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> | |
174cd4b1 | 47 | #include <linux/sched/signal.h> |
1da177e4 LT |
48 | #include <linux/major.h> |
49 | #include <linux/slab.h> | |
50 | #include <linux/poll.h> | |
51 | #include <linux/fcntl.h> | |
52 | #include <linux/init.h> | |
53 | #include <linux/skbuff.h> | |
54 | #include <linux/netdevice.h> | |
55 | #include <linux/etherdevice.h> | |
56 | #include <linux/miscdevice.h> | |
57 | #include <linux/ethtool.h> | |
58 | #include <linux/rtnetlink.h> | |
50857e2a | 59 | #include <linux/compat.h> |
1da177e4 LT |
60 | #include <linux/if.h> |
61 | #include <linux/if_arp.h> | |
62 | #include <linux/if_ether.h> | |
63 | #include <linux/if_tun.h> | |
6680ec68 | 64 | #include <linux/if_vlan.h> |
1da177e4 | 65 | #include <linux/crc32.h> |
d647a591 | 66 | #include <linux/nsproxy.h> |
f43798c2 | 67 | #include <linux/virtio_net.h> |
99405162 | 68 | #include <linux/rcupdate.h> |
881d966b | 69 | #include <net/net_namespace.h> |
79d17604 | 70 | #include <net/netns/generic.h> |
f019a7a5 | 71 | #include <net/rtnetlink.h> |
33dccbb0 | 72 | #include <net/sock.h> |
93e14b6d | 73 | #include <linux/seq_file.h> |
e0b46d0e | 74 | #include <linux/uio.h> |
1576d986 | 75 | #include <linux/skb_array.h> |
761876c8 JW |
76 | #include <linux/bpf.h> |
77 | #include <linux/bpf_trace.h> | |
90e33d45 | 78 | #include <linux/mutex.h> |
1da177e4 | 79 | |
7c0f6ba6 | 80 | #include <linux/uaccess.h> |
1da177e4 | 81 | |
14daa021 RR |
82 | /* Uncomment to enable debugging */ |
83 | /* #define TUN_DEBUG 1 */ | |
84 | ||
1da177e4 LT |
85 | #ifdef TUN_DEBUG |
86 | static int debug; | |
14daa021 | 87 | |
6b8a66ee JP |
88 | #define tun_debug(level, tun, fmt, args...) \ |
89 | do { \ | |
90 | if (tun->debug) \ | |
91 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
92 | } while (0) | |
93 | #define DBG1(level, fmt, args...) \ | |
94 | do { \ | |
95 | if (debug == 2) \ | |
96 | printk(level fmt, ##args); \ | |
97 | } while (0) | |
14daa021 | 98 | #else |
6b8a66ee JP |
99 | #define tun_debug(level, tun, fmt, args...) \ |
100 | do { \ | |
101 | if (0) \ | |
102 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
103 | } while (0) | |
104 | #define DBG1(level, fmt, args...) \ | |
105 | do { \ | |
106 | if (0) \ | |
107 | printk(level fmt, ##args); \ | |
108 | } while (0) | |
14daa021 RR |
109 | #endif |
110 | ||
761876c8 | 111 | #define TUN_HEADROOM 256 |
7df13219 | 112 | #define TUN_RX_PAD (NET_IP_ALIGN + NET_SKB_PAD) |
66ccbc9c | 113 | |
031f5e03 MT |
114 | /* TUN device flags */ |
115 | ||
116 | /* IFF_ATTACH_QUEUE is never stored in device flags, | |
117 | * overload it to mean fasync when stored there. | |
118 | */ | |
119 | #define TUN_FASYNC IFF_ATTACH_QUEUE | |
1cf8e410 MT |
120 | /* High bits in flags field are unused. */ |
121 | #define TUN_VNET_LE 0x80000000 | |
8b8e658b | 122 | #define TUN_VNET_BE 0x40000000 |
031f5e03 MT |
123 | |
124 | #define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \ | |
90e33d45 PP |
125 | IFF_MULTI_QUEUE | IFF_NAPI | IFF_NAPI_FRAGS) |
126 | ||
0690899b MT |
127 | #define GOODCOPY_LEN 128 |
128 | ||
f271b2cc MK |
129 | #define FLT_EXACT_COUNT 8 |
130 | struct tap_filter { | |
131 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
132 | u32 mask[2]; /* Mask of the hashed addrs */ | |
133 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
134 | }; | |
135 | ||
baf71c5c PG |
136 | /* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal |
137 | * to max number of VCPUs in guest. */ | |
138 | #define MAX_TAP_QUEUES 256 | |
b8732fb7 | 139 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 140 | |
96442e42 JW |
141 | #define TUN_FLOW_EXPIRE (3 * HZ) |
142 | ||
608b9977 PA |
143 | struct tun_pcpu_stats { |
144 | u64 rx_packets; | |
145 | u64 rx_bytes; | |
146 | u64 tx_packets; | |
147 | u64 tx_bytes; | |
148 | struct u64_stats_sync syncp; | |
149 | u32 rx_dropped; | |
150 | u32 tx_dropped; | |
151 | u32 rx_frame_errors; | |
152 | }; | |
153 | ||
54f968d6 | 154 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 155 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
156 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
157 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 158 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 159 | * this). |
6e914fc7 JW |
160 | * |
161 | * RCU usage: | |
36fe8c09 | 162 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 163 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 164 | */ |
631ab46b | 165 | struct tun_file { |
54f968d6 JW |
166 | struct sock sk; |
167 | struct socket socket; | |
168 | struct socket_wq wq; | |
6e914fc7 | 169 | struct tun_struct __rcu *tun; |
54f968d6 JW |
170 | struct fasync_struct *fasync; |
171 | /* only used for fasnyc */ | |
172 | unsigned int flags; | |
fb7589a1 PE |
173 | union { |
174 | u16 queue_index; | |
175 | unsigned int ifindex; | |
176 | }; | |
94317099 | 177 | struct napi_struct napi; |
90e33d45 | 178 | struct mutex napi_mutex; /* Protects access to the above napi */ |
4008e97f JW |
179 | struct list_head next; |
180 | struct tun_struct *detached; | |
1576d986 | 181 | struct skb_array tx_array; |
631ab46b EB |
182 | }; |
183 | ||
96442e42 JW |
184 | struct tun_flow_entry { |
185 | struct hlist_node hash_link; | |
186 | struct rcu_head rcu; | |
187 | struct tun_struct *tun; | |
188 | ||
189 | u32 rxhash; | |
9bc88939 | 190 | u32 rps_rxhash; |
96442e42 JW |
191 | int queue_index; |
192 | unsigned long updated; | |
193 | }; | |
194 | ||
195 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
196 | ||
54f968d6 | 197 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 198 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
199 | * file were attached to a persist device. |
200 | */ | |
14daa021 | 201 | struct tun_struct { |
c8d68e6b JW |
202 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
203 | unsigned int numqueues; | |
f271b2cc | 204 | unsigned int flags; |
0625c883 EB |
205 | kuid_t owner; |
206 | kgid_t group; | |
14daa021 | 207 | |
14daa021 | 208 | struct net_device *dev; |
c8f44aff | 209 | netdev_features_t set_features; |
88255375 | 210 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
d591a1f3 | 211 | NETIF_F_TSO6) |
d9d52b51 | 212 | |
eaea34b2 | 213 | int align; |
d9d52b51 | 214 | int vnet_hdr_sz; |
54f968d6 JW |
215 | int sndbuf; |
216 | struct tap_filter txflt; | |
217 | struct sock_fprog fprog; | |
218 | /* protected by rtnl lock */ | |
219 | bool filter_attached; | |
14daa021 RR |
220 | #ifdef TUN_DEBUG |
221 | int debug; | |
1da177e4 | 222 | #endif |
96442e42 | 223 | spinlock_t lock; |
96442e42 JW |
224 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
225 | struct timer_list flow_gc_timer; | |
226 | unsigned long ageing_time; | |
4008e97f JW |
227 | unsigned int numdisabled; |
228 | struct list_head disabled; | |
5dbbaf2d | 229 | void *security; |
b8732fb7 | 230 | u32 flow_count; |
5503fcec | 231 | u32 rx_batched; |
608b9977 | 232 | struct tun_pcpu_stats __percpu *pcpu_stats; |
761876c8 | 233 | struct bpf_prog __rcu *xdp_prog; |
14daa021 | 234 | }; |
1da177e4 | 235 | |
94317099 PP |
236 | static int tun_napi_receive(struct napi_struct *napi, int budget) |
237 | { | |
238 | struct tun_file *tfile = container_of(napi, struct tun_file, napi); | |
239 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
240 | struct sk_buff_head process_queue; | |
241 | struct sk_buff *skb; | |
242 | int received = 0; | |
243 | ||
244 | __skb_queue_head_init(&process_queue); | |
245 | ||
246 | spin_lock(&queue->lock); | |
247 | skb_queue_splice_tail_init(queue, &process_queue); | |
248 | spin_unlock(&queue->lock); | |
249 | ||
250 | while (received < budget && (skb = __skb_dequeue(&process_queue))) { | |
251 | napi_gro_receive(napi, skb); | |
252 | ++received; | |
253 | } | |
254 | ||
255 | if (!skb_queue_empty(&process_queue)) { | |
256 | spin_lock(&queue->lock); | |
257 | skb_queue_splice(&process_queue, queue); | |
258 | spin_unlock(&queue->lock); | |
259 | } | |
260 | ||
261 | return received; | |
262 | } | |
263 | ||
264 | static int tun_napi_poll(struct napi_struct *napi, int budget) | |
265 | { | |
266 | unsigned int received; | |
267 | ||
268 | received = tun_napi_receive(napi, budget); | |
269 | ||
270 | if (received < budget) | |
271 | napi_complete_done(napi, received); | |
272 | ||
273 | return received; | |
274 | } | |
275 | ||
276 | static void tun_napi_init(struct tun_struct *tun, struct tun_file *tfile, | |
277 | bool napi_en) | |
278 | { | |
279 | if (napi_en) { | |
280 | netif_napi_add(tun->dev, &tfile->napi, tun_napi_poll, | |
281 | NAPI_POLL_WEIGHT); | |
282 | napi_enable(&tfile->napi); | |
90e33d45 | 283 | mutex_init(&tfile->napi_mutex); |
94317099 PP |
284 | } |
285 | } | |
286 | ||
287 | static void tun_napi_disable(struct tun_struct *tun, struct tun_file *tfile) | |
288 | { | |
289 | if (tun->flags & IFF_NAPI) | |
290 | napi_disable(&tfile->napi); | |
291 | } | |
292 | ||
293 | static void tun_napi_del(struct tun_struct *tun, struct tun_file *tfile) | |
294 | { | |
295 | if (tun->flags & IFF_NAPI) | |
296 | netif_napi_del(&tfile->napi); | |
297 | } | |
298 | ||
90e33d45 PP |
299 | static bool tun_napi_frags_enabled(const struct tun_struct *tun) |
300 | { | |
301 | return READ_ONCE(tun->flags) & IFF_NAPI_FRAGS; | |
302 | } | |
303 | ||
8b8e658b GK |
304 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
305 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
306 | { | |
307 | return tun->flags & TUN_VNET_BE ? false : | |
308 | virtio_legacy_is_little_endian(); | |
309 | } | |
310 | ||
311 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
312 | { | |
313 | int be = !!(tun->flags & TUN_VNET_BE); | |
314 | ||
315 | if (put_user(be, argp)) | |
316 | return -EFAULT; | |
317 | ||
318 | return 0; | |
319 | } | |
320 | ||
321 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
322 | { | |
323 | int be; | |
324 | ||
325 | if (get_user(be, argp)) | |
326 | return -EFAULT; | |
327 | ||
328 | if (be) | |
329 | tun->flags |= TUN_VNET_BE; | |
330 | else | |
331 | tun->flags &= ~TUN_VNET_BE; | |
332 | ||
333 | return 0; | |
334 | } | |
335 | #else | |
336 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
337 | { | |
338 | return virtio_legacy_is_little_endian(); | |
339 | } | |
340 | ||
341 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
342 | { | |
343 | return -EINVAL; | |
344 | } | |
345 | ||
346 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
347 | { | |
348 | return -EINVAL; | |
349 | } | |
350 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
351 | ||
25bd55bb GK |
352 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
353 | { | |
7d824109 | 354 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 355 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
356 | } |
357 | ||
56f0dcc5 MT |
358 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
359 | { | |
25bd55bb | 360 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
361 | } |
362 | ||
363 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
364 | { | |
25bd55bb | 365 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
366 | } |
367 | ||
96442e42 JW |
368 | static inline u32 tun_hashfn(u32 rxhash) |
369 | { | |
370 | return rxhash & 0x3ff; | |
371 | } | |
372 | ||
373 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
374 | { | |
375 | struct tun_flow_entry *e; | |
96442e42 | 376 | |
b67bfe0d | 377 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
378 | if (e->rxhash == rxhash) |
379 | return e; | |
380 | } | |
381 | return NULL; | |
382 | } | |
383 | ||
384 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
385 | struct hlist_head *head, | |
386 | u32 rxhash, u16 queue_index) | |
387 | { | |
9fdc6bef ED |
388 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
389 | ||
96442e42 JW |
390 | if (e) { |
391 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
392 | rxhash, queue_index); | |
393 | e->updated = jiffies; | |
394 | e->rxhash = rxhash; | |
9bc88939 | 395 | e->rps_rxhash = 0; |
96442e42 JW |
396 | e->queue_index = queue_index; |
397 | e->tun = tun; | |
398 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 399 | ++tun->flow_count; |
96442e42 JW |
400 | } |
401 | return e; | |
402 | } | |
403 | ||
96442e42 JW |
404 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
405 | { | |
406 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
407 | e->rxhash, e->queue_index); | |
408 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 409 | kfree_rcu(e, rcu); |
b8732fb7 | 410 | --tun->flow_count; |
96442e42 JW |
411 | } |
412 | ||
413 | static void tun_flow_flush(struct tun_struct *tun) | |
414 | { | |
415 | int i; | |
416 | ||
417 | spin_lock_bh(&tun->lock); | |
418 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
419 | struct tun_flow_entry *e; | |
b67bfe0d | 420 | struct hlist_node *n; |
96442e42 | 421 | |
b67bfe0d | 422 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
423 | tun_flow_delete(tun, e); |
424 | } | |
425 | spin_unlock_bh(&tun->lock); | |
426 | } | |
427 | ||
428 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
429 | { | |
430 | int i; | |
431 | ||
432 | spin_lock_bh(&tun->lock); | |
433 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
434 | struct tun_flow_entry *e; | |
b67bfe0d | 435 | struct hlist_node *n; |
96442e42 | 436 | |
b67bfe0d | 437 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
438 | if (e->queue_index == queue_index) |
439 | tun_flow_delete(tun, e); | |
440 | } | |
441 | } | |
