructor = tun_free_netdev; |
016adb72 JW |
1562 | /* We prefer our own queue length */ |
1563 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
1564 | } |
1565 | ||
f019a7a5 EB |
1566 | /* Trivial set of netlink ops to allow deleting tun or tap |
1567 | * device with netlink. | |
1568 | */ | |
1569 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1570 | { | |
1571 | return -EINVAL; | |
1572 | } | |
1573 | ||
1574 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
1575 | .kind = DRV_NAME, | |
1576 | .priv_size = sizeof(struct tun_struct), | |
1577 | .setup = tun_setup, | |
1578 | .validate = tun_validate, | |
1579 | }; | |
1580 | ||
33dccbb0 HX |
1581 | static void tun_sock_write_space(struct sock *sk) |
1582 | { | |
54f968d6 | 1583 | struct tun_file *tfile; |
43815482 | 1584 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
1585 | |
1586 | if (!sock_writeable(sk)) | |
1587 | return; | |
1588 | ||
9cd3e072 | 1589 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
1590 | return; |
1591 | ||
43815482 ED |
1592 | wqueue = sk_sleep(sk); |
1593 | if (wqueue && waitqueue_active(wqueue)) | |
1594 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 1595 | POLLWRNORM | POLLWRBAND); |
c722c625 | 1596 | |
54f968d6 JW |
1597 | tfile = container_of(sk, struct tun_file, sk); |
1598 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
1599 | } |
1600 | ||
1b784140 | 1601 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 1602 | { |
54f968d6 JW |
1603 | int ret; |
1604 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1605 | struct tun_struct *tun = __tun_get(tfile); | |
1606 | ||
1607 | if (!tun) | |
1608 | return -EBADFD; | |
f5ff53b4 | 1609 | |
c0371da6 | 1610 | ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter, |
5503fcec JW |
1611 | m->msg_flags & MSG_DONTWAIT, |
1612 | m->msg_flags & MSG_MORE); | |
54f968d6 JW |
1613 | tun_put(tun); |
1614 | return ret; | |
05c2828c MT |
1615 | } |
1616 | ||
1b784140 | 1617 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
1618 | int flags) |
1619 | { | |
54f968d6 JW |
1620 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
1621 | struct tun_struct *tun = __tun_get(tfile); | |
05c2828c | 1622 | int ret; |
54f968d6 JW |
1623 | |
1624 | if (!tun) | |
1625 | return -EBADFD; | |
1626 | ||
eda29772 | 1627 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
1628 | ret = -EINVAL; |
1629 | goto out; | |
1630 | } | |
eda29772 RC |
1631 | if (flags & MSG_ERRQUEUE) { |
1632 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
1633 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
1634 | goto out; | |
1635 | } | |
c0371da6 | 1636 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT); |
87897931 | 1637 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
1638 | m->msg_flags |= MSG_TRUNC; |
1639 | ret = flags & MSG_TRUNC ? ret : total_len; | |
1640 | } | |
3811ae76 | 1641 | out: |
54f968d6 | 1642 | tun_put(tun); |
05c2828c MT |
1643 | return ret; |
1644 | } | |
1645 | ||
1576d986 JW |
1646 | static int tun_peek_len(struct socket *sock) |
1647 | { | |
1648 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
1649 | struct tun_struct *tun; | |
1650 | int ret = 0; | |
1651 | ||
1652 | tun = __tun_get(tfile); | |
1653 | if (!tun) | |
1654 | return 0; | |
1655 | ||
1656 | ret = skb_array_peek_len(&tfile->tx_array); | |
1657 | tun_put(tun); | |
1658 | ||
1659 | return ret; | |
1660 | } | |
1661 | ||
05c2828c MT |
1662 | /* Ops structure to mimic raw sockets with tun */ |
1663 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 1664 | .peek_len = tun_peek_len, |
05c2828c MT |
1665 | .sendmsg = tun_sendmsg, |
1666 | .recvmsg = tun_recvmsg, | |
1667 | }; | |
1668 | ||
33dccbb0 HX |
1669 | static struct proto tun_proto = { |
1670 | .name = "tun", | |
1671 | .owner = THIS_MODULE, | |
54f968d6 | 1672 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 1673 | }; |
f019a7a5 | 1674 | |
980c9e8c DW |
1675 | static int tun_flags(struct tun_struct *tun) |
1676 | { | |
031f5e03 | 1677 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
1678 | } |
1679 | ||
1680 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
1681 | char *buf) | |
1682 | { | |
1683 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
1684 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
1685 | } | |
1686 | ||
1687 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
1688 | char *buf) | |
1689 | { | |
1690 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1691 | return uid_valid(tun->owner)? |
1692 | sprintf(buf, "%u\n", | |
1693 | from_kuid_munged(current_user_ns(), tun->owner)): | |
