Commit | Line | Data |
---|---|---|
c942fddf | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * TUN - Universal TUN/TAP device driver. | |
4 | * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com> | |
5 | * | |
1da177e4 LT |
6 | * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $ |
7 | */ | |
8 | ||
9 | /* | |
10 | * Changes: | |
11 | * | |
ff4cc3ac MK |
12 | * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14 |
13 | * Add TUNSETLINK ioctl to set the link encapsulation | |
14 | * | |
1da177e4 | 15 | * Mark Smith <markzzzsmith@yahoo.com.au> |
344dc8ed | 16 | * Use eth_random_addr() for tap MAC address. |
1da177e4 LT |
17 | * |
18 | * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20 | |
19 | * Fixes in packet dropping, queue length setting and queue wakeup. | |
20 | * Increased default tx queue length. | |
21 | * Added ethtool API. | |
22 | * Minor cleanups | |
23 | * | |
24 | * Daniel Podlejski <underley@underley.eu.org> | |
25 | * Modifications for 2.3.99-pre5 kernel. | |
26 | */ | |
27 | ||
6b8a66ee JP |
28 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
29 | ||
1da177e4 LT |
30 | #define DRV_NAME "tun" |
31 | #define DRV_VERSION "1.6" | |
32 | #define DRV_DESCRIPTION "Universal TUN/TAP device driver" | |
33 | #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>" | |
34 | ||
1da177e4 LT |
35 | #include <linux/module.h> |
36 | #include <linux/errno.h> | |
37 | #include <linux/kernel.h> | |
174cd4b1 | 38 | #include <linux/sched/signal.h> |
1da177e4 LT |
39 | #include <linux/major.h> |
40 | #include <linux/slab.h> | |
41 | #include <linux/poll.h> | |
42 | #include <linux/fcntl.h> | |
43 | #include <linux/init.h> | |
44 | #include <linux/skbuff.h> | |
45 | #include <linux/netdevice.h> | |
46 | #include <linux/etherdevice.h> | |
47 | #include <linux/miscdevice.h> | |
48 | #include <linux/ethtool.h> | |
49 | #include <linux/rtnetlink.h> | |
50857e2a | 50 | #include <linux/compat.h> |
1da177e4 LT |
51 | #include <linux/if.h> |
52 | #include <linux/if_arp.h> | |
53 | #include <linux/if_ether.h> | |
54 | #include <linux/if_tun.h> | |
6680ec68 | 55 | #include <linux/if_vlan.h> |
1da177e4 | 56 | #include <linux/crc32.h> |
d647a591 | 57 | #include <linux/nsproxy.h> |
f43798c2 | 58 | #include <linux/virtio_net.h> |
99405162 | 59 | #include <linux/rcupdate.h> |
881d966b | 60 | #include <net/net_namespace.h> |
79d17604 | 61 | #include <net/netns/generic.h> |
f019a7a5 | 62 | #include <net/rtnetlink.h> |
33dccbb0 | 63 | #include <net/sock.h> |
735fc405 | 64 | #include <net/xdp.h> |
b9815eb1 | 65 | #include <net/ip_tunnels.h> |
93e14b6d | 66 | #include <linux/seq_file.h> |
e0b46d0e | 67 | #include <linux/uio.h> |
1576d986 | 68 | #include <linux/skb_array.h> |
761876c8 JW |
69 | #include <linux/bpf.h> |
70 | #include <linux/bpf_trace.h> | |
90e33d45 | 71 | #include <linux/mutex.h> |
cca8ea3b PP |
72 | #include <linux/ieee802154.h> |
73 | #include <linux/if_ltalk.h> | |
74 | #include <uapi/linux/if_fddi.h> | |
75 | #include <uapi/linux/if_hippi.h> | |
76 | #include <uapi/linux/if_fc.h> | |
77 | #include <net/ax25.h> | |
78 | #include <net/rose.h> | |
79 | #include <net/6lowpan.h> | |
1da177e4 | 80 | |
7c0f6ba6 | 81 | #include <linux/uaccess.h> |
f2780d6d | 82 | #include <linux/proc_fs.h> |
1da177e4 | 83 | |
4e24f2dd CW |
84 | static void tun_default_link_ksettings(struct net_device *dev, |
85 | struct ethtool_link_ksettings *cmd); | |
86 | ||
7df13219 | 87 | #define TUN_RX_PAD (NET_IP_ALIGN + NET_SKB_PAD) |
66ccbc9c | 88 | |
031f5e03 MT |
89 | /* TUN device flags */ |
90 | ||
91 | /* IFF_ATTACH_QUEUE is never stored in device flags, | |
92 | * overload it to mean fasync when stored there. | |
93 | */ | |
94 | #define TUN_FASYNC IFF_ATTACH_QUEUE | |
1cf8e410 MT |
95 | /* High bits in flags field are unused. */ |
96 | #define TUN_VNET_LE 0x80000000 | |
8b8e658b | 97 | #define TUN_VNET_BE 0x40000000 |
031f5e03 MT |
98 | |
99 | #define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \ | |
90e33d45 PP |
100 | IFF_MULTI_QUEUE | IFF_NAPI | IFF_NAPI_FRAGS) |
101 | ||
0690899b MT |
102 | #define GOODCOPY_LEN 128 |
103 | ||
f271b2cc MK |
104 | #define FLT_EXACT_COUNT 8 |
105 | struct tap_filter { | |
106 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
107 | u32 mask[2]; /* Mask of the hashed addrs */ | |
108 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
109 | }; | |
110 | ||
baf71c5c PG |
111 | /* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal |
112 | * to max number of VCPUs in guest. */ | |
113 | #define MAX_TAP_QUEUES 256 | |
b8732fb7 | 114 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 115 | |
96442e42 JW |
116 | #define TUN_FLOW_EXPIRE (3 * HZ) |
117 | ||
54f968d6 | 118 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 119 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
120 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
121 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 122 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 123 | * this). |
6e914fc7 JW |
124 | * |
125 | * RCU usage: | |
36fe8c09 | 126 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 127 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 128 | */ |
631ab46b | 129 | struct tun_file { |
54f968d6 JW |
130 | struct sock sk; |
131 | struct socket socket; | |
6e914fc7 | 132 | struct tun_struct __rcu *tun; |
54f968d6 JW |
133 | struct fasync_struct *fasync; |
134 | /* only used for fasnyc */ | |
135 | unsigned int flags; | |
fb7589a1 PE |
136 | union { |
137 | u16 queue_index; | |
138 | unsigned int ifindex; | |
139 | }; | |
94317099 | 140 | struct napi_struct napi; |
aec72f33 | 141 | bool napi_enabled; |
af3fb24e | 142 | bool napi_frags_enabled; |
90e33d45 | 143 | struct mutex napi_mutex; /* Protects access to the above napi */ |
4008e97f JW |
144 | struct list_head next; |
145 | struct tun_struct *detached; | |
5990a305 | 146 | struct ptr_ring tx_ring; |
8bf5c4ee | 147 | struct xdp_rxq_info xdp_rxq; |
631ab46b EB |
148 | }; |
149 | ||
f9e06c45 JW |
150 | struct tun_page { |
151 | struct page *page; | |
152 | int count; | |
153 | }; | |
154 | ||
96442e42 JW |
155 | struct tun_flow_entry { |
156 | struct hlist_node hash_link; | |
157 | struct rcu_head rcu; | |
158 | struct tun_struct *tun; | |
159 | ||
160 | u32 rxhash; | |
9bc88939 | 161 | u32 rps_rxhash; |
96442e42 | 162 | int queue_index; |
83b1bc12 | 163 | unsigned long updated ____cacheline_aligned_in_smp; |
96442e42 JW |
164 | }; |
165 | ||
166 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
f13b5468 | 167 | #define TUN_MASK_FLOW_ENTRIES (TUN_NUM_FLOW_ENTRIES - 1) |
96442e42 | 168 | |
cd5681d7 | 169 | struct tun_prog { |
96f84061 JW |
170 | struct rcu_head rcu; |
171 | struct bpf_prog *prog; | |
172 | }; | |
173 | ||
54f968d6 | 174 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 175 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
176 | * file were attached to a persist device. |
177 | */ | |
14daa021 | 178 | struct tun_struct { |
c8d68e6b JW |
179 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
180 | unsigned int numqueues; | |
f271b2cc | 181 | unsigned int flags; |
0625c883 EB |
182 | kuid_t owner; |
183 | kgid_t group; | |
14daa021 | 184 | |
14daa021 | 185 | struct net_device *dev; |
c8f44aff | 186 | netdev_features_t set_features; |
88255375 | 187 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
399e0827 | 188 | NETIF_F_TSO6 | NETIF_F_GSO_UDP_L4) |
d9d52b51 | 189 | |
eaea34b2 | 190 | int align; |
d9d52b51 | 191 | int vnet_hdr_sz; |
54f968d6 JW |
192 | int sndbuf; |
193 | struct tap_filter txflt; | |
194 | struct sock_fprog fprog; | |
195 | /* protected by rtnl lock */ | |
196 | bool filter_attached; | |
3424170f | 197 | u32 msg_enable; |
96442e42 | 198 | spinlock_t lock; |
96442e42 JW |
199 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
200 | struct timer_list flow_gc_timer; | |
201 | unsigned long ageing_time; | |
4008e97f JW |
202 | unsigned int numdisabled; |
203 | struct list_head disabled; | |
5dbbaf2d | 204 | void *security; |
b8732fb7 | 205 | u32 flow_count; |
5503fcec | 206 | u32 rx_batched; |
497a5757 | 207 | atomic_long_t rx_frame_errors; |
761876c8 | 208 | struct bpf_prog __rcu *xdp_prog; |
cd5681d7 | 209 | struct tun_prog __rcu *steering_prog; |
aff3d70a | 210 | struct tun_prog __rcu *filter_prog; |
4e24f2dd | 211 | struct ethtool_link_ksettings link_ksettings; |
158b515f GK |
212 | /* init args */ |
213 | struct file *file; | |
214 | struct ifreq *ifr; | |
14daa021 | 215 | }; |
1da177e4 | 216 | |
aff3d70a JW |
217 | struct veth { |
218 | __be16 h_vlan_proto; | |
219 | __be16 h_vlan_TCI; | |
14daa021 | 220 | }; |
1da177e4 | 221 | |
158b515f GK |
222 | static void tun_flow_init(struct tun_struct *tun); |
223 | static void tun_flow_uninit(struct tun_struct *tun); | |
224 | ||
94317099 PP |
225 | static int tun_napi_receive(struct napi_struct *napi, int budget) |
226 | { | |
227 | struct tun_file *tfile = container_of(napi, struct tun_file, napi); | |
228 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
229 | struct sk_buff_head process_queue; | |
230 | struct sk_buff *skb; | |
231 | int received = 0; | |
232 | ||
233 | __skb_queue_head_init(&process_queue); | |
234 | ||
235 | spin_lock(&queue->lock); | |
236 | skb_queue_splice_tail_init(queue, &process_queue); | |
237 | spin_unlock(&queue->lock); | |
238 | ||
239 | while (received < budget && (skb = __skb_dequeue(&process_queue))) { | |
240 | napi_gro_receive(napi, skb); | |
241 | ++received; | |
242 | } | |
243 | ||
244 | if (!skb_queue_empty(&process_queue)) { | |
245 | spin_lock(&queue->lock); | |
246 | skb_queue_splice(&process_queue, queue); | |
247 | spin_unlock(&queue->lock); | |
248 | } | |
249 | ||
250 | return received; | |
251 | } | |
252 | ||
253 | static int tun_napi_poll(struct napi_struct *napi, int budget) | |
254 | { | |
255 | unsigned int received; | |
256 | ||
257 | received = tun_napi_receive(napi, budget); | |
258 | ||
259 | if (received < budget) | |
260 | napi_complete_done(napi, received); | |
261 | ||
262 | return received; | |
263 | } | |
264 | ||
265 | static void tun_napi_init(struct tun_struct *tun, struct tun_file *tfile, | |
af3fb24e | 266 | bool napi_en, bool napi_frags) |
94317099 | 267 | { |
aec72f33 | 268 | tfile->napi_enabled = napi_en; |
af3fb24e | 269 | tfile->napi_frags_enabled = napi_en && napi_frags; |
94317099 | 270 | if (napi_en) { |
16d083e2 | 271 | netif_napi_add_tx(tun->dev, &tfile->napi, tun_napi_poll); |
94317099 PP |
272 | napi_enable(&tfile->napi); |
273 | } | |
274 | } | |
275 | ||
a8fc8cb5 JK |
276 | static void tun_napi_enable(struct tun_file *tfile) |
277 | { | |
278 | if (tfile->napi_enabled) | |
279 | napi_enable(&tfile->napi); | |
280 | } | |
281 | ||
06e55add | 282 | static void tun_napi_disable(struct tun_file *tfile) |
94317099 | 283 | { |
aec72f33 | 284 | if (tfile->napi_enabled) |
94317099 PP |
285 | napi_disable(&tfile->napi); |
286 | } | |
287 | ||
06e55add | 288 | static void tun_napi_del(struct tun_file *tfile) |
94317099 | 289 | { |
aec72f33 | 290 | if (tfile->napi_enabled) |
94317099 PP |
291 | netif_napi_del(&tfile->napi); |
292 | } | |
293 | ||
af3fb24e | 294 | static bool tun_napi_frags_enabled(const struct tun_file *tfile) |
90e33d45 | 295 | { |
af3fb24e | 296 | return tfile->napi_frags_enabled; |
90e33d45 PP |
297 | } |
298 | ||
8b8e658b GK |
299 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
300 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
301 | { | |
302 | return tun->flags & TUN_VNET_BE ? false : | |
303 | virtio_legacy_is_little_endian(); | |
304 | } | |
305 | ||
306 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
307 | { | |
308 | int be = !!(tun->flags & TUN_VNET_BE); | |
309 | ||
310 | if (put_user(be, argp)) | |
311 | return -EFAULT; | |
312 | ||
313 | return 0; | |
314 | } | |
315 | ||
316 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
317 | { | |
318 | int be; | |
319 | ||
320 | if (get_user(be, argp)) | |
321 | return -EFAULT; | |
322 | ||
323 | if (be) | |
324 | tun->flags |= TUN_VNET_BE; | |
325 | else | |
326 | tun->flags &= ~TUN_VNET_BE; | |
327 | ||
328 | return 0; | |
329 | } | |
330 | #else | |
331 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
332 | { | |
333 | return virtio_legacy_is_little_endian(); | |
334 | } | |
335 | ||
336 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
337 | { | |
338 | return -EINVAL; | |
339 | } | |
340 | ||
341 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
342 | { | |
343 | return -EINVAL; | |
344 | } | |
345 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
346 | ||
25bd55bb GK |
347 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
348 | { | |
7d824109 | 349 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 350 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
351 | } |
352 | ||
56f0dcc5 MT |
353 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
354 | { | |
25bd55bb | 355 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
356 | } |
357 | ||
358 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
359 | { | |
25bd55bb | 360 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
361 | } |
362 | ||
96442e42 JW |
363 | static inline u32 tun_hashfn(u32 rxhash) |
364 | { | |
f13b5468 | 365 | return rxhash & TUN_MASK_FLOW_ENTRIES; |
96442e42 JW |
366 | } |
367 | ||
368 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
369 | { | |
370 | struct tun_flow_entry *e; | |
96442e42 | 371 | |
b67bfe0d | 372 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
373 | if (e->rxhash == rxhash) |
374 | return e; | |
375 | } | |
376 | return NULL; | |
377 | } | |
378 | ||
379 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
380 | struct hlist_head *head, | |
381 | u32 rxhash, u16 queue_index) | |
382 | { | |
9fdc6bef ED |
383 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
384 | ||
96442e42 | 385 | if (e) { |
3424170f MK |
386 | netif_info(tun, tx_queued, tun->dev, |
387 | "create flow: hash %u index %u\n", | |
388 | rxhash, queue_index); | |
96442e42 JW |
389 | e->updated = jiffies; |
390 | e->rxhash = rxhash; | |
9bc88939 | 391 | e->rps_rxhash = 0; |
96442e42 JW |
392 | e->queue_index = queue_index; |
393 | e->tun = tun; | |
394 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 395 | ++tun->flow_count; |
96442e42 JW |
396 | } |
397 | return e; | |
398 | } | |
399 | ||
96442e42 JW |
400 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
401 | { | |
3424170f MK |
402 | netif_info(tun, tx_queued, tun->dev, "delete flow: hash %u index %u\n", |
403 | e->rxhash, e->queue_index); | |
96442e42 | 404 | hlist_del_rcu(&e->hash_link); |
9fdc6bef | 405 | kfree_rcu(e, rcu); |
b8732fb7 | 406 | --tun->flow_count; |
96442e42 JW |
407 | } |
408 | ||
409 | static void tun_flow_flush(struct tun_struct *tun) | |
410 | { | |
411 | int i; | |
412 | ||
413 | spin_lock_bh(&tun->lock); | |
414 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
415 | struct tun_flow_entry *e; | |
b67bfe0d | 416 | struct hlist_node *n; |
96442e42 | 417 | |
b67bfe0d | 418 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
419 | tun_flow_delete(tun, e); |
420 | } | |
421 | spin_unlock_bh(&tun->lock); | |
422 | } | |
423 | ||
424 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
425 | { | |
426 | int i; | |
427 | ||
428 | spin_lock_bh(&tun->lock); | |
429 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
430 | struct tun_flow_entry *e; | |
b67bfe0d | 431 | struct hlist_node *n; |
96442e42 | 432 | |
b67bfe0d | 433 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
434 | if (e->queue_index == queue_index) |
435 | tun_flow_delete(tun, e); | |
436 | } | |
437 | } | |
438 | spin_unlock_bh(&tun->lock); | |
439 | } | |
440 | ||
e99e88a9 | 441 | static void tun_flow_cleanup(struct timer_list *t) |
96442e42 | 442 | { |
e99e88a9 | 443 | struct tun_struct *tun = from_timer(tun, t, flow_gc_timer); |
96442e42 JW |
444 | unsigned long delay = tun->ageing_time; |
445 | unsigned long next_timer = jiffies + delay; | |
446 | unsigned long count = 0; | |
447 | int i; | |
448 | ||
7dbfb4ef | 449 | spin_lock(&tun->lock); |
96442e42 JW |
450 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { |
451 | struct tun_flow_entry *e; | |
b67bfe0d | 452 | struct hlist_node *n; |
96442e42 | 453 | |
b67bfe0d | 454 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 | 455 | unsigned long this_timer; |
81d98fa4 | 456 | |
96442e42 | 457 | this_timer = e->updated + delay; |
81d98fa4 | 458 | if (time_before_eq(this_timer, jiffies)) { |
96442e42 | 459 | tun_flow_delete(tun, e); |
81d98fa4 ED |
460 | continue; |
461 | } | |
462 | count++; | |
463 | if (time_before(this_timer, next_timer)) | |
96442e42 JW |
464 | next_timer = this_timer; |
465 | } | |
466 | } | |
467 | ||
468 | if (count) | |
469 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
7dbfb4ef | 470 | spin_unlock(&tun->lock); |
96442e42 JW |
471 | } |
472 | ||
49974420 | 473 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 474 | struct tun_file *tfile) |
96442e42 JW |
475 | { |
476 | struct hlist_head *head; | |
477 | struct tun_flow_entry *e; | |
478 | unsigned long delay = tun->ageing_time; | |
9e85722d | 479 | u16 queue_index = tfile->queue_index; |
96442e42 | 480 | |
5c327f67 | 481 | head = &tun->flows[tun_hashfn(rxhash)]; |
96442e42 JW |
482 | |
483 | rcu_read_lock(); | |
