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) { |
c39e342a PP |
271 | netif_tx_napi_add(tun->dev, &tfile->napi, tun_napi_poll, |
272 | NAPI_POLL_WEIGHT); | |
94317099 PP |
273 | napi_enable(&tfile->napi); |
274 | } | |
275 | } | |
276 | ||
06e55add | 277 | static void tun_napi_disable(struct tun_file *tfile) |
94317099 | 278 | { |
aec72f33 | 279 | if (tfile->napi_enabled) |
94317099 PP |
280 | napi_disable(&tfile->napi); |
281 | } | |
282 | ||
06e55add | 283 | static void tun_napi_del(struct tun_file *tfile) |
94317099 | 284 | { |
aec72f33 | 285 | if (tfile->napi_enabled) |
94317099 PP |
286 | netif_napi_del(&tfile->napi); |
287 | } | |
288 | ||
af3fb24e | 289 | static bool tun_napi_frags_enabled(const struct tun_file *tfile) |
90e33d45 | 290 | { |
af3fb24e | 291 | return tfile->napi_frags_enabled; |
90e33d45 PP |
292 | } |
293 | ||
8b8e658b GK |
294 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
295 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
296 | { | |
297 | return tun->flags & TUN_VNET_BE ? false : | |
298 | virtio_legacy_is_little_endian(); | |
299 | } | |
300 | ||
301 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
302 | { | |
303 | int be = !!(tun->flags & TUN_VNET_BE); | |
304 | ||
305 | if (put_user(be, argp)) | |
306 | return -EFAULT; | |
307 | ||
308 | return 0; | |
309 | } | |
310 | ||
311 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
312 | { | |
313 | int be; | |
314 | ||
315 | if (get_user(be, argp)) | |
316 | return -EFAULT; | |
317 | ||
318 | if (be) | |
319 | tun->flags |= TUN_VNET_BE; | |
320 | else | |
321 | tun->flags &= ~TUN_VNET_BE; | |
322 | ||
323 | return 0; | |
324 | } | |
325 | #else | |
326 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
327 | { | |
328 | return virtio_legacy_is_little_endian(); | |
329 | } | |
330 | ||
331 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
332 | { | |
333 | return -EINVAL; | |
334 | } | |
335 | ||
336 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
337 | { | |
338 | return -EINVAL; | |
339 | } | |
340 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
341 | ||
25bd55bb GK |
342 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
343 | { | |
7d824109 | 344 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 345 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
346 | } |
347 | ||
56f0dcc5 MT |
348 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
349 | { | |
25bd55bb | 350 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
351 | } |
352 | ||
353 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
354 | { | |
25bd55bb | 355 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
356 | } |
357 | ||
96442e42 JW |
358 | static inline u32 tun_hashfn(u32 rxhash) |
359 | { | |
f13b5468 | 360 | return rxhash & TUN_MASK_FLOW_ENTRIES; |
96442e42 JW |
361 | } |
362 | ||
363 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
364 | { | |
365 | struct tun_flow_entry *e; | |
96442e42 | 366 | |
b67bfe0d | 367 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
368 | if (e->rxhash == rxhash) |
369 | return e; | |
370 | } | |
371 | return NULL; | |
372 | } | |
373 | ||
374 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
375 | struct hlist_head *head, | |
376 | u32 rxhash, u16 queue_index) | |
377 | { | |
9fdc6bef ED |
378 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
379 | ||
96442e42 | 380 | if (e) { |
3424170f MK |
381 | netif_info(tun, tx_queued, tun->dev, |
382 | "create flow: hash %u index %u\n", | |
383 | rxhash, queue_index); | |
96442e42 JW |
384 | e->updated = jiffies; |
385 | e->rxhash = rxhash; | |
9bc88939 | 386 | e->rps_rxhash = 0; |
96442e42 JW |
387 | e->queue_index = queue_index; |
388 | e->tun = tun; | |
389 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 390 | ++tun->flow_count; |
96442e42 JW |
391 | } |
392 | return e; | |
393 | } | |
394 | ||
96442e42 JW |
395 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
396 | { | |
3424170f MK |
397 | netif_info(tun, tx_queued, tun->dev, "delete flow: hash %u index %u\n", |
398 | e->rxhash, e->queue_index); | |
96442e42 | 399 | hlist_del_rcu(&e->hash_link); |
9fdc6bef | 400 | kfree_rcu(e, rcu); |
b8732fb7 | 401 | --tun->flow_count; |
96442e42 JW |
402 | } |
403 | ||
404 | static void tun_flow_flush(struct tun_struct *tun) | |
405 | { | |
406 | int i; | |
407 | ||
408 | spin_lock_bh(&tun->lock); | |
409 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
410 | struct tun_flow_entry *e; | |
b67bfe0d | 411 | struct hlist_node *n; |
96442e42 | 412 | |
b67bfe0d | 413 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
414 | tun_flow_delete(tun, e); |
415 | } | |
416 | spin_unlock_bh(&tun->lock); | |
417 | } | |
418 | ||
419 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
420 | { | |
421 | int i; | |
422 | ||
423 | spin_lock_bh(&tun->lock); | |
424 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
425 | struct tun_flow_entry *e; | |
b67bfe0d | 426 | struct hlist_node *n; |
96442e42 | 427 | |
b67bfe0d | 428 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
429 | if (e->queue_index == queue_index) |
430 | tun_flow_delete(tun, e); | |
431 | } | |
432 | } | |
433 | spin_unlock_bh(&tun->lock); | |
434 | } | |
435 | ||
e99e88a9 | 436 | static void tun_flow_cleanup(struct timer_list *t) |
96442e42 | 437 | { |
e99e88a9 | 438 | struct tun_struct *tun = from_timer(tun, t, flow_gc_timer); |
96442e42 JW |
439 | unsigned long delay = tun->ageing_time; |
440 | unsigned long next_timer = jiffies + delay; | |
441 | unsigned long count = 0; | |
442 | int i; | |
443 | ||
7dbfb4ef | 444 | spin_lock(&tun->lock); |
96442e42 JW |
445 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { |
446 | struct tun_flow_entry *e; | |
b67bfe0d | 447 | struct hlist_node *n; |
96442e42 | 448 | |
b67bfe0d | 449 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 | 450 | unsigned long this_timer; |
81d98fa4 | 451 | |
96442e42 | 452 | this_timer = e->updated + delay; |
81d98fa4 | 453 | if (time_before_eq(this_timer, jiffies)) { |
96442e42 | 454 | tun_flow_delete(tun, e); |
81d98fa4 ED |
455 | continue; |
456 | } | |
457 | count++; | |
458 | if (time_before(this_timer, next_timer)) | |
96442e42 JW |
459 | next_timer = this_timer; |
460 | } | |
461 | } | |
462 | ||
463 | if (count) | |
464 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
7dbfb4ef | 465 | spin_unlock(&tun->lock); |
96442e42 JW |
466 | } |
467 | ||
49974420 | 468 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 469 | struct tun_file *tfile) |
96442e42 JW |
470 | { |
471 | struct hlist_head *head; | |
472 | struct tun_flow_entry *e; | |
473 | unsigned long delay = tun->ageing_time; | |
9e85722d | 474 | u16 queue_index = tfile->queue_index; |
96442e42 | 475 | |
5c327f67 | 476 | head = &tun->flows[tun_hashfn(rxhash)]; |
96442e42 JW |
477 | |
478 | rcu_read_lock(); | |
479 | ||
96442e42 JW |
480 | e = tun_flow_find(head, rxhash); |
481 | if (likely(e)) { | |
482 | /* TODO: keep queueing to old queue until it's empty? */ | |
4ffdd22e ED |
483 | if (READ_ONCE(e->queue_index) != queue_index) |
484 | WRITE_ONCE(e->queue_index, queue_index); | |
83b1bc12 LR |
485 | if (e->updated != jiffies) |
486 | e->updated = jiffies; | |
9bc88939 | 487 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
488 | } else { |
489 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
490 | if (!tun_flow_find(head, rxhash) && |
491 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
492 | tun_flow_create(tun, head, rxhash, queue_index); |
493 | ||
494 | if (!timer_pending(&tun->flow_gc_timer)) | |
495 | mod_timer(&tun->flow_gc_timer, | |
496 | round_jiffies_up(jiffies + delay)); | |
497 | spin_unlock_bh(&tun->lock); | |
498 | } | |
499 | ||
96442e42 JW |
500 | rcu_read_unlock(); |
501 | } | |
502 | ||
516c512b | 503 | /* Save the hash received in the stack receive path and update the |
9bc88939 TH |
504 | * flow_hash table accordingly. |
505 | */ | |
506 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
507 | { | |
567e4b79 | 508 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 509 | e->rps_rxhash = hash; |
9bc88939 TH |
510 | } |
511 | ||
4b035271 | 512 | /* We try to identify a flow through its rxhash. The reason that |
92d4ea6e | 513 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
514 | * the rxq based on the txq where the last packet of the flow comes. As |
515 | * the userspace application move between processors, we may get a | |
4b035271 | 516 | * different rxq no. here. |
c8d68e6b | 517 | */ |
96f84061 | 518 | static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
c8d68e6b | 519 | { |
96442e42 | 520 | struct tun_flow_entry *e; |
c8d68e6b JW |
521 | u32 txq = 0; |
522 | u32 numqueues = 0; | |
523 | ||
6aa7de05 | 524 | numqueues = READ_ONCE(tun->numqueues); |
c8d68e6b | 525 | |
feec084a | 526 | txq = __skb_get_hash_symmetric(skb); |
4b035271 WL |
527 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
528 | if (e) { | |
529 | tun_flow_save_rps_rxhash(e, txq); | |
530 | txq = e->queue_index; | |
531 | } else { | |
532 | /* use multiply and shift instead of expensive divide */ | |
533 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
534 | } |
535 | ||
c8d68e6b JW |
536 | return txq; |
537 | } | |
538 | ||
96f84061 JW |
539 | static u16 tun_ebpf_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
540 | { | |
cd5681d7 | 541 | struct tun_prog *prog; |
a35d310f | 542 | u32 numqueues; |
96f84061 JW |
543 | u16 ret = 0; |
544 | ||
a35d310f JW |
545 | numqueues = READ_ONCE(tun->numqueues); |
546 | if (!numqueues) | |
547 | return 0; | |
548 | ||
96f84061 JW |
549 | prog = rcu_dereference(tun->steering_prog); |
550 | if (prog) | |
551 | ret = bpf_prog_run_clear_cb(prog->prog, skb); | |
552 | ||
a35d310f | 553 | return ret % numqueues; |
96f84061 JW |
554 | } |
555 | ||
556 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, | |
a350ecce | 557 | struct net_device *sb_dev) |
96f84061 JW |
558 | { |
559 | struct tun_struct *tun = netdev_priv(dev); | |
560 | u16 ret; | |
561 | ||
562 | rcu_read_lock(); | |
563 | if (rcu_dereference(tun->steering_prog)) | |
564 | ret = tun_ebpf_select_queue(tun, skb); | |
565 | else | |
566 | ret = tun_automq_select_queue(tun, skb); | |
567 | rcu_read_unlock(); | |
568 | ||
569 | return ret; | |
570 | } | |
571 | ||
cde8b15f JW |
572 | static inline bool tun_not_capable(struct tun_struct *tun) |
573 | { | |
574 | const struct cred *cred = current_cred(); | |
c260b772 | 575 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
576 | |
577 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
578 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 579 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
580 | } |
581 | ||
c8d68e6b JW |
582 | static void tun_set_real_num_queues(struct tun_struct *tun) |
583 | { | |
584 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
585 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
586 | } | |
587 | ||
4008e97f JW |
588 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
589 | { | |
590 | tfile->detached = tun; | |
591 | list_add_tail(&tfile->next, &tun->disabled); | |
592 | ++tun->numdisabled; | |
593 | } | |
594 | ||
d32649d1 | 595 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
596 | { |
597 | struct tun_struct *tun = tfile->detached; | |
598 | ||
599 | tfile->detached = NULL; | |
600 | list_del_init(&tfile->next); | |
601 | --tun->numdisabled; | |
602 | return tun; | |
603 | } | |
604 | ||
3a403076 | 605 | void tun_ptr_free(void *ptr) |
fc72d1d5 JW |
606 | { |
607 | if (!ptr) | |
608 | return; | |
1ffcbc85 JDB |
609 | if (tun_is_xdp_frame(ptr)) { |
610 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 611 | |
03993094 | 612 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
613 | } else { |
614 | __skb_array_destroy_skb(ptr); | |
615 | } | |
616 | } | |
3a403076 | 617 | EXPORT_SYMBOL_GPL(tun_ptr_free); |
fc72d1d5 | 618 | |
4bfb0513 JW |
619 | static void tun_queue_purge(struct tun_file *tfile) |
620 | { | |
fc72d1d5 | 621 | void *ptr; |
1576d986 | 622 | |
fc72d1d5 JW |
623 | while ((ptr = ptr_ring_consume(&tfile->tx_ring)) != NULL) |
624 | tun_ptr_free(ptr); | |
1576d986 | 625 | |
5503fcec | 626 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
627 | skb_queue_purge(&tfile->sk.sk_error_queue); |
628 | } | |
629 | ||
c8d68e6b JW |
630 | static void __tun_detach(struct tun_file *tfile, bool clean) |
631 | { | |
632 | struct tun_file *ntfile; | |
633 | struct tun_struct *tun; | |
c8d68e6b | 634 | |
b8deabd3 JW |
635 | tun = rtnl_dereference(tfile->tun); |
636 | ||
94317099 | 637 | if (tun && clean) { |
06e55add ED |
638 | tun_napi_disable(tfile); |
639 | tun_napi_del(tfile); | |
94317099 PP |
640 | } |
641 | ||
9e85722d | 642 | if (tun && !tfile->detached) { |
c8d68e6b JW |
643 | u16 index = tfile->queue_index; |
644 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
645 | |
646 | rcu_assign_pointer(tun->tfiles[index], | |
647 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 648 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b | 649 | ntfile->queue_index = index; |
9871a9e4 JW |
650 | rcu_assign_pointer(tun->tfiles[tun->numqueues - 1], |
651 | NULL); | |
c8d68e6b JW |
652 | |
653 | --tun->numqueues; | |
9e85722d | 654 | if (clean) { |
c956674b | 655 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 656 | sock_put(&tfile->sk); |
9e85722d | 657 | } else |
4008e97f | 658 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
659 | |
660 | synchronize_net(); | |
96442e42 | 661 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 662 | /* Drop read queue */ |
4bfb0513 | 663 | tun_queue_purge(tfile); |
c8d68e6b | 664 | tun_set_real_num_queues(tun); |
dd38bd85 | 665 | } else if (tfile->detached && clean) { |
4008e97f | 666 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
667 | sock_put(&tfile->sk); |
668 | } | |
c8d68e6b JW |
669 | |
670 | if (clean) { | |
af668b3c MT |
671 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
672 | netif_carrier_off(tun->dev); | |
673 | ||
40630b82 | 674 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 675 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 676 | unregister_netdevice(tun->dev); |
af668b3c | 677 | } |
b196d88a JW |
678 | if (tun) |
679 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
7063efd3 | 680 | ptr_ring_cleanup(&tfile->tx_ring, tun_ptr_free); |
140e807d | 681 | sock_put(&tfile->sk); |
c8d68e6b JW |
682 | } |
683 | } | |
684 | ||
685 | static void tun_detach(struct tun_file *tfile, bool clean) | |
686 | { | |
83c1f36f SD |
687 | struct tun_struct *tun; |
688 | struct net_device *dev; | |
689 | ||
c8d68e6b | 690 | rtnl_lock(); |
83c1f36f SD |
691 | tun = rtnl_dereference(tfile->tun); |
692 | dev = tun ? tun->dev : NULL; | |
c8d68e6b | 693 | __tun_detach(tfile, clean); |
83c1f36f SD |
694 | if (dev) |
695 | netdev_state_change(dev); | |
c8d68e6b JW |
696 | rtnl_unlock(); |
697 | } | |
698 | ||
699 | static void tun_detach_all(struct net_device *dev) | |
700 | { | |
701 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 702 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
703 | int i, n = tun->numqueues; |
704 | ||
705 | for (i = 0; i < n; i++) { | |
b8deabd3 | 706 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 707 | BUG_ON(!tfile); |
06e55add | 708 | tun_napi_disable(tfile); |
addf8fc4 | 709 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 710 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 711 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
