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| \ |
d591a1f3 | 188 | NETIF_F_TSO6) |
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 | ||
06e55add | 276 | static void tun_napi_disable(struct tun_file *tfile) |
94317099 | 277 | { |
aec72f33 | 278 | if (tfile->napi_enabled) |
94317099 PP |
279 | napi_disable(&tfile->napi); |
280 | } | |
281 | ||
06e55add | 282 | static void tun_napi_del(struct tun_file *tfile) |
94317099 | 283 | { |
aec72f33 | 284 | if (tfile->napi_enabled) |
94317099 PP |
285 | netif_napi_del(&tfile->napi); |
286 | } | |
287 | ||
af3fb24e | 288 | static bool tun_napi_frags_enabled(const struct tun_file *tfile) |
90e33d45 | 289 | { |
af3fb24e | 290 | return tfile->napi_frags_enabled; |
90e33d45 PP |
291 | } |
292 | ||
8b8e658b GK |
293 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
294 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
295 | { | |
296 | return tun->flags & TUN_VNET_BE ? false : | |
297 | virtio_legacy_is_little_endian(); | |
298 | } | |
299 | ||
300 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
301 | { | |
302 | int be = !!(tun->flags & TUN_VNET_BE); | |
303 | ||
304 | if (put_user(be, argp)) | |
305 | return -EFAULT; | |
306 | ||
307 | return 0; | |
308 | } | |
309 | ||
310 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
311 | { | |
312 | int be; | |
313 | ||
314 | if (get_user(be, argp)) | |
315 | return -EFAULT; | |
316 | ||
317 | if (be) | |
318 | tun->flags |= TUN_VNET_BE; | |
319 | else | |
320 | tun->flags &= ~TUN_VNET_BE; | |
321 | ||
322 | return 0; | |
323 | } | |
324 | #else | |
325 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
326 | { | |
327 | return virtio_legacy_is_little_endian(); | |
328 | } | |
329 | ||
330 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
331 | { | |
332 | return -EINVAL; | |
333 | } | |
334 | ||
335 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
336 | { | |
337 | return -EINVAL; | |
338 | } | |
339 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
340 | ||
25bd55bb GK |
341 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
342 | { | |
7d824109 | 343 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 344 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
345 | } |
346 | ||
56f0dcc5 MT |
347 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
348 | { | |
25bd55bb | 349 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
350 | } |
351 | ||
352 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
353 | { | |
25bd55bb | 354 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
355 | } |
356 | ||
96442e42 JW |
357 | static inline u32 tun_hashfn(u32 rxhash) |
358 | { | |
f13b5468 | 359 | return rxhash & TUN_MASK_FLOW_ENTRIES; |
96442e42 JW |
360 | } |
361 | ||
362 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
363 | { | |
364 | struct tun_flow_entry *e; | |
96442e42 | 365 | |
b67bfe0d | 366 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
367 | if (e->rxhash == rxhash) |
368 | return e; | |
369 | } | |
370 | return NULL; | |
371 | } | |
372 | ||
373 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
374 | struct hlist_head *head, | |
375 | u32 rxhash, u16 queue_index) | |
376 | { | |
9fdc6bef ED |
377 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
378 | ||
96442e42 | 379 | if (e) { |
3424170f MK |
380 | netif_info(tun, tx_queued, tun->dev, |
381 | "create flow: hash %u index %u\n", | |
382 | rxhash, queue_index); | |
96442e42 JW |
383 | e->updated = jiffies; |
384 | e->rxhash = rxhash; | |
9bc88939 | 385 | e->rps_rxhash = 0; |
96442e42 JW |
386 | e->queue_index = queue_index; |
387 | e->tun = tun; | |
388 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 389 | ++tun->flow_count; |
96442e42 JW |
390 | } |
391 | return e; | |
392 | } | |
393 | ||
96442e42 JW |
394 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
395 | { | |
3424170f MK |
396 | netif_info(tun, tx_queued, tun->dev, "delete flow: hash %u index %u\n", |
397 | e->rxhash, e->queue_index); | |
96442e42 | 398 | hlist_del_rcu(&e->hash_link); |
9fdc6bef | 399 | kfree_rcu(e, rcu); |
b8732fb7 | 400 | --tun->flow_count; |
96442e42 JW |
401 | } |
402 | ||
403 | static void tun_flow_flush(struct tun_struct *tun) | |
404 | { | |
405 | int i; | |
406 | ||
407 | spin_lock_bh(&tun->lock); | |
408 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
409 | struct tun_flow_entry *e; | |
b67bfe0d | 410 | struct hlist_node *n; |
96442e42 | 411 | |
b67bfe0d | 412 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
413 | tun_flow_delete(tun, e); |
414 | } | |
415 | spin_unlock_bh(&tun->lock); | |
416 | } | |
417 | ||
418 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
419 | { | |
420 | int i; | |
421 | ||
422 | spin_lock_bh(&tun->lock); | |
423 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
424 | struct tun_flow_entry *e; | |
b67bfe0d | 425 | struct hlist_node *n; |
96442e42 | 426 | |
b67bfe0d | 427 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
428 | if (e->queue_index == queue_index) |
429 | tun_flow_delete(tun, e); | |
430 | } | |
431 | } | |
432 | spin_unlock_bh(&tun->lock); | |
433 | } | |
434 | ||
e99e88a9 | 435 | static void tun_flow_cleanup(struct timer_list *t) |
96442e42 | 436 | { |
e99e88a9 | 437 | struct tun_struct *tun = from_timer(tun, t, flow_gc_timer); |
96442e42 JW |
438 | unsigned long delay = tun->ageing_time; |
439 | unsigned long next_timer = jiffies + delay; | |
440 | unsigned long count = 0; | |
441 | int i; | |
442 | ||
7dbfb4ef | 443 | spin_lock(&tun->lock); |
96442e42 JW |
444 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { |
445 | struct tun_flow_entry *e; | |
b67bfe0d | 446 | struct hlist_node *n; |
96442e42 | 447 | |
b67bfe0d | 448 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 | 449 | unsigned long this_timer; |
81d98fa4 | 450 | |
96442e42 | 451 | this_timer = e->updated + delay; |
81d98fa4 | 452 | if (time_before_eq(this_timer, jiffies)) { |
96442e42 | 453 | tun_flow_delete(tun, e); |
81d98fa4 ED |
454 | continue; |
455 | } | |
456 | count++; | |
457 | if (time_before(this_timer, next_timer)) | |
96442e42 JW |
458 | next_timer = this_timer; |
459 | } | |
460 | } | |
461 | ||
462 | if (count) | |
463 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
7dbfb4ef | 464 | spin_unlock(&tun->lock); |
96442e42 JW |
465 | } |
466 | ||
49974420 | 467 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 468 | struct tun_file *tfile) |
96442e42 JW |
469 | { |
470 | struct hlist_head *head; | |
471 | struct tun_flow_entry *e; | |
472 | unsigned long delay = tun->ageing_time; | |
9e85722d | 473 | u16 queue_index = tfile->queue_index; |
96442e42 | 474 | |
5c327f67 | 475 | head = &tun->flows[tun_hashfn(rxhash)]; |
96442e42 JW |
476 | |
477 | rcu_read_lock(); | |
478 | ||
96442e42 JW |
479 | e = tun_flow_find(head, rxhash); |
480 | if (likely(e)) { | |
481 | /* TODO: keep queueing to old queue until it's empty? */ | |
4ffdd22e ED |
482 | if (READ_ONCE(e->queue_index) != queue_index) |
483 | WRITE_ONCE(e->queue_index, queue_index); | |
83b1bc12 LR |
484 | if (e->updated != jiffies) |
485 | e->updated = jiffies; | |
9bc88939 | 486 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
487 | } else { |
488 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
489 | if (!tun_flow_find(head, rxhash) && |
490 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
491 | tun_flow_create(tun, head, rxhash, queue_index); |
492 | ||
493 | if (!timer_pending(&tun->flow_gc_timer)) | |
494 | mod_timer(&tun->flow_gc_timer, | |
495 | round_jiffies_up(jiffies + delay)); | |
496 | spin_unlock_bh(&tun->lock); | |
497 | } | |
498 | ||
96442e42 JW |
499 | rcu_read_unlock(); |
500 | } | |
501 | ||
516c512b | 502 | /* Save the hash received in the stack receive path and update the |
9bc88939 TH |
503 | * flow_hash table accordingly. |
504 | */ | |
505 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
506 | { | |
567e4b79 | 507 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 508 | e->rps_rxhash = hash; |
9bc88939 TH |
509 | } |
510 | ||
4b035271 | 511 | /* We try to identify a flow through its rxhash. The reason that |
92d4ea6e | 512 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
513 | * the rxq based on the txq where the last packet of the flow comes. As |
514 | * the userspace application move between processors, we may get a | |
4b035271 | 515 | * different rxq no. here. |
c8d68e6b | 516 | */ |
96f84061 | 517 | static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
c8d68e6b | 518 | { |
96442e42 | 519 | struct tun_flow_entry *e; |
c8d68e6b JW |
520 | u32 txq = 0; |
521 | u32 numqueues = 0; | |
522 | ||
6aa7de05 | 523 | numqueues = READ_ONCE(tun->numqueues); |
c8d68e6b | 524 | |
feec084a | 525 | txq = __skb_get_hash_symmetric(skb); |
4b035271 WL |
526 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
527 | if (e) { | |
528 | tun_flow_save_rps_rxhash(e, txq); | |
529 | txq = e->queue_index; | |
530 | } else { | |
531 | /* use multiply and shift instead of expensive divide */ | |
532 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
533 | } |
534 | ||
c8d68e6b JW |
535 | return txq; |
536 | } | |
537 | ||
96f84061 JW |
538 | static u16 tun_ebpf_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
539 | { | |
cd5681d7 | 540 | struct tun_prog *prog; |
a35d310f | 541 | u32 numqueues; |
96f84061 JW |
542 | u16 ret = 0; |
543 | ||
a35d310f JW |
544 | numqueues = READ_ONCE(tun->numqueues); |
545 | if (!numqueues) | |
546 | return 0; | |
547 | ||
96f84061 JW |
548 | prog = rcu_dereference(tun->steering_prog); |
549 | if (prog) | |
550 | ret = bpf_prog_run_clear_cb(prog->prog, skb); | |
551 | ||
a35d310f | 552 | return ret % numqueues; |
96f84061 JW |
553 | } |
554 | ||
555 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, | |
a350ecce | 556 | struct net_device *sb_dev) |
96f84061 JW |
557 | { |
558 | struct tun_struct *tun = netdev_priv(dev); | |
559 | u16 ret; | |
560 | ||
561 | rcu_read_lock(); | |
562 | if (rcu_dereference(tun->steering_prog)) | |
563 | ret = tun_ebpf_select_queue(tun, skb); | |
564 | else | |
565 | ret = tun_automq_select_queue(tun, skb); | |
566 | rcu_read_unlock(); | |
567 | ||
568 | return ret; | |
569 | } | |
570 | ||
cde8b15f JW |
571 | static inline bool tun_not_capable(struct tun_struct *tun) |
572 | { | |
573 | const struct cred *cred = current_cred(); | |
c260b772 | 574 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
575 | |
576 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
577 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 578 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
579 | } |
580 | ||
c8d68e6b JW |
581 | static void tun_set_real_num_queues(struct tun_struct *tun) |
582 | { | |
583 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
584 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
585 | } | |
586 | ||
4008e97f JW |
587 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
588 | { | |
589 | tfile->detached = tun; | |
590 | list_add_tail(&tfile->next, &tun->disabled); | |
591 | ++tun->numdisabled; | |
592 | } | |
593 | ||
d32649d1 | 594 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
595 | { |
596 | struct tun_struct *tun = tfile->detached; | |
597 | ||
598 | tfile->detached = NULL; | |
599 | list_del_init(&tfile->next); | |
600 | --tun->numdisabled; | |
601 | return tun; | |
602 | } | |
603 | ||
3a403076 | 604 | void tun_ptr_free(void *ptr) |
fc72d1d5 JW |
605 | { |
606 | if (!ptr) | |
607 | return; | |
1ffcbc85 JDB |
608 | if (tun_is_xdp_frame(ptr)) { |
609 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 610 | |
03993094 | 611 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
612 | } else { |
613 | __skb_array_destroy_skb(ptr); | |
614 | } | |
615 | } | |
3a403076 | 616 | EXPORT_SYMBOL_GPL(tun_ptr_free); |
fc72d1d5 | 617 | |
4bfb0513 JW |
618 | static void tun_queue_purge(struct tun_file *tfile) |
619 | { | |
fc72d1d5 | 620 | void *ptr; |
1576d986 | 621 | |
fc72d1d5 JW |
622 | while ((ptr = ptr_ring_consume(&tfile->tx_ring)) != NULL) |
623 | tun_ptr_free(ptr); | |
1576d986 | 624 | |
5503fcec | 625 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
626 | skb_queue_purge(&tfile->sk.sk_error_queue); |
627 | } | |
628 | ||
c8d68e6b JW |
629 | static void __tun_detach(struct tun_file *tfile, bool clean) |
630 | { | |
631 | struct tun_file *ntfile; | |
632 | struct tun_struct *tun; | |
c8d68e6b | 633 | |
b8deabd3 JW |
634 | tun = rtnl_dereference(tfile->tun); |
635 | ||
94317099 | 636 | if (tun && clean) { |
06e55add ED |
637 | tun_napi_disable(tfile); |
638 | tun_napi_del(tfile); | |
94317099 PP |
639 | } |
640 | ||
9e85722d | 641 | if (tun && !tfile->detached) { |
c8d68e6b JW |
642 | u16 index = tfile->queue_index; |
643 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
644 | |
645 | rcu_assign_pointer(tun->tfiles[index], | |
646 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 647 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b | 648 | ntfile->queue_index = index; |
9871a9e4 JW |
649 | rcu_assign_pointer(tun->tfiles[tun->numqueues - 1], |
650 | NULL); | |
c8d68e6b JW |
651 | |
652 | --tun->numqueues; | |
9e85722d | 653 | if (clean) { |
c956674b | 654 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 655 | sock_put(&tfile->sk); |
9e85722d | 656 | } else |
4008e97f | 657 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
658 | |
659 | synchronize_net(); | |
96442e42 | 660 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 661 | /* Drop read queue */ |
4bfb0513 | 662 | tun_queue_purge(tfile); |
c8d68e6b | 663 | tun_set_real_num_queues(tun); |
dd38bd85 | 664 | } else if (tfile->detached && clean) { |
4008e97f | 665 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
666 | sock_put(&tfile->sk); |
667 | } | |
c8d68e6b JW |
668 | |
669 | if (clean) { | |
af668b3c MT |
670 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
671 | netif_carrier_off(tun->dev); | |
672 | ||
40630b82 | 673 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 674 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 675 | unregister_netdevice(tun->dev); |
af668b3c | 676 | } |
b196d88a JW |
677 | if (tun) |
678 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
7063efd3 | 679 | ptr_ring_cleanup(&tfile->tx_ring, tun_ptr_free); |
140e807d | 680 | sock_put(&tfile->sk); |
c8d68e6b JW |
681 | } |
682 | } | |
683 | ||
684 | static void tun_detach(struct tun_file *tfile, bool clean) | |
685 | { | |
83c1f36f SD |
686 | struct tun_struct *tun; |
687 | struct net_device *dev; | |
688 | ||
c8d68e6b | 689 | rtnl_lock(); |
83c1f36f SD |
690 | tun = rtnl_dereference(tfile->tun); |
691 | dev = tun ? tun->dev : NULL; | |
c8d68e6b | 692 | __tun_detach(tfile, clean); |
83c1f36f SD |
693 | if (dev) |
694 | netdev_state_change(dev); | |
c8d68e6b JW |
695 | rtnl_unlock(); |
696 | } | |
697 | ||
698 | static void tun_detach_all(struct net_device *dev) | |
699 | { | |
700 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 701 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
702 | int i, n = tun->numqueues; |
703 | ||
704 | for (i = 0; i < n; i++) { | |
b8deabd3 | 705 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 706 | BUG_ON(!tfile); |
06e55add | 707 | tun_napi_disable(tfile); |
addf8fc4 | 708 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 709 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 710 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
711 | --tun->numqueues; |
712 | } | |
9e85722d | 713 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 714 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 715 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 716 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 717 | } |
