]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * TUN - Universal TUN/TAP device driver. | |
3 | * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com> | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify | |
6 | * it under the terms of the GNU General Public License as published by | |
7 | * the Free Software Foundation; either version 2 of the License, or | |
8 | * (at your option) any later version. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, | |
11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
13 | * GNU General Public License for more details. | |
14 | * | |
15 | * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $ | |
16 | */ | |
17 | ||
18 | /* | |
19 | * Changes: | |
20 | * | |
ff4cc3ac MK |
21 | * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14 |
22 | * Add TUNSETLINK ioctl to set the link encapsulation | |
23 | * | |
1da177e4 | 24 | * Mark Smith <markzzzsmith@yahoo.com.au> |
344dc8ed | 25 | * Use eth_random_addr() for tap MAC address. |
1da177e4 LT |
26 | * |
27 | * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20 | |
28 | * Fixes in packet dropping, queue length setting and queue wakeup. | |
29 | * Increased default tx queue length. | |
30 | * Added ethtool API. | |
31 | * Minor cleanups | |
32 | * | |
33 | * Daniel Podlejski <underley@underley.eu.org> | |
34 | * Modifications for 2.3.99-pre5 kernel. | |
35 | */ | |
36 | ||
6b8a66ee JP |
37 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
38 | ||
1da177e4 LT |
39 | #define DRV_NAME "tun" |
40 | #define DRV_VERSION "1.6" | |
41 | #define DRV_DESCRIPTION "Universal TUN/TAP device driver" | |
42 | #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>" | |
43 | ||
1da177e4 LT |
44 | #include <linux/module.h> |
45 | #include <linux/errno.h> | |
46 | #include <linux/kernel.h> | |
174cd4b1 | 47 | #include <linux/sched/signal.h> |
1da177e4 LT |
48 | #include <linux/major.h> |
49 | #include <linux/slab.h> | |
50 | #include <linux/poll.h> | |
51 | #include <linux/fcntl.h> | |
52 | #include <linux/init.h> | |
53 | #include <linux/skbuff.h> | |
54 | #include <linux/netdevice.h> | |
55 | #include <linux/etherdevice.h> | |
56 | #include <linux/miscdevice.h> | |
57 | #include <linux/ethtool.h> | |
58 | #include <linux/rtnetlink.h> | |
50857e2a | 59 | #include <linux/compat.h> |
1da177e4 LT |
60 | #include <linux/if.h> |
61 | #include <linux/if_arp.h> | |
62 | #include <linux/if_ether.h> | |
63 | #include <linux/if_tun.h> | |
6680ec68 | 64 | #include <linux/if_vlan.h> |
1da177e4 | 65 | #include <linux/crc32.h> |
d647a591 | 66 | #include <linux/nsproxy.h> |
f43798c2 | 67 | #include <linux/virtio_net.h> |
99405162 | 68 | #include <linux/rcupdate.h> |
881d966b | 69 | #include <net/net_namespace.h> |
79d17604 | 70 | #include <net/netns/generic.h> |
f019a7a5 | 71 | #include <net/rtnetlink.h> |
33dccbb0 | 72 | #include <net/sock.h> |
93e14b6d | 73 | #include <linux/seq_file.h> |
e0b46d0e | 74 | #include <linux/uio.h> |
1576d986 | 75 | #include <linux/skb_array.h> |
761876c8 JW |
76 | #include <linux/bpf.h> |
77 | #include <linux/bpf_trace.h> | |
90e33d45 | 78 | #include <linux/mutex.h> |
1da177e4 | 79 | |
7c0f6ba6 | 80 | #include <linux/uaccess.h> |
f2780d6d | 81 | #include <linux/proc_fs.h> |
1da177e4 | 82 | |
14daa021 RR |
83 | /* Uncomment to enable debugging */ |
84 | /* #define TUN_DEBUG 1 */ | |
85 | ||
1da177e4 LT |
86 | #ifdef TUN_DEBUG |
87 | static int debug; | |
14daa021 | 88 | |
6b8a66ee JP |
89 | #define tun_debug(level, tun, fmt, args...) \ |
90 | do { \ | |
91 | if (tun->debug) \ | |
92 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
93 | } while (0) | |
94 | #define DBG1(level, fmt, args...) \ | |
95 | do { \ | |
96 | if (debug == 2) \ | |
97 | printk(level fmt, ##args); \ | |
98 | } while (0) | |
14daa021 | 99 | #else |
6b8a66ee JP |
100 | #define tun_debug(level, tun, fmt, args...) \ |
101 | do { \ | |
102 | if (0) \ | |
103 | netdev_printk(level, tun->dev, fmt, ##args); \ | |
104 | } while (0) | |
105 | #define DBG1(level, fmt, args...) \ | |
106 | do { \ | |
107 | if (0) \ | |
108 | printk(level fmt, ##args); \ | |
109 | } while (0) | |
14daa021 RR |
110 | #endif |
111 | ||
761876c8 | 112 | #define TUN_HEADROOM 256 |
7df13219 | 113 | #define TUN_RX_PAD (NET_IP_ALIGN + NET_SKB_PAD) |
66ccbc9c | 114 | |
031f5e03 MT |
115 | /* TUN device flags */ |
116 | ||
117 | /* IFF_ATTACH_QUEUE is never stored in device flags, | |
118 | * overload it to mean fasync when stored there. | |
119 | */ | |
120 | #define TUN_FASYNC IFF_ATTACH_QUEUE | |
1cf8e410 MT |
121 | /* High bits in flags field are unused. */ |
122 | #define TUN_VNET_LE 0x80000000 | |
8b8e658b | 123 | #define TUN_VNET_BE 0x40000000 |
031f5e03 MT |
124 | |
125 | #define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \ | |
90e33d45 PP |
126 | IFF_MULTI_QUEUE | IFF_NAPI | IFF_NAPI_FRAGS) |
127 | ||
0690899b MT |
128 | #define GOODCOPY_LEN 128 |
129 | ||
f271b2cc MK |
130 | #define FLT_EXACT_COUNT 8 |
131 | struct tap_filter { | |
132 | unsigned int count; /* Number of addrs. Zero means disabled */ | |
133 | u32 mask[2]; /* Mask of the hashed addrs */ | |
134 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | |
135 | }; | |
136 | ||
baf71c5c PG |
137 | /* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal |
138 | * to max number of VCPUs in guest. */ | |
139 | #define MAX_TAP_QUEUES 256 | |
b8732fb7 | 140 | #define MAX_TAP_FLOWS 4096 |
c8d68e6b | 141 | |
96442e42 JW |
142 | #define TUN_FLOW_EXPIRE (3 * HZ) |
