]>
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; |
aec72f33 | 178 | bool napi_enabled; |
157f3ad9 | 179 | bool napi_frags_enabled; |
90e33d45 | 180 | struct mutex napi_mutex; /* Protects access to the above napi */ |
4008e97f JW |
181 | struct list_head next; |
182 | struct tun_struct *detached; | |
1576d986 | 183 | struct skb_array tx_array; |
631ab46b EB |
184 | }; |
185 | ||
96442e42 JW |
186 | struct tun_flow_entry { |
187 | struct hlist_node hash_link; | |
188 | struct rcu_head rcu; | |
189 | struct tun_struct *tun; | |
190 | ||
191 | u32 rxhash; | |
9bc88939 | 192 | u32 rps_rxhash; |
96442e42 JW |
193 | int queue_index; |
194 | unsigned long updated; | |
195 | }; | |
196 | ||
197 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
198 | ||
54f968d6 | 199 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 200 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
201 | * file were attached to a persist device. |
202 | */ | |
14daa021 | 203 | struct tun_struct { |
c8d68e6b JW |
204 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
205 | unsigned int numqueues; | |
f271b2cc | 206 | unsigned int flags; |
0625c883 EB |
207 | kuid_t owner; |
208 | kgid_t group; | |
14daa021 | 209 | |
14daa021 | 210 | struct net_device *dev; |
c8f44aff | 211 | netdev_features_t set_features; |
88255375 | 212 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
d591a1f3 | 213 | NETIF_F_TSO6) |
d9d52b51 | 214 | |
eaea34b2 | 215 | int align; |
d9d52b51 | 216 | int vnet_hdr_sz; |
54f968d6 JW |
217 | int sndbuf; |
218 | struct tap_filter txflt; | |
219 | struct sock_fprog fprog; | |
220 | /* protected by rtnl lock */ | |
221 | bool filter_attached; | |
14daa021 RR |
222 | #ifdef TUN_DEBUG |
223 | int debug; | |
1da177e4 | 224 | #endif |
96442e42 | 225 | spinlock_t lock; |
96442e42 JW |
226 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
227 | struct timer_list flow_gc_timer; | |
228 | unsigned long ageing_time; | |
4008e97f JW |
229 | unsigned int numdisabled; |
230 | struct list_head disabled; | |
5dbbaf2d | 231 | void *security; |
b8732fb7 | 232 | u32 flow_count; |
5503fcec | 233 | u32 rx_batched; |
608b9977 | 234 | struct tun_pcpu_stats __percpu *pcpu_stats; |
761876c8 | 235 | struct bpf_prog __rcu *xdp_prog; |
14daa021 | 236 | }; |
1da177e4 | 237 | |
94317099 PP |
238 | static int tun_napi_receive(struct napi_struct *napi, int budget) |
239 | { | |
240 | struct tun_file *tfile = container_of(napi, struct tun_file, napi); | |
241 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
242 | struct sk_buff_head process_queue; | |
243 | struct sk_buff *skb; | |
244 | int received = 0; | |
245 | ||
246 | __skb_queue_head_init(&process_queue); | |
247 | ||
248 | spin_lock(&queue->lock); | |
249 | skb_queue_splice_tail_init(queue, &process_queue); | |
250 | spin_unlock(&queue->lock); | |
251 | ||
252 | while (received < budget && (skb = __skb_dequeue(&process_queue))) { | |
253 | napi_gro_receive(napi, skb); | |
254 | ++received; | |
255 | } | |
256 | ||
257 | if (!skb_queue_empty(&process_queue)) { | |
258 | spin_lock(&queue->lock); | |
259 | skb_queue_splice(&process_queue, queue); | |
260 | spin_unlock(&queue->lock); | |
261 | } | |
262 | ||
263 | return received; | |
264 | } | |
265 | ||
266 | static int tun_napi_poll(struct napi_struct *napi, int budget) | |
267 | { | |
268 | unsigned int received; | |
269 | ||
270 | received = tun_napi_receive(napi, budget); | |
271 | ||
272 | if (received < budget) | |
273 | napi_complete_done(napi, received); | |
274 | ||
275 | return received; | |
276 | } | |
277 | ||
278 | static void tun_napi_init(struct tun_struct *tun, struct tun_file *tfile, | |
157f3ad9 | 279 | bool napi_en, bool napi_frags) |
94317099 | 280 | { |
aec72f33 | 281 | tfile->napi_enabled = napi_en; |
157f3ad9 | 282 | tfile->napi_frags_enabled = napi_en && napi_frags; |
94317099 | 283 | if (napi_en) { |
1b64b21d PP |
284 | netif_tx_napi_add(tun->dev, &tfile->napi, tun_napi_poll, |
285 | NAPI_POLL_WEIGHT); | |
94317099 PP |
286 | napi_enable(&tfile->napi); |
287 | } | |
288 | } | |
289 | ||
be6c1bf6 | 290 | static void tun_napi_disable(struct tun_file *tfile) |
94317099 | 291 | { |
aec72f33 | 292 | if (tfile->napi_enabled) |
94317099 PP |
293 | napi_disable(&tfile->napi); |
294 | } | |
295 | ||
be6c1bf6 | 296 | static void tun_napi_del(struct tun_file *tfile) |
94317099 | 297 | { |
aec72f33 | 298 | if (tfile->napi_enabled) |
94317099 PP |
299 | netif_napi_del(&tfile->napi); |
300 | } | |
301 | ||
157f3ad9 | 302 | static bool tun_napi_frags_enabled(const struct tun_file *tfile) |
90e33d45 | 303 | { |
157f3ad9 | 304 | return tfile->napi_frags_enabled; |
90e33d45 PP |
305 | } |
306 | ||
8b8e658b GK |
307 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
308 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
309 | { | |
310 | return tun->flags & TUN_VNET_BE ? false : | |
311 | virtio_legacy_is_little_endian(); | |
312 | } | |
313 | ||
314 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
315 | { | |
316 | int be = !!(tun->flags & TUN_VNET_BE); | |
317 | ||
318 | if (put_user(be, argp)) | |
319 | return -EFAULT; | |
320 | ||
321 | return 0; | |
322 | } | |
323 | ||
324 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
325 | { | |
326 | int be; | |
327 | ||
328 | if (get_user(be, argp)) | |
329 | return -EFAULT; | |
330 | ||
331 | if (be) | |
332 | tun->flags |= TUN_VNET_BE; | |
333 | else | |
334 | tun->flags &= ~TUN_VNET_BE; | |
335 | ||
336 | return 0; | |
337 | } | |
338 | #else | |
339 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
340 | { | |
341 | return virtio_legacy_is_little_endian(); | |
342 | } | |
343 | ||
344 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
345 | { | |
346 | return -EINVAL; | |
347 | } | |
348 | ||
349 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
350 | { | |
351 | return -EINVAL; | |
352 | } | |
353 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
354 | ||
25bd55bb GK |
355 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
356 | { | |
7d824109 | 357 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 358 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
359 | } |
360 | ||
56f0dcc5 MT |
361 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
362 | { | |
25bd55bb | 363 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
364 | } |
365 | ||
366 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
367 | { | |
25bd55bb | 368 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
369 | } |
370 | ||
96442e42 JW |
371 | static inline u32 tun_hashfn(u32 rxhash) |
372 | { | |
373 | return rxhash & 0x3ff; | |
374 | } | |
375 | ||
376 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
377 | { | |
378 | struct tun_flow_entry *e; | |
96442e42 | 379 | |
b67bfe0d | 380 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
381 | if (e->rxhash == rxhash) |
382 | return e; | |
383 | } | |
384 | return NULL; | |
385 | } | |
386 | ||
387 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
388 | struct hlist_head *head, | |
389 | u32 rxhash, u16 queue_index) | |
390 | { | |
9fdc6bef ED |
391 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
392 | ||
96442e42 JW |
393 | if (e) { |
394 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
395 | rxhash, queue_index); | |
396 | e->updated = jiffies; | |
397 | e->rxhash = rxhash; | |
9bc88939 | 398 | e->rps_rxhash = 0; |
96442e42 JW |
399 | e->queue_index = queue_index; |
400 | e->tun = tun; | |
401 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 402 | ++tun->flow_count; |
96442e42 JW |
403 | } |
404 | return e; | |
405 | } | |
406 | ||
96442e42 JW |
407 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
408 | { | |
409 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
410 | e->rxhash, e->queue_index); | |
411 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 412 | kfree_rcu(e, rcu); |
b8732fb7 | 413 | --tun->flow_count; |
96442e42 JW |
414 | } |
415 | ||
416 | static void tun_flow_flush(struct tun_struct *tun) | |
417 | { | |
418 | int i; | |
419 | ||
420 | spin_lock_bh(&tun->lock); | |
421 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
422 | struct tun_flow_entry *e; | |
b67bfe0d | 423 | struct hlist_node *n; |
96442e42 | 424 | |
b67bfe0d | 425 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
426 | tun_flow_delete(tun, e); |
427 | } | |
428 | spin_unlock_bh(&tun->lock); | |
429 | } | |
430 | ||
431 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
432 | { | |
433 | int i; | |
434 | ||
435 | spin_lock_bh(&tun->lock); | |
436 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
437 | struct tun_flow_entry *e; | |
b67bfe0d | 438 | struct hlist_node *n; |
96442e42 | 439 | |
b67bfe0d | 440 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
441 | if (e->queue_index == queue_index) |
442 | tun_flow_delete(tun, e); | |
443 | } | |
444 | } | |
445 | spin_unlock_bh(&tun->lock); | |
446 | } | |
447 | ||
e99e88a9 | 448 | static void tun_flow_cleanup(struct timer_list *t) |
96442e42 | 449 | { |
e99e88a9 | 450 | struct tun_struct *tun = from_timer(tun, t, flow_gc_timer); |
96442e42 JW |
451 | unsigned long delay = tun->ageing_time; |
452 | unsigned long next_timer = jiffies + delay; | |
453 | unsigned long count = 0; | |
454 | int i; | |
455 | ||
456 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
457 | ||
7dbfb4ef | 458 | spin_lock(&tun->lock); |
96442e42 JW |
459 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { |
460 | struct tun_flow_entry *e; | |
b67bfe0d | 461 | struct hlist_node *n; |
96442e42 | 462 | |
b67bfe0d | 463 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 | 464 | unsigned long this_timer; |
81d98fa4 | 465 | |
96442e42 | 466 | this_timer = e->updated + delay; |
81d98fa4 | 467 | if (time_before_eq(this_timer, jiffies)) { |
96442e42 | 468 | tun_flow_delete(tun, e); |
81d98fa4 ED |
469 | continue; |
470 | } | |
471 | count++; | |
472 | if (time_before(this_timer, next_timer)) | |
96442e42 JW |
473 | next_timer = this_timer; |
474 | } | |
475 | } | |
476 | ||
477 | if (count) | |
478 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
7dbfb4ef | 479 | spin_unlock(&tun->lock); |
96442e42 JW |
480 | } |
481 | ||
49974420 | 482 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 483 | struct tun_file *tfile) |
96442e42 JW |
484 | { |
485 | struct hlist_head *head; | |
486 | struct tun_flow_entry *e; | |
487 | unsigned long delay = tun->ageing_time; | |
9e85722d | 488 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
489 | |
490 | if (!rxhash) | |
491 | return; | |
492 | else | |
493 | head = &tun->flows[tun_hashfn(rxhash)]; | |
494 | ||
495 | rcu_read_lock(); | |
496 | ||
9e85722d JW |
497 | /* We may get a very small possibility of OOO during switching, not |
498 | * worth to optimize.*/ | |
499 | if (tun->numqueues == 1 || tfile->detached) | |
96442e42 JW |
500 | goto unlock; |
501 | ||
502 | e = tun_flow_find(head, rxhash); | |
503 | if (likely(e)) { | |
504 | /* TODO: keep queueing to old queue until it's empty? */ | |
505 | e->queue_index = queue_index; | |
506 | e->updated = jiffies; | |
9bc88939 | 507 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
508 | } else { |
509 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
510 | if (!tun_flow_find(head, rxhash) && |
511 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
512 | tun_flow_create(tun, head, rxhash, queue_index); |
513 | ||
514 | if (!timer_pending(&tun->flow_gc_timer)) | |
515 | mod_timer(&tun->flow_gc_timer, | |
516 | round_jiffies_up(jiffies + delay)); | |
517 | spin_unlock_bh(&tun->lock); | |
518 | } | |
519 | ||
520 | unlock: | |
521 | rcu_read_unlock(); | |
522 | } | |
523 | ||
9bc88939 TH |
524 | /** |
525 | * Save the hash received in the stack receive path and update the | |
526 | * flow_hash table accordingly. | |
527 | */ | |
528 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
529 | { | |
567e4b79 | 530 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 531 | e->rps_rxhash = hash; |
9bc88939 TH |
532 | } |
533 | ||
c8d68e6b | 534 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 535 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
536 | * the rxq based on the txq where the last packet of the flow comes. As |
537 | * the userspace application move between processors, we may get a | |
538 | * different rxq no. here. If we could not get rxhash, then we would | |
539 | * hope the rxq no. may help here. | |
540 | */ | |
f663dd9a | 541 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 542 | void *accel_priv, select_queue_fallback_t fallback) |
c8d68e6b JW |
543 | { |
544 | struct tun_struct *tun = netdev_priv(dev); | |
96442e42 | 545 | struct tun_flow_entry *e; |
c8d68e6b JW |
546 | u32 txq = 0; |
547 | u32 numqueues = 0; | |
548 | ||
549 | rcu_read_lock(); | |
6aa7de05 | 550 | numqueues = READ_ONCE(tun->numqueues); |
c8d68e6b | 551 | |
feec084a | 552 | txq = __skb_get_hash_symmetric(skb); |
c8d68e6b | 553 | if (txq) { |
96442e42 | 554 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 555 | if (e) { |
9bc88939 | 556 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 557 | txq = e->queue_index; |
9bc88939 | 558 | } else |
96442e42 JW |
559 | /* use multiply and shift instead of expensive divide */ |
560 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
561 | } else if (likely(skb_rx_queue_recorded(skb))) { |
562 | txq = skb_get_rx_queue(skb); | |
563 | while (unlikely(txq >= numqueues)) | |
564 | txq -= numqueues; | |
565 | } | |
566 | ||
567 | rcu_read_unlock(); | |