c8d68e6b JW |
718 | BUG_ON(tun->numqueues != 0); |
719 | ||
720 | synchronize_net(); | |
721 | for (i = 0; i < n; i++) { | |
b8deabd3 | 722 | tfile = rtnl_dereference(tun->tfiles[i]); |
06e55add | 723 | tun_napi_del(tfile); |
c8d68e6b | 724 | /* Drop read queue */ |
4bfb0513 | 725 | tun_queue_purge(tfile); |
b196d88a | 726 | xdp_rxq_info_unreg(&tfile->xdp_rxq); |
c8d68e6b JW |
727 | sock_put(&tfile->sk); |
728 | } | |
4008e97f JW |
729 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
730 | tun_enable_queue(tfile); | |
4bfb0513 | 731 | tun_queue_purge(tfile); |
b196d88a | 732 | xdp_rxq_info_unreg(&tfile->xdp_rxq); |
4008e97f JW |
733 | sock_put(&tfile->sk); |
734 | } | |
735 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 736 | |
40630b82 | 737 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 738 | module_put(THIS_MODULE); |
c8d68e6b JW |
739 | } |
740 | ||
94317099 | 741 | static int tun_attach(struct tun_struct *tun, struct file *file, |
77f22f92 YY |
742 | bool skip_filter, bool napi, bool napi_frags, |
743 | bool publish_tun) | |
a7385ba2 | 744 | { |
631ab46b | 745 | struct tun_file *tfile = file->private_data; |
1576d986 | 746 | struct net_device *dev = tun->dev; |
38231b7a | 747 | int err; |
a7385ba2 | 748 | |
5dbbaf2d PM |
749 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
750 | if (err < 0) | |
751 | goto out; | |
752 | ||
38231b7a | 753 | err = -EINVAL; |
9e85722d | 754 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
755 | goto out; |
756 | ||
757 | err = -EBUSY; | |
40630b82 | 758 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
759 | goto out; |
760 | ||
761 | err = -E2BIG; | |
4008e97f JW |
762 | if (!tfile->detached && |
763 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
764 | goto out; |
765 | ||
766 | err = 0; | |
54f968d6 | 767 | |
92d4ea6e | 768 | /* Re-attach the filter to persist device */ |
849c9b6f | 769 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
770 | lock_sock(tfile->socket.sk); |
771 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
772 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
773 | if (!err) |
774 | goto out; | |
775 | } | |
1576d986 JW |
776 | |
777 | if (!tfile->detached && | |
b196d88a JW |
778 | ptr_ring_resize(&tfile->tx_ring, dev->tx_queue_len, |
779 | GFP_KERNEL, tun_ptr_free)) { | |
1576d986 JW |
780 | err = -ENOMEM; |
781 | goto out; | |
782 | } | |
783 | ||
c8d68e6b | 784 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 785 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
8bf5c4ee JDB |
786 | |
787 | if (tfile->detached) { | |
788 | /* Re-attach detached tfile, updating XDP queue_index */ | |
789 | WARN_ON(!xdp_rxq_info_is_reg(&tfile->xdp_rxq)); | |
790 | ||
791 | if (tfile->xdp_rxq.queue_index != tfile->queue_index) | |
792 | tfile->xdp_rxq.queue_index = tfile->queue_index; | |
793 | } else { | |
794 | /* Setup XDP RX-queue info, for new tfile getting attached */ | |
795 | err = xdp_rxq_info_reg(&tfile->xdp_rxq, | |
b02e5a0e | 796 | tun->dev, tfile->queue_index, 0); |
8bf5c4ee JDB |
797 | if (err < 0) |
798 | goto out; | |
8d5d8852 JDB |
799 | err = xdp_rxq_info_reg_mem_model(&tfile->xdp_rxq, |
800 | MEM_TYPE_PAGE_SHARED, NULL); | |
801 | if (err < 0) { | |
802 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
803 | goto out; | |
804 | } | |
8bf5c4ee JDB |
805 | err = 0; |
806 | } | |
807 | ||
94317099 | 808 | if (tfile->detached) { |
4008e97f | 809 | tun_enable_queue(tfile); |
94317099 | 810 | } else { |
4008e97f | 811 | sock_hold(&tfile->sk); |
af3fb24e | 812 | tun_napi_init(tun, tfile, napi, napi_frags); |
94317099 | 813 | } |
4008e97f | 814 | |
e4a2a304 JW |
815 | if (rtnl_dereference(tun->xdp_prog)) |
816 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
817 | ||
c8d68e6b JW |
818 | /* device is allowed to go away first, so no need to hold extra |
819 | * refcnt. | |
820 | */ | |
821 | ||
0b7959b6 SF |
822 | /* Publish tfile->tun and tun->tfiles only after we've fully |
823 | * initialized tfile; otherwise we risk using half-initialized | |
824 | * object. | |
825 | */ | |
77f22f92 YY |
826 | if (publish_tun) |
827 | rcu_assign_pointer(tfile->tun, tun); | |
0b7959b6 SF |
828 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
829 | tun->numqueues++; | |
3a03cb84 | 830 | tun_set_real_num_queues(tun); |
c8d68e6b JW |
831 | out: |
832 | return err; | |
631ab46b EB |
833 | } |
834 | ||
9484dc74 | 835 | static struct tun_struct *tun_get(struct tun_file *tfile) |
631ab46b | 836 | { |
6e914fc7 | 837 | struct tun_struct *tun; |
c70f1829 | 838 | |
6e914fc7 JW |
839 | rcu_read_lock(); |
840 | tun = rcu_dereference(tfile->tun); | |
841 | if (tun) | |
842 | dev_hold(tun->dev); | |
843 | rcu_read_unlock(); | |
c70f1829 EB |
844 | |
845 | return tun; | |
631ab46b EB |
846 | } |
847 | ||
631ab46b EB |
848 | static void tun_put(struct tun_struct *tun) |
849 | { | |
6e914fc7 | 850 | dev_put(tun->dev); |
631ab46b EB |
851 | } |
852 | ||
6b8a66ee | 853 | /* TAP filtering */ |
f271b2cc MK |
854 | static void addr_hash_set(u32 *mask, const u8 *addr) |
855 | { | |
856 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
857 | mask[n >> 5] |= (1 << (n & 31)); | |
858 | } | |
859 | ||
860 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
861 | { | |
862 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
863 | return mask[n >> 5] & (1 << (n & 31)); | |
864 | } | |
865 | ||
866 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
867 | { | |
868 | struct { u8 u[ETH_ALEN]; } *addr; | |
869 | struct tun_filter uf; | |
870 | int err, alen, n, nexact; | |
871 | ||
872 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
873 | return -EFAULT; | |
874 | ||
875 | if (!uf.count) { | |
876 | /* Disabled */ | |
877 | filter->count = 0; | |
878 | return 0; | |
879 | } | |
880 | ||
881 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
882 | addr = memdup_user(arg + sizeof(uf), alen); |
883 | if (IS_ERR(addr)) | |
884 | return PTR_ERR(addr); | |
f271b2cc MK |
885 | |
886 | /* The filter is updated without holding any locks. Which is | |
887 | * perfectly safe. We disable it first and in the worst | |
888 | * case we'll accept a few undesired packets. */ | |
889 | filter->count = 0; | |
890 | wmb(); | |
891 | ||
892 | /* Use first set of addresses as an exact filter */ | |
893 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
894 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
895 | ||
896 | nexact = n; | |
897 | ||
cfbf84fc AW |
898 | /* Remaining multicast addresses are hashed, |
899 | * unicast will leave the filter disabled. */ | |
f271b2cc | 900 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
901 | for (; n < uf.count; n++) { |
902 | if (!is_multicast_ether_addr(addr[n].u)) { | |
903 | err = 0; /* no filter */ | |
3b8d2a69 | 904 | goto free_addr; |
cfbf84fc | 905 | } |
f271b2cc | 906 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 907 | } |
f271b2cc MK |
908 | |
909 | /* For ALLMULTI just set the mask to all ones. | |
910 | * This overrides the mask populated above. */ | |
911 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
912 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
913 | ||
914 | /* Now enable the filter */ | |
915 | wmb(); | |
916 | filter->count = nexact; | |
917 | ||
918 | /* Return the number of exact filters */ | |
919 | err = nexact; | |
3b8d2a69 | 920 | free_addr: |
f271b2cc MK |
921 | kfree(addr); |
922 | return err; | |
923 | } | |
924 | ||
925 | /* Returns: 0 - drop, !=0 - accept */ | |
926 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
927 | { | |
928 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
929 | * at this point. */ | |
930 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
931 | int i; | |
932 | ||
933 | /* Exact match */ | |
934 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 935 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
936 | return 1; |
937 | ||
938 | /* Inexact match (multicast only) */ | |
939 | if (is_multicast_ether_addr(eh->h_dest)) | |
940 | return addr_hash_test(filter->mask, eh->h_dest); | |
941 | ||
942 | return 0; | |
943 | } | |
944 | ||
945 | /* | |
946 | * Checks whether the packet is accepted or not. | |
947 | * Returns: 0 - drop, !=0 - accept | |
948 | */ | |
949 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
950 | { | |
951 | if (!filter->count) | |
952 | return 1; | |
953 | ||
954 | return run_filter(filter, skb); | |
955 | } | |
956 | ||
1da177e4 LT |
957 | /* Network device part of the driver */ |
958 | ||
7282d491 | 959 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 960 | |
158b515f GK |
961 | static int tun_net_init(struct net_device *dev) |
962 | { | |
963 | struct tun_struct *tun = netdev_priv(dev); | |
964 | struct ifreq *ifr = tun->ifr; | |
965 | int err; | |
966 | ||
967 | dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); | |
968 | if (!dev->tstats) | |
969 | return -ENOMEM; | |
970 | ||
971 | spin_lock_init(&tun->lock); | |
972 | ||
973 | err = security_tun_dev_alloc_security(&tun->security); | |
974 | if (err < 0) { | |
975 | free_percpu(dev->tstats); | |
976 | return err; | |
977 | } | |
978 | ||
979 | tun_flow_init(tun); | |
980 | ||
981 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | | |
982 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | | |
983 | NETIF_F_HW_VLAN_STAG_TX; | |
984 | dev->features = dev->hw_features | NETIF_F_LLTX; | |
985 | dev->vlan_features = dev->features & | |
986 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
987 | NETIF_F_HW_VLAN_STAG_TX); | |
988 | ||
989 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
990 | (ifr->ifr_flags & TUN_FEATURES); | |
991 | ||
992 | INIT_LIST_HEAD(&tun->disabled); | |
993 | err = tun_attach(tun, tun->file, false, ifr->ifr_flags & IFF_NAPI, | |
994 | ifr->ifr_flags & IFF_NAPI_FRAGS, false); | |
995 | if (err < 0) { | |
996 | tun_flow_uninit(tun); | |
997 | security_tun_dev_free_security(tun->security); | |
998 | free_percpu(dev->tstats); | |
999 | return err; | |
1000 | } | |
1001 | return 0; | |
1002 | } | |
1003 | ||
c70f1829 EB |
1004 | /* Net device detach from fd. */ |
1005 | static void tun_net_uninit(struct net_device *dev) | |
1006 | { | |
c8d68e6b | 1007 | tun_detach_all(dev); |
c70f1829 EB |
1008 | } |
1009 | ||
1da177e4 LT |
1010 | /* Net device open. */ |
1011 | static int tun_net_open(struct net_device *dev) | |
1012 | { | |
c8d68e6b | 1013 | netif_tx_start_all_queues(dev); |
b20e2d54 | 1014 | |
1da177e4 LT |
1015 | return 0; |
1016 | } | |
1017 | ||
1018 | /* Net device close. */ | |
1019 | static int tun_net_close(struct net_device *dev) | |
1020 | { | |
c8d68e6b | 1021 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
1022 | return 0; |
1023 | } | |
1024 | ||
1025 | /* Net device start xmit */ | |
96f84061 | 1026 | static void tun_automq_xmit(struct tun_struct *tun, struct sk_buff *skb) |
1da177e4 | 1027 | { |
3df97ba8 | 1028 | #ifdef CONFIG_RPS |
dc05360f | 1029 | if (tun->numqueues == 1 && static_branch_unlikely(&rps_needed)) { |
9bc88939 TH |
1030 | /* Select queue was not called for the skbuff, so we extract the |
1031 | * RPS hash and save it into the flow_table here. | |
1032 | */ | |
4b035271 | 1033 | struct tun_flow_entry *e; |
9bc88939 TH |
1034 | __u32 rxhash; |
1035 | ||
feec084a | 1036 | rxhash = __skb_get_hash_symmetric(skb); |
4b035271 WL |
1037 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], rxhash); |
1038 | if (e) | |
1039 | tun_flow_save_rps_rxhash(e, rxhash); | |
9bc88939 | 1040 | } |
3df97ba8 | 1041 | #endif |
96f84061 JW |
1042 | } |
1043 | ||
aff3d70a JW |
1044 | static unsigned int run_ebpf_filter(struct tun_struct *tun, |
1045 | struct sk_buff *skb, | |
1046 | int len) | |
1047 | { | |
1048 | struct tun_prog *prog = rcu_dereference(tun->filter_prog); | |
1049 | ||
1050 | if (prog) | |
1051 | len = bpf_prog_run_clear_cb(prog->prog, skb); | |
1052 | ||
1053 | return len; | |
1054 | } | |
1055 | ||
96f84061 JW |
1056 | /* Net device start xmit */ |
1057 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) | |
1058 | { | |
1059 | struct tun_struct *tun = netdev_priv(dev); | |
4b4f052e | 1060 | enum skb_drop_reason drop_reason; |
96f84061 | 1061 | int txq = skb->queue_mapping; |
a31d27fb | 1062 | struct netdev_queue *queue; |
96f84061 | 1063 | struct tun_file *tfile; |
aff3d70a | 1064 | int len = skb->len; |
96f84061 JW |
1065 | |
1066 | rcu_read_lock(); | |
1067 | tfile = rcu_dereference(tun->tfiles[txq]); | |
96f84061 JW |
1068 | |
1069 | /* Drop packet if interface is not attached */ | |
4b4f052e DZ |
1070 | if (!tfile) { |
1071 | drop_reason = SKB_DROP_REASON_DEV_READY; | |
96f84061 | 1072 | goto drop; |
4b4f052e | 1073 | } |
96f84061 JW |
1074 | |
1075 | if (!rcu_dereference(tun->steering_prog)) | |
1076 | tun_automq_xmit(tun, skb); | |
9bc88939 | 1077 | |
3424170f | 1078 | netif_info(tun, tx_queued, tun->dev, "%s %d\n", __func__, skb->len); |
6e914fc7 | 1079 | |
f271b2cc MK |
1080 | /* Drop if the filter does not like it. |
1081 | * This is a noop if the filter is disabled. | |
1082 | * Filter can be enabled only for the TAP devices. */ | |
4b4f052e DZ |
1083 | if (!check_filter(&tun->txflt, skb)) { |
1084 | drop_reason = SKB_DROP_REASON_TAP_TXFILTER; | |
f271b2cc | 1085 | goto drop; |
4b4f052e | 1086 | } |
f271b2cc | 1087 | |
54f968d6 | 1088 | if (tfile->socket.sk->sk_filter && |
4b4f052e DZ |
1089 | sk_filter(tfile->socket.sk, skb)) { |
1090 | drop_reason = SKB_DROP_REASON_SOCKET_FILTER; | |
99405162 | 1091 | goto drop; |
4b4f052e | 1092 | } |
99405162 | 1093 | |
aff3d70a | 1094 | len = run_ebpf_filter(tun, skb, len); |
4b4f052e DZ |
1095 | if (len == 0) { |
1096 | drop_reason = SKB_DROP_REASON_TAP_FILTER; | |
45a15d89 | 1097 | goto drop; |
4b4f052e | 1098 | } |
45a15d89 | 1099 | |
4b4f052e DZ |
1100 | if (pskb_trim(skb, len)) { |
1101 | drop_reason = SKB_DROP_REASON_NOMEM; | |
aff3d70a | 1102 | goto drop; |
4b4f052e | 1103 | } |
aff3d70a | 1104 | |
4b4f052e DZ |
1105 | if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) { |
1106 | drop_reason = SKB_DROP_REASON_SKB_UCOPY_FAULT; | |
7bf66305 | 1107 | goto drop; |
4b4f052e | 1108 | } |
7bf66305 | 1109 | |
7b996243 | 1110 | skb_tx_timestamp(skb); |
eda29772 | 1111 | |
0110d6f2 | 1112 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
1113 | * for indefinite time. |
1114 | */ | |
0110d6f2 MT |
1115 | skb_orphan(skb); |
1116 | ||
895b5c9f | 1117 | nf_reset_ct(skb); |
f8af75f3 | 1118 | |
4b4f052e DZ |
1119 | if (ptr_ring_produce(&tfile->tx_ring, skb)) { |
1120 | drop_reason = SKB_DROP_REASON_FULL_RING; | |
1576d986 | 1121 | goto drop; |
4b4f052e | 1122 | } |
1da177e4 | 1123 | |
a31d27fb ND |
1124 | /* NETIF_F_LLTX requires to do our own update of trans_start */ |
1125 | queue = netdev_get_tx_queue(dev, txq); | |
968a1a5d | 1126 | txq_trans_cond_update(queue); |
a31d27fb | 1127 | |
1da177e4 | 1128 | /* Notify and wake up reader process */ |
54f968d6 JW |
1129 | if (tfile->flags & TUN_FASYNC) |
1130 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 1131 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
1132 | |
1133 | rcu_read_unlock(); | |
6ed10654 | 1134 | return NETDEV_TX_OK; |
1da177e4 LT |
1135 | |
1136 | drop: | |
625788b5 | 1137 | dev_core_stats_tx_dropped_inc(dev); |
149d36f7 | 1138 | skb_tx_error(skb); |
4b4f052e | 1139 | kfree_skb_reason(skb, drop_reason); |
6e914fc7 | 1140 | rcu_read_unlock(); |
baeababb | 1141 | return NET_XMIT_DROP; |
1da177e4 LT |
1142 | } |
1143 | ||
f271b2cc | 1144 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 1145 | { |
f271b2cc MK |
1146 | /* |
1147 | * This callback is supposed to deal with mc filter in | |
1148 | * _rx_ path and has nothing to do with the _tx_ path. | |