1694 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1695 | } |
1696 | ||
1697 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
1698 | char *buf) | |
1699 | { | |
1700 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
1701 | return gid_valid(tun->group) ? |
1702 | sprintf(buf, "%u\n", | |
1703 | from_kgid_munged(current_user_ns(), tun->group)): | |
1704 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
1705 | } |
1706 | ||
1707 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
1708 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
1709 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
1710 | ||
c4d33e24 TI |
1711 | static struct attribute *tun_dev_attrs[] = { |
1712 | &dev_attr_tun_flags.attr, | |
1713 | &dev_attr_owner.attr, | |
1714 | &dev_attr_group.attr, | |
1715 | NULL | |
1716 | }; | |
1717 | ||
1718 | static const struct attribute_group tun_attr_group = { | |
1719 | .attrs = tun_dev_attrs | |
1720 | }; | |
1721 | ||
d647a591 | 1722 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
1723 | { |
1724 | struct tun_struct *tun; | |
54f968d6 | 1725 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
1726 | struct net_device *dev; |
1727 | int err; | |
1728 | ||
7c0c3b1a JW |
1729 | if (tfile->detached) |
1730 | return -EINVAL; | |
1731 | ||
74a3e5a7 EB |
1732 | dev = __dev_get_by_name(net, ifr->ifr_name); |
1733 | if (dev) { | |
f85ba780 DW |
1734 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
1735 | return -EBUSY; | |
74a3e5a7 EB |
1736 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
1737 | tun = netdev_priv(dev); | |
1738 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
1739 | tun = netdev_priv(dev); | |
1740 | else | |
1741 | return -EINVAL; | |
1742 | ||
8e6d91ae | 1743 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 1744 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
1745 | return -EINVAL; |
1746 | ||
cde8b15f | 1747 | if (tun_not_capable(tun)) |
2b980dbd | 1748 | return -EPERM; |
5dbbaf2d | 1749 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
1750 | if (err < 0) |
1751 | return err; | |
1752 | ||
849c9b6f | 1753 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER); |
a7385ba2 EB |
1754 | if (err < 0) |
1755 | return err; | |
4008e97f | 1756 | |
40630b82 | 1757 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
1758 | (tun->numqueues + tun->numdisabled > 1)) { |
1759 | /* One or more queue has already been attached, no need | |
1760 | * to initialize the device again. | |
1761 | */ | |
1762 | return 0; | |
1763 | } | |
6aa20a22 | 1764 | } |
1da177e4 LT |
1765 | else { |
1766 | char *name; | |
1767 | unsigned long flags = 0; | |
edfb6a14 JW |
1768 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
1769 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 1770 | |
c260b772 | 1771 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 1772 | return -EPERM; |
2b980dbd PM |
1773 | err = security_tun_dev_create(); |
1774 | if (err < 0) | |
1775 | return err; | |
ca6bb5d7 | 1776 | |
1da177e4 LT |
1777 | /* Set dev type */ |
1778 | if (ifr->ifr_flags & IFF_TUN) { | |
1779 | /* TUN device */ | |
40630b82 | 1780 | flags |= IFF_TUN; |
1da177e4 LT |
1781 | name = "tun%d"; |
1782 | } else if (ifr->ifr_flags & IFF_TAP) { | |
1783 | /* TAP device */ | |
40630b82 | 1784 | flags |= IFF_TAP; |
1da177e4 | 1785 | name = "tap%d"; |
6aa20a22 | 1786 | } else |
36989b90 | 1787 | return -EINVAL; |
6aa20a22 | 1788 | |
1da177e4 LT |
1789 | if (*ifr->ifr_name) |
1790 | name = ifr->ifr_name; | |
1791 | ||
c8d68e6b | 1792 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
1793 | NET_NAME_UNKNOWN, tun_setup, queues, |
1794 | queues); | |
edfb6a14 | 1795 | |
1da177e4 LT |
1796 | if (!dev) |
1797 | return -ENOMEM; | |
1798 | ||
fc54c658 | 1799 | dev_net_set(dev, net); |
f019a7a5 | 1800 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 1801 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 1802 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 1803 | |
1da177e4 LT |
1804 | tun = netdev_priv(dev); |
1805 | tun->dev = dev; | |
1806 | tun->flags = flags; | |
f271b2cc | 1807 | tun->txflt.count = 0; |
d9d52b51 | 1808 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 1809 | |
eaea34b2 | 1810 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
1811 | tun->filter_attached = false; |
1812 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 1813 | tun->rx_batched = 0; |
33dccbb0 | 1814 | |
608b9977 PA |
1815 | tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); |
1816 | if (!tun->pcpu_stats) { | |
1817 | err = -ENOMEM; | |
1818 | goto err_free_dev; | |
1819 | } | |
1820 | ||
96442e42 JW |
1821 | spin_lock_init(&tun->lock); |