442 | spin_unlock_bh(&tun->lock); | |
443 | } | |
444 | ||
445 | static void tun_flow_cleanup(unsigned long data) | |
446 | { | |
447 | struct tun_struct *tun = (struct tun_struct *)data; | |
448 | unsigned long delay = tun->ageing_time; | |
449 | unsigned long next_timer = jiffies + delay; | |
450 | unsigned long count = 0; | |
451 | int i; | |
452 | ||
453 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
454 | ||
455 | spin_lock_bh(&tun->lock); | |
456 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
457 | struct tun_flow_entry *e; | |
b67bfe0d | 458 | struct hlist_node *n; |
96442e42 | 459 | |
b67bfe0d | 460 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
461 | unsigned long this_timer; |
462 | count++; | |
463 | this_timer = e->updated + delay; | |
464 | if (time_before_eq(this_timer, jiffies)) | |
465 | tun_flow_delete(tun, e); | |
466 | else if (time_before(this_timer, next_timer)) | |
467 | next_timer = this_timer; | |
468 | } | |
469 | } | |
470 | ||
471 | if (count) | |
472 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
473 | spin_unlock_bh(&tun->lock); | |
474 | } | |
475 | ||
49974420 | 476 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 477 | struct tun_file *tfile) |
96442e42 JW |
478 | { |
479 | struct hlist_head *head; | |
480 | struct tun_flow_entry *e; | |
481 | unsigned long delay = tun->ageing_time; | |
9e85722d | 482 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
483 | |
484 | if (!rxhash) | |
485 | return; | |
486 | else | |
487 | head = &tun->flows[tun_hashfn(rxhash)]; | |
488 | ||
489 | rcu_read_lock(); | |
490 | ||
9e85722d JW |
491 | /* We may get a very small possibility of OOO during switching, not |
492 | * worth to optimize.*/ | |
493 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
494 | goto unlock; |
495 | ||
496 | e = tun_flow_find(head, rxhash); | |
497 | if (likely(e)) { | |
498 | /* TODO: keep queueing to old queue until it's empty? */ | |
499 | e->queue_index = queue_index; | |
500 | e->updated = jiffies; | |
9bc88939 | 501 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
502 | } else { |
503 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
504 | if (!tun_flow_find(head, rxhash) && |
505 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
506 | tun_flow_create(tun, head, rxhash, queue_index); |
507 | ||
508 | if (!timer_pending(&tun->flow_gc_timer)) | |
509 | mod_timer(&tun->flow_gc_timer, | |
510 | round_jiffies_up(jiffies + delay)); | |
511 | spin_unlock_bh(&tun->lock); | |
512 | } | |
513 | ||
514 | unlock: | |
515 | rcu_read_unlock(); | |
516 | } | |
517 | ||
9bc88939 TH |
518 | /** |
519 | * Save the hash received in the stack receive path and update the | |
520 | * flow_hash table accordingly. | |
521 | */ | |
522 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
523 | { | |
567e4b79 | 524 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 525 | e->rps_rxhash = hash; |
9bc88939 TH |
526 | } |
527 | ||
c8d68e6b | 528 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 529 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
530 | * the rxq based on the txq where the last packet of the flow comes. As |
531 | * the userspace application move between processors, we may get a | |
532 | * different rxq no. here. If we could not get rxhash, then we would | |
533 | * hope the rxq no. may help here. | |
534 | */ | |
f663dd9a | 535 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 536 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
537 | { |
538 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 539 | struct tun_flow_entry *e; |
c8d68e6b JW |
540 | u32 txq = 0; |
541 | u32 numqueues = 0; | |
542 | ||
543 | rcu_read_lock(); | |
92bb73ea | 544 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 545 | |
feec084a | 546 | txq = __skb_get_hash_symmetric(skb); |
c8d68e6b | 547 | if (txq) { |
96442e42 | 548 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 549 | if (e) { |
9bc88939 | 550 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 551 | txq = e->queue_index; |
9bc88939 | 552 | } else |
96442e42 JW |
553 | /* use multiply and shift instead of expensive divide */ |
554 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
555 | } else if (likely(skb_rx_queue_recorded(skb))) { |
556 | txq = skb_get_rx_queue(skb); | |
557 | while (unlikely(txq >= numqueues)) | |
558 | txq -= numqueues; | |
559 | } | |
560 | ||
561 | rcu_read_unlock(); | |
562 | return txq; | |
563 | } | |
564 | ||
cde8b15f JW |
565 | static inline bool tun_not_capable(struct tun_struct *tun) |
566 | { | |
567 | const struct cred *cred = current_cred(); | |
c260b772 | 568 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
569 | |
570 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
571 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 572 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
573 | } |
574 | ||
c8d68e6b JW |
575 | static void tun_set_real_num_queues(struct tun_struct *tun) |
576 | { | |
577 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
578 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
579 | } | |
580 | ||
4008e97f JW |
581 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
582 | { | |
583 | tfile->detached = tun; | |
584 | list_add_tail(&tfile->next, &tun->disabled); | |
585 | ++tun->numdisabled; | |
586 | } | |
587 | ||
d32649d1 | 588 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
589 | { |
590 | struct tun_struct *tun = tfile->detached; | |
591 | ||
592 | tfile->detached = NULL; | |
593 | list_del_init(&tfile->next); | |
594 | --tun->numdisabled; | |
595 | return tun; | |
596 | } | |
597 | ||
4bfb0513 JW |
598 | static void tun_queue_purge(struct tun_file *tfile) |
599 | { | |
1576d986 JW |
600 | struct sk_buff *skb; |
601 | ||
602 | while ((skb = skb_array_consume(&tfile->tx_array)) != NULL) | |
603 | kfree_skb(skb); | |
604 | ||
5503fcec | 605 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
606 | skb_queue_purge(&tfile->sk.sk_error_queue); |
607 | } | |
608 | ||
c8d68e6b JW |
609 | static void __tun_detach(struct tun_file *tfile, bool clean) |
610 | { | |
611 | struct tun_file *ntfile; | |
612 | struct tun_struct *tun; | |
c8d68e6b | 613 | |
b8deabd3 JW |
614 | tun = rtnl_dereference(tfile->tun); |
615 | ||
94317099 PP |
616 | if (tun && clean) { |
617 | tun_napi_disable(tun, tfile); | |
618 | tun_napi_del(tun, tfile); | |
619 | } | |
620 | ||
9e85722d | 621 | if (tun && !tfile->detached) { |
c8d68e6b JW |
622 | u16 index = tfile->queue_index; |
623 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
624 | |
625 | rcu_assign_pointer(tun->tfiles[index], | |
626 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 627 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
628 | ntfile->queue_index = index; |
629 | ||
630 | --tun->numqueues; | |
9e85722d | 631 | if (clean) { |
c956674b | 632 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 633 | sock_put(&tfile->sk); |
9e85722d | 634 | } else |
4008e97f | 635 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
636 | |
637 | synchronize_net(); | |
96442e42 | 638 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 639 | /* Drop read queue */ |
4bfb0513 | 640 | tun_queue_purge(tfile); |
c8d68e6b | 641 | tun_set_real_num_queues(tun); |
dd38bd85 | 642 | } else if (tfile->detached && clean) { |
4008e97f | 643 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
644 | sock_put(&tfile->sk); |
645 | } | |
c8d68e6b JW |
646 | |
647 | if (clean) { | |
af668b3c MT |
648 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
649 | netif_carrier_off(tun->dev); | |
650 | ||
40630b82 | 651 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 652 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 653 | unregister_netdevice(tun->dev); |
af668b3c | 654 | } |
1576d986 JW |
655 | if (tun) |
656 | skb_array_cleanup(&tfile->tx_array); | |
140e807d | 657 | sock_put(&tfile->sk); |
c8d68e6b JW |
658 | } |
659 | } | |
660 | ||
661 | static void tun_detach(struct tun_file *tfile, bool clean) | |
662 | { | |
663 | rtnl_lock(); | |
664 | __tun_detach(tfile, clean); | |
665 | rtnl_unlock(); | |
666 | } | |
667 | ||
668 | static void tun_detach_all(struct net_device *dev) | |
669 | { | |
670 | struct tun_struct *tun = netdev_priv(dev); | |
761876c8 | 671 | struct bpf_prog *xdp_prog = rtnl_dereference(tun->xdp_prog); |
4008e97f | 672 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
673 | int i, n = tun->numqueues; |
674 | ||
675 | for (i = 0; i < n; i++) { | |
b8deabd3 | 676 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 677 | BUG_ON(!tfile); |
94317099 | 678 | tun_napi_disable(tun, tfile); |
addf8fc4 | 679 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 680 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 681 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
682 | --tun->numqueues; |
683 | } | |
9e85722d | 684 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 685 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 686 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 687 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 688 | } |
c8d68e6b JW |
689 | BUG_ON(tun->numqueues != 0); |
690 | ||
691 | synchronize_net(); | |
692 | for (i = 0; i < n; i++) { | |
b8deabd3 | 693 | tfile = rtnl_dereference(tun->tfiles[i]); |
94317099 | 694 | tun_napi_del(tun, tfile); |
c8d68e6b | 695 | /* Drop read queue */ |
4bfb0513 | 696 | tun_queue_purge(tfile); |
c8d68e6b JW |
697 | sock_put(&tfile->sk); |
698 | } | |
4008e97f JW |
699 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
700 | tun_enable_queue(tfile); | |
4bfb0513 | 701 | tun_queue_purge(tfile); |
4008e97f JW |
702 | sock_put(&tfile->sk); |
703 | } | |
704 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 705 | |
761876c8 JW |
706 | if (xdp_prog) |
707 | bpf_prog_put(xdp_prog); | |
708 | ||
40630b82 | 709 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 710 | module_put(THIS_MODULE); |
c8d68e6b JW |
711 | } |
712 | ||
94317099 PP |
713 | static int tun_attach(struct tun_struct *tun, struct file *file, |
714 | bool skip_filter, bool napi) | |
a7385ba2 | 715 | { |
631ab46b | 716 | struct tun_file *tfile = file->private_data; |
1576d986 | 717 | struct net_device *dev = tun->dev; |
38231b7a | 718 | int err; |
a7385ba2 | 719 | |
5dbbaf2d PM |
720 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
721 | if (err < 0) | |
722 | goto out; | |
723 | ||
38231b7a | 724 | err = -EINVAL; |
9e85722d | 725 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
726 | goto out; |
727 | ||
728 | err = -EBUSY; | |
40630b82 | 729 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
730 | goto out; |
731 | ||
732 | err = -E2BIG; | |
4008e97f JW |
733 | if (!tfile->detached && |
734 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
735 | goto out; |
736 | ||
737 | err = 0; | |
54f968d6 | 738 | |
92d4ea6e | 739 | /* Re-attach the filter to persist device */ |
849c9b6f | 740 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
741 | lock_sock(tfile->socket.sk); |
742 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
743 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
744 | if (!err) |
745 | goto out; | |
746 | } | |
1576d986 JW |
747 | |
748 | if (!tfile->detached && | |
749 | skb_array_init(&tfile->tx_array, dev->tx_queue_len, GFP_KERNEL)) { | |
750 | err = -ENOMEM; | |
751 | goto out; | |
752 | } | |
753 | ||
c8d68e6b | 754 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 755 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
6e914fc7 | 756 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 757 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 758 | tun->numqueues++; |
a7385ba2 | 759 | |
94317099 | 760 | if (tfile->detached) { |
4008e97f | 761 | tun_enable_queue(tfile); |
94317099 | 762 | } else { |
4008e97f | 763 | sock_hold(&tfile->sk); |
94317099 PP |
764 | tun_napi_init(tun, tfile, napi); |
765 | } | |
4008e97f | 766 | |
c8d68e6b | 767 | tun_set_real_num_queues(tun); |
a7385ba2 | 768 | |
c8d68e6b JW |
769 | /* device is allowed to go away first, so no need to hold extra |
770 | * refcnt. | |
771 | */ | |
772 | ||
773 | out: | |
774 | return err; | |
631ab46b EB |
775 | } |
776 | ||
9484dc74 | 777 | static struct tun_struct *tun_get(struct tun_file *tfile) |
631ab46b | 778 | { |
6e914fc7 | 779 | struct tun_struct *tun; |
c70f1829 | 780 | |
6e914fc7 JW |
781 | rcu_read_lock(); |
782 | tun = rcu_dereference(tfile->tun); | |
783 | if (tun) | |
784 | dev_hold(tun->dev); | |
785 | rcu_read_unlock(); | |
c70f1829 EB |
786 | |
787 | return tun; | |
631ab46b EB |
788 | } |
789 | ||
631ab46b EB |
790 | static void tun_put(struct tun_struct *tun) |
791 | { | |
6e914fc7 | 792 | dev_put(tun->dev); |
631ab46b EB |
793 | } |
794 | ||
6b8a66ee | 795 | /* TAP filtering */ |
f271b2cc MK |
796 | static void addr_hash_set(u32 *mask, const u8 *addr) |
797 | { | |
798 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
799 | mask[n >> 5] |= (1 << (n & 31)); | |
800 | } | |
801 | ||
802 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
803 | { | |
804 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
805 | return mask[n >> 5] & (1 << (n & 31)); | |
806 | } | |
807 | ||
808 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
809 | { | |
810 | struct { u8 u[ETH_ALEN]; } *addr; | |
811 | struct tun_filter uf; | |
812 | int err, alen, n, nexact; | |
813 | ||
814 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
815 | return -EFAULT; | |
816 | ||