143 | ||
608b9977 PA |
144 | struct tun_pcpu_stats { |
145 | u64 rx_packets; | |
146 | u64 rx_bytes; | |
147 | u64 tx_packets; | |
148 | u64 tx_bytes; | |
149 | struct u64_stats_sync syncp; | |
150 | u32 rx_dropped; | |
151 | u32 tx_dropped; | |
152 | u32 rx_frame_errors; | |
153 | }; | |
154 | ||
54f968d6 | 155 | /* A tun_file connects an open character device to a tuntap netdevice. It |
92d4ea6e | 156 | * also contains all socket related structures (except sock_fprog and tap_filter) |
54f968d6 JW |
157 | * to serve as one transmit queue for tuntap device. The sock_fprog and |
158 | * tap_filter were kept in tun_struct since they were used for filtering for the | |
36fe8c09 | 159 | * netdevice not for a specific queue (at least I didn't see the requirement for |
54f968d6 | 160 | * this). |
6e914fc7 JW |
161 | * |
162 | * RCU usage: | |
36fe8c09 | 163 | * The tun_file and tun_struct are loosely coupled, the pointer from one to the |
6e914fc7 | 164 | * other can only be read while rcu_read_lock or rtnl_lock is held. |
54f968d6 | 165 | */ |
631ab46b | 166 | struct tun_file { |
54f968d6 JW |
167 | struct sock sk; |
168 | struct socket socket; | |
169 | struct socket_wq wq; | |
6e914fc7 | 170 | struct tun_struct __rcu *tun; |
54f968d6 JW |
171 | struct fasync_struct *fasync; |
172 | /* only used for fasnyc */ | |
173 | unsigned int flags; | |
fb7589a1 PE |
174 | union { |
175 | u16 queue_index; | |
176 | unsigned int ifindex; | |
177 | }; | |
94317099 | 178 | struct napi_struct napi; |
aec72f33 | 179 | bool napi_enabled; |
90e33d45 | 180 | struct mutex napi_mutex; /* Protects access to the above napi */ |
4008e97f JW |
181 | struct list_head next; |
182 | struct tun_struct *detached; | |
5990a305 | 183 | struct ptr_ring tx_ring; |
8bf5c4ee | 184 | struct xdp_rxq_info xdp_rxq; |
631ab46b EB |
185 | }; |
186 | ||
96442e42 JW |
187 | struct tun_flow_entry { |
188 | struct hlist_node hash_link; | |
189 | struct rcu_head rcu; | |
190 | struct tun_struct *tun; | |
191 | ||
192 | u32 rxhash; | |
9bc88939 | 193 | u32 rps_rxhash; |
96442e42 JW |
194 | int queue_index; |
195 | unsigned long updated; | |
196 | }; | |
197 | ||
198 | #define TUN_NUM_FLOW_ENTRIES 1024 | |
199 | ||
cd5681d7 | 200 | struct tun_prog { |
96f84061 JW |
201 | struct rcu_head rcu; |
202 | struct bpf_prog *prog; | |
203 | }; | |
204 | ||
54f968d6 | 205 | /* Since the socket were moved to tun_file, to preserve the behavior of persist |
36fe8c09 | 206 | * device, socket filter, sndbuf and vnet header size were restore when the |
54f968d6 JW |
207 | * file were attached to a persist device. |
208 | */ | |
14daa021 | 209 | struct tun_struct { |
c8d68e6b JW |
210 | struct tun_file __rcu *tfiles[MAX_TAP_QUEUES]; |
211 | unsigned int numqueues; | |
f271b2cc | 212 | unsigned int flags; |
0625c883 EB |
213 | kuid_t owner; |
214 | kgid_t group; | |
14daa021 | 215 | |
14daa021 | 216 | struct net_device *dev; |
c8f44aff | 217 | netdev_features_t set_features; |
88255375 | 218 | #define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \ |
d591a1f3 | 219 | NETIF_F_TSO6) |
d9d52b51 | 220 | |
eaea34b2 | 221 | int align; |
d9d52b51 | 222 | int vnet_hdr_sz; |
54f968d6 JW |
223 | int sndbuf; |
224 | struct tap_filter txflt; | |
225 | struct sock_fprog fprog; | |
226 | /* protected by rtnl lock */ | |
227 | bool filter_attached; | |
14daa021 RR |
228 | #ifdef TUN_DEBUG |
229 | int debug; | |
1da177e4 | 230 | #endif |
96442e42 | 231 | spinlock_t lock; |
96442e42 JW |
232 | struct hlist_head flows[TUN_NUM_FLOW_ENTRIES]; |
233 | struct timer_list flow_gc_timer; | |
234 | unsigned long ageing_time; | |
4008e97f JW |
235 | unsigned int numdisabled; |
236 | struct list_head disabled; | |
5dbbaf2d | 237 | void *security; |
b8732fb7 | 238 | u32 flow_count; |
5503fcec | 239 | u32 rx_batched; |
608b9977 | 240 | struct tun_pcpu_stats __percpu *pcpu_stats; |
761876c8 | 241 | struct bpf_prog __rcu *xdp_prog; |
cd5681d7 | 242 | struct tun_prog __rcu *steering_prog; |
aff3d70a | 243 | struct tun_prog __rcu *filter_prog; |
14daa021 | 244 | }; |
1da177e4 | 245 | |
aff3d70a JW |
246 | struct veth { |
247 | __be16 h_vlan_proto; | |
248 | __be16 h_vlan_TCI; | |
14daa021 | 249 | }; |
1da177e4 | 250 | |
1ffcbc85 | 251 | bool tun_is_xdp_frame(void *ptr) |
fc72d1d5 JW |
252 | { |
253 | return (unsigned long)ptr & TUN_XDP_FLAG; | |
254 | } | |
1ffcbc85 | 255 | EXPORT_SYMBOL(tun_is_xdp_frame); |
fc72d1d5 JW |
256 | |
257 | void *tun_xdp_to_ptr(void *ptr) | |
258 | { | |
259 | return (void *)((unsigned long)ptr | TUN_XDP_FLAG); | |
260 | } | |
261 | EXPORT_SYMBOL(tun_xdp_to_ptr); | |
262 | ||
263 | void *tun_ptr_to_xdp(void *ptr) | |
264 | { | |
265 | return (void *)((unsigned long)ptr & ~TUN_XDP_FLAG); | |
266 | } | |
267 | EXPORT_SYMBOL(tun_ptr_to_xdp); | |
268 | ||
94317099 PP |
269 | static int tun_napi_receive(struct napi_struct *napi, int budget) |
270 | { | |
271 | struct tun_file *tfile = container_of(napi, struct tun_file, napi); | |
272 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
273 | struct sk_buff_head process_queue; | |
274 | struct sk_buff *skb; | |
275 | int received = 0; | |
276 | ||
277 | __skb_queue_head_init(&process_queue); | |
278 | ||
279 | spin_lock(&queue->lock); | |
280 | skb_queue_splice_tail_init(queue, &process_queue); | |
281 | spin_unlock(&queue->lock); | |
282 | ||
283 | while (received < budget && (skb = __skb_dequeue(&process_queue))) { | |
284 | napi_gro_receive(napi, skb); | |
285 | ++received; | |
286 | } | |
287 | ||