1822 | ||
5dbbaf2d PM |
1823 | err = security_tun_dev_alloc_security(&tun->security); |
1824 | if (err < 0) | |
608b9977 | 1825 | goto err_free_stat; |
2b980dbd | 1826 | |
1da177e4 | 1827 | tun_net_init(dev); |
944a1376 | 1828 | tun_flow_init(tun); |
96442e42 | 1829 | |
88255375 | 1830 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
1831 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
1832 | NETIF_F_HW_VLAN_STAG_TX; | |
2a2bbf17 | 1833 | dev->features = dev->hw_features | NETIF_F_LLTX; |
6671b224 FLVC |
1834 | dev->vlan_features = dev->features & |
1835 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
1836 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 1837 | |
4008e97f | 1838 | INIT_LIST_HEAD(&tun->disabled); |
849c9b6f | 1839 | err = tun_attach(tun, file, false); |
eb0fb363 | 1840 | if (err < 0) |
662ca437 | 1841 | goto err_free_flow; |
eb0fb363 | 1842 | |
1da177e4 LT |
1843 | err = register_netdevice(tun->dev); |
1844 | if (err < 0) | |
662ca437 | 1845 | goto err_detach; |
1da177e4 LT |
1846 | } |
1847 | ||
af668b3c MT |
1848 | netif_carrier_on(tun->dev); |
1849 | ||
6b8a66ee | 1850 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 1851 | |
031f5e03 MT |
1852 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
1853 | (ifr->ifr_flags & TUN_FEATURES); | |
c8d68e6b | 1854 | |
e35259a9 MK |
1855 | /* Make sure persistent devices do not get stuck in |
1856 | * xoff state. | |
1857 | */ | |
1858 | if (netif_running(tun->dev)) | |
c8d68e6b | 1859 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 1860 | |
1da177e4 LT |
1861 | strcpy(ifr->ifr_name, tun->dev->name); |
1862 | return 0; | |
1863 | ||
662ca437 JW |
1864 | err_detach: |
1865 | tun_detach_all(dev); | |
1866 | err_free_flow: | |
1867 | tun_flow_uninit(tun); | |
1868 | security_tun_dev_free_security(tun->security); | |
608b9977 PA |
1869 | err_free_stat: |
1870 | free_percpu(tun->pcpu_stats); | |
662ca437 | 1871 | err_free_dev: |
1da177e4 | 1872 | free_netdev(dev); |
1da177e4 LT |
1873 | return err; |
1874 | } | |
1875 | ||
9ce99cf6 | 1876 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 1877 | struct ifreq *ifr) |
e3b99556 | 1878 | { |
6b8a66ee | 1879 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
1880 | |
1881 | strcpy(ifr->ifr_name, tun->dev->name); | |
1882 | ||
980c9e8c | 1883 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 1884 | |
e3b99556 MM |
1885 | } |
1886 | ||
5228ddc9 RR |
1887 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
1888 | * privs required. */ | |
88255375 | 1889 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 1890 | { |
c8f44aff | 1891 | netdev_features_t features = 0; |
5228ddc9 RR |
1892 | |
1893 | if (arg & TUN_F_CSUM) { | |
88255375 | 1894 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
1895 | arg &= ~TUN_F_CSUM; |
1896 | ||
1897 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
1898 | if (arg & TUN_F_TSO_ECN) { | |
1899 | features |= NETIF_F_TSO_ECN; | |
1900 | arg &= ~TUN_F_TSO_ECN; | |
1901 | } | |
1902 | if (arg & TUN_F_TSO4) | |
1903 | features |= NETIF_F_TSO; | |
1904 | if (arg & TUN_F_TSO6) | |
1905 | features |= NETIF_F_TSO6; | |
1906 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
1907 | } | |
e3e3c423 VY |
1908 | |
1909 | if (arg & TUN_F_UFO) { | |
1910 | features |= NETIF_F_UFO; | |
1911 | arg &= ~TUN_F_UFO; | |
1912 | } | |
5228ddc9 RR |
1913 | } |
1914 | ||
1915 | /* This gives the user a way to test for new features in future by | |
1916 | * trying to set them. */ | |
1917 | if (arg) | |
1918 | return -EINVAL; | |
1919 | ||
88255375 MM |
1920 | tun->set_features = features; |
1921 | netdev_update_features(tun->dev); | |
5228ddc9 RR |
1922 | |
1923 | return 0; | |
1924 | } | |
1925 | ||
c8d68e6b JW |
1926 | static void tun_detach_filter(struct tun_struct *tun, int n) |
1927 | { | |
1928 | int i; | |
1929 | struct tun_file *tfile; | |
1930 | ||
1931 | for (i = 0; i < n; i++) { | |
b8deabd3 | 1932 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
1933 | lock_sock(tfile->socket.sk); |
1934 | sk_detach_filter(tfile->socket.sk); | |
1935 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
1936 | } |
1937 | ||
1938 | tun->filter_attached = false; | |
1939 | } | |
1940 | ||
1941 | static int tun_attach_filter(struct tun_struct *tun) | |
1942 | { | |
1943 | int i, ret = 0; | |
1944 | struct tun_file *tfile; | |
1945 | ||
1946 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1947 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
1948 | lock_sock(tfile->socket.sk); |
1949 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