817 | if (!uf.count) { | |
818 | /* Disabled */ | |
819 | filter->count = 0; | |
820 | return 0; | |
821 | } | |
822 | ||
823 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
824 | addr = memdup_user(arg + sizeof(uf), alen); |
825 | if (IS_ERR(addr)) | |
826 | return PTR_ERR(addr); | |
f271b2cc MK |
827 | |
828 | /* The filter is updated without holding any locks. Which is | |
829 | * perfectly safe. We disable it first and in the worst | |
830 | * case we'll accept a few undesired packets. */ | |
831 | filter->count = 0; | |
832 | wmb(); | |
833 | ||
834 | /* Use first set of addresses as an exact filter */ | |
835 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
836 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
837 | ||
838 | nexact = n; | |
839 | ||
cfbf84fc AW |
840 | /* Remaining multicast addresses are hashed, |
841 | * unicast will leave the filter disabled. */ | |
f271b2cc | 842 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
843 | for (; n < uf.count; n++) { |
844 | if (!is_multicast_ether_addr(addr[n].u)) { | |
845 | err = 0; /* no filter */ | |
3b8d2a69 | 846 | goto free_addr; |
cfbf84fc | 847 | } |
f271b2cc | 848 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 849 | } |
f271b2cc MK |
850 | |
851 | /* For ALLMULTI just set the mask to all ones. | |
852 | * This overrides the mask populated above. */ | |
853 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
854 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
855 | ||
856 | /* Now enable the filter */ | |
857 | wmb(); | |
858 | filter->count = nexact; | |
859 | ||
860 | /* Return the number of exact filters */ | |
861 | err = nexact; | |
3b8d2a69 | 862 | free_addr: |
f271b2cc MK |
863 | kfree(addr); |
864 | return err; | |
865 | } | |
866 | ||
867 | /* Returns: 0 - drop, !=0 - accept */ | |
868 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
869 | { | |
870 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
871 | * at this point. */ | |
872 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
873 | int i; | |
874 | ||
875 | /* Exact match */ | |
876 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 877 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
878 | return 1; |
879 | ||
880 | /* Inexact match (multicast only) */ | |
881 | if (is_multicast_ether_addr(eh->h_dest)) | |
882 | return addr_hash_test(filter->mask, eh->h_dest); | |
883 | ||
884 | return 0; | |
885 | } | |
886 | ||
887 | /* | |
888 | * Checks whether the packet is accepted or not. | |
889 | * Returns: 0 - drop, !=0 - accept | |
890 | */ | |
891 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
892 | { | |
893 | if (!filter->count) | |
894 | return 1; | |
895 | ||
896 | return run_filter(filter, skb); | |
897 | } | |
898 | ||
1da177e4 LT |
899 | /* Network device part of the driver */ |
900 | ||
7282d491 | 901 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 902 | |
c70f1829 EB |
903 | /* Net device detach from fd. */ |
904 | static void tun_net_uninit(struct net_device *dev) | |
905 | { | |
c8d68e6b | 906 | tun_detach_all(dev); |
c70f1829 EB |
907 | } |
908 | ||
1da177e4 LT |
909 | /* Net device open. */ |
910 | static int tun_net_open(struct net_device *dev) | |
911 | { | |
b20e2d54 HFS |
912 | struct tun_struct *tun = netdev_priv(dev); |
913 | int i; | |
914 | ||
c8d68e6b | 915 | netif_tx_start_all_queues(dev); |
b20e2d54 HFS |
916 | |
917 | for (i = 0; i < tun->numqueues; i++) { | |
918 | struct tun_file *tfile; | |
919 | ||
920 | tfile = rtnl_dereference(tun->tfiles[i]); | |
921 | tfile->socket.sk->sk_write_space(tfile->socket.sk); | |
922 | } | |
923 | ||
1da177e4 LT |
924 | return 0; |
925 | } | |
926 | ||
927 | /* Net device close. */ | |
928 | static int tun_net_close(struct net_device *dev) | |
929 | { | |
c8d68e6b | 930 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
931 | return 0; |
932 | } | |
933 | ||
934 | /* Net device start xmit */ | |
424efe9c | 935 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
936 | { |
937 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 938 | int txq = skb->queue_mapping; |
6e914fc7 | 939 | struct tun_file *tfile; |
fa35864e | 940 | u32 numqueues = 0; |
1da177e4 | 941 | |
6e914fc7 | 942 | rcu_read_lock(); |
c8d68e6b | 943 | tfile = rcu_dereference(tun->tfiles[txq]); |
fa35864e | 944 | numqueues = ACCESS_ONCE(tun->numqueues); |
c8d68e6b | 945 | |
1da177e4 | 946 | /* Drop packet if interface is not attached */ |
fa35864e | 947 | if (txq >= numqueues) |
1da177e4 LT |
948 | goto drop; |
949 | ||
3df97ba8 JW |
950 | #ifdef CONFIG_RPS |
951 | if (numqueues == 1 && static_key_false(&rps_needed)) { | |
9bc88939 TH |
952 | /* Select queue was not called for the skbuff, so we extract the |
953 | * RPS hash and save it into the flow_table here. | |
954 | */ | |
955 | __u32 rxhash; | |
956 | ||
feec084a | 957 | rxhash = __skb_get_hash_symmetric(skb); |
9bc88939 TH |
958 | if (rxhash) { |
959 | struct tun_flow_entry *e; | |
960 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
961 | rxhash); | |
962 | if (e) | |
963 | tun_flow_save_rps_rxhash(e, rxhash); | |
964 | } | |
965 | } | |
3df97ba8 | 966 | #endif |
9bc88939 | 967 | |
6e914fc7 JW |
968 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
969 | ||
c8d68e6b JW |
970 | BUG_ON(!tfile); |
971 | ||
f271b2cc MK |
972 | /* Drop if the filter does not like it. |
973 | * This is a noop if the filter is disabled. | |
974 | * Filter can be enabled only for the TAP devices. */ | |
975 | if (!check_filter(&tun->txflt, skb)) | |
976 | goto drop; | |
977 | ||
54f968d6 JW |
978 | if (tfile->socket.sk->sk_filter && |
979 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
980 | goto drop; |
981 | ||
1f8b977a | 982 | if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) |
7bf66305 JW |
983 | goto drop; |
984 | ||
7b996243 | 985 | skb_tx_timestamp(skb); |
eda29772 | 986 | |
0110d6f2 | 987 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
988 | * for indefinite time. |
989 | */ | |
0110d6f2 MT |
990 | skb_orphan(skb); |
991 | ||
f8af75f3 ED |
992 | nf_reset(skb); |
993 | ||
1576d986 JW |
994 | if (skb_array_produce(&tfile->tx_array, skb)) |
995 | goto drop; | |
1da177e4 LT |
996 | |
997 | /* Notify and wake up reader process */ | |
54f968d6 JW |
998 | if (tfile->flags & TUN_FASYNC) |
999 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 1000 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
1001 | |
1002 | rcu_read_unlock(); | |
6ed10654 | 1003 | return NETDEV_TX_OK; |
1da177e4 LT |
1004 | |
1005 | drop: | |
608b9977 | 1006 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
149d36f7 | 1007 | skb_tx_error(skb); |
1da177e4 | 1008 | kfree_skb(skb); |
6e914fc7 | 1009 | rcu_read_unlock(); |
baeababb | 1010 | return NET_XMIT_DROP; |
1da177e4 LT |
1011 | } |
1012 | ||
f271b2cc | 1013 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 1014 | { |
f271b2cc MK |
1015 | /* |
1016 | * This callback is supposed to deal with mc filter in | |
1017 | * _rx_ path and has nothing to do with the _tx_ path. | |
1018 | * In rx path we always accept everything userspace gives us. | |
1019 | */ | |
1da177e4 LT |
1020 | } |
1021 | ||
c8f44aff MM |
1022 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
1023 | netdev_features_t features) | |
88255375 MM |
1024 | { |
1025 | struct tun_struct *tun = netdev_priv(dev); | |
1026 | ||
1027 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
1028 | } | |
bebd097a NH |
1029 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1030 | static void tun_poll_controller(struct net_device *dev) | |
1031 | { | |
1032 | /* | |
1033 | * Tun only receives frames when: | |
1034 | * 1) the char device endpoint gets data from user space | |
1035 | * 2) the tun socket gets a sendmsg call from user space | |
94317099 PP |
1036 | * If NAPI is not enabled, since both of those are synchronous |
1037 | * operations, we are guaranteed never to have pending data when we poll | |
1038 | * for it so there is nothing to do here but return. | |
bebd097a NH |
1039 | * We need this though so netpoll recognizes us as an interface that |
1040 | * supports polling, which enables bridge devices in virt setups to | |
1041 | * still use netconsole | |
94317099 | 1042 | * If NAPI is enabled, however, we need to schedule polling for all |
90e33d45 PP |
1043 | * queues unless we are using napi_gro_frags(), which we call in |
1044 | * process context and not in NAPI context. | |
bebd097a | 1045 | */ |
94317099 PP |
1046 | struct tun_struct *tun = netdev_priv(dev); |
1047 | ||
1048 | if (tun->flags & IFF_NAPI) { | |
1049 | struct tun_file *tfile; | |
1050 | int i; | |
1051 | ||
90e33d45 PP |
1052 | if (tun_napi_frags_enabled(tun)) |
1053 | return; | |
1054 | ||
94317099 PP |
1055 | rcu_read_lock(); |
1056 | for (i = 0; i < tun->numqueues; i++) { | |
1057 | tfile = rcu_dereference(tun->tfiles[i]); | |
1058 | napi_schedule(&tfile->napi); | |
1059 | } | |
1060 | rcu_read_unlock(); | |
1061 | } | |
bebd097a NH |
1062 | return; |
1063 | } | |
1064 | #endif | |
eaea34b2 PA |
1065 | |
1066 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
1067 | { | |
1068 | struct tun_struct *tun = netdev_priv(dev); | |
1069 | ||
1070 | if (new_hr < NET_SKB_PAD) | |
1071 | new_hr = NET_SKB_PAD; | |
1072 | ||
1073 | tun->align = new_hr; | |
1074 | } | |
1075 | ||
bc1f4470 | 1076 | static void |
608b9977 PA |
1077 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
1078 | { | |
1079 | u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; | |
1080 | struct tun_struct *tun = netdev_priv(dev); | |
1081 | struct tun_pcpu_stats *p; | |
1082 | int i; | |
1083 | ||
1084 | for_each_possible_cpu(i) { | |
1085 | u64 rxpackets, rxbytes, txpackets, txbytes; | |
1086 | unsigned int start; | |
1087 | ||
1088 | p = per_cpu_ptr(tun->pcpu_stats, i); | |
1089 | do { | |
1090 | start = u64_stats_fetch_begin(&p->syncp); | |
1091 | rxpackets = p->rx_packets; | |
1092 | rxbytes = p->rx_bytes; | |
1093 | txpackets = p->tx_packets; | |
1094 | txbytes = p->tx_bytes; | |
1095 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
1096 | ||
1097 | stats->rx_packets += rxpackets; | |
1098 | stats->rx_bytes += rxbytes; | |
1099 | stats->tx_packets += txpackets; | |
1100 | stats->tx_bytes += txbytes; | |
1101 | ||
1102 | /* u32 counters */ | |
1103 | rx_dropped += p->rx_dropped; | |
1104 | rx_frame_errors += p->rx_frame_errors; | |
1105 | tx_dropped += p->tx_dropped; | |
1106 | } | |
1107 | stats->rx_dropped = rx_dropped; | |
1108 | stats->rx_frame_errors = rx_frame_errors; | |
1109 | stats->tx_dropped = tx_dropped; | |
608b9977 PA |
1110 | } |
1111 | ||
761876c8 JW |
1112 | static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog, |
1113 | struct netlink_ext_ack *extack) | |
1114 | { | |
1115 | struct tun_struct *tun = netdev_priv(dev); | |
1116 | struct bpf_prog *old_prog; | |
1117 | ||
1118 | old_prog = rtnl_dereference(tun->xdp_prog); | |
1119 | rcu_assign_pointer(tun->xdp_prog, prog); | |
1120 | if (old_prog) | |
1121 | bpf_prog_put(old_prog); | |
1122 | ||
1123 | return 0; | |
1124 | } | |
1125 | ||
1126 | static u32 tun_xdp_query(struct net_device *dev) | |
1127 | { | |
1128 | struct tun_struct *tun = netdev_priv(dev); | |
1129 | const struct bpf_prog *xdp_prog; | |
1130 | ||
1131 | xdp_prog = rtnl_dereference(tun->xdp_prog); | |
1132 | if (xdp_prog) | |
1133 | return xdp_prog->aux->id; | |
1134 | ||
1135 | return 0; | |
1136 | } | |
1137 | ||
1138 | static int tun_xdp(struct net_device *dev, struct netdev_xdp *xdp) | |
1139 | { | |
1140 | switch (xdp->command) { | |
1141 | case XDP_SETUP_PROG: | |
1142 | return tun_xdp_set(dev, xdp->prog, xdp->extack); | |
1143 | case XDP_QUERY_PROG: | |
1144 | xdp->prog_id = tun_xdp_query(dev); | |
1145 | xdp->prog_attached = !!xdp->prog_id; | |
1146 | return 0; | |
1147 | default: | |
1148 | return -EINVAL; | |
1149 | } | |
1150 | } | |
1151 | ||
758e43b7 | 1152 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 1153 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1154 | .ndo_open = tun_net_open, |
1155 | .ndo_stop = tun_net_close, | |
00829823 | 1156 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1157 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1158 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1159 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1160 | .ndo_poll_controller = tun_poll_controller, | |
1161 | #endif | |
eaea34b2 | 1162 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1163 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1164 | }; |
1165 | ||
1166 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 1167 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1168 | .ndo_open = tun_net_open, |
1169 | .ndo_stop = tun_net_close, | |
00829823 | 1170 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1171 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1172 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1173 | .ndo_set_mac_address = eth_mac_addr, |
1174 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1175 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1176 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1177 | .ndo_poll_controller = tun_poll_controller, | |
1178 | #endif | |
5e52796a | 1179 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1180 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1181 | .ndo_get_stats64 = tun_net_get_stats64, |
761876c8 | 1182 | .ndo_xdp = tun_xdp, |
758e43b7 SH |
1183 | }; |
1184 | ||
944a1376 | 1185 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1186 | { |