288 | if (!skb_queue_empty(&process_queue)) { | |
289 | spin_lock(&queue->lock); | |
290 | skb_queue_splice(&process_queue, queue); | |
291 | spin_unlock(&queue->lock); | |
292 | } | |
293 | ||
294 | return received; | |
295 | } | |
296 | ||
297 | static int tun_napi_poll(struct napi_struct *napi, int budget) | |
298 | { | |
299 | unsigned int received; | |
300 | ||
301 | received = tun_napi_receive(napi, budget); | |
302 | ||
303 | if (received < budget) | |
304 | napi_complete_done(napi, received); | |
305 | ||
306 | return received; | |
307 | } | |
308 | ||
309 | static void tun_napi_init(struct tun_struct *tun, struct tun_file *tfile, | |
310 | bool napi_en) | |
311 | { | |
aec72f33 | 312 | tfile->napi_enabled = napi_en; |
94317099 PP |
313 | if (napi_en) { |
314 | netif_napi_add(tun->dev, &tfile->napi, tun_napi_poll, | |
315 | NAPI_POLL_WEIGHT); | |
316 | napi_enable(&tfile->napi); | |
90e33d45 | 317 | mutex_init(&tfile->napi_mutex); |
94317099 PP |
318 | } |
319 | } | |
320 | ||
321 | static void tun_napi_disable(struct tun_struct *tun, struct tun_file *tfile) | |
322 | { | |
aec72f33 | 323 | if (tfile->napi_enabled) |
94317099 PP |
324 | napi_disable(&tfile->napi); |
325 | } | |
326 | ||
327 | static void tun_napi_del(struct tun_struct *tun, struct tun_file *tfile) | |
328 | { | |
aec72f33 | 329 | if (tfile->napi_enabled) |
94317099 PP |
330 | netif_napi_del(&tfile->napi); |
331 | } | |
332 | ||
90e33d45 PP |
333 | static bool tun_napi_frags_enabled(const struct tun_struct *tun) |
334 | { | |
335 | return READ_ONCE(tun->flags) & IFF_NAPI_FRAGS; | |
336 | } | |
337 | ||
8b8e658b GK |
338 | #ifdef CONFIG_TUN_VNET_CROSS_LE |
339 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
340 | { | |
341 | return tun->flags & TUN_VNET_BE ? false : | |
342 | virtio_legacy_is_little_endian(); | |
343 | } | |
344 | ||
345 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
346 | { | |
347 | int be = !!(tun->flags & TUN_VNET_BE); | |
348 | ||
349 | if (put_user(be, argp)) | |
350 | return -EFAULT; | |
351 | ||
352 | return 0; | |
353 | } | |
354 | ||
355 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
356 | { | |
357 | int be; | |
358 | ||
359 | if (get_user(be, argp)) | |
360 | return -EFAULT; | |
361 | ||
362 | if (be) | |
363 | tun->flags |= TUN_VNET_BE; | |
364 | else | |
365 | tun->flags &= ~TUN_VNET_BE; | |
366 | ||
367 | return 0; | |
368 | } | |
369 | #else | |
370 | static inline bool tun_legacy_is_little_endian(struct tun_struct *tun) | |
371 | { | |
372 | return virtio_legacy_is_little_endian(); | |
373 | } | |
374 | ||
375 | static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp) | |
376 | { | |
377 | return -EINVAL; | |
378 | } | |
379 | ||
380 | static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp) | |
381 | { | |
382 | return -EINVAL; | |
383 | } | |
384 | #endif /* CONFIG_TUN_VNET_CROSS_LE */ | |
385 | ||
25bd55bb GK |
386 | static inline bool tun_is_little_endian(struct tun_struct *tun) |
387 | { | |
7d824109 | 388 | return tun->flags & TUN_VNET_LE || |
8b8e658b | 389 | tun_legacy_is_little_endian(tun); |
25bd55bb GK |
390 | } |
391 | ||
56f0dcc5 MT |
392 | static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val) |
393 | { | |
25bd55bb | 394 | return __virtio16_to_cpu(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
395 | } |
396 | ||
397 | static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val) | |
398 | { | |
25bd55bb | 399 | return __cpu_to_virtio16(tun_is_little_endian(tun), val); |
56f0dcc5 MT |
400 | } |
401 | ||
96442e42 JW |
402 | static inline u32 tun_hashfn(u32 rxhash) |
403 | { | |
404 | return rxhash & 0x3ff; | |
405 | } | |
406 | ||
407 | static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash) | |
408 | { | |
409 | struct tun_flow_entry *e; | |
96442e42 | 410 | |
b67bfe0d | 411 | hlist_for_each_entry_rcu(e, head, hash_link) { |
96442e42 JW |
412 | if (e->rxhash == rxhash) |
413 | return e; | |
414 | } | |
415 | return NULL; | |
416 | } | |
417 | ||
418 | static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun, | |
419 | struct hlist_head *head, | |
420 | u32 rxhash, u16 queue_index) | |
421 | { | |
9fdc6bef ED |
422 | struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC); |
423 | ||
96442e42 JW |
424 | if (e) { |
425 | tun_debug(KERN_INFO, tun, "create flow: hash %u index %u\n", | |
426 | rxhash, queue_index); | |
427 | e->updated = jiffies; | |
428 | e->rxhash = rxhash; | |
9bc88939 | 429 | e->rps_rxhash = 0; |
96442e42 JW |
430 | e->queue_index = queue_index; |
431 | e->tun = tun; | |
432 | hlist_add_head_rcu(&e->hash_link, head); | |
b8732fb7 | 433 | ++tun->flow_count; |
96442e42 JW |
434 | } |
435 | return e; | |
436 | } | |
437 | ||
96442e42 JW |
438 | static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e) |
439 | { | |
440 | tun_debug(KERN_INFO, tun, "delete flow: hash %u index %u\n", | |
441 | e->rxhash, e->queue_index); | |
442 | hlist_del_rcu(&e->hash_link); | |
9fdc6bef | 443 | kfree_rcu(e, rcu); |
b8732fb7 | 444 | --tun->flow_count; |
96442e42 JW |
445 | } |
446 | ||
447 | static void tun_flow_flush(struct tun_struct *tun) | |
448 | { | |
449 | int i; | |
450 | ||
451 | spin_lock_bh(&tun->lock); | |
452 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
453 | struct tun_flow_entry *e; | |
b67bfe0d | 454 | struct hlist_node *n; |
96442e42 | 455 | |
b67bfe0d | 456 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) |
96442e42 JW |
457 | tun_flow_delete(tun, e); |
458 | } | |
459 | spin_unlock_bh(&tun->lock); | |