712 | --tun->numqueues; |
713 | } | |
9e85722d | 714 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 715 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 716 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 717 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 718 | } |
c8d68e6b JW |
719 | BUG_ON(tun->numqueues != 0); |
720 | ||
721 | synchronize_net(); | |
722 | for (i = 0; i < n; i++) { | |
b8deabd3 | 723 | tfile = rtnl_dereference(tun->tfiles[i]); |
06e55add | 724 | tun_napi_del(tfile); |
c8d68e6b | 725 | /* Drop read queue */ |
4bfb0513 | 726 | tun_queue_purge(tfile); |
b196d88a | 727 | xdp_rxq_info_unreg(&tfile->xdp_rxq); |
c8d68e6b JW |
728 | sock_put(&tfile->sk); |
729 | } | |
4008e97f JW |
730 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
731 | tun_enable_queue(tfile); | |
4bfb0513 | 732 | tun_queue_purge(tfile); |
b196d88a | 733 | xdp_rxq_info_unreg(&tfile->xdp_rxq); |
4008e97f JW |
734 | sock_put(&tfile->sk); |
735 | } | |
736 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 737 | |
40630b82 | 738 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 739 | module_put(THIS_MODULE); |
c8d68e6b JW |
740 | } |
741 | ||
94317099 | 742 | static int tun_attach(struct tun_struct *tun, struct file *file, |
77f22f92 YY |
743 | bool skip_filter, bool napi, bool napi_frags, |
744 | bool publish_tun) | |
a7385ba2 | 745 | { |
631ab46b | 746 | struct tun_file *tfile = file->private_data; |
1576d986 | 747 | struct net_device *dev = tun->dev; |
38231b7a | 748 | int err; |
a7385ba2 | 749 | |
5dbbaf2d PM |
750 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
751 | if (err < 0) | |
752 | goto out; | |
753 | ||
38231b7a | 754 | err = -EINVAL; |
9e85722d | 755 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
756 | goto out; |
757 | ||
758 | err = -EBUSY; | |
40630b82 | 759 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
760 | goto out; |
761 | ||
762 | err = -E2BIG; | |
4008e97f JW |
763 | if (!tfile->detached && |
764 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
765 | goto out; |
766 | ||
767 | err = 0; | |
54f968d6 | 768 | |
92d4ea6e | 769 | /* Re-attach the filter to persist device */ |
849c9b6f | 770 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
771 | lock_sock(tfile->socket.sk); |
772 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
773 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
774 | if (!err) |
775 | goto out; | |
776 | } | |
1576d986 JW |
777 | |
778 | if (!tfile->detached && | |
b196d88a JW |
779 | ptr_ring_resize(&tfile->tx_ring, dev->tx_queue_len, |
780 | GFP_KERNEL, tun_ptr_free)) { | |
1576d986 JW |
781 | err = -ENOMEM; |
782 | goto out; | |
783 | } | |
784 | ||
c8d68e6b | 785 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 786 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
8bf5c4ee JDB |
787 | |
788 | if (tfile->detached) { | |
789 | /* Re-attach detached tfile, updating XDP queue_index */ | |
790 | WARN_ON(!xdp_rxq_info_is_reg(&tfile->xdp_rxq)); | |
791 | ||
792 | if (tfile->xdp_rxq.queue_index != tfile->queue_index) | |
793 | tfile->xdp_rxq.queue_index = tfile->queue_index; | |
794 | } else { | |
795 | /* Setup XDP RX-queue info, for new tfile getting attached */ | |
796 | err = xdp_rxq_info_reg(&tfile->xdp_rxq, | |
b02e5a0e | 797 | tun->dev, tfile->queue_index, 0); |
8bf5c4ee JDB |
798 | if (err < 0) |
799 | goto out; | |
8d5d8852 JDB |
800 | err = xdp_rxq_info_reg_mem_model(&tfile->xdp_rxq, |
801 | MEM_TYPE_PAGE_SHARED, NULL); | |
802 | if (err < 0) { | |
803 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
804 | goto out; | |
805 | } | |
8bf5c4ee JDB |
806 | err = 0; |
807 | } | |
808 | ||
94317099 | 809 | if (tfile->detached) { |
4008e97f | 810 | tun_enable_queue(tfile); |
94317099 | 811 | } else { |
4008e97f | 812 | sock_hold(&tfile->sk); |
af3fb24e | 813 | tun_napi_init(tun, tfile, napi, napi_frags); |
94317099 | 814 | } |
4008e97f | 815 | |
e4a2a304 JW |
816 | if (rtnl_dereference(tun->xdp_prog)) |
817 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
818 | ||
c8d68e6b JW |
819 | /* device is allowed to go away first, so no need to hold extra |
820 | * refcnt. | |
821 | */ | |
822 | ||
0b7959b6 SF |
823 | /* Publish tfile->tun and tun->tfiles only after we've fully |
824 | * initialized tfile; otherwise we risk using half-initialized | |
825 | * object. | |
826 | */ | |
77f22f92 YY |
827 | if (publish_tun) |
828 | rcu_assign_pointer(tfile->tun, tun); | |
0b7959b6 SF |
829 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
830 | tun->numqueues++; | |
3a03cb84 | 831 | tun_set_real_num_queues(tun); |
c8d68e6b JW |
832 | out: |
833 | return err; | |
631ab46b EB |
834 | } |
835 | ||
9484dc74 | 836 | static struct tun_struct *tun_get(struct tun_file *tfile) |
631ab46b | 837 | { |
6e914fc7 | 838 | struct tun_struct *tun; |
c70f1829 | 839 | |
6e914fc7 JW |
840 | rcu_read_lock(); |
841 | tun = rcu_dereference(tfile->tun); | |
842 | if (tun) | |
843 | dev_hold(tun->dev); | |
844 | rcu_read_unlock(); | |
c70f1829 EB |
845 | |
846 | return tun; | |
631ab46b EB |
847 | } |
848 | ||
631ab46b EB |
849 | static void tun_put(struct tun_struct *tun) |
850 | { | |
6e914fc7 | 851 | dev_put(tun->dev); |
631ab46b EB |
852 | } |
853 | ||
6b8a66ee | 854 | /* TAP filtering */ |
f271b2cc MK |
855 | static void addr_hash_set(u32 *mask, const u8 *addr) |
856 | { | |
857 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
858 | mask[n >> 5] |= (1 << (n & 31)); | |
859 | } | |
860 | ||
861 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
862 | { | |
863 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
864 | return mask[n >> 5] & (1 << (n & 31)); | |
865 | } | |
866 | ||
867 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
868 | { | |
869 | struct { u8 u[ETH_ALEN]; } *addr; | |
870 | struct tun_filter uf; | |
871 | int err, alen, n, nexact; | |
872 | ||
873 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
874 | return -EFAULT; | |
875 | ||
876 | if (!uf.count) { | |
877 | /* Disabled */ | |
878 | filter->count = 0; | |
879 | return 0; | |
880 | } | |
881 | ||
882 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
883 | addr = memdup_user(arg + sizeof(uf), alen); |
884 | if (IS_ERR(addr)) | |
885 | return PTR_ERR(addr); | |
f271b2cc MK |
886 | |
887 | /* The filter is updated without holding any locks. Which is | |
888 | * perfectly safe. We disable it first and in the worst | |
889 | * case we'll accept a few undesired packets. */ | |
890 | filter->count = 0; | |
891 | wmb(); | |
892 | ||
893 | /* Use first set of addresses as an exact filter */ | |
894 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
895 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
896 | ||
897 | nexact = n; | |
898 | ||
cfbf84fc AW |
899 | /* Remaining multicast addresses are hashed, |
900 | * unicast will leave the filter disabled. */ | |
f271b2cc | 901 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
902 | for (; n < uf.count; n++) { |
903 | if (!is_multicast_ether_addr(addr[n].u)) { | |
904 | err = 0; /* no filter */ | |
3b8d2a69 | 905 | goto free_addr; |
cfbf84fc | 906 | } |
f271b2cc | 907 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 908 | } |
f271b2cc MK |
909 | |
910 | /* For ALLMULTI just set the mask to all ones. | |
911 | * This overrides the mask populated above. */ | |
912 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
913 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
914 | ||
915 | /* Now enable the filter */ | |
916 | wmb(); | |
917 | filter->count = nexact; | |
918 | ||
919 | /* Return the number of exact filters */ | |
920 | err = nexact; | |
3b8d2a69 | 921 | free_addr: |
f271b2cc MK |
922 | kfree(addr); |
923 | return err; | |
924 | } | |
925 | ||
926 | /* Returns: 0 - drop, !=0 - accept */ | |
927 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
928 | { | |
929 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
930 | * at this point. */ | |
931 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
932 | int i; | |
933 | ||
934 | /* Exact match */ | |
935 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 936 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
937 | return 1; |
938 | ||
939 | /* Inexact match (multicast only) */ | |
940 | if (is_multicast_ether_addr(eh->h_dest)) | |
941 | return addr_hash_test(filter->mask, eh->h_dest); | |
942 | ||
943 | return 0; | |
944 | } | |
945 | ||
946 | /* | |
947 | * Checks whether the packet is accepted or not. | |
948 | * Returns: 0 - drop, !=0 - accept | |
949 | */ | |
950 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
951 | { | |
952 | if (!filter->count) | |
953 | return 1; | |
954 | ||
955 | return run_filter(filter, skb); | |
956 | } | |
957 | ||
1da177e4 LT |
958 | /* Network device part of the driver */ |
959 | ||
7282d491 | 960 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 961 | |
158b515f GK |
962 | static int tun_net_init(struct net_device *dev) |
963 | { | |
964 | struct tun_struct *tun = netdev_priv(dev); | |
965 | struct ifreq *ifr = tun->ifr; | |
966 | int err; | |
967 | ||
968 | dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); | |
969 | if (!dev->tstats) | |
970 | return -ENOMEM; | |
971 | ||
972 | spin_lock_init(&tun->lock); | |
973 | ||
974 | err = security_tun_dev_alloc_security(&tun->security); | |
975 | if (err < 0) { | |
976 | free_percpu(dev->tstats); | |
977 | return err; | |
978 | } | |
979 | ||
980 | tun_flow_init(tun); | |
981 | ||
982 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | | |
983 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | | |
984 | NETIF_F_HW_VLAN_STAG_TX; | |
985 | dev->features = dev->hw_features | NETIF_F_LLTX; | |
986 | dev->vlan_features = dev->features & | |
987 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
988 | NETIF_F_HW_VLAN_STAG_TX); | |
989 | ||
990 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
991 | (ifr->ifr_flags & TUN_FEATURES); | |
992 | ||
993 | INIT_LIST_HEAD(&tun->disabled); | |
994 | err = tun_attach(tun, tun->file, false, ifr->ifr_flags & IFF_NAPI, | |
995 | ifr->ifr_flags & IFF_NAPI_FRAGS, false); | |
996 | if (err < 0) { | |
997 | tun_flow_uninit(tun); | |
998 | security_tun_dev_free_security(tun->security); | |
999 | free_percpu(dev->tstats); | |
1000 | return err; | |
1001 | } | |
1002 | return 0; | |
1003 | } | |
1004 | ||
c70f1829 EB |
1005 | /* Net device detach from fd. */ |
1006 | static void tun_net_uninit(struct net_device *dev) | |
1007 | { | |
c8d68e6b | 1008 | tun_detach_all(dev); |
c70f1829 EB |
1009 | } |
1010 | ||
1da177e4 LT |
1011 | /* Net device open. */ |
1012 | static int tun_net_open(struct net_device *dev) | |
1013 | { | |
c8d68e6b | 1014 | netif_tx_start_all_queues(dev); |
b20e2d54 | 1015 | |
1da177e4 LT |
1016 | return 0; |
1017 | } | |
1018 | ||
1019 | /* Net device close. */ | |
1020 | static int tun_net_close(struct net_device *dev) | |
1021 | { | |
c8d68e6b | 1022 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
1023 | return 0; |
1024 | } | |
1025 | ||
1026 | /* Net device start xmit */ | |
96f84061 | 1027 | static void tun_automq_xmit(struct tun_struct *tun, struct sk_buff *skb) |
1da177e4 | 1028 | { |
3df97ba8 | 1029 | #ifdef CONFIG_RPS |
dc05360f | 1030 | if (tun->numqueues == 1 && static_branch_unlikely(&rps_needed)) { |
9bc88939 TH |
1031 | /* Select queue was not called for the skbuff, so we extract the |
1032 | * RPS hash and save it into the flow_table here. | |
1033 | */ | |
4b035271 | 1034 | struct tun_flow_entry *e; |
9bc88939 TH |
1035 | __u32 rxhash; |
1036 | ||
feec084a | 1037 | rxhash = __skb_get_hash_symmetric(skb); |
4b035271 WL |
1038 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], rxhash); |
1039 | if (e) | |
1040 | tun_flow_save_rps_rxhash(e, rxhash); | |
9bc88939 | 1041 | } |
3df97ba8 | 1042 | #endif |
96f84061 JW |
1043 | } |
1044 | ||
aff3d70a JW |
1045 | static unsigned int run_ebpf_filter(struct tun_struct *tun, |
1046 | struct sk_buff *skb, | |
1047 | int len) | |
1048 | { | |
1049 | struct tun_prog *prog = rcu_dereference(tun->filter_prog); | |
1050 | ||
1051 | if (prog) | |
1052 | len = bpf_prog_run_clear_cb(prog->prog, skb); | |
1053 | ||
1054 | return len; | |
1055 | } | |
1056 | ||
96f84061 JW |
1057 | /* Net device start xmit */ |
1058 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) | |
1059 | { | |
1060 | struct tun_struct *tun = netdev_priv(dev); | |
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 */ | |
9871a9e4 | 1070 | if (!tfile) |
96f84061 JW |
1071 | goto drop; |
1072 | ||
1073 | if (!rcu_dereference(tun->steering_prog)) | |
1074 | tun_automq_xmit(tun, skb); | |
9bc88939 | 1075 | |
3424170f | 1076 | netif_info(tun, tx_queued, tun->dev, "%s %d\n", __func__, skb->len); |
6e914fc7 | 1077 | |
f271b2cc MK |
1078 | /* Drop if the filter does not like it. |
1079 | * This is a noop if the filter is disabled. | |
1080 | * Filter can be enabled only for the TAP devices. */ | |
1081 | if (!check_filter(&tun->txflt, skb)) | |
1082 | goto drop; | |
1083 | ||
54f968d6 JW |
1084 | if (tfile->socket.sk->sk_filter && |
1085 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
1086 | goto drop; |
1087 | ||
aff3d70a | 1088 | len = run_ebpf_filter(tun, skb, len); |
81c89507 | 1089 | if (len == 0 || pskb_trim(skb, len)) |
aff3d70a JW |
1090 | goto drop; |
1091 | ||
1f8b977a | 1092 | if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) |
7bf66305 JW |
1093 | goto drop; |
1094 | ||
7b996243 | 1095 | skb_tx_timestamp(skb); |
eda29772 | 1096 | |
0110d6f2 | 1097 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
1098 | * for indefinite time. |
1099 | */ | |
0110d6f2 MT |
1100 | skb_orphan(skb); |
1101 | ||
895b5c9f | 1102 | nf_reset_ct(skb); |
f8af75f3 | 1103 | |
5990a305 | 1104 | if (ptr_ring_produce(&tfile->tx_ring, skb)) |
1576d986 | 1105 | goto drop; |
1da177e4 | 1106 | |
a31d27fb ND |
1107 | /* NETIF_F_LLTX requires to do our own update of trans_start */ |
1108 | queue = netdev_get_tx_queue(dev, txq); | |
1109 | queue->trans_start = jiffies; | |
1110 | ||
1da177e4 | 1111 | /* Notify and wake up reader process */ |
54f968d6 JW |
1112 | if (tfile->flags & TUN_FASYNC) |
1113 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 1114 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
1115 | |
1116 | rcu_read_unlock(); | |
6ed10654 | 1117 | return NETDEV_TX_OK; |
1da177e4 LT |
1118 | |
1119 | drop: | |
497a5757 | 1120 | atomic_long_inc(&dev->tx_dropped); |
149d36f7 | 1121 | skb_tx_error(skb); |
1da177e4 | 1122 | kfree_skb(skb); |
6e914fc7 | 1123 | rcu_read_unlock(); |
baeababb | 1124 | return NET_XMIT_DROP; |
1da177e4 LT |
1125 | } |
1126 | ||
f271b2cc | 1127 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 1128 | { |
f271b2cc MK |
1129 | /* |
1130 | * This callback is supposed to deal with mc filter in | |
1131 | * _rx_ path and has nothing to do with the _tx_ path. | |
1132 | * In rx path we always accept everything userspace gives us. | |
1133 | */ | |
1da177e4 LT |
1134 | } |
1135 | ||
c8f44aff MM |
1136 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
1137 | netdev_features_t features) | |
88255375 MM |
1138 | { |
1139 | struct tun_struct *tun = netdev_priv(dev); | |
1140 | ||
1141 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
1142 | } | |
eaea34b2 PA |
1143 | |
1144 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
1145 | { | |
1146 | struct tun_struct *tun = netdev_priv(dev); | |
1147 | ||
1148 | if (new_hr < NET_SKB_PAD) | |
1149 | new_hr = NET_SKB_PAD; | |
1150 | ||