568 | return txq; | |
569 | } | |
570 | ||
cde8b15f JW |
571 | static inline bool tun_not_capable(struct tun_struct *tun) |
572 | { | |
573 | const struct cred *cred = current_cred(); | |
c260b772 | 574 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
575 | |
576 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
577 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 578 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
579 | } |
580 | ||
c8d68e6b JW |
581 | static void tun_set_real_num_queues(struct tun_struct *tun) |
582 | { | |
583 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
584 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
585 | } | |
586 | ||
4008e97f JW |
587 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
588 | { | |
589 | tfile->detached = tun; | |
590 | list_add_tail(&tfile->next, &tun->disabled); | |
591 | ++tun->numdisabled; | |
592 | } | |
593 | ||
d32649d1 | 594 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
595 | { |
596 | struct tun_struct *tun = tfile->detached; | |
597 | ||
598 | tfile->detached = NULL; | |
599 | list_del_init(&tfile->next); | |
600 | --tun->numdisabled; | |
601 | return tun; | |
602 | } | |
603 | ||
4bfb0513 JW |
604 | static void tun_queue_purge(struct tun_file *tfile) |
605 | { | |
1576d986 JW |
606 | struct sk_buff *skb; |
607 | ||
608 | while ((skb = skb_array_consume(&tfile->tx_array)) != NULL) | |
609 | kfree_skb(skb); | |
610 | ||
5503fcec | 611 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
612 | skb_queue_purge(&tfile->sk.sk_error_queue); |
613 | } | |
614 | ||
c8d68e6b JW |
615 | static void __tun_detach(struct tun_file *tfile, bool clean) |
616 | { | |
617 | struct tun_file *ntfile; | |
618 | struct tun_struct *tun; | |
c8d68e6b | 619 | |
b8deabd3 JW |
620 | tun = rtnl_dereference(tfile->tun); |
621 | ||
94317099 | 622 | if (tun && clean) { |
be6c1bf6 ED |
623 | tun_napi_disable(tfile); |
624 | tun_napi_del(tfile); | |
94317099 PP |
625 | } |
626 | ||
9e85722d | 627 | if (tun && !tfile->detached) { |
c8d68e6b JW |
628 | u16 index = tfile->queue_index; |
629 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
630 | |
631 | rcu_assign_pointer(tun->tfiles[index], | |
632 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 633 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
634 | ntfile->queue_index = index; |
635 | ||
636 | --tun->numqueues; | |
9e85722d | 637 | if (clean) { |
c956674b | 638 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 639 | sock_put(&tfile->sk); |
9e85722d | 640 | } else |
4008e97f | 641 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
642 | |
643 | synchronize_net(); | |
96442e42 | 644 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 645 | /* Drop read queue */ |
4bfb0513 | 646 | tun_queue_purge(tfile); |
c8d68e6b | 647 | tun_set_real_num_queues(tun); |
dd38bd85 | 648 | } else if (tfile->detached && clean) { |
4008e97f | 649 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
650 | sock_put(&tfile->sk); |
651 | } | |
c8d68e6b JW |
652 | |
653 | if (clean) { | |
af668b3c MT |
654 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
655 | netif_carrier_off(tun->dev); | |
656 | ||
40630b82 | 657 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 658 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 659 | unregister_netdevice(tun->dev); |
af668b3c | 660 | } |
03230e5a | 661 | skb_array_cleanup(&tfile->tx_array); |
140e807d | 662 | sock_put(&tfile->sk); |
c8d68e6b JW |
663 | } |
664 | } | |
665 | ||
666 | static void tun_detach(struct tun_file *tfile, bool clean) | |
667 | { | |
668 | rtnl_lock(); | |
669 | __tun_detach(tfile, clean); | |
670 | rtnl_unlock(); | |
671 | } | |
672 | ||
673 | static void tun_detach_all(struct net_device *dev) | |
674 | { | |
675 | struct tun_struct *tun = netdev_priv(dev); | |
761876c8 | 676 | struct bpf_prog *xdp_prog = rtnl_dereference(tun->xdp_prog); |
4008e97f | 677 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
678 | int i, n = tun->numqueues; |
679 | ||
680 | for (i = 0; i < n; i++) { | |
b8deabd3 | 681 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 682 | BUG_ON(!tfile); |
be6c1bf6 | 683 | tun_napi_disable(tfile); |
addf8fc4 | 684 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 685 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 686 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
687 | --tun->numqueues; |
688 | } | |
9e85722d | 689 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 690 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 691 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 692 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 693 | } |
c8d68e6b JW |
694 | BUG_ON(tun->numqueues != 0); |
695 | ||
696 | synchronize_net(); | |
697 | for (i = 0; i < n; i++) { | |
b8deabd3 | 698 | tfile = rtnl_dereference(tun->tfiles[i]); |
be6c1bf6 | 699 | tun_napi_del(tfile); |
c8d68e6b | 700 | /* Drop read queue */ |
4bfb0513 | 701 | tun_queue_purge(tfile); |
c8d68e6b JW |
702 | sock_put(&tfile->sk); |
703 | } | |
4008e97f JW |
704 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
705 | tun_enable_queue(tfile); | |
4bfb0513 | 706 | tun_queue_purge(tfile); |
4008e97f JW |
707 | sock_put(&tfile->sk); |
708 | } | |
709 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 710 | |
761876c8 JW |
711 | if (xdp_prog) |
712 | bpf_prog_put(xdp_prog); | |
713 | ||
40630b82 | 714 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 715 | module_put(THIS_MODULE); |
c8d68e6b JW |
716 | } |
717 | ||
94317099 | 718 | static int tun_attach(struct tun_struct *tun, struct file *file, |
fdf434d1 YY |
719 | bool skip_filter, bool napi, bool napi_frags, |
720 | bool publish_tun) | |
a7385ba2 | 721 | { |
631ab46b | 722 | struct tun_file *tfile = file->private_data; |
1576d986 | 723 | struct net_device *dev = tun->dev; |
38231b7a | 724 | int err; |
a7385ba2 | 725 | |
5dbbaf2d PM |
726 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
727 | if (err < 0) | |
728 | goto out; | |
729 | ||
38231b7a | 730 | err = -EINVAL; |
9e85722d | 731 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
732 | goto out; |
733 | ||
734 | err = -EBUSY; | |
40630b82 | 735 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
736 | goto out; |
737 | ||
738 | err = -E2BIG; | |
4008e97f JW |
739 | if (!tfile->detached && |
740 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
741 | goto out; |
742 | ||
743 | err = 0; | |
54f968d6 | 744 | |
92d4ea6e | 745 | /* Re-attach the filter to persist device */ |
849c9b6f | 746 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
747 | lock_sock(tfile->socket.sk); |
748 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
749 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
750 | if (!err) |
751 | goto out; | |
752 | } | |
1576d986 JW |
753 | |
754 | if (!tfile->detached && | |
48ed2557 | 755 | skb_array_resize(&tfile->tx_array, dev->tx_queue_len, GFP_KERNEL)) { |
1576d986 JW |
756 | err = -ENOMEM; |
757 | goto out; | |
758 | } | |
759 | ||
c8d68e6b | 760 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 761 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
94317099 | 762 | if (tfile->detached) { |
4008e97f | 763 | tun_enable_queue(tfile); |
94317099 | 764 | } else { |
4008e97f | 765 | sock_hold(&tfile->sk); |
157f3ad9 | 766 | tun_napi_init(tun, tfile, napi, napi_frags); |
94317099 | 767 | } |
4008e97f | 768 | |
c8d68e6b JW |
769 | /* device is allowed to go away first, so no need to hold extra |
770 | * refcnt. | |
771 | */ | |
772 | ||
44b48ef3 SF |
773 | /* Publish tfile->tun and tun->tfiles only after we've fully |
774 | * initialized tfile; otherwise we risk using half-initialized | |
775 | * object. | |
776 | */ | |
fdf434d1 YY |
777 | if (publish_tun) |
778 | rcu_assign_pointer(tfile->tun, tun); | |
44b48ef3 SF |
779 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
780 | tun->numqueues++; | |
b9cbbf6b | 781 | tun_set_real_num_queues(tun); |
c8d68e6b JW |
782 | out: |
783 | return err; | |
631ab46b EB |
784 | } |
785 | ||
9484dc74 | 786 | static struct tun_struct *tun_get(struct tun_file *tfile) |
631ab46b | 787 | { |
6e914fc7 | 788 | struct tun_struct *tun; |
c70f1829 | 789 | |
6e914fc7 JW |
790 | rcu_read_lock(); |
791 | tun = rcu_dereference(tfile->tun); | |
792 | if (tun) | |
793 | dev_hold(tun->dev); | |
794 | rcu_read_unlock(); | |
c70f1829 EB |
795 | |
796 | return tun; | |
631ab46b EB |
797 | } |
798 | ||
631ab46b EB |
799 | static void tun_put(struct tun_struct *tun) |
800 | { | |
6e914fc7 | 801 | dev_put(tun->dev); |
631ab46b EB |
802 | } |
803 | ||
6b8a66ee | 804 | /* TAP filtering */ |
f271b2cc MK |
805 | static void addr_hash_set(u32 *mask, const u8 *addr) |
806 | { | |
807 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
808 | mask[n >> 5] |= (1 << (n & 31)); | |
809 | } | |
810 | ||
811 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
812 | { | |
813 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
814 | return mask[n >> 5] & (1 << (n & 31)); | |
815 | } | |
816 | ||
817 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
818 | { | |
819 | struct { u8 u[ETH_ALEN]; } *addr; | |
820 | struct tun_filter uf; | |
821 | int err, alen, n, nexact; | |
822 | ||
823 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
824 | return -EFAULT; | |
825 | ||
826 | if (!uf.count) { | |
827 | /* Disabled */ | |
828 | filter->count = 0; | |
829 | return 0; | |
830 | } | |
831 | ||
832 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
833 | addr = memdup_user(arg + sizeof(uf), alen); |
834 | if (IS_ERR(addr)) | |
835 | return PTR_ERR(addr); | |
f271b2cc MK |
836 | |
837 | /* The filter is updated without holding any locks. Which is | |
838 | * perfectly safe. We disable it first and in the worst | |
839 | * case we'll accept a few undesired packets. */ | |
840 | filter->count = 0; | |
841 | wmb(); | |
842 | ||
843 | /* Use first set of addresses as an exact filter */ | |
844 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
845 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
846 | ||
847 | nexact = n; | |
848 | ||
cfbf84fc AW |
849 | /* Remaining multicast addresses are hashed, |
850 | * unicast will leave the filter disabled. */ | |
f271b2cc | 851 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
852 | for (; n < uf.count; n++) { |
853 | if (!is_multicast_ether_addr(addr[n].u)) { | |
854 | err = 0; /* no filter */ | |
3b8d2a69 | 855 | goto free_addr; |
cfbf84fc | 856 | } |
f271b2cc | 857 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 858 | } |
f271b2cc MK |
859 | |
860 | /* For ALLMULTI just set the mask to all ones. | |
861 | * This overrides the mask populated above. */ | |
862 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
863 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
864 | ||
865 | /* Now enable the filter */ | |
866 | wmb(); | |
867 | filter->count = nexact; | |
868 | ||
869 | /* Return the number of exact filters */ | |
870 | err = nexact; | |
3b8d2a69 | 871 | free_addr: |
f271b2cc MK |
872 | kfree(addr); |
873 | return err; | |
874 | } | |
875 | ||
876 | /* Returns: 0 - drop, !=0 - accept */ | |
877 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
878 | { | |
879 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
880 | * at this point. */ | |
881 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
882 | int i; | |
883 | ||
884 | /* Exact match */ | |
885 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 886 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
887 | return 1; |
888 | ||
889 | /* Inexact match (multicast only) */ | |
890 | if (is_multicast_ether_addr(eh->h_dest)) | |
891 | return addr_hash_test(filter->mask, eh->h_dest); | |
892 | ||
893 | return 0; | |
894 | } | |
895 | ||
896 | /* | |
897 | * Checks whether the packet is accepted or not. | |
898 | * Returns: 0 - drop, !=0 - accept | |
899 | */ | |
900 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
901 | { | |
902 | if (!filter->count) | |
903 | return 1; | |
904 | ||
905 | return run_filter(filter, skb); | |
906 | } | |
907 | ||
1da177e4 LT |
908 | /* Network device part of the driver */ |
909 | ||
7282d491 | 910 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 911 | |
c70f1829 EB |
912 | /* Net device detach from fd. */ |
913 | static void tun_net_uninit(struct net_device *dev) | |
914 | { | |
c8d68e6b | 915 | tun_detach_all(dev); |
c70f1829 EB |
916 | } |
917 | ||
1da177e4 LT |
918 | /* Net device open. */ |
919 | static int tun_net_open(struct net_device *dev) | |
920 | { | |
c8d68e6b | 921 | netif_tx_start_all_queues(dev); |
b20e2d54 | 922 | |
1da177e4 LT |
923 | return 0; |
924 | } | |
925 | ||
926 | /* Net device close. */ | |
927 | static int tun_net_close(struct net_device *dev) | |
928 | { | |
c8d68e6b | 929 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
930 | return 0; |
931 | } | |
932 | ||
933 | /* Net device start xmit */ | |
424efe9c | 934 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
1da177e4 LT |
935 | { |
936 | struct tun_struct *tun = netdev_priv(dev); | |
c8d68e6b | 937 | int txq = skb->queue_mapping; |
6e914fc7 | 938 | struct tun_file *tfile; |
fa35864e | 939 | u32 numqueues = 0; |
1da177e4 | 940 | |
6e914fc7 | 941 | rcu_read_lock(); |
c8d68e6b | 942 | tfile = rcu_dereference(tun->tfiles[txq]); |
6aa7de05 | 943 | numqueues = READ_ONCE(tun->numqueues); |
c8d68e6b | 944 | |
1da177e4 | 945 | /* Drop packet if interface is not attached */ |
fa35864e | 946 | if (txq >= numqueues) |
1da177e4 LT |