1187 | int i; | |
1188 | ||
96442e42 JW |
1189 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1190 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1191 | ||
1192 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
1193 | setup_timer(&tun->flow_gc_timer, tun_flow_cleanup, (unsigned long)tun); | |
1194 | mod_timer(&tun->flow_gc_timer, | |
1195 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1196 | } |
1197 | ||
1198 | static void tun_flow_uninit(struct tun_struct *tun) | |
1199 | { | |
1200 | del_timer_sync(&tun->flow_gc_timer); | |
1201 | tun_flow_flush(tun); | |
96442e42 JW |
1202 | } |
1203 | ||
91572088 JW |
1204 | #define MIN_MTU 68 |
1205 | #define MAX_MTU 65535 | |
1206 | ||
1da177e4 LT |
1207 | /* Initialize net device. */ |
1208 | static void tun_net_init(struct net_device *dev) | |
1209 | { | |
1210 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1211 | |
1da177e4 | 1212 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1213 | case IFF_TUN: |
758e43b7 SH |
1214 | dev->netdev_ops = &tun_netdev_ops; |
1215 | ||
1da177e4 LT |
1216 | /* Point-to-Point TUN Device */ |
1217 | dev->hard_header_len = 0; | |
1218 | dev->addr_len = 0; | |
1219 | dev->mtu = 1500; | |
1220 | ||
1221 | /* Zero header length */ | |
6aa20a22 | 1222 | dev->type = ARPHRD_NONE; |
1da177e4 | 1223 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1224 | break; |
1225 | ||
40630b82 | 1226 | case IFF_TAP: |
7a0a9608 | 1227 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1228 | /* Ethernet TAP Device */ |
1da177e4 | 1229 | ether_setup(dev); |
550fd08c | 1230 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1231 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1232 | |
f2cedb63 | 1233 | eth_hw_addr_random(dev); |
36226a8d | 1234 | |
1da177e4 LT |
1235 | break; |
1236 | } | |
91572088 JW |
1237 | |
1238 | dev->min_mtu = MIN_MTU; | |
1239 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1240 | } |
1241 | ||
1242 | /* Character device part */ | |
1243 | ||
1244 | /* Poll */ | |
c8d68e6b | 1245 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1246 | { |
b2430de3 | 1247 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1248 | struct tun_struct *tun = tun_get(tfile); |
3c8a9c63 | 1249 | struct sock *sk; |
33dccbb0 | 1250 | unsigned int mask = 0; |
1da177e4 LT |
1251 | |
1252 | if (!tun) | |
eac9e902 | 1253 | return POLLERR; |
1da177e4 | 1254 | |
54f968d6 | 1255 | sk = tfile->socket.sk; |
3c8a9c63 | 1256 | |
6b8a66ee | 1257 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 1258 | |
9e641bdc | 1259 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1260 | |
1576d986 | 1261 | if (!skb_array_empty(&tfile->tx_array)) |
1da177e4 LT |
1262 | mask |= POLLIN | POLLRDNORM; |
1263 | ||
b20e2d54 HFS |
1264 | if (tun->dev->flags & IFF_UP && |
1265 | (sock_writeable(sk) || | |
1266 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
1267 | sock_writeable(sk)))) | |
33dccbb0 HX |
1268 | mask |= POLLOUT | POLLWRNORM; |
1269 | ||
c70f1829 EB |
1270 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1271 | mask = POLLERR; | |
1272 | ||
631ab46b | 1273 | tun_put(tun); |
1da177e4 LT |
1274 | return mask; |
1275 | } | |
1276 | ||
90e33d45 PP |
1277 | static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile, |
1278 | size_t len, | |
1279 | const struct iov_iter *it) | |
1280 | { | |
1281 | struct sk_buff *skb; | |
1282 | size_t linear; | |
1283 | int err; | |
1284 | int i; | |
1285 | ||
1286 | if (it->nr_segs > MAX_SKB_FRAGS + 1) | |
1287 | return ERR_PTR(-ENOMEM); | |
1288 | ||
1289 | local_bh_disable(); | |
1290 | skb = napi_get_frags(&tfile->napi); | |
1291 | local_bh_enable(); | |
1292 | if (!skb) | |
1293 | return ERR_PTR(-ENOMEM); | |
1294 | ||
1295 | linear = iov_iter_single_seg_count(it); | |
1296 | err = __skb_grow(skb, linear); | |
1297 | if (err) | |
1298 | goto free; | |
1299 | ||
1300 | skb->len = len; | |
1301 | skb->data_len = len - linear; | |
1302 | skb->truesize += skb->data_len; | |
1303 | ||
1304 | for (i = 1; i < it->nr_segs; i++) { | |
1305 | size_t fragsz = it->iov[i].iov_len; | |
1306 | unsigned long offset; | |
1307 | struct page *page; | |
1308 | void *data; | |
1309 | ||
1310 | if (fragsz == 0 || fragsz > PAGE_SIZE) { | |
1311 | err = -EINVAL; | |
1312 | goto free; | |
1313 | } | |
1314 | ||
1315 | local_bh_disable(); | |
1316 | data = napi_alloc_frag(fragsz); | |
1317 | local_bh_enable(); | |
1318 | if (!data) { | |
1319 | err = -ENOMEM; | |
1320 | goto free; | |
1321 | } | |
1322 | ||
1323 | page = virt_to_head_page(data); | |
1324 | offset = data - page_address(page); | |
1325 | skb_fill_page_desc(skb, i - 1, page, offset, fragsz); | |
1326 | } | |
1327 | ||
1328 | return skb; | |
1329 | free: | |
1330 | /* frees skb and all frags allocated with napi_alloc_frag() */ | |
1331 | napi_free_frags(&tfile->napi); | |
1332 | return ERR_PTR(err); | |
1333 | } | |
1334 | ||
f42157cb RR |
1335 | /* prepad is the amount to reserve at front. len is length after that. |
1336 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1337 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1338 | size_t prepad, size_t len, |
1339 | size_t linear, int noblock) | |
f42157cb | 1340 | { |
54f968d6 | 1341 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1342 | struct sk_buff *skb; |
33dccbb0 | 1343 | int err; |
f42157cb RR |
1344 | |
1345 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1346 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1347 | linear = len; |
f42157cb | 1348 | |
33dccbb0 | 1349 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1350 | &err, 0); |
f42157cb | 1351 | if (!skb) |
33dccbb0 | 1352 | return ERR_PTR(err); |
f42157cb RR |
1353 | |
1354 | skb_reserve(skb, prepad); | |
1355 | skb_put(skb, linear); | |
33dccbb0 HX |
1356 | skb->data_len = len - linear; |
1357 | skb->len += len - linear; | |
f42157cb RR |
1358 | |
1359 | return skb; | |
1360 | } | |
1361 | ||
5503fcec JW |
1362 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1363 | struct sk_buff *skb, int more) | |
1364 | { | |
1365 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1366 | struct sk_buff_head process_queue; | |
1367 | u32 rx_batched = tun->rx_batched; | |
1368 | bool rcv = false; | |
1369 | ||
1370 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1371 | local_bh_disable(); | |
1372 | netif_receive_skb(skb); | |
1373 | local_bh_enable(); | |
1374 | return; | |
1375 | } | |
1376 | ||
1377 | spin_lock(&queue->lock); | |
1378 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1379 | __skb_queue_head_init(&process_queue); | |
1380 | skb_queue_splice_tail_init(queue, &process_queue); | |
1381 | rcv = true; | |
1382 | } else { | |
1383 | __skb_queue_tail(queue, skb); | |
1384 | } | |
1385 | spin_unlock(&queue->lock); | |
1386 | ||
1387 | if (rcv) { | |
1388 | struct sk_buff *nskb; | |
1389 | ||
1390 | local_bh_disable(); | |
1391 | while ((nskb = __skb_dequeue(&process_queue))) | |
1392 | netif_receive_skb(nskb); | |
1393 | netif_receive_skb(skb); | |
1394 | local_bh_enable(); | |
1395 | } | |
1396 | } | |
1397 | ||
66ccbc9c JW |
1398 | static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile, |
1399 | int len, int noblock, bool zerocopy) | |
1400 | { | |
1401 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) | |
1402 | return false; | |
1403 | ||
1404 | if (tfile->socket.sk->sk_sndbuf != INT_MAX) | |
1405 | return false; | |
1406 | ||
1407 | if (!noblock) | |
1408 | return false; | |
1409 | ||
1410 | if (zerocopy) | |
1411 | return false; | |
1412 | ||
1413 | if (SKB_DATA_ALIGN(len + TUN_RX_PAD) + | |
1414 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE) | |
1415 | return false; | |
1416 | ||
1417 | return true; | |
1418 | } | |
1419 | ||
761876c8 JW |
1420 | static struct sk_buff *tun_build_skb(struct tun_struct *tun, |
1421 | struct tun_file *tfile, | |
66ccbc9c | 1422 | struct iov_iter *from, |
761876c8 | 1423 | struct virtio_net_hdr *hdr, |
1cfe6e93 | 1424 | int len, int *skb_xdp) |
66ccbc9c | 1425 | { |
0bbd7dad | 1426 | struct page_frag *alloc_frag = ¤t->task_frag; |
66ccbc9c | 1427 | struct sk_buff *skb; |
761876c8 | 1428 | struct bpf_prog *xdp_prog; |
7df13219 | 1429 | int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); |
761876c8 | 1430 | unsigned int delta = 0; |
66ccbc9c JW |
1431 | char *buf; |
1432 | size_t copied; | |
761876c8 | 1433 | bool xdp_xmit = false; |
7df13219 JW |
1434 | int err, pad = TUN_RX_PAD; |
1435 | ||
1436 | rcu_read_lock(); | |
1437 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1438 | if (xdp_prog) | |
1439 | pad += TUN_HEADROOM; | |
1440 | buflen += SKB_DATA_ALIGN(len + pad); | |
1441 | rcu_read_unlock(); | |
66ccbc9c JW |
1442 | |
1443 | if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL))) | |
1444 | return ERR_PTR(-ENOMEM); | |
1445 | ||
1446 | buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; | |
1447 | copied = copy_page_from_iter(alloc_frag->page, | |
7df13219 | 1448 | alloc_frag->offset + pad, |
66ccbc9c JW |
1449 | len, from); |
1450 | if (copied != len) | |
1451 | return ERR_PTR(-EFAULT); | |
1452 | ||
7df13219 JW |
1453 | /* There's a small window that XDP may be set after the check |
1454 | * of xdp_prog above, this should be rare and for simplicity | |
1455 | * we do XDP on skb in case the headroom is not enough. | |
1456 | */ | |
1457 | if (hdr->gso_type || !xdp_prog) | |
1cfe6e93 | 1458 | *skb_xdp = 1; |
761876c8 | 1459 | else |
1cfe6e93 | 1460 | *skb_xdp = 0; |
761876c8 JW |
1461 | |
1462 | rcu_read_lock(); | |
1463 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1cfe6e93 | 1464 | if (xdp_prog && !*skb_xdp) { |
761876c8 JW |
1465 | struct xdp_buff xdp; |
1466 | void *orig_data; | |
1467 | u32 act; | |
1468 | ||
1469 | xdp.data_hard_start = buf; | |
7df13219 | 1470 | xdp.data = buf + pad; |
de8f3a83 | 1471 | xdp_set_data_meta_invalid(&xdp); |
761876c8 JW |
1472 | xdp.data_end = xdp.data + len; |
1473 | orig_data = xdp.data; | |
1474 | act = bpf_prog_run_xdp(xdp_prog, &xdp); | |
1475 | ||
1476 | switch (act) { | |
1477 | case XDP_REDIRECT: | |
1478 | get_page(alloc_frag->page); | |
1479 | alloc_frag->offset += buflen; | |
1480 | err = xdp_do_redirect(tun->dev, &xdp, xdp_prog); | |
1481 | if (err) | |
1482 | goto err_redirect; | |
1483 | return NULL; | |
1484 | case XDP_TX: | |
1485 | xdp_xmit = true; | |
1486 | /* fall through */ | |
1487 | case XDP_PASS: | |
1488 | delta = orig_data - xdp.data; | |
1489 | break; | |
1490 | default: | |
1491 | bpf_warn_invalid_xdp_action(act); | |
1492 | /* fall through */ | |
1493 | case XDP_ABORTED: | |
1494 | trace_xdp_exception(tun->dev, xdp_prog, act); | |
1495 | /* fall through */ | |
1496 | case XDP_DROP: | |
1497 | goto err_xdp; | |
1498 | } | |
1499 | } | |
1500 | ||
66ccbc9c | 1501 | skb = build_skb(buf, buflen); |
761876c8 JW |
1502 | if (!skb) { |
1503 | rcu_read_unlock(); | |
66ccbc9c | 1504 | return ERR_PTR(-ENOMEM); |
761876c8 | 1505 | } |
66ccbc9c | 1506 | |
7df13219 | 1507 | skb_reserve(skb, pad - delta); |
761876c8 | 1508 | skb_put(skb, len + delta); |
66ccbc9c JW |
1509 | get_page(alloc_frag->page); |
1510 | alloc_frag->offset += buflen; | |
1511 | ||
761876c8 JW |
1512 | if (xdp_xmit) { |
1513 | skb->dev = tun->dev; | |
1514 | generic_xdp_tx(skb, xdp_prog); | |
1515 | rcu_read_lock(); | |
1516 | return NULL; | |
1517 | } | |
1518 | ||
1519 | rcu_read_unlock(); | |
1520 | ||
66ccbc9c | 1521 | return skb; |
761876c8 JW |
1522 | |
1523 | err_redirect: | |
1524 | put_page(alloc_frag->page); | |
1525 | err_xdp: | |
1526 | rcu_read_unlock(); | |
1527 | this_cpu_inc(tun->pcpu_stats->rx_dropped); | |
1528 | return NULL; | |
66ccbc9c JW |
1529 | } |
1530 | ||
1da177e4 | 1531 | /* Get packet from user space buffer */ |
54f968d6 | 1532 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1533 | void *msg_control, struct iov_iter *from, |
5503fcec | 1534 | int noblock, bool more) |
1da177e4 | 1535 | { |
09640e63 | 1536 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1537 | struct sk_buff *skb; |
f5ff53b4 | 1538 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1539 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1540 | struct virtio_net_hdr gso = { 0 }; |
608b9977 | 1541 | struct tun_pcpu_stats *stats; |
96f8d9ec | 1542 | int good_linear; |
0690899b MT |
1543 | int copylen; |
1544 | bool zerocopy = false; | |
1545 | int err; | |
49974420 | 1546 | u32 rxhash; |
1cfe6e93 | 1547 | int skb_xdp = 1; |
90e33d45 | 1548 | bool frags = tun_napi_frags_enabled(tun); |
1da177e4 | 1549 | |
1bd4978a ED |
1550 | if (!(tun->dev->flags & IFF_UP)) |
1551 | return -EIO; | |
1552 | ||
40630b82 | 1553 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1554 | if (len < sizeof(pi)) |
1da177e4 | 1555 | return -EINVAL; |
15718ea0 | 1556 | len -= sizeof(pi); |
1da177e4 | 1557 | |
cbbd26b8 | 1558 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1559 | return -EFAULT; |
1560 | } | |
1561 | ||
40630b82 | 1562 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1563 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1564 | ||
1565 | if (len < vnet_hdr_sz) | |
f43798c2 | 1566 | return -EINVAL; |
e1edab87 | 1567 | len -= vnet_hdr_sz; |
f43798c2 | 1568 | |
cbbd26b8 | 1569 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1570 | return -EFAULT; |
1571 | ||
4909122f | 1572 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1573 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1574 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1575 | |