1151 | tun->align = new_hr; | |
1152 | } | |
1153 | ||
bc1f4470 | 1154 | static void |
608b9977 PA |
1155 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
1156 | { | |
608b9977 | 1157 | struct tun_struct *tun = netdev_priv(dev); |
608b9977 | 1158 | |
497a5757 | 1159 | dev_get_tstats64(dev, stats); |
608b9977 | 1160 | |
497a5757 HK |
1161 | stats->rx_frame_errors += |
1162 | (unsigned long)atomic_long_read(&tun->rx_frame_errors); | |
608b9977 PA |
1163 | } |
1164 | ||
761876c8 JW |
1165 | static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog, |
1166 | struct netlink_ext_ack *extack) | |
1167 | { | |
1168 | struct tun_struct *tun = netdev_priv(dev); | |
e4a2a304 | 1169 | struct tun_file *tfile; |
761876c8 | 1170 | struct bpf_prog *old_prog; |
e4a2a304 | 1171 | int i; |
761876c8 JW |
1172 | |
1173 | old_prog = rtnl_dereference(tun->xdp_prog); | |
1174 | rcu_assign_pointer(tun->xdp_prog, prog); | |
1175 | if (old_prog) | |
1176 | bpf_prog_put(old_prog); | |
1177 | ||
e4a2a304 JW |
1178 | for (i = 0; i < tun->numqueues; i++) { |
1179 | tfile = rtnl_dereference(tun->tfiles[i]); | |
1180 | if (prog) | |
1181 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
1182 | else | |
1183 | sock_reset_flag(&tfile->sk, SOCK_XDP); | |
1184 | } | |
1185 | list_for_each_entry(tfile, &tun->disabled, next) { | |
1186 | if (prog) | |
1187 | sock_set_flag(&tfile->sk, SOCK_XDP); | |
1188 | else | |
1189 | sock_reset_flag(&tfile->sk, SOCK_XDP); | |
1190 | } | |
1191 | ||
761876c8 JW |
1192 | return 0; |
1193 | } | |
1194 | ||
f4e63525 | 1195 | static int tun_xdp(struct net_device *dev, struct netdev_bpf *xdp) |
761876c8 JW |
1196 | { |
1197 | switch (xdp->command) { | |
1198 | case XDP_SETUP_PROG: | |
1199 | return tun_xdp_set(dev, xdp->prog, xdp->extack); | |
761876c8 JW |
1200 | default: |
1201 | return -EINVAL; | |
1202 | } | |
1203 | } | |
1204 | ||
26d31925 ND |
1205 | static int tun_net_change_carrier(struct net_device *dev, bool new_carrier) |
1206 | { | |
1207 | if (new_carrier) { | |
1208 | struct tun_struct *tun = netdev_priv(dev); | |
1209 | ||
1210 | if (!tun->numqueues) | |
1211 | return -EPERM; | |
1212 | ||
1213 | netif_carrier_on(dev); | |
1214 | } else { | |
1215 | netif_carrier_off(dev); | |
1216 | } | |
1217 | return 0; | |
1218 | } | |
1219 | ||
758e43b7 | 1220 | static const struct net_device_ops tun_netdev_ops = { |
158b515f | 1221 | .ndo_init = tun_net_init, |
c70f1829 | 1222 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1223 | .ndo_open = tun_net_open, |
1224 | .ndo_stop = tun_net_close, | |
00829823 | 1225 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1226 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1227 | .ndo_select_queue = tun_select_queue, |
eaea34b2 | 1228 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1229 | .ndo_get_stats64 = tun_net_get_stats64, |
26d31925 | 1230 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1231 | }; |
1232 | ||
0c9d917b JDB |
1233 | static void __tun_xdp_flush_tfile(struct tun_file *tfile) |
1234 | { | |
1235 | /* Notify and wake up reader process */ | |
1236 | if (tfile->flags & TUN_FASYNC) | |
1237 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
1238 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); | |
1239 | } | |
1240 | ||
42b33468 JDB |
1241 | static int tun_xdp_xmit(struct net_device *dev, int n, |
1242 | struct xdp_frame **frames, u32 flags) | |
fc72d1d5 JW |
1243 | { |
1244 | struct tun_struct *tun = netdev_priv(dev); | |
fc72d1d5 JW |
1245 | struct tun_file *tfile; |
1246 | u32 numqueues; | |
fdc13979 | 1247 | int nxmit = 0; |
735fc405 | 1248 | int i; |
fc72d1d5 | 1249 | |
0c9d917b | 1250 | if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) |
42b33468 JDB |
1251 | return -EINVAL; |
1252 | ||
fc72d1d5 JW |
1253 | rcu_read_lock(); |
1254 | ||
9871a9e4 | 1255 | resample: |
fc72d1d5 JW |
1256 | numqueues = READ_ONCE(tun->numqueues); |
1257 | if (!numqueues) { | |
735fc405 JDB |
1258 | rcu_read_unlock(); |
1259 | return -ENXIO; /* Caller will free/return all frames */ | |
fc72d1d5 JW |
1260 | } |
1261 | ||
1262 | tfile = rcu_dereference(tun->tfiles[smp_processor_id() % | |
1263 | numqueues]); | |
9871a9e4 JW |
1264 | if (unlikely(!tfile)) |
1265 | goto resample; | |
735fc405 JDB |
1266 | |
1267 | spin_lock(&tfile->tx_ring.producer_lock); | |
1268 | for (i = 0; i < n; i++) { | |
1269 | struct xdp_frame *xdp = frames[i]; | |
1270 | /* Encode the XDP flag into lowest bit for consumer to differ | |
1271 | * XDP buffer from sk_buff. | |
1272 | */ | |
1273 | void *frame = tun_xdp_to_ptr(xdp); | |
1274 | ||
1275 | if (__ptr_ring_produce(&tfile->tx_ring, frame)) { | |
497a5757 | 1276 | atomic_long_inc(&dev->tx_dropped); |
fdc13979 | 1277 | break; |
735fc405 | 1278 | } |
fdc13979 | 1279 | nxmit++; |
fc72d1d5 | 1280 | } |
735fc405 | 1281 | spin_unlock(&tfile->tx_ring.producer_lock); |
fc72d1d5 | 1282 | |
0c9d917b JDB |
1283 | if (flags & XDP_XMIT_FLUSH) |
1284 | __tun_xdp_flush_tfile(tfile); | |
1285 | ||
fc72d1d5 | 1286 | rcu_read_unlock(); |
fdc13979 | 1287 | return nxmit; |
fc72d1d5 JW |
1288 | } |
1289 | ||
44fa2dbd JDB |
1290 | static int tun_xdp_tx(struct net_device *dev, struct xdp_buff *xdp) |
1291 | { | |
1b698fa5 | 1292 | struct xdp_frame *frame = xdp_convert_buff_to_frame(xdp); |
fdc13979 | 1293 | int nxmit; |
44fa2dbd JDB |
1294 | |
1295 | if (unlikely(!frame)) | |
1296 | return -EOVERFLOW; | |
1297 | ||
fdc13979 LB |
1298 | nxmit = tun_xdp_xmit(dev, 1, &frame, XDP_XMIT_FLUSH); |
1299 | if (!nxmit) | |
1300 | xdp_return_frame_rx_napi(frame); | |
1301 | return nxmit; | |
fc72d1d5 JW |
1302 | } |
1303 | ||
758e43b7 | 1304 | static const struct net_device_ops tap_netdev_ops = { |
158b515f | 1305 | .ndo_init = tun_net_init, |
c70f1829 | 1306 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1307 | .ndo_open = tun_net_open, |
1308 | .ndo_stop = tun_net_close, | |
00829823 | 1309 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1310 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1311 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1312 | .ndo_set_mac_address = eth_mac_addr, |
1313 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1314 | .ndo_select_queue = tun_select_queue, |
5e52796a | 1315 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1316 | .ndo_set_rx_headroom = tun_set_headroom, |
497a5757 | 1317 | .ndo_get_stats64 = dev_get_tstats64, |
f4e63525 | 1318 | .ndo_bpf = tun_xdp, |
fc72d1d5 | 1319 | .ndo_xdp_xmit = tun_xdp_xmit, |
26d31925 | 1320 | .ndo_change_carrier = tun_net_change_carrier, |
758e43b7 SH |
1321 | }; |
1322 | ||
944a1376 | 1323 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1324 | { |
1325 | int i; | |
1326 | ||
96442e42 JW |
1327 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1328 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1329 | ||
1330 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
e99e88a9 KC |
1331 | timer_setup(&tun->flow_gc_timer, tun_flow_cleanup, 0); |
1332 | mod_timer(&tun->flow_gc_timer, | |
1333 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1334 | } |
1335 | ||
1336 | static void tun_flow_uninit(struct tun_struct *tun) | |
1337 | { | |
1338 | del_timer_sync(&tun->flow_gc_timer); | |
1339 | tun_flow_flush(tun); | |
96442e42 JW |
1340 | } |
1341 | ||
91572088 JW |
1342 | #define MIN_MTU 68 |
1343 | #define MAX_MTU 65535 | |
1344 | ||
1da177e4 | 1345 | /* Initialize net device. */ |
158b515f | 1346 | static void tun_net_initialize(struct net_device *dev) |
1da177e4 LT |
1347 | { |
1348 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1349 | |
1da177e4 | 1350 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1351 | case IFF_TUN: |
758e43b7 | 1352 | dev->netdev_ops = &tun_netdev_ops; |
b9815eb1 | 1353 | dev->header_ops = &ip_tunnel_header_ops; |
758e43b7 | 1354 | |
1da177e4 LT |
1355 | /* Point-to-Point TUN Device */ |
1356 | dev->hard_header_len = 0; | |
1357 | dev->addr_len = 0; | |
1358 | dev->mtu = 1500; | |
1359 | ||
1360 | /* Zero header length */ | |
6aa20a22 | 1361 | dev->type = ARPHRD_NONE; |
1da177e4 | 1362 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1363 | break; |
1364 | ||
40630b82 | 1365 | case IFF_TAP: |
7a0a9608 | 1366 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1367 | /* Ethernet TAP Device */ |
1da177e4 | 1368 | ether_setup(dev); |
550fd08c | 1369 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1370 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1371 | |
f2cedb63 | 1372 | eth_hw_addr_random(dev); |
36226a8d | 1373 | |
1da177e4 LT |
1374 | break; |
1375 | } | |
91572088 JW |
1376 | |
1377 | dev->min_mtu = MIN_MTU; | |
1378 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1379 | } |
1380 | ||
2f3ab622 JW |
1381 | static bool tun_sock_writeable(struct tun_struct *tun, struct tun_file *tfile) |
1382 | { | |
1383 | struct sock *sk = tfile->socket.sk; | |
1384 | ||
1385 | return (tun->dev->flags & IFF_UP) && sock_writeable(sk); | |
1386 | } | |
1387 | ||
1da177e4 LT |
1388 | /* Character device part */ |
1389 | ||
1390 | /* Poll */ | |
afc9a42b | 1391 | static __poll_t tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1392 | { |
b2430de3 | 1393 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1394 | struct tun_struct *tun = tun_get(tfile); |
3c8a9c63 | 1395 | struct sock *sk; |
afc9a42b | 1396 | __poll_t mask = 0; |
1da177e4 LT |
1397 | |
1398 | if (!tun) | |
a9a08845 | 1399 | return EPOLLERR; |
1da177e4 | 1400 | |
54f968d6 | 1401 | sk = tfile->socket.sk; |
3c8a9c63 | 1402 | |
9e641bdc | 1403 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1404 | |
5990a305 | 1405 | if (!ptr_ring_empty(&tfile->tx_ring)) |
a9a08845 | 1406 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 1407 | |
2f3ab622 JW |
1408 | /* Make sure SOCKWQ_ASYNC_NOSPACE is set if not writable to |
1409 | * guarantee EPOLLOUT to be raised by either here or | |
1410 | * tun_sock_write_space(). Then process could get notification | |
1411 | * after it writes to a down device and meets -EIO. | |
1412 | */ | |
1413 | if (tun_sock_writeable(tun, tfile) || | |
1414 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
1415 | tun_sock_writeable(tun, tfile))) | |
a9a08845 | 1416 | mask |= EPOLLOUT | EPOLLWRNORM; |
33dccbb0 | 1417 | |
c70f1829 | 1418 | if (tun->dev->reg_state != NETREG_REGISTERED) |
a9a08845 | 1419 | mask = EPOLLERR; |
c70f1829 | 1420 | |
631ab46b | 1421 | tun_put(tun); |
1da177e4 LT |
1422 | return mask; |
1423 | } | |
1424 | ||
90e33d45 PP |
1425 | static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile, |
1426 | size_t len, | |
1427 | const struct iov_iter *it) | |
1428 | { | |
1429 | struct sk_buff *skb; | |
1430 | size_t linear; | |
1431 | int err; | |
1432 | int i; | |
1433 | ||
1434 | if (it->nr_segs > MAX_SKB_FRAGS + 1) | |
950271d7 | 1435 | return ERR_PTR(-EMSGSIZE); |
90e33d45 PP |
1436 | |
1437 | local_bh_disable(); | |
1438 | skb = napi_get_frags(&tfile->napi); | |
1439 | local_bh_enable(); | |
1440 | if (!skb) | |
1441 | return ERR_PTR(-ENOMEM); | |
1442 | ||
1443 | linear = iov_iter_single_seg_count(it); | |
1444 | err = __skb_grow(skb, linear); | |
1445 | if (err) | |
1446 | goto free; | |
1447 | ||
1448 | skb->len = len; | |
1449 | skb->data_len = len - linear; | |
1450 | skb->truesize += skb->data_len; | |
1451 | ||
1452 | for (i = 1; i < it->nr_segs; i++) { | |
1453 | size_t fragsz = it->iov[i].iov_len; | |
aa6daaca ED |
1454 | struct page *page; |
1455 | void *frag; | |
90e33d45 PP |
1456 | |
1457 | if (fragsz == 0 || fragsz > PAGE_SIZE) { | |
1458 | err = -EINVAL; | |
1459 | goto free; | |
1460 | } | |
aa6daaca ED |
1461 | frag = netdev_alloc_frag(fragsz); |
1462 | if (!frag) { | |
90e33d45 PP |
1463 | err = -ENOMEM; |
1464 | goto free; | |
1465 | } | |
aa6daaca ED |
1466 | page = virt_to_head_page(frag); |
1467 | skb_fill_page_desc(skb, i - 1, page, | |
1468 | frag - page_address(page), fragsz); | |
90e33d45 PP |
1469 | } |
1470 | ||
1471 | return skb; | |
1472 | free: | |
1473 | /* frees skb and all frags allocated with napi_alloc_frag() */ | |
1474 | napi_free_frags(&tfile->napi); | |
1475 | return ERR_PTR(err); | |
1476 | } | |
1477 | ||
f42157cb RR |
1478 | /* prepad is the amount to reserve at front. len is length after that. |
1479 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1480 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1481 | size_t prepad, size_t len, |
1482 | size_t linear, int noblock) | |
f42157cb | 1483 | { |
54f968d6 | 1484 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1485 | struct sk_buff *skb; |
33dccbb0 | 1486 | int err; |
f42157cb RR |
1487 | |
1488 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1489 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1490 | linear = len; |
f42157cb | 1491 | |
33dccbb0 | 1492 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1493 | &err, 0); |
f42157cb | 1494 | if (!skb) |
33dccbb0 | 1495 | return ERR_PTR(err); |
f42157cb RR |
1496 | |
1497 | skb_reserve(skb, prepad); | |
1498 | skb_put(skb, linear); | |
33dccbb0 HX |
1499 | skb->data_len = len - linear; |
1500 | skb->len += len - linear; | |
f42157cb RR |
1501 | |
1502 | return skb; | |
1503 | } | |
1504 | ||
5503fcec JW |
1505 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1506 | struct sk_buff *skb, int more) | |
1507 | { | |
1508 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1509 | struct sk_buff_head process_queue; | |
1510 | u32 rx_batched = tun->rx_batched; | |
1511 | bool rcv = false; | |
1512 | ||
1513 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1514 | local_bh_disable(); | |
8ebebcba | 1515 | skb_record_rx_queue(skb, tfile->queue_index); |
5503fcec JW |
1516 | netif_receive_skb(skb); |
1517 | local_bh_enable(); | |
1518 | return; | |
1519 | } | |
1520 | ||
1521 | spin_lock(&queue->lock); | |
1522 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1523 | __skb_queue_head_init(&process_queue); | |
1524 | skb_queue_splice_tail_init(queue, &process_queue); | |
1525 | rcv = true; | |
1526 | } else { | |
1527 | __skb_queue_tail(queue, skb); | |
1528 | } | |
1529 | spin_unlock(&queue->lock); | |
1530 | ||
1531 | if (rcv) { | |
1532 | struct sk_buff *nskb; | |
1533 | ||
1534 | local_bh_disable(); | |
8ebebcba MC |
1535 | while ((nskb = __skb_dequeue(&process_queue))) { |
1536 | skb_record_rx_queue(nskb, tfile->queue_index); | |
5503fcec | 1537 | netif_receive_skb(nskb); |
8ebebcba MC |
1538 | } |
1539 | skb_record_rx_queue(skb, tfile->queue_index); | |
5503fcec JW |
1540 | netif_receive_skb(skb); |
1541 | local_bh_enable(); | |
1542 | } | |
1543 | } | |
1544 | ||
66ccbc9c JW |
1545 | static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile, |
1546 | int len, int noblock, bool zerocopy) | |
1547 | { | |
1548 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) | |
1549 | return false; | |
1550 | ||
1551 | if (tfile->socket.sk->sk_sndbuf != INT_MAX) | |
1552 | return false; | |
1553 | ||
1554 | if (!noblock) | |
1555 | return false; | |
1556 | ||
1557 | if (zerocopy) | |
1558 | return false; | |
1559 | ||
1560 | if (SKB_DATA_ALIGN(len + TUN_RX_PAD) + | |
1561 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE) | |
1562 | return false; | |
1563 | ||
1564 | return true; | |
1565 | } | |
1566 | ||
4b663366 AB |
1567 | static struct sk_buff *__tun_build_skb(struct tun_file *tfile, |
1568 | struct page_frag *alloc_frag, char *buf, | |
8ae1aff0 | 1569 | int buflen, int len, int pad) |
ac1f1f6c JW |
1570 | { |
1571 | struct sk_buff *skb = build_skb(buf, buflen); | |
1572 | ||
1573 | if (!skb) | |