947 | goto drop; |
948 | ||
3df97ba8 JW |
949 | #ifdef CONFIG_RPS |
950 | if (numqueues == 1 && static_key_false(&rps_needed)) { | |
9bc88939 TH |
951 | /* Select queue was not called for the skbuff, so we extract the |
952 | * RPS hash and save it into the flow_table here. | |
953 | */ | |
954 | __u32 rxhash; | |
955 | ||
feec084a | 956 | rxhash = __skb_get_hash_symmetric(skb); |
9bc88939 TH |
957 | if (rxhash) { |
958 | struct tun_flow_entry *e; | |
959 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
960 | rxhash); | |
961 | if (e) | |
962 | tun_flow_save_rps_rxhash(e, rxhash); | |
963 | } | |
964 | } | |
3df97ba8 | 965 | #endif |
9bc88939 | 966 | |
6e914fc7 JW |
967 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
968 | ||
c8d68e6b JW |
969 | BUG_ON(!tfile); |
970 | ||
f271b2cc MK |
971 | /* Drop if the filter does not like it. |
972 | * This is a noop if the filter is disabled. | |
973 | * Filter can be enabled only for the TAP devices. */ | |
974 | if (!check_filter(&tun->txflt, skb)) | |
975 | goto drop; | |
976 | ||
54f968d6 JW |
977 | if (tfile->socket.sk->sk_filter && |
978 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
979 | goto drop; |
980 | ||
1f8b977a | 981 | if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) |
7bf66305 JW |
982 | goto drop; |
983 | ||
7b996243 | 984 | skb_tx_timestamp(skb); |
eda29772 | 985 | |
0110d6f2 | 986 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
987 | * for indefinite time. |
988 | */ | |
0110d6f2 MT |
989 | skb_orphan(skb); |
990 | ||
f8af75f3 ED |
991 | nf_reset(skb); |
992 | ||
1576d986 JW |
993 | if (skb_array_produce(&tfile->tx_array, skb)) |
994 | goto drop; | |
1da177e4 LT |
995 | |
996 | /* Notify and wake up reader process */ | |
54f968d6 JW |
997 | if (tfile->flags & TUN_FASYNC) |
998 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 999 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
1000 | |
1001 | rcu_read_unlock(); | |
6ed10654 | 1002 | return NETDEV_TX_OK; |
1da177e4 LT |
1003 | |
1004 | drop: | |
608b9977 | 1005 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
149d36f7 | 1006 | skb_tx_error(skb); |
1da177e4 | 1007 | kfree_skb(skb); |
6e914fc7 | 1008 | rcu_read_unlock(); |
baeababb | 1009 | return NET_XMIT_DROP; |
1da177e4 LT |
1010 | } |
1011 | ||
f271b2cc | 1012 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 1013 | { |
f271b2cc MK |
1014 | /* |
1015 | * This callback is supposed to deal with mc filter in | |
1016 | * _rx_ path and has nothing to do with the _tx_ path. | |
1017 | * In rx path we always accept everything userspace gives us. | |
1018 | */ | |
1da177e4 LT |
1019 | } |
1020 | ||
c8f44aff MM |
1021 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
1022 | netdev_features_t features) | |
88255375 MM |
1023 | { |
1024 | struct tun_struct *tun = netdev_priv(dev); | |
1025 | ||
1026 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
1027 | } | |
bebd097a NH |
1028 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1029 | static void tun_poll_controller(struct net_device *dev) | |
1030 | { | |
1031 | /* | |
1032 | * Tun only receives frames when: | |
1033 | * 1) the char device endpoint gets data from user space | |
1034 | * 2) the tun socket gets a sendmsg call from user space | |
94317099 PP |
1035 | * If NAPI is not enabled, since both of those are synchronous |
1036 | * operations, we are guaranteed never to have pending data when we poll | |
1037 | * for it so there is nothing to do here but return. | |
bebd097a NH |
1038 | * We need this though so netpoll recognizes us as an interface that |
1039 | * supports polling, which enables bridge devices in virt setups to | |
1040 | * still use netconsole | |
94317099 | 1041 | * If NAPI is enabled, however, we need to schedule polling for all |
90e33d45 PP |
1042 | * queues unless we are using napi_gro_frags(), which we call in |
1043 | * process context and not in NAPI context. | |
bebd097a | 1044 | */ |
94317099 PP |
1045 | struct tun_struct *tun = netdev_priv(dev); |
1046 | ||
1047 | if (tun->flags & IFF_NAPI) { | |
1048 | struct tun_file *tfile; | |
1049 | int i; | |
1050 | ||
1051 | rcu_read_lock(); | |
1052 | for (i = 0; i < tun->numqueues; i++) { | |
1053 | tfile = rcu_dereference(tun->tfiles[i]); | |
157f3ad9 ED |
1054 | if (!tun_napi_frags_enabled(tfile) && |
1055 | tfile->napi_enabled) | |
aec72f33 | 1056 | napi_schedule(&tfile->napi); |
94317099 PP |
1057 | } |
1058 | rcu_read_unlock(); | |
1059 | } | |
bebd097a NH |
1060 | return; |
1061 | } | |
1062 | #endif | |
eaea34b2 PA |
1063 | |
1064 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
1065 | { | |
1066 | struct tun_struct *tun = netdev_priv(dev); | |
1067 | ||
1068 | if (new_hr < NET_SKB_PAD) | |
1069 | new_hr = NET_SKB_PAD; | |
1070 | ||
1071 | tun->align = new_hr; | |
1072 | } | |
1073 | ||
bc1f4470 | 1074 | static void |
608b9977 PA |
1075 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
1076 | { | |
1077 | u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; | |
1078 | struct tun_struct *tun = netdev_priv(dev); | |
1079 | struct tun_pcpu_stats *p; | |
1080 | int i; | |
1081 | ||
1082 | for_each_possible_cpu(i) { | |
1083 | u64 rxpackets, rxbytes, txpackets, txbytes; | |
1084 | unsigned int start; | |
1085 | ||
1086 | p = per_cpu_ptr(tun->pcpu_stats, i); | |
1087 | do { | |
1088 | start = u64_stats_fetch_begin(&p->syncp); | |
1089 | rxpackets = p->rx_packets; | |
1090 | rxbytes = p->rx_bytes; | |
1091 | txpackets = p->tx_packets; | |
1092 | txbytes = p->tx_bytes; | |
1093 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
1094 | ||
1095 | stats->rx_packets += rxpackets; | |
1096 | stats->rx_bytes += rxbytes; | |
1097 | stats->tx_packets += txpackets; | |
1098 | stats->tx_bytes += txbytes; | |
1099 | ||
1100 | /* u32 counters */ | |
1101 | rx_dropped += p->rx_dropped; | |
1102 | rx_frame_errors += p->rx_frame_errors; | |
1103 | tx_dropped += p->tx_dropped; | |
1104 | } | |
1105 | stats->rx_dropped = rx_dropped; | |
1106 | stats->rx_frame_errors = rx_frame_errors; | |
1107 | stats->tx_dropped = tx_dropped; | |
608b9977 PA |
1108 | } |
1109 | ||
761876c8 JW |
1110 | static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog, |
1111 | struct netlink_ext_ack *extack) | |
1112 | { | |
1113 | struct tun_struct *tun = netdev_priv(dev); | |
1114 | struct bpf_prog *old_prog; | |
1115 | ||
1116 | old_prog = rtnl_dereference(tun->xdp_prog); | |
1117 | rcu_assign_pointer(tun->xdp_prog, prog); | |
1118 | if (old_prog) | |
1119 | bpf_prog_put(old_prog); | |
1120 | ||
1121 | return 0; | |
1122 | } | |
1123 | ||
1124 | static u32 tun_xdp_query(struct net_device *dev) | |
1125 | { | |
1126 | struct tun_struct *tun = netdev_priv(dev); | |
1127 | const struct bpf_prog *xdp_prog; | |
1128 | ||
1129 | xdp_prog = rtnl_dereference(tun->xdp_prog); | |
1130 | if (xdp_prog) | |
1131 | return xdp_prog->aux->id; | |
1132 | ||
1133 | return 0; | |
1134 | } | |
1135 | ||
f4e63525 | 1136 | static int tun_xdp(struct net_device *dev, struct netdev_bpf *xdp) |
761876c8 JW |
1137 | { |
1138 | switch (xdp->command) { | |
1139 | case XDP_SETUP_PROG: | |
1140 | return tun_xdp_set(dev, xdp->prog, xdp->extack); | |
1141 | case XDP_QUERY_PROG: | |
1142 | xdp->prog_id = tun_xdp_query(dev); | |
1143 | xdp->prog_attached = !!xdp->prog_id; | |
1144 | return 0; | |
1145 | default: | |
1146 | return -EINVAL; | |
1147 | } | |
1148 | } | |
1149 | ||
76db66f7 ND |
1150 | static int tun_net_change_carrier(struct net_device *dev, bool new_carrier) |
1151 | { | |
1152 | if (new_carrier) { | |
1153 | struct tun_struct *tun = netdev_priv(dev); | |
1154 | ||
1155 | if (!tun->numqueues) | |
1156 | return -EPERM; | |
1157 | ||
1158 | netif_carrier_on(dev); | |
1159 | } else { | |
1160 | netif_carrier_off(dev); | |
1161 | } | |
1162 | return 0; | |
1163 | } | |
1164 | ||
758e43b7 | 1165 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 1166 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1167 | .ndo_open = tun_net_open, |
1168 | .ndo_stop = tun_net_close, | |
00829823 | 1169 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1170 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1171 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1172 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1173 | .ndo_poll_controller = tun_poll_controller, | |
1174 | #endif | |
eaea34b2 | 1175 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1176 | .ndo_get_stats64 = tun_net_get_stats64, |
76db66f7 | 1177 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1178 | }; |
1179 | ||
1180 | static const struct net_device_ops tap_netdev_ops = { | |
c70f1829 | 1181 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1182 | .ndo_open = tun_net_open, |
1183 | .ndo_stop = tun_net_close, | |
00829823 | 1184 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1185 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1186 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1187 | .ndo_set_mac_address = eth_mac_addr, |
1188 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1189 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1190 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1191 | .ndo_poll_controller = tun_poll_controller, | |
1192 | #endif | |
5e52796a | 1193 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1194 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1195 | .ndo_get_stats64 = tun_net_get_stats64, |
f4e63525 | 1196 | .ndo_bpf = tun_xdp, |
76db66f7 | 1197 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1198 | }; |
1199 | ||
944a1376 | 1200 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1201 | { |
1202 | int i; | |
1203 | ||
96442e42 JW |
1204 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1205 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1206 | ||
1207 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
e99e88a9 KC |
1208 | timer_setup(&tun->flow_gc_timer, tun_flow_cleanup, 0); |
1209 | mod_timer(&tun->flow_gc_timer, | |
1210 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1211 | } |
1212 | ||
1213 | static void tun_flow_uninit(struct tun_struct *tun) | |
1214 | { | |
1215 | del_timer_sync(&tun->flow_gc_timer); | |
1216 | tun_flow_flush(tun); | |
96442e42 JW |
1217 | } |
1218 | ||
91572088 JW |
1219 | #define MIN_MTU 68 |
1220 | #define MAX_MTU 65535 | |
1221 | ||
1da177e4 LT |
1222 | /* Initialize net device. */ |
1223 | static void tun_net_init(struct net_device *dev) | |
1224 | { | |
1225 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1226 | |
1da177e4 | 1227 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1228 | case IFF_TUN: |
758e43b7 SH |
1229 | dev->netdev_ops = &tun_netdev_ops; |
1230 | ||
1da177e4 LT |
1231 | /* Point-to-Point TUN Device */ |
1232 | dev->hard_header_len = 0; | |
1233 | dev->addr_len = 0; | |
1234 | dev->mtu = 1500; | |
1235 | ||
1236 | /* Zero header length */ | |
6aa20a22 | 1237 | dev->type = ARPHRD_NONE; |
1da177e4 | 1238 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1239 | break; |
1240 | ||
40630b82 | 1241 | case IFF_TAP: |
7a0a9608 | 1242 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1243 | /* Ethernet TAP Device */ |
1da177e4 | 1244 | ether_setup(dev); |
550fd08c | 1245 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1246 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1247 | |
f2cedb63 | 1248 | eth_hw_addr_random(dev); |
36226a8d | 1249 | |
1da177e4 LT |
1250 | break; |
1251 | } | |
91572088 JW |
1252 | |
1253 | dev->min_mtu = MIN_MTU; | |
1254 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1255 | } |
1256 | ||
1475a1b5 JW |
1257 | static bool tun_sock_writeable(struct tun_struct *tun, struct tun_file *tfile) |
1258 | { | |
1259 | struct sock *sk = tfile->socket.sk; | |
1260 | ||
1261 | return (tun->dev->flags & IFF_UP) && sock_writeable(sk); | |
1262 | } | |
1263 | ||
1da177e4 LT |
1264 | /* Character device part */ |
1265 | ||
1266 | /* Poll */ | |
c8d68e6b | 1267 | static unsigned int tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1268 | { |
b2430de3 | 1269 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1270 | struct tun_struct *tun = tun_get(tfile); |
3c8a9c63 | 1271 | struct sock *sk; |
33dccbb0 | 1272 | unsigned int mask = 0; |
1da177e4 LT |
1273 | |
1274 | if (!tun) | |
eac9e902 | 1275 | return POLLERR; |
1da177e4 | 1276 | |
54f968d6 | 1277 | sk = tfile->socket.sk; |
3c8a9c63 | 1278 | |
6b8a66ee | 1279 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 1280 | |
9e641bdc | 1281 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1282 | |
1576d986 | 1283 | if (!skb_array_empty(&tfile->tx_array)) |
1da177e4 LT |
1284 | mask |= POLLIN | POLLRDNORM; |
1285 | ||
1475a1b5 JW |
1286 | /* Make sure SOCKWQ_ASYNC_NOSPACE is set if not writable to |
1287 | * guarantee EPOLLOUT to be raised by either here or | |
1288 | * tun_sock_write_space(). Then process could get notification | |
1289 | * after it writes to a down device and meets -EIO. | |
1290 | */ | |
1291 | if (tun_sock_writeable(tun, tfile) || | |
1292 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
1293 | tun_sock_writeable(tun, tfile))) | |
33dccbb0 HX |
1294 | mask |= POLLOUT | POLLWRNORM; |
1295 | ||
c70f1829 EB |
1296 | if (tun->dev->reg_state != NETREG_REGISTERED) |
1297 | mask = POLLERR; | |
1298 | ||
631ab46b | 1299 | tun_put(tun); |
1da177e4 LT |
1300 | return mask; |
1301 | } | |
1302 | ||
90e33d45 PP |
1303 | static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile, |
1304 | size_t len, | |
1305 | const struct iov_iter *it) | |