484 | ||
96442e42 JW |
485 | e = tun_flow_find(head, rxhash); |
486 | if (likely(e)) { | |
487 | /* TODO: keep queueing to old queue until it's empty? */ | |
4ffdd22e ED |
488 | if (READ_ONCE(e->queue_index) != queue_index) |
489 | WRITE_ONCE(e->queue_index, queue_index); | |
83b1bc12 LR |
490 | if (e->updated != jiffies) |
491 | e->updated = jiffies; | |
9bc88939 | 492 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
493 | } else { |
494 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
495 | if (!tun_flow_find(head, rxhash) && |
496 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
497 | tun_flow_create(tun, head, rxhash, queue_index); |
498 | ||
499 | if (!timer_pending(&tun->flow_gc_timer)) | |
500 | mod_timer(&tun->flow_gc_timer, | |
501 | round_jiffies_up(jiffies + delay)); | |
502 | spin_unlock_bh(&tun->lock); | |
503 | } | |
504 | ||
96442e42 JW |
505 | rcu_read_unlock(); |
506 | } | |
507 | ||
516c512b | 508 | /* Save the hash received in the stack receive path and update the |
9bc88939 TH |
509 | * flow_hash table accordingly. |
510 | */ | |
511 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
512 | { | |
567e4b79 | 513 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 514 | e->rps_rxhash = hash; |
9bc88939 TH |
515 | } |
516 | ||
4b035271 | 517 | /* We try to identify a flow through its rxhash. The reason that |
92d4ea6e | 518 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
519 | * the rxq based on the txq where the last packet of the flow comes. As |
520 | * the userspace application move between processors, we may get a | |
4b035271 | 521 | * different rxq no. here. |
c8d68e6b | 522 | */ |
96f84061 | 523 | static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
c8d68e6b | 524 | { |
96442e42 | 525 | struct tun_flow_entry *e; |
c8d68e6b JW |
526 | u32 txq = 0; |
527 | u32 numqueues = 0; | |
528 | ||
6aa7de05 | 529 | numqueues = READ_ONCE(tun->numqueues); |
c8d68e6b | 530 | |
feec084a | 531 | txq = __skb_get_hash_symmetric(skb); |
4b035271 WL |
532 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
533 | if (e) { | |
534 | tun_flow_save_rps_rxhash(e, txq); | |
535 | txq = e->queue_index; | |
536 | } else { | |
537 | /* use multiply and shift instead of expensive divide */ | |
538 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
539 | } |
540 | ||
c8d68e6b JW |
541 | return txq; |
542 | } | |
543 | ||
96f84061 JW |
544 | static u16 tun_ebpf_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
545 | { | |
cd5681d7 | 546 | struct tun_prog *prog; |
a35d310f | 547 | u32 numqueues; |
96f84061 JW |
548 | u16 ret = 0; |
549 | ||
a35d310f JW |
550 | numqueues = READ_ONCE(tun->numqueues); |
551 | if (!numqueues) | |
552 | return 0; | |
553 | ||
96f84061 JW |
554 | prog = rcu_dereference(tun->steering_prog); |
555 | if (prog) | |
556 | ret = bpf_prog_run_clear_cb(prog->prog, skb); | |
557 | ||
a35d310f | 558 | return ret % numqueues; |
96f84061 JW |
559 | } |
560 | ||
561 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, | |
a350ecce | 562 | struct net_device *sb_dev) |
96f84061 JW |
563 | { |
564 | struct tun_struct *tun = netdev_priv(dev); | |
565 | u16 ret; | |
566 | ||
567 | rcu_read_lock(); | |
568 | if (rcu_dereference(tun->steering_prog)) | |
569 | ret = tun_ebpf_select_queue(tun, skb); | |
570 | else | |
571 | ret = tun_automq_select_queue(tun, skb); | |
572 | rcu_read_unlock(); | |
573 | ||
574 | return ret; | |
575 | } | |
576 | ||
cde8b15f JW |
577 | static inline bool tun_not_capable(struct tun_struct *tun) |
578 | { | |
579 | const struct cred *cred = current_cred(); | |
c260b772 | 580 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
581 | |
582 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
583 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 584 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
585 | } |
586 | ||
c8d68e6b JW |
587 | static void tun_set_real_num_queues(struct tun_struct *tun) |
588 | { | |
589 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
590 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
591 | } | |
592 | ||
4008e97f JW |
593 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
594 | { | |
595 | tfile->detached = tun; | |
596 | list_add_tail(&tfile->next, &tun->disabled); | |
597 | ++tun->numdisabled; | |
598 | } | |
599 | ||
d32649d1 | 600 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
601 | { |
602 | struct tun_struct *tun = tfile->detached; | |
603 | ||
604 | tfile->detached = NULL; | |
605 | list_del_init(&tfile->next); | |
606 | --tun->numdisabled; | |
607 | return tun; | |
608 | } | |
609 | ||
3a403076 | 610 | void tun_ptr_free(void *ptr) |
fc72d1d5 JW |
611 | { |
612 | if (!ptr) | |
613 | return; | |
1ffcbc85 JDB |
614 | if (tun_is_xdp_frame(ptr)) { |
615 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 616 | |
03993094 | 617 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
618 | } else { |
619 | __skb_array_destroy_skb(ptr); | |
620 | } | |
621 | } | |
3a403076 | 622 | EXPORT_SYMBOL_GPL(tun_ptr_free); |
fc72d1d5 | 623 | |
4bfb0513 JW |
624 | static void tun_queue_purge(struct tun_file *tfile) |
625 | { | |
fc72d1d5 | 626 | void *ptr; |
1576d986 | 627 | |
fc72d1d5 JW |
628 | while ((ptr = ptr_ring_consume(&tfile->tx_ring)) != NULL) |
629 | tun_ptr_free(ptr); | |
1576d986 | 630 | |
5503fcec | 631 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
632 | skb_queue_purge(&tfile->sk.sk_error_queue); |
633 | } | |
634 | ||
c8d68e6b JW |
635 | static void __tun_detach(struct tun_file *tfile, bool clean) |
636 | { | |
637 | struct tun_file *ntfile; | |
638 | struct tun_struct *tun; | |
c8d68e6b | 639 | |
b8deabd3 JW |
640 | tun = rtnl_dereference(tfile->tun); |
641 | ||
94317099 | 642 | if (tun && clean) { |
ff1fa208 JK |
643 | if (!tfile->detached) |
644 | tun_napi_disable(tfile); | |
06e55add | 645 | tun_napi_del(tfile); |
94317099 PP |
646 | } |
647 | ||
9e85722d | 648 | if (tun && !tfile->detached) { |
c8d68e6b JW |
649 | u16 index = tfile->queue_index; |
650 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
651 | |
652 | rcu_assign_pointer(tun->tfiles[index], | |
653 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 654 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b | 655 | ntfile->queue_index = index; |
2a770cdc | 656 | ntfile->xdp_rxq.queue_index = index; |
9871a9e4 JW |
657 | rcu_assign_pointer(tun->tfiles[tun->numqueues - 1], |
658 | NULL); | |
c8d68e6b JW |
659 | |
660 | --tun->numqueues; | |
9e85722d | 661 | if (clean) { |
c956674b | 662 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 663 | sock_put(&tfile->sk); |
a8fc8cb5 | 664 | } else { |
4008e97f | 665 | tun_disable_queue(tun, tfile); |
a8fc8cb5 JK |
666 | tun_napi_disable(tfile); |
667 | } | |
c8d68e6b JW |
668 | |
669 | synchronize_net(); | |
96442e42 | 670 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 671 | /* Drop read queue */ |
4bfb0513 | 672 | tun_queue_purge(tfile); |
c8d68e6b | 673 | tun_set_real_num_queues(tun); |
dd38bd85 | 674 | } else if (tfile->detached && clean) { |
4008e97f | 675 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
676 | sock_put(&tfile->sk); |
677 | } | |
c8d68e6b JW |
678 | |
679 | if (clean) { | |
af668b3c MT |
680 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
681 | netif_carrier_off(tun->dev); | |
682 | ||
40630b82 | 683 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 684 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 685 | unregister_netdevice(tun->dev); |
af668b3c | 686 | } |
b196d88a JW |
687 | if (tun) |
688 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
7063efd3 | 689 | ptr_ring_cleanup(&tfile->tx_ring, tun_ptr_free); |
c8d68e6b JW |
690 | } |
691 | } | |
692 | ||
693 | static void tun_detach(struct tun_file *tfile, bool clean) | |
694 | { | |
83c1f36f SD |
695 | struct tun_struct *tun; |
696 | struct net_device *dev; | |
697 | ||
c8d68e6b | 698 | rtnl_lock(); |
83c1f36f SD |
699 | tun = rtnl_dereference(tfile->tun); |
700 | dev = tun ? tun->dev : NULL; | |
c8d68e6b | 701 | __tun_detach(tfile, clean); |
83c1f36f SD |
702 | if (dev) |
703 | netdev_state_change(dev); | |
c8d68e6b | 704 | rtnl_unlock(); |
5daadc86 SY |
705 | |
706 | if (clean) | |
707 | sock_put(&tfile->sk); | |
c8d68e6b JW |
708 | } |
709 | ||
710 | static void tun_detach_all(struct net_device *dev) | |
711 | { | |
712 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 713 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
714 | int i, n = tun->numqueues; |
715 | ||
716 | for (i = 0; i < n; i++) { | |
b8deabd3 | 717 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 718 | BUG_ON(!tfile); |
06e55add | 719 | tun_napi_disable(tfile); |
addf8fc4 | 720 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 721 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 722 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
723 | --tun->numqueues; |
724 | } | |
9e85722d | 725 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 726 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 727 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 728 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 729 | } |
c8d68e6b JW |
730 | BUG_ON(tun->numqueues != 0); |
731 | ||
732 | synchronize_net(); | |
733 | for (i = 0; i < n; i++) { | |
b8deabd3 | 734 | tfile = rtnl_dereference(tun->tfiles[i]); |
06e55add | 735 | tun_napi_del(tfile); |
c8d68e6b | 736 | /* Drop read queue */ |
4bfb0513 | 737 | tun_queue_purge(tfile); |
b196d88a | 738 | xdp_rxq_info_unreg(&tfile->xdp_rxq); |
c8d68e6b JW |
739 | sock_put(&tfile->sk); |
740 | } | |
4008e97f | 741 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
3b9bc84d | 742 | tun_napi_del(tfile); |
4008e97f | 743 | tun_enable_queue(tfile); |
4bfb0513 | 744 | tun_queue_purge(tfile); |
b196d88a | 745 | xdp_rxq_info_unreg(&tfile->xdp_rxq); |
4008e97f JW |
746 | sock_put(&tfile->sk); |
747 | } | |
748 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 749 | |
40630b82 | 750 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 751 | module_put(THIS_MODULE); |
c8d68e6b JW |
752 | } |
753 | ||
94317099 | 754 | static int tun_attach(struct tun_struct *tun, struct file *file, |
77f22f92 YY |
755 | bool skip_filter, bool napi, bool napi_frags, |
756 | bool publish_tun) | |
a7385ba2 | 757 | { |
631ab46b | 758 | struct tun_file *tfile = file->private_data; |
1576d986 | 759 | struct net_device *dev = tun->dev; |
38231b7a | 760 | int err; |
a7385ba2 | 761 | |
5dbbaf2d PM |
762 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
763 | if (err < 0) | |
764 | goto out; | |
765 | ||
38231b7a | 766 | err = -EINVAL; |
9e85722d | 767 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
768 | goto out; |
769 | ||
770 | err = -EBUSY; | |
40630b82 | 771 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
772 | goto out; |
773 | ||
774 | err = -E2BIG; | |
4008e97f JW |
775 | if (!tfile->detached && |
776 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
777 | goto out; |
778 | ||
779 | err = 0; | |
54f968d6 | 780 | |
92d4ea6e | 781 | /* Re-attach the filter to persist device */ |
849c9b6f | 782 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
783 | lock_sock(tfile->socket.sk); |
784 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
785 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
786 | if (!err) |
787 | goto out; | |
788 | } | |
1576d986 JW |
789 | |
790 | if (!tfile->detached && | |
b196d88a JW |
791 | ptr_ring_resize(&tfile->tx_ring, dev->tx_queue_len, |
792 | GFP_KERNEL, tun_ptr_free)) { | |
1576d986 JW |
793 | err = -ENOMEM; |
794 | goto out; | |
795 | } | |
796 | ||
c8d68e6b | 797 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 798 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
8bf5c4ee JDB |
799 | |
800 | if (tfile->detached) { | |
801 | /* Re-attach detached tfile, updating XDP queue_index */ | |
802 | WARN_ON(!xdp_rxq_info_is_reg(&tfile->xdp_rxq)); | |
803 | ||
804 | if (tfile->xdp_rxq.queue_index != tfile->queue_index) | |
805 | tfile->xdp_rxq.queue_index = tfile->queue_index; | |
806 | } else { | |
807 | /* Setup XDP RX-queue info, for new tfile getting attached */ | |
808 | err = xdp_rxq_info_reg(&tfile->xdp_rxq, | |
b02e5a0e | 809 | tun->dev, tfile->queue_index, 0); |
8bf5c4ee JDB |
810 | if (err < 0) |
811 | goto out; | |
8d5d8852 JDB |
812 | err = xdp_rxq_info_reg_mem_model(&tfile->xdp_rxq, |
813 | MEM_TYPE_PAGE_SHARED, NULL); | |
814 | if (err < 0) { | |
815 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
816 | goto out; | |
817 | } | |
8bf5c4ee JDB |
818 | err = 0; |
819 | } | |
820 | ||
94317099 | 821 | if (tfile->detached) { |
4008e97f | 822 | tun_enable_queue(tfile); |
a8fc8cb5 | 823 | tun_napi_enable(tfile); |
94317099 | 824 | } else { |
4008e97f | 825 | sock_hold(&tfile->sk); |
af3fb24e | 826 | tun_napi_init(tun, tfile, napi, napi_frags); |
94317099 | 827 | } |
4008e97f | 828 | |
e4a2a304 JW |
829 | if (rtnl_dereference(tun->xdp_prog)) |
830 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
831 | ||
c8d68e6b JW |
832 | /* device is allowed to go away first, so no need to hold extra |
833 | * refcnt. | |
834 | */ | |
835 | ||
0b7959b6 SF |
836 | /* Publish tfile->tun and tun->tfiles only after we've fully |
837 | * initialized tfile; otherwise we risk using half-initialized | |
838 | * object. | |
839 | */ | |
77f22f92 YY |
840 | if (publish_tun) |
841 | rcu_assign_pointer(tfile->tun, tun); | |
0b7959b6 SF |
842 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
843 | tun->numqueues++; | |
3a03cb84 | 844 | tun_set_real_num_queues(tun); |
c8d68e6b JW |
845 | out: |
846 | return err; | |
631ab46b EB |
847 | } |
848 | ||
9484dc74 | 849 | static struct tun_struct *tun_get(struct tun_file *tfile) |
631ab46b | 850 | { |
6e914fc7 | 851 | struct tun_struct *tun; |
c70f1829 | 852 | |
6e914fc7 JW |
853 | rcu_read_lock(); |
854 | tun = rcu_dereference(tfile->tun); | |
855 | if (tun) | |
856 | dev_hold(tun->dev); | |
857 | rcu_read_unlock(); | |
c70f1829 EB |
858 | |
859 | return tun; | |
631ab46b EB |
860 | } |
861 | ||
631ab46b EB |
862 | static void tun_put(struct tun_struct *tun) |
863 | { | |
6e914fc7 | 864 | dev_put(tun->dev); |
631ab46b EB |
865 | } |
866 | ||
6b8a66ee | 867 | /* TAP filtering */ |
f271b2cc MK |
868 | static void addr_hash_set(u32 *mask, const u8 *addr) |
869 | { | |
870 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
871 | mask[n >> 5] |= (1 << (n & 31)); | |
872 | } | |
873 | ||
874 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
875 | { | |
876 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
877 | return mask[n >> 5] & (1 << (n & 31)); | |
878 | } | |
879 | ||
880 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
881 | { | |
882 | struct { u8 u[ETH_ALEN]; } *addr; | |
883 | struct tun_filter uf; | |
884 | int err, alen, n, nexact; | |
885 | ||
886 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
887 | return -EFAULT; | |
888 | ||
889 | if (!uf.count) { | |
890 | /* Disabled */ | |
891 | filter->count = 0; | |
892 | return 0; | |
893 | } | |
894 | ||
895 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
896 | addr = memdup_user(arg + sizeof(uf), alen); |
897 | if (IS_ERR(addr)) | |
898 | return PTR_ERR(addr); | |
f271b2cc MK |
899 | |
900 | /* The filter is updated without holding any locks. Which is | |
901 | * perfectly safe. We disable it first and in the worst | |
902 | * case we'll accept a few undesired packets. */ | |
903 | filter->count = 0; | |
904 | wmb(); | |
905 | ||
906 | /* Use first set of addresses as an exact filter */ | |
907 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
908 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
909 | ||
910 | nexact = n; | |
911 | ||
cfbf84fc AW |
912 | /* Remaining multicast addresses are hashed, |
913 | * unicast will leave the filter disabled. */ | |
f271b2cc | 914 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
915 | for (; n < uf.count; n++) { |
916 | if (!is_multicast_ether_addr(addr[n].u)) { | |
917 | err = 0; /* no filter */ | |
3b8d2a69 | 918 | goto free_addr; |
cfbf84fc | 919 | } |
f271b2cc | 920 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 921 | } |
f271b2cc MK |
922 | |
923 | /* For ALLMULTI just set the mask to all ones. | |
924 | * This overrides the mask populated above. */ | |
925 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
926 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
927 | ||
928 | /* Now enable the filter */ | |
929 | wmb(); | |
930 | filter->count = nexact; | |
931 | ||
932 | /* Return the number of exact filters */ | |
933 | err = nexact; | |
3b8d2a69 | 934 | free_addr: |
f271b2cc MK |
935 | kfree(addr); |
936 | return err; | |
937 | } | |
938 | ||
939 | /* Returns: 0 - drop, !=0 - accept */ | |
940 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
941 | { | |
942 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
943 | * at this point. */ | |
944 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
945 | int i; | |
946 | ||
947 | /* Exact match */ | |
948 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 949 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
950 | return 1; |
951 | ||
952 | /* Inexact match (multicast only) */ | |
953 | if (is_multicast_ether_addr(eh->h_dest)) | |
954 | return addr_hash_test(filter->mask, eh->h_dest); | |
955 | ||