1950 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
1951 | if (ret) { |
1952 | tun_detach_filter(tun, i); | |
1953 | return ret; | |
1954 | } | |
1955 | } | |
1956 | ||
1957 | tun->filter_attached = true; | |
1958 | return ret; | |
1959 | } | |
1960 | ||
1961 | static void tun_set_sndbuf(struct tun_struct *tun) | |
1962 | { | |
1963 | struct tun_file *tfile; | |
1964 | int i; | |
1965 | ||
1966 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 1967 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
1968 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
1969 | } | |
1970 | } | |
1971 | ||
cde8b15f JW |
1972 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
1973 | { | |
1974 | struct tun_file *tfile = file->private_data; | |
1975 | struct tun_struct *tun; | |
cde8b15f JW |
1976 | int ret = 0; |
1977 | ||
1978 | rtnl_lock(); | |
1979 | ||
1980 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 1981 | tun = tfile->detached; |
5dbbaf2d | 1982 | if (!tun) { |
cde8b15f | 1983 | ret = -EINVAL; |
5dbbaf2d PM |
1984 | goto unlock; |
1985 | } | |
1986 | ret = security_tun_dev_attach_queue(tun->security); | |
1987 | if (ret < 0) | |
1988 | goto unlock; | |
849c9b6f | 1989 | ret = tun_attach(tun, file, false); |
4008e97f | 1990 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 1991 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 1992 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
1993 | ret = -EINVAL; |
1994 | else | |
1995 | __tun_detach(tfile, false); | |
1996 | } else | |
cde8b15f JW |
1997 | ret = -EINVAL; |
1998 | ||
5dbbaf2d | 1999 | unlock: |
cde8b15f JW |
2000 | rtnl_unlock(); |
2001 | return ret; | |
2002 | } | |
2003 | ||
50857e2a AB |
2004 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
2005 | unsigned long arg, int ifreq_len) | |
1da177e4 | 2006 | { |
36b50bab | 2007 | struct tun_file *tfile = file->private_data; |
631ab46b | 2008 | struct tun_struct *tun; |
1da177e4 LT |
2009 | void __user* argp = (void __user*)arg; |
2010 | struct ifreq ifr; | |
0625c883 EB |
2011 | kuid_t owner; |
2012 | kgid_t group; | |
33dccbb0 | 2013 | int sndbuf; |
d9d52b51 | 2014 | int vnet_hdr_sz; |
fb7589a1 | 2015 | unsigned int ifindex; |
1cf8e410 | 2016 | int le; |
f271b2cc | 2017 | int ret; |
1da177e4 | 2018 | |
20861f26 | 2019 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == SOCK_IOC_TYPE) { |
50857e2a | 2020 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 2021 | return -EFAULT; |
8bbb1813 | 2022 | } else { |
a117dacd | 2023 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 2024 | } |
631ab46b EB |
2025 | if (cmd == TUNGETFEATURES) { |
2026 | /* Currently this just means: "what IFF flags are valid?". | |
2027 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
2028 | * TUNSETIFF. |
2029 | */ | |
2030 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 2031 | (unsigned int __user*)argp); |
cde8b15f JW |
2032 | } else if (cmd == TUNSETQUEUE) |
2033 | return tun_set_queue(file, &ifr); | |
631ab46b | 2034 | |
c8d68e6b | 2035 | ret = 0; |
876bfd4d HX |
2036 | rtnl_lock(); |
2037 | ||
36b50bab | 2038 | tun = __tun_get(tfile); |
0f16bc13 GF |
2039 | if (cmd == TUNSETIFF) { |
2040 | ret = -EEXIST; | |
2041 | if (tun) | |
2042 | goto unlock; | |
2043 | ||
1da177e4 LT |
2044 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
2045 | ||
140e807d | 2046 | ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr); |
1da177e4 | 2047 | |
876bfd4d HX |
2048 | if (ret) |
2049 | goto unlock; | |
1da177e4 | 2050 | |
50857e2a | 2051 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
2052 | ret = -EFAULT; |
2053 | goto unlock; | |
1da177e4 | 2054 | } |
fb7589a1 PE |
2055 | if (cmd == TUNSETIFINDEX) { |
2056 | ret = -EPERM; | |
2057 | if (tun) | |
2058 | goto unlock; | |
2059 | ||
2060 | ret = -EFAULT; | |
2061 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
2062 | goto unlock; | |
2063 | ||
2064 | ret = 0; | |
2065 | tfile->ifindex = ifindex; | |
2066 | goto unlock; | |
2067 | } | |
1da177e4 | 2068 | |
876bfd4d | 2069 | ret = -EBADFD; |
1da177e4 | 2070 | if (!tun) |
876bfd4d | 2071 | goto unlock; |
1da177e4 | 2072 | |
1e588338 | 2073 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 2074 | |
631ab46b | 2075 | ret = 0; |
1da177e4 | 2076 | switch (cmd) { |
e3b99556 | 2077 | case TUNGETIFF: |
9ce99cf6 | 2078 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 2079 | |
3d407a80 PE |
2080 | if (tfile->detached) |
2081 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
2082 | if (!tfile->socket.sk->sk_filter) |
2083 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 2084 | |
50857e2a | 2085 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 2086 | ret = -EFAULT; |