1306 | { | |
1307 | struct sk_buff *skb; | |
1308 | size_t linear; | |
1309 | int err; | |
1310 | int i; | |
1311 | ||
1312 | if (it->nr_segs > MAX_SKB_FRAGS + 1) | |
1313 | return ERR_PTR(-ENOMEM); | |
1314 | ||
1315 | local_bh_disable(); | |
1316 | skb = napi_get_frags(&tfile->napi); | |
1317 | local_bh_enable(); | |
1318 | if (!skb) | |
1319 | return ERR_PTR(-ENOMEM); | |
1320 | ||
1321 | linear = iov_iter_single_seg_count(it); | |
1322 | err = __skb_grow(skb, linear); | |
1323 | if (err) | |
1324 | goto free; | |
1325 | ||
1326 | skb->len = len; | |
1327 | skb->data_len = len - linear; | |
1328 | skb->truesize += skb->data_len; | |
1329 | ||
1330 | for (i = 1; i < it->nr_segs; i++) { | |
bf0c8520 | 1331 | struct page_frag *pfrag = ¤t->task_frag; |
90e33d45 | 1332 | size_t fragsz = it->iov[i].iov_len; |
90e33d45 PP |
1333 | |
1334 | if (fragsz == 0 || fragsz > PAGE_SIZE) { | |
1335 | err = -EINVAL; | |
1336 | goto free; | |
1337 | } | |
1338 | ||
bf0c8520 | 1339 | if (!skb_page_frag_refill(fragsz, pfrag, GFP_KERNEL)) { |
90e33d45 PP |
1340 | err = -ENOMEM; |
1341 | goto free; | |
1342 | } | |
1343 | ||
bf0c8520 ED |
1344 | skb_fill_page_desc(skb, i - 1, pfrag->page, |
1345 | pfrag->offset, fragsz); | |
1346 | page_ref_inc(pfrag->page); | |
1347 | pfrag->offset += fragsz; | |
90e33d45 PP |
1348 | } |
1349 | ||
1350 | return skb; | |
1351 | free: | |
1352 | /* frees skb and all frags allocated with napi_alloc_frag() */ | |
1353 | napi_free_frags(&tfile->napi); | |
1354 | return ERR_PTR(err); | |
1355 | } | |
1356 | ||
f42157cb RR |
1357 | /* prepad is the amount to reserve at front. len is length after that. |
1358 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1359 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1360 | size_t prepad, size_t len, |
1361 | size_t linear, int noblock) | |
f42157cb | 1362 | { |
54f968d6 | 1363 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1364 | struct sk_buff *skb; |
33dccbb0 | 1365 | int err; |
f42157cb RR |
1366 | |
1367 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1368 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1369 | linear = len; |
f42157cb | 1370 | |
33dccbb0 | 1371 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1372 | &err, 0); |
f42157cb | 1373 | if (!skb) |
33dccbb0 | 1374 | return ERR_PTR(err); |
f42157cb RR |
1375 | |
1376 | skb_reserve(skb, prepad); | |
1377 | skb_put(skb, linear); | |
33dccbb0 HX |
1378 | skb->data_len = len - linear; |
1379 | skb->len += len - linear; | |
f42157cb RR |
1380 | |
1381 | return skb; | |
1382 | } | |
1383 | ||
5503fcec JW |
1384 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1385 | struct sk_buff *skb, int more) | |
1386 | { | |
1387 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1388 | struct sk_buff_head process_queue; | |
1389 | u32 rx_batched = tun->rx_batched; | |
1390 | bool rcv = false; | |
1391 | ||
1392 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1393 | local_bh_disable(); | |
22e20b77 | 1394 | skb_record_rx_queue(skb, tfile->queue_index); |
5503fcec JW |
1395 | netif_receive_skb(skb); |
1396 | local_bh_enable(); | |
1397 | return; | |
1398 | } | |
1399 | ||
1400 | spin_lock(&queue->lock); | |
1401 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1402 | __skb_queue_head_init(&process_queue); | |
1403 | skb_queue_splice_tail_init(queue, &process_queue); | |
1404 | rcv = true; | |
1405 | } else { | |
1406 | __skb_queue_tail(queue, skb); | |
1407 | } | |
1408 | spin_unlock(&queue->lock); | |
1409 | ||
1410 | if (rcv) { | |
1411 | struct sk_buff *nskb; | |
1412 | ||
1413 | local_bh_disable(); | |
22e20b77 MC |
1414 | while ((nskb = __skb_dequeue(&process_queue))) { |
1415 | skb_record_rx_queue(nskb, tfile->queue_index); | |
5503fcec | 1416 | netif_receive_skb(nskb); |
22e20b77 MC |
1417 | } |
1418 | skb_record_rx_queue(skb, tfile->queue_index); | |
5503fcec JW |
1419 | netif_receive_skb(skb); |
1420 | local_bh_enable(); | |
1421 | } | |
1422 | } | |
1423 | ||
66ccbc9c JW |
1424 | static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile, |
1425 | int len, int noblock, bool zerocopy) | |
1426 | { | |
1427 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) | |
1428 | return false; | |
1429 | ||
1430 | if (tfile->socket.sk->sk_sndbuf != INT_MAX) | |
1431 | return false; | |
1432 | ||
1433 | if (!noblock) | |
1434 | return false; | |
1435 | ||
1436 | if (zerocopy) | |
1437 | return false; | |
1438 | ||
1439 | if (SKB_DATA_ALIGN(len + TUN_RX_PAD) + | |
1440 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE) | |
1441 | return false; | |
1442 | ||
1443 | return true; | |
1444 | } | |
1445 | ||
761876c8 JW |
1446 | static struct sk_buff *tun_build_skb(struct tun_struct *tun, |
1447 | struct tun_file *tfile, | |
66ccbc9c | 1448 | struct iov_iter *from, |
761876c8 | 1449 | struct virtio_net_hdr *hdr, |
1cfe6e93 | 1450 | int len, int *skb_xdp) |
66ccbc9c | 1451 | { |
0bbd7dad | 1452 | struct page_frag *alloc_frag = ¤t->task_frag; |
66ccbc9c | 1453 | struct sk_buff *skb; |
761876c8 | 1454 | struct bpf_prog *xdp_prog; |
7df13219 | 1455 | int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); |
761876c8 | 1456 | unsigned int delta = 0; |
66ccbc9c JW |
1457 | char *buf; |
1458 | size_t copied; | |
761876c8 | 1459 | bool xdp_xmit = false; |
7df13219 JW |
1460 | int err, pad = TUN_RX_PAD; |
1461 | ||
1462 | rcu_read_lock(); | |
1463 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1464 | if (xdp_prog) | |
1465 | pad += TUN_HEADROOM; | |
1466 | buflen += SKB_DATA_ALIGN(len + pad); | |
1467 | rcu_read_unlock(); | |
66ccbc9c | 1468 | |
63b9ab65 | 1469 | alloc_frag->offset = ALIGN((u64)alloc_frag->offset, SMP_CACHE_BYTES); |
66ccbc9c JW |
1470 | if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL))) |
1471 | return ERR_PTR(-ENOMEM); | |
1472 | ||
1473 | buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; | |
1474 | copied = copy_page_from_iter(alloc_frag->page, | |
7df13219 | 1475 | alloc_frag->offset + pad, |
66ccbc9c JW |
1476 | len, from); |
1477 | if (copied != len) | |
1478 | return ERR_PTR(-EFAULT); | |
1479 | ||
7df13219 JW |
1480 | /* There's a small window that XDP may be set after the check |
1481 | * of xdp_prog above, this should be rare and for simplicity | |
1482 | * we do XDP on skb in case the headroom is not enough. | |
1483 | */ | |
1484 | if (hdr->gso_type || !xdp_prog) | |
1cfe6e93 | 1485 | *skb_xdp = 1; |
761876c8 | 1486 | else |
1cfe6e93 | 1487 | *skb_xdp = 0; |
761876c8 | 1488 | |
eac5c3cc | 1489 | local_bh_disable(); |
761876c8 JW |
1490 | rcu_read_lock(); |
1491 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1cfe6e93 | 1492 | if (xdp_prog && !*skb_xdp) { |
761876c8 JW |
1493 | struct xdp_buff xdp; |
1494 | void *orig_data; | |
1495 | u32 act; | |
1496 | ||
1497 | xdp.data_hard_start = buf; | |
7df13219 | 1498 | xdp.data = buf + pad; |
de8f3a83 | 1499 | xdp_set_data_meta_invalid(&xdp); |
761876c8 JW |
1500 | xdp.data_end = xdp.data + len; |
1501 | orig_data = xdp.data; | |
1502 | act = bpf_prog_run_xdp(xdp_prog, &xdp); | |
1503 | ||
1504 | switch (act) { | |
1505 | case XDP_REDIRECT: | |
1506 | get_page(alloc_frag->page); | |
1507 | alloc_frag->offset += buflen; | |
1508 | err = xdp_do_redirect(tun->dev, &xdp, xdp_prog); | |
bf2b7190 | 1509 | xdp_do_flush_map(); |
761876c8 JW |
1510 | if (err) |
1511 | goto err_redirect; | |
654d5738 | 1512 | rcu_read_unlock(); |
eac5c3cc | 1513 | local_bh_enable(); |
761876c8 JW |
1514 | return NULL; |
1515 | case XDP_TX: | |
1516 | xdp_xmit = true; | |
1517 | /* fall through */ | |
1518 | case XDP_PASS: | |
1519 | delta = orig_data - xdp.data; | |
1520 | break; | |
1521 | default: | |
1522 | bpf_warn_invalid_xdp_action(act); | |
1523 | /* fall through */ | |
1524 | case XDP_ABORTED: | |
1525 | trace_xdp_exception(tun->dev, xdp_prog, act); | |
1526 | /* fall through */ | |
1527 | case XDP_DROP: | |
1528 | goto err_xdp; | |
1529 | } | |
1530 | } | |
1531 | ||
66ccbc9c | 1532 | skb = build_skb(buf, buflen); |
761876c8 JW |
1533 | if (!skb) { |
1534 | rcu_read_unlock(); | |
eac5c3cc | 1535 | local_bh_enable(); |
66ccbc9c | 1536 | return ERR_PTR(-ENOMEM); |
761876c8 | 1537 | } |
66ccbc9c | 1538 | |
7df13219 | 1539 | skb_reserve(skb, pad - delta); |
761876c8 | 1540 | skb_put(skb, len + delta); |
d84f7de0 | 1541 | skb_set_owner_w(skb, tfile->socket.sk); |
66ccbc9c JW |
1542 | get_page(alloc_frag->page); |
1543 | alloc_frag->offset += buflen; | |
1544 | ||
761876c8 JW |
1545 | if (xdp_xmit) { |
1546 | skb->dev = tun->dev; | |
1547 | generic_xdp_tx(skb, xdp_prog); | |
654d5738 | 1548 | rcu_read_unlock(); |
eac5c3cc | 1549 | local_bh_enable(); |
761876c8 JW |
1550 | return NULL; |
1551 | } | |
1552 | ||
1553 | rcu_read_unlock(); | |
eac5c3cc | 1554 | local_bh_enable(); |
761876c8 | 1555 | |
66ccbc9c | 1556 | return skb; |
761876c8 JW |
1557 | |
1558 | err_redirect: | |
1559 | put_page(alloc_frag->page); | |
1560 | err_xdp: | |
1561 | rcu_read_unlock(); | |
eac5c3cc | 1562 | local_bh_enable(); |
761876c8 JW |
1563 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
1564 | return NULL; | |
66ccbc9c JW |
1565 | } |
1566 | ||
1da177e4 | 1567 | /* Get packet from user space buffer */ |
54f968d6 | 1568 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1569 | void *msg_control, struct iov_iter *from, |
5503fcec | 1570 | int noblock, bool more) |
1da177e4 | 1571 | { |
09640e63 | 1572 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1573 | struct sk_buff *skb; |
f5ff53b4 | 1574 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1575 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1576 | struct virtio_net_hdr gso = { 0 }; |
608b9977 | 1577 | struct tun_pcpu_stats *stats; |
96f8d9ec | 1578 | int good_linear; |
0690899b MT |
1579 | int copylen; |
1580 | bool zerocopy = false; | |
1581 | int err; | |
49974420 | 1582 | u32 rxhash; |
1cfe6e93 | 1583 | int skb_xdp = 1; |
157f3ad9 | 1584 | bool frags = tun_napi_frags_enabled(tfile); |
1da177e4 | 1585 | |
40630b82 | 1586 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1587 | if (len < sizeof(pi)) |
1da177e4 | 1588 | return -EINVAL; |
15718ea0 | 1589 | len -= sizeof(pi); |
1da177e4 | 1590 | |
cbbd26b8 | 1591 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1592 | return -EFAULT; |
1593 | } | |
1594 | ||
40630b82 | 1595 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1596 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1597 | ||
1598 | if (len < vnet_hdr_sz) | |
f43798c2 | 1599 | return -EINVAL; |
e1edab87 | 1600 | len -= vnet_hdr_sz; |
f43798c2 | 1601 | |
cbbd26b8 | 1602 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1603 | return -EFAULT; |
1604 | ||
4909122f | 1605 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1606 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1607 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1608 | |
56f0dcc5 | 1609 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1610 | return -EINVAL; |
e1edab87 | 1611 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1612 | } |
1613 | ||
40630b82 | 1614 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1615 | align += NET_IP_ALIGN; |
0eca93bc | 1616 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1617 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1618 | return -EINVAL; |
1619 | } | |
6aa20a22 | 1620 | |
96f8d9ec JW |
1621 | good_linear = SKB_MAX_HEAD(align); |
1622 | ||
88529176 | 1623 | if (msg_control) { |
f5ff53b4 AV |
1624 | struct iov_iter i = *from; |
1625 | ||
88529176 JW |
1626 | /* There are 256 bytes to be copied in skb, so there is |
1627 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1628 | * The rest of the buffer is mapped from userspace. |
1629 | */ | |
56f0dcc5 | 1630 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1631 | if (copylen > good_linear) |
1632 | copylen = good_linear; | |
3dd5c330 | 1633 | linear = copylen; |
f5ff53b4 AV |
1634 | iov_iter_advance(&i, copylen); |
1635 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1636 | zerocopy = true; |
1637 | } | |
1638 | ||
90e33d45 | 1639 | if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) { |
1cfe6e93 JW |
1640 | /* For the packet that is not easy to be processed |
1641 | * (e.g gso or jumbo packet), we will do it at after | |
1642 | * skb was created with generic XDP routine. | |
1643 | */ | |
1644 | skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp); | |
66ccbc9c | 1645 | if (IS_ERR(skb)) { |
608b9977 | 1646 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
66ccbc9c JW |
1647 | return PTR_ERR(skb); |
1648 | } | |
761876c8 JW |
1649 | if (!skb) |
1650 | return total_len; | |
66ccbc9c JW |
1651 | } else { |
1652 | if (!zerocopy) { | |
1653 | copylen = len; | |
1654 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) | |
1655 | linear = good_linear; | |
1656 | else | |
1657 | linear = tun16_to_cpu(tun, gso.hdr_len); | |
1658 | } | |
1da177e4 | 1659 | |
90e33d45 PP |
1660 | if (frags) { |
1661 | mutex_lock(&tfile->napi_mutex); | |
1662 | skb = tun_napi_alloc_frags(tfile, copylen, from); | |
1663 | /* tun_napi_alloc_frags() enforces a layout for the skb. | |
1664 | * If zerocopy is enabled, then this layout will be | |
1665 | * overwritten by zerocopy_sg_from_iter(). | |
1666 | */ | |
1667 | zerocopy = false; | |
1668 | } else { | |
1669 | skb = tun_alloc_skb(tfile, align, copylen, linear, | |
1670 | noblock); | |
1671 | } | |
1672 | ||
66ccbc9c JW |
1673 | if (IS_ERR(skb)) { |
1674 | if (PTR_ERR(skb) != -EAGAIN) | |
1675 | this_cpu_inc(tun->pcpu_stats->rx_dropped); | |