956 | return 0; | |
957 | } | |
958 | ||
959 | /* | |
960 | * Checks whether the packet is accepted or not. | |
961 | * Returns: 0 - drop, !=0 - accept | |
962 | */ | |
963 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
964 | { | |
965 | if (!filter->count) | |
966 | return 1; | |
967 | ||
968 | return run_filter(filter, skb); | |
969 | } | |
970 | ||
1da177e4 LT |
971 | /* Network device part of the driver */ |
972 | ||
7282d491 | 973 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 974 | |
158b515f GK |
975 | static int tun_net_init(struct net_device *dev) |
976 | { | |
977 | struct tun_struct *tun = netdev_priv(dev); | |
978 | struct ifreq *ifr = tun->ifr; | |
979 | int err; | |
980 | ||
981 | dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); | |
982 | if (!dev->tstats) | |
983 | return -ENOMEM; | |
984 | ||
985 | spin_lock_init(&tun->lock); | |
986 | ||
987 | err = security_tun_dev_alloc_security(&tun->security); | |
988 | if (err < 0) { | |
989 | free_percpu(dev->tstats); | |
990 | return err; | |
991 | } | |
992 | ||
993 | tun_flow_init(tun); | |
994 | ||
995 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | | |
996 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | | |
997 | NETIF_F_HW_VLAN_STAG_TX; | |
998 | dev->features = dev->hw_features | NETIF_F_LLTX; | |
999 | dev->vlan_features = dev->features & | |
1000 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
1001 | NETIF_F_HW_VLAN_STAG_TX); | |
1002 | ||
1003 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
1004 | (ifr->ifr_flags & TUN_FEATURES); | |
1005 | ||
1006 | INIT_LIST_HEAD(&tun->disabled); | |
1007 | err = tun_attach(tun, tun->file, false, ifr->ifr_flags & IFF_NAPI, | |
1008 | ifr->ifr_flags & IFF_NAPI_FRAGS, false); | |
1009 | if (err < 0) { | |
1010 | tun_flow_uninit(tun); | |
1011 | security_tun_dev_free_security(tun->security); | |
1012 | free_percpu(dev->tstats); | |
1013 | return err; | |
1014 | } | |
1015 | return 0; | |
1016 | } | |
1017 | ||
c70f1829 EB |
1018 | /* Net device detach from fd. */ |
1019 | static void tun_net_uninit(struct net_device *dev) | |
1020 | { | |
c8d68e6b | 1021 | tun_detach_all(dev); |
c70f1829 EB |
1022 | } |
1023 | ||
1da177e4 LT |
1024 | /* Net device open. */ |
1025 | static int tun_net_open(struct net_device *dev) | |
1026 | { | |
c8d68e6b | 1027 | netif_tx_start_all_queues(dev); |
b20e2d54 | 1028 | |
1da177e4 LT |
1029 | return 0; |
1030 | } | |
1031 | ||
1032 | /* Net device close. */ | |
1033 | static int tun_net_close(struct net_device *dev) | |
1034 | { | |
c8d68e6b | 1035 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
1036 | return 0; |
1037 | } | |
1038 | ||
1039 | /* Net device start xmit */ | |
96f84061 | 1040 | static void tun_automq_xmit(struct tun_struct *tun, struct sk_buff *skb) |
1da177e4 | 1041 | { |
3df97ba8 | 1042 | #ifdef CONFIG_RPS |
dc05360f | 1043 | if (tun->numqueues == 1 && static_branch_unlikely(&rps_needed)) { |
9bc88939 TH |
1044 | /* Select queue was not called for the skbuff, so we extract the |
1045 | * RPS hash and save it into the flow_table here. | |
1046 | */ | |
4b035271 | 1047 | struct tun_flow_entry *e; |
9bc88939 TH |
1048 | __u32 rxhash; |
1049 | ||
feec084a | 1050 | rxhash = __skb_get_hash_symmetric(skb); |
4b035271 WL |
1051 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], rxhash); |
1052 | if (e) | |
1053 | tun_flow_save_rps_rxhash(e, rxhash); | |
9bc88939 | 1054 | } |
3df97ba8 | 1055 | #endif |
96f84061 JW |
1056 | } |
1057 | ||
aff3d70a JW |
1058 | static unsigned int run_ebpf_filter(struct tun_struct *tun, |
1059 | struct sk_buff *skb, | |
1060 | int len) | |
1061 | { | |
1062 | struct tun_prog *prog = rcu_dereference(tun->filter_prog); | |
1063 | ||
1064 | if (prog) | |
1065 | len = bpf_prog_run_clear_cb(prog->prog, skb); | |
1066 | ||
1067 | return len; | |
1068 | } | |
1069 | ||
96f84061 JW |
1070 | /* Net device start xmit */ |
1071 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) | |
1072 | { | |
1073 | struct tun_struct *tun = netdev_priv(dev); | |
4b4f052e | 1074 | enum skb_drop_reason drop_reason; |
96f84061 | 1075 | int txq = skb->queue_mapping; |
a31d27fb | 1076 | struct netdev_queue *queue; |
96f84061 | 1077 | struct tun_file *tfile; |
aff3d70a | 1078 | int len = skb->len; |
96f84061 JW |
1079 | |
1080 | rcu_read_lock(); | |
1081 | tfile = rcu_dereference(tun->tfiles[txq]); | |
96f84061 JW |
1082 | |
1083 | /* Drop packet if interface is not attached */ | |
4b4f052e DZ |
1084 | if (!tfile) { |
1085 | drop_reason = SKB_DROP_REASON_DEV_READY; | |
96f84061 | 1086 | goto drop; |
4b4f052e | 1087 | } |
96f84061 JW |
1088 | |
1089 | if (!rcu_dereference(tun->steering_prog)) | |
1090 | tun_automq_xmit(tun, skb); | |
9bc88939 | 1091 | |
3424170f | 1092 | netif_info(tun, tx_queued, tun->dev, "%s %d\n", __func__, skb->len); |
6e914fc7 | 1093 | |
f271b2cc MK |
1094 | /* Drop if the filter does not like it. |
1095 | * This is a noop if the filter is disabled. | |
1096 | * Filter can be enabled only for the TAP devices. */ | |
4b4f052e DZ |
1097 | if (!check_filter(&tun->txflt, skb)) { |
1098 | drop_reason = SKB_DROP_REASON_TAP_TXFILTER; | |
f271b2cc | 1099 | goto drop; |
4b4f052e | 1100 | } |
f271b2cc | 1101 | |
54f968d6 | 1102 | if (tfile->socket.sk->sk_filter && |
4b4f052e DZ |
1103 | sk_filter(tfile->socket.sk, skb)) { |
1104 | drop_reason = SKB_DROP_REASON_SOCKET_FILTER; | |
99405162 | 1105 | goto drop; |
4b4f052e | 1106 | } |
99405162 | 1107 | |
aff3d70a | 1108 | len = run_ebpf_filter(tun, skb, len); |
4b4f052e DZ |
1109 | if (len == 0) { |
1110 | drop_reason = SKB_DROP_REASON_TAP_FILTER; | |
45a15d89 | 1111 | goto drop; |
4b4f052e | 1112 | } |
45a15d89 | 1113 | |
4b4f052e DZ |
1114 | if (pskb_trim(skb, len)) { |
1115 | drop_reason = SKB_DROP_REASON_NOMEM; | |
aff3d70a | 1116 | goto drop; |
4b4f052e | 1117 | } |
aff3d70a | 1118 | |
4b4f052e DZ |
1119 | if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) { |
1120 | drop_reason = SKB_DROP_REASON_SKB_UCOPY_FAULT; | |
7bf66305 | 1121 | goto drop; |
4b4f052e | 1122 | } |
7bf66305 | 1123 | |
7b996243 | 1124 | skb_tx_timestamp(skb); |
eda29772 | 1125 | |
0110d6f2 | 1126 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
1127 | * for indefinite time. |
1128 | */ | |
0110d6f2 MT |
1129 | skb_orphan(skb); |
1130 | ||
895b5c9f | 1131 | nf_reset_ct(skb); |
f8af75f3 | 1132 | |
4b4f052e DZ |
1133 | if (ptr_ring_produce(&tfile->tx_ring, skb)) { |
1134 | drop_reason = SKB_DROP_REASON_FULL_RING; | |
1576d986 | 1135 | goto drop; |
4b4f052e | 1136 | } |
1da177e4 | 1137 | |
a31d27fb ND |
1138 | /* NETIF_F_LLTX requires to do our own update of trans_start */ |
1139 | queue = netdev_get_tx_queue(dev, txq); | |
968a1a5d | 1140 | txq_trans_cond_update(queue); |
a31d27fb | 1141 | |
1da177e4 | 1142 | /* Notify and wake up reader process */ |
54f968d6 JW |
1143 | if (tfile->flags & TUN_FASYNC) |
1144 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 1145 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
1146 | |
1147 | rcu_read_unlock(); | |
6ed10654 | 1148 | return NETDEV_TX_OK; |
1da177e4 LT |
1149 | |
1150 | drop: | |
625788b5 | 1151 | dev_core_stats_tx_dropped_inc(dev); |
149d36f7 | 1152 | skb_tx_error(skb); |
4b4f052e | 1153 | kfree_skb_reason(skb, drop_reason); |
6e914fc7 | 1154 | rcu_read_unlock(); |
baeababb | 1155 | return NET_XMIT_DROP; |
1da177e4 LT |
1156 | } |
1157 | ||
f271b2cc | 1158 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 1159 | { |
f271b2cc MK |
1160 | /* |
1161 | * This callback is supposed to deal with mc filter in | |
1162 | * _rx_ path and has nothing to do with the _tx_ path. | |
1163 | * In rx path we always accept everything userspace gives us. | |
1164 | */ | |
1da177e4 LT |
1165 | } |
1166 | ||
c8f44aff MM |
1167 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
1168 | netdev_features_t features) | |
88255375 MM |
1169 | { |
1170 | struct tun_struct *tun = netdev_priv(dev); | |
1171 | ||
1172 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
1173 | } | |
eaea34b2 PA |
1174 | |
1175 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
1176 | { | |
1177 | struct tun_struct *tun = netdev_priv(dev); | |
1178 | ||
1179 | if (new_hr < NET_SKB_PAD) | |
1180 | new_hr = NET_SKB_PAD; | |
1181 | ||
1182 | tun->align = new_hr; | |
1183 | } | |
1184 | ||
bc1f4470 | 1185 | static void |
608b9977 PA |
1186 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
1187 | { | |
608b9977 | 1188 | struct tun_struct *tun = netdev_priv(dev); |
608b9977 | 1189 | |
497a5757 | 1190 | dev_get_tstats64(dev, stats); |
608b9977 | 1191 | |
497a5757 HK |
1192 | stats->rx_frame_errors += |
1193 | (unsigned long)atomic_long_read(&tun->rx_frame_errors); | |
608b9977 PA |
1194 | } |
1195 | ||
761876c8 JW |
1196 | static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog, |
1197 | struct netlink_ext_ack *extack) | |
1198 | { | |
1199 | struct tun_struct *tun = netdev_priv(dev); | |
e4a2a304 | 1200 | struct tun_file *tfile; |
761876c8 | 1201 | struct bpf_prog *old_prog; |
e4a2a304 | 1202 | int i; |
761876c8 JW |
1203 | |
1204 | old_prog = rtnl_dereference(tun->xdp_prog); | |
1205 | rcu_assign_pointer(tun->xdp_prog, prog); | |
1206 | if (old_prog) | |
1207 | bpf_prog_put(old_prog); | |
1208 | ||
e4a2a304 JW |
1209 | for (i = 0; i < tun->numqueues; i++) { |
1210 | tfile = rtnl_dereference(tun->tfiles[i]); | |
1211 | if (prog) | |
1212 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
1213 | else | |
1214 | sock_reset_flag(&tfile->sk, SOCK_XDP); | |
1215 | } | |
1216 | list_for_each_entry(tfile, &tun->disabled, next) { | |
1217 | if (prog) | |
1218 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
1219 | else | |
1220 | sock_reset_flag(&tfile->sk, SOCK_XDP); | |
1221 | } | |
1222 | ||
761876c8 JW |
1223 | return 0; |
1224 | } | |
1225 | ||
f4e63525 | 1226 | static int tun_xdp(struct net_device *dev, struct netdev_bpf *xdp) |
761876c8 JW |
1227 | { |
1228 | switch (xdp->command) { | |
1229 | case XDP_SETUP_PROG: | |
1230 | return tun_xdp_set(dev, xdp->prog, xdp->extack); | |
761876c8 JW |
1231 | default: |
1232 | return -EINVAL; | |
1233 | } | |
1234 | } | |
1235 | ||
26d31925 ND |
1236 | static int tun_net_change_carrier(struct net_device *dev, bool new_carrier) |
1237 | { | |
1238 | if (new_carrier) { | |
1239 | struct tun_struct *tun = netdev_priv(dev); | |
1240 | ||
1241 | if (!tun->numqueues) | |
1242 | return -EPERM; | |
1243 | ||
1244 | netif_carrier_on(dev); | |
1245 | } else { | |
1246 | netif_carrier_off(dev); | |
1247 | } | |
1248 | return 0; | |
1249 | } | |
1250 | ||
758e43b7 | 1251 | static const struct net_device_ops tun_netdev_ops = { |
158b515f | 1252 | .ndo_init = tun_net_init, |
c70f1829 | 1253 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1254 | .ndo_open = tun_net_open, |
1255 | .ndo_stop = tun_net_close, | |
00829823 | 1256 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1257 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1258 | .ndo_select_queue = tun_select_queue, |
eaea34b2 | 1259 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1260 | .ndo_get_stats64 = tun_net_get_stats64, |
26d31925 | 1261 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1262 | }; |
1263 | ||
0c9d917b JDB |
1264 | static void __tun_xdp_flush_tfile(struct tun_file *tfile) |
1265 | { | |
1266 | /* Notify and wake up reader process */ | |
1267 | if (tfile->flags & TUN_FASYNC) | |
1268 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
1269 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); | |
1270 | } | |
1271 | ||
42b33468 JDB |
1272 | static int tun_xdp_xmit(struct net_device *dev, int n, |
1273 | struct xdp_frame **frames, u32 flags) | |
fc72d1d5 JW |
1274 | { |
1275 | struct tun_struct *tun = netdev_priv(dev); | |
fc72d1d5 JW |
1276 | struct tun_file *tfile; |
1277 | u32 numqueues; | |
fdc13979 | 1278 | int nxmit = 0; |
735fc405 | 1279 | int i; |
fc72d1d5 | 1280 | |
0c9d917b | 1281 | if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) |
42b33468 JDB |
1282 | return -EINVAL; |
1283 | ||
fc72d1d5 JW |
1284 | rcu_read_lock(); |
1285 | ||
9871a9e4 | 1286 | resample: |
fc72d1d5 JW |
1287 | numqueues = READ_ONCE(tun->numqueues); |
1288 | if (!numqueues) { | |
735fc405 JDB |
1289 | rcu_read_unlock(); |
1290 | return -ENXIO; /* Caller will free/return all frames */ | |
fc72d1d5 JW |
1291 | } |
1292 | ||
1293 | tfile = rcu_dereference(tun->tfiles[smp_processor_id() % | |
1294 | numqueues]); | |
9871a9e4 JW |
1295 | if (unlikely(!tfile)) |
1296 | goto resample; | |
735fc405 JDB |
1297 | |
1298 | spin_lock(&tfile->tx_ring.producer_lock); | |
1299 | for (i = 0; i < n; i++) { | |
1300 | struct xdp_frame *xdp = frames[i]; | |
1301 | /* Encode the XDP flag into lowest bit for consumer to differ | |
1302 | * XDP buffer from sk_buff. | |
1303 | */ | |
1304 | void *frame = tun_xdp_to_ptr(xdp); | |
1305 | ||
1306 | if (__ptr_ring_produce(&tfile->tx_ring, frame)) { | |
625788b5 | 1307 | dev_core_stats_tx_dropped_inc(dev); |
fdc13979 | 1308 | break; |
735fc405 | 1309 | } |
fdc13979 | 1310 | nxmit++; |
fc72d1d5 | 1311 | } |
735fc405 | 1312 | spin_unlock(&tfile->tx_ring.producer_lock); |
fc72d1d5 | 1313 | |
0c9d917b JDB |
1314 | if (flags & XDP_XMIT_FLUSH) |
1315 | __tun_xdp_flush_tfile(tfile); | |
1316 | ||
fc72d1d5 | 1317 | rcu_read_unlock(); |
fdc13979 | 1318 | return nxmit; |
fc72d1d5 JW |
1319 | } |
1320 | ||
44fa2dbd JDB |
1321 | static int tun_xdp_tx(struct net_device *dev, struct xdp_buff *xdp) |
1322 | { | |
1b698fa5 | 1323 | struct xdp_frame *frame = xdp_convert_buff_to_frame(xdp); |
fdc13979 | 1324 | int nxmit; |
44fa2dbd JDB |
1325 | |
1326 | if (unlikely(!frame)) | |
1327 | return -EOVERFLOW; | |
1328 | ||
fdc13979 LB |
1329 | nxmit = tun_xdp_xmit(dev, 1, &frame, XDP_XMIT_FLUSH); |
1330 | if (!nxmit) | |
1331 | xdp_return_frame_rx_napi(frame); | |
1332 | return nxmit; | |
fc72d1d5 JW |
1333 | } |
1334 | ||
758e43b7 | 1335 | static const struct net_device_ops tap_netdev_ops = { |
158b515f | 1336 | .ndo_init = tun_net_init, |
c70f1829 | 1337 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1338 | .ndo_open = tun_net_open, |
1339 | .ndo_stop = tun_net_close, | |
00829823 | 1340 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1341 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1342 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1343 | .ndo_set_mac_address = eth_mac_addr, |
1344 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1345 | .ndo_select_queue = tun_select_queue, |
5e52796a | 1346 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1347 | .ndo_set_rx_headroom = tun_set_headroom, |
497a5757 | 1348 | .ndo_get_stats64 = dev_get_tstats64, |
f4e63525 | 1349 | .ndo_bpf = tun_xdp, |
fc72d1d5 | 1350 | .ndo_xdp_xmit = tun_xdp_xmit, |
26d31925 | 1351 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1352 | }; |
1353 | ||
944a1376 | 1354 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1355 | { |
1356 | int i; | |
1357 | ||
96442e42 JW |
1358 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1359 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1360 | ||
1361 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
e99e88a9 KC |
1362 | timer_setup(&tun->flow_gc_timer, tun_flow_cleanup, 0); |
1363 | mod_timer(&tun->flow_gc_timer, | |
1364 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1365 | } |
1366 | ||
1367 | static void tun_flow_uninit(struct tun_struct *tun) | |
1368 | { | |
1369 | del_timer_sync(&tun->flow_gc_timer); | |
1370 | tun_flow_flush(tun); | |
96442e42 JW |
1371 | } |
1372 | ||
91572088 JW |
1373 | #define MIN_MTU 68 |
1374 | #define MAX_MTU 65535 | |
1375 | ||
1da177e4 | 1376 | /* Initialize net device. */ |
158b515f | 1377 | static void tun_net_initialize(struct net_device *dev) |
1da177e4 LT |
1378 | { |
1379 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1380 | |
1da177e4 | 1381 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1382 | case IFF_TUN: |
758e43b7 | 1383 | dev->netdev_ops = &tun_netdev_ops; |
b9815eb1 | 1384 | dev->header_ops = &ip_tunnel_header_ops; |
758e43b7 | 1385 | |
1da177e4 LT |
1386 | /* Point-to-Point TUN Device */ |
1387 | dev->hard_header_len = 0; | |
1388 | dev->addr_len = 0; | |
1389 | dev->mtu = 1500; | |
1390 | ||
1391 | /* Zero header length */ | |
6aa20a22 | 1392 | dev->type = ARPHRD_NONE; |
1da177e4 | 1393 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1394 | break; |
1395 | ||
40630b82 | 1396 | case IFF_TAP: |
7a0a9608 | 1397 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1398 | /* Ethernet TAP Device */ |
1da177e4 | 1399 | ether_setup(dev); |
550fd08c | 1400 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1401 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1402 | |
f2cedb63 | 1403 | eth_hw_addr_random(dev); |
36226a8d | 1404 | |
66c0e13a MM |
1405 | /* Currently tun does not support XDP, only tap does. */ |
1406 | dev->xdp_features = NETDEV_XDP_ACT_BASIC | | |
1407 | NETDEV_XDP_ACT_REDIRECT | | |
1408 | NETDEV_XDP_ACT_NDO_XMIT; | |
1409 | ||
1da177e4 LT |
1410 | break; |
1411 | } | |
91572088 JW |
1412 | |
1413 | dev->min_mtu = MIN_MTU; | |
1414 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1415 | } |
1416 | ||
2f3ab622 JW |
1417 | static bool tun_sock_writeable(struct tun_struct *tun, struct tun_file *tfile) |
1418 | { | |
1419 | struct sock *sk = tfile->socket.sk; | |