56f0dcc5 | 1576 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1577 | return -EINVAL; |
e1edab87 | 1578 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1579 | } |
1580 | ||
40630b82 | 1581 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1582 | align += NET_IP_ALIGN; |
0eca93bc | 1583 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1584 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1585 | return -EINVAL; |
1586 | } | |
6aa20a22 | 1587 | |
96f8d9ec JW |
1588 | good_linear = SKB_MAX_HEAD(align); |
1589 | ||
88529176 | 1590 | if (msg_control) { |
f5ff53b4 AV |
1591 | struct iov_iter i = *from; |
1592 | ||
88529176 JW |
1593 | /* There are 256 bytes to be copied in skb, so there is |
1594 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1595 | * The rest of the buffer is mapped from userspace. |
1596 | */ | |
56f0dcc5 | 1597 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1598 | if (copylen > good_linear) |
1599 | copylen = good_linear; | |
3dd5c330 | 1600 | linear = copylen; |
f5ff53b4 AV |
1601 | iov_iter_advance(&i, copylen); |
1602 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1603 | zerocopy = true; |
1604 | } | |
1605 | ||
90e33d45 | 1606 | if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) { |
1cfe6e93 JW |
1607 | /* For the packet that is not easy to be processed |
1608 | * (e.g gso or jumbo packet), we will do it at after | |
1609 | * skb was created with generic XDP routine. | |
1610 | */ | |
1611 | skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp); | |
66ccbc9c | 1612 | if (IS_ERR(skb)) { |
608b9977 | 1613 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
66ccbc9c JW |
1614 | return PTR_ERR(skb); |
1615 | } | |
761876c8 JW |
1616 | if (!skb) |
1617 | return total_len; | |
66ccbc9c JW |
1618 | } else { |
1619 | if (!zerocopy) { | |
1620 | copylen = len; | |
1621 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) | |
1622 | linear = good_linear; | |
1623 | else | |
1624 | linear = tun16_to_cpu(tun, gso.hdr_len); | |
1625 | } | |
1da177e4 | 1626 | |
90e33d45 PP |
1627 | if (frags) { |
1628 | mutex_lock(&tfile->napi_mutex); | |
1629 | skb = tun_napi_alloc_frags(tfile, copylen, from); | |
1630 | /* tun_napi_alloc_frags() enforces a layout for the skb. | |
1631 | * If zerocopy is enabled, then this layout will be | |
1632 | * overwritten by zerocopy_sg_from_iter(). | |
1633 | */ | |
1634 | zerocopy = false; | |
1635 | } else { | |
1636 | skb = tun_alloc_skb(tfile, align, copylen, linear, | |
1637 | noblock); | |
1638 | } | |
1639 | ||
66ccbc9c JW |
1640 | if (IS_ERR(skb)) { |
1641 | if (PTR_ERR(skb) != -EAGAIN) | |
1642 | this_cpu_inc(tun->pcpu_stats->rx_dropped); | |
90e33d45 PP |
1643 | if (frags) |
1644 | mutex_unlock(&tfile->napi_mutex); | |
66ccbc9c JW |
1645 | return PTR_ERR(skb); |
1646 | } | |
0690899b | 1647 | |
66ccbc9c JW |
1648 | if (zerocopy) |
1649 | err = zerocopy_sg_from_iter(skb, from); | |
1650 | else | |
1651 | err = skb_copy_datagram_from_iter(skb, 0, from, len); | |
1652 | ||
1653 | if (err) { | |
1654 | this_cpu_inc(tun->pcpu_stats->rx_dropped); | |
1655 | kfree_skb(skb); | |
90e33d45 PP |
1656 | if (frags) { |
1657 | tfile->napi.skb = NULL; | |
1658 | mutex_unlock(&tfile->napi_mutex); | |
1659 | } | |
1660 | ||
66ccbc9c JW |
1661 | return -EFAULT; |
1662 | } | |
8f22757e | 1663 | } |
1da177e4 | 1664 | |
3e9e40e7 | 1665 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
df10db98 PA |
1666 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
1667 | kfree_skb(skb); | |
90e33d45 PP |
1668 | if (frags) { |
1669 | tfile->napi.skb = NULL; | |
1670 | mutex_unlock(&tfile->napi_mutex); | |
1671 | } | |
1672 | ||
df10db98 PA |
1673 | return -EINVAL; |
1674 | } | |
1675 | ||
1da177e4 | 1676 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1677 | case IFF_TUN: |
1678 | if (tun->flags & IFF_NO_PI) { | |
2580c4c1 AP |
1679 | u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0; |
1680 | ||
1681 | switch (ip_version) { | |
1682 | case 4: | |
f09f7ee2 AWC |
1683 | pi.proto = htons(ETH_P_IP); |
1684 | break; | |
2580c4c1 | 1685 | case 6: |
f09f7ee2 AWC |
1686 | pi.proto = htons(ETH_P_IPV6); |
1687 | break; | |
1688 | default: | |
608b9977 | 1689 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
f09f7ee2 AWC |
1690 | kfree_skb(skb); |
1691 | return -EINVAL; | |
1692 | } | |
1693 | } | |
1694 | ||
459a98ed | 1695 | skb_reset_mac_header(skb); |
1da177e4 | 1696 | skb->protocol = pi.proto; |
4c13eb66 | 1697 | skb->dev = tun->dev; |
1da177e4 | 1698 | break; |
40630b82 | 1699 | case IFF_TAP: |
90e33d45 PP |
1700 | if (!frags) |
1701 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1da177e4 | 1702 | break; |
6403eab1 | 1703 | } |
1da177e4 | 1704 | |
0690899b MT |
1705 | /* copy skb_ubuf_info for callback when skb has no error */ |
1706 | if (zerocopy) { | |
1707 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1708 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1709 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
af1cc7a2 JW |
1710 | } else if (msg_control) { |
1711 | struct ubuf_info *uarg = msg_control; | |
1712 | uarg->callback(uarg, false); | |
0690899b MT |
1713 | } |
1714 | ||
72f65107 | 1715 | skb_reset_network_header(skb); |
40893fd0 | 1716 | skb_probe_transport_header(skb, 0); |
38502af7 | 1717 | |
1cfe6e93 | 1718 | if (skb_xdp) { |
761876c8 JW |
1719 | struct bpf_prog *xdp_prog; |
1720 | int ret; | |
1721 | ||
1722 | rcu_read_lock(); | |
1723 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1724 | if (xdp_prog) { | |
1725 | ret = do_xdp_generic(xdp_prog, skb); | |
1726 | if (ret != XDP_PASS) { | |
1727 | rcu_read_unlock(); | |
1728 | return total_len; | |
1729 | } | |
1730 | } | |
1731 | rcu_read_unlock(); | |
1732 | } | |
1733 | ||
feec084a | 1734 | rxhash = __skb_get_hash_symmetric(skb); |
94317099 | 1735 | |
90e33d45 PP |
1736 | if (frags) { |
1737 | /* Exercise flow dissector code path. */ | |
1738 | u32 headlen = eth_get_headlen(skb->data, skb_headlen(skb)); | |
1739 | ||
010f245b | 1740 | if (unlikely(headlen > skb_headlen(skb))) { |
90e33d45 PP |
1741 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
1742 | napi_free_frags(&tfile->napi); | |
1743 | mutex_unlock(&tfile->napi_mutex); | |
1744 | WARN_ON(1); | |
1745 | return -ENOMEM; | |
1746 | } | |
1747 | ||
1748 | local_bh_disable(); | |
1749 | napi_gro_frags(&tfile->napi); | |
1750 | local_bh_enable(); | |
1751 | mutex_unlock(&tfile->napi_mutex); | |
1752 | } else if (tun->flags & IFF_NAPI) { | |
94317099 PP |
1753 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; |
1754 | int queue_len; | |
1755 | ||
1756 | spin_lock_bh(&queue->lock); | |
1757 | __skb_queue_tail(queue, skb); | |
1758 | queue_len = skb_queue_len(queue); | |
1759 | spin_unlock(&queue->lock); | |
1760 | ||
1761 | if (!more || queue_len > NAPI_POLL_WEIGHT) | |
1762 | napi_schedule(&tfile->napi); | |
1763 | ||
1764 | local_bh_enable(); | |
1765 | } else if (!IS_ENABLED(CONFIG_4KSTACKS)) { | |
1766 | tun_rx_batched(tun, tfile, skb, more); | |
1767 | } else { | |
1768 | netif_rx_ni(skb); | |
1769 | } | |
6aa20a22 | 1770 | |
608b9977 PA |
1771 | stats = get_cpu_ptr(tun->pcpu_stats); |
1772 | u64_stats_update_begin(&stats->syncp); | |
1773 | stats->rx_packets++; | |
1774 | stats->rx_bytes += len; | |
1775 | u64_stats_update_end(&stats->syncp); | |
1776 | put_cpu_ptr(stats); | |
1da177e4 | 1777 | |
9e85722d | 1778 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1779 | return total_len; |
6aa20a22 | 1780 | } |
1da177e4 | 1781 | |
f5ff53b4 | 1782 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1783 | { |
33dccbb0 | 1784 | struct file *file = iocb->ki_filp; |
54f968d6 | 1785 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1786 | struct tun_struct *tun = tun_get(tfile); |
631ab46b | 1787 | ssize_t result; |
1da177e4 LT |
1788 | |
1789 | if (!tun) | |
1790 | return -EBADFD; | |
1791 | ||
5503fcec JW |
1792 | result = tun_get_user(tun, tfile, NULL, from, |
1793 | file->f_flags & O_NONBLOCK, false); | |
631ab46b EB |
1794 | |
1795 | tun_put(tun); | |
1796 | return result; | |
1da177e4 LT |
1797 | } |
1798 | ||
1da177e4 | 1799 | /* Put packet to the user space buffer */ |
6f7c156c | 1800 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1801 | struct tun_file *tfile, |
6f7c156c | 1802 | struct sk_buff *skb, |
e0b46d0e | 1803 | struct iov_iter *iter) |
1da177e4 LT |
1804 | { |
1805 | struct tun_pi pi = { 0, skb->protocol }; | |
608b9977 | 1806 | struct tun_pcpu_stats *stats; |
e0b46d0e | 1807 | ssize_t total; |
8c847d25 | 1808 | int vlan_offset = 0; |
a8f9bfdf | 1809 | int vlan_hlen = 0; |
2eb783c4 | 1810 | int vnet_hdr_sz = 0; |
a8f9bfdf | 1811 | |
df8a39de | 1812 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 1813 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 1814 | |
40630b82 | 1815 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 1816 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 1817 | |
e0b46d0e HX |
1818 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
1819 | ||
40630b82 | 1820 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 1821 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
1822 | return -EINVAL; |
1823 | ||
e0b46d0e HX |
1824 | total += sizeof(pi); |
1825 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
1826 | /* Packet will be striped */ |
1827 | pi.flags |= TUN_PKT_STRIP; | |
1828 | } | |
6aa20a22 | 1829 | |
e0b46d0e | 1830 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 1831 | return -EFAULT; |
6aa20a22 | 1832 | } |
1da177e4 | 1833 | |
2eb783c4 | 1834 | if (vnet_hdr_sz) { |
9403cd7c | 1835 | struct virtio_net_hdr gso; |
34166093 | 1836 | |
e0b46d0e | 1837 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
1838 | return -EINVAL; |
1839 | ||
3e9e40e7 | 1840 | if (virtio_net_hdr_from_skb(skb, &gso, |
6391a448 | 1841 | tun_is_little_endian(tun), true)) { |
f43798c2 | 1842 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
1843 | pr_err("unexpected GSO type: " |
1844 | "0x%x, gso_size %d, hdr_len %d\n", | |
1845 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
1846 | tun16_to_cpu(tun, gso.hdr_len)); | |
1847 | print_hex_dump(KERN_ERR, "tun: ", | |
1848 | DUMP_PREFIX_NONE, | |
1849 | 16, 1, skb->head, | |
1850 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
1851 | WARN_ON_ONCE(1); | |
1852 | return -EINVAL; | |
1853 | } | |
f43798c2 | 1854 | |
e0b46d0e | 1855 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 1856 | return -EFAULT; |
8c847d25 JW |
1857 | |
1858 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
1859 | } |
1860 | ||
a8f9bfdf | 1861 | if (vlan_hlen) { |
e0b46d0e | 1862 | int ret; |
6680ec68 JW |
1863 | struct { |
1864 | __be16 h_vlan_proto; | |
1865 | __be16 h_vlan_TCI; | |
1866 | } veth; | |
1867 | ||
1868 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 1869 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
1870 | |
1871 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 1872 | |
e0b46d0e HX |
1873 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
1874 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
1875 | goto done; |
1876 | ||
e0b46d0e HX |
1877 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
1878 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
1879 | goto done; |
1880 | } | |
1da177e4 | 1881 | |
e0b46d0e | 1882 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 1883 | |
6680ec68 | 1884 | done: |
608b9977 PA |
1885 | /* caller is in process context, */ |
1886 | stats = get_cpu_ptr(tun->pcpu_stats); | |
1887 | u64_stats_update_begin(&stats->syncp); | |
1888 | stats->tx_packets++; | |
1889 | stats->tx_bytes += skb->len + vlan_hlen; | |
1890 | u64_stats_update_end(&stats->syncp); | |
1891 | put_cpu_ptr(tun->pcpu_stats); | |
1da177e4 LT |
1892 | |
1893 | return total; | |
1894 | } | |
1895 | ||
1576d986 JW |
1896 | static struct sk_buff *tun_ring_recv(struct tun_file *tfile, int noblock, |
1897 | int *err) | |
1898 | { | |
1899 | DECLARE_WAITQUEUE(wait, current); | |
1900 | struct sk_buff *skb = NULL; | |
f48cc6b2 | 1901 | int error = 0; |
1576d986 JW |
1902 | |
1903 | skb = skb_array_consume(&tfile->tx_array); | |
1904 | if (skb) | |
1905 | goto out; | |
1906 | if (noblock) { | |
f48cc6b2 | 1907 | error = -EAGAIN; |
1576d986 JW |
1908 | goto out; |
1909 | } | |
1910 | ||
1911 | add_wait_queue(&tfile->wq.wait, &wait); | |
1912 | current->state = TASK_INTERRUPTIBLE; | |
1913 | ||
1914 | while (1) { | |
1915 | skb = skb_array_consume(&tfile->tx_array); | |
1916 | if (skb) | |
1917 | break; | |
1918 | if (signal_pending(current)) { | |
f48cc6b2 | 1919 | error = -ERESTARTSYS; |
1576d986 JW |
1920 | break; |
1921 | } | |
1922 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 1923 | error = -EFAULT; |
1576d986 JW |
1924 | break; |
1925 | } | |
1926 | ||
1927 | schedule(); | |
1928 | } | |
1929 | ||
1930 | current->state = TASK_RUNNING; | |
1931 | remove_wait_queue(&tfile->wq.wait, &wait); | |
1932 | ||
1933 | out: | |
f48cc6b2 | 1934 | *err = error; |
1576d986 JW |
1935 | return skb; |
1936 | } | |
1937 | ||
54f968d6 | 1938 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 | 1939 | struct iov_iter *to, |
ac77cfd4 | 1940 | int noblock, struct sk_buff *skb) |
1da177e4 | 1941 | { |
9b067034 | 1942 | ssize_t ret; |
1576d986 | 1943 | int err; |
1da177e4 | 1944 | |
3872baf6 | 1945 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1946 | |
9b067034 AV |
1947 | if (!iov_iter_count(to)) |
1948 | return 0; | |
1da177e4 | 1949 | |