1149 | * In rx path we always accept everything userspace gives us. | |
1150 | */ | |
1da177e4 LT |
1151 | } |
1152 | ||
c8f44aff MM |
1153 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
1154 | netdev_features_t features) | |
88255375 MM |
1155 | { |
1156 | struct tun_struct *tun = netdev_priv(dev); | |
1157 | ||
1158 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
1159 | } | |
eaea34b2 PA |
1160 | |
1161 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
1162 | { | |
1163 | struct tun_struct *tun = netdev_priv(dev); | |
1164 | ||
1165 | if (new_hr < NET_SKB_PAD) | |
1166 | new_hr = NET_SKB_PAD; | |
1167 | ||
1168 | tun->align = new_hr; | |
1169 | } | |
1170 | ||
bc1f4470 | 1171 | static void |
608b9977 PA |
1172 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
1173 | { | |
608b9977 | 1174 | struct tun_struct *tun = netdev_priv(dev); |
608b9977 | 1175 | |
497a5757 | 1176 | dev_get_tstats64(dev, stats); |
608b9977 | 1177 | |
497a5757 HK |
1178 | stats->rx_frame_errors += |
1179 | (unsigned long)atomic_long_read(&tun->rx_frame_errors); | |
608b9977 PA |
1180 | } |
1181 | ||
761876c8 JW |
1182 | static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog, |
1183 | struct netlink_ext_ack *extack) | |
1184 | { | |
1185 | struct tun_struct *tun = netdev_priv(dev); | |
e4a2a304 | 1186 | struct tun_file *tfile; |
761876c8 | 1187 | struct bpf_prog *old_prog; |
e4a2a304 | 1188 | int i; |
761876c8 JW |
1189 | |
1190 | old_prog = rtnl_dereference(tun->xdp_prog); | |
1191 | rcu_assign_pointer(tun->xdp_prog, prog); | |
1192 | if (old_prog) | |
1193 | bpf_prog_put(old_prog); | |
1194 | ||
e4a2a304 JW |
1195 | for (i = 0; i < tun->numqueues; i++) { |
1196 | tfile = rtnl_dereference(tun->tfiles[i]); | |
1197 | if (prog) | |
1198 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
1199 | else | |
1200 | sock_reset_flag(&tfile->sk, SOCK_XDP); | |
1201 | } | |
1202 | list_for_each_entry(tfile, &tun->disabled, next) { | |
1203 | if (prog) | |
1204 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
1205 | else | |
1206 | sock_reset_flag(&tfile->sk, SOCK_XDP); | |
1207 | } | |
1208 | ||
761876c8 JW |
1209 | return 0; |
1210 | } | |
1211 | ||
f4e63525 | 1212 | static int tun_xdp(struct net_device *dev, struct netdev_bpf *xdp) |
761876c8 JW |
1213 | { |
1214 | switch (xdp->command) { | |
1215 | case XDP_SETUP_PROG: | |
1216 | return tun_xdp_set(dev, xdp->prog, xdp->extack); | |
761876c8 JW |
1217 | default: |
1218 | return -EINVAL; | |
1219 | } | |
1220 | } | |
1221 | ||
26d31925 ND |
1222 | static int tun_net_change_carrier(struct net_device *dev, bool new_carrier) |
1223 | { | |
1224 | if (new_carrier) { | |
1225 | struct tun_struct *tun = netdev_priv(dev); | |
1226 | ||
1227 | if (!tun->numqueues) | |
1228 | return -EPERM; | |
1229 | ||
1230 | netif_carrier_on(dev); | |
1231 | } else { | |
1232 | netif_carrier_off(dev); | |
1233 | } | |
1234 | return 0; | |
1235 | } | |
1236 | ||
758e43b7 | 1237 | static const struct net_device_ops tun_netdev_ops = { |
158b515f | 1238 | .ndo_init = tun_net_init, |
c70f1829 | 1239 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1240 | .ndo_open = tun_net_open, |
1241 | .ndo_stop = tun_net_close, | |
00829823 | 1242 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1243 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1244 | .ndo_select_queue = tun_select_queue, |
eaea34b2 | 1245 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1246 | .ndo_get_stats64 = tun_net_get_stats64, |
26d31925 | 1247 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1248 | }; |
1249 | ||
0c9d917b JDB |
1250 | static void __tun_xdp_flush_tfile(struct tun_file *tfile) |
1251 | { | |
1252 | /* Notify and wake up reader process */ | |
1253 | if (tfile->flags & TUN_FASYNC) | |
1254 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
1255 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); | |
1256 | } | |
1257 | ||
42b33468 JDB |
1258 | static int tun_xdp_xmit(struct net_device *dev, int n, |
1259 | struct xdp_frame **frames, u32 flags) | |
fc72d1d5 JW |
1260 | { |
1261 | struct tun_struct *tun = netdev_priv(dev); | |
fc72d1d5 JW |
1262 | struct tun_file *tfile; |
1263 | u32 numqueues; | |
fdc13979 | 1264 | int nxmit = 0; |
735fc405 | 1265 | int i; |
fc72d1d5 | 1266 | |
0c9d917b | 1267 | if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) |
42b33468 JDB |
1268 | return -EINVAL; |
1269 | ||
fc72d1d5 JW |
1270 | rcu_read_lock(); |
1271 | ||
9871a9e4 | 1272 | resample: |
fc72d1d5 JW |
1273 | numqueues = READ_ONCE(tun->numqueues); |
1274 | if (!numqueues) { | |
735fc405 JDB |
1275 | rcu_read_unlock(); |
1276 | return -ENXIO; /* Caller will free/return all frames */ | |
fc72d1d5 JW |
1277 | } |
1278 | ||
1279 | tfile = rcu_dereference(tun->tfiles[smp_processor_id() % | |
1280 | numqueues]); | |
9871a9e4 JW |
1281 | if (unlikely(!tfile)) |
1282 | goto resample; | |
735fc405 JDB |
1283 | |
1284 | spin_lock(&tfile->tx_ring.producer_lock); | |
1285 | for (i = 0; i < n; i++) { | |
1286 | struct xdp_frame *xdp = frames[i]; | |
1287 | /* Encode the XDP flag into lowest bit for consumer to differ | |
1288 | * XDP buffer from sk_buff. | |
1289 | */ | |
1290 | void *frame = tun_xdp_to_ptr(xdp); | |
1291 | ||
1292 | if (__ptr_ring_produce(&tfile->tx_ring, frame)) { | |
625788b5 | 1293 | dev_core_stats_tx_dropped_inc(dev); |
fdc13979 | 1294 | break; |
735fc405 | 1295 | } |
fdc13979 | 1296 | nxmit++; |
fc72d1d5 | 1297 | } |
735fc405 | 1298 | spin_unlock(&tfile->tx_ring.producer_lock); |
fc72d1d5 | 1299 | |
0c9d917b JDB |
1300 | if (flags & XDP_XMIT_FLUSH) |
1301 | __tun_xdp_flush_tfile(tfile); | |
1302 | ||
fc72d1d5 | 1303 | rcu_read_unlock(); |
fdc13979 | 1304 | return nxmit; |
fc72d1d5 JW |
1305 | } |
1306 | ||
44fa2dbd JDB |
1307 | static int tun_xdp_tx(struct net_device *dev, struct xdp_buff *xdp) |
1308 | { | |
1b698fa5 | 1309 | struct xdp_frame *frame = xdp_convert_buff_to_frame(xdp); |
fdc13979 | 1310 | int nxmit; |
44fa2dbd JDB |
1311 | |
1312 | if (unlikely(!frame)) | |
1313 | return -EOVERFLOW; | |
1314 | ||
fdc13979 LB |
1315 | nxmit = tun_xdp_xmit(dev, 1, &frame, XDP_XMIT_FLUSH); |
1316 | if (!nxmit) | |
1317 | xdp_return_frame_rx_napi(frame); | |
1318 | return nxmit; | |
fc72d1d5 JW |
1319 | } |
1320 | ||
758e43b7 | 1321 | static const struct net_device_ops tap_netdev_ops = { |
158b515f | 1322 | .ndo_init = tun_net_init, |
c70f1829 | 1323 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1324 | .ndo_open = tun_net_open, |
1325 | .ndo_stop = tun_net_close, | |
00829823 | 1326 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1327 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1328 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1329 | .ndo_set_mac_address = eth_mac_addr, |
1330 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1331 | .ndo_select_queue = tun_select_queue, |
5e52796a | 1332 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1333 | .ndo_set_rx_headroom = tun_set_headroom, |
497a5757 | 1334 | .ndo_get_stats64 = dev_get_tstats64, |
f4e63525 | 1335 | .ndo_bpf = tun_xdp, |
fc72d1d5 | 1336 | .ndo_xdp_xmit = tun_xdp_xmit, |
26d31925 | 1337 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1338 | }; |
1339 | ||
944a1376 | 1340 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1341 | { |
1342 | int i; | |
1343 | ||
96442e42 JW |
1344 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1345 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1346 | ||
1347 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
e99e88a9 KC |
1348 | timer_setup(&tun->flow_gc_timer, tun_flow_cleanup, 0); |
1349 | mod_timer(&tun->flow_gc_timer, | |
1350 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1351 | } |
1352 | ||
1353 | static void tun_flow_uninit(struct tun_struct *tun) | |
1354 | { | |
1355 | del_timer_sync(&tun->flow_gc_timer); | |
1356 | tun_flow_flush(tun); | |
96442e42 JW |
1357 | } |
1358 | ||
91572088 JW |
1359 | #define MIN_MTU 68 |
1360 | #define MAX_MTU 65535 | |
1361 | ||
1da177e4 | 1362 | /* Initialize net device. */ |
158b515f | 1363 | static void tun_net_initialize(struct net_device *dev) |
1da177e4 LT |
1364 | { |
1365 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1366 | |
1da177e4 | 1367 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1368 | case IFF_TUN: |
758e43b7 | 1369 | dev->netdev_ops = &tun_netdev_ops; |
b9815eb1 | 1370 | dev->header_ops = &ip_tunnel_header_ops; |
758e43b7 | 1371 | |
1da177e4 LT |
1372 | /* Point-to-Point TUN Device */ |
1373 | dev->hard_header_len = 0; | |
1374 | dev->addr_len = 0; | |
1375 | dev->mtu = 1500; | |
1376 | ||
1377 | /* Zero header length */ | |
6aa20a22 | 1378 | dev->type = ARPHRD_NONE; |
1da177e4 | 1379 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1380 | break; |
1381 | ||
40630b82 | 1382 | case IFF_TAP: |
7a0a9608 | 1383 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1384 | /* Ethernet TAP Device */ |
1da177e4 | 1385 | ether_setup(dev); |
550fd08c | 1386 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1387 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1388 | |
f2cedb63 | 1389 | eth_hw_addr_random(dev); |
36226a8d | 1390 | |
1da177e4 LT |
1391 | break; |
1392 | } | |
91572088 JW |
1393 | |
1394 | dev->min_mtu = MIN_MTU; | |
1395 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1396 | } |
1397 | ||
2f3ab622 JW |
1398 | static bool tun_sock_writeable(struct tun_struct *tun, struct tun_file *tfile) |
1399 | { | |
1400 | struct sock *sk = tfile->socket.sk; | |
1401 | ||
1402 | return (tun->dev->flags & IFF_UP) && sock_writeable(sk); | |
1403 | } | |
1404 | ||
1da177e4 LT |
1405 | /* Character device part */ |
1406 | ||
1407 | /* Poll */ | |
afc9a42b | 1408 | static __poll_t tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1409 | { |
b2430de3 | 1410 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1411 | struct tun_struct *tun = tun_get(tfile); |
3c8a9c63 | 1412 | struct sock *sk; |
afc9a42b | 1413 | __poll_t mask = 0; |
1da177e4 LT |
1414 | |
1415 | if (!tun) | |
a9a08845 | 1416 | return EPOLLERR; |
1da177e4 | 1417 | |
54f968d6 | 1418 | sk = tfile->socket.sk; |
3c8a9c63 | 1419 | |
9e641bdc | 1420 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1421 | |
5990a305 | 1422 | if (!ptr_ring_empty(&tfile->tx_ring)) |
a9a08845 | 1423 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 1424 | |
2f3ab622 JW |
1425 | /* Make sure SOCKWQ_ASYNC_NOSPACE is set if not writable to |
1426 | * guarantee EPOLLOUT to be raised by either here or | |
1427 | * tun_sock_write_space(). Then process could get notification | |
1428 | * after it writes to a down device and meets -EIO. | |
1429 | */ | |
1430 | if (tun_sock_writeable(tun, tfile) || | |
1431 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
1432 | tun_sock_writeable(tun, tfile))) | |
a9a08845 | 1433 | mask |= EPOLLOUT | EPOLLWRNORM; |
33dccbb0 | 1434 | |
c70f1829 | 1435 | if (tun->dev->reg_state != NETREG_REGISTERED) |
a9a08845 | 1436 | mask = EPOLLERR; |
c70f1829 | 1437 | |
631ab46b | 1438 | tun_put(tun); |
1da177e4 LT |
1439 | return mask; |
1440 | } | |
1441 | ||
90e33d45 PP |
1442 | static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile, |
1443 | size_t len, | |
1444 | const struct iov_iter *it) | |
1445 | { | |
1446 | struct sk_buff *skb; | |
1447 | size_t linear; | |
1448 | int err; | |
1449 | int i; | |
1450 | ||
1451 | if (it->nr_segs > MAX_SKB_FRAGS + 1) | |
950271d7 | 1452 | return ERR_PTR(-EMSGSIZE); |
90e33d45 PP |
1453 | |
1454 | local_bh_disable(); | |
1455 | skb = napi_get_frags(&tfile->napi); | |
1456 | local_bh_enable(); | |
1457 | if (!skb) | |
1458 | return ERR_PTR(-ENOMEM); | |
1459 | ||
1460 | linear = iov_iter_single_seg_count(it); | |
1461 | err = __skb_grow(skb, linear); | |
1462 | if (err) | |
1463 | goto free; | |
1464 | ||
1465 | skb->len = len; | |
1466 | skb->data_len = len - linear; | |
1467 | skb->truesize += skb->data_len; | |
1468 | ||
1469 | for (i = 1; i < it->nr_segs; i++) { | |
1470 | size_t fragsz = it->iov[i].iov_len; | |
aa6daaca ED |
1471 | struct page *page; |
1472 | void *frag; | |
90e33d45 PP |
1473 | |
1474 | if (fragsz == 0 || fragsz > PAGE_SIZE) { | |
1475 | err = -EINVAL; | |
1476 | goto free; | |
1477 | } | |
aa6daaca ED |
1478 | frag = netdev_alloc_frag(fragsz); |
1479 | if (!frag) { | |
90e33d45 PP |
1480 | err = -ENOMEM; |
1481 | goto free; | |
1482 | } | |
aa6daaca ED |
1483 | page = virt_to_head_page(frag); |
1484 | skb_fill_page_desc(skb, i - 1, page, | |
1485 | frag - page_address(page), fragsz); | |
90e33d45 PP |
1486 | } |
1487 | ||
1488 | return skb; | |
1489 | free: | |
1490 | /* frees skb and all frags allocated with napi_alloc_frag() */ | |
1491 | napi_free_frags(&tfile->napi); | |
1492 | return ERR_PTR(err); | |
1493 | } | |
1494 | ||
f42157cb RR |
1495 | /* prepad is the amount to reserve at front. len is length after that. |
1496 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1497 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1498 | size_t prepad, size_t len, |
1499 | size_t linear, int noblock) | |
f42157cb | 1500 | { |
54f968d6 | 1501 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1502 | struct sk_buff *skb; |
33dccbb0 | 1503 | int err; |
f42157cb RR |
1504 | |
1505 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1506 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1507 | linear = len; |
f42157cb | 1508 | |
33dccbb0 | 1509 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1510 | &err, 0); |
f42157cb | 1511 | if (!skb) |
33dccbb0 | 1512 | return ERR_PTR(err); |
f42157cb RR |
1513 | |
1514 | skb_reserve(skb, prepad); | |
1515 | skb_put(skb, linear); | |
33dccbb0 HX |
1516 | skb->data_len = len - linear; |
1517 | skb->len += len - linear; | |
f42157cb RR |
1518 | |
1519 | return skb; | |
1520 | } | |
1521 | ||
5503fcec JW |
1522 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1523 | struct sk_buff *skb, int more) | |
1524 | { | |
1525 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1526 | struct sk_buff_head process_queue; | |
1527 | u32 rx_batched = tun->rx_batched; | |
1528 | bool rcv = false; | |
1529 | ||
1530 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1531 | local_bh_disable(); | |
8ebebcba | 1532 | skb_record_rx_queue(skb, tfile->queue_index); |
5503fcec JW |
1533 | netif_receive_skb(skb); |
1534 | local_bh_enable(); | |
1535 | return; | |
1536 | } | |
1537 | ||
1538 | spin_lock(&queue->lock); | |
1539 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1540 | __skb_queue_head_init(&process_queue); | |
1541 | skb_queue_splice_tail_init(queue, &process_queue); | |
1542 | rcv = true; | |
1543 | } else { | |
1544 | __skb_queue_tail(queue, skb); | |
1545 | } | |
1546 | spin_unlock(&queue->lock); | |
1547 | ||
1548 | if (rcv) { | |
1549 | struct sk_buff *nskb; | |
1550 | ||
1551 | local_bh_disable(); | |
8ebebcba MC |
1552 | while ((nskb = __skb_dequeue(&process_queue))) { |
1553 | skb_record_rx_queue(nskb, tfile->queue_index); | |
5503fcec | 1554 | netif_receive_skb(nskb); |
8ebebcba MC |
1555 | } |
1556 | skb_record_rx_queue(skb, tfile->queue_index); | |
5503fcec JW |
1557 | netif_receive_skb(skb); |
1558 | local_bh_enable(); | |
1559 | } | |
1560 | } | |
1561 | ||
66ccbc9c JW |
1562 | static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile, |
1563 | int len, int noblock, bool zerocopy) | |
1564 | { | |
1565 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) | |
1566 | return false; | |
1567 | ||
1568 | if (tfile->socket.sk->sk_sndbuf != INT_MAX) | |
1569 | return false; | |
1570 | ||
1571 | if (!noblock) | |
1572 | return false; | |
1573 | ||
1574 | if (zerocopy) | |
1575 | return false; | |