1420 | ||
1421 | return (tun->dev->flags & IFF_UP) && sock_writeable(sk); | |
1422 | } | |
1423 | ||
1da177e4 LT |
1424 | /* Character device part */ |
1425 | ||
1426 | /* Poll */ | |
afc9a42b | 1427 | static __poll_t tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1428 | { |
b2430de3 | 1429 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1430 | struct tun_struct *tun = tun_get(tfile); |
3c8a9c63 | 1431 | struct sock *sk; |
afc9a42b | 1432 | __poll_t mask = 0; |
1da177e4 LT |
1433 | |
1434 | if (!tun) | |
a9a08845 | 1435 | return EPOLLERR; |
1da177e4 | 1436 | |
54f968d6 | 1437 | sk = tfile->socket.sk; |
3c8a9c63 | 1438 | |
9e641bdc | 1439 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1440 | |
5990a305 | 1441 | if (!ptr_ring_empty(&tfile->tx_ring)) |
a9a08845 | 1442 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 1443 | |
2f3ab622 JW |
1444 | /* Make sure SOCKWQ_ASYNC_NOSPACE is set if not writable to |
1445 | * guarantee EPOLLOUT to be raised by either here or | |
1446 | * tun_sock_write_space(). Then process could get notification | |
1447 | * after it writes to a down device and meets -EIO. | |
1448 | */ | |
1449 | if (tun_sock_writeable(tun, tfile) || | |
1450 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
1451 | tun_sock_writeable(tun, tfile))) | |
a9a08845 | 1452 | mask |= EPOLLOUT | EPOLLWRNORM; |
33dccbb0 | 1453 | |
c70f1829 | 1454 | if (tun->dev->reg_state != NETREG_REGISTERED) |
a9a08845 | 1455 | mask = EPOLLERR; |
c70f1829 | 1456 | |
631ab46b | 1457 | tun_put(tun); |
1da177e4 LT |
1458 | return mask; |
1459 | } | |
1460 | ||
90e33d45 PP |
1461 | static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile, |
1462 | size_t len, | |
1463 | const struct iov_iter *it) | |
1464 | { | |
1465 | struct sk_buff *skb; | |
1466 | size_t linear; | |
1467 | int err; | |
1468 | int i; | |
1469 | ||
363a5328 ZX |
1470 | if (it->nr_segs > MAX_SKB_FRAGS + 1 || |
1471 | len > (ETH_MAX_MTU - NET_SKB_PAD - NET_IP_ALIGN)) | |
950271d7 | 1472 | return ERR_PTR(-EMSGSIZE); |
90e33d45 PP |
1473 | |
1474 | local_bh_disable(); | |
1475 | skb = napi_get_frags(&tfile->napi); | |
1476 | local_bh_enable(); | |
1477 | if (!skb) | |
1478 | return ERR_PTR(-ENOMEM); | |
1479 | ||
1480 | linear = iov_iter_single_seg_count(it); | |
1481 | err = __skb_grow(skb, linear); | |
1482 | if (err) | |
1483 | goto free; | |
1484 | ||
1485 | skb->len = len; | |
1486 | skb->data_len = len - linear; | |
1487 | skb->truesize += skb->data_len; | |
1488 | ||
1489 | for (i = 1; i < it->nr_segs; i++) { | |
de4f5fed JA |
1490 | const struct iovec *iov = iter_iov(it); |
1491 | size_t fragsz = iov->iov_len; | |
aa6daaca ED |
1492 | struct page *page; |
1493 | void *frag; | |
90e33d45 PP |
1494 | |
1495 | if (fragsz == 0 || fragsz > PAGE_SIZE) { | |
1496 | err = -EINVAL; | |
1497 | goto free; | |
1498 | } | |
aa6daaca ED |
1499 | frag = netdev_alloc_frag(fragsz); |
1500 | if (!frag) { | |
90e33d45 PP |
1501 | err = -ENOMEM; |
1502 | goto free; | |
1503 | } | |
aa6daaca ED |
1504 | page = virt_to_head_page(frag); |
1505 | skb_fill_page_desc(skb, i - 1, page, | |
1506 | frag - page_address(page), fragsz); | |
90e33d45 PP |
1507 | } |
1508 | ||
1509 | return skb; | |
1510 | free: | |
1511 | /* frees skb and all frags allocated with napi_alloc_frag() */ | |
1512 | napi_free_frags(&tfile->napi); | |
1513 | return ERR_PTR(err); | |
1514 | } | |
1515 | ||
f42157cb RR |
1516 | /* prepad is the amount to reserve at front. len is length after that. |
1517 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1518 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1519 | size_t prepad, size_t len, |
1520 | size_t linear, int noblock) | |
f42157cb | 1521 | { |
54f968d6 | 1522 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1523 | struct sk_buff *skb; |
33dccbb0 | 1524 | int err; |
f42157cb RR |
1525 | |
1526 | /* Under a page? Don't bother with paged skb. */ | |
6231e47b | 1527 | if (prepad + len < PAGE_SIZE) |
33dccbb0 | 1528 | linear = len; |
f42157cb | 1529 | |
ce7c7fef ED |
1530 | if (len - linear > MAX_SKB_FRAGS * (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) |
1531 | linear = len - MAX_SKB_FRAGS * (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER); | |
33dccbb0 | 1532 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
ce7c7fef | 1533 | &err, PAGE_ALLOC_COSTLY_ORDER); |
f42157cb | 1534 | if (!skb) |
33dccbb0 | 1535 | return ERR_PTR(err); |
f42157cb RR |
1536 | |
1537 | skb_reserve(skb, prepad); | |
1538 | skb_put(skb, linear); | |
33dccbb0 HX |
1539 | skb->data_len = len - linear; |
1540 | skb->len += len - linear; | |
f42157cb RR |
1541 | |
1542 | return skb; | |
1543 | } | |
1544 | ||
5503fcec JW |
1545 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1546 | struct sk_buff *skb, int more) | |
1547 | { | |
1548 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1549 | struct sk_buff_head process_queue; | |
1550 | u32 rx_batched = tun->rx_batched; | |
1551 | bool rcv = false; | |
1552 | ||
1553 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1554 | local_bh_disable(); | |
8ebebcba | 1555 | skb_record_rx_queue(skb, tfile->queue_index); |
5503fcec JW |
1556 | netif_receive_skb(skb); |
1557 | local_bh_enable(); | |
1558 | return; | |
1559 | } | |
1560 | ||
1561 | spin_lock(&queue->lock); | |
1562 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1563 | __skb_queue_head_init(&process_queue); | |
1564 | skb_queue_splice_tail_init(queue, &process_queue); | |
1565 | rcv = true; | |
1566 | } else { | |
1567 | __skb_queue_tail(queue, skb); | |
1568 | } | |
1569 | spin_unlock(&queue->lock); | |
1570 | ||
1571 | if (rcv) { | |
1572 | struct sk_buff *nskb; | |
1573 | ||
1574 | local_bh_disable(); | |
8ebebcba MC |
1575 | while ((nskb = __skb_dequeue(&process_queue))) { |
1576 | skb_record_rx_queue(nskb, tfile->queue_index); | |
5503fcec | 1577 | netif_receive_skb(nskb); |
8ebebcba MC |
1578 | } |
1579 | skb_record_rx_queue(skb, tfile->queue_index); | |
5503fcec JW |
1580 | netif_receive_skb(skb); |
1581 | local_bh_enable(); | |
1582 | } | |
1583 | } | |
1584 | ||
66ccbc9c JW |
1585 | static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile, |
1586 | int len, int noblock, bool zerocopy) | |
1587 | { | |
1588 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) | |
1589 | return false; | |
1590 | ||
1591 | if (tfile->socket.sk->sk_sndbuf != INT_MAX) | |
1592 | return false; | |
1593 | ||
1594 | if (!noblock) | |
1595 | return false; | |
1596 | ||
1597 | if (zerocopy) | |
1598 | return false; | |
1599 | ||
59eeb232 | 1600 | if (SKB_DATA_ALIGN(len + TUN_RX_PAD + XDP_PACKET_HEADROOM) + |
66ccbc9c JW |
1601 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE) |
1602 | return false; | |
1603 | ||
1604 | return true; | |
1605 | } | |
1606 | ||
4b663366 AB |
1607 | static struct sk_buff *__tun_build_skb(struct tun_file *tfile, |
1608 | struct page_frag *alloc_frag, char *buf, | |
8ae1aff0 | 1609 | int buflen, int len, int pad) |
ac1f1f6c JW |
1610 | { |
1611 | struct sk_buff *skb = build_skb(buf, buflen); | |
1612 | ||
1613 | if (!skb) | |
1614 | return ERR_PTR(-ENOMEM); | |
1615 | ||
8ae1aff0 | 1616 | skb_reserve(skb, pad); |
ac1f1f6c | 1617 | skb_put(skb, len); |
4b663366 | 1618 | skb_set_owner_w(skb, tfile->socket.sk); |
ac1f1f6c JW |
1619 | |
1620 | get_page(alloc_frag->page); | |
1621 | alloc_frag->offset += buflen; | |
1622 | ||
1623 | return skb; | |
1624 | } | |
1625 | ||
8ae1aff0 JW |
1626 | static int tun_xdp_act(struct tun_struct *tun, struct bpf_prog *xdp_prog, |
1627 | struct xdp_buff *xdp, u32 act) | |
1628 | { | |
1629 | int err; | |
1630 | ||
1631 | switch (act) { | |
1632 | case XDP_REDIRECT: | |
1633 | err = xdp_do_redirect(tun->dev, xdp, xdp_prog); | |
5744ba05 YW |
1634 | if (err) { |
1635 | dev_core_stats_rx_dropped_inc(tun->dev); | |
8ae1aff0 | 1636 | return err; |
5744ba05 | 1637 | } |
f1084c42 | 1638 | dev_sw_netstats_rx_add(tun->dev, xdp->data_end - xdp->data); |
8ae1aff0 JW |
1639 | break; |
1640 | case XDP_TX: | |
1641 | err = tun_xdp_tx(tun->dev, xdp); | |
5744ba05 YW |
1642 | if (err < 0) { |
1643 | dev_core_stats_rx_dropped_inc(tun->dev); | |
8ae1aff0 | 1644 | return err; |
5744ba05 | 1645 | } |
f1084c42 | 1646 | dev_sw_netstats_rx_add(tun->dev, xdp->data_end - xdp->data); |
8ae1aff0 JW |
1647 | break; |
1648 | case XDP_PASS: | |
1649 | break; | |
1650 | default: | |
c8064e5b | 1651 | bpf_warn_invalid_xdp_action(tun->dev, xdp_prog, act); |
df561f66 | 1652 | fallthrough; |
8ae1aff0 JW |
1653 | case XDP_ABORTED: |
1654 | trace_xdp_exception(tun->dev, xdp_prog, act); | |
df561f66 | 1655 | fallthrough; |
8ae1aff0 | 1656 | case XDP_DROP: |
625788b5 | 1657 | dev_core_stats_rx_dropped_inc(tun->dev); |
8ae1aff0 JW |
1658 | break; |
1659 | } | |
1660 | ||
1661 | return act; | |
1662 | } | |
1663 | ||
761876c8 JW |
1664 | static struct sk_buff *tun_build_skb(struct tun_struct *tun, |
1665 | struct tun_file *tfile, | |
66ccbc9c | 1666 | struct iov_iter *from, |
761876c8 | 1667 | struct virtio_net_hdr *hdr, |
1cfe6e93 | 1668 | int len, int *skb_xdp) |
66ccbc9c | 1669 | { |
0bbd7dad | 1670 | struct page_frag *alloc_frag = ¤t->task_frag; |
761876c8 | 1671 | struct bpf_prog *xdp_prog; |
7df13219 | 1672 | int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); |
66ccbc9c JW |
1673 | char *buf; |
1674 | size_t copied; | |
8ae1aff0 JW |
1675 | int pad = TUN_RX_PAD; |
1676 | int err = 0; | |
7df13219 JW |
1677 | |
1678 | rcu_read_lock(); | |
1679 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1680 | if (xdp_prog) | |
4f23aff8 | 1681 | pad += XDP_PACKET_HEADROOM; |
7df13219 JW |
1682 | buflen += SKB_DATA_ALIGN(len + pad); |
1683 | rcu_read_unlock(); | |
66ccbc9c | 1684 | |
63b9ab65 | 1685 | alloc_frag->offset = ALIGN((u64)alloc_frag->offset, SMP_CACHE_BYTES); |
66ccbc9c JW |
1686 | if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL))) |
1687 | return ERR_PTR(-ENOMEM); | |
1688 | ||
1689 | buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; | |
1690 | copied = copy_page_from_iter(alloc_frag->page, | |
7df13219 | 1691 | alloc_frag->offset + pad, |
66ccbc9c JW |
1692 | len, from); |
1693 | if (copied != len) | |
1694 | return ERR_PTR(-EFAULT); | |
1695 | ||
7df13219 JW |
1696 | /* There's a small window that XDP may be set after the check |
1697 | * of xdp_prog above, this should be rare and for simplicity | |
1698 | * we do XDP on skb in case the headroom is not enough. | |
1699 | */ | |
ac1f1f6c | 1700 | if (hdr->gso_type || !xdp_prog) { |
1cfe6e93 | 1701 | *skb_xdp = 1; |
4b663366 AB |
1702 | return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, |
1703 | pad); | |
ac1f1f6c JW |
1704 | } |
1705 | ||
1706 | *skb_xdp = 0; | |
761876c8 | 1707 | |
6547e387 | 1708 | local_bh_disable(); |
761876c8 JW |
1709 | rcu_read_lock(); |
1710 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
8ae1aff0 | 1711 | if (xdp_prog) { |
761876c8 | 1712 | struct xdp_buff xdp; |
761876c8 JW |
1713 | u32 act; |
1714 | ||
43b5169d | 1715 | xdp_init_buff(&xdp, buflen, &tfile->xdp_rxq); |
be9df4af | 1716 | xdp_prepare_buff(&xdp, buf, pad, len, false); |
761876c8 | 1717 | |
8ae1aff0 JW |
1718 | act = bpf_prog_run_xdp(xdp_prog, &xdp); |
1719 | if (act == XDP_REDIRECT || act == XDP_TX) { | |
59655a5b JW |
1720 | get_page(alloc_frag->page); |
1721 | alloc_frag->offset += buflen; | |
761876c8 | 1722 | } |
8ae1aff0 | 1723 | err = tun_xdp_act(tun, xdp_prog, &xdp, act); |
bee34890 WD |
1724 | if (err < 0) { |
1725 | if (act == XDP_REDIRECT || act == XDP_TX) | |
1726 | put_page(alloc_frag->page); | |
1727 | goto out; | |
1728 | } | |
1729 | ||
1a097910 | 1730 | if (err == XDP_REDIRECT) |
1d233886 | 1731 | xdp_do_flush(); |
8ae1aff0 JW |
1732 | if (err != XDP_PASS) |
1733 | goto out; | |
1734 | ||
1735 | pad = xdp.data - xdp.data_hard_start; | |
1736 | len = xdp.data_end - xdp.data; | |
761876c8 | 1737 | } |
291aeb2b JW |
1738 | rcu_read_unlock(); |
1739 | local_bh_enable(); | |
761876c8 | 1740 | |
4b663366 | 1741 | return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, pad); |
761876c8 | 1742 | |
f7053b6c | 1743 | out: |
761876c8 | 1744 | rcu_read_unlock(); |
6547e387 | 1745 | local_bh_enable(); |
761876c8 | 1746 | return NULL; |
66ccbc9c JW |
1747 | } |
1748 | ||
1da177e4 | 1749 | /* Get packet from user space buffer */ |
54f968d6 | 1750 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1751 | void *msg_control, struct iov_iter *from, |
5503fcec | 1752 | int noblock, bool more) |
1da177e4 | 1753 | { |
09640e63 | 1754 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1755 | struct sk_buff *skb; |
f5ff53b4 | 1756 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1757 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1758 | struct virtio_net_hdr gso = { 0 }; |
96f8d9ec | 1759 | int good_linear; |
0690899b MT |
1760 | int copylen; |
1761 | bool zerocopy = false; | |
1762 | int err; | |
96f84061 | 1763 | u32 rxhash = 0; |
1cfe6e93 | 1764 | int skb_xdp = 1; |
af3fb24e | 1765 | bool frags = tun_napi_frags_enabled(tfile); |
ab00af85 | 1766 | enum skb_drop_reason drop_reason = SKB_DROP_REASON_NOT_SPECIFIED; |
1da177e4 | 1767 | |
40630b82 | 1768 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1769 | if (len < sizeof(pi)) |
1da177e4 | 1770 | return -EINVAL; |
15718ea0 | 1771 | len -= sizeof(pi); |
1da177e4 | 1772 | |
cbbd26b8 | 1773 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1774 | return -EFAULT; |
1775 | } | |
1776 | ||
40630b82 | 1777 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1778 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1779 | ||
1780 | if (len < vnet_hdr_sz) | |
f43798c2 | 1781 | return -EINVAL; |
e1edab87 | 1782 | len -= vnet_hdr_sz; |
f43798c2 | 1783 | |
cbbd26b8 | 1784 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1785 | return -EFAULT; |
1786 | ||
4909122f | 1787 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1788 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1789 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1790 | |
56f0dcc5 | 1791 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1792 | return -EINVAL; |
e1edab87 | 1793 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1794 | } |
1795 | ||
40630b82 | 1796 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1797 | align += NET_IP_ALIGN; |
0eca93bc | 1798 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1799 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1800 | return -EINVAL; |
1801 | } | |
6aa20a22 | 1802 | |
96f8d9ec JW |
1803 | good_linear = SKB_MAX_HEAD(align); |
1804 | ||
88529176 | 1805 | if (msg_control) { |
f5ff53b4 AV |
1806 | struct iov_iter i = *from; |
1807 | ||
88529176 JW |
1808 | /* There are 256 bytes to be copied in skb, so there is |
1809 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1810 | * The rest of the buffer is mapped from userspace. |
1811 | */ | |
56f0dcc5 | 1812 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1813 | if (copylen > good_linear) |
1814 | copylen = good_linear; | |
3dd5c330 | 1815 | linear = copylen; |
f5ff53b4 AV |
1816 | iov_iter_advance(&i, copylen); |
1817 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1818 | zerocopy = true; |
1819 | } | |
1820 | ||
90e33d45 | 1821 | if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) { |
1cfe6e93 JW |
1822 | /* For the packet that is not easy to be processed |
1823 | * (e.g gso or jumbo packet), we will do it at after | |
1824 | * skb was created with generic XDP routine. | |
1825 | */ | |
1826 | skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp); | |
ab00af85 CW |
1827 | err = PTR_ERR_OR_ZERO(skb); |
1828 | if (err) | |
1829 | goto drop; | |
761876c8 JW |
1830 | if (!skb) |
1831 | return total_len; | |
66ccbc9c JW |
1832 | } else { |
1833 | if (!zerocopy) { | |
1834 | copylen = len; | |
1835 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) | |
1836 | linear = good_linear; | |
1837 | else | |
1838 | linear = tun16_to_cpu(tun, gso.hdr_len); | |
1839 | } | |
1da177e4 | 1840 | |
90e33d45 PP |
1841 | if (frags) { |
1842 | mutex_lock(&tfile->napi_mutex); | |
1843 | skb = tun_napi_alloc_frags(tfile, copylen, from); | |
1844 | /* tun_napi_alloc_frags() enforces a layout for the skb. | |
1845 | * If zerocopy is enabled, then this layout will be | |
1846 | * overwritten by zerocopy_sg_from_iter(). | |
1847 | */ | |
1848 | zerocopy = false; | |
1849 | } else { | |
6231e47b TE |
1850 | if (!linear) |
1851 | linear = min_t(size_t, good_linear, copylen); | |
1852 | ||
90e33d45 PP |
1853 | skb = tun_alloc_skb(tfile, align, copylen, linear, |
1854 | noblock); | |
1855 | } | |
1856 | ||
ab00af85 CW |
1857 | err = PTR_ERR_OR_ZERO(skb); |
1858 | if (err) | |
1859 | goto drop; | |
0690899b | 1860 | |
66ccbc9c JW |
1861 | if (zerocopy) |
1862 | err = zerocopy_sg_from_iter(skb, from); | |
1863 | else | |
1864 | err = skb_copy_datagram_from_iter(skb, 0, from, len); | |
1865 | ||
1866 | if (err) { | |
4477138f | 1867 | err = -EFAULT; |
4b4f052e | 1868 | drop_reason = SKB_DROP_REASON_SKB_UCOPY_FAULT; |
ab00af85 | 1869 | goto drop; |
66ccbc9c | 1870 | } |
8f22757e | 1871 | } |
1da177e4 | 1872 | |
3e9e40e7 | 1873 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
497a5757 | 1874 | atomic_long_inc(&tun->rx_frame_errors); |
ab00af85 CW |
1875 | err = -EINVAL; |
1876 | goto free_skb; | |
df10db98 PA |
1877 | } |