ac77cfd4 JW |
1950 | if (!skb) { |
1951 | /* Read frames from ring */ | |
1952 | skb = tun_ring_recv(tfile, noblock, &err); | |
1953 | if (!skb) | |
1954 | return err; | |
1955 | } | |
e0b46d0e | 1956 | |
9b067034 | 1957 | ret = tun_put_user(tun, tfile, skb, to); |
f51a5e82 | 1958 | if (unlikely(ret < 0)) |
f271b2cc | 1959 | kfree_skb(skb); |
f51a5e82 JW |
1960 | else |
1961 | consume_skb(skb); | |
1da177e4 | 1962 | |
05c2828c MT |
1963 | return ret; |
1964 | } | |
1965 | ||
9b067034 | 1966 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
1967 | { |
1968 | struct file *file = iocb->ki_filp; | |
1969 | struct tun_file *tfile = file->private_data; | |
9484dc74 | 1970 | struct tun_struct *tun = tun_get(tfile); |
9b067034 | 1971 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
1972 | |
1973 | if (!tun) | |
1974 | return -EBADFD; | |
ac77cfd4 | 1975 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK, NULL); |
42404c09 | 1976 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
1977 | if (ret > 0) |
1978 | iocb->ki_pos = ret; | |
631ab46b | 1979 | tun_put(tun); |
1da177e4 LT |
1980 | return ret; |
1981 | } | |
1982 | ||
96442e42 JW |
1983 | static void tun_free_netdev(struct net_device *dev) |
1984 | { | |
1985 | struct tun_struct *tun = netdev_priv(dev); | |
1986 | ||
4008e97f | 1987 | BUG_ON(!(list_empty(&tun->disabled))); |
608b9977 | 1988 | free_percpu(tun->pcpu_stats); |
96442e42 | 1989 | tun_flow_uninit(tun); |
5dbbaf2d | 1990 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
1991 | } |
1992 | ||
1da177e4 LT |
1993 | static void tun_setup(struct net_device *dev) |
1994 | { | |
1995 | struct tun_struct *tun = netdev_priv(dev); | |
1996 | ||
0625c883 EB |
1997 | tun->owner = INVALID_UID; |
1998 | tun->group = INVALID_GID; | |
1da177e4 | 1999 | |
1da177e4 | 2000 | dev->ethtool_ops = &tun_ethtool_ops; |
cf124db5 DM |
2001 | dev->needs_free_netdev = true; |
2002 | dev->priv_destructor = tun_free_netdev; | |
016adb72 JW |
2003 | /* We prefer our own queue length */ |
2004 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
2005 | } |
2006 | ||
f019a7a5 EB |
2007 | /* Trivial set of netlink ops to allow deleting tun or tap |
2008 | * device with netlink. | |
2009 | */ | |
a8b8a889 MS |
2010 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[], |
2011 | struct netlink_ext_ack *extack) | |
f019a7a5 EB |
2012 | { |
2013 | return -EINVAL; | |
2014 | } | |
2015 | ||
2016 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
2017 | .kind = DRV_NAME, | |
2018 | .priv_size = sizeof(struct tun_struct), | |
2019 | .setup = tun_setup, | |
2020 | .validate = tun_validate, | |
2021 | }; | |
2022 | ||
33dccbb0 HX |
2023 | static void tun_sock_write_space(struct sock *sk) |
2024 | { | |
54f968d6 | 2025 | struct tun_file *tfile; |
43815482 | 2026 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
2027 | |
2028 | if (!sock_writeable(sk)) | |
2029 | return; | |
2030 | ||
9cd3e072 | 2031 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
2032 | return; |
2033 | ||
43815482 ED |
2034 | wqueue = sk_sleep(sk); |
2035 | if (wqueue && waitqueue_active(wqueue)) | |
2036 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 2037 | POLLWRNORM | POLLWRBAND); |
c722c625 | 2038 | |
54f968d6 JW |
2039 | tfile = container_of(sk, struct tun_file, sk); |
2040 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
2041 | } |
2042 | ||
1b784140 | 2043 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 2044 | { |
54f968d6 JW |
2045 | int ret; |
2046 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
9484dc74 | 2047 | struct tun_struct *tun = tun_get(tfile); |
54f968d6 JW |
2048 | |
2049 | if (!tun) | |
2050 | return -EBADFD; | |
f5ff53b4 | 2051 | |
c0371da6 | 2052 | ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter, |
5503fcec JW |
2053 | m->msg_flags & MSG_DONTWAIT, |
2054 | m->msg_flags & MSG_MORE); | |
54f968d6 JW |
2055 | tun_put(tun); |
2056 | return ret; | |
05c2828c MT |
2057 | } |
2058 | ||
1b784140 | 2059 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
2060 | int flags) |
2061 | { | |
54f968d6 | 2062 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2063 | struct tun_struct *tun = tun_get(tfile); |
05c2828c | 2064 | int ret; |
54f968d6 JW |
2065 | |
2066 | if (!tun) | |
2067 | return -EBADFD; | |
2068 | ||
eda29772 | 2069 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 G |
2070 | ret = -EINVAL; |
2071 | goto out; | |
2072 | } | |
eda29772 RC |
2073 | if (flags & MSG_ERRQUEUE) { |
2074 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
2075 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
2076 | goto out; | |
2077 | } | |
ac77cfd4 JW |
2078 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, |
2079 | m->msg_control); | |
87897931 | 2080 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
2081 | m->msg_flags |= MSG_TRUNC; |
2082 | ret = flags & MSG_TRUNC ? ret : total_len; | |
2083 | } | |
3811ae76 | 2084 | out: |
54f968d6 | 2085 | tun_put(tun); |
05c2828c MT |
2086 | return ret; |
2087 | } | |
2088 | ||
1576d986 JW |
2089 | static int tun_peek_len(struct socket *sock) |
2090 | { | |
2091 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
2092 | struct tun_struct *tun; | |
2093 | int ret = 0; | |
2094 | ||
9484dc74 | 2095 | tun = tun_get(tfile); |
1576d986 JW |
2096 | if (!tun) |
2097 | return 0; | |
2098 | ||
2099 | ret = skb_array_peek_len(&tfile->tx_array); | |
2100 | tun_put(tun); | |
2101 | ||
2102 | return ret; | |
2103 | } | |
2104 | ||
05c2828c MT |
2105 | /* Ops structure to mimic raw sockets with tun */ |
2106 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 2107 | .peek_len = tun_peek_len, |
05c2828c MT |
2108 | .sendmsg = tun_sendmsg, |
2109 | .recvmsg = tun_recvmsg, | |
2110 | }; | |
2111 | ||
33dccbb0 HX |
2112 | static struct proto tun_proto = { |
2113 | .name = "tun", | |
2114 | .owner = THIS_MODULE, | |
54f968d6 | 2115 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 2116 | }; |
f019a7a5 | 2117 | |
980c9e8c DW |
2118 | static int tun_flags(struct tun_struct *tun) |
2119 | { | |
031f5e03 | 2120 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
2121 | } |
2122 | ||
2123 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
2124 | char *buf) | |
2125 | { | |
2126 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
2127 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
2128 | } | |
2129 | ||
2130 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
2131 | char *buf) | |
2132 | { | |
2133 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2134 | return uid_valid(tun->owner)? |
2135 | sprintf(buf, "%u\n", | |
2136 | from_kuid_munged(current_user_ns(), tun->owner)): | |
2137 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2138 | } |
2139 | ||
2140 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
2141 | char *buf) | |
2142 | { | |
2143 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2144 | return gid_valid(tun->group) ? |
2145 | sprintf(buf, "%u\n", | |
2146 | from_kgid_munged(current_user_ns(), tun->group)): | |
2147 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2148 | } |
2149 | ||
2150 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
2151 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
2152 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
2153 | ||
c4d33e24 TI |
2154 | static struct attribute *tun_dev_attrs[] = { |
2155 | &dev_attr_tun_flags.attr, | |
2156 | &dev_attr_owner.attr, | |
2157 | &dev_attr_group.attr, | |
2158 | NULL | |
2159 | }; | |
2160 | ||
2161 | static const struct attribute_group tun_attr_group = { | |
2162 | .attrs = tun_dev_attrs | |
2163 | }; | |
2164 | ||
d647a591 | 2165 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
2166 | { |
2167 | struct tun_struct *tun; | |
54f968d6 | 2168 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2169 | struct net_device *dev; |
2170 | int err; | |
2171 | ||
7c0c3b1a JW |
2172 | if (tfile->detached) |
2173 | return -EINVAL; | |
2174 | ||
90e33d45 PP |
2175 | if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) { |
2176 | if (!capable(CAP_NET_ADMIN)) | |
2177 | return -EPERM; | |
2178 | ||
2179 | if (!(ifr->ifr_flags & IFF_NAPI) || | |
2180 | (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP) | |
2181 | return -EINVAL; | |
2182 | } | |
2183 | ||
74a3e5a7 EB |
2184 | dev = __dev_get_by_name(net, ifr->ifr_name); |
2185 | if (dev) { | |
f85ba780 DW |
2186 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
2187 | return -EBUSY; | |
74a3e5a7 EB |
2188 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
2189 | tun = netdev_priv(dev); | |
2190 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
2191 | tun = netdev_priv(dev); | |
2192 | else | |
2193 | return -EINVAL; | |
2194 | ||
8e6d91ae | 2195 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 2196 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
2197 | return -EINVAL; |
2198 | ||
cde8b15f | 2199 | if (tun_not_capable(tun)) |
2b980dbd | 2200 | return -EPERM; |
5dbbaf2d | 2201 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
2202 | if (err < 0) |
2203 | return err; | |
2204 | ||
94317099 PP |
2205 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER, |
2206 | ifr->ifr_flags & IFF_NAPI); | |
a7385ba2 EB |
2207 | if (err < 0) |
2208 | return err; | |
4008e97f | 2209 | |
40630b82 | 2210 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
2211 | (tun->numqueues + tun->numdisabled > 1)) { |
2212 | /* One or more queue has already been attached, no need | |
2213 | * to initialize the device again. | |
2214 | */ | |
2215 | return 0; | |
2216 | } | |
6aa20a22 | 2217 | } |
1da177e4 LT |
2218 | else { |
2219 | char *name; | |
2220 | unsigned long flags = 0; | |
edfb6a14 JW |
2221 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
2222 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 2223 | |
c260b772 | 2224 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 2225 | return -EPERM; |
2b980dbd PM |
2226 | err = security_tun_dev_create(); |
2227 | if (err < 0) | |
2228 | return err; | |
ca6bb5d7 | 2229 | |
1da177e4 LT |
2230 | /* Set dev type */ |
2231 | if (ifr->ifr_flags & IFF_TUN) { | |
2232 | /* TUN device */ | |
40630b82 | 2233 | flags |= IFF_TUN; |
1da177e4 LT |
2234 | name = "tun%d"; |
2235 | } else if (ifr->ifr_flags & IFF_TAP) { | |
2236 | /* TAP device */ | |
40630b82 | 2237 | flags |= IFF_TAP; |
1da177e4 | 2238 | name = "tap%d"; |
6aa20a22 | 2239 | } else |
36989b90 | 2240 | return -EINVAL; |
6aa20a22 | 2241 | |
1da177e4 LT |
2242 | if (*ifr->ifr_name) |
2243 | name = ifr->ifr_name; | |
2244 | ||
c8d68e6b | 2245 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
2246 | NET_NAME_UNKNOWN, tun_setup, queues, |
2247 | queues); | |
edfb6a14 | 2248 | |
1da177e4 LT |
2249 | if (!dev) |
2250 | return -ENOMEM; | |
2251 | ||
fc54c658 | 2252 | dev_net_set(dev, net); |
f019a7a5 | 2253 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 2254 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 2255 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 2256 | |
1da177e4 LT |
2257 | tun = netdev_priv(dev); |
2258 | tun->dev = dev; | |
2259 | tun->flags = flags; | |
f271b2cc | 2260 | tun->txflt.count = 0; |
d9d52b51 | 2261 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 2262 | |
eaea34b2 | 2263 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
2264 | tun->filter_attached = false; |
2265 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 2266 | tun->rx_batched = 0; |
33dccbb0 | 2267 | |
608b9977 PA |
2268 | tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); |
2269 | if (!tun->pcpu_stats) { | |
2270 | err = -ENOMEM; | |
2271 | goto err_free_dev; | |
2272 | } | |
2273 | ||
96442e42 JW |
2274 | spin_lock_init(&tun->lock); |
2275 | ||
5dbbaf2d PM |
2276 | err = security_tun_dev_alloc_security(&tun->security); |
2277 | if (err < 0) | |
608b9977 | 2278 | goto err_free_stat; |
2b980dbd | 2279 | |
1da177e4 | 2280 | tun_net_init(dev); |
944a1376 | 2281 | tun_flow_init(tun); |
96442e42 | 2282 | |
88255375 | 2283 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
2284 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
2285 | NETIF_F_HW_VLAN_STAG_TX; | |
2a2bbf17 | 2286 | dev->features = dev->hw_features | NETIF_F_LLTX; |
6671b224 FLVC |
2287 | dev->vlan_features = dev->features & |
2288 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
2289 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 2290 | |
4008e97f | 2291 | INIT_LIST_HEAD(&tun->disabled); |
94317099 | 2292 | err = tun_attach(tun, file, false, ifr->ifr_flags & IFF_NAPI); |
eb0fb363 | 2293 | if (err < 0) |
662ca437 | 2294 | goto err_free_flow; |
eb0fb363 | 2295 | |
1da177e4 LT |
2296 | err = register_netdevice(tun->dev); |
2297 | if (err < 0) | |
662ca437 | 2298 | goto err_detach; |
1da177e4 LT |
2299 | } |
2300 | ||
af668b3c MT |
2301 | netif_carrier_on(tun->dev); |
2302 | ||
6b8a66ee | 2303 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 2304 | |
031f5e03 MT |
2305 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
2306 | (ifr->ifr_flags & TUN_FEATURES); | |
c8d68e6b | 2307 | |
e35259a9 MK |
2308 | /* Make sure persistent devices do not get stuck in |
2309 | * xoff state. | |
2310 | */ | |
2311 | if (netif_running(tun->dev)) | |
c8d68e6b | 2312 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 2313 | |
1da177e4 LT |
2314 | strcpy(ifr->ifr_name, tun->dev->name); |
2315 | return 0; | |
2316 | ||
662ca437 JW |
2317 | err_detach: |
2318 | tun_detach_all(dev); | |
ff244c6b ED |