90e33d45 PP |
1676 | if (frags) |
1677 | mutex_unlock(&tfile->napi_mutex); | |
66ccbc9c JW |
1678 | return PTR_ERR(skb); |
1679 | } | |
0690899b | 1680 | |
66ccbc9c JW |
1681 | if (zerocopy) |
1682 | err = zerocopy_sg_from_iter(skb, from); | |
1683 | else | |
1684 | err = skb_copy_datagram_from_iter(skb, 0, from, len); | |
1685 | ||
1686 | if (err) { | |
09e99eef ED |
1687 | err = -EFAULT; |
1688 | drop: | |
66ccbc9c JW |
1689 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
1690 | kfree_skb(skb); | |
90e33d45 PP |
1691 | if (frags) { |
1692 | tfile->napi.skb = NULL; | |
1693 | mutex_unlock(&tfile->napi_mutex); | |
1694 | } | |
1695 | ||
09e99eef | 1696 | return err; |
66ccbc9c | 1697 | } |
8f22757e | 1698 | } |
1da177e4 | 1699 | |
3e9e40e7 | 1700 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
df10db98 PA |
1701 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
1702 | kfree_skb(skb); | |
90e33d45 PP |
1703 | if (frags) { |
1704 | tfile->napi.skb = NULL; | |
1705 | mutex_unlock(&tfile->napi_mutex); | |
1706 | } | |
1707 | ||
df10db98 PA |
1708 | return -EINVAL; |
1709 | } | |
1710 | ||
1da177e4 | 1711 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1712 | case IFF_TUN: |
1713 | if (tun->flags & IFF_NO_PI) { | |
2580c4c1 AP |
1714 | u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0; |
1715 | ||
1716 | switch (ip_version) { | |
1717 | case 4: | |
f09f7ee2 AWC |
1718 | pi.proto = htons(ETH_P_IP); |
1719 | break; | |
2580c4c1 | 1720 | case 6: |
f09f7ee2 AWC |
1721 | pi.proto = htons(ETH_P_IPV6); |
1722 | break; | |
1723 | default: | |
608b9977 | 1724 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
f09f7ee2 AWC |
1725 | kfree_skb(skb); |
1726 | return -EINVAL; | |
1727 | } | |
1728 | } | |
1729 | ||
459a98ed | 1730 | skb_reset_mac_header(skb); |
1da177e4 | 1731 | skb->protocol = pi.proto; |
4c13eb66 | 1732 | skb->dev = tun->dev; |
1da177e4 | 1733 | break; |
40630b82 | 1734 | case IFF_TAP: |
90e33d45 PP |
1735 | if (!frags) |
1736 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1da177e4 | 1737 | break; |
6403eab1 | 1738 | } |
1da177e4 | 1739 | |
0690899b MT |
1740 | /* copy skb_ubuf_info for callback when skb has no error */ |
1741 | if (zerocopy) { | |
1742 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1743 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1744 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
af1cc7a2 JW |
1745 | } else if (msg_control) { |
1746 | struct ubuf_info *uarg = msg_control; | |
1747 | uarg->callback(uarg, false); | |
0690899b MT |
1748 | } |
1749 | ||
72f65107 | 1750 | skb_reset_network_header(skb); |
40893fd0 | 1751 | skb_probe_transport_header(skb, 0); |
38502af7 | 1752 | |
1cfe6e93 | 1753 | if (skb_xdp) { |
761876c8 JW |
1754 | struct bpf_prog *xdp_prog; |
1755 | int ret; | |
1756 | ||
eac5c3cc | 1757 | local_bh_disable(); |
761876c8 JW |
1758 | rcu_read_lock(); |
1759 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1760 | if (xdp_prog) { | |
1761 | ret = do_xdp_generic(xdp_prog, skb); | |
1762 | if (ret != XDP_PASS) { | |
1763 | rcu_read_unlock(); | |
eac5c3cc | 1764 | local_bh_enable(); |
761876c8 JW |
1765 | return total_len; |
1766 | } | |
1767 | } | |
1768 | rcu_read_unlock(); | |
eac5c3cc | 1769 | local_bh_enable(); |
761876c8 JW |
1770 | } |
1771 | ||
feec084a | 1772 | rxhash = __skb_get_hash_symmetric(skb); |
94317099 | 1773 | |
09e99eef ED |
1774 | rcu_read_lock(); |
1775 | if (unlikely(!(tun->dev->flags & IFF_UP))) { | |
1776 | err = -EIO; | |
77268819 | 1777 | rcu_read_unlock(); |
09e99eef ED |
1778 | goto drop; |
1779 | } | |
1780 | ||
90e33d45 PP |
1781 | if (frags) { |
1782 | /* Exercise flow dissector code path. */ | |
1783 | u32 headlen = eth_get_headlen(skb->data, skb_headlen(skb)); | |
1784 | ||
010f245b | 1785 | if (unlikely(headlen > skb_headlen(skb))) { |
90e33d45 PP |
1786 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
1787 | napi_free_frags(&tfile->napi); | |
09e99eef | 1788 | rcu_read_unlock(); |
90e33d45 PP |
1789 | mutex_unlock(&tfile->napi_mutex); |
1790 | WARN_ON(1); | |
1791 | return -ENOMEM; | |
1792 | } | |
1793 | ||
1794 | local_bh_disable(); | |
1795 | napi_gro_frags(&tfile->napi); | |
1796 | local_bh_enable(); | |
1797 | mutex_unlock(&tfile->napi_mutex); | |
aec72f33 | 1798 | } else if (tfile->napi_enabled) { |
94317099 PP |
1799 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; |
1800 | int queue_len; | |
1801 | ||
1802 | spin_lock_bh(&queue->lock); | |
1803 | __skb_queue_tail(queue, skb); | |
1804 | queue_len = skb_queue_len(queue); | |
1805 | spin_unlock(&queue->lock); | |
1806 | ||
1807 | if (!more || queue_len > NAPI_POLL_WEIGHT) | |
1808 | napi_schedule(&tfile->napi); | |
1809 | ||
1810 | local_bh_enable(); | |
1811 | } else if (!IS_ENABLED(CONFIG_4KSTACKS)) { | |
1812 | tun_rx_batched(tun, tfile, skb, more); | |
1813 | } else { | |
1814 | netif_rx_ni(skb); | |
1815 | } | |
09e99eef | 1816 | rcu_read_unlock(); |
6aa20a22 | 1817 | |
608b9977 PA |
1818 | stats = get_cpu_ptr(tun->pcpu_stats); |
1819 | u64_stats_update_begin(&stats->syncp); | |
1820 | stats->rx_packets++; | |
1821 | stats->rx_bytes += len; | |
1822 | u64_stats_update_end(&stats->syncp); | |
1823 | put_cpu_ptr(stats); | |
1da177e4 | 1824 | |
9e85722d | 1825 | tun_flow_update(tun, rxhash, tfile); |
0690899b | 1826 | return total_len; |
6aa20a22 | 1827 | } |
1da177e4 | 1828 | |
f5ff53b4 | 1829 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1830 | { |
33dccbb0 | 1831 | struct file *file = iocb->ki_filp; |
54f968d6 | 1832 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1833 | struct tun_struct *tun = tun_get(tfile); |
631ab46b | 1834 | ssize_t result; |
1da177e4 LT |
1835 | |
1836 | if (!tun) | |
1837 | return -EBADFD; | |
1838 | ||
5503fcec JW |
1839 | result = tun_get_user(tun, tfile, NULL, from, |
1840 | file->f_flags & O_NONBLOCK, false); | |
631ab46b EB |
1841 | |
1842 | tun_put(tun); | |
1843 | return result; | |
1da177e4 LT |
1844 | } |
1845 | ||
1da177e4 | 1846 | /* Put packet to the user space buffer */ |
6f7c156c | 1847 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 1848 | struct tun_file *tfile, |
6f7c156c | 1849 | struct sk_buff *skb, |
e0b46d0e | 1850 | struct iov_iter *iter) |
1da177e4 LT |
1851 | { |
1852 | struct tun_pi pi = { 0, skb->protocol }; | |
608b9977 | 1853 | struct tun_pcpu_stats *stats; |
e0b46d0e | 1854 | ssize_t total; |
8c847d25 | 1855 | int vlan_offset = 0; |
a8f9bfdf | 1856 | int vlan_hlen = 0; |
2eb783c4 | 1857 | int vnet_hdr_sz = 0; |
a8f9bfdf | 1858 | |
df8a39de | 1859 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 1860 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 1861 | |
40630b82 | 1862 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 1863 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 1864 | |
e0b46d0e HX |
1865 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
1866 | ||
40630b82 | 1867 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 1868 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
1869 | return -EINVAL; |
1870 | ||
e0b46d0e HX |
1871 | total += sizeof(pi); |
1872 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
1873 | /* Packet will be striped */ |
1874 | pi.flags |= TUN_PKT_STRIP; | |
1875 | } | |
6aa20a22 | 1876 | |
e0b46d0e | 1877 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 1878 | return -EFAULT; |
6aa20a22 | 1879 | } |
1da177e4 | 1880 | |
2eb783c4 | 1881 | if (vnet_hdr_sz) { |
9403cd7c | 1882 | struct virtio_net_hdr gso; |
34166093 | 1883 | |
e0b46d0e | 1884 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
1885 | return -EINVAL; |
1886 | ||
3e9e40e7 | 1887 | if (virtio_net_hdr_from_skb(skb, &gso, |
34d736aa WB |
1888 | tun_is_little_endian(tun), true, |
1889 | vlan_hlen)) { | |
f43798c2 | 1890 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
1891 | pr_err("unexpected GSO type: " |
1892 | "0x%x, gso_size %d, hdr_len %d\n", | |
1893 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
1894 | tun16_to_cpu(tun, gso.hdr_len)); | |
1895 | print_hex_dump(KERN_ERR, "tun: ", | |
1896 | DUMP_PREFIX_NONE, | |
1897 | 16, 1, skb->head, | |
1898 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
1899 | WARN_ON_ONCE(1); | |
1900 | return -EINVAL; | |
1901 | } | |
f43798c2 | 1902 | |
e0b46d0e | 1903 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 1904 | return -EFAULT; |
8c847d25 JW |
1905 | |
1906 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
1907 | } |
1908 | ||
a8f9bfdf | 1909 | if (vlan_hlen) { |
e0b46d0e | 1910 | int ret; |
6680ec68 JW |
1911 | struct { |
1912 | __be16 h_vlan_proto; | |
1913 | __be16 h_vlan_TCI; | |
1914 | } veth; | |
1915 | ||
1916 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 1917 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
1918 | |
1919 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 1920 | |
e0b46d0e HX |
1921 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
1922 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
1923 | goto done; |
1924 | ||
e0b46d0e HX |
1925 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
1926 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
1927 | goto done; |
1928 | } | |
1da177e4 | 1929 | |
e0b46d0e | 1930 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 1931 | |
6680ec68 | 1932 | done: |
608b9977 PA |
1933 | /* caller is in process context, */ |
1934 | stats = get_cpu_ptr(tun->pcpu_stats); | |
1935 | u64_stats_update_begin(&stats->syncp); | |
1936 | stats->tx_packets++; | |
1937 | stats->tx_bytes += skb->len + vlan_hlen; | |
1938 | u64_stats_update_end(&stats->syncp); | |
1939 | put_cpu_ptr(tun->pcpu_stats); | |
1da177e4 LT |
1940 | |
1941 | return total; | |
1942 | } | |
1943 | ||
1576d986 JW |
1944 | static struct sk_buff *tun_ring_recv(struct tun_file *tfile, int noblock, |
1945 | int *err) | |
1946 | { | |
1947 | DECLARE_WAITQUEUE(wait, current); | |
1948 | struct sk_buff *skb = NULL; | |
f48cc6b2 | 1949 | int error = 0; |
1576d986 JW |
1950 | |
1951 | skb = skb_array_consume(&tfile->tx_array); | |
1952 | if (skb) | |
1953 | goto out; | |
1954 | if (noblock) { | |
f48cc6b2 | 1955 | error = -EAGAIN; |
1576d986 JW |
1956 | goto out; |
1957 | } | |
1958 | ||
1959 | add_wait_queue(&tfile->wq.wait, &wait); | |
1576d986 JW |
1960 | |
1961 | while (1) { | |
1f47139b | 1962 | set_current_state(TASK_INTERRUPTIBLE); |
1576d986 JW |
1963 | skb = skb_array_consume(&tfile->tx_array); |
1964 | if (skb) | |
1965 | break; | |
1966 | if (signal_pending(current)) { | |
f48cc6b2 | 1967 | error = -ERESTARTSYS; |
1576d986 JW |
1968 | break; |
1969 | } | |
1970 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 1971 | error = -EFAULT; |
1576d986 JW |
1972 | break; |
1973 | } | |
1974 | ||
1975 | schedule(); | |
1976 | } | |
1977 | ||
d68a2046 | 1978 | __set_current_state(TASK_RUNNING); |
1576d986 JW |
1979 | remove_wait_queue(&tfile->wq.wait, &wait); |
1980 | ||
1981 | out: | |
f48cc6b2 | 1982 | *err = error; |
1576d986 JW |
1983 | return skb; |
1984 | } | |
1985 | ||
54f968d6 | 1986 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 | 1987 | struct iov_iter *to, |
ac77cfd4 | 1988 | int noblock, struct sk_buff *skb) |
1da177e4 | 1989 | { |
9b067034 | 1990 | ssize_t ret; |
1576d986 | 1991 | int err; |
1da177e4 | 1992 | |
3872baf6 | 1993 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 1994 | |
c33ee15b WX |
1995 | if (!iov_iter_count(to)) { |
1996 | if (skb) | |
1997 | kfree_skb(skb); | |
9b067034 | 1998 | return 0; |
c33ee15b | 1999 | } |
1da177e4 | 2000 | |
ac77cfd4 JW |
2001 | if (!skb) { |
2002 | /* Read frames from ring */ | |
2003 | skb = tun_ring_recv(tfile, noblock, &err); | |
2004 | if (!skb) | |
2005 | return err; | |
2006 | } | |
e0b46d0e | 2007 | |
9b067034 | 2008 | ret = tun_put_user(tun, tfile, skb, to); |
f51a5e82 | 2009 | if (unlikely(ret < 0)) |
f271b2cc | 2010 | kfree_skb(skb); |
f51a5e82 JW |
2011 | else |
2012 | consume_skb(skb); | |
1da177e4 | 2013 | |
05c2828c MT |
2014 | return ret; |
2015 | } | |
2016 | ||
9b067034 | 2017 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
2018 | { |
2019 | struct file *file = iocb->ki_filp; | |
2020 | struct tun_file *tfile = file->private_data; | |
9484dc74 | 2021 | struct tun_struct *tun = tun_get(tfile); |
9b067034 | 2022 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
2023 | |
2024 | if (!tun) | |
2025 | return -EBADFD; | |
ac77cfd4 | 2026 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK, NULL); |
42404c09 | 2027 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
2028 | if (ret > 0) |
2029 | iocb->ki_pos = ret; | |
631ab46b | 2030 | tun_put(tun); |
1da177e4 LT |
2031 | return ret; |
2032 | } | |
2033 | ||
96442e42 JW |
2034 | static void tun_free_netdev(struct net_device *dev) |
2035 | { | |
2036 | struct tun_struct *tun = netdev_priv(dev); | |
2037 | ||
4008e97f | 2038 | BUG_ON(!(list_empty(&tun->disabled))); |
608b9977 | 2039 | free_percpu(tun->pcpu_stats); |
96442e42 | 2040 | tun_flow_uninit(tun); |
5dbbaf2d | 2041 | security_tun_dev_free_security(tun->security); |
96442e42 JW |
2042 | } |
2043 | ||
1da177e4 LT |
2044 | static void tun_setup(struct net_device *dev) |
2045 | { | |
2046 | struct tun_struct *tun = netdev_priv(dev); | |
2047 | ||
0625c883 EB |
2048 | tun->owner = INVALID_UID; |
2049 | tun->group = INVALID_GID; | |
1da177e4 | 2050 | |
1da177e4 | 2051 | dev->ethtool_ops = &tun_ethtool_ops; |
cf124db5 DM |
2052 | dev->needs_free_netdev = true; |
2053 | dev->priv_destructor = tun_free_netdev; | |
016adb72 JW |