460 | } | |
461 | ||
462 | static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index) | |
463 | { | |
464 | int i; | |
465 | ||
466 | spin_lock_bh(&tun->lock); | |
467 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { | |
468 | struct tun_flow_entry *e; | |
b67bfe0d | 469 | struct hlist_node *n; |
96442e42 | 470 | |
b67bfe0d | 471 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 JW |
472 | if (e->queue_index == queue_index) |
473 | tun_flow_delete(tun, e); | |
474 | } | |
475 | } | |
476 | spin_unlock_bh(&tun->lock); | |
477 | } | |
478 | ||
e99e88a9 | 479 | static void tun_flow_cleanup(struct timer_list *t) |
96442e42 | 480 | { |
e99e88a9 | 481 | struct tun_struct *tun = from_timer(tun, t, flow_gc_timer); |
96442e42 JW |
482 | unsigned long delay = tun->ageing_time; |
483 | unsigned long next_timer = jiffies + delay; | |
484 | unsigned long count = 0; | |
485 | int i; | |
486 | ||
487 | tun_debug(KERN_INFO, tun, "tun_flow_cleanup\n"); | |
488 | ||
7dbfb4ef | 489 | spin_lock(&tun->lock); |
96442e42 JW |
490 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) { |
491 | struct tun_flow_entry *e; | |
b67bfe0d | 492 | struct hlist_node *n; |
96442e42 | 493 | |
b67bfe0d | 494 | hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) { |
96442e42 | 495 | unsigned long this_timer; |
81d98fa4 | 496 | |
96442e42 | 497 | this_timer = e->updated + delay; |
81d98fa4 | 498 | if (time_before_eq(this_timer, jiffies)) { |
96442e42 | 499 | tun_flow_delete(tun, e); |
81d98fa4 ED |
500 | continue; |
501 | } | |
502 | count++; | |
503 | if (time_before(this_timer, next_timer)) | |
96442e42 JW |
504 | next_timer = this_timer; |
505 | } | |
506 | } | |
507 | ||
508 | if (count) | |
509 | mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer)); | |
7dbfb4ef | 510 | spin_unlock(&tun->lock); |
96442e42 JW |
511 | } |
512 | ||
49974420 | 513 | static void tun_flow_update(struct tun_struct *tun, u32 rxhash, |
9e85722d | 514 | struct tun_file *tfile) |
96442e42 JW |
515 | { |
516 | struct hlist_head *head; | |
517 | struct tun_flow_entry *e; | |
518 | unsigned long delay = tun->ageing_time; | |
9e85722d | 519 | u16 queue_index = tfile->queue_index; |
96442e42 JW |
520 | |
521 | if (!rxhash) | |
522 | return; | |
523 | else | |
524 | head = &tun->flows[tun_hashfn(rxhash)]; | |
525 | ||
526 | rcu_read_lock(); | |
527 | ||
96442e42 JW |
528 | e = tun_flow_find(head, rxhash); |
529 | if (likely(e)) { | |
530 | /* TODO: keep queueing to old queue until it's empty? */ | |
531 | e->queue_index = queue_index; | |
532 | e->updated = jiffies; | |
9bc88939 | 533 | sock_rps_record_flow_hash(e->rps_rxhash); |
96442e42 JW |
534 | } else { |
535 | spin_lock_bh(&tun->lock); | |
b8732fb7 JW |
536 | if (!tun_flow_find(head, rxhash) && |
537 | tun->flow_count < MAX_TAP_FLOWS) | |
96442e42 JW |
538 | tun_flow_create(tun, head, rxhash, queue_index); |
539 | ||
540 | if (!timer_pending(&tun->flow_gc_timer)) | |
541 | mod_timer(&tun->flow_gc_timer, | |
542 | round_jiffies_up(jiffies + delay)); | |
543 | spin_unlock_bh(&tun->lock); | |
544 | } | |
545 | ||
96442e42 JW |
546 | rcu_read_unlock(); |
547 | } | |
548 | ||
9bc88939 TH |
549 | /** |
550 | * Save the hash received in the stack receive path and update the | |
551 | * flow_hash table accordingly. | |
552 | */ | |
553 | static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash) | |
554 | { | |
567e4b79 | 555 | if (unlikely(e->rps_rxhash != hash)) |
9bc88939 | 556 | e->rps_rxhash = hash; |
9bc88939 TH |
557 | } |
558 | ||
c8d68e6b | 559 | /* We try to identify a flow through its rxhash first. The reason that |
92d4ea6e | 560 | * we do not check rxq no. is because some cards(e.g 82599), chooses |
c8d68e6b JW |
561 | * the rxq based on the txq where the last packet of the flow comes. As |
562 | * the userspace application move between processors, we may get a | |
563 | * different rxq no. here. If we could not get rxhash, then we would | |
564 | * hope the rxq no. may help here. | |
565 | */ | |
96f84061 | 566 | static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
c8d68e6b | 567 | { |
96442e42 | 568 | struct tun_flow_entry *e; |
c8d68e6b JW |
569 | u32 txq = 0; |
570 | u32 numqueues = 0; | |
571 | ||
6aa7de05 | 572 | numqueues = READ_ONCE(tun->numqueues); |
c8d68e6b | 573 | |
feec084a | 574 | txq = __skb_get_hash_symmetric(skb); |
c8d68e6b | 575 | if (txq) { |
96442e42 | 576 | e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq); |
9bc88939 | 577 | if (e) { |
9bc88939 | 578 | tun_flow_save_rps_rxhash(e, txq); |
fbe4d456 | 579 | txq = e->queue_index; |
9bc88939 | 580 | } else |
96442e42 JW |
581 | /* use multiply and shift instead of expensive divide */ |
582 | txq = ((u64)txq * numqueues) >> 32; | |
c8d68e6b JW |
583 | } else if (likely(skb_rx_queue_recorded(skb))) { |
584 | txq = skb_get_rx_queue(skb); | |
585 | while (unlikely(txq >= numqueues)) | |
586 | txq -= numqueues; | |
587 | } | |
588 | ||
c8d68e6b JW |
589 | return txq; |
590 | } | |
591 | ||
96f84061 JW |
592 | static u16 tun_ebpf_select_queue(struct tun_struct *tun, struct sk_buff *skb) |
593 | { | |
cd5681d7 | 594 | struct tun_prog *prog; |
96f84061 JW |
595 | u16 ret = 0; |
596 | ||
597 | prog = rcu_dereference(tun->steering_prog); | |
598 | if (prog) | |
599 | ret = bpf_prog_run_clear_cb(prog->prog, skb); | |
600 | ||
601 | return ret % tun->numqueues; | |
602 | } | |
603 | ||
604 | static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb, | |