2319 | /* register_netdevice() already called tun_free_netdev() */ |
2320 | goto err_free_dev; | |
2321 | ||
662ca437 JW |
2322 | err_free_flow: |
2323 | tun_flow_uninit(tun); | |
2324 | security_tun_dev_free_security(tun->security); | |
608b9977 PA |
2325 | err_free_stat: |
2326 | free_percpu(tun->pcpu_stats); | |
662ca437 | 2327 | err_free_dev: |
1da177e4 | 2328 | free_netdev(dev); |
1da177e4 LT |
2329 | return err; |
2330 | } | |
2331 | ||
9ce99cf6 | 2332 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 2333 | struct ifreq *ifr) |
e3b99556 | 2334 | { |
6b8a66ee | 2335 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
2336 | |
2337 | strcpy(ifr->ifr_name, tun->dev->name); | |
2338 | ||
980c9e8c | 2339 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 2340 | |
e3b99556 MM |
2341 | } |
2342 | ||
5228ddc9 RR |
2343 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
2344 | * privs required. */ | |
88255375 | 2345 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 2346 | { |
c8f44aff | 2347 | netdev_features_t features = 0; |
5228ddc9 RR |
2348 | |
2349 | if (arg & TUN_F_CSUM) { | |
88255375 | 2350 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
2351 | arg &= ~TUN_F_CSUM; |
2352 | ||
2353 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
2354 | if (arg & TUN_F_TSO_ECN) { | |
2355 | features |= NETIF_F_TSO_ECN; | |
2356 | arg &= ~TUN_F_TSO_ECN; | |
2357 | } | |
2358 | if (arg & TUN_F_TSO4) | |
2359 | features |= NETIF_F_TSO; | |
2360 | if (arg & TUN_F_TSO6) | |
2361 | features |= NETIF_F_TSO6; | |
2362 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
2363 | } | |
2364 | } | |
2365 | ||
2366 | /* This gives the user a way to test for new features in future by | |
2367 | * trying to set them. */ | |
2368 | if (arg) | |
2369 | return -EINVAL; | |
2370 | ||
88255375 | 2371 | tun->set_features = features; |
09050957 YI |
2372 | tun->dev->wanted_features &= ~TUN_USER_FEATURES; |
2373 | tun->dev->wanted_features |= features; | |
88255375 | 2374 | netdev_update_features(tun->dev); |
5228ddc9 RR |
2375 | |
2376 | return 0; | |
2377 | } | |
2378 | ||
c8d68e6b JW |
2379 | static void tun_detach_filter(struct tun_struct *tun, int n) |
2380 | { | |
2381 | int i; | |
2382 | struct tun_file *tfile; | |
2383 | ||
2384 | for (i = 0; i < n; i++) { | |
b8deabd3 | 2385 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2386 | lock_sock(tfile->socket.sk); |
2387 | sk_detach_filter(tfile->socket.sk); | |
2388 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2389 | } |
2390 | ||
2391 | tun->filter_attached = false; | |
2392 | } | |
2393 | ||
2394 | static int tun_attach_filter(struct tun_struct *tun) | |
2395 | { | |
2396 | int i, ret = 0; | |
2397 | struct tun_file *tfile; | |
2398 | ||
2399 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2400 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2401 | lock_sock(tfile->socket.sk); |
2402 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
2403 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2404 | if (ret) { |
2405 | tun_detach_filter(tun, i); | |
2406 | return ret; | |
2407 | } | |
2408 | } | |
2409 | ||
2410 | tun->filter_attached = true; | |
2411 | return ret; | |
2412 | } | |
2413 | ||
2414 | static void tun_set_sndbuf(struct tun_struct *tun) | |
2415 | { | |
2416 | struct tun_file *tfile; | |
2417 | int i; | |
2418 | ||
2419 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2420 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
2421 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
2422 | } | |
2423 | } | |
2424 | ||
cde8b15f JW |
2425 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
2426 | { | |
2427 | struct tun_file *tfile = file->private_data; | |
2428 | struct tun_struct *tun; | |
cde8b15f JW |
2429 | int ret = 0; |
2430 | ||
2431 | rtnl_lock(); | |
2432 | ||
2433 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 2434 | tun = tfile->detached; |
5dbbaf2d | 2435 | if (!tun) { |
cde8b15f | 2436 | ret = -EINVAL; |
5dbbaf2d PM |
2437 | goto unlock; |
2438 | } | |
2439 | ret = security_tun_dev_attach_queue(tun->security); | |
2440 | if (ret < 0) | |
2441 | goto unlock; | |
94317099 | 2442 | ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI); |
4008e97f | 2443 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 2444 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 2445 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
2446 | ret = -EINVAL; |
2447 | else | |
2448 | __tun_detach(tfile, false); | |
2449 | } else | |
cde8b15f JW |
2450 | ret = -EINVAL; |
2451 | ||
5dbbaf2d | 2452 | unlock: |
cde8b15f JW |
2453 | rtnl_unlock(); |
2454 | return ret; | |
2455 | } | |
2456 | ||
50857e2a AB |
2457 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
2458 | unsigned long arg, int ifreq_len) | |
1da177e4 | 2459 | { |
36b50bab | 2460 | struct tun_file *tfile = file->private_data; |
631ab46b | 2461 | struct tun_struct *tun; |
1da177e4 LT |
2462 | void __user* argp = (void __user*)arg; |
2463 | struct ifreq ifr; | |
0625c883 EB |
2464 | kuid_t owner; |
2465 | kgid_t group; | |
33dccbb0 | 2466 | int sndbuf; |
d9d52b51 | 2467 | int vnet_hdr_sz; |
fb7589a1 | 2468 | unsigned int ifindex; |
1cf8e410 | 2469 | int le; |
f271b2cc | 2470 | int ret; |
1da177e4 | 2471 | |
20861f26 | 2472 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == SOCK_IOC_TYPE) { |
50857e2a | 2473 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 2474 | return -EFAULT; |
8bbb1813 | 2475 | } else { |
a117dacd | 2476 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 2477 | } |
631ab46b EB |
2478 | if (cmd == TUNGETFEATURES) { |
2479 | /* Currently this just means: "what IFF flags are valid?". | |
2480 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
2481 | * TUNSETIFF. |
2482 | */ | |
2483 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 2484 | (unsigned int __user*)argp); |
cde8b15f JW |
2485 | } else if (cmd == TUNSETQUEUE) |
2486 | return tun_set_queue(file, &ifr); | |
631ab46b | 2487 | |
c8d68e6b | 2488 | ret = 0; |
876bfd4d HX |
2489 | rtnl_lock(); |
2490 | ||
9484dc74 | 2491 | tun = tun_get(tfile); |
0f16bc13 GF |
2492 | if (cmd == TUNSETIFF) { |
2493 | ret = -EEXIST; | |
2494 | if (tun) | |
2495 | goto unlock; | |
2496 | ||
1da177e4 LT |
2497 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
2498 | ||
140e807d | 2499 | ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr); |
1da177e4 | 2500 | |
876bfd4d HX |
2501 | if (ret) |
2502 | goto unlock; | |
1da177e4 | 2503 | |
50857e2a | 2504 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
2505 | ret = -EFAULT; |
2506 | goto unlock; | |
1da177e4 | 2507 | } |
fb7589a1 PE |
2508 | if (cmd == TUNSETIFINDEX) { |
2509 | ret = -EPERM; | |
2510 | if (tun) | |
2511 | goto unlock; | |
2512 | ||
2513 | ret = -EFAULT; | |
2514 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
2515 | goto unlock; | |
2516 | ||
2517 | ret = 0; | |
2518 | tfile->ifindex = ifindex; | |
2519 | goto unlock; | |
2520 | } | |
1da177e4 | 2521 | |
876bfd4d | 2522 | ret = -EBADFD; |
1da177e4 | 2523 | if (!tun) |
876bfd4d | 2524 | goto unlock; |
1da177e4 | 2525 | |
1e588338 | 2526 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 2527 | |
631ab46b | 2528 | ret = 0; |
1da177e4 | 2529 | switch (cmd) { |
e3b99556 | 2530 | case TUNGETIFF: |
9ce99cf6 | 2531 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 2532 | |
3d407a80 PE |
2533 | if (tfile->detached) |
2534 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
2535 | if (!tfile->socket.sk->sk_filter) |
2536 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 2537 | |
50857e2a | 2538 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 2539 | ret = -EFAULT; |
e3b99556 MM |
2540 | break; |
2541 | ||
1da177e4 LT |
2542 | case TUNSETNOCSUM: |
2543 | /* Disable/Enable checksum */ | |
1da177e4 | 2544 | |
88255375 MM |
2545 | /* [unimplemented] */ |
2546 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 2547 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
2548 | break; |
2549 | ||
2550 | case TUNSETPERSIST: | |
54f968d6 JW |
2551 | /* Disable/Enable persist mode. Keep an extra reference to the |
2552 | * module to prevent the module being unprobed. | |
2553 | */ | |
40630b82 MT |
2554 | if (arg && !(tun->flags & IFF_PERSIST)) { |
2555 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 2556 | __module_get(THIS_MODULE); |
dd38bd85 | 2557 | } |
40630b82 MT |
2558 | if (!arg && (tun->flags & IFF_PERSIST)) { |
2559 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 JW |
2560 | module_put(THIS_MODULE); |
2561 | } | |
1da177e4 | 2562 | |
6b8a66ee JP |
2563 | tun_debug(KERN_INFO, tun, "persist %s\n", |
2564 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
2565 | break; |
2566 | ||
2567 | case TUNSETOWNER: | |
2568 | /* Set owner of the device */ | |
0625c883 EB |
2569 | owner = make_kuid(current_user_ns(), arg); |
2570 | if (!uid_valid(owner)) { | |
2571 | ret = -EINVAL; | |
2572 | break; | |
2573 | } | |
2574 | tun->owner = owner; | |
1e588338 | 2575 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 2576 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
2577 | break; |
2578 | ||
8c644623 GG |
2579 | case TUNSETGROUP: |
2580 | /* Set group of the device */ | |
0625c883 EB |
2581 | group = make_kgid(current_user_ns(), arg); |
2582 | if (!gid_valid(group)) { | |
2583 | ret = -EINVAL; | |
2584 | break; | |
2585 | } | |
2586 | tun->group = group; | |
1e588338 | 2587 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 2588 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
2589 | break; |
2590 | ||
ff4cc3ac MK |
2591 | case TUNSETLINK: |
2592 | /* Only allow setting the type when the interface is down */ | |
2593 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
2594 | tun_debug(KERN_INFO, tun, |
2595 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 2596 | ret = -EBUSY; |
ff4cc3ac MK |
2597 | } else { |
2598 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2599 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2600 | tun->dev->type); | |
48abfe05 | 2601 | ret = 0; |
ff4cc3ac | 2602 | } |
631ab46b | 2603 | break; |
ff4cc3ac | 2604 | |
1da177e4 LT |
2605 | #ifdef TUN_DEBUG |
2606 | case TUNSETDEBUG: | |
2607 | tun->debug = arg; | |
2608 | break; | |
2609 | #endif | |
5228ddc9 | 2610 | case TUNSETOFFLOAD: |
88255375 | 2611 | ret = set_offload(tun, arg); |
631ab46b | 2612 | break; |
5228ddc9 | 2613 | |
f271b2cc MK |
2614 | case TUNSETTXFILTER: |
2615 | /* Can be set only for TAPs */ | |
631ab46b | 2616 | ret = -EINVAL; |
40630b82 | 2617 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 2618 | break; |
c0e5a8c2 | 2619 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2620 | break; |
1da177e4 LT |
2621 | |
2622 | case SIOCGIFHWADDR: | |
b595076a | 2623 | /* Get hw address */ |
f271b2cc MK |
2624 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2625 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2626 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2627 | ret = -EFAULT; |
2628 | break; | |
1da177e4 LT |
2629 | |
2630 | case SIOCSIFHWADDR: | |
f271b2cc | 2631 | /* Set hw address */ |
6b8a66ee JP |
2632 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2633 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2634 | |
40102371 | 2635 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2636 | break; |
33dccbb0 HX |
2637 | |
2638 | case TUNGETSNDBUF: | |
54f968d6 | 2639 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2640 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2641 | ret = -EFAULT; | |
2642 | break; | |
2643 | ||
2644 | case TUNSETSNDBUF: | |
2645 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2646 | ret = -EFAULT; | |
2647 | break; | |
2648 | } | |
2649 | ||
c8d68e6b JW |
2650 | tun->sndbuf = sndbuf; |
2651 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2652 | break; |
2653 | ||
d9d52b51 MT |
2654 | case TUNGETVNETHDRSZ: |
2655 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2656 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2657 | ret = -EFAULT; | |
2658 | break; | |
2659 | ||
2660 | case TUNSETVNETHDRSZ: | |
2661 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2662 | ret = -EFAULT; | |
2663 | break; | |
2664 | } | |
2665 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2666 | ret = -EINVAL; | |
2667 | break; | |
2668 | } | |
2669 | ||
2670 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2671 | break; | |
2672 | ||
1cf8e410 MT |
2673 | case TUNGETVNETLE: |
2674 | le = !!(tun->flags & TUN_VNET_LE); | |
2675 | if (put_user(le, (int __user *)argp)) | |
2676 | ret = -EFAULT; | |
2677 | break; | |
2678 | ||
2679 | case TUNSETVNETLE: | |
2680 | if (get_user(le, (int __user *)argp)) { | |
2681 | ret = -EFAULT; | |
2682 | break; | |
2683 | } | |
2684 | if (le) | |
2685 | tun->flags |= TUN_VNET_LE; | |
2686 | else | |
2687 | tun->flags &= ~TUN_VNET_LE; | |
2688 | break; | |
2689 | ||
8b8e658b GK |
2690 | case TUNGETVNETBE: |
2691 | ret = tun_get_vnet_be(tun, argp); | |
2692 | break; | |
2693 | ||
2694 | case TUNSETVNETBE: | |
2695 | ret = tun_set_vnet_be(tun, argp); | |
2696 | break; | |
2697 | ||
99405162 MT |
2698 | case TUNATTACHFILTER: |
2699 | /* Can be set only for TAPs */ | |
2700 | ret = -EINVAL; | |
40630b82 | 2701 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
2702 | break; |
2703 | ret = -EFAULT; | |
54f968d6 | 2704 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2705 | break; |
2706 | ||
c8d68e6b | 2707 | ret = tun_attach_filter(tun); |