1574 | return ERR_PTR(-ENOMEM); | |
1575 | ||
8ae1aff0 | 1576 | skb_reserve(skb, pad); |
ac1f1f6c | 1577 | skb_put(skb, len); |
4b663366 | 1578 | skb_set_owner_w(skb, tfile->socket.sk); |
ac1f1f6c JW |
1579 | |
1580 | get_page(alloc_frag->page); | |
1581 | alloc_frag->offset += buflen; | |
1582 | ||
1583 | return skb; | |
1584 | } | |
1585 | ||
8ae1aff0 JW |
1586 | static int tun_xdp_act(struct tun_struct *tun, struct bpf_prog *xdp_prog, |
1587 | struct xdp_buff *xdp, u32 act) | |
1588 | { | |
1589 | int err; | |
1590 | ||
1591 | switch (act) { | |
1592 | case XDP_REDIRECT: | |
1593 | err = xdp_do_redirect(tun->dev, xdp, xdp_prog); | |
8ae1aff0 JW |
1594 | if (err) |
1595 | return err; | |
1596 | break; | |
1597 | case XDP_TX: | |
1598 | err = tun_xdp_tx(tun->dev, xdp); | |
1599 | if (err < 0) | |
1600 | return err; | |
1601 | break; | |
1602 | case XDP_PASS: | |
1603 | break; | |
1604 | default: | |
c8064e5b | 1605 | bpf_warn_invalid_xdp_action(tun->dev, xdp_prog, act); |
df561f66 | 1606 | fallthrough; |
8ae1aff0 JW |
1607 | case XDP_ABORTED: |
1608 | trace_xdp_exception(tun->dev, xdp_prog, act); | |
df561f66 | 1609 | fallthrough; |
8ae1aff0 | 1610 | case XDP_DROP: |
497a5757 | 1611 | atomic_long_inc(&tun->dev->rx_dropped); |
8ae1aff0 JW |
1612 | break; |
1613 | } | |
1614 | ||
1615 | return act; | |
1616 | } | |
1617 | ||
761876c8 JW |
1618 | static struct sk_buff *tun_build_skb(struct tun_struct *tun, |
1619 | struct tun_file *tfile, | |
66ccbc9c | 1620 | struct iov_iter *from, |
761876c8 | 1621 | struct virtio_net_hdr *hdr, |
1cfe6e93 | 1622 | int len, int *skb_xdp) |
66ccbc9c | 1623 | { |
0bbd7dad | 1624 | struct page_frag *alloc_frag = ¤t->task_frag; |
761876c8 | 1625 | struct bpf_prog *xdp_prog; |
7df13219 | 1626 | int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); |
66ccbc9c JW |
1627 | char *buf; |
1628 | size_t copied; | |
8ae1aff0 JW |
1629 | int pad = TUN_RX_PAD; |
1630 | int err = 0; | |
7df13219 JW |
1631 | |
1632 | rcu_read_lock(); | |
1633 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1634 | if (xdp_prog) | |
4f23aff8 | 1635 | pad += XDP_PACKET_HEADROOM; |
7df13219 JW |
1636 | buflen += SKB_DATA_ALIGN(len + pad); |
1637 | rcu_read_unlock(); | |
66ccbc9c | 1638 | |
63b9ab65 | 1639 | alloc_frag->offset = ALIGN((u64)alloc_frag->offset, SMP_CACHE_BYTES); |
66ccbc9c JW |
1640 | if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL))) |
1641 | return ERR_PTR(-ENOMEM); | |
1642 | ||
1643 | buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; | |
1644 | copied = copy_page_from_iter(alloc_frag->page, | |
7df13219 | 1645 | alloc_frag->offset + pad, |
66ccbc9c JW |
1646 | len, from); |
1647 | if (copied != len) | |
1648 | return ERR_PTR(-EFAULT); | |
1649 | ||
7df13219 JW |
1650 | /* There's a small window that XDP may be set after the check |
1651 | * of xdp_prog above, this should be rare and for simplicity | |
1652 | * we do XDP on skb in case the headroom is not enough. | |
1653 | */ | |
ac1f1f6c | 1654 | if (hdr->gso_type || !xdp_prog) { |
1cfe6e93 | 1655 | *skb_xdp = 1; |
4b663366 AB |
1656 | return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, |
1657 | pad); | |
ac1f1f6c JW |
1658 | } |
1659 | ||
1660 | *skb_xdp = 0; | |
761876c8 | 1661 | |
6547e387 | 1662 | local_bh_disable(); |
761876c8 JW |
1663 | rcu_read_lock(); |
1664 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
8ae1aff0 | 1665 | if (xdp_prog) { |
761876c8 | 1666 | struct xdp_buff xdp; |
761876c8 JW |
1667 | u32 act; |
1668 | ||
43b5169d | 1669 | xdp_init_buff(&xdp, buflen, &tfile->xdp_rxq); |
be9df4af | 1670 | xdp_prepare_buff(&xdp, buf, pad, len, false); |
761876c8 | 1671 | |
8ae1aff0 JW |
1672 | act = bpf_prog_run_xdp(xdp_prog, &xdp); |
1673 | if (act == XDP_REDIRECT || act == XDP_TX) { | |
59655a5b JW |
1674 | get_page(alloc_frag->page); |
1675 | alloc_frag->offset += buflen; | |
761876c8 | 1676 | } |
8ae1aff0 | 1677 | err = tun_xdp_act(tun, xdp_prog, &xdp, act); |
bee34890 WD |
1678 | if (err < 0) { |
1679 | if (act == XDP_REDIRECT || act == XDP_TX) | |
1680 | put_page(alloc_frag->page); | |
1681 | goto out; | |
1682 | } | |
1683 | ||
1a097910 | 1684 | if (err == XDP_REDIRECT) |
1d233886 | 1685 | xdp_do_flush(); |
8ae1aff0 JW |
1686 | if (err != XDP_PASS) |
1687 | goto out; | |
1688 | ||
1689 | pad = xdp.data - xdp.data_hard_start; | |
1690 | len = xdp.data_end - xdp.data; | |
761876c8 | 1691 | } |
291aeb2b JW |
1692 | rcu_read_unlock(); |
1693 | local_bh_enable(); | |
761876c8 | 1694 | |
4b663366 | 1695 | return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, pad); |
761876c8 | 1696 | |
f7053b6c | 1697 | out: |
761876c8 | 1698 | rcu_read_unlock(); |
6547e387 | 1699 | local_bh_enable(); |
761876c8 | 1700 | return NULL; |
66ccbc9c JW |
1701 | } |
1702 | ||
1da177e4 | 1703 | /* Get packet from user space buffer */ |
54f968d6 | 1704 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1705 | void *msg_control, struct iov_iter *from, |
5503fcec | 1706 | int noblock, bool more) |
1da177e4 | 1707 | { |
09640e63 | 1708 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1709 | struct sk_buff *skb; |
f5ff53b4 | 1710 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1711 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1712 | struct virtio_net_hdr gso = { 0 }; |
96f8d9ec | 1713 | int good_linear; |
0690899b MT |
1714 | int copylen; |
1715 | bool zerocopy = false; | |
1716 | int err; | |
96f84061 | 1717 | u32 rxhash = 0; |
1cfe6e93 | 1718 | int skb_xdp = 1; |
af3fb24e | 1719 | bool frags = tun_napi_frags_enabled(tfile); |
1da177e4 | 1720 | |
40630b82 | 1721 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1722 | if (len < sizeof(pi)) |
1da177e4 | 1723 | return -EINVAL; |
15718ea0 | 1724 | len -= sizeof(pi); |
1da177e4 | 1725 | |
cbbd26b8 | 1726 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1727 | return -EFAULT; |
1728 | } | |
1729 | ||
40630b82 | 1730 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1731 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1732 | ||
1733 | if (len < vnet_hdr_sz) | |
f43798c2 | 1734 | return -EINVAL; |
e1edab87 | 1735 | len -= vnet_hdr_sz; |
f43798c2 | 1736 | |
cbbd26b8 | 1737 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1738 | return -EFAULT; |
1739 | ||
4909122f | 1740 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1741 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1742 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1743 | |
56f0dcc5 | 1744 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1745 | return -EINVAL; |
e1edab87 | 1746 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1747 | } |
1748 | ||
40630b82 | 1749 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1750 | align += NET_IP_ALIGN; |
0eca93bc | 1751 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1752 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1753 | return -EINVAL; |
1754 | } | |
6aa20a22 | 1755 | |
96f8d9ec JW |
1756 | good_linear = SKB_MAX_HEAD(align); |
1757 | ||
88529176 | 1758 | if (msg_control) { |
f5ff53b4 AV |
1759 | struct iov_iter i = *from; |
1760 | ||
88529176 JW |
1761 | /* There are 256 bytes to be copied in skb, so there is |
1762 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1763 | * The rest of the buffer is mapped from userspace. |
1764 | */ | |
56f0dcc5 | 1765 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1766 | if (copylen > good_linear) |
1767 | copylen = good_linear; | |
3dd5c330 | 1768 | linear = copylen; |
f5ff53b4 AV |
1769 | iov_iter_advance(&i, copylen); |
1770 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1771 | zerocopy = true; |
1772 | } | |
1773 | ||
90e33d45 | 1774 | if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) { |
1cfe6e93 JW |
1775 | /* For the packet that is not easy to be processed |
1776 | * (e.g gso or jumbo packet), we will do it at after | |
1777 | * skb was created with generic XDP routine. | |
1778 | */ | |
1779 | skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp); | |
66ccbc9c | 1780 | if (IS_ERR(skb)) { |
497a5757 | 1781 | atomic_long_inc(&tun->dev->rx_dropped); |
66ccbc9c JW |
1782 | return PTR_ERR(skb); |
1783 | } | |
761876c8 JW |
1784 | if (!skb) |
1785 | return total_len; | |
66ccbc9c JW |
1786 | } else { |
1787 | if (!zerocopy) { | |
1788 | copylen = len; | |
1789 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) | |
1790 | linear = good_linear; | |
1791 | else | |
1792 | linear = tun16_to_cpu(tun, gso.hdr_len); | |
1793 | } | |
1da177e4 | 1794 | |
90e33d45 PP |
1795 | if (frags) { |
1796 | mutex_lock(&tfile->napi_mutex); | |
1797 | skb = tun_napi_alloc_frags(tfile, copylen, from); | |
1798 | /* tun_napi_alloc_frags() enforces a layout for the skb. | |
1799 | * If zerocopy is enabled, then this layout will be | |
1800 | * overwritten by zerocopy_sg_from_iter(). | |
1801 | */ | |
1802 | zerocopy = false; | |
1803 | } else { | |
1804 | skb = tun_alloc_skb(tfile, align, copylen, linear, | |
1805 | noblock); | |
1806 | } | |
1807 | ||
66ccbc9c JW |
1808 | if (IS_ERR(skb)) { |
1809 | if (PTR_ERR(skb) != -EAGAIN) | |
497a5757 | 1810 | atomic_long_inc(&tun->dev->rx_dropped); |
90e33d45 PP |
1811 | if (frags) |
1812 | mutex_unlock(&tfile->napi_mutex); | |
66ccbc9c JW |
1813 | return PTR_ERR(skb); |
1814 | } | |
0690899b | 1815 | |
66ccbc9c JW |
1816 | if (zerocopy) |
1817 | err = zerocopy_sg_from_iter(skb, from); | |
1818 | else | |
1819 | err = skb_copy_datagram_from_iter(skb, 0, from, len); | |
1820 | ||
1821 | if (err) { | |
4477138f ED |
1822 | err = -EFAULT; |
1823 | drop: | |
497a5757 | 1824 | atomic_long_inc(&tun->dev->rx_dropped); |
66ccbc9c | 1825 | kfree_skb(skb); |
90e33d45 PP |
1826 | if (frags) { |
1827 | tfile->napi.skb = NULL; | |
1828 | mutex_unlock(&tfile->napi_mutex); | |
1829 | } | |
1830 | ||
4477138f | 1831 | return err; |
66ccbc9c | 1832 | } |
8f22757e | 1833 | } |
1da177e4 | 1834 | |
3e9e40e7 | 1835 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
497a5757 | 1836 | atomic_long_inc(&tun->rx_frame_errors); |
df10db98 | 1837 | kfree_skb(skb); |
90e33d45 PP |
1838 | if (frags) { |
1839 | tfile->napi.skb = NULL; | |
1840 | mutex_unlock(&tfile->napi_mutex); | |
1841 | } | |
1842 | ||
df10db98 PA |
1843 | return -EINVAL; |
1844 | } | |
1845 | ||
1da177e4 | 1846 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1847 | case IFF_TUN: |
1848 | if (tun->flags & IFF_NO_PI) { | |
2580c4c1 AP |
1849 | u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0; |
1850 | ||
1851 | switch (ip_version) { | |
1852 | case 4: | |
f09f7ee2 AWC |
1853 | pi.proto = htons(ETH_P_IP); |
1854 | break; | |
2580c4c1 | 1855 | case 6: |
f09f7ee2 AWC |
1856 | pi.proto = htons(ETH_P_IPV6); |
1857 | break; | |
1858 | default: | |
497a5757 | 1859 | atomic_long_inc(&tun->dev->rx_dropped); |
f09f7ee2 AWC |
1860 | kfree_skb(skb); |
1861 | return -EINVAL; | |
1862 | } | |
1863 | } | |
1864 | ||
459a98ed | 1865 | skb_reset_mac_header(skb); |
1da177e4 | 1866 | skb->protocol = pi.proto; |
4c13eb66 | 1867 | skb->dev = tun->dev; |
1da177e4 | 1868 | break; |
40630b82 | 1869 | case IFF_TAP: |
96aa1b22 WB |
1870 | if (frags && !pskb_may_pull(skb, ETH_HLEN)) { |
1871 | err = -ENOMEM; | |
1872 | goto drop; | |
1873 | } | |
1874 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1da177e4 | 1875 | break; |
6403eab1 | 1876 | } |
1da177e4 | 1877 | |
0690899b MT |
1878 | /* copy skb_ubuf_info for callback when skb has no error */ |
1879 | if (zerocopy) { | |
9ee5e5ad | 1880 | skb_zcopy_init(skb, msg_control); |
af1cc7a2 JW |
1881 | } else if (msg_control) { |
1882 | struct ubuf_info *uarg = msg_control; | |
36177832 | 1883 | uarg->callback(NULL, uarg, false); |
0690899b MT |
1884 | } |
1885 | ||
72f65107 | 1886 | skb_reset_network_header(skb); |
d2aa125d | 1887 | skb_probe_transport_header(skb); |
3fe260e0 | 1888 | skb_record_rx_queue(skb, tfile->queue_index); |
38502af7 | 1889 | |
1cfe6e93 | 1890 | if (skb_xdp) { |
761876c8 JW |
1891 | struct bpf_prog *xdp_prog; |
1892 | int ret; | |
1893 | ||
6547e387 | 1894 | local_bh_disable(); |
761876c8 JW |
1895 | rcu_read_lock(); |
1896 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1897 | if (xdp_prog) { | |
1898 | ret = do_xdp_generic(xdp_prog, skb); | |
1899 | if (ret != XDP_PASS) { | |
1900 | rcu_read_unlock(); | |
6547e387 | 1901 | local_bh_enable(); |
1efba987 ED |
1902 | if (frags) { |
1903 | tfile->napi.skb = NULL; | |
1904 | mutex_unlock(&tfile->napi_mutex); | |
1905 | } | |
761876c8 JW |
1906 | return total_len; |
1907 | } | |
1908 | } | |
1909 | rcu_read_unlock(); | |
6547e387 | 1910 | local_bh_enable(); |
761876c8 JW |
1911 | } |
1912 | ||
cf1a1e07 PA |
1913 | /* Compute the costly rx hash only if needed for flow updates. |
1914 | * We may get a very small possibility of OOO during switching, not | |
1915 | * worth to optimize. | |
1916 | */ | |
1917 | if (!rcu_access_pointer(tun->steering_prog) && tun->numqueues > 1 && | |
1918 | !tfile->detached) | |
96f84061 | 1919 | rxhash = __skb_get_hash_symmetric(skb); |
94317099 | 1920 | |
4477138f ED |
1921 | rcu_read_lock(); |
1922 | if (unlikely(!(tun->dev->flags & IFF_UP))) { | |
1923 | err = -EIO; | |
9180bb4f | 1924 | rcu_read_unlock(); |
4477138f ED |
1925 | goto drop; |
1926 | } | |
1927 | ||
90e33d45 | 1928 | if (frags) { |
96aa1b22 WB |
1929 | u32 headlen; |
1930 | ||
90e33d45 | 1931 | /* Exercise flow dissector code path. */ |
96aa1b22 WB |
1932 | skb_push(skb, ETH_HLEN); |
1933 | headlen = eth_get_headlen(tun->dev, skb->data, | |
1934 | skb_headlen(skb)); | |
90e33d45 | 1935 | |
010f245b | 1936 | if (unlikely(headlen > skb_headlen(skb))) { |
497a5757 | 1937 | atomic_long_inc(&tun->dev->rx_dropped); |
90e33d45 | 1938 | napi_free_frags(&tfile->napi); |
4477138f | 1939 | rcu_read_unlock(); |
90e33d45 PP |
1940 | mutex_unlock(&tfile->napi_mutex); |
1941 | WARN_ON(1); | |
1942 | return -ENOMEM; | |
1943 | } | |
1944 | ||
1945 | local_bh_disable(); | |
1946 | napi_gro_frags(&tfile->napi); | |
1947 | local_bh_enable(); | |
1948 | mutex_unlock(&tfile->napi_mutex); | |
aec72f33 | 1949 | } else if (tfile->napi_enabled) { |
94317099 PP |
1950 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; |
1951 | int queue_len; | |
1952 | ||
1953 | spin_lock_bh(&queue->lock); | |
1954 | __skb_queue_tail(queue, skb); | |
1955 | queue_len = skb_queue_len(queue); | |
1956 | spin_unlock(&queue->lock); | |
1957 | ||
1958 | if (!more || queue_len > NAPI_POLL_WEIGHT) | |
1959 | napi_schedule(&tfile->napi); | |
1960 | ||
1961 | local_bh_enable(); | |
1962 | } else if (!IS_ENABLED(CONFIG_4KSTACKS)) { | |
1963 | tun_rx_batched(tun, tfile, skb, more); | |
1964 | } else { | |
1965 | netif_rx_ni(skb); | |
1966 | } | |
4477138f | 1967 | rcu_read_unlock(); |
6aa20a22 | 1968 | |
497a5757 HK |
1969 | preempt_disable(); |
1970 | dev_sw_netstats_rx_add(tun->dev, len); | |
1971 | preempt_enable(); | |
1da177e4 | 1972 | |
96f84061 JW |
1973 | if (rxhash) |
1974 | tun_flow_update(tun, rxhash, tfile); | |
1975 | ||
0690899b | 1976 | return total_len; |
6aa20a22 | 1977 | } |
1da177e4 | 1978 | |
f5ff53b4 | 1979 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1980 | { |
33dccbb0 | 1981 | struct file *file = iocb->ki_filp; |
54f968d6 | 1982 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1983 | struct tun_struct *tun = tun_get(tfile); |
631ab46b | 1984 | ssize_t result; |
5aac0390 | 1985 | int noblock = 0; |
1da177e4 LT |
1986 | |
1987 | if (!tun) | |
1988 | return -EBADFD; | |
1989 | ||
5aac0390 JA |
1990 | if ((file->f_flags & O_NONBLOCK) || (iocb->ki_flags & IOCB_NOWAIT)) |