1576 | ||
1577 | if (SKB_DATA_ALIGN(len + TUN_RX_PAD) + | |
1578 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE) | |
1579 | return false; | |
1580 | ||
1581 | return true; | |
1582 | } | |
1583 | ||
4b663366 AB |
1584 | static struct sk_buff *__tun_build_skb(struct tun_file *tfile, |
1585 | struct page_frag *alloc_frag, char *buf, | |
8ae1aff0 | 1586 | int buflen, int len, int pad) |
ac1f1f6c JW |
1587 | { |
1588 | struct sk_buff *skb = build_skb(buf, buflen); | |
1589 | ||
1590 | if (!skb) | |
1591 | return ERR_PTR(-ENOMEM); | |
1592 | ||
8ae1aff0 | 1593 | skb_reserve(skb, pad); |
ac1f1f6c | 1594 | skb_put(skb, len); |
4b663366 | 1595 | skb_set_owner_w(skb, tfile->socket.sk); |
ac1f1f6c JW |
1596 | |
1597 | get_page(alloc_frag->page); | |
1598 | alloc_frag->offset += buflen; | |
1599 | ||
1600 | return skb; | |
1601 | } | |
1602 | ||
8ae1aff0 JW |
1603 | static int tun_xdp_act(struct tun_struct *tun, struct bpf_prog *xdp_prog, |
1604 | struct xdp_buff *xdp, u32 act) | |
1605 | { | |
1606 | int err; | |
1607 | ||
1608 | switch (act) { | |
1609 | case XDP_REDIRECT: | |
1610 | err = xdp_do_redirect(tun->dev, xdp, xdp_prog); | |
8ae1aff0 JW |
1611 | if (err) |
1612 | return err; | |
1613 | break; | |
1614 | case XDP_TX: | |
1615 | err = tun_xdp_tx(tun->dev, xdp); | |
1616 | if (err < 0) | |
1617 | return err; | |
1618 | break; | |
1619 | case XDP_PASS: | |
1620 | break; | |
1621 | default: | |
c8064e5b | 1622 | bpf_warn_invalid_xdp_action(tun->dev, xdp_prog, act); |
df561f66 | 1623 | fallthrough; |
8ae1aff0 JW |
1624 | case XDP_ABORTED: |
1625 | trace_xdp_exception(tun->dev, xdp_prog, act); | |
df561f66 | 1626 | fallthrough; |
8ae1aff0 | 1627 | case XDP_DROP: |
625788b5 | 1628 | dev_core_stats_rx_dropped_inc(tun->dev); |
8ae1aff0 JW |
1629 | break; |
1630 | } | |
1631 | ||
1632 | return act; | |
1633 | } | |
1634 | ||
761876c8 JW |
1635 | static struct sk_buff *tun_build_skb(struct tun_struct *tun, |
1636 | struct tun_file *tfile, | |
66ccbc9c | 1637 | struct iov_iter *from, |
761876c8 | 1638 | struct virtio_net_hdr *hdr, |
1cfe6e93 | 1639 | int len, int *skb_xdp) |
66ccbc9c | 1640 | { |
0bbd7dad | 1641 | struct page_frag *alloc_frag = ¤t->task_frag; |
761876c8 | 1642 | struct bpf_prog *xdp_prog; |
7df13219 | 1643 | int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); |
66ccbc9c JW |
1644 | char *buf; |
1645 | size_t copied; | |
8ae1aff0 JW |
1646 | int pad = TUN_RX_PAD; |
1647 | int err = 0; | |
7df13219 JW |
1648 | |
1649 | rcu_read_lock(); | |
1650 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1651 | if (xdp_prog) | |
4f23aff8 | 1652 | pad += XDP_PACKET_HEADROOM; |
7df13219 JW |
1653 | buflen += SKB_DATA_ALIGN(len + pad); |
1654 | rcu_read_unlock(); | |
66ccbc9c | 1655 | |
63b9ab65 | 1656 | alloc_frag->offset = ALIGN((u64)alloc_frag->offset, SMP_CACHE_BYTES); |
66ccbc9c JW |
1657 | if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL))) |
1658 | return ERR_PTR(-ENOMEM); | |
1659 | ||
1660 | buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; | |
1661 | copied = copy_page_from_iter(alloc_frag->page, | |
7df13219 | 1662 | alloc_frag->offset + pad, |
66ccbc9c JW |
1663 | len, from); |
1664 | if (copied != len) | |
1665 | return ERR_PTR(-EFAULT); | |
1666 | ||
7df13219 JW |
1667 | /* There's a small window that XDP may be set after the check |
1668 | * of xdp_prog above, this should be rare and for simplicity | |
1669 | * we do XDP on skb in case the headroom is not enough. | |
1670 | */ | |
ac1f1f6c | 1671 | if (hdr->gso_type || !xdp_prog) { |
1cfe6e93 | 1672 | *skb_xdp = 1; |
4b663366 AB |
1673 | return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, |
1674 | pad); | |
ac1f1f6c JW |
1675 | } |
1676 | ||
1677 | *skb_xdp = 0; | |
761876c8 | 1678 | |
6547e387 | 1679 | local_bh_disable(); |
761876c8 JW |
1680 | rcu_read_lock(); |
1681 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
8ae1aff0 | 1682 | if (xdp_prog) { |
761876c8 | 1683 | struct xdp_buff xdp; |
761876c8 JW |
1684 | u32 act; |
1685 | ||
43b5169d | 1686 | xdp_init_buff(&xdp, buflen, &tfile->xdp_rxq); |
be9df4af | 1687 | xdp_prepare_buff(&xdp, buf, pad, len, false); |
761876c8 | 1688 | |
8ae1aff0 JW |
1689 | act = bpf_prog_run_xdp(xdp_prog, &xdp); |
1690 | if (act == XDP_REDIRECT || act == XDP_TX) { | |
59655a5b JW |
1691 | get_page(alloc_frag->page); |
1692 | alloc_frag->offset += buflen; | |
761876c8 | 1693 | } |
8ae1aff0 | 1694 | err = tun_xdp_act(tun, xdp_prog, &xdp, act); |
bee34890 WD |
1695 | if (err < 0) { |
1696 | if (act == XDP_REDIRECT || act == XDP_TX) | |
1697 | put_page(alloc_frag->page); | |
1698 | goto out; | |
1699 | } | |
1700 | ||
1a097910 | 1701 | if (err == XDP_REDIRECT) |
1d233886 | 1702 | xdp_do_flush(); |
8ae1aff0 JW |
1703 | if (err != XDP_PASS) |
1704 | goto out; | |
1705 | ||
1706 | pad = xdp.data - xdp.data_hard_start; | |
1707 | len = xdp.data_end - xdp.data; | |
761876c8 | 1708 | } |
291aeb2b JW |
1709 | rcu_read_unlock(); |
1710 | local_bh_enable(); | |
761876c8 | 1711 | |
4b663366 | 1712 | return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, pad); |
761876c8 | 1713 | |
f7053b6c | 1714 | out: |
761876c8 | 1715 | rcu_read_unlock(); |
6547e387 | 1716 | local_bh_enable(); |
761876c8 | 1717 | return NULL; |
66ccbc9c JW |
1718 | } |
1719 | ||
1da177e4 | 1720 | /* Get packet from user space buffer */ |
54f968d6 | 1721 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1722 | void *msg_control, struct iov_iter *from, |
5503fcec | 1723 | int noblock, bool more) |
1da177e4 | 1724 | { |
09640e63 | 1725 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1726 | struct sk_buff *skb; |
f5ff53b4 | 1727 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1728 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1729 | struct virtio_net_hdr gso = { 0 }; |
96f8d9ec | 1730 | int good_linear; |
0690899b MT |
1731 | int copylen; |
1732 | bool zerocopy = false; | |
1733 | int err; | |
96f84061 | 1734 | u32 rxhash = 0; |
1cfe6e93 | 1735 | int skb_xdp = 1; |
af3fb24e | 1736 | bool frags = tun_napi_frags_enabled(tfile); |
4b4f052e | 1737 | enum skb_drop_reason drop_reason; |
1da177e4 | 1738 | |
40630b82 | 1739 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1740 | if (len < sizeof(pi)) |
1da177e4 | 1741 | return -EINVAL; |
15718ea0 | 1742 | len -= sizeof(pi); |
1da177e4 | 1743 | |
cbbd26b8 | 1744 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1745 | return -EFAULT; |
1746 | } | |
1747 | ||
40630b82 | 1748 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1749 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1750 | ||
1751 | if (len < vnet_hdr_sz) | |
f43798c2 | 1752 | return -EINVAL; |
e1edab87 | 1753 | len -= vnet_hdr_sz; |
f43798c2 | 1754 | |
cbbd26b8 | 1755 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1756 | return -EFAULT; |
1757 | ||
4909122f | 1758 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1759 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1760 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1761 | |
56f0dcc5 | 1762 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1763 | return -EINVAL; |
e1edab87 | 1764 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1765 | } |
1766 | ||
40630b82 | 1767 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1768 | align += NET_IP_ALIGN; |
0eca93bc | 1769 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1770 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1771 | return -EINVAL; |
1772 | } | |
6aa20a22 | 1773 | |
96f8d9ec JW |
1774 | good_linear = SKB_MAX_HEAD(align); |
1775 | ||
88529176 | 1776 | if (msg_control) { |
f5ff53b4 AV |
1777 | struct iov_iter i = *from; |
1778 | ||
88529176 JW |
1779 | /* There are 256 bytes to be copied in skb, so there is |
1780 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1781 | * The rest of the buffer is mapped from userspace. |
1782 | */ | |
56f0dcc5 | 1783 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1784 | if (copylen > good_linear) |
1785 | copylen = good_linear; | |
3dd5c330 | 1786 | linear = copylen; |
f5ff53b4 AV |
1787 | iov_iter_advance(&i, copylen); |
1788 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1789 | zerocopy = true; |
1790 | } | |
1791 | ||
90e33d45 | 1792 | if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) { |
1cfe6e93 JW |
1793 | /* For the packet that is not easy to be processed |
1794 | * (e.g gso or jumbo packet), we will do it at after | |
1795 | * skb was created with generic XDP routine. | |
1796 | */ | |
1797 | skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp); | |
66ccbc9c | 1798 | if (IS_ERR(skb)) { |
625788b5 | 1799 | dev_core_stats_rx_dropped_inc(tun->dev); |
66ccbc9c JW |
1800 | return PTR_ERR(skb); |
1801 | } | |
761876c8 JW |
1802 | if (!skb) |
1803 | return total_len; | |
66ccbc9c JW |
1804 | } else { |
1805 | if (!zerocopy) { | |
1806 | copylen = len; | |
1807 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) | |
1808 | linear = good_linear; | |
1809 | else | |
1810 | linear = tun16_to_cpu(tun, gso.hdr_len); | |
1811 | } | |
1da177e4 | 1812 | |
90e33d45 PP |
1813 | if (frags) { |
1814 | mutex_lock(&tfile->napi_mutex); | |
1815 | skb = tun_napi_alloc_frags(tfile, copylen, from); | |
1816 | /* tun_napi_alloc_frags() enforces a layout for the skb. | |
1817 | * If zerocopy is enabled, then this layout will be | |
1818 | * overwritten by zerocopy_sg_from_iter(). | |
1819 | */ | |
1820 | zerocopy = false; | |
1821 | } else { | |
1822 | skb = tun_alloc_skb(tfile, align, copylen, linear, | |
1823 | noblock); | |
1824 | } | |
1825 | ||
66ccbc9c JW |
1826 | if (IS_ERR(skb)) { |
1827 | if (PTR_ERR(skb) != -EAGAIN) | |
625788b5 | 1828 | dev_core_stats_rx_dropped_inc(tun->dev); |
90e33d45 PP |
1829 | if (frags) |
1830 | mutex_unlock(&tfile->napi_mutex); | |
66ccbc9c JW |
1831 | return PTR_ERR(skb); |
1832 | } | |
0690899b | 1833 | |
66ccbc9c JW |
1834 | if (zerocopy) |
1835 | err = zerocopy_sg_from_iter(skb, from); | |
1836 | else | |
1837 | err = skb_copy_datagram_from_iter(skb, 0, from, len); | |
1838 | ||
1839 | if (err) { | |
4477138f | 1840 | err = -EFAULT; |
4b4f052e | 1841 | drop_reason = SKB_DROP_REASON_SKB_UCOPY_FAULT; |
4477138f | 1842 | drop: |
625788b5 | 1843 | dev_core_stats_rx_dropped_inc(tun->dev); |
4b4f052e | 1844 | kfree_skb_reason(skb, drop_reason); |
90e33d45 PP |
1845 | if (frags) { |
1846 | tfile->napi.skb = NULL; | |
1847 | mutex_unlock(&tfile->napi_mutex); | |
1848 | } | |
1849 | ||
4477138f | 1850 | return err; |
66ccbc9c | 1851 | } |
8f22757e | 1852 | } |
1da177e4 | 1853 | |
3e9e40e7 | 1854 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
497a5757 | 1855 | atomic_long_inc(&tun->rx_frame_errors); |
df10db98 | 1856 | kfree_skb(skb); |
90e33d45 PP |
1857 | if (frags) { |
1858 | tfile->napi.skb = NULL; | |
1859 | mutex_unlock(&tfile->napi_mutex); | |
1860 | } | |
1861 | ||
df10db98 PA |
1862 | return -EINVAL; |
1863 | } | |
1864 | ||
1da177e4 | 1865 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1866 | case IFF_TUN: |
1867 | if (tun->flags & IFF_NO_PI) { | |
2580c4c1 AP |
1868 | u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0; |
1869 | ||
1870 | switch (ip_version) { | |
1871 | case 4: | |
f09f7ee2 AWC |
1872 | pi.proto = htons(ETH_P_IP); |
1873 | break; | |
2580c4c1 | 1874 | case 6: |
f09f7ee2 AWC |
1875 | pi.proto = htons(ETH_P_IPV6); |
1876 | break; | |
1877 | default: | |
625788b5 | 1878 | dev_core_stats_rx_dropped_inc(tun->dev); |
f09f7ee2 AWC |
1879 | kfree_skb(skb); |
1880 | return -EINVAL; | |
1881 | } | |
1882 | } | |
1883 | ||
459a98ed | 1884 | skb_reset_mac_header(skb); |
1da177e4 | 1885 | skb->protocol = pi.proto; |
4c13eb66 | 1886 | skb->dev = tun->dev; |
1da177e4 | 1887 | break; |
40630b82 | 1888 | case IFF_TAP: |
96aa1b22 WB |
1889 | if (frags && !pskb_may_pull(skb, ETH_HLEN)) { |
1890 | err = -ENOMEM; | |
4b4f052e | 1891 | drop_reason = SKB_DROP_REASON_HDR_TRUNC; |
96aa1b22 WB |
1892 | goto drop; |
1893 | } | |
1894 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1da177e4 | 1895 | break; |
6403eab1 | 1896 | } |
1da177e4 | 1897 | |
0690899b MT |
1898 | /* copy skb_ubuf_info for callback when skb has no error */ |
1899 | if (zerocopy) { | |
9ee5e5ad | 1900 | skb_zcopy_init(skb, msg_control); |
af1cc7a2 JW |
1901 | } else if (msg_control) { |
1902 | struct ubuf_info *uarg = msg_control; | |
36177832 | 1903 | uarg->callback(NULL, uarg, false); |
0690899b MT |
1904 | } |
1905 | ||
72f65107 | 1906 | skb_reset_network_header(skb); |
d2aa125d | 1907 | skb_probe_transport_header(skb); |
3fe260e0 | 1908 | skb_record_rx_queue(skb, tfile->queue_index); |
38502af7 | 1909 | |
1cfe6e93 | 1910 | if (skb_xdp) { |
761876c8 JW |
1911 | struct bpf_prog *xdp_prog; |
1912 | int ret; | |
1913 | ||
6547e387 | 1914 | local_bh_disable(); |
761876c8 JW |
1915 | rcu_read_lock(); |
1916 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1917 | if (xdp_prog) { | |
1918 | ret = do_xdp_generic(xdp_prog, skb); | |
1919 | if (ret != XDP_PASS) { | |
1920 | rcu_read_unlock(); | |
6547e387 | 1921 | local_bh_enable(); |
1efba987 ED |
1922 | if (frags) { |
1923 | tfile->napi.skb = NULL; | |
1924 | mutex_unlock(&tfile->napi_mutex); | |
1925 | } | |
761876c8 JW |
1926 | return total_len; |
1927 | } | |
1928 | } | |
1929 | rcu_read_unlock(); | |
6547e387 | 1930 | local_bh_enable(); |
761876c8 JW |
1931 | } |
1932 | ||
cf1a1e07 PA |
1933 | /* Compute the costly rx hash only if needed for flow updates. |
1934 | * We may get a very small possibility of OOO during switching, not | |
1935 | * worth to optimize. | |
1936 | */ | |
1937 | if (!rcu_access_pointer(tun->steering_prog) && tun->numqueues > 1 && | |
1938 | !tfile->detached) | |
96f84061 | 1939 | rxhash = __skb_get_hash_symmetric(skb); |
94317099 | 1940 | |
4477138f ED |
1941 | rcu_read_lock(); |
1942 | if (unlikely(!(tun->dev->flags & IFF_UP))) { | |
1943 | err = -EIO; | |
9180bb4f | 1944 | rcu_read_unlock(); |
4b4f052e | 1945 | drop_reason = SKB_DROP_REASON_DEV_READY; |
4477138f ED |
1946 | goto drop; |
1947 | } | |
1948 | ||
90e33d45 | 1949 | if (frags) { |
96aa1b22 WB |
1950 | u32 headlen; |
1951 | ||
90e33d45 | 1952 | /* Exercise flow dissector code path. */ |
96aa1b22 WB |
1953 | skb_push(skb, ETH_HLEN); |
1954 | headlen = eth_get_headlen(tun->dev, skb->data, | |
1955 | skb_headlen(skb)); | |
90e33d45 | 1956 | |
010f245b | 1957 | if (unlikely(headlen > skb_headlen(skb))) { |
625788b5 | 1958 | dev_core_stats_rx_dropped_inc(tun->dev); |
90e33d45 | 1959 | napi_free_frags(&tfile->napi); |
4477138f | 1960 | rcu_read_unlock(); |
90e33d45 PP |
1961 | mutex_unlock(&tfile->napi_mutex); |
1962 | WARN_ON(1); | |
1963 | return -ENOMEM; | |
1964 | } | |
1965 | ||
1966 | local_bh_disable(); | |
1967 | napi_gro_frags(&tfile->napi); | |
1968 | local_bh_enable(); | |
1969 | mutex_unlock(&tfile->napi_mutex); | |
aec72f33 | 1970 | } else if (tfile->napi_enabled) { |
94317099 PP |
1971 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; |
1972 | int queue_len; | |
1973 | ||
1974 | spin_lock_bh(&queue->lock); | |
1975 | __skb_queue_tail(queue, skb); | |
1976 | queue_len = skb_queue_len(queue); | |
1977 | spin_unlock(&queue->lock); | |
1978 | ||
1979 | if (!more || queue_len > NAPI_POLL_WEIGHT) | |
1980 | napi_schedule(&tfile->napi); | |
1981 | ||
1982 | local_bh_enable(); | |
1983 | } else if (!IS_ENABLED(CONFIG_4KSTACKS)) { | |
1984 | tun_rx_batched(tun, tfile, skb, more); | |
1985 | } else { | |
3d391f65 | 1986 | netif_rx(skb); |
94317099 | 1987 | } |
4477138f | 1988 | rcu_read_unlock(); |
6aa20a22 | 1989 | |
497a5757 HK |
1990 | preempt_disable(); |
1991 | dev_sw_netstats_rx_add(tun->dev, len); | |
1992 | preempt_enable(); | |
1da177e4 | 1993 | |
96f84061 JW |
1994 | if (rxhash) |