e3b99556 MM |
2087 | break; |
2088 | ||
1da177e4 LT |
2089 | case TUNSETNOCSUM: |
2090 | /* Disable/Enable checksum */ | |
1da177e4 | 2091 | |
88255375 MM |
2092 | /* [unimplemented] */ |
2093 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 2094 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
2095 | break; |
2096 | ||
2097 | case TUNSETPERSIST: | |
54f968d6 JW |
2098 | /* Disable/Enable persist mode. Keep an extra reference to the |
2099 | * module to prevent the module being unprobed. | |
2100 | */ | |
40630b82 MT |
2101 | if (arg && !(tun->flags & IFF_PERSIST)) { |
2102 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 2103 | __module_get(THIS_MODULE); |
dd38bd85 | 2104 | } |
40630b82 MT |
2105 | if (!arg && (tun->flags & IFF_PERSIST)) { |
2106 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 JW |
2107 | module_put(THIS_MODULE); |
2108 | } | |
1da177e4 | 2109 | |
6b8a66ee JP |
2110 | tun_debug(KERN_INFO, tun, "persist %s\n", |
2111 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
2112 | break; |
2113 | ||
2114 | case TUNSETOWNER: | |
2115 | /* Set owner of the device */ | |
0625c883 EB |
2116 | owner = make_kuid(current_user_ns(), arg); |
2117 | if (!uid_valid(owner)) { | |
2118 | ret = -EINVAL; | |
2119 | break; | |
2120 | } | |
2121 | tun->owner = owner; | |
1e588338 | 2122 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 2123 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
2124 | break; |
2125 | ||
8c644623 GG |
2126 | case TUNSETGROUP: |
2127 | /* Set group of the device */ | |
0625c883 EB |
2128 | group = make_kgid(current_user_ns(), arg); |
2129 | if (!gid_valid(group)) { | |
2130 | ret = -EINVAL; | |
2131 | break; | |
2132 | } | |
2133 | tun->group = group; | |
1e588338 | 2134 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 2135 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
2136 | break; |
2137 | ||
ff4cc3ac MK |
2138 | case TUNSETLINK: |
2139 | /* Only allow setting the type when the interface is down */ | |
2140 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
2141 | tun_debug(KERN_INFO, tun, |
2142 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 2143 | ret = -EBUSY; |
ff4cc3ac MK |
2144 | } else { |
2145 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2146 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2147 | tun->dev->type); | |
48abfe05 | 2148 | ret = 0; |
ff4cc3ac | 2149 | } |
631ab46b | 2150 | break; |
ff4cc3ac | 2151 | |
1da177e4 LT |
2152 | #ifdef TUN_DEBUG |
2153 | case TUNSETDEBUG: | |
2154 | tun->debug = arg; | |
2155 | break; | |
2156 | #endif | |
5228ddc9 | 2157 | case TUNSETOFFLOAD: |
88255375 | 2158 | ret = set_offload(tun, arg); |
631ab46b | 2159 | break; |
5228ddc9 | 2160 | |
f271b2cc MK |
2161 | case TUNSETTXFILTER: |
2162 | /* Can be set only for TAPs */ | |
631ab46b | 2163 | ret = -EINVAL; |
40630b82 | 2164 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 2165 | break; |
c0e5a8c2 | 2166 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2167 | break; |
1da177e4 LT |
2168 | |
2169 | case SIOCGIFHWADDR: | |
b595076a | 2170 | /* Get hw address */ |
f271b2cc MK |
2171 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2172 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2173 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2174 | ret = -EFAULT; |
2175 | break; | |
1da177e4 LT |
2176 | |
2177 | case SIOCSIFHWADDR: | |
f271b2cc | 2178 | /* Set hw address */ |
6b8a66ee JP |
2179 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2180 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2181 | |
40102371 | 2182 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2183 | break; |
33dccbb0 HX |
2184 | |
2185 | case TUNGETSNDBUF: | |
54f968d6 | 2186 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2187 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2188 | ret = -EFAULT; | |
2189 | break; | |
2190 | ||
2191 | case TUNSETSNDBUF: | |
2192 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2193 | ret = -EFAULT; | |
2194 | break; | |
2195 | } | |
2196 | ||
c8d68e6b JW |
2197 | tun->sndbuf = sndbuf; |
2198 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2199 | break; |
2200 | ||
d9d52b51 MT |
2201 | case TUNGETVNETHDRSZ: |
2202 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2203 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2204 | ret = -EFAULT; | |
2205 | break; | |
2206 | ||
2207 | case TUNSETVNETHDRSZ: | |
2208 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2209 | ret = -EFAULT; | |
2210 | break; | |
2211 | } | |
2212 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2213 | ret = -EINVAL; | |