1991 | noblock = 1; | |
1992 | ||
1993 | result = tun_get_user(tun, tfile, NULL, from, noblock, false); | |
631ab46b EB |
1994 | |
1995 | tun_put(tun); | |
1996 | return result; | |
1da177e4 LT |
1997 | } |
1998 | ||
fc72d1d5 JW |
1999 | static ssize_t tun_put_user_xdp(struct tun_struct *tun, |
2000 | struct tun_file *tfile, | |
1ffcbc85 | 2001 | struct xdp_frame *xdp_frame, |
fc72d1d5 JW |
2002 | struct iov_iter *iter) |
2003 | { | |
2004 | int vnet_hdr_sz = 0; | |
1ffcbc85 | 2005 | size_t size = xdp_frame->len; |
fc72d1d5 JW |
2006 | size_t ret; |
2007 | ||
2008 | if (tun->flags & IFF_VNET_HDR) { | |
2009 | struct virtio_net_hdr gso = { 0 }; | |
2010 | ||
2011 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); | |
2012 | if (unlikely(iov_iter_count(iter) < vnet_hdr_sz)) | |
2013 | return -EINVAL; | |
2014 | if (unlikely(copy_to_iter(&gso, sizeof(gso), iter) != | |
2015 | sizeof(gso))) | |
2016 | return -EFAULT; | |
2017 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
2018 | } | |
2019 | ||
1ffcbc85 | 2020 | ret = copy_to_iter(xdp_frame->data, size, iter) + vnet_hdr_sz; |
fc72d1d5 | 2021 | |
497a5757 HK |
2022 | preempt_disable(); |
2023 | dev_sw_netstats_tx_add(tun->dev, 1, ret); | |
2024 | preempt_enable(); | |
fc72d1d5 JW |
2025 | |
2026 | return ret; | |
2027 | } | |
2028 | ||
1da177e4 | 2029 | /* Put packet to the user space buffer */ |
6f7c156c | 2030 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 2031 | struct tun_file *tfile, |
6f7c156c | 2032 | struct sk_buff *skb, |
e0b46d0e | 2033 | struct iov_iter *iter) |
1da177e4 LT |
2034 | { |
2035 | struct tun_pi pi = { 0, skb->protocol }; | |
e0b46d0e | 2036 | ssize_t total; |
8c847d25 | 2037 | int vlan_offset = 0; |
a8f9bfdf | 2038 | int vlan_hlen = 0; |
2eb783c4 | 2039 | int vnet_hdr_sz = 0; |
a8f9bfdf | 2040 | |
df8a39de | 2041 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 2042 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 2043 | |
40630b82 | 2044 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 2045 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 2046 | |
e0b46d0e HX |
2047 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
2048 | ||
40630b82 | 2049 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 2050 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
2051 | return -EINVAL; |
2052 | ||
e0b46d0e HX |
2053 | total += sizeof(pi); |
2054 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
2055 | /* Packet will be striped */ |
2056 | pi.flags |= TUN_PKT_STRIP; | |
2057 | } | |
6aa20a22 | 2058 | |
e0b46d0e | 2059 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 2060 | return -EFAULT; |
6aa20a22 | 2061 | } |
1da177e4 | 2062 | |
2eb783c4 | 2063 | if (vnet_hdr_sz) { |
9403cd7c | 2064 | struct virtio_net_hdr gso; |
34166093 | 2065 | |
e0b46d0e | 2066 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
2067 | return -EINVAL; |
2068 | ||
3e9e40e7 | 2069 | if (virtio_net_hdr_from_skb(skb, &gso, |
fd3a8862 WB |
2070 | tun_is_little_endian(tun), true, |
2071 | vlan_hlen)) { | |
f43798c2 | 2072 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
2073 | pr_err("unexpected GSO type: " |
2074 | "0x%x, gso_size %d, hdr_len %d\n", | |
2075 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
2076 | tun16_to_cpu(tun, gso.hdr_len)); | |
2077 | print_hex_dump(KERN_ERR, "tun: ", | |
2078 | DUMP_PREFIX_NONE, | |
2079 | 16, 1, skb->head, | |
2080 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
2081 | WARN_ON_ONCE(1); | |
2082 | return -EINVAL; | |
2083 | } | |
f43798c2 | 2084 | |
e0b46d0e | 2085 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 2086 | return -EFAULT; |
8c847d25 JW |
2087 | |
2088 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
2089 | } |
2090 | ||
a8f9bfdf | 2091 | if (vlan_hlen) { |
e0b46d0e | 2092 | int ret; |
aff3d70a | 2093 | struct veth veth; |
6680ec68 JW |
2094 | |
2095 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 2096 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
2097 | |
2098 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 2099 | |
e0b46d0e HX |
2100 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
2101 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
2102 | goto done; |
2103 | ||
e0b46d0e HX |
2104 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
2105 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
2106 | goto done; |
2107 | } | |
1da177e4 | 2108 | |
e0b46d0e | 2109 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 2110 | |
6680ec68 | 2111 | done: |
608b9977 | 2112 | /* caller is in process context, */ |
497a5757 HK |
2113 | preempt_disable(); |
2114 | dev_sw_netstats_tx_add(tun->dev, 1, skb->len + vlan_hlen); | |
2115 | preempt_enable(); | |
1da177e4 LT |
2116 | |
2117 | return total; | |
2118 | } | |
2119 | ||
fc72d1d5 | 2120 | static void *tun_ring_recv(struct tun_file *tfile, int noblock, int *err) |
1576d986 JW |
2121 | { |
2122 | DECLARE_WAITQUEUE(wait, current); | |
fc72d1d5 | 2123 | void *ptr = NULL; |
f48cc6b2 | 2124 | int error = 0; |
1576d986 | 2125 | |
fc72d1d5 JW |
2126 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2127 | if (ptr) | |
1576d986 JW |
2128 | goto out; |
2129 | if (noblock) { | |
f48cc6b2 | 2130 | error = -EAGAIN; |
1576d986 JW |
2131 | goto out; |
2132 | } | |
2133 | ||
333f7909 | 2134 | add_wait_queue(&tfile->socket.wq.wait, &wait); |
1576d986 JW |
2135 | |
2136 | while (1) { | |
71828b22 | 2137 | set_current_state(TASK_INTERRUPTIBLE); |
fc72d1d5 JW |
2138 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2139 | if (ptr) | |
1576d986 JW |
2140 | break; |
2141 | if (signal_pending(current)) { | |
f48cc6b2 | 2142 | error = -ERESTARTSYS; |
1576d986 JW |
2143 | break; |
2144 | } | |
2145 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 2146 | error = -EFAULT; |
1576d986 JW |
2147 | break; |
2148 | } | |
2149 | ||
2150 | schedule(); | |
2151 | } | |
2152 | ||
ecef67cb | 2153 | __set_current_state(TASK_RUNNING); |
333f7909 | 2154 | remove_wait_queue(&tfile->socket.wq.wait, &wait); |
1576d986 JW |
2155 | |
2156 | out: | |
f48cc6b2 | 2157 | *err = error; |
fc72d1d5 | 2158 | return ptr; |
1576d986 JW |
2159 | } |
2160 | ||
54f968d6 | 2161 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 | 2162 | struct iov_iter *to, |
fc72d1d5 | 2163 | int noblock, void *ptr) |
1da177e4 | 2164 | { |
9b067034 | 2165 | ssize_t ret; |
1576d986 | 2166 | int err; |
1da177e4 | 2167 | |
c33ee15b | 2168 | if (!iov_iter_count(to)) { |
fc72d1d5 | 2169 | tun_ptr_free(ptr); |
9b067034 | 2170 | return 0; |
c33ee15b | 2171 | } |
1da177e4 | 2172 | |
fc72d1d5 | 2173 | if (!ptr) { |
ac77cfd4 | 2174 | /* Read frames from ring */ |
fc72d1d5 JW |
2175 | ptr = tun_ring_recv(tfile, noblock, &err); |
2176 | if (!ptr) | |
ac77cfd4 JW |
2177 | return err; |
2178 | } | |
e0b46d0e | 2179 | |
1ffcbc85 JDB |
2180 | if (tun_is_xdp_frame(ptr)) { |
2181 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2182 | |
1ffcbc85 | 2183 | ret = tun_put_user_xdp(tun, tfile, xdpf, to); |
03993094 | 2184 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
2185 | } else { |
2186 | struct sk_buff *skb = ptr; | |
2187 | ||
2188 | ret = tun_put_user(tun, tfile, skb, to); | |
2189 | if (unlikely(ret < 0)) | |
2190 | kfree_skb(skb); | |
2191 | else | |
2192 | consume_skb(skb); | |
2193 | } | |
1da177e4 | 2194 | |
05c2828c MT |
2195 | return ret; |
2196 | } | |
2197 | ||
9b067034 | 2198 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
2199 | { |
2200 | struct file *file = iocb->ki_filp; | |
2201 | struct tun_file *tfile = file->private_data; | |
9484dc74 | 2202 | struct tun_struct *tun = tun_get(tfile); |
9b067034 | 2203 | ssize_t len = iov_iter_count(to), ret; |
5aac0390 | 2204 | int noblock = 0; |
05c2828c MT |
2205 | |
2206 | if (!tun) | |
2207 | return -EBADFD; | |
5aac0390 JA |
2208 | |
2209 | if ((file->f_flags & O_NONBLOCK) || (iocb->ki_flags & IOCB_NOWAIT)) | |
2210 | noblock = 1; | |
2211 | ||
2212 | ret = tun_do_read(tun, tfile, to, noblock, NULL); | |
42404c09 | 2213 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
2214 | if (ret > 0) |
2215 | iocb->ki_pos = ret; | |
631ab46b | 2216 | tun_put(tun); |
1da177e4 LT |
2217 | return ret; |
2218 | } | |
2219 | ||
cd5681d7 | 2220 | static void tun_prog_free(struct rcu_head *rcu) |
96f84061 | 2221 | { |
cd5681d7 | 2222 | struct tun_prog *prog = container_of(rcu, struct tun_prog, rcu); |
96f84061 JW |
2223 | |
2224 | bpf_prog_destroy(prog->prog); | |
2225 | kfree(prog); | |
2226 | } | |
2227 | ||
9d6474e4 JW |
2228 | static int __tun_set_ebpf(struct tun_struct *tun, |
2229 | struct tun_prog __rcu **prog_p, | |
cd5681d7 | 2230 | struct bpf_prog *prog) |
96f84061 | 2231 | { |
cd5681d7 | 2232 | struct tun_prog *old, *new = NULL; |
96f84061 JW |
2233 | |
2234 | if (prog) { | |
2235 | new = kmalloc(sizeof(*new), GFP_KERNEL); | |
2236 | if (!new) | |
2237 | return -ENOMEM; | |
2238 | new->prog = prog; | |
2239 | } | |
2240 | ||
124da8f6 | 2241 | spin_lock_bh(&tun->lock); |
cd5681d7 | 2242 | old = rcu_dereference_protected(*prog_p, |
124da8f6 | 2243 | lockdep_is_held(&tun->lock)); |
cd5681d7 | 2244 | rcu_assign_pointer(*prog_p, new); |
124da8f6 | 2245 | spin_unlock_bh(&tun->lock); |
96f84061 JW |
2246 | |
2247 | if (old) | |
cd5681d7 | 2248 | call_rcu(&old->rcu, tun_prog_free); |
96f84061 JW |
2249 | |
2250 | return 0; | |
2251 | } | |
2252 | ||
96442e42 JW |
2253 | static void tun_free_netdev(struct net_device *dev) |
2254 | { | |
2255 | struct tun_struct *tun = netdev_priv(dev); | |
2256 | ||
4008e97f | 2257 | BUG_ON(!(list_empty(&tun->disabled))); |
11fc7d5a | 2258 | |
497a5757 | 2259 | free_percpu(dev->tstats); |
96442e42 | 2260 | tun_flow_uninit(tun); |
5dbbaf2d | 2261 | security_tun_dev_free_security(tun->security); |
cd5681d7 | 2262 | __tun_set_ebpf(tun, &tun->steering_prog, NULL); |
aff3d70a | 2263 | __tun_set_ebpf(tun, &tun->filter_prog, NULL); |
96442e42 JW |
2264 | } |
2265 | ||
1da177e4 LT |
2266 | static void tun_setup(struct net_device *dev) |
2267 | { | |
2268 | struct tun_struct *tun = netdev_priv(dev); | |
2269 | ||
0625c883 EB |
2270 | tun->owner = INVALID_UID; |
2271 | tun->group = INVALID_GID; | |
4e24f2dd | 2272 | tun_default_link_ksettings(dev, &tun->link_ksettings); |
1da177e4 | 2273 | |
1da177e4 | 2274 | dev->ethtool_ops = &tun_ethtool_ops; |
cf124db5 DM |
2275 | dev->needs_free_netdev = true; |
2276 | dev->priv_destructor = tun_free_netdev; | |
016adb72 JW |
2277 | /* We prefer our own queue length */ |
2278 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
2279 | } |
2280 | ||
f019a7a5 EB |
2281 | /* Trivial set of netlink ops to allow deleting tun or tap |
2282 | * device with netlink. | |
2283 | */ | |
a8b8a889 MS |
2284 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[], |
2285 | struct netlink_ext_ack *extack) | |
f019a7a5 | 2286 | { |
35b827b6 ND |
2287 | NL_SET_ERR_MSG(extack, |
2288 | "tun/tap creation via rtnetlink is not supported."); | |
2289 | return -EOPNOTSUPP; | |
f019a7a5 EB |
2290 | } |
2291 | ||
1ec010e7 SD |
2292 | static size_t tun_get_size(const struct net_device *dev) |
2293 | { | |
2294 | BUILD_BUG_ON(sizeof(u32) != sizeof(uid_t)); | |
2295 | BUILD_BUG_ON(sizeof(u32) != sizeof(gid_t)); | |
2296 | ||
2297 | return nla_total_size(sizeof(uid_t)) + /* OWNER */ | |
2298 | nla_total_size(sizeof(gid_t)) + /* GROUP */ | |
2299 | nla_total_size(sizeof(u8)) + /* TYPE */ | |
2300 | nla_total_size(sizeof(u8)) + /* PI */ | |
2301 | nla_total_size(sizeof(u8)) + /* VNET_HDR */ | |
2302 | nla_total_size(sizeof(u8)) + /* PERSIST */ | |
2303 | nla_total_size(sizeof(u8)) + /* MULTI_QUEUE */ | |
2304 | nla_total_size(sizeof(u32)) + /* NUM_QUEUES */ | |
2305 | nla_total_size(sizeof(u32)) + /* NUM_DISABLED_QUEUES */ | |
2306 | 0; | |
2307 | } | |
2308 | ||
2309 | static int tun_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
2310 | { | |
2311 | struct tun_struct *tun = netdev_priv(dev); | |
2312 | ||
2313 | if (nla_put_u8(skb, IFLA_TUN_TYPE, tun->flags & TUN_TYPE_MASK)) | |
2314 | goto nla_put_failure; | |
2315 | if (uid_valid(tun->owner) && | |
2316 | nla_put_u32(skb, IFLA_TUN_OWNER, | |
2317 | from_kuid_munged(current_user_ns(), tun->owner))) | |
2318 | goto nla_put_failure; | |
2319 | if (gid_valid(tun->group) && | |
2320 | nla_put_u32(skb, IFLA_TUN_GROUP, | |
2321 | from_kgid_munged(current_user_ns(), tun->group))) | |
2322 | goto nla_put_failure; | |
2323 | if (nla_put_u8(skb, IFLA_TUN_PI, !(tun->flags & IFF_NO_PI))) | |
2324 | goto nla_put_failure; | |
2325 | if (nla_put_u8(skb, IFLA_TUN_VNET_HDR, !!(tun->flags & IFF_VNET_HDR))) | |
2326 | goto nla_put_failure; | |
2327 | if (nla_put_u8(skb, IFLA_TUN_PERSIST, !!(tun->flags & IFF_PERSIST))) | |
2328 | goto nla_put_failure; | |
2329 | if (nla_put_u8(skb, IFLA_TUN_MULTI_QUEUE, | |
2330 | !!(tun->flags & IFF_MULTI_QUEUE))) | |
2331 | goto nla_put_failure; | |
2332 | if (tun->flags & IFF_MULTI_QUEUE) { | |
2333 | if (nla_put_u32(skb, IFLA_TUN_NUM_QUEUES, tun->numqueues)) | |
2334 | goto nla_put_failure; | |
2335 | if (nla_put_u32(skb, IFLA_TUN_NUM_DISABLED_QUEUES, | |
2336 | tun->numdisabled)) | |
2337 | goto nla_put_failure; | |
2338 | } | |
2339 | ||
2340 | return 0; | |
2341 | ||
2342 | nla_put_failure: | |
2343 | return -EMSGSIZE; | |
2344 | } | |
2345 | ||
f019a7a5 EB |
2346 | static struct rtnl_link_ops tun_link_ops __read_mostly = { |
2347 | .kind = DRV_NAME, | |
2348 | .priv_size = sizeof(struct tun_struct), | |
2349 | .setup = tun_setup, | |
2350 | .validate = tun_validate, | |
1ec010e7 SD |
2351 | .get_size = tun_get_size, |
2352 | .fill_info = tun_fill_info, | |
f019a7a5 EB |
2353 | }; |
2354 | ||
33dccbb0 HX |
2355 | static void tun_sock_write_space(struct sock *sk) |
2356 | { | |
54f968d6 | 2357 | struct tun_file *tfile; |
43815482 | 2358 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
2359 | |
2360 | if (!sock_writeable(sk)) | |
2361 | return; | |
2362 | ||
9cd3e072 | 2363 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
2364 | return; |
2365 | ||
43815482 ED |
2366 | wqueue = sk_sleep(sk); |
2367 | if (wqueue && waitqueue_active(wqueue)) | |
a9a08845 LT |
2368 | wake_up_interruptible_sync_poll(wqueue, EPOLLOUT | |
2369 | EPOLLWRNORM | EPOLLWRBAND); | |
c722c625 | 2370 | |
54f968d6 JW |
2371 | tfile = container_of(sk, struct tun_file, sk); |
2372 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
2373 | } |
2374 | ||
f9e06c45 JW |
2375 | static void tun_put_page(struct tun_page *tpage) |
2376 | { | |
2377 | if (tpage->page) | |
2378 | __page_frag_cache_drain(tpage->page, tpage->count); | |
2379 | } | |
2380 | ||
043d222f JW |
2381 | static int tun_xdp_one(struct tun_struct *tun, |
2382 | struct tun_file *tfile, | |
f9e06c45 JW |
2383 | struct xdp_buff *xdp, int *flush, |
2384 | struct tun_page *tpage) | |
043d222f | 2385 | { |
4e4b08e5 | 2386 | unsigned int datasize = xdp->data_end - xdp->data; |
043d222f JW |
2387 | struct tun_xdp_hdr *hdr = xdp->data_hard_start; |
2388 | struct virtio_net_hdr *gso = &hdr->gso; | |
043d222f JW |
2389 | struct bpf_prog *xdp_prog; |
2390 | struct sk_buff *skb = NULL; | |
fb3f9037 | 2391 | struct sk_buff_head *queue; |
043d222f JW |
2392 | u32 rxhash = 0, act; |
2393 | int buflen = hdr->buflen; | |
fb3f9037 | 2394 | int ret = 0; |
043d222f | 2395 | bool skb_xdp = false; |
f9e06c45 | 2396 | struct page *page; |
043d222f JW |
2397 | |
2398 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
2399 | if (xdp_prog) { | |
2400 | if (gso->gso_type) { | |
2401 | skb_xdp = true; | |
2402 | goto build; | |
2403 | } | |
43b5169d LB |
2404 | |