2054 | /* We prefer our own queue length */ |
2055 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
2056 | } |
2057 | ||
f019a7a5 EB |
2058 | /* Trivial set of netlink ops to allow deleting tun or tap |
2059 | * device with netlink. | |
2060 | */ | |
a8b8a889 MS |
2061 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[], |
2062 | struct netlink_ext_ack *extack) | |
f019a7a5 | 2063 | { |
6bcf49b9 ND |
2064 | NL_SET_ERR_MSG(extack, |
2065 | "tun/tap creation via rtnetlink is not supported."); | |
2066 | return -EOPNOTSUPP; | |
f019a7a5 EB |
2067 | } |
2068 | ||
2069 | static struct rtnl_link_ops tun_link_ops __read_mostly = { | |
2070 | .kind = DRV_NAME, | |
2071 | .priv_size = sizeof(struct tun_struct), | |
2072 | .setup = tun_setup, | |
2073 | .validate = tun_validate, | |
2074 | }; | |
2075 | ||
33dccbb0 HX |
2076 | static void tun_sock_write_space(struct sock *sk) |
2077 | { | |
54f968d6 | 2078 | struct tun_file *tfile; |
43815482 | 2079 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
2080 | |
2081 | if (!sock_writeable(sk)) | |
2082 | return; | |
2083 | ||
9cd3e072 | 2084 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
2085 | return; |
2086 | ||
43815482 ED |
2087 | wqueue = sk_sleep(sk); |
2088 | if (wqueue && waitqueue_active(wqueue)) | |
2089 | wake_up_interruptible_sync_poll(wqueue, POLLOUT | | |
05c2828c | 2090 | POLLWRNORM | POLLWRBAND); |
c722c625 | 2091 | |
54f968d6 JW |
2092 | tfile = container_of(sk, struct tun_file, sk); |
2093 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
2094 | } |
2095 | ||
1b784140 | 2096 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 2097 | { |
54f968d6 JW |
2098 | int ret; |
2099 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
9484dc74 | 2100 | struct tun_struct *tun = tun_get(tfile); |
54f968d6 JW |
2101 | |
2102 | if (!tun) | |
2103 | return -EBADFD; | |
f5ff53b4 | 2104 | |
c0371da6 | 2105 | ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter, |
5503fcec JW |
2106 | m->msg_flags & MSG_DONTWAIT, |
2107 | m->msg_flags & MSG_MORE); | |
54f968d6 JW |
2108 | tun_put(tun); |
2109 | return ret; | |
05c2828c MT |
2110 | } |
2111 | ||
1b784140 | 2112 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
2113 | int flags) |
2114 | { | |
54f968d6 | 2115 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2116 | struct tun_struct *tun = tun_get(tfile); |
c33ee15b | 2117 | struct sk_buff *skb = m->msg_control; |
05c2828c | 2118 | int ret; |
54f968d6 | 2119 | |
c33ee15b WX |
2120 | if (!tun) { |
2121 | ret = -EBADFD; | |
2122 | goto out_free_skb; | |
2123 | } | |
54f968d6 | 2124 | |
eda29772 | 2125 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 | 2126 | ret = -EINVAL; |
c33ee15b | 2127 | goto out_put_tun; |
3811ae76 | 2128 | } |
eda29772 RC |
2129 | if (flags & MSG_ERRQUEUE) { |
2130 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
2131 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
2132 | goto out; | |
2133 | } | |
c33ee15b | 2134 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, skb); |
87897931 | 2135 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
2136 | m->msg_flags |= MSG_TRUNC; |
2137 | ret = flags & MSG_TRUNC ? ret : total_len; | |
2138 | } | |
3811ae76 | 2139 | out: |
54f968d6 | 2140 | tun_put(tun); |
05c2828c | 2141 | return ret; |
c33ee15b WX |
2142 | |
2143 | out_put_tun: | |
2144 | tun_put(tun); | |
2145 | out_free_skb: | |
2146 | if (skb) | |
2147 | kfree_skb(skb); | |
2148 | return ret; | |
05c2828c MT |
2149 | } |
2150 | ||
1576d986 JW |
2151 | static int tun_peek_len(struct socket *sock) |
2152 | { | |
2153 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
2154 | struct tun_struct *tun; | |
2155 | int ret = 0; | |
2156 | ||
9484dc74 | 2157 | tun = tun_get(tfile); |
1576d986 JW |
2158 | if (!tun) |
2159 | return 0; | |
2160 | ||
2161 | ret = skb_array_peek_len(&tfile->tx_array); | |
2162 | tun_put(tun); | |
2163 | ||
2164 | return ret; | |
2165 | } | |
2166 | ||
05c2828c MT |
2167 | /* Ops structure to mimic raw sockets with tun */ |
2168 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 2169 | .peek_len = tun_peek_len, |
05c2828c MT |
2170 | .sendmsg = tun_sendmsg, |
2171 | .recvmsg = tun_recvmsg, | |
2172 | }; | |
2173 | ||
33dccbb0 HX |
2174 | static struct proto tun_proto = { |
2175 | .name = "tun", | |
2176 | .owner = THIS_MODULE, | |
54f968d6 | 2177 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 2178 | }; |
f019a7a5 | 2179 | |
980c9e8c DW |
2180 | static int tun_flags(struct tun_struct *tun) |
2181 | { | |
031f5e03 | 2182 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
2183 | } |
2184 | ||
2185 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
2186 | char *buf) | |
2187 | { | |
2188 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
2189 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
2190 | } | |
2191 | ||
2192 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
2193 | char *buf) | |
2194 | { | |
2195 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2196 | return uid_valid(tun->owner)? |
2197 | sprintf(buf, "%u\n", | |
2198 | from_kuid_munged(current_user_ns(), tun->owner)): | |
2199 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2200 | } |
2201 | ||
2202 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
2203 | char *buf) | |
2204 | { | |
2205 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2206 | return gid_valid(tun->group) ? |
2207 | sprintf(buf, "%u\n", | |
2208 | from_kgid_munged(current_user_ns(), tun->group)): | |
2209 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2210 | } |
2211 | ||
2212 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
2213 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
2214 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
2215 | ||
c4d33e24 TI |
2216 | static struct attribute *tun_dev_attrs[] = { |
2217 | &dev_attr_tun_flags.attr, | |
2218 | &dev_attr_owner.attr, | |
2219 | &dev_attr_group.attr, | |
2220 | NULL | |
2221 | }; | |
2222 | ||
2223 | static const struct attribute_group tun_attr_group = { | |
2224 | .attrs = tun_dev_attrs | |
2225 | }; | |
2226 | ||
d647a591 | 2227 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
2228 | { |
2229 | struct tun_struct *tun; | |
54f968d6 | 2230 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2231 | struct net_device *dev; |
2232 | int err; | |
2233 | ||
7c0c3b1a JW |
2234 | if (tfile->detached) |
2235 | return -EINVAL; | |
2236 | ||
90e33d45 PP |
2237 | if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) { |
2238 | if (!capable(CAP_NET_ADMIN)) | |
2239 | return -EPERM; | |
2240 | ||
2241 | if (!(ifr->ifr_flags & IFF_NAPI) || | |
2242 | (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP) | |
2243 | return -EINVAL; | |
2244 | } | |
2245 | ||
74a3e5a7 EB |
2246 | dev = __dev_get_by_name(net, ifr->ifr_name); |
2247 | if (dev) { | |
f85ba780 DW |
2248 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
2249 | return -EBUSY; | |
74a3e5a7 EB |
2250 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
2251 | tun = netdev_priv(dev); | |
2252 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
2253 | tun = netdev_priv(dev); | |
2254 | else | |
2255 | return -EINVAL; | |
2256 | ||
8e6d91ae | 2257 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 2258 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
2259 | return -EINVAL; |
2260 | ||
cde8b15f | 2261 | if (tun_not_capable(tun)) |
2b980dbd | 2262 | return -EPERM; |
5dbbaf2d | 2263 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
2264 | if (err < 0) |
2265 | return err; | |
2266 | ||
94317099 | 2267 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER, |
157f3ad9 | 2268 | ifr->ifr_flags & IFF_NAPI, |
fdf434d1 | 2269 | ifr->ifr_flags & IFF_NAPI_FRAGS, true); |
a7385ba2 EB |
2270 | if (err < 0) |
2271 | return err; | |
4008e97f | 2272 | |
40630b82 | 2273 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
2274 | (tun->numqueues + tun->numdisabled > 1)) { |
2275 | /* One or more queue has already been attached, no need | |
2276 | * to initialize the device again. | |
2277 | */ | |
2278 | return 0; | |
2279 | } | |
6aa20a22 | 2280 | } |
1da177e4 LT |
2281 | else { |
2282 | char *name; | |
2283 | unsigned long flags = 0; | |
edfb6a14 JW |
2284 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
2285 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 2286 | |
c260b772 | 2287 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 2288 | return -EPERM; |
2b980dbd PM |
2289 | err = security_tun_dev_create(); |
2290 | if (err < 0) | |
2291 | return err; | |
ca6bb5d7 | 2292 | |
1da177e4 LT |
2293 | /* Set dev type */ |
2294 | if (ifr->ifr_flags & IFF_TUN) { | |
2295 | /* TUN device */ | |
40630b82 | 2296 | flags |= IFF_TUN; |
1da177e4 LT |
2297 | name = "tun%d"; |
2298 | } else if (ifr->ifr_flags & IFF_TAP) { | |
2299 | /* TAP device */ | |
40630b82 | 2300 | flags |= IFF_TAP; |
1da177e4 | 2301 | name = "tap%d"; |
6aa20a22 | 2302 | } else |
36989b90 | 2303 | return -EINVAL; |
6aa20a22 | 2304 | |
1da177e4 LT |
2305 | if (*ifr->ifr_name) |
2306 | name = ifr->ifr_name; | |
2307 | ||
c8d68e6b | 2308 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
2309 | NET_NAME_UNKNOWN, tun_setup, queues, |
2310 | queues); | |
edfb6a14 | 2311 | |
1da177e4 LT |
2312 | if (!dev) |
2313 | return -ENOMEM; | |
0ad646c8 | 2314 | err = dev_get_valid_name(net, dev, name); |
5c25f65f | 2315 | if (err < 0) |
0ad646c8 | 2316 | goto err_free_dev; |
1da177e4 | 2317 | |
fc54c658 | 2318 | dev_net_set(dev, net); |
f019a7a5 | 2319 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 2320 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 2321 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 2322 | |
1da177e4 LT |
2323 | tun = netdev_priv(dev); |
2324 | tun->dev = dev; | |
2325 | tun->flags = flags; | |
f271b2cc | 2326 | tun->txflt.count = 0; |
d9d52b51 | 2327 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 2328 | |
eaea34b2 | 2329 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
2330 | tun->filter_attached = false; |
2331 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 2332 | tun->rx_batched = 0; |
33dccbb0 | 2333 | |
608b9977 PA |
2334 | tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); |
2335 | if (!tun->pcpu_stats) { | |
2336 | err = -ENOMEM; | |
2337 | goto err_free_dev; | |
2338 | } | |
2339 | ||
96442e42 JW |
2340 | spin_lock_init(&tun->lock); |
2341 | ||
5dbbaf2d PM |
2342 | err = security_tun_dev_alloc_security(&tun->security); |
2343 | if (err < 0) | |
608b9977 | 2344 | goto err_free_stat; |
2b980dbd | 2345 | |
1da177e4 | 2346 | tun_net_init(dev); |
944a1376 | 2347 | tun_flow_init(tun); |
96442e42 | 2348 | |
88255375 | 2349 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
2350 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
2351 | NETIF_F_HW_VLAN_STAG_TX; | |
2a2bbf17 | 2352 | dev->features = dev->hw_features | NETIF_F_LLTX; |
6671b224 FLVC |
2353 | dev->vlan_features = dev->features & |
2354 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
2355 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 2356 | |
4008e97f | 2357 | INIT_LIST_HEAD(&tun->disabled); |
157f3ad9 | 2358 | err = tun_attach(tun, file, false, ifr->ifr_flags & IFF_NAPI, |
fdf434d1 | 2359 | ifr->ifr_flags & IFF_NAPI_FRAGS, false); |
eb0fb363 | 2360 | if (err < 0) |
662ca437 | 2361 | goto err_free_flow; |
eb0fb363 | 2362 | |
1da177e4 LT |
2363 | err = register_netdevice(tun->dev); |
2364 | if (err < 0) | |
662ca437 | 2365 | goto err_detach; |
fdf434d1 YY |
2366 | /* free_netdev() won't check refcnt, to aovid race |
2367 | * with dev_put() we need publish tun after registration. | |
2368 | */ | |
2369 | rcu_assign_pointer(tfile->tun, tun); | |
1da177e4 LT |
2370 | } |
2371 | ||
af668b3c MT |
2372 | netif_carrier_on(tun->dev); |
2373 | ||
6b8a66ee | 2374 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 2375 | |
031f5e03 MT |
2376 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
2377 | (ifr->ifr_flags & TUN_FEATURES); | |
c8d68e6b | 2378 | |
e35259a9 MK |
2379 | /* Make sure persistent devices do not get stuck in |
2380 | * xoff state. | |
2381 | */ | |
2382 | if (netif_running(tun->dev)) | |
c8d68e6b | 2383 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 2384 | |
1da177e4 LT |
2385 | strcpy(ifr->ifr_name, tun->dev->name); |
2386 | return 0; | |
2387 | ||
662ca437 JW |
2388 | err_detach: |
2389 | tun_detach_all(dev); | |
ff244c6b ED |
2390 | /* register_netdevice() already called tun_free_netdev() */ |
2391 | goto err_free_dev; | |
2392 | ||
662ca437 JW |
2393 | err_free_flow: |
2394 | tun_flow_uninit(tun); | |
2395 | security_tun_dev_free_security(tun->security); | |
608b9977 PA |
2396 | err_free_stat: |
2397 | free_percpu(tun->pcpu_stats); | |
662ca437 | 2398 | err_free_dev: |
1da177e4 | 2399 | free_netdev(dev); |
1da177e4 LT |
2400 | return err; |
2401 | } | |
2402 | ||
9ce99cf6 | 2403 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 2404 | struct ifreq *ifr) |
e3b99556 | 2405 | { |
6b8a66ee | 2406 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
2407 | |
2408 | strcpy(ifr->ifr_name, tun->dev->name); | |
2409 | ||
980c9e8c | 2410 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 2411 | |
e3b99556 MM |
2412 | } |
2413 | ||
5228ddc9 RR |
2414 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
2415 | * privs required. */ | |
88255375 | 2416 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 2417 | { |
c8f44aff | 2418 | netdev_features_t features = 0; |
5228ddc9 RR |
2419 | |