1995 | tun_flow_update(tun, rxhash, tfile); | |
1996 | ||
0690899b | 1997 | return total_len; |
6aa20a22 | 1998 | } |
1da177e4 | 1999 | |
f5ff53b4 | 2000 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 2001 | { |
33dccbb0 | 2002 | struct file *file = iocb->ki_filp; |
54f968d6 | 2003 | struct tun_file *tfile = file->private_data; |
9484dc74 | 2004 | struct tun_struct *tun = tun_get(tfile); |
631ab46b | 2005 | ssize_t result; |
5aac0390 | 2006 | int noblock = 0; |
1da177e4 LT |
2007 | |
2008 | if (!tun) | |
2009 | return -EBADFD; | |
2010 | ||
5aac0390 JA |
2011 | if ((file->f_flags & O_NONBLOCK) || (iocb->ki_flags & IOCB_NOWAIT)) |
2012 | noblock = 1; | |
2013 | ||
2014 | result = tun_get_user(tun, tfile, NULL, from, noblock, false); | |
631ab46b EB |
2015 | |
2016 | tun_put(tun); | |
2017 | return result; | |
1da177e4 LT |
2018 | } |
2019 | ||
fc72d1d5 JW |
2020 | static ssize_t tun_put_user_xdp(struct tun_struct *tun, |
2021 | struct tun_file *tfile, | |
1ffcbc85 | 2022 | struct xdp_frame *xdp_frame, |
fc72d1d5 JW |
2023 | struct iov_iter *iter) |
2024 | { | |
2025 | int vnet_hdr_sz = 0; | |
1ffcbc85 | 2026 | size_t size = xdp_frame->len; |
fc72d1d5 JW |
2027 | size_t ret; |
2028 | ||
2029 | if (tun->flags & IFF_VNET_HDR) { | |
2030 | struct virtio_net_hdr gso = { 0 }; | |
2031 | ||
2032 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); | |
2033 | if (unlikely(iov_iter_count(iter) < vnet_hdr_sz)) | |
2034 | return -EINVAL; | |
2035 | if (unlikely(copy_to_iter(&gso, sizeof(gso), iter) != | |
2036 | sizeof(gso))) | |
2037 | return -EFAULT; | |
2038 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
2039 | } | |
2040 | ||
1ffcbc85 | 2041 | ret = copy_to_iter(xdp_frame->data, size, iter) + vnet_hdr_sz; |
fc72d1d5 | 2042 | |
497a5757 HK |
2043 | preempt_disable(); |
2044 | dev_sw_netstats_tx_add(tun->dev, 1, ret); | |
2045 | preempt_enable(); | |
fc72d1d5 JW |
2046 | |
2047 | return ret; | |
2048 | } | |
2049 | ||
1da177e4 | 2050 | /* Put packet to the user space buffer */ |
6f7c156c | 2051 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 2052 | struct tun_file *tfile, |
6f7c156c | 2053 | struct sk_buff *skb, |
e0b46d0e | 2054 | struct iov_iter *iter) |
1da177e4 LT |
2055 | { |
2056 | struct tun_pi pi = { 0, skb->protocol }; | |
e0b46d0e | 2057 | ssize_t total; |
8c847d25 | 2058 | int vlan_offset = 0; |
a8f9bfdf | 2059 | int vlan_hlen = 0; |
2eb783c4 | 2060 | int vnet_hdr_sz = 0; |
a8f9bfdf | 2061 | |
df8a39de | 2062 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 2063 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 2064 | |
40630b82 | 2065 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 2066 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 2067 | |
e0b46d0e HX |
2068 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
2069 | ||
40630b82 | 2070 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 2071 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
2072 | return -EINVAL; |
2073 | ||
e0b46d0e HX |
2074 | total += sizeof(pi); |
2075 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
2076 | /* Packet will be striped */ |
2077 | pi.flags |= TUN_PKT_STRIP; | |
2078 | } | |
6aa20a22 | 2079 | |
e0b46d0e | 2080 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 2081 | return -EFAULT; |
6aa20a22 | 2082 | } |
1da177e4 | 2083 | |
2eb783c4 | 2084 | if (vnet_hdr_sz) { |
9403cd7c | 2085 | struct virtio_net_hdr gso; |
34166093 | 2086 | |
e0b46d0e | 2087 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
2088 | return -EINVAL; |
2089 | ||
3e9e40e7 | 2090 | if (virtio_net_hdr_from_skb(skb, &gso, |
fd3a8862 WB |
2091 | tun_is_little_endian(tun), true, |
2092 | vlan_hlen)) { | |
f43798c2 | 2093 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
2094 | pr_err("unexpected GSO type: " |
2095 | "0x%x, gso_size %d, hdr_len %d\n", | |
2096 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
2097 | tun16_to_cpu(tun, gso.hdr_len)); | |
2098 | print_hex_dump(KERN_ERR, "tun: ", | |
2099 | DUMP_PREFIX_NONE, | |
2100 | 16, 1, skb->head, | |
2101 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
2102 | WARN_ON_ONCE(1); | |
2103 | return -EINVAL; | |
2104 | } | |
f43798c2 | 2105 | |
e0b46d0e | 2106 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 2107 | return -EFAULT; |
8c847d25 JW |
2108 | |
2109 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
2110 | } |
2111 | ||
a8f9bfdf | 2112 | if (vlan_hlen) { |
e0b46d0e | 2113 | int ret; |
aff3d70a | 2114 | struct veth veth; |
6680ec68 JW |
2115 | |
2116 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 2117 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
2118 | |
2119 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 2120 | |
e0b46d0e HX |
2121 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
2122 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
2123 | goto done; |
2124 | ||
e0b46d0e HX |
2125 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
2126 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
2127 | goto done; |
2128 | } | |
1da177e4 | 2129 | |
e0b46d0e | 2130 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 2131 | |
6680ec68 | 2132 | done: |
608b9977 | 2133 | /* caller is in process context, */ |
497a5757 HK |
2134 | preempt_disable(); |
2135 | dev_sw_netstats_tx_add(tun->dev, 1, skb->len + vlan_hlen); | |
2136 | preempt_enable(); | |
1da177e4 LT |
2137 | |
2138 | return total; | |
2139 | } | |
2140 | ||
fc72d1d5 | 2141 | static void *tun_ring_recv(struct tun_file *tfile, int noblock, int *err) |
1576d986 JW |
2142 | { |
2143 | DECLARE_WAITQUEUE(wait, current); | |
fc72d1d5 | 2144 | void *ptr = NULL; |
f48cc6b2 | 2145 | int error = 0; |
1576d986 | 2146 | |
fc72d1d5 JW |
2147 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2148 | if (ptr) | |
1576d986 JW |
2149 | goto out; |
2150 | if (noblock) { | |
f48cc6b2 | 2151 | error = -EAGAIN; |
1576d986 JW |
2152 | goto out; |
2153 | } | |
2154 | ||
333f7909 | 2155 | add_wait_queue(&tfile->socket.wq.wait, &wait); |
1576d986 JW |
2156 | |
2157 | while (1) { | |
71828b22 | 2158 | set_current_state(TASK_INTERRUPTIBLE); |
fc72d1d5 JW |
2159 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2160 | if (ptr) | |
1576d986 JW |
2161 | break; |
2162 | if (signal_pending(current)) { | |
f48cc6b2 | 2163 | error = -ERESTARTSYS; |
1576d986 JW |
2164 | break; |
2165 | } | |
2166 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 2167 | error = -EFAULT; |
1576d986 JW |
2168 | break; |
2169 | } | |
2170 | ||
2171 | schedule(); | |
2172 | } | |
2173 | ||
ecef67cb | 2174 | __set_current_state(TASK_RUNNING); |
333f7909 | 2175 | remove_wait_queue(&tfile->socket.wq.wait, &wait); |
1576d986 JW |
2176 | |
2177 | out: | |
f48cc6b2 | 2178 | *err = error; |
fc72d1d5 | 2179 | return ptr; |
1576d986 JW |
2180 | } |
2181 | ||
54f968d6 | 2182 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 | 2183 | struct iov_iter *to, |
fc72d1d5 | 2184 | int noblock, void *ptr) |
1da177e4 | 2185 | { |
9b067034 | 2186 | ssize_t ret; |
1576d986 | 2187 | int err; |
1da177e4 | 2188 | |
c33ee15b | 2189 | if (!iov_iter_count(to)) { |
fc72d1d5 | 2190 | tun_ptr_free(ptr); |
9b067034 | 2191 | return 0; |
c33ee15b | 2192 | } |
1da177e4 | 2193 | |
fc72d1d5 | 2194 | if (!ptr) { |
ac77cfd4 | 2195 | /* Read frames from ring */ |
fc72d1d5 JW |
2196 | ptr = tun_ring_recv(tfile, noblock, &err); |
2197 | if (!ptr) | |
ac77cfd4 JW |
2198 | return err; |
2199 | } | |
e0b46d0e | 2200 | |
1ffcbc85 JDB |
2201 | if (tun_is_xdp_frame(ptr)) { |
2202 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2203 | |
1ffcbc85 | 2204 | ret = tun_put_user_xdp(tun, tfile, xdpf, to); |
03993094 | 2205 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
2206 | } else { |
2207 | struct sk_buff *skb = ptr; | |
2208 | ||
2209 | ret = tun_put_user(tun, tfile, skb, to); | |
2210 | if (unlikely(ret < 0)) | |
2211 | kfree_skb(skb); | |
2212 | else | |
2213 | consume_skb(skb); | |
2214 | } | |
1da177e4 | 2215 | |
05c2828c MT |
2216 | return ret; |
2217 | } | |
2218 | ||
9b067034 | 2219 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
2220 | { |
2221 | struct file *file = iocb->ki_filp; | |
2222 | struct tun_file *tfile = file->private_data; | |
9484dc74 | 2223 | struct tun_struct *tun = tun_get(tfile); |
9b067034 | 2224 | ssize_t len = iov_iter_count(to), ret; |
5aac0390 | 2225 | int noblock = 0; |
05c2828c MT |
2226 | |
2227 | if (!tun) | |
2228 | return -EBADFD; | |
5aac0390 JA |
2229 | |
2230 | if ((file->f_flags & O_NONBLOCK) || (iocb->ki_flags & IOCB_NOWAIT)) | |
2231 | noblock = 1; | |
2232 | ||
2233 | ret = tun_do_read(tun, tfile, to, noblock, NULL); | |
42404c09 | 2234 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
2235 | if (ret > 0) |
2236 | iocb->ki_pos = ret; | |
631ab46b | 2237 | tun_put(tun); |
1da177e4 LT |
2238 | return ret; |
2239 | } | |
2240 | ||
cd5681d7 | 2241 | static void tun_prog_free(struct rcu_head *rcu) |
96f84061 | 2242 | { |
cd5681d7 | 2243 | struct tun_prog *prog = container_of(rcu, struct tun_prog, rcu); |
96f84061 JW |
2244 | |
2245 | bpf_prog_destroy(prog->prog); | |
2246 | kfree(prog); | |
2247 | } | |
2248 | ||
9d6474e4 JW |
2249 | static int __tun_set_ebpf(struct tun_struct *tun, |
2250 | struct tun_prog __rcu **prog_p, | |
cd5681d7 | 2251 | struct bpf_prog *prog) |
96f84061 | 2252 | { |
cd5681d7 | 2253 | struct tun_prog *old, *new = NULL; |
96f84061 JW |
2254 | |
2255 | if (prog) { | |
2256 | new = kmalloc(sizeof(*new), GFP_KERNEL); | |
2257 | if (!new) | |
2258 | return -ENOMEM; | |
2259 | new->prog = prog; | |
2260 | } | |
2261 | ||
124da8f6 | 2262 | spin_lock_bh(&tun->lock); |
cd5681d7 | 2263 | old = rcu_dereference_protected(*prog_p, |
124da8f6 | 2264 | lockdep_is_held(&tun->lock)); |
cd5681d7 | 2265 | rcu_assign_pointer(*prog_p, new); |
124da8f6 | 2266 | spin_unlock_bh(&tun->lock); |
96f84061 JW |
2267 | |
2268 | if (old) | |
cd5681d7 | 2269 | call_rcu(&old->rcu, tun_prog_free); |
96f84061 JW |
2270 | |
2271 | return 0; | |
2272 | } | |
2273 | ||
96442e42 JW |
2274 | static void tun_free_netdev(struct net_device *dev) |
2275 | { | |
2276 | struct tun_struct *tun = netdev_priv(dev); | |
2277 | ||
4008e97f | 2278 | BUG_ON(!(list_empty(&tun->disabled))); |
11fc7d5a | 2279 | |
497a5757 | 2280 | free_percpu(dev->tstats); |
96442e42 | 2281 | tun_flow_uninit(tun); |
5dbbaf2d | 2282 | security_tun_dev_free_security(tun->security); |
cd5681d7 | 2283 | __tun_set_ebpf(tun, &tun->steering_prog, NULL); |
aff3d70a | 2284 | __tun_set_ebpf(tun, &tun->filter_prog, NULL); |
96442e42 JW |
2285 | } |
2286 | ||
1da177e4 LT |
2287 | static void tun_setup(struct net_device *dev) |
2288 | { | |
2289 | struct tun_struct *tun = netdev_priv(dev); | |
2290 | ||
0625c883 EB |
2291 | tun->owner = INVALID_UID; |
2292 | tun->group = INVALID_GID; | |
4e24f2dd | 2293 | tun_default_link_ksettings(dev, &tun->link_ksettings); |
1da177e4 | 2294 | |
1da177e4 | 2295 | dev->ethtool_ops = &tun_ethtool_ops; |
cf124db5 DM |
2296 | dev->needs_free_netdev = true; |
2297 | dev->priv_destructor = tun_free_netdev; | |
016adb72 JW |
2298 | /* We prefer our own queue length */ |
2299 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
2300 | } |
2301 | ||
f019a7a5 EB |
2302 | /* Trivial set of netlink ops to allow deleting tun or tap |
2303 | * device with netlink. | |
2304 | */ | |
a8b8a889 MS |
2305 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[], |
2306 | struct netlink_ext_ack *extack) | |
f019a7a5 | 2307 | { |
35b827b6 ND |
2308 | NL_SET_ERR_MSG(extack, |
2309 | "tun/tap creation via rtnetlink is not supported."); | |
2310 | return -EOPNOTSUPP; | |
f019a7a5 EB |
2311 | } |
2312 | ||
1ec010e7 SD |
2313 | static size_t tun_get_size(const struct net_device *dev) |
2314 | { | |
2315 | BUILD_BUG_ON(sizeof(u32) != sizeof(uid_t)); | |
2316 | BUILD_BUG_ON(sizeof(u32) != sizeof(gid_t)); | |
2317 | ||
2318 | return nla_total_size(sizeof(uid_t)) + /* OWNER */ | |
2319 | nla_total_size(sizeof(gid_t)) + /* GROUP */ | |
2320 | nla_total_size(sizeof(u8)) + /* TYPE */ | |
2321 | nla_total_size(sizeof(u8)) + /* PI */ | |
2322 | nla_total_size(sizeof(u8)) + /* VNET_HDR */ | |
2323 | nla_total_size(sizeof(u8)) + /* PERSIST */ | |
2324 | nla_total_size(sizeof(u8)) + /* MULTI_QUEUE */ | |
2325 | nla_total_size(sizeof(u32)) + /* NUM_QUEUES */ | |
2326 | nla_total_size(sizeof(u32)) + /* NUM_DISABLED_QUEUES */ | |
2327 | 0; | |
2328 | } | |
2329 | ||
2330 | static int tun_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
2331 | { | |
2332 | struct tun_struct *tun = netdev_priv(dev); | |
2333 | ||
2334 | if (nla_put_u8(skb, IFLA_TUN_TYPE, tun->flags & TUN_TYPE_MASK)) | |
2335 | goto nla_put_failure; | |
2336 | if (uid_valid(tun->owner) && | |
2337 | nla_put_u32(skb, IFLA_TUN_OWNER, | |
2338 | from_kuid_munged(current_user_ns(), tun->owner))) | |
2339 | goto nla_put_failure; | |
2340 | if (gid_valid(tun->group) && | |
2341 | nla_put_u32(skb, IFLA_TUN_GROUP, | |
2342 | from_kgid_munged(current_user_ns(), tun->group))) | |
2343 | goto nla_put_failure; | |
2344 | if (nla_put_u8(skb, IFLA_TUN_PI, !(tun->flags & IFF_NO_PI))) | |
2345 | goto nla_put_failure; | |
2346 | if (nla_put_u8(skb, IFLA_TUN_VNET_HDR, !!(tun->flags & IFF_VNET_HDR))) | |
2347 | goto nla_put_failure; | |
2348 | if (nla_put_u8(skb, IFLA_TUN_PERSIST, !!(tun->flags & IFF_PERSIST))) | |
2349 | goto nla_put_failure; | |
2350 | if (nla_put_u8(skb, IFLA_TUN_MULTI_QUEUE, | |
2351 | !!(tun->flags & IFF_MULTI_QUEUE))) | |
2352 | goto nla_put_failure; | |
2353 | if (tun->flags & IFF_MULTI_QUEUE) { | |
2354 | if (nla_put_u32(skb, IFLA_TUN_NUM_QUEUES, tun->numqueues)) | |
2355 | goto nla_put_failure; | |
2356 | if (nla_put_u32(skb, IFLA_TUN_NUM_DISABLED_QUEUES, | |
2357 | tun->numdisabled)) | |
2358 | goto nla_put_failure; | |
2359 | } | |
2360 | ||
2361 | return 0; | |
2362 | ||
2363 | nla_put_failure: | |
2364 | return -EMSGSIZE; | |
2365 | } | |
2366 | ||
f019a7a5 EB |
2367 | static struct rtnl_link_ops tun_link_ops __read_mostly = { |
2368 | .kind = DRV_NAME, | |
2369 | .priv_size = sizeof(struct tun_struct), | |
2370 | .setup = tun_setup, | |
2371 | .validate = tun_validate, | |
1ec010e7 SD |
2372 | .get_size = tun_get_size, |
2373 | .fill_info = tun_fill_info, | |
f019a7a5 EB |
2374 | }; |
2375 | ||
33dccbb0 HX |
2376 | static void tun_sock_write_space(struct sock *sk) |
2377 | { | |
54f968d6 | 2378 | struct tun_file *tfile; |
43815482 | 2379 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
2380 | |
2381 | if (!sock_writeable(sk)) | |
2382 | return; | |
2383 | ||
9cd3e072 | 2384 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
2385 | return; |
2386 | ||
43815482 ED |
2387 | wqueue = sk_sleep(sk); |
2388 | if (wqueue && waitqueue_active(wqueue)) | |
a9a08845 LT |
2389 | wake_up_interruptible_sync_poll(wqueue, EPOLLOUT | |
2390 | EPOLLWRNORM | EPOLLWRBAND); | |
c722c625 | 2391 | |
54f968d6 JW |
2392 | tfile = container_of(sk, struct tun_file, sk); |
2393 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
2394 | } |
2395 | ||
f9e06c45 JW |
2396 | static void tun_put_page(struct tun_page *tpage) |
2397 | { | |
2398 | if (tpage->page) | |
2399 | __page_frag_cache_drain(tpage->page, tpage->count); | |
2400 | } | |
2401 | ||
043d222f JW |
2402 | static int tun_xdp_one(struct tun_struct *tun, |
2403 | struct tun_file *tfile, | |
f9e06c45 JW |
2404 | struct xdp_buff *xdp, int *flush, |
2405 | struct tun_page *tpage) | |
043d222f | 2406 | { |
4e4b08e5 | 2407 | unsigned int datasize = xdp->data_end - xdp->data; |
043d222f JW |
2408 | struct tun_xdp_hdr *hdr = xdp->data_hard_start; |
2409 | struct virtio_net_hdr *gso = &hdr->gso; | |
043d222f JW |
2410 | struct bpf_prog *xdp_prog; |
2411 | struct sk_buff *skb = NULL; | |