1878 | ||
1da177e4 | 1879 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1880 | case IFF_TUN: |
1881 | if (tun->flags & IFF_NO_PI) { | |
2580c4c1 AP |
1882 | u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0; |
1883 | ||
1884 | switch (ip_version) { | |
1885 | case 4: | |
f09f7ee2 AWC |
1886 | pi.proto = htons(ETH_P_IP); |
1887 | break; | |
2580c4c1 | 1888 | case 6: |
f09f7ee2 AWC |
1889 | pi.proto = htons(ETH_P_IPV6); |
1890 | break; | |
1891 | default: | |
ab00af85 CW |
1892 | err = -EINVAL; |
1893 | goto drop; | |
f09f7ee2 AWC |
1894 | } |
1895 | } | |
1896 | ||
459a98ed | 1897 | skb_reset_mac_header(skb); |
1da177e4 | 1898 | skb->protocol = pi.proto; |
4c13eb66 | 1899 | skb->dev = tun->dev; |
1da177e4 | 1900 | break; |
40630b82 | 1901 | case IFF_TAP: |
96aa1b22 WB |
1902 | if (frags && !pskb_may_pull(skb, ETH_HLEN)) { |
1903 | err = -ENOMEM; | |
4b4f052e | 1904 | drop_reason = SKB_DROP_REASON_HDR_TRUNC; |
96aa1b22 WB |
1905 | goto drop; |
1906 | } | |
1907 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1da177e4 | 1908 | break; |
6403eab1 | 1909 | } |
1da177e4 | 1910 | |
0690899b MT |
1911 | /* copy skb_ubuf_info for callback when skb has no error */ |
1912 | if (zerocopy) { | |
9ee5e5ad | 1913 | skb_zcopy_init(skb, msg_control); |
af1cc7a2 JW |
1914 | } else if (msg_control) { |
1915 | struct ubuf_info *uarg = msg_control; | |
36177832 | 1916 | uarg->callback(NULL, uarg, false); |
0690899b MT |
1917 | } |
1918 | ||
72f65107 | 1919 | skb_reset_network_header(skb); |
d2aa125d | 1920 | skb_probe_transport_header(skb); |
3fe260e0 | 1921 | skb_record_rx_queue(skb, tfile->queue_index); |
38502af7 | 1922 | |
1cfe6e93 | 1923 | if (skb_xdp) { |
761876c8 JW |
1924 | struct bpf_prog *xdp_prog; |
1925 | int ret; | |
1926 | ||
6547e387 | 1927 | local_bh_disable(); |
761876c8 JW |
1928 | rcu_read_lock(); |
1929 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1930 | if (xdp_prog) { | |
1931 | ret = do_xdp_generic(xdp_prog, skb); | |
1932 | if (ret != XDP_PASS) { | |
1933 | rcu_read_unlock(); | |
6547e387 | 1934 | local_bh_enable(); |
ab00af85 | 1935 | goto unlock_frags; |
761876c8 JW |
1936 | } |
1937 | } | |
1938 | rcu_read_unlock(); | |
6547e387 | 1939 | local_bh_enable(); |
761876c8 JW |
1940 | } |
1941 | ||
cf1a1e07 PA |
1942 | /* Compute the costly rx hash only if needed for flow updates. |
1943 | * We may get a very small possibility of OOO during switching, not | |
1944 | * worth to optimize. | |
1945 | */ | |
1946 | if (!rcu_access_pointer(tun->steering_prog) && tun->numqueues > 1 && | |
1947 | !tfile->detached) | |
96f84061 | 1948 | rxhash = __skb_get_hash_symmetric(skb); |
94317099 | 1949 | |
4477138f ED |
1950 | rcu_read_lock(); |
1951 | if (unlikely(!(tun->dev->flags & IFF_UP))) { | |
1952 | err = -EIO; | |
9180bb4f | 1953 | rcu_read_unlock(); |
4b4f052e | 1954 | drop_reason = SKB_DROP_REASON_DEV_READY; |
4477138f ED |
1955 | goto drop; |
1956 | } | |
1957 | ||
90e33d45 | 1958 | if (frags) { |
96aa1b22 WB |
1959 | u32 headlen; |
1960 | ||
90e33d45 | 1961 | /* Exercise flow dissector code path. */ |
96aa1b22 WB |
1962 | skb_push(skb, ETH_HLEN); |
1963 | headlen = eth_get_headlen(tun->dev, skb->data, | |
1964 | skb_headlen(skb)); | |
90e33d45 | 1965 | |
010f245b | 1966 | if (unlikely(headlen > skb_headlen(skb))) { |
07d120aa ED |
1967 | WARN_ON_ONCE(1); |
1968 | err = -ENOMEM; | |
625788b5 | 1969 | dev_core_stats_rx_dropped_inc(tun->dev); |
07d120aa | 1970 | napi_busy: |
90e33d45 | 1971 | napi_free_frags(&tfile->napi); |
4477138f | 1972 | rcu_read_unlock(); |
90e33d45 | 1973 | mutex_unlock(&tfile->napi_mutex); |
07d120aa | 1974 | return err; |
90e33d45 PP |
1975 | } |
1976 | ||
07d120aa ED |
1977 | if (likely(napi_schedule_prep(&tfile->napi))) { |
1978 | local_bh_disable(); | |
1979 | napi_gro_frags(&tfile->napi); | |
1980 | napi_complete(&tfile->napi); | |
1981 | local_bh_enable(); | |
1982 | } else { | |
1983 | err = -EBUSY; | |
1984 | goto napi_busy; | |
1985 | } | |
90e33d45 | 1986 | mutex_unlock(&tfile->napi_mutex); |
aec72f33 | 1987 | } else if (tfile->napi_enabled) { |
94317099 PP |
1988 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; |
1989 | int queue_len; | |
1990 | ||
1991 | spin_lock_bh(&queue->lock); | |
82b2bc27 KI |
1992 | |
1993 | if (unlikely(tfile->detached)) { | |
1994 | spin_unlock_bh(&queue->lock); | |
1995 | rcu_read_unlock(); | |
1996 | err = -EBUSY; | |
1997 | goto free_skb; | |
1998 | } | |
1999 | ||
94317099 PP |
2000 | __skb_queue_tail(queue, skb); |
2001 | queue_len = skb_queue_len(queue); | |
2002 | spin_unlock(&queue->lock); | |
2003 | ||
2004 | if (!more || queue_len > NAPI_POLL_WEIGHT) | |
2005 | napi_schedule(&tfile->napi); | |
2006 | ||
2007 | local_bh_enable(); | |
2008 | } else if (!IS_ENABLED(CONFIG_4KSTACKS)) { | |
2009 | tun_rx_batched(tun, tfile, skb, more); | |
2010 | } else { | |
3d391f65 | 2011 | netif_rx(skb); |
94317099 | 2012 | } |
4477138f | 2013 | rcu_read_unlock(); |
6aa20a22 | 2014 | |
497a5757 HK |
2015 | preempt_disable(); |
2016 | dev_sw_netstats_rx_add(tun->dev, len); | |
2017 | preempt_enable(); | |
1da177e4 | 2018 | |
96f84061 JW |
2019 | if (rxhash) |
2020 | tun_flow_update(tun, rxhash, tfile); | |
2021 | ||
0690899b | 2022 | return total_len; |
ab00af85 CW |
2023 | |
2024 | drop: | |
2025 | if (err != -EAGAIN) | |
2026 | dev_core_stats_rx_dropped_inc(tun->dev); | |
2027 | ||
2028 | free_skb: | |
2029 | if (!IS_ERR_OR_NULL(skb)) | |
2030 | kfree_skb_reason(skb, drop_reason); | |
2031 | ||
2032 | unlock_frags: | |
2033 | if (frags) { | |
2034 | tfile->napi.skb = NULL; | |
2035 | mutex_unlock(&tfile->napi_mutex); | |
2036 | } | |
2037 | ||
2038 | return err ?: total_len; | |
6aa20a22 | 2039 | } |
1da177e4 | 2040 | |
f5ff53b4 | 2041 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 2042 | { |
33dccbb0 | 2043 | struct file *file = iocb->ki_filp; |
54f968d6 | 2044 | struct tun_file *tfile = file->private_data; |
9484dc74 | 2045 | struct tun_struct *tun = tun_get(tfile); |
631ab46b | 2046 | ssize_t result; |
5aac0390 | 2047 | int noblock = 0; |
1da177e4 LT |
2048 | |
2049 | if (!tun) | |
2050 | return -EBADFD; | |
2051 | ||
5aac0390 JA |
2052 | if ((file->f_flags & O_NONBLOCK) || (iocb->ki_flags & IOCB_NOWAIT)) |
2053 | noblock = 1; | |
2054 | ||
2055 | result = tun_get_user(tun, tfile, NULL, from, noblock, false); | |
631ab46b EB |
2056 | |
2057 | tun_put(tun); | |
2058 | return result; | |
1da177e4 LT |
2059 | } |
2060 | ||
fc72d1d5 JW |
2061 | static ssize_t tun_put_user_xdp(struct tun_struct *tun, |
2062 | struct tun_file *tfile, | |
1ffcbc85 | 2063 | struct xdp_frame *xdp_frame, |
fc72d1d5 JW |
2064 | struct iov_iter *iter) |
2065 | { | |
2066 | int vnet_hdr_sz = 0; | |
1ffcbc85 | 2067 | size_t size = xdp_frame->len; |
fc72d1d5 JW |
2068 | size_t ret; |
2069 | ||
2070 | if (tun->flags & IFF_VNET_HDR) { | |
2071 | struct virtio_net_hdr gso = { 0 }; | |
2072 | ||
2073 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); | |
2074 | if (unlikely(iov_iter_count(iter) < vnet_hdr_sz)) | |
2075 | return -EINVAL; | |
2076 | if (unlikely(copy_to_iter(&gso, sizeof(gso), iter) != | |
2077 | sizeof(gso))) | |
2078 | return -EFAULT; | |
2079 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
2080 | } | |
2081 | ||
1ffcbc85 | 2082 | ret = copy_to_iter(xdp_frame->data, size, iter) + vnet_hdr_sz; |
fc72d1d5 | 2083 | |
497a5757 HK |
2084 | preempt_disable(); |
2085 | dev_sw_netstats_tx_add(tun->dev, 1, ret); | |
2086 | preempt_enable(); | |
fc72d1d5 JW |
2087 | |
2088 | return ret; | |
2089 | } | |
2090 | ||
1da177e4 | 2091 | /* Put packet to the user space buffer */ |
6f7c156c | 2092 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 2093 | struct tun_file *tfile, |
6f7c156c | 2094 | struct sk_buff *skb, |
e0b46d0e | 2095 | struct iov_iter *iter) |
1da177e4 LT |
2096 | { |
2097 | struct tun_pi pi = { 0, skb->protocol }; | |
e0b46d0e | 2098 | ssize_t total; |
8c847d25 | 2099 | int vlan_offset = 0; |
a8f9bfdf | 2100 | int vlan_hlen = 0; |
2eb783c4 | 2101 | int vnet_hdr_sz = 0; |
a8f9bfdf | 2102 | |
df8a39de | 2103 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 2104 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 2105 | |
40630b82 | 2106 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 2107 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 2108 | |
e0b46d0e HX |
2109 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
2110 | ||
40630b82 | 2111 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 2112 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
2113 | return -EINVAL; |
2114 | ||
e0b46d0e HX |
2115 | total += sizeof(pi); |
2116 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
2117 | /* Packet will be striped */ |
2118 | pi.flags |= TUN_PKT_STRIP; | |
2119 | } | |
6aa20a22 | 2120 | |
e0b46d0e | 2121 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 2122 | return -EFAULT; |
6aa20a22 | 2123 | } |
1da177e4 | 2124 | |
2eb783c4 | 2125 | if (vnet_hdr_sz) { |
9403cd7c | 2126 | struct virtio_net_hdr gso; |
34166093 | 2127 | |
e0b46d0e | 2128 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
2129 | return -EINVAL; |
2130 | ||
3e9e40e7 | 2131 | if (virtio_net_hdr_from_skb(skb, &gso, |
fd3a8862 WB |
2132 | tun_is_little_endian(tun), true, |
2133 | vlan_hlen)) { | |
f43798c2 | 2134 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
2135 | pr_err("unexpected GSO type: " |
2136 | "0x%x, gso_size %d, hdr_len %d\n", | |
2137 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
2138 | tun16_to_cpu(tun, gso.hdr_len)); | |
2139 | print_hex_dump(KERN_ERR, "tun: ", | |
2140 | DUMP_PREFIX_NONE, | |
2141 | 16, 1, skb->head, | |
2142 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
2143 | WARN_ON_ONCE(1); | |
2144 | return -EINVAL; | |
2145 | } | |
f43798c2 | 2146 | |
e0b46d0e | 2147 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 2148 | return -EFAULT; |
8c847d25 JW |
2149 | |
2150 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
2151 | } |
2152 | ||
a8f9bfdf | 2153 | if (vlan_hlen) { |
e0b46d0e | 2154 | int ret; |
aff3d70a | 2155 | struct veth veth; |
6680ec68 JW |
2156 | |
2157 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 2158 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
2159 | |
2160 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 2161 | |
e0b46d0e HX |
2162 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
2163 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
2164 | goto done; |
2165 | ||
e0b46d0e HX |
2166 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
2167 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
2168 | goto done; |
2169 | } | |
1da177e4 | 2170 | |
e0b46d0e | 2171 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 2172 | |
6680ec68 | 2173 | done: |
608b9977 | 2174 | /* caller is in process context, */ |
497a5757 HK |
2175 | preempt_disable(); |
2176 | dev_sw_netstats_tx_add(tun->dev, 1, skb->len + vlan_hlen); | |
2177 | preempt_enable(); | |
1da177e4 LT |
2178 | |
2179 | return total; | |
2180 | } | |
2181 | ||
fc72d1d5 | 2182 | static void *tun_ring_recv(struct tun_file *tfile, int noblock, int *err) |
1576d986 JW |
2183 | { |
2184 | DECLARE_WAITQUEUE(wait, current); | |
fc72d1d5 | 2185 | void *ptr = NULL; |
f48cc6b2 | 2186 | int error = 0; |
1576d986 | 2187 | |
fc72d1d5 JW |
2188 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2189 | if (ptr) | |
1576d986 JW |
2190 | goto out; |
2191 | if (noblock) { | |
f48cc6b2 | 2192 | error = -EAGAIN; |
1576d986 JW |
2193 | goto out; |
2194 | } | |
2195 | ||
333f7909 | 2196 | add_wait_queue(&tfile->socket.wq.wait, &wait); |
1576d986 JW |
2197 | |
2198 | while (1) { | |
71828b22 | 2199 | set_current_state(TASK_INTERRUPTIBLE); |
fc72d1d5 JW |
2200 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2201 | if (ptr) | |
1576d986 JW |
2202 | break; |
2203 | if (signal_pending(current)) { | |
f48cc6b2 | 2204 | error = -ERESTARTSYS; |
1576d986 JW |
2205 | break; |
2206 | } | |
2207 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 2208 | error = -EFAULT; |
1576d986 JW |
2209 | break; |
2210 | } | |
2211 | ||
2212 | schedule(); | |
2213 | } | |
2214 | ||
ecef67cb | 2215 | __set_current_state(TASK_RUNNING); |
333f7909 | 2216 | remove_wait_queue(&tfile->socket.wq.wait, &wait); |
1576d986 JW |
2217 | |
2218 | out: | |
f48cc6b2 | 2219 | *err = error; |
fc72d1d5 | 2220 | return ptr; |
1576d986 JW |
2221 | } |
2222 | ||
54f968d6 | 2223 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 | 2224 | struct iov_iter *to, |
fc72d1d5 | 2225 | int noblock, void *ptr) |
1da177e4 | 2226 | { |
9b067034 | 2227 | ssize_t ret; |
1576d986 | 2228 | int err; |
1da177e4 | 2229 | |
c33ee15b | 2230 | if (!iov_iter_count(to)) { |
fc72d1d5 | 2231 | tun_ptr_free(ptr); |
9b067034 | 2232 | return 0; |
c33ee15b | 2233 | } |
1da177e4 | 2234 | |
fc72d1d5 | 2235 | if (!ptr) { |
ac77cfd4 | 2236 | /* Read frames from ring */ |
fc72d1d5 JW |
2237 | ptr = tun_ring_recv(tfile, noblock, &err); |
2238 | if (!ptr) | |
ac77cfd4 JW |
2239 | return err; |
2240 | } | |
e0b46d0e | 2241 | |
1ffcbc85 JDB |
2242 | if (tun_is_xdp_frame(ptr)) { |
2243 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2244 | |
1ffcbc85 | 2245 | ret = tun_put_user_xdp(tun, tfile, xdpf, to); |
03993094 | 2246 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
2247 | } else { |
2248 | struct sk_buff *skb = ptr; | |
2249 | ||
2250 | ret = tun_put_user(tun, tfile, skb, to); | |
2251 | if (unlikely(ret < 0)) | |
2252 | kfree_skb(skb); | |
2253 | else | |
2254 | consume_skb(skb); | |
2255 | } | |
1da177e4 | 2256 | |
05c2828c MT |
2257 | return ret; |
2258 | } | |
2259 | ||
9b067034 | 2260 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
2261 | { |
2262 | struct file *file = iocb->ki_filp; | |
2263 | struct tun_file *tfile = file->private_data; | |
9484dc74 | 2264 | struct tun_struct *tun = tun_get(tfile); |
9b067034 | 2265 | ssize_t len = iov_iter_count(to), ret; |
5aac0390 | 2266 | int noblock = 0; |
05c2828c MT |
2267 | |
2268 | if (!tun) | |
2269 | return -EBADFD; | |
5aac0390 JA |
2270 | |
2271 | if ((file->f_flags & O_NONBLOCK) || (iocb->ki_flags & IOCB_NOWAIT)) | |
2272 | noblock = 1; | |
2273 | ||
2274 | ret = tun_do_read(tun, tfile, to, noblock, NULL); | |
42404c09 | 2275 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
2276 | if (ret > 0) |
2277 | iocb->ki_pos = ret; | |
631ab46b | 2278 | tun_put(tun); |
1da177e4 LT |
2279 | return ret; |
2280 | } | |
2281 | ||
cd5681d7 | 2282 | static void tun_prog_free(struct rcu_head *rcu) |
96f84061 | 2283 | { |
cd5681d7 | 2284 | struct tun_prog *prog = container_of(rcu, struct tun_prog, rcu); |
96f84061 JW |
2285 | |
2286 | bpf_prog_destroy(prog->prog); | |
2287 | kfree(prog); | |
2288 | } | |
2289 | ||
9d6474e4 JW |
2290 | static int __tun_set_ebpf(struct tun_struct *tun, |
2291 | struct tun_prog __rcu **prog_p, | |
cd5681d7 | 2292 | struct bpf_prog *prog) |
96f84061 | 2293 | { |
cd5681d7 | 2294 | struct tun_prog *old, *new = NULL; |
96f84061 JW |
2295 | |
2296 | if (prog) { | |
2297 | new = kmalloc(sizeof(*new), GFP_KERNEL); | |
2298 | if (!new) | |
2299 | return -ENOMEM; | |
2300 | new->prog = prog; | |
2301 | } | |
2302 | ||
124da8f6 | 2303 | spin_lock_bh(&tun->lock); |
cd5681d7 | 2304 | old = rcu_dereference_protected(*prog_p, |
124da8f6 | 2305 | lockdep_is_held(&tun->lock)); |
cd5681d7 | 2306 | rcu_assign_pointer(*prog_p, new); |
124da8f6 | 2307 | spin_unlock_bh(&tun->lock); |
96f84061 JW |
2308 | |
2309 | if (old) | |
cd5681d7 | 2310 | call_rcu(&old->rcu, tun_prog_free); |
96f84061 JW |
2311 | |
2312 | return 0; | |
2313 | } | |
2314 | ||
96442e42 JW |
2315 | static void tun_free_netdev(struct net_device *dev) |
2316 | { | |
2317 | struct tun_struct *tun = netdev_priv(dev); | |
2318 | ||
4008e97f | 2319 | BUG_ON(!(list_empty(&tun->disabled))); |
11fc7d5a | 2320 | |
497a5757 | 2321 | free_percpu(dev->tstats); |
96442e42 | 2322 | tun_flow_uninit(tun); |
5dbbaf2d | 2323 | security_tun_dev_free_security(tun->security); |
cd5681d7 | 2324 | __tun_set_ebpf(tun, &tun->steering_prog, NULL); |
aff3d70a | 2325 | __tun_set_ebpf(tun, &tun->filter_prog, NULL); |
96442e42 JW |
2326 | } |
2327 | ||
1da177e4 LT |
2328 | static void tun_setup(struct net_device *dev) |
2329 | { | |
2330 | struct tun_struct *tun = netdev_priv(dev); | |
2331 | ||
0625c883 EB |
2332 | tun->owner = INVALID_UID; |
2333 | tun->group = INVALID_GID; | |
4e24f2dd | 2334 | tun_default_link_ksettings(dev, &tun->link_ksettings); |
1da177e4 | 2335 | |
1da177e4 | 2336 | dev->ethtool_ops = &tun_ethtool_ops; |
cf124db5 DM |
2337 | dev->needs_free_netdev = true; |
2338 | dev->priv_destructor = tun_free_netdev; | |
016adb72 JW |
2339 | /* We prefer our own queue length */ |
2340 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
2341 | } |
2342 | ||
f019a7a5 EB |
2343 | /* Trivial set of netlink ops to allow deleting tun or tap |
2344 | * device with netlink. | |
2345 | */ | |
a8b8a889 MS |
2346 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[], |
2347 | struct netlink_ext_ack *extack) | |
f019a7a5 | 2348 | { |
35b827b6 ND |
2349 | NL_SET_ERR_MSG(extack, |
2350 | "tun/tap creation via rtnetlink is not supported."); | |
2351 | return -EOPNOTSUPP; | |
f019a7a5 EB |
2352 | } |
2353 | ||
1ec010e7 SD |
2354 | static size_t tun_get_size(const struct net_device *dev) |
2355 | { | |