2405 | xdp_init_buff(xdp, buflen, &tfile->xdp_rxq); | |
043d222f | 2406 | xdp_set_data_meta_invalid(xdp); |
043d222f JW |
2407 | |
2408 | act = bpf_prog_run_xdp(xdp_prog, xdp); | |
fb3f9037 HH |
2409 | ret = tun_xdp_act(tun, xdp_prog, xdp, act); |
2410 | if (ret < 0) { | |
043d222f | 2411 | put_page(virt_to_head_page(xdp->data)); |
fb3f9037 | 2412 | return ret; |
043d222f JW |
2413 | } |
2414 | ||
fb3f9037 | 2415 | switch (ret) { |
043d222f JW |
2416 | case XDP_REDIRECT: |
2417 | *flush = true; | |
df561f66 | 2418 | fallthrough; |
043d222f JW |
2419 | case XDP_TX: |
2420 | return 0; | |
2421 | case XDP_PASS: | |
2422 | break; | |
2423 | default: | |
f9e06c45 JW |
2424 | page = virt_to_head_page(xdp->data); |
2425 | if (tpage->page == page) { | |
2426 | ++tpage->count; | |
2427 | } else { | |
2428 | tun_put_page(tpage); | |
2429 | tpage->page = page; | |
2430 | tpage->count = 1; | |
2431 | } | |
043d222f JW |
2432 | return 0; |
2433 | } | |
2434 | } | |
2435 | ||
2436 | build: | |
2437 | skb = build_skb(xdp->data_hard_start, buflen); | |
2438 | if (!skb) { | |
fb3f9037 | 2439 | ret = -ENOMEM; |
043d222f JW |
2440 | goto out; |
2441 | } | |
2442 | ||
2443 | skb_reserve(skb, xdp->data - xdp->data_hard_start); | |
2444 | skb_put(skb, xdp->data_end - xdp->data); | |
2445 | ||
2446 | if (virtio_net_hdr_to_skb(skb, gso, tun_is_little_endian(tun))) { | |
497a5757 | 2447 | atomic_long_inc(&tun->rx_frame_errors); |
043d222f | 2448 | kfree_skb(skb); |
fb3f9037 | 2449 | ret = -EINVAL; |
043d222f JW |
2450 | goto out; |
2451 | } | |
2452 | ||
2453 | skb->protocol = eth_type_trans(skb, tun->dev); | |
2454 | skb_reset_network_header(skb); | |
d2aa125d | 2455 | skb_probe_transport_header(skb); |
3fe260e0 | 2456 | skb_record_rx_queue(skb, tfile->queue_index); |
043d222f JW |
2457 | |
2458 | if (skb_xdp) { | |
fb3f9037 HH |
2459 | ret = do_xdp_generic(xdp_prog, skb); |
2460 | if (ret != XDP_PASS) { | |
2461 | ret = 0; | |
043d222f | 2462 | goto out; |
fb3f9037 | 2463 | } |
043d222f JW |
2464 | } |
2465 | ||
f29eb2a9 PA |
2466 | if (!rcu_dereference(tun->steering_prog) && tun->numqueues > 1 && |
2467 | !tfile->detached) | |
043d222f JW |
2468 | rxhash = __skb_get_hash_symmetric(skb); |
2469 | ||
fb3f9037 HH |
2470 | if (tfile->napi_enabled) { |
2471 | queue = &tfile->sk.sk_write_queue; | |
2472 | spin_lock(&queue->lock); | |
2473 | __skb_queue_tail(queue, skb); | |
2474 | spin_unlock(&queue->lock); | |
2475 | ret = 1; | |
2476 | } else { | |
2477 | netif_receive_skb(skb); | |
2478 | ret = 0; | |
2479 | } | |
043d222f | 2480 | |
497a5757 | 2481 | /* No need to disable preemption here since this function is |
6342ca64 PB |
2482 | * always called with bh disabled |
2483 | */ | |
497a5757 | 2484 | dev_sw_netstats_rx_add(tun->dev, datasize); |
043d222f JW |
2485 | |
2486 | if (rxhash) | |
2487 | tun_flow_update(tun, rxhash, tfile); | |
2488 | ||
2489 | out: | |
fb3f9037 | 2490 | return ret; |
043d222f JW |
2491 | } |
2492 | ||
1b784140 | 2493 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 2494 | { |
043d222f | 2495 | int ret, i; |
54f968d6 | 2496 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2497 | struct tun_struct *tun = tun_get(tfile); |
fe8dd45b | 2498 | struct tun_msg_ctl *ctl = m->msg_control; |
043d222f | 2499 | struct xdp_buff *xdp; |
54f968d6 JW |
2500 | |
2501 | if (!tun) | |
2502 | return -EBADFD; | |
f5ff53b4 | 2503 | |
74a335a0 HH |
2504 | if (m->msg_controllen == sizeof(struct tun_msg_ctl) && |
2505 | ctl && ctl->type == TUN_MSG_PTR) { | |
6f0271d9 | 2506 | struct tun_page tpage; |
043d222f | 2507 | int n = ctl->num; |
fb3f9037 | 2508 | int flush = 0, queued = 0; |
043d222f | 2509 | |
6f0271d9 DM |
2510 | memset(&tpage, 0, sizeof(tpage)); |
2511 | ||
043d222f JW |
2512 | local_bh_disable(); |
2513 | rcu_read_lock(); | |
2514 | ||
2515 | for (i = 0; i < n; i++) { | |
2516 | xdp = &((struct xdp_buff *)ctl->ptr)[i]; | |
fb3f9037 HH |
2517 | ret = tun_xdp_one(tun, tfile, xdp, &flush, &tpage); |
2518 | if (ret > 0) | |
2519 | queued += ret; | |
043d222f JW |
2520 | } |
2521 | ||
2522 | if (flush) | |
1d233886 | 2523 | xdp_do_flush(); |
043d222f | 2524 | |
fb3f9037 HH |
2525 | if (tfile->napi_enabled && queued > 0) |
2526 | napi_schedule(&tfile->napi); | |
2527 | ||
043d222f JW |
2528 | rcu_read_unlock(); |
2529 | local_bh_enable(); | |
2530 | ||
f9e06c45 JW |
2531 | tun_put_page(&tpage); |
2532 | ||
043d222f JW |
2533 | ret = total_len; |
2534 | goto out; | |
2535 | } | |
fe8dd45b JW |
2536 | |
2537 | ret = tun_get_user(tun, tfile, ctl ? ctl->ptr : NULL, &m->msg_iter, | |
5503fcec JW |
2538 | m->msg_flags & MSG_DONTWAIT, |
2539 | m->msg_flags & MSG_MORE); | |
043d222f | 2540 | out: |
54f968d6 JW |
2541 | tun_put(tun); |
2542 | return ret; | |
05c2828c MT |
2543 | } |
2544 | ||
1b784140 | 2545 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
2546 | int flags) |
2547 | { | |
54f968d6 | 2548 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2549 | struct tun_struct *tun = tun_get(tfile); |
fc72d1d5 | 2550 | void *ptr = m->msg_control; |
05c2828c | 2551 | int ret; |
54f968d6 | 2552 | |
c33ee15b WX |
2553 | if (!tun) { |
2554 | ret = -EBADFD; | |
fc72d1d5 | 2555 | goto out_free; |
c33ee15b | 2556 | } |
54f968d6 | 2557 | |
eda29772 | 2558 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 | 2559 | ret = -EINVAL; |
c33ee15b | 2560 | goto out_put_tun; |
3811ae76 | 2561 | } |
eda29772 RC |
2562 | if (flags & MSG_ERRQUEUE) { |
2563 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
2564 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
2565 | goto out; | |
2566 | } | |
fc72d1d5 | 2567 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, ptr); |
87897931 | 2568 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
2569 | m->msg_flags |= MSG_TRUNC; |
2570 | ret = flags & MSG_TRUNC ? ret : total_len; | |
2571 | } | |
3811ae76 | 2572 | out: |
54f968d6 | 2573 | tun_put(tun); |
05c2828c | 2574 | return ret; |
c33ee15b WX |
2575 | |
2576 | out_put_tun: | |
2577 | tun_put(tun); | |
fc72d1d5 JW |
2578 | out_free: |
2579 | tun_ptr_free(ptr); | |
c33ee15b | 2580 | return ret; |
05c2828c MT |
2581 | } |
2582 | ||
fc72d1d5 JW |
2583 | static int tun_ptr_peek_len(void *ptr) |
2584 | { | |
2585 | if (likely(ptr)) { | |
1ffcbc85 JDB |
2586 | if (tun_is_xdp_frame(ptr)) { |
2587 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2588 | |
1ffcbc85 | 2589 | return xdpf->len; |
fc72d1d5 JW |
2590 | } |
2591 | return __skb_array_len_with_tag(ptr); | |
2592 | } else { | |
2593 | return 0; | |
2594 | } | |
2595 | } | |
2596 | ||
1576d986 JW |
2597 | static int tun_peek_len(struct socket *sock) |
2598 | { | |
2599 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
2600 | struct tun_struct *tun; | |
2601 | int ret = 0; | |
2602 | ||
9484dc74 | 2603 | tun = tun_get(tfile); |
1576d986 JW |
2604 | if (!tun) |
2605 | return 0; | |
2606 | ||
fc72d1d5 | 2607 | ret = PTR_RING_PEEK_CALL(&tfile->tx_ring, tun_ptr_peek_len); |
1576d986 JW |
2608 | tun_put(tun); |
2609 | ||
2610 | return ret; | |
2611 | } | |
2612 | ||
05c2828c MT |
2613 | /* Ops structure to mimic raw sockets with tun */ |
2614 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 2615 | .peek_len = tun_peek_len, |
05c2828c MT |
2616 | .sendmsg = tun_sendmsg, |
2617 | .recvmsg = tun_recvmsg, | |
2618 | }; | |
2619 | ||
33dccbb0 HX |
2620 | static struct proto tun_proto = { |
2621 | .name = "tun", | |
2622 | .owner = THIS_MODULE, | |
54f968d6 | 2623 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 2624 | }; |
f019a7a5 | 2625 | |
980c9e8c DW |
2626 | static int tun_flags(struct tun_struct *tun) |
2627 | { | |
031f5e03 | 2628 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
2629 | } |
2630 | ||
bc6d076d | 2631 | static ssize_t tun_flags_show(struct device *dev, struct device_attribute *attr, |
980c9e8c DW |
2632 | char *buf) |
2633 | { | |
2634 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
2635 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
2636 | } | |
2637 | ||
bc6d076d Y |
2638 | static ssize_t owner_show(struct device *dev, struct device_attribute *attr, |
2639 | char *buf) | |
980c9e8c DW |
2640 | { |
2641 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2642 | return uid_valid(tun->owner)? |
2643 | sprintf(buf, "%u\n", | |
2644 | from_kuid_munged(current_user_ns(), tun->owner)): | |
2645 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2646 | } |
2647 | ||
bc6d076d Y |
2648 | static ssize_t group_show(struct device *dev, struct device_attribute *attr, |
2649 | char *buf) | |
980c9e8c DW |
2650 | { |
2651 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2652 | return gid_valid(tun->group) ? |
2653 | sprintf(buf, "%u\n", | |
2654 | from_kgid_munged(current_user_ns(), tun->group)): | |
2655 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2656 | } |
2657 | ||
bc6d076d Y |
2658 | static DEVICE_ATTR_RO(tun_flags); |
2659 | static DEVICE_ATTR_RO(owner); | |
2660 | static DEVICE_ATTR_RO(group); | |
980c9e8c | 2661 | |
c4d33e24 TI |
2662 | static struct attribute *tun_dev_attrs[] = { |
2663 | &dev_attr_tun_flags.attr, | |
2664 | &dev_attr_owner.attr, | |
2665 | &dev_attr_group.attr, | |
2666 | NULL | |
2667 | }; | |
2668 | ||
2669 | static const struct attribute_group tun_attr_group = { | |
2670 | .attrs = tun_dev_attrs | |
2671 | }; | |
2672 | ||
d647a591 | 2673 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
2674 | { |
2675 | struct tun_struct *tun; | |
54f968d6 | 2676 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2677 | struct net_device *dev; |
2678 | int err; | |
2679 | ||
7c0c3b1a JW |
2680 | if (tfile->detached) |
2681 | return -EINVAL; | |
2682 | ||
90e33d45 PP |
2683 | if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) { |
2684 | if (!capable(CAP_NET_ADMIN)) | |
2685 | return -EPERM; | |
2686 | ||
2687 | if (!(ifr->ifr_flags & IFF_NAPI) || | |
2688 | (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP) | |
2689 | return -EINVAL; | |
2690 | } | |
2691 | ||
74a3e5a7 EB |
2692 | dev = __dev_get_by_name(net, ifr->ifr_name); |
2693 | if (dev) { | |
f85ba780 DW |
2694 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
2695 | return -EBUSY; | |
74a3e5a7 EB |
2696 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
2697 | tun = netdev_priv(dev); | |
2698 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
2699 | tun = netdev_priv(dev); | |
2700 | else | |
2701 | return -EINVAL; | |
2702 | ||
8e6d91ae | 2703 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 2704 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
2705 | return -EINVAL; |
2706 | ||
cde8b15f | 2707 | if (tun_not_capable(tun)) |
2b980dbd | 2708 | return -EPERM; |
5dbbaf2d | 2709 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
2710 | if (err < 0) |
2711 | return err; | |
2712 | ||
94317099 | 2713 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER, |
af3fb24e | 2714 | ifr->ifr_flags & IFF_NAPI, |
77f22f92 | 2715 | ifr->ifr_flags & IFF_NAPI_FRAGS, true); |
a7385ba2 EB |
2716 | if (err < 0) |
2717 | return err; | |
4008e97f | 2718 | |
40630b82 | 2719 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
2720 | (tun->numqueues + tun->numdisabled > 1)) { |
2721 | /* One or more queue has already been attached, no need | |
2722 | * to initialize the device again. | |
2723 | */ | |
83c1f36f | 2724 | netdev_state_change(dev); |
e8dbad66 JW |
2725 | return 0; |
2726 | } | |
9fffc5c6 SD |
2727 | |
2728 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
2729 | (ifr->ifr_flags & TUN_FEATURES); | |
83c1f36f SD |
2730 | |
2731 | netdev_state_change(dev); | |
2732 | } else { | |
1da177e4 LT |
2733 | char *name; |
2734 | unsigned long flags = 0; | |
edfb6a14 JW |
2735 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
2736 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 2737 | |
c260b772 | 2738 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 2739 | return -EPERM; |
2b980dbd PM |
2740 | err = security_tun_dev_create(); |
2741 | if (err < 0) | |
2742 | return err; | |
ca6bb5d7 | 2743 | |
1da177e4 LT |
2744 | /* Set dev type */ |
2745 | if (ifr->ifr_flags & IFF_TUN) { | |
2746 | /* TUN device */ | |
40630b82 | 2747 | flags |= IFF_TUN; |
1da177e4 LT |
2748 | name = "tun%d"; |
2749 | } else if (ifr->ifr_flags & IFF_TAP) { | |
2750 | /* TAP device */ | |
40630b82 | 2751 | flags |= IFF_TAP; |
1da177e4 | 2752 | name = "tap%d"; |
6aa20a22 | 2753 | } else |
36989b90 | 2754 | return -EINVAL; |
6aa20a22 | 2755 | |
1da177e4 LT |
2756 | if (*ifr->ifr_name) |
2757 | name = ifr->ifr_name; | |
2758 | ||
c8d68e6b | 2759 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
2760 | NET_NAME_UNKNOWN, tun_setup, queues, |
2761 | queues); | |
edfb6a14 | 2762 | |
1da177e4 LT |
2763 | if (!dev) |
2764 | return -ENOMEM; | |
2765 | ||
fc54c658 | 2766 | dev_net_set(dev, net); |
f019a7a5 | 2767 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 2768 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 2769 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 2770 | |
1da177e4 LT |
2771 | tun = netdev_priv(dev); |
2772 | tun->dev = dev; | |
2773 | tun->flags = flags; | |
f271b2cc | 2774 | tun->txflt.count = 0; |
d9d52b51 | 2775 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 2776 | |
eaea34b2 | 2777 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
2778 | tun->filter_attached = false; |
2779 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 2780 | tun->rx_batched = 0; |
96f84061 | 2781 | RCU_INIT_POINTER(tun->steering_prog, NULL); |
33dccbb0 | 2782 | |
158b515f GK |
2783 | tun->ifr = ifr; |
2784 | tun->file = file; | |
608b9977 | 2785 | |
158b515f | 2786 | tun_net_initialize(dev); |
eb0fb363 | 2787 | |
1da177e4 | 2788 | err = register_netdevice(tun->dev); |
158b515f GK |
2789 | if (err < 0) { |
2790 | free_netdev(dev); | |
2791 | return err; | |
2792 | } | |
c2e315b8 | 2793 | /* free_netdev() won't check refcnt, to avoid race |
77f22f92 YY |
2794 | * with dev_put() we need publish tun after registration. |
2795 | */ | |
2796 | rcu_assign_pointer(tfile->tun, tun); | |
1da177e4 LT |
2797 | } |
2798 | ||
af668b3c MT |
2799 | netif_carrier_on(tun->dev); |
2800 | ||
e35259a9 MK |
2801 | /* Make sure persistent devices do not get stuck in |
2802 | * xoff state. | |
2803 | */ | |
2804 | if (netif_running(tun->dev)) | |
c8d68e6b | 2805 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 2806 | |
1da177e4 LT |
2807 | strcpy(ifr->ifr_name, tun->dev->name); |
2808 | return 0; | |
1da177e4 LT |
2809 | } |
2810 | ||
12132768 | 2811 | static void tun_get_iff(struct tun_struct *tun, struct ifreq *ifr) |
e3b99556 | 2812 | { |
e3b99556 MM |
2813 | strcpy(ifr->ifr_name, tun->dev->name); |
2814 | ||
980c9e8c | 2815 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 2816 | |
e3b99556 MM |
2817 | } |
2818 | ||
5228ddc9 RR |
2819 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
2820 | * privs required. */ | |
88255375 | 2821 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 2822 | { |
c8f44aff | 2823 | netdev_features_t features = 0; |
5228ddc9 RR |
2824 | |
2825 | if (arg & TUN_F_CSUM) { | |
88255375 | 2826 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
2827 | arg &= ~TUN_F_CSUM; |
2828 | ||
2829 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
2830 | if (arg & TUN_F_TSO_ECN) { | |
2831 | features |= NETIF_F_TSO_ECN; | |
2832 | arg &= ~TUN_F_TSO_ECN; | |
2833 | } | |
2834 | if (arg & TUN_F_TSO4) | |
2835 | features |= NETIF_F_TSO; | |
2836 | if (arg & TUN_F_TSO6) | |
2837 | features |= NETIF_F_TSO6; | |