2214 | break; | |
2215 | } | |
2216 | ||
2217 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2218 | break; | |
2219 | ||
1cf8e410 MT |
2220 | case TUNGETVNETLE: |
2221 | le = !!(tun->flags & TUN_VNET_LE); | |
2222 | if (put_user(le, (int __user *)argp)) | |
2223 | ret = -EFAULT; | |
2224 | break; | |
2225 | ||
2226 | case TUNSETVNETLE: | |
2227 | if (get_user(le, (int __user *)argp)) { | |
2228 | ret = -EFAULT; | |
2229 | break; | |
2230 | } | |
2231 | if (le) | |
2232 | tun->flags |= TUN_VNET_LE; | |
2233 | else | |
2234 | tun->flags &= ~TUN_VNET_LE; | |
2235 | break; | |
2236 | ||
8b8e658b GK |
2237 | case TUNGETVNETBE: |
2238 | ret = tun_get_vnet_be(tun, argp); | |
2239 | break; | |
2240 | ||
2241 | case TUNSETVNETBE: | |
2242 | ret = tun_set_vnet_be(tun, argp); | |
2243 | break; | |
2244 | ||
99405162 MT |
2245 | case TUNATTACHFILTER: |
2246 | /* Can be set only for TAPs */ | |
2247 | ret = -EINVAL; | |
40630b82 | 2248 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
2249 | break; |
2250 | ret = -EFAULT; | |
54f968d6 | 2251 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2252 | break; |
2253 | ||
c8d68e6b | 2254 | ret = tun_attach_filter(tun); |
99405162 MT |
2255 | break; |
2256 | ||
2257 | case TUNDETACHFILTER: | |
2258 | /* Can be set only for TAPs */ | |
2259 | ret = -EINVAL; | |
40630b82 | 2260 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 2261 | break; |
c8d68e6b JW |
2262 | ret = 0; |
2263 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2264 | break; |
2265 | ||
76975e9c PE |
2266 | case TUNGETFILTER: |
2267 | ret = -EINVAL; | |
40630b82 | 2268 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
2269 | break; |
2270 | ret = -EFAULT; | |
2271 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2272 | break; | |
2273 | ret = 0; | |
2274 | break; | |
2275 | ||
1da177e4 | 2276 | default: |
631ab46b EB |
2277 | ret = -EINVAL; |
2278 | break; | |
ee289b64 | 2279 | } |
1da177e4 | 2280 | |
876bfd4d HX |
2281 | unlock: |
2282 | rtnl_unlock(); | |
2283 | if (tun) | |
2284 | tun_put(tun); | |
631ab46b | 2285 | return ret; |
1da177e4 LT |
2286 | } |
2287 | ||
50857e2a AB |
2288 | static long tun_chr_ioctl(struct file *file, |
2289 | unsigned int cmd, unsigned long arg) | |
2290 | { | |
2291 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2292 | } | |
2293 | ||
2294 | #ifdef CONFIG_COMPAT | |
2295 | static long tun_chr_compat_ioctl(struct file *file, | |
2296 | unsigned int cmd, unsigned long arg) | |
2297 | { | |
2298 | switch (cmd) { | |
2299 | case TUNSETIFF: | |
2300 | case TUNGETIFF: | |
2301 | case TUNSETTXFILTER: | |
2302 | case TUNGETSNDBUF: | |
2303 | case TUNSETSNDBUF: | |
2304 | case SIOCGIFHWADDR: | |
2305 | case SIOCSIFHWADDR: | |
2306 | arg = (unsigned long)compat_ptr(arg); | |
2307 | break; | |
2308 | default: | |
2309 | arg = (compat_ulong_t)arg; | |
2310 | break; | |
2311 | } | |
2312 | ||
2313 | /* | |
2314 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2315 | * the end of that structure. All fields that are used in this | |
2316 | * driver are compatible though, we don't need to convert the | |
2317 | * contents. | |
2318 | */ | |
2319 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2320 | } | |
2321 | #endif /* CONFIG_COMPAT */ | |
2322 | ||
1da177e4 LT |
2323 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2324 | { | |
54f968d6 | 2325 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2326 | int ret; |
2327 | ||
54f968d6 | 2328 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2329 | goto out; |
6aa20a22 | 2330 | |
1da177e4 | 2331 | if (on) { |
e0b93edd | 2332 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2333 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2334 | } else |
54f968d6 | 2335 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2336 | ret = 0; |
2337 | out: | |
9d319522 | 2338 | return ret; |
1da177e4 LT |
2339 | } |
2340 | ||
2341 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2342 | { | |
140e807d | 2343 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 2344 | struct tun_file *tfile; |
deed49fb | 2345 | |
6b8a66ee | 2346 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2347 | |
140e807d | 2348 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 2349 | &tun_proto, 0); |
631ab46b EB |
2350 | if (!tfile) |
2351 | return -ENOMEM; | |
c956674b | 2352 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 2353 | tfile->flags = 0; |
fb7589a1 | 2354 | tfile->ifindex = 0; |
54f968d6 | 2355 | |
54f968d6 | 2356 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2357 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2358 | |