605 | void *accel_priv, select_queue_fallback_t fallback) | |
606 | { | |
607 | struct tun_struct *tun = netdev_priv(dev); | |
608 | u16 ret; | |
609 | ||
610 | rcu_read_lock(); | |
611 | if (rcu_dereference(tun->steering_prog)) | |
612 | ret = tun_ebpf_select_queue(tun, skb); | |
613 | else | |
614 | ret = tun_automq_select_queue(tun, skb); | |
615 | rcu_read_unlock(); | |
616 | ||
617 | return ret; | |
618 | } | |
619 | ||
cde8b15f JW |
620 | static inline bool tun_not_capable(struct tun_struct *tun) |
621 | { | |
622 | const struct cred *cred = current_cred(); | |
c260b772 | 623 | struct net *net = dev_net(tun->dev); |
cde8b15f JW |
624 | |
625 | return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) || | |
626 | (gid_valid(tun->group) && !in_egroup_p(tun->group))) && | |
c260b772 | 627 | !ns_capable(net->user_ns, CAP_NET_ADMIN); |
cde8b15f JW |
628 | } |
629 | ||
c8d68e6b JW |
630 | static void tun_set_real_num_queues(struct tun_struct *tun) |
631 | { | |
632 | netif_set_real_num_tx_queues(tun->dev, tun->numqueues); | |
633 | netif_set_real_num_rx_queues(tun->dev, tun->numqueues); | |
634 | } | |
635 | ||
4008e97f JW |
636 | static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile) |
637 | { | |
638 | tfile->detached = tun; | |
639 | list_add_tail(&tfile->next, &tun->disabled); | |
640 | ++tun->numdisabled; | |
641 | } | |
642 | ||
d32649d1 | 643 | static struct tun_struct *tun_enable_queue(struct tun_file *tfile) |
4008e97f JW |
644 | { |
645 | struct tun_struct *tun = tfile->detached; | |
646 | ||
647 | tfile->detached = NULL; | |
648 | list_del_init(&tfile->next); | |
649 | --tun->numdisabled; | |
650 | return tun; | |
651 | } | |
652 | ||
3a403076 | 653 | void tun_ptr_free(void *ptr) |
fc72d1d5 JW |
654 | { |
655 | if (!ptr) | |
656 | return; | |
1ffcbc85 JDB |
657 | if (tun_is_xdp_frame(ptr)) { |
658 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 659 | |
03993094 | 660 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
661 | } else { |
662 | __skb_array_destroy_skb(ptr); | |
663 | } | |
664 | } | |
3a403076 | 665 | EXPORT_SYMBOL_GPL(tun_ptr_free); |
fc72d1d5 | 666 | |
4bfb0513 JW |
667 | static void tun_queue_purge(struct tun_file *tfile) |
668 | { | |
fc72d1d5 | 669 | void *ptr; |
1576d986 | 670 | |
fc72d1d5 JW |
671 | while ((ptr = ptr_ring_consume(&tfile->tx_ring)) != NULL) |
672 | tun_ptr_free(ptr); | |
1576d986 | 673 | |
5503fcec | 674 | skb_queue_purge(&tfile->sk.sk_write_queue); |
4bfb0513 JW |
675 | skb_queue_purge(&tfile->sk.sk_error_queue); |
676 | } | |
677 | ||
8565d26b | 678 | static void tun_cleanup_tx_ring(struct tun_file *tfile) |
4df0bfc7 | 679 | { |
8565d26b DM |
680 | if (tfile->tx_ring.queue) { |
681 | ptr_ring_cleanup(&tfile->tx_ring, tun_ptr_free); | |
682 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
683 | memset(&tfile->tx_ring, 0, sizeof(tfile->tx_ring)); | |
4df0bfc7 CW |
684 | } |
685 | } | |
686 | ||
c8d68e6b JW |
687 | static void __tun_detach(struct tun_file *tfile, bool clean) |
688 | { | |
689 | struct tun_file *ntfile; | |
690 | struct tun_struct *tun; | |
c8d68e6b | 691 | |
b8deabd3 JW |
692 | tun = rtnl_dereference(tfile->tun); |
693 | ||
94317099 PP |
694 | if (tun && clean) { |
695 | tun_napi_disable(tun, tfile); | |
696 | tun_napi_del(tun, tfile); | |
697 | } | |
698 | ||
9e85722d | 699 | if (tun && !tfile->detached) { |
c8d68e6b JW |
700 | u16 index = tfile->queue_index; |
701 | BUG_ON(index >= tun->numqueues); | |
c8d68e6b JW |
702 | |
703 | rcu_assign_pointer(tun->tfiles[index], | |
704 | tun->tfiles[tun->numqueues - 1]); | |
b8deabd3 | 705 | ntfile = rtnl_dereference(tun->tfiles[index]); |
c8d68e6b JW |
706 | ntfile->queue_index = index; |
707 | ||
708 | --tun->numqueues; | |
9e85722d | 709 | if (clean) { |
c956674b | 710 | RCU_INIT_POINTER(tfile->tun, NULL); |
4008e97f | 711 | sock_put(&tfile->sk); |
9e85722d | 712 | } else |
4008e97f | 713 | tun_disable_queue(tun, tfile); |
c8d68e6b JW |
714 | |
715 | synchronize_net(); | |
96442e42 | 716 | tun_flow_delete_by_queue(tun, tun->numqueues + 1); |
c8d68e6b | 717 | /* Drop read queue */ |
4bfb0513 | 718 | tun_queue_purge(tfile); |
c8d68e6b | 719 | tun_set_real_num_queues(tun); |
dd38bd85 | 720 | } else if (tfile->detached && clean) { |
4008e97f | 721 | tun = tun_enable_queue(tfile); |
dd38bd85 JW |
722 | sock_put(&tfile->sk); |
723 | } | |
c8d68e6b JW |
724 | |
725 | if (clean) { | |
af668b3c MT |
726 | if (tun && tun->numqueues == 0 && tun->numdisabled == 0) { |
727 | netif_carrier_off(tun->dev); | |
728 | ||
40630b82 | 729 | if (!(tun->flags & IFF_PERSIST) && |
af668b3c | 730 | tun->dev->reg_state == NETREG_REGISTERED) |
4008e97f | 731 | unregister_netdevice(tun->dev); |
af668b3c | 732 | } |
8565d26b | 733 | tun_cleanup_tx_ring(tfile); |
140e807d | 734 | sock_put(&tfile->sk); |
c8d68e6b JW |
735 | } |
736 | } | |
737 | ||
738 | static void tun_detach(struct tun_file *tfile, bool clean) | |
739 | { | |
83c1f36f SD |
740 | struct tun_struct *tun; |
741 | struct net_device *dev; | |
742 | ||
c8d68e6b | 743 | rtnl_lock(); |
83c1f36f SD |
744 | tun = rtnl_dereference(tfile->tun); |
745 | dev = tun ? tun->dev : NULL; | |
c8d68e6b | 746 | __tun_detach(tfile, clean); |
83c1f36f SD |
747 | if (dev) |
748 | netdev_state_change(dev); | |
c8d68e6b JW |
749 | rtnl_unlock(); |
750 | } | |
751 | ||
752 | static void tun_detach_all(struct net_device *dev) | |
753 | { | |
754 | struct tun_struct *tun = netdev_priv(dev); | |
4008e97f | 755 | struct tun_file *tfile, *tmp; |
c8d68e6b JW |