2838 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
2839 | } | |
0c19f846 WB |
2840 | |
2841 | arg &= ~TUN_F_UFO; | |
5228ddc9 RR |
2842 | } |
2843 | ||
2844 | /* This gives the user a way to test for new features in future by | |
2845 | * trying to set them. */ | |
2846 | if (arg) | |
2847 | return -EINVAL; | |
2848 | ||
88255375 | 2849 | tun->set_features = features; |
09050957 YI |
2850 | tun->dev->wanted_features &= ~TUN_USER_FEATURES; |
2851 | tun->dev->wanted_features |= features; | |
88255375 | 2852 | netdev_update_features(tun->dev); |
5228ddc9 RR |
2853 | |
2854 | return 0; | |
2855 | } | |
2856 | ||
c8d68e6b JW |
2857 | static void tun_detach_filter(struct tun_struct *tun, int n) |
2858 | { | |
2859 | int i; | |
2860 | struct tun_file *tfile; | |
2861 | ||
2862 | for (i = 0; i < n; i++) { | |
b8deabd3 | 2863 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2864 | lock_sock(tfile->socket.sk); |
2865 | sk_detach_filter(tfile->socket.sk); | |
2866 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2867 | } |
2868 | ||
2869 | tun->filter_attached = false; | |
2870 | } | |
2871 | ||
2872 | static int tun_attach_filter(struct tun_struct *tun) | |
2873 | { | |
2874 | int i, ret = 0; | |
2875 | struct tun_file *tfile; | |
2876 | ||
2877 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2878 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2879 | lock_sock(tfile->socket.sk); |
2880 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
2881 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2882 | if (ret) { |
2883 | tun_detach_filter(tun, i); | |
2884 | return ret; | |
2885 | } | |
2886 | } | |
2887 | ||
2888 | tun->filter_attached = true; | |
2889 | return ret; | |
2890 | } | |
2891 | ||
2892 | static void tun_set_sndbuf(struct tun_struct *tun) | |
2893 | { | |
2894 | struct tun_file *tfile; | |
2895 | int i; | |
2896 | ||
2897 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2898 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
2899 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
2900 | } | |
2901 | } | |
2902 | ||
cde8b15f JW |
2903 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
2904 | { | |
2905 | struct tun_file *tfile = file->private_data; | |
2906 | struct tun_struct *tun; | |
cde8b15f JW |
2907 | int ret = 0; |
2908 | ||
2909 | rtnl_lock(); | |
2910 | ||
2911 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 2912 | tun = tfile->detached; |
5dbbaf2d | 2913 | if (!tun) { |
cde8b15f | 2914 | ret = -EINVAL; |
5dbbaf2d PM |
2915 | goto unlock; |
2916 | } | |
2917 | ret = security_tun_dev_attach_queue(tun->security); | |
2918 | if (ret < 0) | |
2919 | goto unlock; | |
af3fb24e | 2920 | ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI, |
77f22f92 | 2921 | tun->flags & IFF_NAPI_FRAGS, true); |
4008e97f | 2922 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 2923 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 2924 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
2925 | ret = -EINVAL; |
2926 | else | |
2927 | __tun_detach(tfile, false); | |
2928 | } else | |
cde8b15f JW |
2929 | ret = -EINVAL; |
2930 | ||
83c1f36f SD |
2931 | if (ret >= 0) |
2932 | netdev_state_change(tun->dev); | |
2933 | ||
5dbbaf2d | 2934 | unlock: |
cde8b15f JW |
2935 | rtnl_unlock(); |
2936 | return ret; | |
2937 | } | |
2938 | ||
8f3f330d | 2939 | static int tun_set_ebpf(struct tun_struct *tun, struct tun_prog __rcu **prog_p, |
cd5681d7 | 2940 | void __user *data) |
96f84061 JW |
2941 | { |
2942 | struct bpf_prog *prog; | |
2943 | int fd; | |
2944 | ||
2945 | if (copy_from_user(&fd, data, sizeof(fd))) | |
2946 | return -EFAULT; | |
2947 | ||
2948 | if (fd == -1) { | |
2949 | prog = NULL; | |
2950 | } else { | |
2951 | prog = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); | |
2952 | if (IS_ERR(prog)) | |
2953 | return PTR_ERR(prog); | |
2954 | } | |
2955 | ||
cd5681d7 | 2956 | return __tun_set_ebpf(tun, prog_p, prog); |
96f84061 JW |
2957 | } |
2958 | ||
cca8ea3b PP |
2959 | /* Return correct value for tun->dev->addr_len based on tun->dev->type. */ |
2960 | static unsigned char tun_get_addr_len(unsigned short type) | |
2961 | { | |
2962 | switch (type) { | |
2963 | case ARPHRD_IP6GRE: | |
2964 | case ARPHRD_TUNNEL6: | |
2965 | return sizeof(struct in6_addr); | |
2966 | case ARPHRD_IPGRE: | |
2967 | case ARPHRD_TUNNEL: | |
2968 | case ARPHRD_SIT: | |
2969 | return 4; | |
2970 | case ARPHRD_ETHER: | |
2971 | return ETH_ALEN; | |
2972 | case ARPHRD_IEEE802154: | |
2973 | case ARPHRD_IEEE802154_MONITOR: | |
2974 | return IEEE802154_EXTENDED_ADDR_LEN; | |
2975 | case ARPHRD_PHONET_PIPE: | |
2976 | case ARPHRD_PPP: | |
2977 | case ARPHRD_NONE: | |
2978 | return 0; | |
2979 | case ARPHRD_6LOWPAN: | |
2980 | return EUI64_ADDR_LEN; | |
2981 | case ARPHRD_FDDI: | |
2982 | return FDDI_K_ALEN; | |
2983 | case ARPHRD_HIPPI: | |
2984 | return HIPPI_ALEN; | |
2985 | case ARPHRD_IEEE802: | |
2986 | return FC_ALEN; | |
2987 | case ARPHRD_ROSE: | |
2988 | return ROSE_ADDR_LEN; | |
2989 | case ARPHRD_NETROM: | |
2990 | return AX25_ADDR_LEN; | |
2991 | case ARPHRD_LOCALTLK: | |
2992 | return LTALK_ALEN; | |
2993 | default: | |
2994 | return 0; | |
2995 | } | |
2996 | } | |
2997 | ||
50857e2a AB |
2998 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
2999 | unsigned long arg, int ifreq_len) | |
1da177e4 | 3000 | { |
36b50bab | 3001 | struct tun_file *tfile = file->private_data; |
f663706a | 3002 | struct net *net = sock_net(&tfile->sk); |
631ab46b | 3003 | struct tun_struct *tun; |
1da177e4 | 3004 | void __user* argp = (void __user*)arg; |
26d31925 | 3005 | unsigned int ifindex, carrier; |
1da177e4 | 3006 | struct ifreq ifr; |
0625c883 EB |
3007 | kuid_t owner; |
3008 | kgid_t group; | |
33dccbb0 | 3009 | int sndbuf; |
d9d52b51 | 3010 | int vnet_hdr_sz; |
1cf8e410 | 3011 | int le; |
f271b2cc | 3012 | int ret; |
83c1f36f | 3013 | bool do_notify = false; |
1da177e4 | 3014 | |
f2780d6d KT |
3015 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || |
3016 | (_IOC_TYPE(cmd) == SOCK_IOC_TYPE && cmd != SIOCGSKNS)) { | |
50857e2a | 3017 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 3018 | return -EFAULT; |
8bbb1813 | 3019 | } else { |
a117dacd | 3020 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 3021 | } |
631ab46b EB |
3022 | if (cmd == TUNGETFEATURES) { |
3023 | /* Currently this just means: "what IFF flags are valid?". | |
3024 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
3025 | * TUNSETIFF. |
3026 | */ | |
3027 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 3028 | (unsigned int __user*)argp); |
f663706a | 3029 | } else if (cmd == TUNSETQUEUE) { |
cde8b15f | 3030 | return tun_set_queue(file, &ifr); |
f663706a KT |
3031 | } else if (cmd == SIOCGSKNS) { |
3032 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
3033 | return -EPERM; | |
3034 | return open_related_ns(&net->ns, get_net_ns); | |
3035 | } | |
631ab46b | 3036 | |
876bfd4d HX |
3037 | rtnl_lock(); |
3038 | ||
9484dc74 | 3039 | tun = tun_get(tfile); |
0f16bc13 GF |
3040 | if (cmd == TUNSETIFF) { |
3041 | ret = -EEXIST; | |
3042 | if (tun) | |
3043 | goto unlock; | |
3044 | ||
1da177e4 LT |
3045 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
3046 | ||
f2780d6d | 3047 | ret = tun_set_iff(net, file, &ifr); |
1da177e4 | 3048 | |
876bfd4d HX |
3049 | if (ret) |
3050 | goto unlock; | |
1da177e4 | 3051 | |
50857e2a | 3052 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
3053 | ret = -EFAULT; |
3054 | goto unlock; | |
1da177e4 | 3055 | } |
fb7589a1 PE |
3056 | if (cmd == TUNSETIFINDEX) { |
3057 | ret = -EPERM; | |
3058 | if (tun) | |
3059 | goto unlock; | |
3060 | ||
3061 | ret = -EFAULT; | |
3062 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
3063 | goto unlock; | |
3064 | ||
3065 | ret = 0; | |
3066 | tfile->ifindex = ifindex; | |
3067 | goto unlock; | |
3068 | } | |
1da177e4 | 3069 | |
876bfd4d | 3070 | ret = -EBADFD; |
1da177e4 | 3071 | if (!tun) |
876bfd4d | 3072 | goto unlock; |
1da177e4 | 3073 | |
3424170f | 3074 | netif_info(tun, drv, tun->dev, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 3075 | |
0c3e0e3b | 3076 | net = dev_net(tun->dev); |
631ab46b | 3077 | ret = 0; |
1da177e4 | 3078 | switch (cmd) { |
e3b99556 | 3079 | case TUNGETIFF: |
12132768 | 3080 | tun_get_iff(tun, &ifr); |
e3b99556 | 3081 | |
3d407a80 PE |
3082 | if (tfile->detached) |
3083 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
3084 | if (!tfile->socket.sk->sk_filter) |
3085 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 3086 | |
50857e2a | 3087 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 3088 | ret = -EFAULT; |
e3b99556 MM |
3089 | break; |
3090 | ||
1da177e4 LT |
3091 | case TUNSETNOCSUM: |
3092 | /* Disable/Enable checksum */ | |
1da177e4 | 3093 | |
88255375 | 3094 | /* [unimplemented] */ |
3424170f MK |
3095 | netif_info(tun, drv, tun->dev, "ignored: set checksum %s\n", |
3096 | arg ? "disabled" : "enabled"); | |
1da177e4 LT |
3097 | break; |
3098 | ||
3099 | case TUNSETPERSIST: | |
54f968d6 JW |
3100 | /* Disable/Enable persist mode. Keep an extra reference to the |
3101 | * module to prevent the module being unprobed. | |
3102 | */ | |
40630b82 MT |
3103 | if (arg && !(tun->flags & IFF_PERSIST)) { |
3104 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 3105 | __module_get(THIS_MODULE); |
83c1f36f | 3106 | do_notify = true; |
dd38bd85 | 3107 | } |
40630b82 MT |
3108 | if (!arg && (tun->flags & IFF_PERSIST)) { |
3109 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 | 3110 | module_put(THIS_MODULE); |
83c1f36f | 3111 | do_notify = true; |
54f968d6 | 3112 | } |
1da177e4 | 3113 | |
3424170f MK |
3114 | netif_info(tun, drv, tun->dev, "persist %s\n", |
3115 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
3116 | break; |
3117 | ||
3118 | case TUNSETOWNER: | |
3119 | /* Set owner of the device */ | |
0625c883 EB |
3120 | owner = make_kuid(current_user_ns(), arg); |
3121 | if (!uid_valid(owner)) { | |
3122 | ret = -EINVAL; | |
3123 | break; | |
3124 | } | |
3125 | tun->owner = owner; | |
83c1f36f | 3126 | do_notify = true; |
3424170f MK |
3127 | netif_info(tun, drv, tun->dev, "owner set to %u\n", |
3128 | from_kuid(&init_user_ns, tun->owner)); | |
1da177e4 LT |
3129 | break; |
3130 | ||
8c644623 GG |
3131 | case TUNSETGROUP: |
3132 | /* Set group of the device */ | |
0625c883 EB |
3133 | group = make_kgid(current_user_ns(), arg); |
3134 | if (!gid_valid(group)) { | |
3135 | ret = -EINVAL; | |
3136 | break; | |
3137 | } | |
3138 | tun->group = group; | |
83c1f36f | 3139 | do_notify = true; |
3424170f MK |
3140 | netif_info(tun, drv, tun->dev, "group set to %u\n", |
3141 | from_kgid(&init_user_ns, tun->group)); | |
8c644623 GG |
3142 | break; |
3143 | ||
ff4cc3ac MK |
3144 | case TUNSETLINK: |
3145 | /* Only allow setting the type when the interface is down */ | |
3146 | if (tun->dev->flags & IFF_UP) { | |
3424170f MK |
3147 | netif_info(tun, drv, tun->dev, |
3148 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 3149 | ret = -EBUSY; |
ff4cc3ac | 3150 | } else { |
8e1e33ff MS |
3151 | ret = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
3152 | tun->dev); | |
3153 | ret = notifier_to_errno(ret); | |
3154 | if (ret) { | |
3155 | netif_info(tun, drv, tun->dev, | |
3156 | "Refused to change device type\n"); | |
3157 | break; | |
3158 | } | |
ff4cc3ac | 3159 | tun->dev->type = (int) arg; |
cca8ea3b | 3160 | tun->dev->addr_len = tun_get_addr_len(tun->dev->type); |
3424170f MK |
3161 | netif_info(tun, drv, tun->dev, "linktype set to %d\n", |
3162 | tun->dev->type); | |
8e1e33ff MS |
3163 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
3164 | tun->dev); | |
ff4cc3ac | 3165 | } |
631ab46b | 3166 | break; |
ff4cc3ac | 3167 | |
1da177e4 | 3168 | case TUNSETDEBUG: |
3424170f | 3169 | tun->msg_enable = (u32)arg; |
1da177e4 | 3170 | break; |
3424170f | 3171 | |
5228ddc9 | 3172 | case TUNSETOFFLOAD: |
88255375 | 3173 | ret = set_offload(tun, arg); |
631ab46b | 3174 | break; |
5228ddc9 | 3175 | |
f271b2cc MK |
3176 | case TUNSETTXFILTER: |
3177 | /* Can be set only for TAPs */ | |
631ab46b | 3178 | ret = -EINVAL; |
40630b82 | 3179 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 3180 | break; |
c0e5a8c2 | 3181 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 3182 | break; |
1da177e4 LT |
3183 | |
3184 | case SIOCGIFHWADDR: | |
b595076a | 3185 | /* Get hw address */ |
3b23a32a | 3186 | dev_get_mac_address(&ifr.ifr_hwaddr, net, tun->dev->name); |
50857e2a | 3187 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
3188 | ret = -EFAULT; |
3189 | break; | |
1da177e4 LT |
3190 | |
3191 | case SIOCSIFHWADDR: | |
f271b2cc | 3192 | /* Set hw address */ |
3b23a32a | 3193 | ret = dev_set_mac_address_user(tun->dev, &ifr.ifr_hwaddr, NULL); |
631ab46b | 3194 | break; |
33dccbb0 HX |
3195 | |
3196 | case TUNGETSNDBUF: | |
54f968d6 | 3197 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
3198 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
3199 | ret = -EFAULT; | |
3200 | break; | |
3201 | ||
3202 | case TUNSETSNDBUF: | |
3203 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
3204 | ret = -EFAULT; | |
3205 | break; | |
3206 | } | |
93161922 CG |
3207 | if (sndbuf <= 0) { |
3208 | ret = -EINVAL; | |
3209 | break; | |
3210 | } | |
33dccbb0 | 3211 | |
c8d68e6b JW |
3212 | tun->sndbuf = sndbuf; |
3213 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
3214 | break; |
3215 | ||
d9d52b51 MT |
3216 | case TUNGETVNETHDRSZ: |
3217 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
3218 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
3219 | ret = -EFAULT; | |
3220 | break; | |
3221 | ||
3222 | case TUNSETVNETHDRSZ: | |
3223 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
3224 | ret = -EFAULT; | |
3225 | break; | |
3226 | } | |
3227 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
3228 | ret = -EINVAL; | |
3229 | break; | |
3230 | } | |
3231 | ||
3232 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
3233 | break; | |
3234 | ||
1cf8e410 MT |
3235 | case TUNGETVNETLE: |
3236 | le = !!(tun->flags & TUN_VNET_LE); | |
3237 | if (put_user(le, (int __user *)argp)) | |
3238 | ret = -EFAULT; | |
3239 | break; | |
3240 | ||
3241 | case TUNSETVNETLE: | |
3242 | if (get_user(le, (int __user *)argp)) { | |
3243 | ret = -EFAULT; | |
3244 | break; | |
3245 | } | |
3246 | if (le) | |
3247 | tun->flags |= TUN_VNET_LE; | |
3248 | else | |
3249 | tun->flags &= ~TUN_VNET_LE; | |
3250 | break; | |
3251 | ||
8b8e658b GK |
3252 | case TUNGETVNETBE: |
3253 | ret = tun_get_vnet_be(tun, argp); | |
3254 | break; | |
3255 | ||
3256 | case TUNSETVNETBE: | |
3257 | ret = tun_set_vnet_be(tun, argp); | |
3258 | break; | |
3259 | ||
99405162 MT |
3260 | case TUNATTACHFILTER: |
3261 | /* Can be set only for TAPs */ | |
3262 | ret = -EINVAL; | |
40630b82 | 3263 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
3264 | break; |
3265 | ret = -EFAULT; | |
54f968d6 | 3266 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
3267 | break; |
3268 | ||
c8d68e6b | 3269 | ret = tun_attach_filter(tun); |
99405162 MT |
3270 | break; |
3271 | ||
3272 | case TUNDETACHFILTER: | |
3273 | /* Can be set only for TAPs */ | |
3274 | ret = -EINVAL; | |
40630b82 | 3275 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 3276 | break; |
c8d68e6b JW |
3277 | ret = 0; |
3278 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
3279 | break; |
3280 | ||
76975e9c PE |
3281 | case TUNGETFILTER: |
3282 | ret = -EINVAL; | |
40630b82 | 3283 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