fb3f9037 | 2412 | struct sk_buff_head *queue; |
043d222f JW |
2413 | u32 rxhash = 0, act; |
2414 | int buflen = hdr->buflen; | |
fb3f9037 | 2415 | int ret = 0; |
043d222f | 2416 | bool skb_xdp = false; |
f9e06c45 | 2417 | struct page *page; |
043d222f JW |
2418 | |
2419 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
2420 | if (xdp_prog) { | |
2421 | if (gso->gso_type) { | |
2422 | skb_xdp = true; | |
2423 | goto build; | |
2424 | } | |
43b5169d LB |
2425 | |
2426 | xdp_init_buff(xdp, buflen, &tfile->xdp_rxq); | |
043d222f | 2427 | xdp_set_data_meta_invalid(xdp); |
043d222f JW |
2428 | |
2429 | act = bpf_prog_run_xdp(xdp_prog, xdp); | |
fb3f9037 HH |
2430 | ret = tun_xdp_act(tun, xdp_prog, xdp, act); |
2431 | if (ret < 0) { | |
043d222f | 2432 | put_page(virt_to_head_page(xdp->data)); |
fb3f9037 | 2433 | return ret; |
043d222f JW |
2434 | } |
2435 | ||
fb3f9037 | 2436 | switch (ret) { |
043d222f JW |
2437 | case XDP_REDIRECT: |
2438 | *flush = true; | |
df561f66 | 2439 | fallthrough; |
043d222f JW |
2440 | case XDP_TX: |
2441 | return 0; | |
2442 | case XDP_PASS: | |
2443 | break; | |
2444 | default: | |
f9e06c45 JW |
2445 | page = virt_to_head_page(xdp->data); |
2446 | if (tpage->page == page) { | |
2447 | ++tpage->count; | |
2448 | } else { | |
2449 | tun_put_page(tpage); | |
2450 | tpage->page = page; | |
2451 | tpage->count = 1; | |
2452 | } | |
043d222f JW |
2453 | return 0; |
2454 | } | |
2455 | } | |
2456 | ||
2457 | build: | |
2458 | skb = build_skb(xdp->data_hard_start, buflen); | |
2459 | if (!skb) { | |
fb3f9037 | 2460 | ret = -ENOMEM; |
043d222f JW |
2461 | goto out; |
2462 | } | |
2463 | ||
2464 | skb_reserve(skb, xdp->data - xdp->data_hard_start); | |
2465 | skb_put(skb, xdp->data_end - xdp->data); | |
2466 | ||
2467 | if (virtio_net_hdr_to_skb(skb, gso, tun_is_little_endian(tun))) { | |
497a5757 | 2468 | atomic_long_inc(&tun->rx_frame_errors); |
043d222f | 2469 | kfree_skb(skb); |
fb3f9037 | 2470 | ret = -EINVAL; |
043d222f JW |
2471 | goto out; |
2472 | } | |
2473 | ||
2474 | skb->protocol = eth_type_trans(skb, tun->dev); | |
2475 | skb_reset_network_header(skb); | |
d2aa125d | 2476 | skb_probe_transport_header(skb); |
3fe260e0 | 2477 | skb_record_rx_queue(skb, tfile->queue_index); |
043d222f JW |
2478 | |
2479 | if (skb_xdp) { | |
fb3f9037 HH |
2480 | ret = do_xdp_generic(xdp_prog, skb); |
2481 | if (ret != XDP_PASS) { | |
2482 | ret = 0; | |
043d222f | 2483 | goto out; |
fb3f9037 | 2484 | } |
043d222f JW |
2485 | } |
2486 | ||
f29eb2a9 PA |
2487 | if (!rcu_dereference(tun->steering_prog) && tun->numqueues > 1 && |
2488 | !tfile->detached) | |
043d222f JW |
2489 | rxhash = __skb_get_hash_symmetric(skb); |
2490 | ||
fb3f9037 HH |
2491 | if (tfile->napi_enabled) { |
2492 | queue = &tfile->sk.sk_write_queue; | |
2493 | spin_lock(&queue->lock); | |
2494 | __skb_queue_tail(queue, skb); | |
2495 | spin_unlock(&queue->lock); | |
2496 | ret = 1; | |
2497 | } else { | |
2498 | netif_receive_skb(skb); | |
2499 | ret = 0; | |
2500 | } | |
043d222f | 2501 | |
497a5757 | 2502 | /* No need to disable preemption here since this function is |
6342ca64 PB |
2503 | * always called with bh disabled |
2504 | */ | |
497a5757 | 2505 | dev_sw_netstats_rx_add(tun->dev, datasize); |
043d222f JW |
2506 | |
2507 | if (rxhash) | |
2508 | tun_flow_update(tun, rxhash, tfile); | |
2509 | ||
2510 | out: | |
fb3f9037 | 2511 | return ret; |
043d222f JW |
2512 | } |
2513 | ||
1b784140 | 2514 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 2515 | { |
043d222f | 2516 | int ret, i; |
54f968d6 | 2517 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2518 | struct tun_struct *tun = tun_get(tfile); |
fe8dd45b | 2519 | struct tun_msg_ctl *ctl = m->msg_control; |
043d222f | 2520 | struct xdp_buff *xdp; |
54f968d6 JW |
2521 | |
2522 | if (!tun) | |
2523 | return -EBADFD; | |
f5ff53b4 | 2524 | |
74a335a0 HH |
2525 | if (m->msg_controllen == sizeof(struct tun_msg_ctl) && |
2526 | ctl && ctl->type == TUN_MSG_PTR) { | |
6f0271d9 | 2527 | struct tun_page tpage; |
043d222f | 2528 | int n = ctl->num; |
fb3f9037 | 2529 | int flush = 0, queued = 0; |
043d222f | 2530 | |
6f0271d9 DM |
2531 | memset(&tpage, 0, sizeof(tpage)); |
2532 | ||
043d222f JW |
2533 | local_bh_disable(); |
2534 | rcu_read_lock(); | |
2535 | ||
2536 | for (i = 0; i < n; i++) { | |
2537 | xdp = &((struct xdp_buff *)ctl->ptr)[i]; | |
fb3f9037 HH |
2538 | ret = tun_xdp_one(tun, tfile, xdp, &flush, &tpage); |
2539 | if (ret > 0) | |
2540 | queued += ret; | |
043d222f JW |
2541 | } |
2542 | ||
2543 | if (flush) | |
1d233886 | 2544 | xdp_do_flush(); |
043d222f | 2545 | |
fb3f9037 HH |
2546 | if (tfile->napi_enabled && queued > 0) |
2547 | napi_schedule(&tfile->napi); | |
2548 | ||
043d222f JW |
2549 | rcu_read_unlock(); |
2550 | local_bh_enable(); | |
2551 | ||
f9e06c45 JW |
2552 | tun_put_page(&tpage); |
2553 | ||
043d222f JW |
2554 | ret = total_len; |
2555 | goto out; | |
2556 | } | |
fe8dd45b JW |
2557 | |
2558 | ret = tun_get_user(tun, tfile, ctl ? ctl->ptr : NULL, &m->msg_iter, | |
5503fcec JW |
2559 | m->msg_flags & MSG_DONTWAIT, |
2560 | m->msg_flags & MSG_MORE); | |
043d222f | 2561 | out: |
54f968d6 JW |
2562 | tun_put(tun); |
2563 | return ret; | |
05c2828c MT |
2564 | } |
2565 | ||
1b784140 | 2566 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
2567 | int flags) |
2568 | { | |
54f968d6 | 2569 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2570 | struct tun_struct *tun = tun_get(tfile); |
fc72d1d5 | 2571 | void *ptr = m->msg_control; |
05c2828c | 2572 | int ret; |
54f968d6 | 2573 | |
c33ee15b WX |
2574 | if (!tun) { |
2575 | ret = -EBADFD; | |
fc72d1d5 | 2576 | goto out_free; |
c33ee15b | 2577 | } |
54f968d6 | 2578 | |
eda29772 | 2579 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 | 2580 | ret = -EINVAL; |
c33ee15b | 2581 | goto out_put_tun; |
3811ae76 | 2582 | } |
eda29772 RC |
2583 | if (flags & MSG_ERRQUEUE) { |
2584 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
2585 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
2586 | goto out; | |
2587 | } | |
fc72d1d5 | 2588 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, ptr); |
87897931 | 2589 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
2590 | m->msg_flags |= MSG_TRUNC; |
2591 | ret = flags & MSG_TRUNC ? ret : total_len; | |
2592 | } | |
3811ae76 | 2593 | out: |
54f968d6 | 2594 | tun_put(tun); |
05c2828c | 2595 | return ret; |
c33ee15b WX |
2596 | |
2597 | out_put_tun: | |
2598 | tun_put(tun); | |
fc72d1d5 JW |
2599 | out_free: |
2600 | tun_ptr_free(ptr); | |
c33ee15b | 2601 | return ret; |
05c2828c MT |
2602 | } |
2603 | ||
fc72d1d5 JW |
2604 | static int tun_ptr_peek_len(void *ptr) |
2605 | { | |
2606 | if (likely(ptr)) { | |
1ffcbc85 JDB |
2607 | if (tun_is_xdp_frame(ptr)) { |
2608 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2609 | |
1ffcbc85 | 2610 | return xdpf->len; |
fc72d1d5 JW |
2611 | } |
2612 | return __skb_array_len_with_tag(ptr); | |
2613 | } else { | |
2614 | return 0; | |
2615 | } | |
2616 | } | |
2617 | ||
1576d986 JW |
2618 | static int tun_peek_len(struct socket *sock) |
2619 | { | |
2620 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
2621 | struct tun_struct *tun; | |
2622 | int ret = 0; | |
2623 | ||
9484dc74 | 2624 | tun = tun_get(tfile); |
1576d986 JW |
2625 | if (!tun) |
2626 | return 0; | |
2627 | ||
fc72d1d5 | 2628 | ret = PTR_RING_PEEK_CALL(&tfile->tx_ring, tun_ptr_peek_len); |
1576d986 JW |
2629 | tun_put(tun); |
2630 | ||
2631 | return ret; | |
2632 | } | |
2633 | ||
05c2828c MT |
2634 | /* Ops structure to mimic raw sockets with tun */ |
2635 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 2636 | .peek_len = tun_peek_len, |
05c2828c MT |
2637 | .sendmsg = tun_sendmsg, |
2638 | .recvmsg = tun_recvmsg, | |
2639 | }; | |
2640 | ||
33dccbb0 HX |
2641 | static struct proto tun_proto = { |
2642 | .name = "tun", | |
2643 | .owner = THIS_MODULE, | |
54f968d6 | 2644 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 2645 | }; |
f019a7a5 | 2646 | |
980c9e8c DW |
2647 | static int tun_flags(struct tun_struct *tun) |
2648 | { | |
031f5e03 | 2649 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
2650 | } |
2651 | ||
bc6d076d | 2652 | static ssize_t tun_flags_show(struct device *dev, struct device_attribute *attr, |
980c9e8c DW |
2653 | char *buf) |
2654 | { | |
2655 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
2656 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
2657 | } | |
2658 | ||
bc6d076d Y |
2659 | static ssize_t owner_show(struct device *dev, struct device_attribute *attr, |
2660 | char *buf) | |
980c9e8c DW |
2661 | { |
2662 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2663 | return uid_valid(tun->owner)? |
2664 | sprintf(buf, "%u\n", | |
2665 | from_kuid_munged(current_user_ns(), tun->owner)): | |
2666 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2667 | } |
2668 | ||
bc6d076d Y |
2669 | static ssize_t group_show(struct device *dev, struct device_attribute *attr, |
2670 | char *buf) | |
980c9e8c DW |
2671 | { |
2672 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2673 | return gid_valid(tun->group) ? |
2674 | sprintf(buf, "%u\n", | |
2675 | from_kgid_munged(current_user_ns(), tun->group)): | |
2676 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2677 | } |
2678 | ||
bc6d076d Y |
2679 | static DEVICE_ATTR_RO(tun_flags); |
2680 | static DEVICE_ATTR_RO(owner); | |
2681 | static DEVICE_ATTR_RO(group); | |
980c9e8c | 2682 | |
c4d33e24 TI |
2683 | static struct attribute *tun_dev_attrs[] = { |
2684 | &dev_attr_tun_flags.attr, | |
2685 | &dev_attr_owner.attr, | |
2686 | &dev_attr_group.attr, | |
2687 | NULL | |
2688 | }; | |
2689 | ||
2690 | static const struct attribute_group tun_attr_group = { | |
2691 | .attrs = tun_dev_attrs | |
2692 | }; | |
2693 | ||
d647a591 | 2694 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
2695 | { |
2696 | struct tun_struct *tun; | |
54f968d6 | 2697 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2698 | struct net_device *dev; |
2699 | int err; | |
2700 | ||
7c0c3b1a JW |
2701 | if (tfile->detached) |
2702 | return -EINVAL; | |
2703 | ||
90e33d45 PP |
2704 | if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) { |
2705 | if (!capable(CAP_NET_ADMIN)) | |
2706 | return -EPERM; | |
2707 | ||
2708 | if (!(ifr->ifr_flags & IFF_NAPI) || | |
2709 | (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP) | |
2710 | return -EINVAL; | |
2711 | } | |
2712 | ||
74a3e5a7 EB |
2713 | dev = __dev_get_by_name(net, ifr->ifr_name); |
2714 | if (dev) { | |
f85ba780 DW |
2715 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
2716 | return -EBUSY; | |
74a3e5a7 EB |
2717 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
2718 | tun = netdev_priv(dev); | |
2719 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
2720 | tun = netdev_priv(dev); | |
2721 | else | |
2722 | return -EINVAL; | |
2723 | ||
8e6d91ae | 2724 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 2725 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
2726 | return -EINVAL; |
2727 | ||
cde8b15f | 2728 | if (tun_not_capable(tun)) |
2b980dbd | 2729 | return -EPERM; |
5dbbaf2d | 2730 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
2731 | if (err < 0) |
2732 | return err; | |
2733 | ||
94317099 | 2734 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER, |
af3fb24e | 2735 | ifr->ifr_flags & IFF_NAPI, |
77f22f92 | 2736 | ifr->ifr_flags & IFF_NAPI_FRAGS, true); |
a7385ba2 EB |
2737 | if (err < 0) |
2738 | return err; | |
4008e97f | 2739 | |
40630b82 | 2740 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
2741 | (tun->numqueues + tun->numdisabled > 1)) { |
2742 | /* One or more queue has already been attached, no need | |
2743 | * to initialize the device again. | |
2744 | */ | |
83c1f36f | 2745 | netdev_state_change(dev); |
e8dbad66 JW |
2746 | return 0; |
2747 | } | |
9fffc5c6 SD |
2748 | |
2749 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
2750 | (ifr->ifr_flags & TUN_FEATURES); | |
83c1f36f SD |
2751 | |
2752 | netdev_state_change(dev); | |
2753 | } else { | |
1da177e4 LT |
2754 | char *name; |
2755 | unsigned long flags = 0; | |
edfb6a14 JW |
2756 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
2757 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 2758 | |
c260b772 | 2759 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 2760 | return -EPERM; |
2b980dbd PM |
2761 | err = security_tun_dev_create(); |
2762 | if (err < 0) | |
2763 | return err; | |
ca6bb5d7 | 2764 | |
1da177e4 LT |
2765 | /* Set dev type */ |
2766 | if (ifr->ifr_flags & IFF_TUN) { | |
2767 | /* TUN device */ | |
40630b82 | 2768 | flags |= IFF_TUN; |
1da177e4 LT |
2769 | name = "tun%d"; |
2770 | } else if (ifr->ifr_flags & IFF_TAP) { | |
2771 | /* TAP device */ | |
40630b82 | 2772 | flags |= IFF_TAP; |
1da177e4 | 2773 | name = "tap%d"; |
6aa20a22 | 2774 | } else |
36989b90 | 2775 | return -EINVAL; |
6aa20a22 | 2776 | |
1da177e4 LT |
2777 | if (*ifr->ifr_name) |
2778 | name = ifr->ifr_name; | |
2779 | ||
c8d68e6b | 2780 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
2781 | NET_NAME_UNKNOWN, tun_setup, queues, |
2782 | queues); | |
edfb6a14 | 2783 | |
1da177e4 LT |
2784 | if (!dev) |
2785 | return -ENOMEM; | |
2786 | ||
fc54c658 | 2787 | dev_net_set(dev, net); |
f019a7a5 | 2788 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 2789 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 2790 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 2791 | |
1da177e4 LT |
2792 | tun = netdev_priv(dev); |
2793 | tun->dev = dev; | |
2794 | tun->flags = flags; | |
f271b2cc | 2795 | tun->txflt.count = 0; |
d9d52b51 | 2796 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 2797 | |
eaea34b2 | 2798 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
2799 | tun->filter_attached = false; |
2800 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 2801 | tun->rx_batched = 0; |
96f84061 | 2802 | RCU_INIT_POINTER(tun->steering_prog, NULL); |
33dccbb0 | 2803 | |
158b515f GK |
2804 | tun->ifr = ifr; |
2805 | tun->file = file; | |
608b9977 | 2806 | |
158b515f | 2807 | tun_net_initialize(dev); |
eb0fb363 | 2808 | |
1da177e4 | 2809 | err = register_netdevice(tun->dev); |
158b515f GK |
2810 | if (err < 0) { |
2811 | free_netdev(dev); | |
2812 | return err; | |
2813 | } | |
c2e315b8 | 2814 | /* free_netdev() won't check refcnt, to avoid race |
77f22f92 YY |
2815 | * with dev_put() we need publish tun after registration. |
2816 | */ | |
2817 | rcu_assign_pointer(tfile->tun, tun); | |
1da177e4 LT |
2818 | } |
2819 | ||
af668b3c MT |
2820 | netif_carrier_on(tun->dev); |
2821 | ||
e35259a9 MK |
2822 | /* Make sure persistent devices do not get stuck in |
2823 | * xoff state. | |
2824 | */ | |
2825 | if (netif_running(tun->dev)) | |
c8d68e6b | 2826 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 2827 | |
1da177e4 LT |
2828 | strcpy(ifr->ifr_name, tun->dev->name); |
2829 | return 0; | |
1da177e4 LT |
2830 | } |
2831 | ||
12132768 | 2832 | static void tun_get_iff(struct tun_struct *tun, struct ifreq *ifr) |
e3b99556 | 2833 | { |
e3b99556 MM |
2834 | strcpy(ifr->ifr_name, tun->dev->name); |
2835 | ||
980c9e8c | 2836 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 2837 | |
e3b99556 MM |
2838 | } |
2839 | ||
5228ddc9 RR |
2840 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
2841 | * privs required. */ | |
88255375 | 2842 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 2843 | { |
c8f44aff | 2844 | netdev_features_t features = 0; |
5228ddc9 RR |
2845 | |
2846 | if (arg & TUN_F_CSUM) { | |
88255375 | 2847 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
2848 | arg &= ~TUN_F_CSUM; |
2849 | ||