2359 | tfile->socket.file = file; | |
2360 | tfile->socket.ops = &tun_socket_ops; | |
2361 | ||
2362 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
2363 | |
2364 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2365 | tfile->sk.sk_sndbuf = INT_MAX; | |
2366 | ||
631ab46b | 2367 | file->private_data = tfile; |
4008e97f | 2368 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2369 | |
19a6afb2 JW |
2370 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2371 | ||
1da177e4 LT |
2372 | return 0; |
2373 | } | |
2374 | ||
2375 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2376 | { | |
631ab46b | 2377 | struct tun_file *tfile = file->private_data; |
1da177e4 | 2378 | |
c8d68e6b | 2379 | tun_detach(tfile, true); |
1da177e4 LT |
2380 | |
2381 | return 0; | |
2382 | } | |
2383 | ||
93e14b6d | 2384 | #ifdef CONFIG_PROC_FS |
a3816ab0 | 2385 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *f) |
93e14b6d MY |
2386 | { |
2387 | struct tun_struct *tun; | |
2388 | struct ifreq ifr; | |
2389 | ||
2390 | memset(&ifr, 0, sizeof(ifr)); | |
2391 | ||
2392 | rtnl_lock(); | |
2393 | tun = tun_get(f); | |
2394 | if (tun) | |
2395 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2396 | rtnl_unlock(); | |
2397 | ||
2398 | if (tun) | |
2399 | tun_put(tun); | |
2400 | ||
a3816ab0 | 2401 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
2402 | } |
2403 | #endif | |
2404 | ||
d54b1fdb | 2405 | static const struct file_operations tun_fops = { |
6aa20a22 | 2406 | .owner = THIS_MODULE, |
1da177e4 | 2407 | .llseek = no_llseek, |
9b067034 | 2408 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 2409 | .write_iter = tun_chr_write_iter, |
1da177e4 | 2410 | .poll = tun_chr_poll, |
50857e2a AB |
2411 | .unlocked_ioctl = tun_chr_ioctl, |
2412 | #ifdef CONFIG_COMPAT | |
2413 | .compat_ioctl = tun_chr_compat_ioctl, | |
2414 | #endif | |
1da177e4 LT |
2415 | .open = tun_chr_open, |
2416 | .release = tun_chr_close, | |
93e14b6d MY |
2417 | .fasync = tun_chr_fasync, |
2418 | #ifdef CONFIG_PROC_FS | |
2419 | .show_fdinfo = tun_chr_show_fdinfo, | |
2420 | #endif | |
1da177e4 LT |
2421 | }; |
2422 | ||
2423 | static struct miscdevice tun_miscdev = { | |
2424 | .minor = TUN_MINOR, | |
2425 | .name = "tun", | |
e454cea2 | 2426 | .nodename = "net/tun", |
1da177e4 | 2427 | .fops = &tun_fops, |
1da177e4 LT |
2428 | }; |
2429 | ||
2430 | /* ethtool interface */ | |
2431 | ||
2432 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) | |
2433 | { | |
2434 | cmd->supported = 0; | |
2435 | cmd->advertising = 0; | |
70739497 | 2436 | ethtool_cmd_speed_set(cmd, SPEED_10); |
1da177e4 LT |
2437 | cmd->duplex = DUPLEX_FULL; |
2438 | cmd->port = PORT_TP; | |
2439 | cmd->phy_address = 0; | |
2440 | cmd->transceiver = XCVR_INTERNAL; | |
2441 | cmd->autoneg = AUTONEG_DISABLE; | |
2442 | cmd->maxtxpkt = 0; | |
2443 | cmd->maxrxpkt = 0; | |
2444 | return 0; | |
2445 | } | |
2446 | ||
2447 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2448 | { | |
2449 | struct tun_struct *tun = netdev_priv(dev); | |
2450 | ||
33a5ba14 RJ |
2451 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2452 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2453 | |
2454 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 2455 | case IFF_TUN: |
33a5ba14 | 2456 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 2457 | break; |
40630b82 | 2458 | case IFF_TAP: |
33a5ba14 | 2459 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2460 | break; |
2461 | } | |
2462 | } | |
2463 | ||
2464 | static u32 tun_get_msglevel(struct net_device *dev) | |
2465 | { | |
2466 | #ifdef TUN_DEBUG | |
2467 | struct tun_struct *tun = netdev_priv(dev); | |
2468 | return tun->debug; | |
2469 | #else | |
2470 | return -EOPNOTSUPP; | |
2471 | #endif | |
2472 | } | |
2473 | ||
2474 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2475 | { | |
2476 | #ifdef TUN_DEBUG | |
2477 | struct tun_struct *tun = netdev_priv(dev); | |
2478 | tun->debug = value; | |
2479 | #endif | |
2480 | } | |
2481 | ||
5503fcec JW |
2482 | static int tun_get_coalesce(struct net_device *dev, |
2483 | struct ethtool_coalesce *ec) | |
2484 | { | |
2485 | struct tun_struct *tun = netdev_priv(dev); | |
2486 | ||
2487 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
2488 | ||
2489 | return 0; | |
2490 | } | |
2491 | ||
2492 | static int tun_set_coalesce(struct net_device *dev, | |
2493 | struct ethtool_coalesce *ec) | |
2494 | { | |
2495 | struct tun_struct *tun = netdev_priv(dev); | |
2496 | ||
2497 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
2498 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