756 | int i, n = tun->numqueues; |
757 | ||
758 | for (i = 0; i < n; i++) { | |
b8deabd3 | 759 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b | 760 | BUG_ON(!tfile); |
94317099 | 761 | tun_napi_disable(tun, tfile); |
addf8fc4 | 762 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 763 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 764 | RCU_INIT_POINTER(tfile->tun, NULL); |
c8d68e6b JW |
765 | --tun->numqueues; |
766 | } | |
9e85722d | 767 | list_for_each_entry(tfile, &tun->disabled, next) { |
addf8fc4 | 768 | tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN; |
9e641bdc | 769 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
c956674b | 770 | RCU_INIT_POINTER(tfile->tun, NULL); |
9e85722d | 771 | } |
c8d68e6b JW |
772 | BUG_ON(tun->numqueues != 0); |
773 | ||
774 | synchronize_net(); | |
775 | for (i = 0; i < n; i++) { | |
b8deabd3 | 776 | tfile = rtnl_dereference(tun->tfiles[i]); |
94317099 | 777 | tun_napi_del(tun, tfile); |
c8d68e6b | 778 | /* Drop read queue */ |
4bfb0513 | 779 | tun_queue_purge(tfile); |
c8d68e6b | 780 | sock_put(&tfile->sk); |
8565d26b | 781 | tun_cleanup_tx_ring(tfile); |
c8d68e6b | 782 | } |
4008e97f JW |
783 | list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) { |
784 | tun_enable_queue(tfile); | |
4bfb0513 | 785 | tun_queue_purge(tfile); |
4008e97f | 786 | sock_put(&tfile->sk); |
8565d26b | 787 | tun_cleanup_tx_ring(tfile); |
4008e97f JW |
788 | } |
789 | BUG_ON(tun->numdisabled != 0); | |
dd38bd85 | 790 | |
40630b82 | 791 | if (tun->flags & IFF_PERSIST) |
dd38bd85 | 792 | module_put(THIS_MODULE); |
c8d68e6b JW |
793 | } |
794 | ||
94317099 PP |
795 | static int tun_attach(struct tun_struct *tun, struct file *file, |
796 | bool skip_filter, bool napi) | |
a7385ba2 | 797 | { |
631ab46b | 798 | struct tun_file *tfile = file->private_data; |
1576d986 | 799 | struct net_device *dev = tun->dev; |
38231b7a | 800 | int err; |
a7385ba2 | 801 | |
5dbbaf2d PM |
802 | err = security_tun_dev_attach(tfile->socket.sk, tun->security); |
803 | if (err < 0) | |
804 | goto out; | |
805 | ||
38231b7a | 806 | err = -EINVAL; |
9e85722d | 807 | if (rtnl_dereference(tfile->tun) && !tfile->detached) |
38231b7a EB |
808 | goto out; |
809 | ||
810 | err = -EBUSY; | |
40630b82 | 811 | if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1) |
c8d68e6b JW |
812 | goto out; |
813 | ||
814 | err = -E2BIG; | |
4008e97f JW |
815 | if (!tfile->detached && |
816 | tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES) | |
38231b7a EB |
817 | goto out; |
818 | ||
819 | err = 0; | |
54f968d6 | 820 | |
92d4ea6e | 821 | /* Re-attach the filter to persist device */ |
849c9b6f | 822 | if (!skip_filter && (tun->filter_attached == true)) { |
8ced425e HFS |
823 | lock_sock(tfile->socket.sk); |
824 | err = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
825 | release_sock(tfile->socket.sk); | |
54f968d6 JW |
826 | if (!err) |
827 | goto out; | |
828 | } | |
1576d986 JW |
829 | |
830 | if (!tfile->detached && | |
5990a305 | 831 | ptr_ring_init(&tfile->tx_ring, dev->tx_queue_len, GFP_KERNEL)) { |
1576d986 JW |
832 | err = -ENOMEM; |
833 | goto out; | |
834 | } | |
835 | ||
c8d68e6b | 836 | tfile->queue_index = tun->numqueues; |
addf8fc4 | 837 | tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN; |
8bf5c4ee JDB |
838 | |
839 | if (tfile->detached) { | |
840 | /* Re-attach detached tfile, updating XDP queue_index */ | |
841 | WARN_ON(!xdp_rxq_info_is_reg(&tfile->xdp_rxq)); | |
842 | ||
843 | if (tfile->xdp_rxq.queue_index != tfile->queue_index) | |
844 | tfile->xdp_rxq.queue_index = tfile->queue_index; | |
845 | } else { | |
846 | /* Setup XDP RX-queue info, for new tfile getting attached */ | |
847 | err = xdp_rxq_info_reg(&tfile->xdp_rxq, | |
848 | tun->dev, tfile->queue_index); | |
849 | if (err < 0) | |
850 | goto out; | |
8d5d8852 JDB |
851 | err = xdp_rxq_info_reg_mem_model(&tfile->xdp_rxq, |
852 | MEM_TYPE_PAGE_SHARED, NULL); | |
853 | if (err < 0) { | |
854 | xdp_rxq_info_unreg(&tfile->xdp_rxq); | |
855 | goto out; | |
856 | } | |
8bf5c4ee JDB |
857 | err = 0; |
858 | } | |
859 | ||
6e914fc7 | 860 | rcu_assign_pointer(tfile->tun, tun); |
c8d68e6b | 861 | rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile); |
c8d68e6b | 862 | tun->numqueues++; |
a7385ba2 | 863 | |
94317099 | 864 | if (tfile->detached) { |
4008e97f | 865 | tun_enable_queue(tfile); |
94317099 | 866 | } else { |
4008e97f | 867 | sock_hold(&tfile->sk); |
94317099 PP |
868 | tun_napi_init(tun, tfile, napi); |
869 | } | |
4008e97f | 870 | |
c8d68e6b | 871 | tun_set_real_num_queues(tun); |
a7385ba2 | 872 | |
c8d68e6b JW |
873 | /* device is allowed to go away first, so no need to hold extra |
874 | * refcnt. | |
875 | */ | |
876 | ||
877 | out: | |
878 | return err; | |
631ab46b EB |
879 | } |
880 | ||
9484dc74 | 881 | static struct tun_struct *tun_get(struct tun_file *tfile) |
631ab46b | 882 | { |
6e914fc7 | 883 | struct tun_struct *tun; |
c70f1829 | 884 | |
6e914fc7 JW |
885 | rcu_read_lock(); |
886 | tun = rcu_dereference(tfile->tun); | |
887 | if (tun) | |
888 | dev_hold(tun->dev); | |
889 | rcu_read_unlock(); | |
c70f1829 EB |
890 | |
891 | return tun; | |
631ab46b EB |
892 | } |
893 | ||
631ab46b EB |
894 | static void tun_put(struct tun_struct *tun) |
895 | { | |
6e914fc7 | 896 | dev_put(tun->dev); |
631ab46b EB |
897 | } |
898 | ||
6b8a66ee | 899 | /* TAP filtering */ |
f271b2cc MK |
900 | static void addr_hash_set(u32 *mask, const u8 *addr) |
901 | { | |
902 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