2850 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
2851 | if (arg & TUN_F_TSO_ECN) { | |
2852 | features |= NETIF_F_TSO_ECN; | |
2853 | arg &= ~TUN_F_TSO_ECN; | |
2854 | } | |
2855 | if (arg & TUN_F_TSO4) | |
2856 | features |= NETIF_F_TSO; | |
2857 | if (arg & TUN_F_TSO6) | |
2858 | features |= NETIF_F_TSO6; | |
2859 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
2860 | } | |
0c19f846 WB |
2861 | |
2862 | arg &= ~TUN_F_UFO; | |
5228ddc9 RR |
2863 | } |
2864 | ||
2865 | /* This gives the user a way to test for new features in future by | |
2866 | * trying to set them. */ | |
2867 | if (arg) | |
2868 | return -EINVAL; | |
2869 | ||
88255375 | 2870 | tun->set_features = features; |
09050957 YI |
2871 | tun->dev->wanted_features &= ~TUN_USER_FEATURES; |
2872 | tun->dev->wanted_features |= features; | |
88255375 | 2873 | netdev_update_features(tun->dev); |
5228ddc9 RR |
2874 | |
2875 | return 0; | |
2876 | } | |
2877 | ||
c8d68e6b JW |
2878 | static void tun_detach_filter(struct tun_struct *tun, int n) |
2879 | { | |
2880 | int i; | |
2881 | struct tun_file *tfile; | |
2882 | ||
2883 | for (i = 0; i < n; i++) { | |
b8deabd3 | 2884 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2885 | lock_sock(tfile->socket.sk); |
2886 | sk_detach_filter(tfile->socket.sk); | |
2887 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2888 | } |
2889 | ||
2890 | tun->filter_attached = false; | |
2891 | } | |
2892 | ||
2893 | static int tun_attach_filter(struct tun_struct *tun) | |
2894 | { | |
2895 | int i, ret = 0; | |
2896 | struct tun_file *tfile; | |
2897 | ||
2898 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2899 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2900 | lock_sock(tfile->socket.sk); |
2901 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
2902 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2903 | if (ret) { |
2904 | tun_detach_filter(tun, i); | |
2905 | return ret; | |
2906 | } | |
2907 | } | |
2908 | ||
2909 | tun->filter_attached = true; | |
2910 | return ret; | |
2911 | } | |
2912 | ||
2913 | static void tun_set_sndbuf(struct tun_struct *tun) | |
2914 | { | |
2915 | struct tun_file *tfile; | |
2916 | int i; | |
2917 | ||
2918 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2919 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
2920 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
2921 | } | |
2922 | } | |
2923 | ||
cde8b15f JW |
2924 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
2925 | { | |
2926 | struct tun_file *tfile = file->private_data; | |
2927 | struct tun_struct *tun; | |
cde8b15f JW |
2928 | int ret = 0; |
2929 | ||
2930 | rtnl_lock(); | |
2931 | ||
2932 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 2933 | tun = tfile->detached; |
5dbbaf2d | 2934 | if (!tun) { |
cde8b15f | 2935 | ret = -EINVAL; |
5dbbaf2d PM |
2936 | goto unlock; |
2937 | } | |
2938 | ret = security_tun_dev_attach_queue(tun->security); | |
2939 | if (ret < 0) | |
2940 | goto unlock; | |
af3fb24e | 2941 | ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI, |
77f22f92 | 2942 | tun->flags & IFF_NAPI_FRAGS, true); |
4008e97f | 2943 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 2944 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 2945 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
2946 | ret = -EINVAL; |
2947 | else | |
2948 | __tun_detach(tfile, false); | |
2949 | } else | |
cde8b15f JW |
2950 | ret = -EINVAL; |
2951 | ||
83c1f36f SD |
2952 | if (ret >= 0) |
2953 | netdev_state_change(tun->dev); | |
2954 | ||
5dbbaf2d | 2955 | unlock: |
cde8b15f JW |
2956 | rtnl_unlock(); |
2957 | return ret; | |
2958 | } | |
2959 | ||
8f3f330d | 2960 | static int tun_set_ebpf(struct tun_struct *tun, struct tun_prog __rcu **prog_p, |
cd5681d7 | 2961 | void __user *data) |
96f84061 JW |
2962 | { |
2963 | struct bpf_prog *prog; | |
2964 | int fd; | |
2965 | ||
2966 | if (copy_from_user(&fd, data, sizeof(fd))) | |
2967 | return -EFAULT; | |
2968 | ||
2969 | if (fd == -1) { | |
2970 | prog = NULL; | |
2971 | } else { | |
2972 | prog = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); | |
2973 | if (IS_ERR(prog)) | |
2974 | return PTR_ERR(prog); | |
2975 | } | |
2976 | ||
cd5681d7 | 2977 | return __tun_set_ebpf(tun, prog_p, prog); |
96f84061 JW |
2978 | } |
2979 | ||
cca8ea3b PP |
2980 | /* Return correct value for tun->dev->addr_len based on tun->dev->type. */ |
2981 | static unsigned char tun_get_addr_len(unsigned short type) | |
2982 | { | |
2983 | switch (type) { | |
2984 | case ARPHRD_IP6GRE: | |
2985 | case ARPHRD_TUNNEL6: | |
2986 | return sizeof(struct in6_addr); | |
2987 | case ARPHRD_IPGRE: | |
2988 | case ARPHRD_TUNNEL: | |
2989 | case ARPHRD_SIT: | |
2990 | return 4; | |
2991 | case ARPHRD_ETHER: | |
2992 | return ETH_ALEN; | |
2993 | case ARPHRD_IEEE802154: | |
2994 | case ARPHRD_IEEE802154_MONITOR: | |
2995 | return IEEE802154_EXTENDED_ADDR_LEN; | |
2996 | case ARPHRD_PHONET_PIPE: | |
2997 | case ARPHRD_PPP: | |
2998 | case ARPHRD_NONE: | |
2999 | return 0; | |
3000 | case ARPHRD_6LOWPAN: | |
3001 | return EUI64_ADDR_LEN; | |
3002 | case ARPHRD_FDDI: | |
3003 | return FDDI_K_ALEN; | |
3004 | case ARPHRD_HIPPI: | |
3005 | return HIPPI_ALEN; | |
3006 | case ARPHRD_IEEE802: | |
3007 | return FC_ALEN; | |
3008 | case ARPHRD_ROSE: | |
3009 | return ROSE_ADDR_LEN; | |
3010 | case ARPHRD_NETROM: | |
3011 | return AX25_ADDR_LEN; | |
3012 | case ARPHRD_LOCALTLK: | |
3013 | return LTALK_ALEN; | |
3014 | default: | |
3015 | return 0; | |
3016 | } | |
3017 | } | |
3018 | ||
50857e2a AB |
3019 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
3020 | unsigned long arg, int ifreq_len) | |
1da177e4 | 3021 | { |
36b50bab | 3022 | struct tun_file *tfile = file->private_data; |
f663706a | 3023 | struct net *net = sock_net(&tfile->sk); |
631ab46b | 3024 | struct tun_struct *tun; |
1da177e4 | 3025 | void __user* argp = (void __user*)arg; |
26d31925 | 3026 | unsigned int ifindex, carrier; |
1da177e4 | 3027 | struct ifreq ifr; |
0625c883 EB |
3028 | kuid_t owner; |
3029 | kgid_t group; | |
33dccbb0 | 3030 | int sndbuf; |
d9d52b51 | 3031 | int vnet_hdr_sz; |
1cf8e410 | 3032 | int le; |
f271b2cc | 3033 | int ret; |
83c1f36f | 3034 | bool do_notify = false; |
1da177e4 | 3035 | |
f2780d6d KT |
3036 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || |
3037 | (_IOC_TYPE(cmd) == SOCK_IOC_TYPE && cmd != SIOCGSKNS)) { | |
50857e2a | 3038 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 3039 | return -EFAULT; |
8bbb1813 | 3040 | } else { |
a117dacd | 3041 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 3042 | } |
631ab46b EB |
3043 | if (cmd == TUNGETFEATURES) { |
3044 | /* Currently this just means: "what IFF flags are valid?". | |
3045 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
3046 | * TUNSETIFF. |
3047 | */ | |
3048 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 3049 | (unsigned int __user*)argp); |
f663706a | 3050 | } else if (cmd == TUNSETQUEUE) { |
cde8b15f | 3051 | return tun_set_queue(file, &ifr); |
f663706a KT |
3052 | } else if (cmd == SIOCGSKNS) { |
3053 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
3054 | return -EPERM; | |
3055 | return open_related_ns(&net->ns, get_net_ns); | |
3056 | } | |
631ab46b | 3057 | |
876bfd4d HX |
3058 | rtnl_lock(); |
3059 | ||
9484dc74 | 3060 | tun = tun_get(tfile); |
0f16bc13 GF |
3061 | if (cmd == TUNSETIFF) { |
3062 | ret = -EEXIST; | |
3063 | if (tun) | |
3064 | goto unlock; | |
3065 | ||
1da177e4 LT |
3066 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
3067 | ||
f2780d6d | 3068 | ret = tun_set_iff(net, file, &ifr); |
1da177e4 | 3069 | |
876bfd4d HX |
3070 | if (ret) |
3071 | goto unlock; | |
1da177e4 | 3072 | |
50857e2a | 3073 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
3074 | ret = -EFAULT; |
3075 | goto unlock; | |
1da177e4 | 3076 | } |
fb7589a1 PE |
3077 | if (cmd == TUNSETIFINDEX) { |
3078 | ret = -EPERM; | |
3079 | if (tun) | |
3080 | goto unlock; | |
3081 | ||
3082 | ret = -EFAULT; | |
3083 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
3084 | goto unlock; | |
3085 | ||
3086 | ret = 0; | |
3087 | tfile->ifindex = ifindex; | |
3088 | goto unlock; | |
3089 | } | |
1da177e4 | 3090 | |
876bfd4d | 3091 | ret = -EBADFD; |
1da177e4 | 3092 | if (!tun) |
876bfd4d | 3093 | goto unlock; |
1da177e4 | 3094 | |
3424170f | 3095 | netif_info(tun, drv, tun->dev, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 3096 | |
0c3e0e3b | 3097 | net = dev_net(tun->dev); |
631ab46b | 3098 | ret = 0; |
1da177e4 | 3099 | switch (cmd) { |
e3b99556 | 3100 | case TUNGETIFF: |
12132768 | 3101 | tun_get_iff(tun, &ifr); |
e3b99556 | 3102 | |
3d407a80 PE |
3103 | if (tfile->detached) |
3104 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
3105 | if (!tfile->socket.sk->sk_filter) |
3106 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 3107 | |
50857e2a | 3108 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 3109 | ret = -EFAULT; |
e3b99556 MM |
3110 | break; |
3111 | ||
1da177e4 LT |
3112 | case TUNSETNOCSUM: |
3113 | /* Disable/Enable checksum */ | |
1da177e4 | 3114 | |
88255375 | 3115 | /* [unimplemented] */ |
3424170f MK |
3116 | netif_info(tun, drv, tun->dev, "ignored: set checksum %s\n", |
3117 | arg ? "disabled" : "enabled"); | |
1da177e4 LT |
3118 | break; |
3119 | ||
3120 | case TUNSETPERSIST: | |
54f968d6 JW |
3121 | /* Disable/Enable persist mode. Keep an extra reference to the |
3122 | * module to prevent the module being unprobed. | |
3123 | */ | |
40630b82 MT |
3124 | if (arg && !(tun->flags & IFF_PERSIST)) { |
3125 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 3126 | __module_get(THIS_MODULE); |
83c1f36f | 3127 | do_notify = true; |
dd38bd85 | 3128 | } |
40630b82 MT |
3129 | if (!arg && (tun->flags & IFF_PERSIST)) { |
3130 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 | 3131 | module_put(THIS_MODULE); |
83c1f36f | 3132 | do_notify = true; |
54f968d6 | 3133 | } |
1da177e4 | 3134 | |
3424170f MK |
3135 | netif_info(tun, drv, tun->dev, "persist %s\n", |
3136 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
3137 | break; |
3138 | ||
3139 | case TUNSETOWNER: | |
3140 | /* Set owner of the device */ | |
0625c883 EB |
3141 | owner = make_kuid(current_user_ns(), arg); |
3142 | if (!uid_valid(owner)) { | |
3143 | ret = -EINVAL; | |
3144 | break; | |
3145 | } | |
3146 | tun->owner = owner; | |
83c1f36f | 3147 | do_notify = true; |
3424170f MK |
3148 | netif_info(tun, drv, tun->dev, "owner set to %u\n", |
3149 | from_kuid(&init_user_ns, tun->owner)); | |
1da177e4 LT |
3150 | break; |
3151 | ||
8c644623 GG |
3152 | case TUNSETGROUP: |
3153 | /* Set group of the device */ | |
0625c883 EB |
3154 | group = make_kgid(current_user_ns(), arg); |
3155 | if (!gid_valid(group)) { | |
3156 | ret = -EINVAL; | |
3157 | break; | |
3158 | } | |
3159 | tun->group = group; | |
83c1f36f | 3160 | do_notify = true; |
3424170f MK |
3161 | netif_info(tun, drv, tun->dev, "group set to %u\n", |
3162 | from_kgid(&init_user_ns, tun->group)); | |
8c644623 GG |
3163 | break; |
3164 | ||
ff4cc3ac MK |
3165 | case TUNSETLINK: |
3166 | /* Only allow setting the type when the interface is down */ | |
3167 | if (tun->dev->flags & IFF_UP) { | |
3424170f MK |
3168 | netif_info(tun, drv, tun->dev, |
3169 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 3170 | ret = -EBUSY; |
ff4cc3ac | 3171 | } else { |
8e1e33ff MS |
3172 | ret = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
3173 | tun->dev); | |
3174 | ret = notifier_to_errno(ret); | |
3175 | if (ret) { | |
3176 | netif_info(tun, drv, tun->dev, | |
3177 | "Refused to change device type\n"); | |
3178 | break; | |
3179 | } | |
ff4cc3ac | 3180 | tun->dev->type = (int) arg; |
cca8ea3b | 3181 | tun->dev->addr_len = tun_get_addr_len(tun->dev->type); |
3424170f MK |
3182 | netif_info(tun, drv, tun->dev, "linktype set to %d\n", |
3183 | tun->dev->type); | |
8e1e33ff MS |
3184 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
3185 | tun->dev); | |
ff4cc3ac | 3186 | } |
631ab46b | 3187 | break; |
ff4cc3ac | 3188 | |
1da177e4 | 3189 | case TUNSETDEBUG: |
3424170f | 3190 | tun->msg_enable = (u32)arg; |
1da177e4 | 3191 | break; |
3424170f | 3192 | |
5228ddc9 | 3193 | case TUNSETOFFLOAD: |
88255375 | 3194 | ret = set_offload(tun, arg); |
631ab46b | 3195 | break; |
5228ddc9 | 3196 | |
f271b2cc MK |
3197 | case TUNSETTXFILTER: |
3198 | /* Can be set only for TAPs */ | |
631ab46b | 3199 | ret = -EINVAL; |
40630b82 | 3200 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 3201 | break; |
c0e5a8c2 | 3202 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 3203 | break; |
1da177e4 LT |
3204 | |
3205 | case SIOCGIFHWADDR: | |
b595076a | 3206 | /* Get hw address */ |
3b23a32a | 3207 | dev_get_mac_address(&ifr.ifr_hwaddr, net, tun->dev->name); |
50857e2a | 3208 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
3209 | ret = -EFAULT; |
3210 | break; | |
1da177e4 LT |
3211 | |
3212 | case SIOCSIFHWADDR: | |
f271b2cc | 3213 | /* Set hw address */ |
3b23a32a | 3214 | ret = dev_set_mac_address_user(tun->dev, &ifr.ifr_hwaddr, NULL); |
631ab46b | 3215 | break; |
33dccbb0 HX |
3216 | |
3217 | case TUNGETSNDBUF: | |
54f968d6 | 3218 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
3219 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
3220 | ret = -EFAULT; | |
3221 | break; | |
3222 | ||
3223 | case TUNSETSNDBUF: | |
3224 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
3225 | ret = -EFAULT; | |
3226 | break; | |
3227 | } | |
93161922 CG |
3228 | if (sndbuf <= 0) { |
3229 | ret = -EINVAL; | |
3230 | break; | |
3231 | } | |
33dccbb0 | 3232 | |
c8d68e6b JW |
3233 | tun->sndbuf = sndbuf; |
3234 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
3235 | break; |
3236 | ||
d9d52b51 MT |
3237 | case TUNGETVNETHDRSZ: |
3238 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
3239 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
3240 | ret = -EFAULT; | |
3241 | break; | |
3242 | ||
3243 | case TUNSETVNETHDRSZ: | |
3244 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
3245 | ret = -EFAULT; | |
3246 | break; | |
3247 | } | |
3248 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
3249 | ret = -EINVAL; | |
3250 | break; | |
3251 | } | |
3252 | ||
3253 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
3254 | break; | |
3255 | ||
1cf8e410 MT |
3256 | case TUNGETVNETLE: |
3257 | le = !!(tun->flags & TUN_VNET_LE); | |
3258 | if (put_user(le, (int __user *)argp)) | |
3259 | ret = -EFAULT; | |
3260 | break; | |
3261 | ||
3262 | case TUNSETVNETLE: | |
3263 | if (get_user(le, (int __user *)argp)) { | |
3264 | ret = -EFAULT; | |
3265 | break; | |
3266 | } | |
3267 | if (le) | |
3268 | tun->flags |= TUN_VNET_LE; | |
3269 | else | |
3270 | tun->flags &= ~TUN_VNET_LE; | |
3271 | break; | |
3272 | ||
8b8e658b GK |
3273 | case TUNGETVNETBE: |
3274 | ret = tun_get_vnet_be(tun, argp); | |
3275 | break; | |
3276 | ||
3277 | case TUNSETVNETBE: | |
3278 | ret = tun_set_vnet_be(tun, argp); | |
3279 | break; | |
3280 | ||
99405162 MT |
3281 | case TUNATTACHFILTER: |
3282 | /* Can be set only for TAPs */ | |
3283 | ret = -EINVAL; | |
40630b82 | 3284 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
3285 | break; |
3286 | ret = -EFAULT; | |
54f968d6 | 3287 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
3288 | break; |
3289 | ||
c8d68e6b | 3290 | ret = tun_attach_filter(tun); |
99405162 MT |
3291 | break; |
3292 | ||
3293 | case TUNDETACHFILTER: | |
3294 | /* Can be set only for TAPs */ | |
3295 | ret = -EINVAL; | |
40630b82 | 3296 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 3297 | break; |
c8d68e6b JW |
3298 | ret = 0; |
3299 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