2420 | if (arg & TUN_F_CSUM) { | |
88255375 | 2421 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
2422 | arg &= ~TUN_F_CSUM; |
2423 | ||
2424 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
2425 | if (arg & TUN_F_TSO_ECN) { | |
2426 | features |= NETIF_F_TSO_ECN; | |
2427 | arg &= ~TUN_F_TSO_ECN; | |
2428 | } | |
2429 | if (arg & TUN_F_TSO4) | |
2430 | features |= NETIF_F_TSO; | |
2431 | if (arg & TUN_F_TSO6) | |
2432 | features |= NETIF_F_TSO6; | |
2433 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
2434 | } | |
0c19f846 WB |
2435 | |
2436 | arg &= ~TUN_F_UFO; | |
5228ddc9 RR |
2437 | } |
2438 | ||
2439 | /* This gives the user a way to test for new features in future by | |
2440 | * trying to set them. */ | |
2441 | if (arg) | |
2442 | return -EINVAL; | |
2443 | ||
88255375 | 2444 | tun->set_features = features; |
09050957 YI |
2445 | tun->dev->wanted_features &= ~TUN_USER_FEATURES; |
2446 | tun->dev->wanted_features |= features; | |
88255375 | 2447 | netdev_update_features(tun->dev); |
5228ddc9 RR |
2448 | |
2449 | return 0; | |
2450 | } | |
2451 | ||
c8d68e6b JW |
2452 | static void tun_detach_filter(struct tun_struct *tun, int n) |
2453 | { | |
2454 | int i; | |
2455 | struct tun_file *tfile; | |
2456 | ||
2457 | for (i = 0; i < n; i++) { | |
b8deabd3 | 2458 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2459 | lock_sock(tfile->socket.sk); |
2460 | sk_detach_filter(tfile->socket.sk); | |
2461 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2462 | } |
2463 | ||
2464 | tun->filter_attached = false; | |
2465 | } | |
2466 | ||
2467 | static int tun_attach_filter(struct tun_struct *tun) | |
2468 | { | |
2469 | int i, ret = 0; | |
2470 | struct tun_file *tfile; | |
2471 | ||
2472 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2473 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2474 | lock_sock(tfile->socket.sk); |
2475 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
2476 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2477 | if (ret) { |
2478 | tun_detach_filter(tun, i); | |
2479 | return ret; | |
2480 | } | |
2481 | } | |
2482 | ||
2483 | tun->filter_attached = true; | |
2484 | return ret; | |
2485 | } | |
2486 | ||
2487 | static void tun_set_sndbuf(struct tun_struct *tun) | |
2488 | { | |
2489 | struct tun_file *tfile; | |
2490 | int i; | |
2491 | ||
2492 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2493 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
2494 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
2495 | } | |
2496 | } | |
2497 | ||
cde8b15f JW |
2498 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
2499 | { | |
2500 | struct tun_file *tfile = file->private_data; | |
2501 | struct tun_struct *tun; | |
cde8b15f JW |
2502 | int ret = 0; |
2503 | ||
2504 | rtnl_lock(); | |
2505 | ||
2506 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 2507 | tun = tfile->detached; |
5dbbaf2d | 2508 | if (!tun) { |
cde8b15f | 2509 | ret = -EINVAL; |
5dbbaf2d PM |
2510 | goto unlock; |
2511 | } | |
2512 | ret = security_tun_dev_attach_queue(tun->security); | |
2513 | if (ret < 0) | |
2514 | goto unlock; | |
157f3ad9 | 2515 | ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI, |
fdf434d1 | 2516 | tun->flags & IFF_NAPI_FRAGS, true); |
4008e97f | 2517 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 2518 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 2519 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
2520 | ret = -EINVAL; |
2521 | else | |
2522 | __tun_detach(tfile, false); | |
2523 | } else | |
cde8b15f JW |
2524 | ret = -EINVAL; |
2525 | ||
5dbbaf2d | 2526 | unlock: |
cde8b15f JW |
2527 | rtnl_unlock(); |
2528 | return ret; | |
2529 | } | |
2530 | ||
50857e2a AB |
2531 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
2532 | unsigned long arg, int ifreq_len) | |
1da177e4 | 2533 | { |
36b50bab | 2534 | struct tun_file *tfile = file->private_data; |
631ab46b | 2535 | struct tun_struct *tun; |
1da177e4 | 2536 | void __user* argp = (void __user*)arg; |
76db66f7 | 2537 | unsigned int ifindex, carrier; |
1da177e4 | 2538 | struct ifreq ifr; |
0625c883 EB |
2539 | kuid_t owner; |
2540 | kgid_t group; | |
33dccbb0 | 2541 | int sndbuf; |
d9d52b51 | 2542 | int vnet_hdr_sz; |
1cf8e410 | 2543 | int le; |
f271b2cc | 2544 | int ret; |
1da177e4 | 2545 | |
20861f26 | 2546 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || _IOC_TYPE(cmd) == SOCK_IOC_TYPE) { |
50857e2a | 2547 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 2548 | return -EFAULT; |
8bbb1813 | 2549 | } else { |
a117dacd | 2550 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 2551 | } |
631ab46b EB |
2552 | if (cmd == TUNGETFEATURES) { |
2553 | /* Currently this just means: "what IFF flags are valid?". | |
2554 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
2555 | * TUNSETIFF. |
2556 | */ | |
2557 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 2558 | (unsigned int __user*)argp); |
cde8b15f JW |
2559 | } else if (cmd == TUNSETQUEUE) |
2560 | return tun_set_queue(file, &ifr); | |
631ab46b | 2561 | |
c8d68e6b | 2562 | ret = 0; |
876bfd4d HX |
2563 | rtnl_lock(); |
2564 | ||
9484dc74 | 2565 | tun = tun_get(tfile); |
0f16bc13 GF |
2566 | if (cmd == TUNSETIFF) { |
2567 | ret = -EEXIST; | |
2568 | if (tun) | |
2569 | goto unlock; | |
2570 | ||
1da177e4 LT |
2571 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
2572 | ||
140e807d | 2573 | ret = tun_set_iff(sock_net(&tfile->sk), file, &ifr); |
1da177e4 | 2574 | |
876bfd4d HX |
2575 | if (ret) |
2576 | goto unlock; | |
1da177e4 | 2577 | |
50857e2a | 2578 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
2579 | ret = -EFAULT; |
2580 | goto unlock; | |
1da177e4 | 2581 | } |
fb7589a1 PE |
2582 | if (cmd == TUNSETIFINDEX) { |
2583 | ret = -EPERM; | |
2584 | if (tun) | |
2585 | goto unlock; | |
2586 | ||
2587 | ret = -EFAULT; | |
2588 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
2589 | goto unlock; | |
2590 | ||
2591 | ret = 0; | |
2592 | tfile->ifindex = ifindex; | |
2593 | goto unlock; | |
2594 | } | |
1da177e4 | 2595 | |
876bfd4d | 2596 | ret = -EBADFD; |
1da177e4 | 2597 | if (!tun) |
876bfd4d | 2598 | goto unlock; |
1da177e4 | 2599 | |
1e588338 | 2600 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 2601 | |
631ab46b | 2602 | ret = 0; |
1da177e4 | 2603 | switch (cmd) { |
e3b99556 | 2604 | case TUNGETIFF: |
9ce99cf6 | 2605 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 2606 | |
3d407a80 PE |
2607 | if (tfile->detached) |
2608 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
2609 | if (!tfile->socket.sk->sk_filter) |
2610 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 2611 | |
50857e2a | 2612 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 2613 | ret = -EFAULT; |
e3b99556 MM |
2614 | break; |
2615 | ||
1da177e4 LT |
2616 | case TUNSETNOCSUM: |
2617 | /* Disable/Enable checksum */ | |
1da177e4 | 2618 | |
88255375 MM |
2619 | /* [unimplemented] */ |
2620 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 2621 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
2622 | break; |
2623 | ||
2624 | case TUNSETPERSIST: | |
54f968d6 JW |
2625 | /* Disable/Enable persist mode. Keep an extra reference to the |
2626 | * module to prevent the module being unprobed. | |
2627 | */ | |
40630b82 MT |
2628 | if (arg && !(tun->flags & IFF_PERSIST)) { |
2629 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 2630 | __module_get(THIS_MODULE); |
dd38bd85 | 2631 | } |
40630b82 MT |
2632 | if (!arg && (tun->flags & IFF_PERSIST)) { |
2633 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 JW |
2634 | module_put(THIS_MODULE); |
2635 | } | |
1da177e4 | 2636 | |
6b8a66ee JP |
2637 | tun_debug(KERN_INFO, tun, "persist %s\n", |
2638 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
2639 | break; |
2640 | ||
2641 | case TUNSETOWNER: | |
2642 | /* Set owner of the device */ | |
0625c883 EB |
2643 | owner = make_kuid(current_user_ns(), arg); |
2644 | if (!uid_valid(owner)) { | |
2645 | ret = -EINVAL; | |
2646 | break; | |
2647 | } | |
2648 | tun->owner = owner; | |
1e588338 | 2649 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 2650 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
2651 | break; |
2652 | ||
8c644623 GG |
2653 | case TUNSETGROUP: |
2654 | /* Set group of the device */ | |
0625c883 EB |
2655 | group = make_kgid(current_user_ns(), arg); |
2656 | if (!gid_valid(group)) { | |
2657 | ret = -EINVAL; | |
2658 | break; | |
2659 | } | |
2660 | tun->group = group; | |
1e588338 | 2661 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 2662 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
2663 | break; |
2664 | ||
ff4cc3ac MK |
2665 | case TUNSETLINK: |
2666 | /* Only allow setting the type when the interface is down */ | |
2667 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
2668 | tun_debug(KERN_INFO, tun, |
2669 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 2670 | ret = -EBUSY; |
ff4cc3ac MK |
2671 | } else { |
2672 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
2673 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
2674 | tun->dev->type); | |
48abfe05 | 2675 | ret = 0; |
ff4cc3ac | 2676 | } |
631ab46b | 2677 | break; |
ff4cc3ac | 2678 | |
1da177e4 LT |
2679 | #ifdef TUN_DEBUG |
2680 | case TUNSETDEBUG: | |
2681 | tun->debug = arg; | |
2682 | break; | |
2683 | #endif | |
5228ddc9 | 2684 | case TUNSETOFFLOAD: |
88255375 | 2685 | ret = set_offload(tun, arg); |
631ab46b | 2686 | break; |
5228ddc9 | 2687 | |
f271b2cc MK |
2688 | case TUNSETTXFILTER: |
2689 | /* Can be set only for TAPs */ | |
631ab46b | 2690 | ret = -EINVAL; |
40630b82 | 2691 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 2692 | break; |
c0e5a8c2 | 2693 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 2694 | break; |
1da177e4 LT |
2695 | |
2696 | case SIOCGIFHWADDR: | |
b595076a | 2697 | /* Get hw address */ |
f271b2cc MK |
2698 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
2699 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 2700 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
2701 | ret = -EFAULT; |
2702 | break; | |
1da177e4 LT |
2703 | |
2704 | case SIOCSIFHWADDR: | |
f271b2cc | 2705 | /* Set hw address */ |
6b8a66ee JP |
2706 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
2707 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 2708 | |
40102371 | 2709 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 2710 | break; |
33dccbb0 HX |
2711 | |
2712 | case TUNGETSNDBUF: | |
54f968d6 | 2713 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
2714 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
2715 | ret = -EFAULT; | |
2716 | break; | |
2717 | ||
2718 | case TUNSETSNDBUF: | |
2719 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
2720 | ret = -EFAULT; | |
2721 | break; | |
2722 | } | |
93161922 CG |
2723 | if (sndbuf <= 0) { |
2724 | ret = -EINVAL; | |
2725 | break; | |
2726 | } | |
33dccbb0 | 2727 | |
c8d68e6b JW |
2728 | tun->sndbuf = sndbuf; |
2729 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
2730 | break; |
2731 | ||
d9d52b51 MT |
2732 | case TUNGETVNETHDRSZ: |
2733 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
2734 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
2735 | ret = -EFAULT; | |
2736 | break; | |
2737 | ||
2738 | case TUNSETVNETHDRSZ: | |
2739 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
2740 | ret = -EFAULT; | |
2741 | break; | |
2742 | } | |
2743 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
2744 | ret = -EINVAL; | |
2745 | break; | |
2746 | } | |
2747 | ||
2748 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
2749 | break; | |
2750 | ||
1cf8e410 MT |
2751 | case TUNGETVNETLE: |
2752 | le = !!(tun->flags & TUN_VNET_LE); | |
2753 | if (put_user(le, (int __user *)argp)) | |
2754 | ret = -EFAULT; | |
2755 | break; | |
2756 | ||
2757 | case TUNSETVNETLE: | |
2758 | if (get_user(le, (int __user *)argp)) { | |
2759 | ret = -EFAULT; | |
2760 | break; | |
2761 | } | |
2762 | if (le) | |
2763 | tun->flags |= TUN_VNET_LE; | |
2764 | else | |
2765 | tun->flags &= ~TUN_VNET_LE; | |
2766 | break; | |
2767 | ||
8b8e658b GK |
2768 | case TUNGETVNETBE: |
2769 | ret = tun_get_vnet_be(tun, argp); | |
2770 | break; | |
2771 | ||
2772 | case TUNSETVNETBE: | |
2773 | ret = tun_set_vnet_be(tun, argp); | |
2774 | break; | |
2775 | ||
99405162 MT |
2776 | case TUNATTACHFILTER: |
2777 | /* Can be set only for TAPs */ | |
2778 | ret = -EINVAL; | |
40630b82 | 2779 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
2780 | break; |
2781 | ret = -EFAULT; | |
54f968d6 | 2782 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
2783 | break; |
2784 | ||
c8d68e6b | 2785 | ret = tun_attach_filter(tun); |
99405162 MT |
2786 | break; |
2787 | ||
2788 | case TUNDETACHFILTER: | |
2789 | /* Can be set only for TAPs */ | |
2790 | ret = -EINVAL; | |
40630b82 | 2791 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 2792 | break; |
c8d68e6b JW |
2793 | ret = 0; |
2794 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
2795 | break; |
2796 | ||
76975e9c PE |
2797 | case TUNGETFILTER: |
2798 | ret = -EINVAL; | |
40630b82 | 2799 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
2800 | break; |
2801 | ret = -EFAULT; | |
2802 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
2803 | break; | |
2804 | ret = 0; | |
2805 | break; | |
2806 | ||
76db66f7 ND |
2807 | case TUNSETCARRIER: |
2808 | ret = -EFAULT; | |
2809 | if (copy_from_user(&carrier, argp, sizeof(carrier))) | |
2810 | goto unlock; | |
2811 | ||