903 | mask[n >> 5] |= (1 << (n & 31)); | |
904 | } | |
905 | ||
906 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | |
907 | { | |
908 | int n = ether_crc(ETH_ALEN, addr) >> 26; | |
909 | return mask[n >> 5] & (1 << (n & 31)); | |
910 | } | |
911 | ||
912 | static int update_filter(struct tap_filter *filter, void __user *arg) | |
913 | { | |
914 | struct { u8 u[ETH_ALEN]; } *addr; | |
915 | struct tun_filter uf; | |
916 | int err, alen, n, nexact; | |
917 | ||
918 | if (copy_from_user(&uf, arg, sizeof(uf))) | |
919 | return -EFAULT; | |
920 | ||
921 | if (!uf.count) { | |
922 | /* Disabled */ | |
923 | filter->count = 0; | |
924 | return 0; | |
925 | } | |
926 | ||
927 | alen = ETH_ALEN * uf.count; | |
28e8190d ME |
928 | addr = memdup_user(arg + sizeof(uf), alen); |
929 | if (IS_ERR(addr)) | |
930 | return PTR_ERR(addr); | |
f271b2cc MK |
931 | |
932 | /* The filter is updated without holding any locks. Which is | |
933 | * perfectly safe. We disable it first and in the worst | |
934 | * case we'll accept a few undesired packets. */ | |
935 | filter->count = 0; | |
936 | wmb(); | |
937 | ||
938 | /* Use first set of addresses as an exact filter */ | |
939 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | |
940 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | |
941 | ||
942 | nexact = n; | |
943 | ||
cfbf84fc AW |
944 | /* Remaining multicast addresses are hashed, |
945 | * unicast will leave the filter disabled. */ | |
f271b2cc | 946 | memset(filter->mask, 0, sizeof(filter->mask)); |
cfbf84fc AW |
947 | for (; n < uf.count; n++) { |
948 | if (!is_multicast_ether_addr(addr[n].u)) { | |
949 | err = 0; /* no filter */ | |
3b8d2a69 | 950 | goto free_addr; |
cfbf84fc | 951 | } |
f271b2cc | 952 | addr_hash_set(filter->mask, addr[n].u); |
cfbf84fc | 953 | } |
f271b2cc MK |
954 | |
955 | /* For ALLMULTI just set the mask to all ones. | |
956 | * This overrides the mask populated above. */ | |
957 | if ((uf.flags & TUN_FLT_ALLMULTI)) | |
958 | memset(filter->mask, ~0, sizeof(filter->mask)); | |
959 | ||
960 | /* Now enable the filter */ | |
961 | wmb(); | |
962 | filter->count = nexact; | |
963 | ||
964 | /* Return the number of exact filters */ | |
965 | err = nexact; | |
3b8d2a69 | 966 | free_addr: |
f271b2cc MK |
967 | kfree(addr); |
968 | return err; | |
969 | } | |
970 | ||
971 | /* Returns: 0 - drop, !=0 - accept */ | |
972 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
973 | { | |
974 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | |
975 | * at this point. */ | |
976 | struct ethhdr *eh = (struct ethhdr *) skb->data; | |
977 | int i; | |
978 | ||
979 | /* Exact match */ | |
980 | for (i = 0; i < filter->count; i++) | |
2e42e474 | 981 | if (ether_addr_equal(eh->h_dest, filter->addr[i])) |
f271b2cc MK |
982 | return 1; |
983 | ||
984 | /* Inexact match (multicast only) */ | |
985 | if (is_multicast_ether_addr(eh->h_dest)) | |
986 | return addr_hash_test(filter->mask, eh->h_dest); | |
987 | ||
988 | return 0; | |
989 | } | |
990 | ||
991 | /* | |
992 | * Checks whether the packet is accepted or not. | |
993 | * Returns: 0 - drop, !=0 - accept | |
994 | */ | |
995 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | |
996 | { | |
997 | if (!filter->count) | |
998 | return 1; | |
999 | ||
1000 | return run_filter(filter, skb); | |
1001 | } | |
1002 | ||
1da177e4 LT |
1003 | /* Network device part of the driver */ |
1004 | ||
7282d491 | 1005 | static const struct ethtool_ops tun_ethtool_ops; |
1da177e4 | 1006 | |
c70f1829 EB |
1007 | /* Net device detach from fd. */ |
1008 | static void tun_net_uninit(struct net_device *dev) | |
1009 | { | |
c8d68e6b | 1010 | tun_detach_all(dev); |
c70f1829 EB |
1011 | } |
1012 | ||
1da177e4 LT |
1013 | /* Net device open. */ |
1014 | static int tun_net_open(struct net_device *dev) | |
1015 | { | |
b20e2d54 HFS |
1016 | struct tun_struct *tun = netdev_priv(dev); |
1017 | int i; | |
1018 | ||
c8d68e6b | 1019 | netif_tx_start_all_queues(dev); |
b20e2d54 HFS |
1020 | |
1021 | for (i = 0; i < tun->numqueues; i++) { | |
1022 | struct tun_file *tfile; | |
1023 | ||
1024 | tfile = rtnl_dereference(tun->tfiles[i]); | |
1025 | tfile->socket.sk->sk_write_space(tfile->socket.sk); | |
1026 | } | |
1027 | ||
1da177e4 LT |
1028 | return 0; |
1029 | } | |
1030 | ||
1031 | /* Net device close. */ | |
1032 | static int tun_net_close(struct net_device *dev) | |
1033 | { | |
c8d68e6b | 1034 | netif_tx_stop_all_queues(dev); |
1da177e4 LT |
1035 | return 0; |
1036 | } | |
1037 | ||
1038 | /* Net device start xmit */ | |
96f84061 | 1039 | static void tun_automq_xmit(struct tun_struct *tun, struct sk_buff *skb) |
1da177e4 | 1040 | { |
3df97ba8 | 1041 | #ifdef CONFIG_RPS |
96f84061 | 1042 | if (tun->numqueues == 1 && static_key_false(&rps_needed)) { |
9bc88939 TH |
1043 | /* Select queue was not called for the skbuff, so we extract the |
1044 | * RPS hash and save it into the flow_table here. | |
1045 | */ | |
1046 | __u32 rxhash; | |
1047 | ||
feec084a | 1048 | rxhash = __skb_get_hash_symmetric(skb); |
9bc88939 TH |
1049 | if (rxhash) { |
1050 | struct tun_flow_entry *e; | |
1051 | e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], | |
1052 | rxhash); | |
1053 | if (e) | |
1054 | tun_flow_save_rps_rxhash(e, rxhash); | |
1055 | } | |
1056 | } | |
3df97ba8 | 1057 | #endif |
96f84061 JW |
1058 | } |
1059 | ||
aff3d70a JW |
1060 | static unsigned int run_ebpf_filter(struct tun_struct *tun, |
1061 | struct sk_buff *skb, | |