2356 | BUILD_BUG_ON(sizeof(u32) != sizeof(uid_t)); | |
2357 | BUILD_BUG_ON(sizeof(u32) != sizeof(gid_t)); | |
2358 | ||
2359 | return nla_total_size(sizeof(uid_t)) + /* OWNER */ | |
2360 | nla_total_size(sizeof(gid_t)) + /* GROUP */ | |
2361 | nla_total_size(sizeof(u8)) + /* TYPE */ | |
2362 | nla_total_size(sizeof(u8)) + /* PI */ | |
2363 | nla_total_size(sizeof(u8)) + /* VNET_HDR */ | |
2364 | nla_total_size(sizeof(u8)) + /* PERSIST */ | |
2365 | nla_total_size(sizeof(u8)) + /* MULTI_QUEUE */ | |
2366 | nla_total_size(sizeof(u32)) + /* NUM_QUEUES */ | |
2367 | nla_total_size(sizeof(u32)) + /* NUM_DISABLED_QUEUES */ | |
2368 | 0; | |
2369 | } | |
2370 | ||
2371 | static int tun_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
2372 | { | |
2373 | struct tun_struct *tun = netdev_priv(dev); | |
2374 | ||
2375 | if (nla_put_u8(skb, IFLA_TUN_TYPE, tun->flags & TUN_TYPE_MASK)) | |
2376 | goto nla_put_failure; | |
2377 | if (uid_valid(tun->owner) && | |
2378 | nla_put_u32(skb, IFLA_TUN_OWNER, | |
2379 | from_kuid_munged(current_user_ns(), tun->owner))) | |
2380 | goto nla_put_failure; | |
2381 | if (gid_valid(tun->group) && | |
2382 | nla_put_u32(skb, IFLA_TUN_GROUP, | |
2383 | from_kgid_munged(current_user_ns(), tun->group))) | |
2384 | goto nla_put_failure; | |
2385 | if (nla_put_u8(skb, IFLA_TUN_PI, !(tun->flags & IFF_NO_PI))) | |
2386 | goto nla_put_failure; | |
2387 | if (nla_put_u8(skb, IFLA_TUN_VNET_HDR, !!(tun->flags & IFF_VNET_HDR))) | |
2388 | goto nla_put_failure; | |
2389 | if (nla_put_u8(skb, IFLA_TUN_PERSIST, !!(tun->flags & IFF_PERSIST))) | |
2390 | goto nla_put_failure; | |
2391 | if (nla_put_u8(skb, IFLA_TUN_MULTI_QUEUE, | |
2392 | !!(tun->flags & IFF_MULTI_QUEUE))) | |
2393 | goto nla_put_failure; | |
2394 | if (tun->flags & IFF_MULTI_QUEUE) { | |
2395 | if (nla_put_u32(skb, IFLA_TUN_NUM_QUEUES, tun->numqueues)) | |
2396 | goto nla_put_failure; | |
2397 | if (nla_put_u32(skb, IFLA_TUN_NUM_DISABLED_QUEUES, | |
2398 | tun->numdisabled)) | |
2399 | goto nla_put_failure; | |
2400 | } | |
2401 | ||
2402 | return 0; | |
2403 | ||
2404 | nla_put_failure: | |
2405 | return -EMSGSIZE; | |
2406 | } | |
2407 | ||
f019a7a5 EB |
2408 | static struct rtnl_link_ops tun_link_ops __read_mostly = { |
2409 | .kind = DRV_NAME, | |
2410 | .priv_size = sizeof(struct tun_struct), | |
2411 | .setup = tun_setup, | |
2412 | .validate = tun_validate, | |
1ec010e7 SD |
2413 | .get_size = tun_get_size, |
2414 | .fill_info = tun_fill_info, | |
f019a7a5 EB |
2415 | }; |
2416 | ||
33dccbb0 HX |
2417 | static void tun_sock_write_space(struct sock *sk) |
2418 | { | |
54f968d6 | 2419 | struct tun_file *tfile; |
43815482 | 2420 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
2421 | |
2422 | if (!sock_writeable(sk)) | |
2423 | return; | |
2424 | ||
9cd3e072 | 2425 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
2426 | return; |
2427 | ||
43815482 ED |
2428 | wqueue = sk_sleep(sk); |
2429 | if (wqueue && waitqueue_active(wqueue)) | |
a9a08845 LT |
2430 | wake_up_interruptible_sync_poll(wqueue, EPOLLOUT | |
2431 | EPOLLWRNORM | EPOLLWRBAND); | |
c722c625 | 2432 | |
54f968d6 JW |
2433 | tfile = container_of(sk, struct tun_file, sk); |
2434 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
2435 | } |
2436 | ||
f9e06c45 JW |
2437 | static void tun_put_page(struct tun_page *tpage) |
2438 | { | |
2439 | if (tpage->page) | |
2440 | __page_frag_cache_drain(tpage->page, tpage->count); | |
2441 | } | |
2442 | ||
043d222f JW |
2443 | static int tun_xdp_one(struct tun_struct *tun, |
2444 | struct tun_file *tfile, | |
f9e06c45 JW |
2445 | struct xdp_buff *xdp, int *flush, |
2446 | struct tun_page *tpage) | |
043d222f | 2447 | { |
4e4b08e5 | 2448 | unsigned int datasize = xdp->data_end - xdp->data; |
043d222f JW |
2449 | struct tun_xdp_hdr *hdr = xdp->data_hard_start; |
2450 | struct virtio_net_hdr *gso = &hdr->gso; | |
043d222f JW |
2451 | struct bpf_prog *xdp_prog; |
2452 | struct sk_buff *skb = NULL; | |
fb3f9037 | 2453 | struct sk_buff_head *queue; |
043d222f JW |
2454 | u32 rxhash = 0, act; |
2455 | int buflen = hdr->buflen; | |
fb3f9037 | 2456 | int ret = 0; |
043d222f | 2457 | bool skb_xdp = false; |
f9e06c45 | 2458 | struct page *page; |
043d222f JW |
2459 | |
2460 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
2461 | if (xdp_prog) { | |
2462 | if (gso->gso_type) { | |
2463 | skb_xdp = true; | |
2464 | goto build; | |
2465 | } | |
43b5169d LB |
2466 | |
2467 | xdp_init_buff(xdp, buflen, &tfile->xdp_rxq); | |
043d222f | 2468 | xdp_set_data_meta_invalid(xdp); |
043d222f JW |
2469 | |
2470 | act = bpf_prog_run_xdp(xdp_prog, xdp); | |
fb3f9037 HH |
2471 | ret = tun_xdp_act(tun, xdp_prog, xdp, act); |
2472 | if (ret < 0) { | |
043d222f | 2473 | put_page(virt_to_head_page(xdp->data)); |
fb3f9037 | 2474 | return ret; |
043d222f JW |
2475 | } |
2476 | ||
fb3f9037 | 2477 | switch (ret) { |
043d222f JW |
2478 | case XDP_REDIRECT: |
2479 | *flush = true; | |
df561f66 | 2480 | fallthrough; |
043d222f JW |
2481 | case XDP_TX: |
2482 | return 0; | |
2483 | case XDP_PASS: | |
2484 | break; | |
2485 | default: | |
f9e06c45 JW |
2486 | page = virt_to_head_page(xdp->data); |
2487 | if (tpage->page == page) { | |
2488 | ++tpage->count; | |
2489 | } else { | |
2490 | tun_put_page(tpage); | |
2491 | tpage->page = page; | |
2492 | tpage->count = 1; | |
2493 | } | |
043d222f JW |
2494 | return 0; |
2495 | } | |
2496 | } | |
2497 | ||
2498 | build: | |
2499 | skb = build_skb(xdp->data_hard_start, buflen); | |
2500 | if (!skb) { | |
fb3f9037 | 2501 | ret = -ENOMEM; |
043d222f JW |
2502 | goto out; |
2503 | } | |
2504 | ||
2505 | skb_reserve(skb, xdp->data - xdp->data_hard_start); | |
2506 | skb_put(skb, xdp->data_end - xdp->data); | |
2507 | ||
2508 | if (virtio_net_hdr_to_skb(skb, gso, tun_is_little_endian(tun))) { | |
497a5757 | 2509 | atomic_long_inc(&tun->rx_frame_errors); |
043d222f | 2510 | kfree_skb(skb); |
fb3f9037 | 2511 | ret = -EINVAL; |
043d222f JW |
2512 | goto out; |
2513 | } | |
2514 | ||
2515 | skb->protocol = eth_type_trans(skb, tun->dev); | |
2516 | skb_reset_network_header(skb); | |
d2aa125d | 2517 | skb_probe_transport_header(skb); |
3fe260e0 | 2518 | skb_record_rx_queue(skb, tfile->queue_index); |
043d222f JW |
2519 | |
2520 | if (skb_xdp) { | |
fb3f9037 HH |
2521 | ret = do_xdp_generic(xdp_prog, skb); |
2522 | if (ret != XDP_PASS) { | |
2523 | ret = 0; | |
043d222f | 2524 | goto out; |
fb3f9037 | 2525 | } |
043d222f JW |
2526 | } |
2527 | ||
f29eb2a9 PA |
2528 | if (!rcu_dereference(tun->steering_prog) && tun->numqueues > 1 && |
2529 | !tfile->detached) | |
043d222f JW |
2530 | rxhash = __skb_get_hash_symmetric(skb); |
2531 | ||
fb3f9037 HH |
2532 | if (tfile->napi_enabled) { |
2533 | queue = &tfile->sk.sk_write_queue; | |
2534 | spin_lock(&queue->lock); | |
82b2bc27 KI |
2535 | |
2536 | if (unlikely(tfile->detached)) { | |
2537 | spin_unlock(&queue->lock); | |
2538 | kfree_skb(skb); | |
2539 | return -EBUSY; | |
2540 | } | |
2541 | ||
fb3f9037 HH |
2542 | __skb_queue_tail(queue, skb); |
2543 | spin_unlock(&queue->lock); | |
2544 | ret = 1; | |
2545 | } else { | |
2546 | netif_receive_skb(skb); | |
2547 | ret = 0; | |
2548 | } | |
043d222f | 2549 | |
497a5757 | 2550 | /* No need to disable preemption here since this function is |
6342ca64 PB |
2551 | * always called with bh disabled |
2552 | */ | |
497a5757 | 2553 | dev_sw_netstats_rx_add(tun->dev, datasize); |
043d222f JW |
2554 | |
2555 | if (rxhash) | |
2556 | tun_flow_update(tun, rxhash, tfile); | |
2557 | ||
2558 | out: | |
fb3f9037 | 2559 | return ret; |
043d222f JW |
2560 | } |
2561 | ||
1b784140 | 2562 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 2563 | { |
043d222f | 2564 | int ret, i; |
54f968d6 | 2565 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2566 | struct tun_struct *tun = tun_get(tfile); |
fe8dd45b | 2567 | struct tun_msg_ctl *ctl = m->msg_control; |
043d222f | 2568 | struct xdp_buff *xdp; |
54f968d6 JW |
2569 | |
2570 | if (!tun) | |
2571 | return -EBADFD; | |
f5ff53b4 | 2572 | |
74a335a0 HH |
2573 | if (m->msg_controllen == sizeof(struct tun_msg_ctl) && |
2574 | ctl && ctl->type == TUN_MSG_PTR) { | |
6f0271d9 | 2575 | struct tun_page tpage; |
043d222f | 2576 | int n = ctl->num; |
fb3f9037 | 2577 | int flush = 0, queued = 0; |
043d222f | 2578 | |
6f0271d9 DM |
2579 | memset(&tpage, 0, sizeof(tpage)); |
2580 | ||
043d222f JW |
2581 | local_bh_disable(); |
2582 | rcu_read_lock(); | |
2583 | ||
2584 | for (i = 0; i < n; i++) { | |
2585 | xdp = &((struct xdp_buff *)ctl->ptr)[i]; | |
fb3f9037 HH |
2586 | ret = tun_xdp_one(tun, tfile, xdp, &flush, &tpage); |
2587 | if (ret > 0) | |
2588 | queued += ret; | |
043d222f JW |
2589 | } |
2590 | ||
2591 | if (flush) | |
1d233886 | 2592 | xdp_do_flush(); |
043d222f | 2593 | |
fb3f9037 HH |
2594 | if (tfile->napi_enabled && queued > 0) |
2595 | napi_schedule(&tfile->napi); | |
2596 | ||
043d222f JW |
2597 | rcu_read_unlock(); |
2598 | local_bh_enable(); | |
2599 | ||
f9e06c45 JW |
2600 | tun_put_page(&tpage); |
2601 | ||
043d222f JW |
2602 | ret = total_len; |
2603 | goto out; | |
2604 | } | |
fe8dd45b JW |
2605 | |
2606 | ret = tun_get_user(tun, tfile, ctl ? ctl->ptr : NULL, &m->msg_iter, | |
5503fcec JW |
2607 | m->msg_flags & MSG_DONTWAIT, |
2608 | m->msg_flags & MSG_MORE); | |
043d222f | 2609 | out: |
54f968d6 JW |
2610 | tun_put(tun); |
2611 | return ret; | |
05c2828c MT |
2612 | } |
2613 | ||
1b784140 | 2614 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
2615 | int flags) |
2616 | { | |
54f968d6 | 2617 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2618 | struct tun_struct *tun = tun_get(tfile); |
fc72d1d5 | 2619 | void *ptr = m->msg_control; |
05c2828c | 2620 | int ret; |
54f968d6 | 2621 | |
c33ee15b WX |
2622 | if (!tun) { |
2623 | ret = -EBADFD; | |
fc72d1d5 | 2624 | goto out_free; |
c33ee15b | 2625 | } |
54f968d6 | 2626 | |
eda29772 | 2627 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 | 2628 | ret = -EINVAL; |
c33ee15b | 2629 | goto out_put_tun; |
3811ae76 | 2630 | } |
eda29772 RC |
2631 | if (flags & MSG_ERRQUEUE) { |
2632 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
2633 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
2634 | goto out; | |
2635 | } | |
fc72d1d5 | 2636 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, ptr); |
87897931 | 2637 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
2638 | m->msg_flags |= MSG_TRUNC; |
2639 | ret = flags & MSG_TRUNC ? ret : total_len; | |
2640 | } | |
3811ae76 | 2641 | out: |
54f968d6 | 2642 | tun_put(tun); |
05c2828c | 2643 | return ret; |
c33ee15b WX |
2644 | |
2645 | out_put_tun: | |
2646 | tun_put(tun); | |
fc72d1d5 JW |
2647 | out_free: |
2648 | tun_ptr_free(ptr); | |
c33ee15b | 2649 | return ret; |
05c2828c MT |
2650 | } |
2651 | ||
fc72d1d5 JW |
2652 | static int tun_ptr_peek_len(void *ptr) |
2653 | { | |
2654 | if (likely(ptr)) { | |
1ffcbc85 JDB |
2655 | if (tun_is_xdp_frame(ptr)) { |
2656 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2657 | |
1ffcbc85 | 2658 | return xdpf->len; |
fc72d1d5 JW |
2659 | } |
2660 | return __skb_array_len_with_tag(ptr); | |
2661 | } else { | |
2662 | return 0; | |
2663 | } | |
2664 | } | |
2665 | ||
1576d986 JW |
2666 | static int tun_peek_len(struct socket *sock) |
2667 | { | |
2668 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
2669 | struct tun_struct *tun; | |
2670 | int ret = 0; | |
2671 | ||
9484dc74 | 2672 | tun = tun_get(tfile); |
1576d986 JW |
2673 | if (!tun) |
2674 | return 0; | |
2675 | ||
fc72d1d5 | 2676 | ret = PTR_RING_PEEK_CALL(&tfile->tx_ring, tun_ptr_peek_len); |
1576d986 JW |
2677 | tun_put(tun); |
2678 | ||
2679 | return ret; | |
2680 | } | |
2681 | ||
05c2828c MT |
2682 | /* Ops structure to mimic raw sockets with tun */ |
2683 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 2684 | .peek_len = tun_peek_len, |
05c2828c MT |
2685 | .sendmsg = tun_sendmsg, |
2686 | .recvmsg = tun_recvmsg, | |
2687 | }; | |
2688 | ||
33dccbb0 HX |
2689 | static struct proto tun_proto = { |
2690 | .name = "tun", | |
2691 | .owner = THIS_MODULE, | |
54f968d6 | 2692 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 2693 | }; |
f019a7a5 | 2694 | |
980c9e8c DW |
2695 | static int tun_flags(struct tun_struct *tun) |
2696 | { | |
031f5e03 | 2697 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
2698 | } |
2699 | ||
bc6d076d | 2700 | static ssize_t tun_flags_show(struct device *dev, struct device_attribute *attr, |
980c9e8c DW |
2701 | char *buf) |
2702 | { | |
2703 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
aff30699 | 2704 | return sysfs_emit(buf, "0x%x\n", tun_flags(tun)); |
980c9e8c DW |
2705 | } |
2706 | ||
bc6d076d Y |
2707 | static ssize_t owner_show(struct device *dev, struct device_attribute *attr, |
2708 | char *buf) | |
980c9e8c DW |
2709 | { |
2710 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 | 2711 | return uid_valid(tun->owner)? |
aff30699 WY |
2712 | sysfs_emit(buf, "%u\n", |
2713 | from_kuid_munged(current_user_ns(), tun->owner)) : | |
2714 | sysfs_emit(buf, "-1\n"); | |
980c9e8c DW |
2715 | } |
2716 | ||
bc6d076d Y |
2717 | static ssize_t group_show(struct device *dev, struct device_attribute *attr, |
2718 | char *buf) | |
980c9e8c DW |
2719 | { |
2720 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 | 2721 | return gid_valid(tun->group) ? |
aff30699 WY |
2722 | sysfs_emit(buf, "%u\n", |
2723 | from_kgid_munged(current_user_ns(), tun->group)) : | |
2724 | sysfs_emit(buf, "-1\n"); | |
980c9e8c DW |
2725 | } |
2726 | ||
bc6d076d Y |
2727 | static DEVICE_ATTR_RO(tun_flags); |
2728 | static DEVICE_ATTR_RO(owner); | |
2729 | static DEVICE_ATTR_RO(group); | |
980c9e8c | 2730 | |
c4d33e24 TI |
2731 | static struct attribute *tun_dev_attrs[] = { |
2732 | &dev_attr_tun_flags.attr, | |
2733 | &dev_attr_owner.attr, | |
2734 | &dev_attr_group.attr, | |
2735 | NULL | |
2736 | }; | |
2737 | ||
2738 | static const struct attribute_group tun_attr_group = { | |
2739 | .attrs = tun_dev_attrs | |
2740 | }; | |
2741 | ||
d647a591 | 2742 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
2743 | { |
2744 | struct tun_struct *tun; | |
54f968d6 | 2745 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2746 | struct net_device *dev; |
2747 | int err; | |
2748 | ||
7c0c3b1a JW |
2749 | if (tfile->detached) |
2750 | return -EINVAL; | |
2751 | ||
90e33d45 PP |
2752 | if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) { |
2753 | if (!capable(CAP_NET_ADMIN)) | |
2754 | return -EPERM; | |
2755 | ||
2756 | if (!(ifr->ifr_flags & IFF_NAPI) || | |
2757 | (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP) | |
2758 | return -EINVAL; | |
2759 | } | |
2760 | ||
74a3e5a7 EB |
2761 | dev = __dev_get_by_name(net, ifr->ifr_name); |
2762 | if (dev) { | |
f85ba780 DW |
2763 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
2764 | return -EBUSY; | |
74a3e5a7 EB |
2765 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
2766 | tun = netdev_priv(dev); | |
2767 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
2768 | tun = netdev_priv(dev); | |
2769 | else | |
2770 | return -EINVAL; | |
2771 | ||
8e6d91ae | 2772 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 2773 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
2774 | return -EINVAL; |
2775 | ||
cde8b15f | 2776 | if (tun_not_capable(tun)) |
2b980dbd | 2777 | return -EPERM; |
5dbbaf2d | 2778 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
2779 | if (err < 0) |
2780 | return err; | |
2781 | ||
94317099 | 2782 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER, |
af3fb24e | 2783 | ifr->ifr_flags & IFF_NAPI, |
77f22f92 | 2784 | ifr->ifr_flags & IFF_NAPI_FRAGS, true); |
a7385ba2 EB |
2785 | if (err < 0) |
2786 | return err; | |
4008e97f | 2787 | |
40630b82 | 2788 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
2789 | (tun->numqueues + tun->numdisabled > 1)) { |
2790 | /* One or more queue has already been attached, no need | |
2791 | * to initialize the device again. | |
2792 | */ | |
83c1f36f | 2793 | netdev_state_change(dev); |
e8dbad66 JW |
2794 | return 0; |
2795 | } | |
9fffc5c6 SD |
2796 | |
2797 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
2798 | (ifr->ifr_flags & TUN_FEATURES); | |
83c1f36f SD |
2799 | |
2800 | netdev_state_change(dev); | |
2801 | } else { | |
1da177e4 LT |
2802 | char *name; |
2803 | unsigned long flags = 0; | |
edfb6a14 JW |
2804 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
2805 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 2806 | |
c260b772 | 2807 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 2808 | return -EPERM; |
2b980dbd PM |
2809 | err = security_tun_dev_create(); |
2810 | if (err < 0) | |
2811 | return err; | |
ca6bb5d7 | 2812 | |
1da177e4 LT |
2813 | /* Set dev type */ |
2814 | if (ifr->ifr_flags & IFF_TUN) { | |
2815 | /* TUN device */ | |
40630b82 | 2816 | flags |= IFF_TUN; |
1da177e4 LT |
2817 | name = "tun%d"; |
2818 | } else if (ifr->ifr_flags & IFF_TAP) { | |
2819 | /* TAP device */ | |
40630b82 | 2820 | flags |= IFF_TAP; |
1da177e4 | 2821 | name = "tap%d"; |
6aa20a22 | 2822 | } else |
36989b90 | 2823 | return -EINVAL; |
6aa20a22 | 2824 | |
1da177e4 LT |
2825 | if (*ifr->ifr_name) |
2826 | name = ifr->ifr_name; | |
2827 | ||
c8d68e6b | 2828 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
2829 | NET_NAME_UNKNOWN, tun_setup, queues, |
2830 | queues); | |