2499 | else | |
2500 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
2501 | ||
2502 | return 0; | |
2503 | } | |
2504 | ||
7282d491 | 2505 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2506 | .get_settings = tun_get_settings, |
2507 | .get_drvinfo = tun_get_drvinfo, | |
2508 | .get_msglevel = tun_get_msglevel, | |
2509 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2510 | .get_link = ethtool_op_get_link, |
eda29772 | 2511 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
2512 | .get_coalesce = tun_get_coalesce, |
2513 | .set_coalesce = tun_set_coalesce, | |
1da177e4 LT |
2514 | }; |
2515 | ||
1576d986 JW |
2516 | static int tun_queue_resize(struct tun_struct *tun) |
2517 | { | |
2518 | struct net_device *dev = tun->dev; | |
2519 | struct tun_file *tfile; | |
2520 | struct skb_array **arrays; | |
2521 | int n = tun->numqueues + tun->numdisabled; | |
2522 | int ret, i; | |
2523 | ||
2524 | arrays = kmalloc(sizeof *arrays * n, GFP_KERNEL); | |
2525 | if (!arrays) | |
2526 | return -ENOMEM; | |
2527 | ||
2528 | for (i = 0; i < tun->numqueues; i++) { | |
2529 | tfile = rtnl_dereference(tun->tfiles[i]); | |
2530 | arrays[i] = &tfile->tx_array; | |
2531 | } | |
2532 | list_for_each_entry(tfile, &tun->disabled, next) | |
2533 | arrays[i++] = &tfile->tx_array; | |
2534 | ||
2535 | ret = skb_array_resize_multiple(arrays, n, | |
2536 | dev->tx_queue_len, GFP_KERNEL); | |
2537 | ||
2538 | kfree(arrays); | |
2539 | return ret; | |
2540 | } | |
2541 | ||
2542 | static int tun_device_event(struct notifier_block *unused, | |
2543 | unsigned long event, void *ptr) | |
2544 | { | |
2545 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
2546 | struct tun_struct *tun = netdev_priv(dev); | |
2547 | ||
86dfb4ac CG |
2548 | if (dev->rtnl_link_ops != &tun_link_ops) |
2549 | return NOTIFY_DONE; | |
2550 | ||
1576d986 JW |
2551 | switch (event) { |
2552 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
2553 | if (tun_queue_resize(tun)) | |
2554 | return NOTIFY_BAD; | |
2555 | break; | |
2556 | default: | |
2557 | break; | |
2558 | } | |
2559 | ||
2560 | return NOTIFY_DONE; | |
2561 | } | |
2562 | ||
2563 | static struct notifier_block tun_notifier_block __read_mostly = { | |
2564 | .notifier_call = tun_device_event, | |
2565 | }; | |
79d17604 | 2566 | |
1da177e4 LT |
2567 | static int __init tun_init(void) |
2568 | { | |
2569 | int ret = 0; | |
2570 | ||
6b8a66ee JP |
2571 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
2572 | pr_info("%s\n", DRV_COPYRIGHT); | |
1da177e4 | 2573 | |
f019a7a5 | 2574 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 2575 | if (ret) { |
6b8a66ee | 2576 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 2577 | goto err_linkops; |
79d17604 PE |
2578 | } |
2579 | ||
1da177e4 | 2580 | ret = misc_register(&tun_miscdev); |
79d17604 | 2581 | if (ret) { |
6b8a66ee | 2582 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
2583 | goto err_misc; |
2584 | } | |
1576d986 JW |
2585 | |
2586 | register_netdevice_notifier(&tun_notifier_block); | |
f019a7a5 | 2587 | return 0; |
79d17604 | 2588 | err_misc: |
f019a7a5 EB |
2589 | rtnl_link_unregister(&tun_link_ops); |
2590 | err_linkops: | |
1da177e4 LT |
2591 | return ret; |
2592 | } | |
2593 | ||
2594 | static void tun_cleanup(void) | |
2595 | { | |
6aa20a22 | 2596 | misc_deregister(&tun_miscdev); |
f019a7a5 | 2597 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 2598 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
2599 | } |
2600 | ||
05c2828c MT |
2601 | /* Get an underlying socket object from tun file. Returns error unless file is |
2602 | * attached to a device. The returned object works like a packet socket, it | |
2603 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
2604 | * holding a reference to the file for as long as the socket is in use. */ | |
2605 | struct socket *tun_get_socket(struct file *file) | |
2606 | { | |
6e914fc7 | 2607 | struct tun_file *tfile; |
05c2828c MT |
2608 | if (file->f_op != &tun_fops) |
2609 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
2610 | tfile = file->private_data; |
2611 | if (!tfile) | |
05c2828c | 2612 | return ERR_PTR(-EBADFD); |
54f968d6 | 2613 | return &tfile->socket; |
05c2828c MT |
2614 | } |
2615 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
2616 | ||
1da177e4 LT |
2617 | module_init(tun_init); |
2618 | module_exit(tun_cleanup); | |
2619 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
2620 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
2621 | MODULE_LICENSE("GPL"); | |
2622 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 2623 | MODULE_ALIAS("devname:net/tun"); |