99405162 MT |
2708 | break; |
2709 | ||
2710 | case TUNDETACHFILTER: | |
2711 | /* Can be set only for TAPs */ | |
2712 | ret = -EINVAL; | |
40630b82 | 2713 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 2714 | break; |
c8d68e6b JW |
2715 | ret = 0; |
2716 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2717 | break; |
2718 | ||
76975e9c PE |
2719 | case TUNGETFILTER: |
2720 | ret = -EINVAL; | |
40630b82 | 2721 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
2722 | break; |
2723 | ret = -EFAULT; | |
2724 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2725 | break; | |
2726 | ret = 0; | |
2727 | break; | |
2728 | ||
1da177e4 | 2729 | default: |
631ab46b EB |
2730 | ret = -EINVAL; |
2731 | break; | |
ee289b64 | 2732 | } |
1da177e4 | 2733 | |
876bfd4d HX |
2734 | unlock: |
2735 | rtnl_unlock(); | |
2736 | if (tun) | |
2737 | tun_put(tun); | |
631ab46b | 2738 | return ret; |
1da177e4 LT |
2739 | } |
2740 | ||
50857e2a AB |
2741 | static long tun_chr_ioctl(struct file *file, |
2742 | unsigned int cmd, unsigned long arg) | |
2743 | { | |
2744 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2745 | } | |
2746 | ||
2747 | #ifdef CONFIG_COMPAT | |
2748 | static long tun_chr_compat_ioctl(struct file *file, | |
2749 | unsigned int cmd, unsigned long arg) | |
2750 | { | |
2751 | switch (cmd) { | |
2752 | case TUNSETIFF: | |
2753 | case TUNGETIFF: | |
2754 | case TUNSETTXFILTER: | |
2755 | case TUNGETSNDBUF: | |
2756 | case TUNSETSNDBUF: | |
2757 | case SIOCGIFHWADDR: | |
2758 | case SIOCSIFHWADDR: | |
2759 | arg = (unsigned long)compat_ptr(arg); | |
2760 | break; | |
2761 | default: | |
2762 | arg = (compat_ulong_t)arg; | |
2763 | break; | |
2764 | } | |
2765 | ||
2766 | /* | |
2767 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2768 | * the end of that structure. All fields that are used in this | |
2769 | * driver are compatible though, we don't need to convert the | |
2770 | * contents. | |
2771 | */ | |
2772 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2773 | } | |
2774 | #endif /* CONFIG_COMPAT */ | |
2775 | ||
1da177e4 LT |
2776 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2777 | { | |
54f968d6 | 2778 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2779 | int ret; |
2780 | ||
54f968d6 | 2781 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2782 | goto out; |
6aa20a22 | 2783 | |
1da177e4 | 2784 | if (on) { |
e0b93edd | 2785 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2786 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2787 | } else |
54f968d6 | 2788 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2789 | ret = 0; |
2790 | out: | |
9d319522 | 2791 | return ret; |
1da177e4 LT |
2792 | } |
2793 | ||
2794 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2795 | { | |
140e807d | 2796 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 2797 | struct tun_file *tfile; |
deed49fb | 2798 | |
6b8a66ee | 2799 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2800 | |
140e807d | 2801 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 2802 | &tun_proto, 0); |
631ab46b EB |
2803 | if (!tfile) |
2804 | return -ENOMEM; | |
c956674b | 2805 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 2806 | tfile->flags = 0; |
fb7589a1 | 2807 | tfile->ifindex = 0; |
54f968d6 | 2808 | |
54f968d6 | 2809 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2810 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2811 | |
2812 | tfile->socket.file = file; | |
2813 | tfile->socket.ops = &tun_socket_ops; | |
2814 | ||
2815 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
2816 | |
2817 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2818 | tfile->sk.sk_sndbuf = INT_MAX; | |
2819 | ||
631ab46b | 2820 | file->private_data = tfile; |
4008e97f | 2821 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2822 | |
19a6afb2 JW |
2823 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2824 | ||
1da177e4 LT |
2825 | return 0; |
2826 | } | |
2827 | ||
2828 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2829 | { | |
631ab46b | 2830 | struct tun_file *tfile = file->private_data; |
1da177e4 | 2831 | |
c8d68e6b | 2832 | tun_detach(tfile, true); |
1da177e4 LT |
2833 | |
2834 | return 0; | |
2835 | } | |
2836 | ||
93e14b6d | 2837 | #ifdef CONFIG_PROC_FS |
9484dc74 | 2838 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file) |
93e14b6d | 2839 | { |
9484dc74 | 2840 | struct tun_file *tfile = file->private_data; |
93e14b6d MY |
2841 | struct tun_struct *tun; |
2842 | struct ifreq ifr; | |
2843 | ||
2844 | memset(&ifr, 0, sizeof(ifr)); | |
2845 | ||
2846 | rtnl_lock(); | |
9484dc74 | 2847 | tun = tun_get(tfile); |
93e14b6d MY |
2848 | if (tun) |
2849 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2850 | rtnl_unlock(); | |
2851 | ||
2852 | if (tun) | |
2853 | tun_put(tun); | |
2854 | ||
a3816ab0 | 2855 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
2856 | } |
2857 | #endif | |
2858 | ||
d54b1fdb | 2859 | static const struct file_operations tun_fops = { |
6aa20a22 | 2860 | .owner = THIS_MODULE, |
1da177e4 | 2861 | .llseek = no_llseek, |
9b067034 | 2862 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 2863 | .write_iter = tun_chr_write_iter, |
1da177e4 | 2864 | .poll = tun_chr_poll, |
50857e2a AB |
2865 | .unlocked_ioctl = tun_chr_ioctl, |
2866 | #ifdef CONFIG_COMPAT | |
2867 | .compat_ioctl = tun_chr_compat_ioctl, | |
2868 | #endif | |
1da177e4 LT |
2869 | .open = tun_chr_open, |
2870 | .release = tun_chr_close, | |
93e14b6d MY |
2871 | .fasync = tun_chr_fasync, |
2872 | #ifdef CONFIG_PROC_FS | |
2873 | .show_fdinfo = tun_chr_show_fdinfo, | |
2874 | #endif | |
1da177e4 LT |
2875 | }; |
2876 | ||
2877 | static struct miscdevice tun_miscdev = { | |
2878 | .minor = TUN_MINOR, | |
2879 | .name = "tun", | |
e454cea2 | 2880 | .nodename = "net/tun", |
1da177e4 | 2881 | .fops = &tun_fops, |
1da177e4 LT |
2882 | }; |
2883 | ||
2884 | /* ethtool interface */ | |
2885 | ||
29ccc49d PR |
2886 | static int tun_get_link_ksettings(struct net_device *dev, |
2887 | struct ethtool_link_ksettings *cmd) | |
2888 | { | |
2889 | ethtool_link_ksettings_zero_link_mode(cmd, supported); | |
2890 | ethtool_link_ksettings_zero_link_mode(cmd, advertising); | |
2891 | cmd->base.speed = SPEED_10; | |
2892 | cmd->base.duplex = DUPLEX_FULL; | |
2893 | cmd->base.port = PORT_TP; | |
2894 | cmd->base.phy_address = 0; | |
2895 | cmd->base.autoneg = AUTONEG_DISABLE; | |
1da177e4 LT |
2896 | return 0; |
2897 | } | |
2898 | ||
2899 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2900 | { | |
2901 | struct tun_struct *tun = netdev_priv(dev); | |
2902 | ||
33a5ba14 RJ |
2903 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2904 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2905 | |
2906 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 2907 | case IFF_TUN: |
33a5ba14 | 2908 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 2909 | break; |
40630b82 | 2910 | case IFF_TAP: |
33a5ba14 | 2911 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
2912 | break; |
2913 | } | |
2914 | } | |
2915 | ||
2916 | static u32 tun_get_msglevel(struct net_device *dev) | |
2917 | { | |
2918 | #ifdef TUN_DEBUG | |
2919 | struct tun_struct *tun = netdev_priv(dev); | |
2920 | return tun->debug; | |
2921 | #else | |
2922 | return -EOPNOTSUPP; | |
2923 | #endif | |
2924 | } | |
2925 | ||
2926 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
2927 | { | |
2928 | #ifdef TUN_DEBUG | |
2929 | struct tun_struct *tun = netdev_priv(dev); | |
2930 | tun->debug = value; | |
2931 | #endif | |
2932 | } | |
2933 | ||
5503fcec JW |
2934 | static int tun_get_coalesce(struct net_device *dev, |
2935 | struct ethtool_coalesce *ec) | |
2936 | { | |
2937 | struct tun_struct *tun = netdev_priv(dev); | |
2938 | ||
2939 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
2940 | ||
2941 | return 0; | |
2942 | } | |
2943 | ||
2944 | static int tun_set_coalesce(struct net_device *dev, | |
2945 | struct ethtool_coalesce *ec) | |
2946 | { | |
2947 | struct tun_struct *tun = netdev_priv(dev); | |
2948 | ||
2949 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
2950 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
2951 | else | |
2952 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
2953 | ||
2954 | return 0; | |
2955 | } | |
2956 | ||
7282d491 | 2957 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
2958 | .get_drvinfo = tun_get_drvinfo, |
2959 | .get_msglevel = tun_get_msglevel, | |
2960 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 2961 | .get_link = ethtool_op_get_link, |
eda29772 | 2962 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
2963 | .get_coalesce = tun_get_coalesce, |
2964 | .set_coalesce = tun_set_coalesce, | |
29ccc49d | 2965 | .get_link_ksettings = tun_get_link_ksettings, |
1da177e4 LT |
2966 | }; |
2967 | ||
1576d986 JW |
2968 | static int tun_queue_resize(struct tun_struct *tun) |
2969 | { | |
2970 | struct net_device *dev = tun->dev; | |
2971 | struct tun_file *tfile; | |
2972 | struct skb_array **arrays; | |
2973 | int n = tun->numqueues + tun->numdisabled; | |
2974 | int ret, i; | |
2975 | ||
12039046 | 2976 | arrays = kmalloc_array(n, sizeof(*arrays), GFP_KERNEL); |
1576d986 JW |
2977 | if (!arrays) |
2978 | return -ENOMEM; | |
2979 | ||
2980 | for (i = 0; i < tun->numqueues; i++) { | |
2981 | tfile = rtnl_dereference(tun->tfiles[i]); | |
2982 | arrays[i] = &tfile->tx_array; | |
2983 | } | |
2984 | list_for_each_entry(tfile, &tun->disabled, next) | |
2985 | arrays[i++] = &tfile->tx_array; | |
2986 | ||
2987 | ret = skb_array_resize_multiple(arrays, n, | |
2988 | dev->tx_queue_len, GFP_KERNEL); | |
2989 | ||
2990 | kfree(arrays); | |
2991 | return ret; | |
2992 | } | |
2993 | ||
2994 | static int tun_device_event(struct notifier_block *unused, | |
2995 | unsigned long event, void *ptr) | |
2996 | { | |
2997 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
2998 | struct tun_struct *tun = netdev_priv(dev); | |
2999 | ||
86dfb4ac CG |
3000 | if (dev->rtnl_link_ops != &tun_link_ops) |
3001 | return NOTIFY_DONE; | |
3002 | ||
1576d986 JW |
3003 | switch (event) { |
3004 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
3005 | if (tun_queue_resize(tun)) | |
3006 | return NOTIFY_BAD; | |
3007 | break; | |
3008 | default: | |
3009 | break; | |
3010 | } | |
3011 | ||
3012 | return NOTIFY_DONE; | |
3013 | } | |
3014 | ||
3015 | static struct notifier_block tun_notifier_block __read_mostly = { | |
3016 | .notifier_call = tun_device_event, | |
3017 | }; | |
79d17604 | 3018 | |
1da177e4 LT |
3019 | static int __init tun_init(void) |
3020 | { | |
3021 | int ret = 0; | |
3022 | ||
6b8a66ee | 3023 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
1da177e4 | 3024 | |
f019a7a5 | 3025 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 3026 | if (ret) { |
6b8a66ee | 3027 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 3028 | goto err_linkops; |
79d17604 PE |
3029 | } |
3030 | ||
1da177e4 | 3031 | ret = misc_register(&tun_miscdev); |
79d17604 | 3032 | if (ret) { |
6b8a66ee | 3033 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
3034 | goto err_misc; |
3035 | } | |
1576d986 | 3036 | |
5edfbd3c TZ |
3037 | ret = register_netdevice_notifier(&tun_notifier_block); |
3038 | if (ret) { | |
3039 | pr_err("Can't register netdevice notifier\n"); | |
3040 | goto err_notifier; | |
3041 | } | |
3042 | ||
f019a7a5 | 3043 | return 0; |
5edfbd3c TZ |
3044 | |
3045 | err_notifier: | |
3046 | misc_deregister(&tun_miscdev); | |
79d17604 | 3047 | err_misc: |
f019a7a5 EB |
3048 | rtnl_link_unregister(&tun_link_ops); |
3049 | err_linkops: | |
1da177e4 LT |
3050 | return ret; |
3051 | } | |
3052 | ||
3053 | static void tun_cleanup(void) | |
3054 | { | |
6aa20a22 | 3055 | misc_deregister(&tun_miscdev); |
f019a7a5 | 3056 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 3057 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
3058 | } |
3059 | ||
05c2828c MT |
3060 | /* Get an underlying socket object from tun file. Returns error unless file is |
3061 | * attached to a device. The returned object works like a packet socket, it | |
3062 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
3063 | * holding a reference to the file for as long as the socket is in use. */ | |
3064 | struct socket *tun_get_socket(struct file *file) | |
3065 | { | |
6e914fc7 | 3066 | struct tun_file *tfile; |
05c2828c MT |
3067 | if (file->f_op != &tun_fops) |
3068 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
3069 | tfile = file->private_data; |
3070 | if (!tfile) | |
05c2828c | 3071 | return ERR_PTR(-EBADFD); |
54f968d6 | 3072 | return &tfile->socket; |
05c2828c MT |
3073 | } |
3074 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
3075 | ||
83339c6b JW |
3076 | struct skb_array *tun_get_skb_array(struct file *file) |
3077 | { | |
3078 | struct tun_file *tfile; | |
3079 | ||
3080 | if (file->f_op != &tun_fops) | |
3081 | return ERR_PTR(-EINVAL); | |
3082 | tfile = file->private_data; | |
3083 | if (!tfile) | |
3084 | return ERR_PTR(-EBADFD); | |
3085 | return &tfile->tx_array; | |
3086 | } | |
3087 | EXPORT_SYMBOL_GPL(tun_get_skb_array); | |
3088 | ||
1da177e4 LT |
3089 | module_init(tun_init); |
3090 | module_exit(tun_cleanup); | |
3091 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
3092 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
3093 | MODULE_LICENSE("GPL"); | |
3094 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 3095 | MODULE_ALIAS("devname:net/tun"); |