edfb6a14 | 2831 | |
1da177e4 LT |
2832 | if (!dev) |
2833 | return -ENOMEM; | |
2834 | ||
fc54c658 | 2835 | dev_net_set(dev, net); |
f019a7a5 | 2836 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 2837 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 2838 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 2839 | |
1da177e4 LT |
2840 | tun = netdev_priv(dev); |
2841 | tun->dev = dev; | |
2842 | tun->flags = flags; | |
f271b2cc | 2843 | tun->txflt.count = 0; |
d9d52b51 | 2844 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 2845 | |
eaea34b2 | 2846 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
2847 | tun->filter_attached = false; |
2848 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 2849 | tun->rx_batched = 0; |
96f84061 | 2850 | RCU_INIT_POINTER(tun->steering_prog, NULL); |
33dccbb0 | 2851 | |
158b515f GK |
2852 | tun->ifr = ifr; |
2853 | tun->file = file; | |
608b9977 | 2854 | |
158b515f | 2855 | tun_net_initialize(dev); |
eb0fb363 | 2856 | |
1da177e4 | 2857 | err = register_netdevice(tun->dev); |
158b515f GK |
2858 | if (err < 0) { |
2859 | free_netdev(dev); | |
2860 | return err; | |
2861 | } | |
c2e315b8 | 2862 | /* free_netdev() won't check refcnt, to avoid race |
77f22f92 YY |
2863 | * with dev_put() we need publish tun after registration. |
2864 | */ | |
2865 | rcu_assign_pointer(tfile->tun, tun); | |
1da177e4 LT |
2866 | } |
2867 | ||
195624d9 PR |
2868 | if (ifr->ifr_flags & IFF_NO_CARRIER) |
2869 | netif_carrier_off(tun->dev); | |
2870 | else | |
2871 | netif_carrier_on(tun->dev); | |
af668b3c | 2872 | |
e35259a9 MK |
2873 | /* Make sure persistent devices do not get stuck in |
2874 | * xoff state. | |
2875 | */ | |
2876 | if (netif_running(tun->dev)) | |
c8d68e6b | 2877 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 2878 | |
1da177e4 LT |
2879 | strcpy(ifr->ifr_name, tun->dev->name); |
2880 | return 0; | |
1da177e4 LT |
2881 | } |
2882 | ||
12132768 | 2883 | static void tun_get_iff(struct tun_struct *tun, struct ifreq *ifr) |
e3b99556 | 2884 | { |
e3b99556 MM |
2885 | strcpy(ifr->ifr_name, tun->dev->name); |
2886 | ||
980c9e8c | 2887 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 2888 | |
e3b99556 MM |
2889 | } |
2890 | ||
5228ddc9 RR |
2891 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
2892 | * privs required. */ | |
88255375 | 2893 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 2894 | { |
c8f44aff | 2895 | netdev_features_t features = 0; |
5228ddc9 RR |
2896 | |
2897 | if (arg & TUN_F_CSUM) { | |
88255375 | 2898 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
2899 | arg &= ~TUN_F_CSUM; |
2900 | ||
2901 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
2902 | if (arg & TUN_F_TSO_ECN) { | |
2903 | features |= NETIF_F_TSO_ECN; | |
2904 | arg &= ~TUN_F_TSO_ECN; | |
2905 | } | |
2906 | if (arg & TUN_F_TSO4) | |
2907 | features |= NETIF_F_TSO; | |
2908 | if (arg & TUN_F_TSO6) | |
2909 | features |= NETIF_F_TSO6; | |
2910 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
2911 | } | |
0c19f846 WB |
2912 | |
2913 | arg &= ~TUN_F_UFO; | |
399e0827 AM |
2914 | |
2915 | /* TODO: for now USO4 and USO6 should work simultaneously */ | |
2916 | if (arg & TUN_F_USO4 && arg & TUN_F_USO6) { | |
2917 | features |= NETIF_F_GSO_UDP_L4; | |
2918 | arg &= ~(TUN_F_USO4 | TUN_F_USO6); | |
2919 | } | |
5228ddc9 RR |
2920 | } |
2921 | ||
2922 | /* This gives the user a way to test for new features in future by | |
2923 | * trying to set them. */ | |
2924 | if (arg) | |
2925 | return -EINVAL; | |
2926 | ||
88255375 | 2927 | tun->set_features = features; |
09050957 YI |
2928 | tun->dev->wanted_features &= ~TUN_USER_FEATURES; |
2929 | tun->dev->wanted_features |= features; | |
88255375 | 2930 | netdev_update_features(tun->dev); |
5228ddc9 RR |
2931 | |
2932 | return 0; | |
2933 | } | |
2934 | ||
c8d68e6b JW |
2935 | static void tun_detach_filter(struct tun_struct *tun, int n) |
2936 | { | |
2937 | int i; | |
2938 | struct tun_file *tfile; | |
2939 | ||
2940 | for (i = 0; i < n; i++) { | |
b8deabd3 | 2941 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2942 | lock_sock(tfile->socket.sk); |
2943 | sk_detach_filter(tfile->socket.sk); | |
2944 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2945 | } |
2946 | ||
2947 | tun->filter_attached = false; | |
2948 | } | |
2949 | ||
2950 | static int tun_attach_filter(struct tun_struct *tun) | |
2951 | { | |
2952 | int i, ret = 0; | |
2953 | struct tun_file *tfile; | |
2954 | ||
2955 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2956 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2957 | lock_sock(tfile->socket.sk); |
2958 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
2959 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2960 | if (ret) { |
2961 | tun_detach_filter(tun, i); | |
2962 | return ret; | |
2963 | } | |
2964 | } | |
2965 | ||
2966 | tun->filter_attached = true; | |
2967 | return ret; | |
2968 | } | |
2969 | ||
2970 | static void tun_set_sndbuf(struct tun_struct *tun) | |
2971 | { | |
2972 | struct tun_file *tfile; | |
2973 | int i; | |
2974 | ||
2975 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2976 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
2977 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
2978 | } | |
2979 | } | |
2980 | ||
cde8b15f JW |
2981 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
2982 | { | |
2983 | struct tun_file *tfile = file->private_data; | |
2984 | struct tun_struct *tun; | |
cde8b15f JW |
2985 | int ret = 0; |
2986 | ||
2987 | rtnl_lock(); | |
2988 | ||
2989 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 2990 | tun = tfile->detached; |
5dbbaf2d | 2991 | if (!tun) { |
cde8b15f | 2992 | ret = -EINVAL; |
5dbbaf2d PM |
2993 | goto unlock; |
2994 | } | |
2995 | ret = security_tun_dev_attach_queue(tun->security); | |
2996 | if (ret < 0) | |
2997 | goto unlock; | |
af3fb24e | 2998 | ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI, |
77f22f92 | 2999 | tun->flags & IFF_NAPI_FRAGS, true); |
4008e97f | 3000 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 3001 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 3002 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
3003 | ret = -EINVAL; |
3004 | else | |
3005 | __tun_detach(tfile, false); | |
3006 | } else | |
cde8b15f JW |
3007 | ret = -EINVAL; |
3008 | ||
83c1f36f SD |
3009 | if (ret >= 0) |
3010 | netdev_state_change(tun->dev); | |
3011 | ||
5dbbaf2d | 3012 | unlock: |
cde8b15f JW |
3013 | rtnl_unlock(); |
3014 | return ret; | |
3015 | } | |
3016 | ||
8f3f330d | 3017 | static int tun_set_ebpf(struct tun_struct *tun, struct tun_prog __rcu **prog_p, |
cd5681d7 | 3018 | void __user *data) |
96f84061 JW |
3019 | { |
3020 | struct bpf_prog *prog; | |
3021 | int fd; | |
3022 | ||
3023 | if (copy_from_user(&fd, data, sizeof(fd))) | |
3024 | return -EFAULT; | |
3025 | ||
3026 | if (fd == -1) { | |
3027 | prog = NULL; | |
3028 | } else { | |
3029 | prog = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); | |
3030 | if (IS_ERR(prog)) | |
3031 | return PTR_ERR(prog); | |
3032 | } | |
3033 | ||
cd5681d7 | 3034 | return __tun_set_ebpf(tun, prog_p, prog); |
96f84061 JW |
3035 | } |
3036 | ||
cca8ea3b PP |
3037 | /* Return correct value for tun->dev->addr_len based on tun->dev->type. */ |
3038 | static unsigned char tun_get_addr_len(unsigned short type) | |
3039 | { | |
3040 | switch (type) { | |
3041 | case ARPHRD_IP6GRE: | |
3042 | case ARPHRD_TUNNEL6: | |
3043 | return sizeof(struct in6_addr); | |
3044 | case ARPHRD_IPGRE: | |
3045 | case ARPHRD_TUNNEL: | |
3046 | case ARPHRD_SIT: | |
3047 | return 4; | |
3048 | case ARPHRD_ETHER: | |
3049 | return ETH_ALEN; | |
3050 | case ARPHRD_IEEE802154: | |
3051 | case ARPHRD_IEEE802154_MONITOR: | |
3052 | return IEEE802154_EXTENDED_ADDR_LEN; | |
3053 | case ARPHRD_PHONET_PIPE: | |
3054 | case ARPHRD_PPP: | |
3055 | case ARPHRD_NONE: | |
3056 | return 0; | |
3057 | case ARPHRD_6LOWPAN: | |
3058 | return EUI64_ADDR_LEN; | |
3059 | case ARPHRD_FDDI: | |
3060 | return FDDI_K_ALEN; | |
3061 | case ARPHRD_HIPPI: | |
3062 | return HIPPI_ALEN; | |
3063 | case ARPHRD_IEEE802: | |
3064 | return FC_ALEN; | |
3065 | case ARPHRD_ROSE: | |
3066 | return ROSE_ADDR_LEN; | |
3067 | case ARPHRD_NETROM: | |
3068 | return AX25_ADDR_LEN; | |
3069 | case ARPHRD_LOCALTLK: | |
3070 | return LTALK_ALEN; | |
3071 | default: | |
3072 | return 0; | |
3073 | } | |
3074 | } | |
3075 | ||
50857e2a AB |
3076 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
3077 | unsigned long arg, int ifreq_len) | |
1da177e4 | 3078 | { |
36b50bab | 3079 | struct tun_file *tfile = file->private_data; |
f663706a | 3080 | struct net *net = sock_net(&tfile->sk); |
631ab46b | 3081 | struct tun_struct *tun; |
1da177e4 | 3082 | void __user* argp = (void __user*)arg; |
cbfbfe3a | 3083 | unsigned int carrier; |
1da177e4 | 3084 | struct ifreq ifr; |
0625c883 EB |
3085 | kuid_t owner; |
3086 | kgid_t group; | |
cbfbfe3a | 3087 | int ifindex; |
33dccbb0 | 3088 | int sndbuf; |
d9d52b51 | 3089 | int vnet_hdr_sz; |
1cf8e410 | 3090 | int le; |
f271b2cc | 3091 | int ret; |
83c1f36f | 3092 | bool do_notify = false; |
1da177e4 | 3093 | |
f2780d6d KT |
3094 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || |
3095 | (_IOC_TYPE(cmd) == SOCK_IOC_TYPE && cmd != SIOCGSKNS)) { | |
50857e2a | 3096 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 3097 | return -EFAULT; |
8bbb1813 | 3098 | } else { |
a117dacd | 3099 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 3100 | } |
631ab46b EB |
3101 | if (cmd == TUNGETFEATURES) { |
3102 | /* Currently this just means: "what IFF flags are valid?". | |
3103 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
3104 | * TUNSETIFF. |
3105 | */ | |
195624d9 PR |
3106 | return put_user(IFF_TUN | IFF_TAP | IFF_NO_CARRIER | |
3107 | TUN_FEATURES, (unsigned int __user*)argp); | |
f663706a | 3108 | } else if (cmd == TUNSETQUEUE) { |
cde8b15f | 3109 | return tun_set_queue(file, &ifr); |
f663706a KT |
3110 | } else if (cmd == SIOCGSKNS) { |
3111 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
3112 | return -EPERM; | |
3113 | return open_related_ns(&net->ns, get_net_ns); | |
3114 | } | |
631ab46b | 3115 | |
876bfd4d HX |
3116 | rtnl_lock(); |
3117 | ||
9484dc74 | 3118 | tun = tun_get(tfile); |
0f16bc13 GF |
3119 | if (cmd == TUNSETIFF) { |
3120 | ret = -EEXIST; | |
3121 | if (tun) | |
3122 | goto unlock; | |
3123 | ||
1da177e4 LT |
3124 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
3125 | ||
f2780d6d | 3126 | ret = tun_set_iff(net, file, &ifr); |
1da177e4 | 3127 | |
876bfd4d HX |
3128 | if (ret) |
3129 | goto unlock; | |
1da177e4 | 3130 | |
50857e2a | 3131 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
3132 | ret = -EFAULT; |
3133 | goto unlock; | |
1da177e4 | 3134 | } |
fb7589a1 PE |
3135 | if (cmd == TUNSETIFINDEX) { |
3136 | ret = -EPERM; | |
3137 | if (tun) | |
3138 | goto unlock; | |
3139 | ||
3140 | ret = -EFAULT; | |
3141 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
3142 | goto unlock; | |
cbfbfe3a ED |
3143 | ret = -EINVAL; |
3144 | if (ifindex < 0) | |
3145 | goto unlock; | |
fb7589a1 PE |
3146 | ret = 0; |
3147 | tfile->ifindex = ifindex; | |
3148 | goto unlock; | |
3149 | } | |
1da177e4 | 3150 | |
876bfd4d | 3151 | ret = -EBADFD; |
1da177e4 | 3152 | if (!tun) |
876bfd4d | 3153 | goto unlock; |
1da177e4 | 3154 | |
3424170f | 3155 | netif_info(tun, drv, tun->dev, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 3156 | |
0c3e0e3b | 3157 | net = dev_net(tun->dev); |
631ab46b | 3158 | ret = 0; |
1da177e4 | 3159 | switch (cmd) { |
e3b99556 | 3160 | case TUNGETIFF: |
12132768 | 3161 | tun_get_iff(tun, &ifr); |
e3b99556 | 3162 | |
3d407a80 PE |
3163 | if (tfile->detached) |
3164 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
3165 | if (!tfile->socket.sk->sk_filter) |
3166 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 3167 | |
50857e2a | 3168 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 3169 | ret = -EFAULT; |
e3b99556 MM |
3170 | break; |
3171 | ||
1da177e4 LT |
3172 | case TUNSETNOCSUM: |
3173 | /* Disable/Enable checksum */ | |
1da177e4 | 3174 | |
88255375 | 3175 | /* [unimplemented] */ |
3424170f MK |
3176 | netif_info(tun, drv, tun->dev, "ignored: set checksum %s\n", |
3177 | arg ? "disabled" : "enabled"); | |
1da177e4 LT |
3178 | break; |
3179 | ||
3180 | case TUNSETPERSIST: | |
54f968d6 JW |
3181 | /* Disable/Enable persist mode. Keep an extra reference to the |
3182 | * module to prevent the module being unprobed. | |
3183 | */ | |
40630b82 MT |
3184 | if (arg && !(tun->flags & IFF_PERSIST)) { |
3185 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 3186 | __module_get(THIS_MODULE); |
83c1f36f | 3187 | do_notify = true; |
dd38bd85 | 3188 | } |
40630b82 MT |
3189 | if (!arg && (tun->flags & IFF_PERSIST)) { |
3190 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 | 3191 | module_put(THIS_MODULE); |
83c1f36f | 3192 | do_notify = true; |
54f968d6 | 3193 | } |
1da177e4 | 3194 | |
3424170f MK |
3195 | netif_info(tun, drv, tun->dev, "persist %s\n", |
3196 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
3197 | break; |
3198 | ||
3199 | case TUNSETOWNER: | |
3200 | /* Set owner of the device */ | |
0625c883 EB |
3201 | owner = make_kuid(current_user_ns(), arg); |
3202 | if (!uid_valid(owner)) { | |
3203 | ret = -EINVAL; | |
3204 | break; | |
3205 | } | |
3206 | tun->owner = owner; | |
83c1f36f | 3207 | do_notify = true; |
3424170f MK |
3208 | netif_info(tun, drv, tun->dev, "owner set to %u\n", |
3209 | from_kuid(&init_user_ns, tun->owner)); | |
1da177e4 LT |
3210 | break; |
3211 | ||
8c644623 GG |
3212 | case TUNSETGROUP: |
3213 | /* Set group of the device */ | |
0625c883 EB |
3214 | group = make_kgid(current_user_ns(), arg); |
3215 | if (!gid_valid(group)) { | |
3216 | ret = -EINVAL; | |
3217 | break; | |
3218 | } | |
3219 | tun->group = group; | |
83c1f36f | 3220 | do_notify = true; |
3424170f MK |
3221 | netif_info(tun, drv, tun->dev, "group set to %u\n", |
3222 | from_kgid(&init_user_ns, tun->group)); | |
8c644623 GG |
3223 | break; |
3224 | ||
ff4cc3ac MK |
3225 | case TUNSETLINK: |
3226 | /* Only allow setting the type when the interface is down */ | |
3227 | if (tun->dev->flags & IFF_UP) { | |
3424170f MK |
3228 | netif_info(tun, drv, tun->dev, |
3229 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 3230 | ret = -EBUSY; |
ff4cc3ac | 3231 | } else { |
8e1e33ff MS |
3232 | ret = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
3233 | tun->dev); | |
3234 | ret = notifier_to_errno(ret); | |
3235 | if (ret) { | |
3236 | netif_info(tun, drv, tun->dev, | |
3237 | "Refused to change device type\n"); | |
3238 | break; | |
3239 | } | |
ff4cc3ac | 3240 | tun->dev->type = (int) arg; |
cca8ea3b | 3241 | tun->dev->addr_len = tun_get_addr_len(tun->dev->type); |
3424170f MK |
3242 | netif_info(tun, drv, tun->dev, "linktype set to %d\n", |
3243 | tun->dev->type); | |
8e1e33ff MS |
3244 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
3245 | tun->dev); | |
ff4cc3ac | 3246 | } |
631ab46b | 3247 | break; |
ff4cc3ac | 3248 | |
1da177e4 | 3249 | case TUNSETDEBUG: |
3424170f | 3250 | tun->msg_enable = (u32)arg; |
1da177e4 | 3251 | break; |
3424170f | 3252 | |
5228ddc9 | 3253 | case TUNSETOFFLOAD: |
88255375 | 3254 | ret = set_offload(tun, arg); |
631ab46b | 3255 | break; |
5228ddc9 | 3256 | |
f271b2cc MK |
3257 | case TUNSETTXFILTER: |
3258 | /* Can be set only for TAPs */ | |
631ab46b | 3259 | ret = -EINVAL; |
40630b82 | 3260 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 3261 | break; |
c0e5a8c2 | 3262 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 3263 | break; |
1da177e4 LT |
3264 | |
3265 | case SIOCGIFHWADDR: | |
b595076a | 3266 | /* Get hw address */ |
3b23a32a | 3267 | dev_get_mac_address(&ifr.ifr_hwaddr, net, tun->dev->name); |
50857e2a | 3268 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
3269 | ret = -EFAULT; |
3270 | break; | |
1da177e4 LT |
3271 | |
3272 | case SIOCSIFHWADDR: | |
f271b2cc | 3273 | /* Set hw address */ |
3b23a32a | 3274 | ret = dev_set_mac_address_user(tun->dev, &ifr.ifr_hwaddr, NULL); |
631ab46b | 3275 | break; |
33dccbb0 HX |
3276 | |
3277 | case TUNGETSNDBUF: | |
54f968d6 | 3278 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
3279 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
3280 | ret = -EFAULT; | |
3281 | break; | |
3282 | ||
3283 | case TUNSETSNDBUF: | |
3284 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
3285 | ret = -EFAULT; | |
3286 | break; | |
3287 | } | |
93161922 CG |
3288 | if (sndbuf <= 0) { |
3289 | ret = -EINVAL; | |
3290 | break; | |
3291 | } | |
33dccbb0 | 3292 | |
c8d68e6b JW |
3293 | tun->sndbuf = sndbuf; |
3294 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
3295 | break; |
3296 | ||
d9d52b51 MT |
3297 | case TUNGETVNETHDRSZ: |
3298 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
3299 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
3300 | ret = -EFAULT; | |
3301 | break; | |
3302 | ||
3303 | case TUNSETVNETHDRSZ: | |
3304 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
3305 | ret = -EFAULT; | |
3306 | break; | |
3307 | } | |
3308 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
3309 | ret = -EINVAL; | |
3310 | break; | |
3311 | } | |
3312 | ||
3313 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