3284 | break; |
3285 | ret = -EFAULT; | |
3286 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
3287 | break; | |
3288 | ret = 0; | |
3289 | break; | |
3290 | ||
96f84061 | 3291 | case TUNSETSTEERINGEBPF: |
cd5681d7 | 3292 | ret = tun_set_ebpf(tun, &tun->steering_prog, argp); |
96f84061 JW |
3293 | break; |
3294 | ||
aff3d70a JW |
3295 | case TUNSETFILTEREBPF: |
3296 | ret = tun_set_ebpf(tun, &tun->filter_prog, argp); | |
3297 | break; | |
3298 | ||
26d31925 ND |
3299 | case TUNSETCARRIER: |
3300 | ret = -EFAULT; | |
3301 | if (copy_from_user(&carrier, argp, sizeof(carrier))) | |
3302 | goto unlock; | |
3303 | ||
3304 | ret = tun_net_change_carrier(tun->dev, (bool)carrier); | |
3305 | break; | |
3306 | ||
0c3e0e3b KT |
3307 | case TUNGETDEVNETNS: |
3308 | ret = -EPERM; | |
3309 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
3310 | goto unlock; | |
3311 | ret = open_related_ns(&net->ns, get_net_ns); | |
3312 | break; | |
3313 | ||
1da177e4 | 3314 | default: |
631ab46b EB |
3315 | ret = -EINVAL; |
3316 | break; | |
ee289b64 | 3317 | } |
1da177e4 | 3318 | |
83c1f36f SD |
3319 | if (do_notify) |
3320 | netdev_state_change(tun->dev); | |
3321 | ||
876bfd4d HX |
3322 | unlock: |
3323 | rtnl_unlock(); | |
3324 | if (tun) | |
3325 | tun_put(tun); | |
631ab46b | 3326 | return ret; |
1da177e4 LT |
3327 | } |
3328 | ||
50857e2a AB |
3329 | static long tun_chr_ioctl(struct file *file, |
3330 | unsigned int cmd, unsigned long arg) | |
3331 | { | |
3332 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
3333 | } | |
3334 | ||
3335 | #ifdef CONFIG_COMPAT | |
3336 | static long tun_chr_compat_ioctl(struct file *file, | |
3337 | unsigned int cmd, unsigned long arg) | |
3338 | { | |
3339 | switch (cmd) { | |
3340 | case TUNSETIFF: | |
3341 | case TUNGETIFF: | |
3342 | case TUNSETTXFILTER: | |
3343 | case TUNGETSNDBUF: | |
3344 | case TUNSETSNDBUF: | |
3345 | case SIOCGIFHWADDR: | |
3346 | case SIOCSIFHWADDR: | |
3347 | arg = (unsigned long)compat_ptr(arg); | |
3348 | break; | |
3349 | default: | |
3350 | arg = (compat_ulong_t)arg; | |
3351 | break; | |
3352 | } | |
3353 | ||
3354 | /* | |
3355 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
3356 | * the end of that structure. All fields that are used in this | |
3357 | * driver are compatible though, we don't need to convert the | |
3358 | * contents. | |
3359 | */ | |
3360 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
3361 | } | |
3362 | #endif /* CONFIG_COMPAT */ | |
3363 | ||
1da177e4 LT |
3364 | static int tun_chr_fasync(int fd, struct file *file, int on) |
3365 | { | |
54f968d6 | 3366 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
3367 | int ret; |
3368 | ||
54f968d6 | 3369 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 3370 | goto out; |
6aa20a22 | 3371 | |
1da177e4 | 3372 | if (on) { |
01919134 | 3373 | __f_setown(file, task_pid(current), PIDTYPE_TGID, 0); |
54f968d6 | 3374 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 3375 | } else |
54f968d6 | 3376 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
3377 | ret = 0; |
3378 | out: | |
9d319522 | 3379 | return ret; |
1da177e4 LT |
3380 | } |
3381 | ||
3382 | static int tun_chr_open(struct inode *inode, struct file * file) | |
3383 | { | |
140e807d | 3384 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 3385 | struct tun_file *tfile; |
deed49fb | 3386 | |
140e807d | 3387 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 3388 | &tun_proto, 0); |
631ab46b EB |
3389 | if (!tfile) |
3390 | return -ENOMEM; | |
b196d88a JW |
3391 | if (ptr_ring_init(&tfile->tx_ring, 0, GFP_KERNEL)) { |
3392 | sk_free(&tfile->sk); | |
3393 | return -ENOMEM; | |
3394 | } | |
3395 | ||
c7256f57 | 3396 | mutex_init(&tfile->napi_mutex); |
c956674b | 3397 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 3398 | tfile->flags = 0; |
fb7589a1 | 3399 | tfile->ifindex = 0; |
54f968d6 | 3400 | |
333f7909 | 3401 | init_waitqueue_head(&tfile->socket.wq.wait); |
54f968d6 JW |
3402 | |
3403 | tfile->socket.file = file; | |
3404 | tfile->socket.ops = &tun_socket_ops; | |
3405 | ||
3406 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
3407 | |
3408 | tfile->sk.sk_write_space = tun_sock_write_space; | |
3409 | tfile->sk.sk_sndbuf = INT_MAX; | |
3410 | ||
631ab46b | 3411 | file->private_data = tfile; |
4008e97f | 3412 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 3413 | |
19a6afb2 JW |
3414 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
3415 | ||
1da177e4 LT |
3416 | return 0; |
3417 | } | |
3418 | ||
3419 | static int tun_chr_close(struct inode *inode, struct file *file) | |
3420 | { | |
631ab46b | 3421 | struct tun_file *tfile = file->private_data; |
1da177e4 | 3422 | |
c8d68e6b | 3423 | tun_detach(tfile, true); |
1da177e4 LT |
3424 | |
3425 | return 0; | |
3426 | } | |
3427 | ||
93e14b6d | 3428 | #ifdef CONFIG_PROC_FS |
9484dc74 | 3429 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file) |
93e14b6d | 3430 | { |
9484dc74 | 3431 | struct tun_file *tfile = file->private_data; |
93e14b6d MY |
3432 | struct tun_struct *tun; |
3433 | struct ifreq ifr; | |
3434 | ||
3435 | memset(&ifr, 0, sizeof(ifr)); | |
3436 | ||
3437 | rtnl_lock(); | |
9484dc74 | 3438 | tun = tun_get(tfile); |
93e14b6d | 3439 | if (tun) |
12132768 | 3440 | tun_get_iff(tun, &ifr); |
93e14b6d MY |
3441 | rtnl_unlock(); |
3442 | ||
3443 | if (tun) | |
3444 | tun_put(tun); | |
3445 | ||
a3816ab0 | 3446 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
3447 | } |
3448 | #endif | |
3449 | ||
d54b1fdb | 3450 | static const struct file_operations tun_fops = { |
6aa20a22 | 3451 | .owner = THIS_MODULE, |
1da177e4 | 3452 | .llseek = no_llseek, |
9b067034 | 3453 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 3454 | .write_iter = tun_chr_write_iter, |
1da177e4 | 3455 | .poll = tun_chr_poll, |
50857e2a AB |
3456 | .unlocked_ioctl = tun_chr_ioctl, |
3457 | #ifdef CONFIG_COMPAT | |
3458 | .compat_ioctl = tun_chr_compat_ioctl, | |
3459 | #endif | |
1da177e4 LT |
3460 | .open = tun_chr_open, |
3461 | .release = tun_chr_close, | |
93e14b6d MY |
3462 | .fasync = tun_chr_fasync, |
3463 | #ifdef CONFIG_PROC_FS | |
3464 | .show_fdinfo = tun_chr_show_fdinfo, | |
3465 | #endif | |
1da177e4 LT |
3466 | }; |
3467 | ||
3468 | static struct miscdevice tun_miscdev = { | |
3469 | .minor = TUN_MINOR, | |
3470 | .name = "tun", | |
e454cea2 | 3471 | .nodename = "net/tun", |
1da177e4 | 3472 | .fops = &tun_fops, |
1da177e4 LT |
3473 | }; |
3474 | ||
3475 | /* ethtool interface */ | |
3476 | ||
4e24f2dd CW |
3477 | static void tun_default_link_ksettings(struct net_device *dev, |
3478 | struct ethtool_link_ksettings *cmd) | |
29ccc49d PR |
3479 | { |
3480 | ethtool_link_ksettings_zero_link_mode(cmd, supported); | |
3481 | ethtool_link_ksettings_zero_link_mode(cmd, advertising); | |
3482 | cmd->base.speed = SPEED_10; | |
3483 | cmd->base.duplex = DUPLEX_FULL; | |
3484 | cmd->base.port = PORT_TP; | |
3485 | cmd->base.phy_address = 0; | |
3486 | cmd->base.autoneg = AUTONEG_DISABLE; | |
4e24f2dd CW |
3487 | } |
3488 | ||
3489 | static int tun_get_link_ksettings(struct net_device *dev, | |
3490 | struct ethtool_link_ksettings *cmd) | |
3491 | { | |
3492 | struct tun_struct *tun = netdev_priv(dev); | |
3493 | ||
3494 | memcpy(cmd, &tun->link_ksettings, sizeof(*cmd)); | |
3495 | return 0; | |
3496 | } | |
3497 | ||
3498 | static int tun_set_link_ksettings(struct net_device *dev, | |
3499 | const struct ethtool_link_ksettings *cmd) | |
3500 | { | |
3501 | struct tun_struct *tun = netdev_priv(dev); | |
3502 | ||
3503 | memcpy(&tun->link_ksettings, cmd, sizeof(*cmd)); | |
1da177e4 LT |
3504 | return 0; |
3505 | } | |
3506 | ||
3507 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
3508 | { | |
3509 | struct tun_struct *tun = netdev_priv(dev); | |
3510 | ||
33a5ba14 RJ |
3511 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
3512 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
3513 | |
3514 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 3515 | case IFF_TUN: |
33a5ba14 | 3516 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 3517 | break; |
40630b82 | 3518 | case IFF_TAP: |
33a5ba14 | 3519 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
3520 | break; |
3521 | } | |
3522 | } | |
3523 | ||
3524 | static u32 tun_get_msglevel(struct net_device *dev) | |
3525 | { | |
1da177e4 | 3526 | struct tun_struct *tun = netdev_priv(dev); |
3424170f MK |
3527 | |
3528 | return tun->msg_enable; | |
1da177e4 LT |
3529 | } |
3530 | ||
3531 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
3532 | { | |
1da177e4 | 3533 | struct tun_struct *tun = netdev_priv(dev); |
3424170f MK |
3534 | |
3535 | tun->msg_enable = value; | |
1da177e4 LT |
3536 | } |
3537 | ||
5503fcec | 3538 | static int tun_get_coalesce(struct net_device *dev, |
f3ccfda1 YM |
3539 | struct ethtool_coalesce *ec, |
3540 | struct kernel_ethtool_coalesce *kernel_coal, | |
3541 | struct netlink_ext_ack *extack) | |
5503fcec JW |
3542 | { |
3543 | struct tun_struct *tun = netdev_priv(dev); | |
3544 | ||
3545 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
3546 | ||
3547 | return 0; | |
3548 | } | |
3549 | ||
3550 | static int tun_set_coalesce(struct net_device *dev, | |
f3ccfda1 YM |
3551 | struct ethtool_coalesce *ec, |
3552 | struct kernel_ethtool_coalesce *kernel_coal, | |
3553 | struct netlink_ext_ack *extack) | |
5503fcec JW |
3554 | { |
3555 | struct tun_struct *tun = netdev_priv(dev); | |
3556 | ||
3557 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
3558 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
3559 | else | |
3560 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
3561 | ||
3562 | return 0; | |
3563 | } | |
3564 | ||
7282d491 | 3565 | static const struct ethtool_ops tun_ethtool_ops = { |
e5ad00b3 | 3566 | .supported_coalesce_params = ETHTOOL_COALESCE_RX_MAX_FRAMES, |
1da177e4 LT |
3567 | .get_drvinfo = tun_get_drvinfo, |
3568 | .get_msglevel = tun_get_msglevel, | |
3569 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 3570 | .get_link = ethtool_op_get_link, |
eda29772 | 3571 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
3572 | .get_coalesce = tun_get_coalesce, |
3573 | .set_coalesce = tun_set_coalesce, | |
29ccc49d | 3574 | .get_link_ksettings = tun_get_link_ksettings, |
4e24f2dd | 3575 | .set_link_ksettings = tun_set_link_ksettings, |
1da177e4 LT |
3576 | }; |
3577 | ||
1576d986 JW |
3578 | static int tun_queue_resize(struct tun_struct *tun) |
3579 | { | |
3580 | struct net_device *dev = tun->dev; | |
3581 | struct tun_file *tfile; | |
5990a305 | 3582 | struct ptr_ring **rings; |
1576d986 JW |
3583 | int n = tun->numqueues + tun->numdisabled; |
3584 | int ret, i; | |
3585 | ||
5990a305 JW |
3586 | rings = kmalloc_array(n, sizeof(*rings), GFP_KERNEL); |
3587 | if (!rings) | |
1576d986 JW |
3588 | return -ENOMEM; |
3589 | ||
3590 | for (i = 0; i < tun->numqueues; i++) { | |
3591 | tfile = rtnl_dereference(tun->tfiles[i]); | |
5990a305 | 3592 | rings[i] = &tfile->tx_ring; |
1576d986 JW |
3593 | } |
3594 | list_for_each_entry(tfile, &tun->disabled, next) | |
5990a305 | 3595 | rings[i++] = &tfile->tx_ring; |
1576d986 | 3596 | |
5990a305 JW |
3597 | ret = ptr_ring_resize_multiple(rings, n, |
3598 | dev->tx_queue_len, GFP_KERNEL, | |
fc72d1d5 | 3599 | tun_ptr_free); |
1576d986 | 3600 | |
5990a305 | 3601 | kfree(rings); |
1576d986 JW |
3602 | return ret; |
3603 | } | |
3604 | ||
3605 | static int tun_device_event(struct notifier_block *unused, | |
3606 | unsigned long event, void *ptr) | |
3607 | { | |
3608 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
3609 | struct tun_struct *tun = netdev_priv(dev); | |
72b319dc | 3610 | int i; |
1576d986 | 3611 | |
86dfb4ac CG |
3612 | if (dev->rtnl_link_ops != &tun_link_ops) |
3613 | return NOTIFY_DONE; | |
3614 | ||
1576d986 JW |
3615 | switch (event) { |
3616 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
3617 | if (tun_queue_resize(tun)) | |
3618 | return NOTIFY_BAD; | |
3619 | break; | |
72b319dc FL |
3620 | case NETDEV_UP: |
3621 | for (i = 0; i < tun->numqueues; i++) { | |
3622 | struct tun_file *tfile; | |
3623 | ||
3624 | tfile = rtnl_dereference(tun->tfiles[i]); | |
3625 | tfile->socket.sk->sk_write_space(tfile->socket.sk); | |
3626 | } | |
3627 | break; | |
1576d986 JW |
3628 | default: |
3629 | break; | |
3630 | } | |
3631 | ||
3632 | return NOTIFY_DONE; | |
3633 | } | |
3634 | ||
3635 | static struct notifier_block tun_notifier_block __read_mostly = { | |
3636 | .notifier_call = tun_device_event, | |
3637 | }; | |
79d17604 | 3638 | |
1da177e4 LT |
3639 | static int __init tun_init(void) |
3640 | { | |
3641 | int ret = 0; | |
3642 | ||
6b8a66ee | 3643 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
1da177e4 | 3644 | |
f019a7a5 | 3645 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 3646 | if (ret) { |
6b8a66ee | 3647 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 3648 | goto err_linkops; |
79d17604 PE |
3649 | } |
3650 | ||
1da177e4 | 3651 | ret = misc_register(&tun_miscdev); |
79d17604 | 3652 | if (ret) { |
6b8a66ee | 3653 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
3654 | goto err_misc; |
3655 | } | |
1576d986 | 3656 | |
5edfbd3c TZ |
3657 | ret = register_netdevice_notifier(&tun_notifier_block); |
3658 | if (ret) { | |
3659 | pr_err("Can't register netdevice notifier\n"); | |
3660 | goto err_notifier; | |
3661 | } | |
3662 | ||
f019a7a5 | 3663 | return 0; |
5edfbd3c TZ |
3664 | |
3665 | err_notifier: | |
3666 | misc_deregister(&tun_miscdev); | |
79d17604 | 3667 | err_misc: |
f019a7a5 EB |
3668 | rtnl_link_unregister(&tun_link_ops); |
3669 | err_linkops: | |
1da177e4 LT |
3670 | return ret; |
3671 | } | |
3672 | ||
3673 | static void tun_cleanup(void) | |
3674 | { | |
6aa20a22 | 3675 | misc_deregister(&tun_miscdev); |
f019a7a5 | 3676 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 3677 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
3678 | } |
3679 | ||
05c2828c MT |
3680 | /* Get an underlying socket object from tun file. Returns error unless file is |
3681 | * attached to a device. The returned object works like a packet socket, it | |
3682 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
3683 | * holding a reference to the file for as long as the socket is in use. */ | |
3684 | struct socket *tun_get_socket(struct file *file) | |
3685 | { | |
6e914fc7 | 3686 | struct tun_file *tfile; |
05c2828c MT |
3687 | if (file->f_op != &tun_fops) |
3688 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
3689 | tfile = file->private_data; |
3690 | if (!tfile) | |
05c2828c | 3691 | return ERR_PTR(-EBADFD); |
54f968d6 | 3692 | return &tfile->socket; |
05c2828c MT |
3693 | } |
3694 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
3695 | ||
5990a305 | 3696 | struct ptr_ring *tun_get_tx_ring(struct file *file) |
83339c6b JW |
3697 | { |
3698 | struct tun_file *tfile; | |
3699 | ||
3700 | if (file->f_op != &tun_fops) | |
3701 | return ERR_PTR(-EINVAL); | |
3702 | tfile = file->private_data; | |
3703 | if (!tfile) | |
3704 | return ERR_PTR(-EBADFD); | |
5990a305 | 3705 | return &tfile->tx_ring; |
83339c6b | 3706 | } |
5990a305 | 3707 | EXPORT_SYMBOL_GPL(tun_get_tx_ring); |
83339c6b | 3708 | |
1da177e4 LT |
3709 | module_init(tun_init); |
3710 | module_exit(tun_cleanup); | |
3711 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
3712 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
3713 | MODULE_LICENSE("GPL"); | |
3714 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 3715 | MODULE_ALIAS("devname:net/tun"); |