3300 | break; |
3301 | ||
76975e9c PE |
3302 | case TUNGETFILTER: |
3303 | ret = -EINVAL; | |
40630b82 | 3304 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
3305 | break; |
3306 | ret = -EFAULT; | |
3307 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
3308 | break; | |
3309 | ret = 0; | |
3310 | break; | |
3311 | ||
96f84061 | 3312 | case TUNSETSTEERINGEBPF: |
cd5681d7 | 3313 | ret = tun_set_ebpf(tun, &tun->steering_prog, argp); |
96f84061 JW |
3314 | break; |
3315 | ||
aff3d70a JW |
3316 | case TUNSETFILTEREBPF: |
3317 | ret = tun_set_ebpf(tun, &tun->filter_prog, argp); | |
3318 | break; | |
3319 | ||
26d31925 ND |
3320 | case TUNSETCARRIER: |
3321 | ret = -EFAULT; | |
3322 | if (copy_from_user(&carrier, argp, sizeof(carrier))) | |
3323 | goto unlock; | |
3324 | ||
3325 | ret = tun_net_change_carrier(tun->dev, (bool)carrier); | |
3326 | break; | |
3327 | ||
0c3e0e3b KT |
3328 | case TUNGETDEVNETNS: |
3329 | ret = -EPERM; | |
3330 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
3331 | goto unlock; | |
3332 | ret = open_related_ns(&net->ns, get_net_ns); | |
3333 | break; | |
3334 | ||
1da177e4 | 3335 | default: |
631ab46b EB |
3336 | ret = -EINVAL; |
3337 | break; | |
ee289b64 | 3338 | } |
1da177e4 | 3339 | |
83c1f36f SD |
3340 | if (do_notify) |
3341 | netdev_state_change(tun->dev); | |
3342 | ||
876bfd4d HX |
3343 | unlock: |
3344 | rtnl_unlock(); | |
3345 | if (tun) | |
3346 | tun_put(tun); | |
631ab46b | 3347 | return ret; |
1da177e4 LT |
3348 | } |
3349 | ||
50857e2a AB |
3350 | static long tun_chr_ioctl(struct file *file, |
3351 | unsigned int cmd, unsigned long arg) | |
3352 | { | |
3353 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
3354 | } | |
3355 | ||
3356 | #ifdef CONFIG_COMPAT | |
3357 | static long tun_chr_compat_ioctl(struct file *file, | |
3358 | unsigned int cmd, unsigned long arg) | |
3359 | { | |
3360 | switch (cmd) { | |
3361 | case TUNSETIFF: | |
3362 | case TUNGETIFF: | |
3363 | case TUNSETTXFILTER: | |
3364 | case TUNGETSNDBUF: | |
3365 | case TUNSETSNDBUF: | |
3366 | case SIOCGIFHWADDR: | |
3367 | case SIOCSIFHWADDR: | |
3368 | arg = (unsigned long)compat_ptr(arg); | |
3369 | break; | |
3370 | default: | |
3371 | arg = (compat_ulong_t)arg; | |
3372 | break; | |
3373 | } | |
3374 | ||
3375 | /* | |
3376 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
3377 | * the end of that structure. All fields that are used in this | |
3378 | * driver are compatible though, we don't need to convert the | |
3379 | * contents. | |
3380 | */ | |
3381 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
3382 | } | |
3383 | #endif /* CONFIG_COMPAT */ | |
3384 | ||
1da177e4 LT |
3385 | static int tun_chr_fasync(int fd, struct file *file, int on) |
3386 | { | |
54f968d6 | 3387 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
3388 | int ret; |
3389 | ||
54f968d6 | 3390 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 3391 | goto out; |
6aa20a22 | 3392 | |
1da177e4 | 3393 | if (on) { |
01919134 | 3394 | __f_setown(file, task_pid(current), PIDTYPE_TGID, 0); |
54f968d6 | 3395 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 3396 | } else |
54f968d6 | 3397 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
3398 | ret = 0; |
3399 | out: | |
9d319522 | 3400 | return ret; |
1da177e4 LT |
3401 | } |
3402 | ||
3403 | static int tun_chr_open(struct inode *inode, struct file * file) | |
3404 | { | |
140e807d | 3405 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 3406 | struct tun_file *tfile; |
deed49fb | 3407 | |
140e807d | 3408 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 3409 | &tun_proto, 0); |
631ab46b EB |
3410 | if (!tfile) |
3411 | return -ENOMEM; | |
b196d88a JW |
3412 | if (ptr_ring_init(&tfile->tx_ring, 0, GFP_KERNEL)) { |
3413 | sk_free(&tfile->sk); | |
3414 | return -ENOMEM; | |
3415 | } | |
3416 | ||
c7256f57 | 3417 | mutex_init(&tfile->napi_mutex); |
c956674b | 3418 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 3419 | tfile->flags = 0; |
fb7589a1 | 3420 | tfile->ifindex = 0; |
54f968d6 | 3421 | |
333f7909 | 3422 | init_waitqueue_head(&tfile->socket.wq.wait); |
54f968d6 JW |
3423 | |
3424 | tfile->socket.file = file; | |
3425 | tfile->socket.ops = &tun_socket_ops; | |
3426 | ||
3427 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
3428 | |
3429 | tfile->sk.sk_write_space = tun_sock_write_space; | |
3430 | tfile->sk.sk_sndbuf = INT_MAX; | |
3431 | ||
631ab46b | 3432 | file->private_data = tfile; |
4008e97f | 3433 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 3434 | |
19a6afb2 JW |
3435 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
3436 | ||
1da177e4 LT |
3437 | return 0; |
3438 | } | |
3439 | ||
3440 | static int tun_chr_close(struct inode *inode, struct file *file) | |
3441 | { | |
631ab46b | 3442 | struct tun_file *tfile = file->private_data; |
1da177e4 | 3443 | |
c8d68e6b | 3444 | tun_detach(tfile, true); |
1da177e4 LT |
3445 | |
3446 | return 0; | |
3447 | } | |
3448 | ||
93e14b6d | 3449 | #ifdef CONFIG_PROC_FS |
9484dc74 | 3450 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file) |
93e14b6d | 3451 | { |
9484dc74 | 3452 | struct tun_file *tfile = file->private_data; |
93e14b6d MY |
3453 | struct tun_struct *tun; |
3454 | struct ifreq ifr; | |
3455 | ||
3456 | memset(&ifr, 0, sizeof(ifr)); | |
3457 | ||
3458 | rtnl_lock(); | |
9484dc74 | 3459 | tun = tun_get(tfile); |
93e14b6d | 3460 | if (tun) |
12132768 | 3461 | tun_get_iff(tun, &ifr); |
93e14b6d MY |
3462 | rtnl_unlock(); |
3463 | ||
3464 | if (tun) | |
3465 | tun_put(tun); | |
3466 | ||
a3816ab0 | 3467 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
3468 | } |
3469 | #endif | |
3470 | ||
d54b1fdb | 3471 | static const struct file_operations tun_fops = { |
6aa20a22 | 3472 | .owner = THIS_MODULE, |
1da177e4 | 3473 | .llseek = no_llseek, |
9b067034 | 3474 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 3475 | .write_iter = tun_chr_write_iter, |
1da177e4 | 3476 | .poll = tun_chr_poll, |
50857e2a AB |
3477 | .unlocked_ioctl = tun_chr_ioctl, |
3478 | #ifdef CONFIG_COMPAT | |
3479 | .compat_ioctl = tun_chr_compat_ioctl, | |
3480 | #endif | |
1da177e4 LT |
3481 | .open = tun_chr_open, |
3482 | .release = tun_chr_close, | |
93e14b6d MY |
3483 | .fasync = tun_chr_fasync, |
3484 | #ifdef CONFIG_PROC_FS | |
3485 | .show_fdinfo = tun_chr_show_fdinfo, | |
3486 | #endif | |
1da177e4 LT |
3487 | }; |
3488 | ||
3489 | static struct miscdevice tun_miscdev = { | |
3490 | .minor = TUN_MINOR, | |
3491 | .name = "tun", | |
e454cea2 | 3492 | .nodename = "net/tun", |
1da177e4 | 3493 | .fops = &tun_fops, |
1da177e4 LT |
3494 | }; |
3495 | ||
3496 | /* ethtool interface */ | |
3497 | ||
4e24f2dd CW |
3498 | static void tun_default_link_ksettings(struct net_device *dev, |
3499 | struct ethtool_link_ksettings *cmd) | |
29ccc49d PR |
3500 | { |
3501 | ethtool_link_ksettings_zero_link_mode(cmd, supported); | |
3502 | ethtool_link_ksettings_zero_link_mode(cmd, advertising); | |
3503 | cmd->base.speed = SPEED_10; | |
3504 | cmd->base.duplex = DUPLEX_FULL; | |
3505 | cmd->base.port = PORT_TP; | |
3506 | cmd->base.phy_address = 0; | |
3507 | cmd->base.autoneg = AUTONEG_DISABLE; | |
4e24f2dd CW |
3508 | } |
3509 | ||
3510 | static int tun_get_link_ksettings(struct net_device *dev, | |
3511 | struct ethtool_link_ksettings *cmd) | |
3512 | { | |
3513 | struct tun_struct *tun = netdev_priv(dev); | |
3514 | ||
3515 | memcpy(cmd, &tun->link_ksettings, sizeof(*cmd)); | |
3516 | return 0; | |
3517 | } | |
3518 | ||
3519 | static int tun_set_link_ksettings(struct net_device *dev, | |
3520 | const struct ethtool_link_ksettings *cmd) | |
3521 | { | |
3522 | struct tun_struct *tun = netdev_priv(dev); | |
3523 | ||
3524 | memcpy(&tun->link_ksettings, cmd, sizeof(*cmd)); | |
1da177e4 LT |
3525 | return 0; |
3526 | } | |
3527 | ||
3528 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
3529 | { | |
3530 | struct tun_struct *tun = netdev_priv(dev); | |
3531 | ||
33a5ba14 RJ |
3532 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
3533 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
3534 | |
3535 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 3536 | case IFF_TUN: |
33a5ba14 | 3537 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 3538 | break; |
40630b82 | 3539 | case IFF_TAP: |
33a5ba14 | 3540 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
3541 | break; |
3542 | } | |
3543 | } | |
3544 | ||
3545 | static u32 tun_get_msglevel(struct net_device *dev) | |
3546 | { | |
1da177e4 | 3547 | struct tun_struct *tun = netdev_priv(dev); |
3424170f MK |
3548 | |
3549 | return tun->msg_enable; | |
1da177e4 LT |
3550 | } |
3551 | ||
3552 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
3553 | { | |
1da177e4 | 3554 | struct tun_struct *tun = netdev_priv(dev); |
3424170f MK |
3555 | |
3556 | tun->msg_enable = value; | |
1da177e4 LT |
3557 | } |
3558 | ||
5503fcec | 3559 | static int tun_get_coalesce(struct net_device *dev, |
f3ccfda1 YM |
3560 | struct ethtool_coalesce *ec, |
3561 | struct kernel_ethtool_coalesce *kernel_coal, | |
3562 | struct netlink_ext_ack *extack) | |
5503fcec JW |
3563 | { |
3564 | struct tun_struct *tun = netdev_priv(dev); | |
3565 | ||
3566 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
3567 | ||
3568 | return 0; | |
3569 | } | |
3570 | ||
3571 | static int tun_set_coalesce(struct net_device *dev, | |
f3ccfda1 YM |
3572 | struct ethtool_coalesce *ec, |
3573 | struct kernel_ethtool_coalesce *kernel_coal, | |
3574 | struct netlink_ext_ack *extack) | |
5503fcec JW |
3575 | { |
3576 | struct tun_struct *tun = netdev_priv(dev); | |
3577 | ||
3578 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
3579 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
3580 | else | |
3581 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
3582 | ||
3583 | return 0; | |
3584 | } | |
3585 | ||
7282d491 | 3586 | static const struct ethtool_ops tun_ethtool_ops = { |
e5ad00b3 | 3587 | .supported_coalesce_params = ETHTOOL_COALESCE_RX_MAX_FRAMES, |
1da177e4 LT |
3588 | .get_drvinfo = tun_get_drvinfo, |
3589 | .get_msglevel = tun_get_msglevel, | |
3590 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 3591 | .get_link = ethtool_op_get_link, |
eda29772 | 3592 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
3593 | .get_coalesce = tun_get_coalesce, |
3594 | .set_coalesce = tun_set_coalesce, | |
29ccc49d | 3595 | .get_link_ksettings = tun_get_link_ksettings, |
4e24f2dd | 3596 | .set_link_ksettings = tun_set_link_ksettings, |
1da177e4 LT |
3597 | }; |
3598 | ||
1576d986 JW |
3599 | static int tun_queue_resize(struct tun_struct *tun) |
3600 | { | |
3601 | struct net_device *dev = tun->dev; | |
3602 | struct tun_file *tfile; | |
5990a305 | 3603 | struct ptr_ring **rings; |
1576d986 JW |
3604 | int n = tun->numqueues + tun->numdisabled; |
3605 | int ret, i; | |
3606 | ||
5990a305 JW |
3607 | rings = kmalloc_array(n, sizeof(*rings), GFP_KERNEL); |
3608 | if (!rings) | |
1576d986 JW |
3609 | return -ENOMEM; |
3610 | ||
3611 | for (i = 0; i < tun->numqueues; i++) { | |
3612 | tfile = rtnl_dereference(tun->tfiles[i]); | |
5990a305 | 3613 | rings[i] = &tfile->tx_ring; |
1576d986 JW |
3614 | } |
3615 | list_for_each_entry(tfile, &tun->disabled, next) | |
5990a305 | 3616 | rings[i++] = &tfile->tx_ring; |
1576d986 | 3617 | |
5990a305 JW |
3618 | ret = ptr_ring_resize_multiple(rings, n, |
3619 | dev->tx_queue_len, GFP_KERNEL, | |
fc72d1d5 | 3620 | tun_ptr_free); |
1576d986 | 3621 | |
5990a305 | 3622 | kfree(rings); |
1576d986 JW |
3623 | return ret; |
3624 | } | |
3625 | ||
3626 | static int tun_device_event(struct notifier_block *unused, | |
3627 | unsigned long event, void *ptr) | |
3628 | { | |
3629 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
3630 | struct tun_struct *tun = netdev_priv(dev); | |
72b319dc | 3631 | int i; |
1576d986 | 3632 | |
86dfb4ac CG |
3633 | if (dev->rtnl_link_ops != &tun_link_ops) |
3634 | return NOTIFY_DONE; | |
3635 | ||
1576d986 JW |
3636 | switch (event) { |
3637 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
3638 | if (tun_queue_resize(tun)) | |
3639 | return NOTIFY_BAD; | |
3640 | break; | |
72b319dc FL |
3641 | case NETDEV_UP: |
3642 | for (i = 0; i < tun->numqueues; i++) { | |
3643 | struct tun_file *tfile; | |
3644 | ||
3645 | tfile = rtnl_dereference(tun->tfiles[i]); | |
3646 | tfile->socket.sk->sk_write_space(tfile->socket.sk); | |
3647 | } | |
3648 | break; | |
1576d986 JW |
3649 | default: |
3650 | break; | |
3651 | } | |
3652 | ||
3653 | return NOTIFY_DONE; | |
3654 | } | |
3655 | ||
3656 | static struct notifier_block tun_notifier_block __read_mostly = { | |
3657 | .notifier_call = tun_device_event, | |
3658 | }; | |
79d17604 | 3659 | |
1da177e4 LT |
3660 | static int __init tun_init(void) |
3661 | { | |
3662 | int ret = 0; | |
3663 | ||
6b8a66ee | 3664 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
1da177e4 | 3665 | |
f019a7a5 | 3666 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 3667 | if (ret) { |
6b8a66ee | 3668 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 3669 | goto err_linkops; |
79d17604 PE |
3670 | } |
3671 | ||
1da177e4 | 3672 | ret = misc_register(&tun_miscdev); |
79d17604 | 3673 | if (ret) { |
6b8a66ee | 3674 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
3675 | goto err_misc; |
3676 | } | |
1576d986 | 3677 | |
5edfbd3c TZ |
3678 | ret = register_netdevice_notifier(&tun_notifier_block); |
3679 | if (ret) { | |
3680 | pr_err("Can't register netdevice notifier\n"); | |
3681 | goto err_notifier; | |
3682 | } | |
3683 | ||
f019a7a5 | 3684 | return 0; |
5edfbd3c TZ |
3685 | |
3686 | err_notifier: | |
3687 | misc_deregister(&tun_miscdev); | |
79d17604 | 3688 | err_misc: |
f019a7a5 EB |
3689 | rtnl_link_unregister(&tun_link_ops); |
3690 | err_linkops: | |
1da177e4 LT |
3691 | return ret; |
3692 | } | |
3693 | ||
3694 | static void tun_cleanup(void) | |
3695 | { | |
6aa20a22 | 3696 | misc_deregister(&tun_miscdev); |
f019a7a5 | 3697 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 3698 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
3699 | } |
3700 | ||
05c2828c MT |
3701 | /* Get an underlying socket object from tun file. Returns error unless file is |
3702 | * attached to a device. The returned object works like a packet socket, it | |
3703 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
3704 | * holding a reference to the file for as long as the socket is in use. */ | |
3705 | struct socket *tun_get_socket(struct file *file) | |
3706 | { | |
6e914fc7 | 3707 | struct tun_file *tfile; |
05c2828c MT |
3708 | if (file->f_op != &tun_fops) |
3709 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
3710 | tfile = file->private_data; |
3711 | if (!tfile) | |
05c2828c | 3712 | return ERR_PTR(-EBADFD); |
54f968d6 | 3713 | return &tfile->socket; |
05c2828c MT |
3714 | } |
3715 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
3716 | ||
5990a305 | 3717 | struct ptr_ring *tun_get_tx_ring(struct file *file) |
83339c6b JW |
3718 | { |
3719 | struct tun_file *tfile; | |
3720 | ||
3721 | if (file->f_op != &tun_fops) | |
3722 | return ERR_PTR(-EINVAL); | |
3723 | tfile = file->private_data; | |
3724 | if (!tfile) | |
3725 | return ERR_PTR(-EBADFD); | |
5990a305 | 3726 | return &tfile->tx_ring; |
83339c6b | 3727 | } |
5990a305 | 3728 | EXPORT_SYMBOL_GPL(tun_get_tx_ring); |
83339c6b | 3729 | |
1da177e4 LT |
3730 | module_init(tun_init); |
3731 | module_exit(tun_cleanup); | |
3732 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
3733 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
3734 | MODULE_LICENSE("GPL"); | |
3735 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 3736 | MODULE_ALIAS("devname:net/tun"); |