2812 | ret = tun_net_change_carrier(tun->dev, (bool)carrier); | |
2813 | break; | |
2814 | ||
1da177e4 | 2815 | default: |
631ab46b EB |
2816 | ret = -EINVAL; |
2817 | break; | |
ee289b64 | 2818 | } |
1da177e4 | 2819 | |
876bfd4d HX |
2820 | unlock: |
2821 | rtnl_unlock(); | |
2822 | if (tun) | |
2823 | tun_put(tun); | |
631ab46b | 2824 | return ret; |
1da177e4 LT |
2825 | } |
2826 | ||
50857e2a AB |
2827 | static long tun_chr_ioctl(struct file *file, |
2828 | unsigned int cmd, unsigned long arg) | |
2829 | { | |
2830 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
2831 | } | |
2832 | ||
2833 | #ifdef CONFIG_COMPAT | |
2834 | static long tun_chr_compat_ioctl(struct file *file, | |
2835 | unsigned int cmd, unsigned long arg) | |
2836 | { | |
2837 | switch (cmd) { | |
2838 | case TUNSETIFF: | |
2839 | case TUNGETIFF: | |
2840 | case TUNSETTXFILTER: | |
2841 | case TUNGETSNDBUF: | |
2842 | case TUNSETSNDBUF: | |
2843 | case SIOCGIFHWADDR: | |
2844 | case SIOCSIFHWADDR: | |
2845 | arg = (unsigned long)compat_ptr(arg); | |
2846 | break; | |
2847 | default: | |
2848 | arg = (compat_ulong_t)arg; | |
2849 | break; | |
2850 | } | |
2851 | ||
2852 | /* | |
2853 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
2854 | * the end of that structure. All fields that are used in this | |
2855 | * driver are compatible though, we don't need to convert the | |
2856 | * contents. | |
2857 | */ | |
2858 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
2859 | } | |
2860 | #endif /* CONFIG_COMPAT */ | |
2861 | ||
1da177e4 LT |
2862 | static int tun_chr_fasync(int fd, struct file *file, int on) |
2863 | { | |
54f968d6 | 2864 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2865 | int ret; |
2866 | ||
54f968d6 | 2867 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 2868 | goto out; |
6aa20a22 | 2869 | |
1da177e4 | 2870 | if (on) { |
e0b93edd | 2871 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 2872 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 2873 | } else |
54f968d6 | 2874 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
2875 | ret = 0; |
2876 | out: | |
9d319522 | 2877 | return ret; |
1da177e4 LT |
2878 | } |
2879 | ||
2880 | static int tun_chr_open(struct inode *inode, struct file * file) | |
2881 | { | |
140e807d | 2882 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 2883 | struct tun_file *tfile; |
deed49fb | 2884 | |
6b8a66ee | 2885 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 2886 | |
140e807d | 2887 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 2888 | &tun_proto, 0); |
631ab46b EB |
2889 | if (!tfile) |
2890 | return -ENOMEM; | |
48ed2557 JW |
2891 | if (skb_array_init(&tfile->tx_array, 0, GFP_KERNEL)) { |
2892 | sk_free(&tfile->sk); | |
2893 | return -ENOMEM; | |
2894 | } | |
2895 | ||
ea65e06d | 2896 | mutex_init(&tfile->napi_mutex); |
c956674b | 2897 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 2898 | tfile->flags = 0; |
fb7589a1 | 2899 | tfile->ifindex = 0; |
54f968d6 | 2900 | |
54f968d6 | 2901 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 2902 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
2903 | |
2904 | tfile->socket.file = file; | |
2905 | tfile->socket.ops = &tun_socket_ops; | |
2906 | ||
2907 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
2908 | |
2909 | tfile->sk.sk_write_space = tun_sock_write_space; | |
2910 | tfile->sk.sk_sndbuf = INT_MAX; | |
2911 | ||
631ab46b | 2912 | file->private_data = tfile; |
4008e97f | 2913 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 2914 | |
19a6afb2 JW |
2915 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
2916 | ||
1da177e4 LT |
2917 | return 0; |
2918 | } | |
2919 | ||
2920 | static int tun_chr_close(struct inode *inode, struct file *file) | |
2921 | { | |
631ab46b | 2922 | struct tun_file *tfile = file->private_data; |
1da177e4 | 2923 | |
c8d68e6b | 2924 | tun_detach(tfile, true); |
1da177e4 LT |
2925 | |
2926 | return 0; | |
2927 | } | |
2928 | ||
93e14b6d | 2929 | #ifdef CONFIG_PROC_FS |
9484dc74 | 2930 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file) |
93e14b6d | 2931 | { |
9484dc74 | 2932 | struct tun_file *tfile = file->private_data; |
93e14b6d MY |
2933 | struct tun_struct *tun; |
2934 | struct ifreq ifr; | |
2935 | ||
2936 | memset(&ifr, 0, sizeof(ifr)); | |
2937 | ||
2938 | rtnl_lock(); | |
9484dc74 | 2939 | tun = tun_get(tfile); |
93e14b6d MY |
2940 | if (tun) |
2941 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
2942 | rtnl_unlock(); | |
2943 | ||
2944 | if (tun) | |
2945 | tun_put(tun); | |
2946 | ||
a3816ab0 | 2947 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
2948 | } |
2949 | #endif | |
2950 | ||
d54b1fdb | 2951 | static const struct file_operations tun_fops = { |
6aa20a22 | 2952 | .owner = THIS_MODULE, |
1da177e4 | 2953 | .llseek = no_llseek, |
9b067034 | 2954 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 2955 | .write_iter = tun_chr_write_iter, |
1da177e4 | 2956 | .poll = tun_chr_poll, |
50857e2a AB |
2957 | .unlocked_ioctl = tun_chr_ioctl, |
2958 | #ifdef CONFIG_COMPAT | |
2959 | .compat_ioctl = tun_chr_compat_ioctl, | |
2960 | #endif | |
1da177e4 LT |
2961 | .open = tun_chr_open, |
2962 | .release = tun_chr_close, | |
93e14b6d MY |
2963 | .fasync = tun_chr_fasync, |
2964 | #ifdef CONFIG_PROC_FS | |
2965 | .show_fdinfo = tun_chr_show_fdinfo, | |
2966 | #endif | |
1da177e4 LT |
2967 | }; |
2968 | ||
2969 | static struct miscdevice tun_miscdev = { | |
2970 | .minor = TUN_MINOR, | |
2971 | .name = "tun", | |
e454cea2 | 2972 | .nodename = "net/tun", |
1da177e4 | 2973 | .fops = &tun_fops, |
1da177e4 LT |
2974 | }; |
2975 | ||
2976 | /* ethtool interface */ | |
2977 | ||
29ccc49d PR |
2978 | static int tun_get_link_ksettings(struct net_device *dev, |
2979 | struct ethtool_link_ksettings *cmd) | |
2980 | { | |
2981 | ethtool_link_ksettings_zero_link_mode(cmd, supported); | |
2982 | ethtool_link_ksettings_zero_link_mode(cmd, advertising); | |
2983 | cmd->base.speed = SPEED_10; | |
2984 | cmd->base.duplex = DUPLEX_FULL; | |
2985 | cmd->base.port = PORT_TP; | |
2986 | cmd->base.phy_address = 0; | |
2987 | cmd->base.autoneg = AUTONEG_DISABLE; | |
1da177e4 LT |
2988 | return 0; |
2989 | } | |
2990 | ||
2991 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
2992 | { | |
2993 | struct tun_struct *tun = netdev_priv(dev); | |
2994 | ||
33a5ba14 RJ |
2995 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
2996 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
2997 | |
2998 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 2999 | case IFF_TUN: |
33a5ba14 | 3000 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 3001 | break; |
40630b82 | 3002 | case IFF_TAP: |
33a5ba14 | 3003 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
3004 | break; |
3005 | } | |
3006 | } | |
3007 | ||
3008 | static u32 tun_get_msglevel(struct net_device *dev) | |
3009 | { | |
3010 | #ifdef TUN_DEBUG | |
3011 | struct tun_struct *tun = netdev_priv(dev); | |
3012 | return tun->debug; | |
3013 | #else | |
3014 | return -EOPNOTSUPP; | |
3015 | #endif | |
3016 | } | |
3017 | ||
3018 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
3019 | { | |
3020 | #ifdef TUN_DEBUG | |
3021 | struct tun_struct *tun = netdev_priv(dev); | |
3022 | tun->debug = value; | |
3023 | #endif | |
3024 | } | |
3025 | ||
5503fcec JW |
3026 | static int tun_get_coalesce(struct net_device *dev, |
3027 | struct ethtool_coalesce *ec) | |
3028 | { | |
3029 | struct tun_struct *tun = netdev_priv(dev); | |
3030 | ||
3031 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
3032 | ||
3033 | return 0; | |
3034 | } | |
3035 | ||
3036 | static int tun_set_coalesce(struct net_device *dev, | |
3037 | struct ethtool_coalesce *ec) | |
3038 | { | |
3039 | struct tun_struct *tun = netdev_priv(dev); | |
3040 | ||
3041 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
3042 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
3043 | else | |
3044 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
3045 | ||
3046 | return 0; | |
3047 | } | |
3048 | ||
7282d491 | 3049 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
3050 | .get_drvinfo = tun_get_drvinfo, |
3051 | .get_msglevel = tun_get_msglevel, | |
3052 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 3053 | .get_link = ethtool_op_get_link, |
eda29772 | 3054 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
3055 | .get_coalesce = tun_get_coalesce, |
3056 | .set_coalesce = tun_set_coalesce, | |
29ccc49d | 3057 | .get_link_ksettings = tun_get_link_ksettings, |
1da177e4 LT |
3058 | }; |
3059 | ||
1576d986 JW |
3060 | static int tun_queue_resize(struct tun_struct *tun) |
3061 | { | |
3062 | struct net_device *dev = tun->dev; | |
3063 | struct tun_file *tfile; | |
3064 | struct skb_array **arrays; | |
3065 | int n = tun->numqueues + tun->numdisabled; | |
3066 | int ret, i; | |
3067 | ||
12039046 | 3068 | arrays = kmalloc_array(n, sizeof(*arrays), GFP_KERNEL); |
1576d986 JW |
3069 | if (!arrays) |
3070 | return -ENOMEM; | |
3071 | ||
3072 | for (i = 0; i < tun->numqueues; i++) { | |
3073 | tfile = rtnl_dereference(tun->tfiles[i]); | |
3074 | arrays[i] = &tfile->tx_array; | |
3075 | } | |
3076 | list_for_each_entry(tfile, &tun->disabled, next) | |
3077 | arrays[i++] = &tfile->tx_array; | |
3078 | ||
3079 | ret = skb_array_resize_multiple(arrays, n, | |
3080 | dev->tx_queue_len, GFP_KERNEL); | |
3081 | ||
3082 | kfree(arrays); | |
3083 | return ret; | |
3084 | } | |
3085 | ||
3086 | static int tun_device_event(struct notifier_block *unused, | |
3087 | unsigned long event, void *ptr) | |
3088 | { | |
3089 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
3090 | struct tun_struct *tun = netdev_priv(dev); | |
97f6a877 | 3091 | int i; |
1576d986 | 3092 | |
86dfb4ac CG |
3093 | if (dev->rtnl_link_ops != &tun_link_ops) |
3094 | return NOTIFY_DONE; | |
3095 | ||
1576d986 JW |
3096 | switch (event) { |
3097 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
3098 | if (tun_queue_resize(tun)) | |
3099 | return NOTIFY_BAD; | |
3100 | break; | |
97f6a877 FL |
3101 | case NETDEV_UP: |
3102 | for (i = 0; i < tun->numqueues; i++) { | |
3103 | struct tun_file *tfile; | |
3104 | ||
3105 | tfile = rtnl_dereference(tun->tfiles[i]); | |
3106 | tfile->socket.sk->sk_write_space(tfile->socket.sk); | |
3107 | } | |
3108 | break; | |
1576d986 JW |
3109 | default: |
3110 | break; | |
3111 | } | |
3112 | ||
3113 | return NOTIFY_DONE; | |
3114 | } | |
3115 | ||
3116 | static struct notifier_block tun_notifier_block __read_mostly = { | |
3117 | .notifier_call = tun_device_event, | |
3118 | }; | |
79d17604 | 3119 | |
1da177e4 LT |
3120 | static int __init tun_init(void) |
3121 | { | |
3122 | int ret = 0; | |
3123 | ||
6b8a66ee | 3124 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
1da177e4 | 3125 | |
f019a7a5 | 3126 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 3127 | if (ret) { |
6b8a66ee | 3128 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 3129 | goto err_linkops; |
79d17604 PE |
3130 | } |
3131 | ||
1da177e4 | 3132 | ret = misc_register(&tun_miscdev); |
79d17604 | 3133 | if (ret) { |
6b8a66ee | 3134 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
3135 | goto err_misc; |
3136 | } | |
1576d986 | 3137 | |
5edfbd3c TZ |
3138 | ret = register_netdevice_notifier(&tun_notifier_block); |
3139 | if (ret) { | |
3140 | pr_err("Can't register netdevice notifier\n"); | |
3141 | goto err_notifier; | |
3142 | } | |
3143 | ||
f019a7a5 | 3144 | return 0; |
5edfbd3c TZ |
3145 | |
3146 | err_notifier: | |
3147 | misc_deregister(&tun_miscdev); | |
79d17604 | 3148 | err_misc: |
f019a7a5 EB |
3149 | rtnl_link_unregister(&tun_link_ops); |
3150 | err_linkops: | |
1da177e4 LT |
3151 | return ret; |
3152 | } | |
3153 | ||
3154 | static void tun_cleanup(void) | |
3155 | { | |
6aa20a22 | 3156 | misc_deregister(&tun_miscdev); |
f019a7a5 | 3157 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 3158 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
3159 | } |
3160 | ||
05c2828c MT |
3161 | /* Get an underlying socket object from tun file. Returns error unless file is |
3162 | * attached to a device. The returned object works like a packet socket, it | |
3163 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
3164 | * holding a reference to the file for as long as the socket is in use. */ | |
3165 | struct socket *tun_get_socket(struct file *file) | |
3166 | { | |
6e914fc7 | 3167 | struct tun_file *tfile; |
05c2828c MT |
3168 | if (file->f_op != &tun_fops) |
3169 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
3170 | tfile = file->private_data; |
3171 | if (!tfile) | |
05c2828c | 3172 | return ERR_PTR(-EBADFD); |
54f968d6 | 3173 | return &tfile->socket; |
05c2828c MT |
3174 | } |
3175 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
3176 | ||
83339c6b JW |
3177 | struct skb_array *tun_get_skb_array(struct file *file) |
3178 | { | |
3179 | struct tun_file *tfile; | |
3180 | ||
3181 | if (file->f_op != &tun_fops) | |
3182 | return ERR_PTR(-EINVAL); | |
3183 | tfile = file->private_data; | |
3184 | if (!tfile) | |
3185 | return ERR_PTR(-EBADFD); | |
3186 | return &tfile->tx_array; | |
3187 | } | |
3188 | EXPORT_SYMBOL_GPL(tun_get_skb_array); | |
3189 | ||
1da177e4 LT |
3190 | module_init(tun_init); |
3191 | module_exit(tun_cleanup); | |
3192 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
3193 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
3194 | MODULE_LICENSE("GPL"); | |
3195 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 3196 | MODULE_ALIAS("devname:net/tun"); |