3314 | break; | |
3315 | ||
1cf8e410 MT |
3316 | case TUNGETVNETLE: |
3317 | le = !!(tun->flags & TUN_VNET_LE); | |
3318 | if (put_user(le, (int __user *)argp)) | |
3319 | ret = -EFAULT; | |
3320 | break; | |
3321 | ||
3322 | case TUNSETVNETLE: | |
3323 | if (get_user(le, (int __user *)argp)) { | |
3324 | ret = -EFAULT; | |
3325 | break; | |
3326 | } | |
3327 | if (le) | |
3328 | tun->flags |= TUN_VNET_LE; | |
3329 | else | |
3330 | tun->flags &= ~TUN_VNET_LE; | |
3331 | break; | |
3332 | ||
8b8e658b GK |
3333 | case TUNGETVNETBE: |
3334 | ret = tun_get_vnet_be(tun, argp); | |
3335 | break; | |
3336 | ||
3337 | case TUNSETVNETBE: | |
3338 | ret = tun_set_vnet_be(tun, argp); | |
3339 | break; | |
3340 | ||
99405162 MT |
3341 | case TUNATTACHFILTER: |
3342 | /* Can be set only for TAPs */ | |
3343 | ret = -EINVAL; | |
40630b82 | 3344 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
3345 | break; |
3346 | ret = -EFAULT; | |
54f968d6 | 3347 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
3348 | break; |
3349 | ||
c8d68e6b | 3350 | ret = tun_attach_filter(tun); |
99405162 MT |
3351 | break; |
3352 | ||
3353 | case TUNDETACHFILTER: | |
3354 | /* Can be set only for TAPs */ | |
3355 | ret = -EINVAL; | |
40630b82 | 3356 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 3357 | break; |
c8d68e6b JW |
3358 | ret = 0; |
3359 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
3360 | break; |
3361 | ||
76975e9c PE |
3362 | case TUNGETFILTER: |
3363 | ret = -EINVAL; | |
40630b82 | 3364 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
3365 | break; |
3366 | ret = -EFAULT; | |
3367 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
3368 | break; | |
3369 | ret = 0; | |
3370 | break; | |
3371 | ||
96f84061 | 3372 | case TUNSETSTEERINGEBPF: |
cd5681d7 | 3373 | ret = tun_set_ebpf(tun, &tun->steering_prog, argp); |
96f84061 JW |
3374 | break; |
3375 | ||
aff3d70a JW |
3376 | case TUNSETFILTEREBPF: |
3377 | ret = tun_set_ebpf(tun, &tun->filter_prog, argp); | |
3378 | break; | |
3379 | ||
26d31925 ND |
3380 | case TUNSETCARRIER: |
3381 | ret = -EFAULT; | |
3382 | if (copy_from_user(&carrier, argp, sizeof(carrier))) | |
3383 | goto unlock; | |
3384 | ||
3385 | ret = tun_net_change_carrier(tun->dev, (bool)carrier); | |
3386 | break; | |
3387 | ||
0c3e0e3b KT |
3388 | case TUNGETDEVNETNS: |
3389 | ret = -EPERM; | |
3390 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
3391 | goto unlock; | |
3392 | ret = open_related_ns(&net->ns, get_net_ns); | |
3393 | break; | |
3394 | ||
1da177e4 | 3395 | default: |
631ab46b EB |
3396 | ret = -EINVAL; |
3397 | break; | |
ee289b64 | 3398 | } |
1da177e4 | 3399 | |
83c1f36f SD |
3400 | if (do_notify) |
3401 | netdev_state_change(tun->dev); | |
3402 | ||
876bfd4d HX |
3403 | unlock: |
3404 | rtnl_unlock(); | |
3405 | if (tun) | |
3406 | tun_put(tun); | |
631ab46b | 3407 | return ret; |
1da177e4 LT |
3408 | } |
3409 | ||
50857e2a AB |
3410 | static long tun_chr_ioctl(struct file *file, |
3411 | unsigned int cmd, unsigned long arg) | |
3412 | { | |
3413 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
3414 | } | |
3415 | ||
3416 | #ifdef CONFIG_COMPAT | |
3417 | static long tun_chr_compat_ioctl(struct file *file, | |
3418 | unsigned int cmd, unsigned long arg) | |
3419 | { | |
3420 | switch (cmd) { | |
3421 | case TUNSETIFF: | |
3422 | case TUNGETIFF: | |
3423 | case TUNSETTXFILTER: | |
3424 | case TUNGETSNDBUF: | |
3425 | case TUNSETSNDBUF: | |
3426 | case SIOCGIFHWADDR: | |
3427 | case SIOCSIFHWADDR: | |
3428 | arg = (unsigned long)compat_ptr(arg); | |
3429 | break; | |
3430 | default: | |
3431 | arg = (compat_ulong_t)arg; | |
3432 | break; | |
3433 | } | |
3434 | ||
3435 | /* | |
3436 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
3437 | * the end of that structure. All fields that are used in this | |
3438 | * driver are compatible though, we don't need to convert the | |
3439 | * contents. | |
3440 | */ | |
3441 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
3442 | } | |
3443 | #endif /* CONFIG_COMPAT */ | |
3444 | ||
1da177e4 LT |
3445 | static int tun_chr_fasync(int fd, struct file *file, int on) |
3446 | { | |
54f968d6 | 3447 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
3448 | int ret; |
3449 | ||
54f968d6 | 3450 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 3451 | goto out; |
6aa20a22 | 3452 | |
1da177e4 | 3453 | if (on) { |
01919134 | 3454 | __f_setown(file, task_pid(current), PIDTYPE_TGID, 0); |
54f968d6 | 3455 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 3456 | } else |
54f968d6 | 3457 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
3458 | ret = 0; |
3459 | out: | |
9d319522 | 3460 | return ret; |
1da177e4 LT |
3461 | } |
3462 | ||
3463 | static int tun_chr_open(struct inode *inode, struct file * file) | |
3464 | { | |
140e807d | 3465 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 3466 | struct tun_file *tfile; |
deed49fb | 3467 | |
140e807d | 3468 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 3469 | &tun_proto, 0); |
631ab46b EB |
3470 | if (!tfile) |
3471 | return -ENOMEM; | |
b196d88a JW |
3472 | if (ptr_ring_init(&tfile->tx_ring, 0, GFP_KERNEL)) { |
3473 | sk_free(&tfile->sk); | |
3474 | return -ENOMEM; | |
3475 | } | |
3476 | ||
c7256f57 | 3477 | mutex_init(&tfile->napi_mutex); |
c956674b | 3478 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 3479 | tfile->flags = 0; |
fb7589a1 | 3480 | tfile->ifindex = 0; |
54f968d6 | 3481 | |
333f7909 | 3482 | init_waitqueue_head(&tfile->socket.wq.wait); |
54f968d6 JW |
3483 | |
3484 | tfile->socket.file = file; | |
3485 | tfile->socket.ops = &tun_socket_ops; | |
3486 | ||
9bc30473 | 3487 | sock_init_data_uid(&tfile->socket, &tfile->sk, current_fsuid()); |
54f968d6 JW |
3488 | |
3489 | tfile->sk.sk_write_space = tun_sock_write_space; | |
3490 | tfile->sk.sk_sndbuf = INT_MAX; | |
3491 | ||
631ab46b | 3492 | file->private_data = tfile; |
4008e97f | 3493 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 3494 | |
19a6afb2 JW |
3495 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
3496 | ||
438b4060 JA |
3497 | /* tun groks IOCB_NOWAIT just fine, mark it as such */ |
3498 | file->f_mode |= FMODE_NOWAIT; | |
1da177e4 LT |
3499 | return 0; |
3500 | } | |
3501 | ||
3502 | static int tun_chr_close(struct inode *inode, struct file *file) | |
3503 | { | |
631ab46b | 3504 | struct tun_file *tfile = file->private_data; |
1da177e4 | 3505 | |
c8d68e6b | 3506 | tun_detach(tfile, true); |
1da177e4 LT |
3507 | |
3508 | return 0; | |
3509 | } | |
3510 | ||
93e14b6d | 3511 | #ifdef CONFIG_PROC_FS |
9484dc74 | 3512 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file) |
93e14b6d | 3513 | { |
9484dc74 | 3514 | struct tun_file *tfile = file->private_data; |
93e14b6d MY |
3515 | struct tun_struct *tun; |
3516 | struct ifreq ifr; | |
3517 | ||
3518 | memset(&ifr, 0, sizeof(ifr)); | |
3519 | ||
3520 | rtnl_lock(); | |
9484dc74 | 3521 | tun = tun_get(tfile); |
93e14b6d | 3522 | if (tun) |
12132768 | 3523 | tun_get_iff(tun, &ifr); |
93e14b6d MY |
3524 | rtnl_unlock(); |
3525 | ||
3526 | if (tun) | |
3527 | tun_put(tun); | |
3528 | ||
a3816ab0 | 3529 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
3530 | } |
3531 | #endif | |
3532 | ||
d54b1fdb | 3533 | static const struct file_operations tun_fops = { |
6aa20a22 | 3534 | .owner = THIS_MODULE, |
1da177e4 | 3535 | .llseek = no_llseek, |
9b067034 | 3536 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 3537 | .write_iter = tun_chr_write_iter, |
1da177e4 | 3538 | .poll = tun_chr_poll, |
50857e2a AB |
3539 | .unlocked_ioctl = tun_chr_ioctl, |
3540 | #ifdef CONFIG_COMPAT | |
3541 | .compat_ioctl = tun_chr_compat_ioctl, | |
3542 | #endif | |
1da177e4 LT |
3543 | .open = tun_chr_open, |
3544 | .release = tun_chr_close, | |
93e14b6d MY |
3545 | .fasync = tun_chr_fasync, |
3546 | #ifdef CONFIG_PROC_FS | |
3547 | .show_fdinfo = tun_chr_show_fdinfo, | |
3548 | #endif | |
1da177e4 LT |
3549 | }; |
3550 | ||
3551 | static struct miscdevice tun_miscdev = { | |
3552 | .minor = TUN_MINOR, | |
3553 | .name = "tun", | |
e454cea2 | 3554 | .nodename = "net/tun", |
1da177e4 | 3555 | .fops = &tun_fops, |
1da177e4 LT |
3556 | }; |
3557 | ||
3558 | /* ethtool interface */ | |
3559 | ||
4e24f2dd CW |
3560 | static void tun_default_link_ksettings(struct net_device *dev, |
3561 | struct ethtool_link_ksettings *cmd) | |
29ccc49d PR |
3562 | { |
3563 | ethtool_link_ksettings_zero_link_mode(cmd, supported); | |
3564 | ethtool_link_ksettings_zero_link_mode(cmd, advertising); | |
598d2982 | 3565 | cmd->base.speed = SPEED_10000; |
29ccc49d PR |
3566 | cmd->base.duplex = DUPLEX_FULL; |
3567 | cmd->base.port = PORT_TP; | |
3568 | cmd->base.phy_address = 0; | |
3569 | cmd->base.autoneg = AUTONEG_DISABLE; | |
4e24f2dd CW |
3570 | } |
3571 | ||
3572 | static int tun_get_link_ksettings(struct net_device *dev, | |
3573 | struct ethtool_link_ksettings *cmd) | |
3574 | { | |
3575 | struct tun_struct *tun = netdev_priv(dev); | |
3576 | ||
3577 | memcpy(cmd, &tun->link_ksettings, sizeof(*cmd)); | |
3578 | return 0; | |
3579 | } | |
3580 | ||
3581 | static int tun_set_link_ksettings(struct net_device *dev, | |
3582 | const struct ethtool_link_ksettings *cmd) | |
3583 | { | |
3584 | struct tun_struct *tun = netdev_priv(dev); | |
3585 | ||
3586 | memcpy(&tun->link_ksettings, cmd, sizeof(*cmd)); | |
1da177e4 LT |
3587 | return 0; |
3588 | } | |
3589 | ||
3590 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
3591 | { | |
3592 | struct tun_struct *tun = netdev_priv(dev); | |
3593 | ||
fb3ceec1 WS |
3594 | strscpy(info->driver, DRV_NAME, sizeof(info->driver)); |
3595 | strscpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
3596 | |
3597 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 3598 | case IFF_TUN: |
fb3ceec1 | 3599 | strscpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 3600 | break; |
40630b82 | 3601 | case IFF_TAP: |
fb3ceec1 | 3602 | strscpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
3603 | break; |
3604 | } | |
3605 | } | |
3606 | ||
3607 | static u32 tun_get_msglevel(struct net_device *dev) | |
3608 | { | |
1da177e4 | 3609 | struct tun_struct *tun = netdev_priv(dev); |
3424170f MK |
3610 | |
3611 | return tun->msg_enable; | |
1da177e4 LT |
3612 | } |
3613 | ||
3614 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
3615 | { | |
1da177e4 | 3616 | struct tun_struct *tun = netdev_priv(dev); |
3424170f MK |
3617 | |
3618 | tun->msg_enable = value; | |
1da177e4 LT |
3619 | } |
3620 | ||
5503fcec | 3621 | static int tun_get_coalesce(struct net_device *dev, |
f3ccfda1 YM |
3622 | struct ethtool_coalesce *ec, |
3623 | struct kernel_ethtool_coalesce *kernel_coal, | |
3624 | struct netlink_ext_ack *extack) | |
5503fcec JW |
3625 | { |
3626 | struct tun_struct *tun = netdev_priv(dev); | |
3627 | ||
3628 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
3629 | ||
3630 | return 0; | |
3631 | } | |
3632 | ||
3633 | static int tun_set_coalesce(struct net_device *dev, | |
f3ccfda1 YM |
3634 | struct ethtool_coalesce *ec, |
3635 | struct kernel_ethtool_coalesce *kernel_coal, | |
3636 | struct netlink_ext_ack *extack) | |
5503fcec JW |
3637 | { |
3638 | struct tun_struct *tun = netdev_priv(dev); | |
3639 | ||
3640 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
3641 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
3642 | else | |
3643 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
3644 | ||
3645 | return 0; | |
3646 | } | |
3647 | ||
7282d491 | 3648 | static const struct ethtool_ops tun_ethtool_ops = { |
e5ad00b3 | 3649 | .supported_coalesce_params = ETHTOOL_COALESCE_RX_MAX_FRAMES, |
1da177e4 LT |
3650 | .get_drvinfo = tun_get_drvinfo, |
3651 | .get_msglevel = tun_get_msglevel, | |
3652 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 3653 | .get_link = ethtool_op_get_link, |
eda29772 | 3654 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
3655 | .get_coalesce = tun_get_coalesce, |
3656 | .set_coalesce = tun_set_coalesce, | |
29ccc49d | 3657 | .get_link_ksettings = tun_get_link_ksettings, |
4e24f2dd | 3658 | .set_link_ksettings = tun_set_link_ksettings, |
1da177e4 LT |
3659 | }; |
3660 | ||
1576d986 JW |
3661 | static int tun_queue_resize(struct tun_struct *tun) |
3662 | { | |
3663 | struct net_device *dev = tun->dev; | |
3664 | struct tun_file *tfile; | |
5990a305 | 3665 | struct ptr_ring **rings; |
1576d986 JW |
3666 | int n = tun->numqueues + tun->numdisabled; |
3667 | int ret, i; | |
3668 | ||
5990a305 JW |
3669 | rings = kmalloc_array(n, sizeof(*rings), GFP_KERNEL); |
3670 | if (!rings) | |
1576d986 JW |
3671 | return -ENOMEM; |
3672 | ||
3673 | for (i = 0; i < tun->numqueues; i++) { | |
3674 | tfile = rtnl_dereference(tun->tfiles[i]); | |
5990a305 | 3675 | rings[i] = &tfile->tx_ring; |
1576d986 JW |
3676 | } |
3677 | list_for_each_entry(tfile, &tun->disabled, next) | |
5990a305 | 3678 | rings[i++] = &tfile->tx_ring; |
1576d986 | 3679 | |
5990a305 JW |
3680 | ret = ptr_ring_resize_multiple(rings, n, |
3681 | dev->tx_queue_len, GFP_KERNEL, | |
fc72d1d5 | 3682 | tun_ptr_free); |
1576d986 | 3683 | |
5990a305 | 3684 | kfree(rings); |
1576d986 JW |
3685 | return ret; |
3686 | } | |
3687 | ||
3688 | static int tun_device_event(struct notifier_block *unused, | |
3689 | unsigned long event, void *ptr) | |
3690 | { | |
3691 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
3692 | struct tun_struct *tun = netdev_priv(dev); | |
72b319dc | 3693 | int i; |
1576d986 | 3694 | |
86dfb4ac CG |
3695 | if (dev->rtnl_link_ops != &tun_link_ops) |
3696 | return NOTIFY_DONE; | |
3697 | ||
1576d986 JW |
3698 | switch (event) { |
3699 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
3700 | if (tun_queue_resize(tun)) | |
3701 | return NOTIFY_BAD; | |
3702 | break; | |
72b319dc FL |
3703 | case NETDEV_UP: |
3704 | for (i = 0; i < tun->numqueues; i++) { | |
3705 | struct tun_file *tfile; | |
3706 | ||
3707 | tfile = rtnl_dereference(tun->tfiles[i]); | |
3708 | tfile->socket.sk->sk_write_space(tfile->socket.sk); | |
3709 | } | |
3710 | break; | |
1576d986 JW |
3711 | default: |
3712 | break; | |
3713 | } | |
3714 | ||
3715 | return NOTIFY_DONE; | |
3716 | } | |
3717 | ||
3718 | static struct notifier_block tun_notifier_block __read_mostly = { | |
3719 | .notifier_call = tun_device_event, | |
3720 | }; | |
79d17604 | 3721 | |
1da177e4 LT |
3722 | static int __init tun_init(void) |
3723 | { | |
3724 | int ret = 0; | |
3725 | ||
6b8a66ee | 3726 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
1da177e4 | 3727 | |
f019a7a5 | 3728 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 3729 | if (ret) { |
6b8a66ee | 3730 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 3731 | goto err_linkops; |
79d17604 PE |
3732 | } |
3733 | ||
1da177e4 | 3734 | ret = misc_register(&tun_miscdev); |
79d17604 | 3735 | if (ret) { |
6b8a66ee | 3736 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
3737 | goto err_misc; |
3738 | } | |
1576d986 | 3739 | |
5edfbd3c TZ |
3740 | ret = register_netdevice_notifier(&tun_notifier_block); |
3741 | if (ret) { | |
3742 | pr_err("Can't register netdevice notifier\n"); | |
3743 | goto err_notifier; | |
3744 | } | |
3745 | ||
f019a7a5 | 3746 | return 0; |
5edfbd3c TZ |
3747 | |
3748 | err_notifier: | |
3749 | misc_deregister(&tun_miscdev); | |
79d17604 | 3750 | err_misc: |
f019a7a5 EB |
3751 | rtnl_link_unregister(&tun_link_ops); |
3752 | err_linkops: | |
1da177e4 LT |
3753 | return ret; |
3754 | } | |
3755 | ||
b2f83233 | 3756 | static void __exit tun_cleanup(void) |
1da177e4 | 3757 | { |
6aa20a22 | 3758 | misc_deregister(&tun_miscdev); |
f019a7a5 | 3759 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 3760 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
3761 | } |
3762 | ||
05c2828c MT |
3763 | /* Get an underlying socket object from tun file. Returns error unless file is |
3764 | * attached to a device. The returned object works like a packet socket, it | |
3765 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
3766 | * holding a reference to the file for as long as the socket is in use. */ | |
3767 | struct socket *tun_get_socket(struct file *file) | |
3768 | { | |
6e914fc7 | 3769 | struct tun_file *tfile; |
05c2828c MT |
3770 | if (file->f_op != &tun_fops) |
3771 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
3772 | tfile = file->private_data; |
3773 | if (!tfile) | |
05c2828c | 3774 | return ERR_PTR(-EBADFD); |
54f968d6 | 3775 | return &tfile->socket; |
05c2828c MT |
3776 | } |
3777 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
3778 | ||
5990a305 | 3779 | struct ptr_ring *tun_get_tx_ring(struct file *file) |
83339c6b JW |
3780 | { |
3781 | struct tun_file *tfile; | |
3782 | ||
3783 | if (file->f_op != &tun_fops) | |
3784 | return ERR_PTR(-EINVAL); | |
3785 | tfile = file->private_data; | |
3786 | if (!tfile) | |
3787 | return ERR_PTR(-EBADFD); | |
5990a305 | 3788 | return &tfile->tx_ring; |
83339c6b | 3789 | } |
5990a305 | 3790 | EXPORT_SYMBOL_GPL(tun_get_tx_ring); |
83339c6b | 3791 | |
1da177e4 LT |
3792 | module_init(tun_init); |
3793 | module_exit(tun_cleanup); | |
3794 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
3795 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
3796 | MODULE_LICENSE("GPL"); | |
3797 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 3798 | MODULE_ALIAS("devname:net/tun"); |