1062 | int len) | |
1063 | { | |
1064 | struct tun_prog *prog = rcu_dereference(tun->filter_prog); | |
1065 | ||
1066 | if (prog) | |
1067 | len = bpf_prog_run_clear_cb(prog->prog, skb); | |
1068 | ||
1069 | return len; | |
1070 | } | |
1071 | ||
96f84061 JW |
1072 | /* Net device start xmit */ |
1073 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) | |
1074 | { | |
1075 | struct tun_struct *tun = netdev_priv(dev); | |
1076 | int txq = skb->queue_mapping; | |
1077 | struct tun_file *tfile; | |
aff3d70a | 1078 | int len = skb->len; |
96f84061 JW |
1079 | |
1080 | rcu_read_lock(); | |
1081 | tfile = rcu_dereference(tun->tfiles[txq]); | |
96f84061 JW |
1082 | |
1083 | /* Drop packet if interface is not attached */ | |
cc166427 | 1084 | if (txq >= tun->numqueues) |
96f84061 JW |
1085 | goto drop; |
1086 | ||
1087 | if (!rcu_dereference(tun->steering_prog)) | |
1088 | tun_automq_xmit(tun, skb); | |
9bc88939 | 1089 | |
6e914fc7 JW |
1090 | tun_debug(KERN_INFO, tun, "tun_net_xmit %d\n", skb->len); |
1091 | ||
c8d68e6b JW |
1092 | BUG_ON(!tfile); |
1093 | ||
f271b2cc MK |
1094 | /* Drop if the filter does not like it. |
1095 | * This is a noop if the filter is disabled. | |
1096 | * Filter can be enabled only for the TAP devices. */ | |
1097 | if (!check_filter(&tun->txflt, skb)) | |
1098 | goto drop; | |
1099 | ||
54f968d6 JW |
1100 | if (tfile->socket.sk->sk_filter && |
1101 | sk_filter(tfile->socket.sk, skb)) | |
99405162 MT |
1102 | goto drop; |
1103 | ||
aff3d70a | 1104 | len = run_ebpf_filter(tun, skb, len); |
81c89507 | 1105 | if (len == 0 || pskb_trim(skb, len)) |
aff3d70a JW |
1106 | goto drop; |
1107 | ||
1f8b977a | 1108 | if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) |
7bf66305 JW |
1109 | goto drop; |
1110 | ||
7b996243 | 1111 | skb_tx_timestamp(skb); |
eda29772 | 1112 | |
0110d6f2 | 1113 | /* Orphan the skb - required as we might hang on to it |
7bf66305 JW |
1114 | * for indefinite time. |
1115 | */ | |
0110d6f2 MT |
1116 | skb_orphan(skb); |
1117 | ||
f8af75f3 ED |
1118 | nf_reset(skb); |
1119 | ||
5990a305 | 1120 | if (ptr_ring_produce(&tfile->tx_ring, skb)) |
1576d986 | 1121 | goto drop; |
1da177e4 LT |
1122 | |
1123 | /* Notify and wake up reader process */ | |
54f968d6 JW |
1124 | if (tfile->flags & TUN_FASYNC) |
1125 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
9e641bdc | 1126 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); |
6e914fc7 JW |
1127 | |
1128 | rcu_read_unlock(); | |
6ed10654 | 1129 | return NETDEV_TX_OK; |
1da177e4 LT |
1130 | |
1131 | drop: | |
608b9977 | 1132 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
149d36f7 | 1133 | skb_tx_error(skb); |
1da177e4 | 1134 | kfree_skb(skb); |
6e914fc7 | 1135 | rcu_read_unlock(); |
baeababb | 1136 | return NET_XMIT_DROP; |
1da177e4 LT |
1137 | } |
1138 | ||
f271b2cc | 1139 | static void tun_net_mclist(struct net_device *dev) |
1da177e4 | 1140 | { |
f271b2cc MK |
1141 | /* |
1142 | * This callback is supposed to deal with mc filter in | |
1143 | * _rx_ path and has nothing to do with the _tx_ path. | |
1144 | * In rx path we always accept everything userspace gives us. | |
1145 | */ | |
1da177e4 LT |
1146 | } |
1147 | ||
c8f44aff MM |
1148 | static netdev_features_t tun_net_fix_features(struct net_device *dev, |
1149 | netdev_features_t features) | |
88255375 MM |
1150 | { |
1151 | struct tun_struct *tun = netdev_priv(dev); | |
1152 | ||
1153 | return (features & tun->set_features) | (features & ~TUN_USER_FEATURES); | |
1154 | } | |
bebd097a NH |
1155 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1156 | static void tun_poll_controller(struct net_device *dev) | |
1157 | { | |
1158 | /* | |
1159 | * Tun only receives frames when: | |
1160 | * 1) the char device endpoint gets data from user space | |
1161 | * 2) the tun socket gets a sendmsg call from user space | |
94317099 PP |
1162 | * If NAPI is not enabled, since both of those are synchronous |
1163 | * operations, we are guaranteed never to have pending data when we poll | |
1164 | * for it so there is nothing to do here but return. | |
bebd097a NH |
1165 | * We need this though so netpoll recognizes us as an interface that |
1166 | * supports polling, which enables bridge devices in virt setups to | |
1167 | * still use netconsole | |
94317099 | 1168 | * If NAPI is enabled, however, we need to schedule polling for all |
90e33d45 PP |
1169 | * queues unless we are using napi_gro_frags(), which we call in |
1170 | * process context and not in NAPI context. | |
bebd097a | 1171 | */ |
94317099 PP |
1172 | struct tun_struct *tun = netdev_priv(dev); |
1173 | ||
1174 | if (tun->flags & IFF_NAPI) { | |
1175 | struct tun_file *tfile; | |
1176 | int i; | |
1177 | ||
90e33d45 PP |
1178 | if (tun_napi_frags_enabled(tun)) |
1179 | return; | |
1180 | ||
94317099 PP |
1181 | rcu_read_lock(); |
1182 | for (i = 0; i < tun->numqueues; i++) { | |
1183 | tfile = rcu_dereference(tun->tfiles[i]); | |
aec72f33 ED |
1184 | if (tfile->napi_enabled) |
1185 | napi_schedule(&tfile->napi); | |
94317099 PP |
1186 | } |
1187 | rcu_read_unlock(); | |
1188 | } | |
bebd097a NH |
1189 | return; |
1190 | } | |
1191 | #endif | |
eaea34b2 PA |
1192 | |
1193 | static void tun_set_headroom(struct net_device *dev, int new_hr) | |
1194 | { | |
1195 | struct tun_struct *tun = netdev_priv(dev); | |
1196 | ||
1197 | if (new_hr < NET_SKB_PAD) | |
1198 | new_hr = NET_SKB_PAD; | |
1199 | ||
1200 | tun->align = new_hr; | |
1201 | } | |
1202 | ||
bc1f4470 | 1203 | static void |
608b9977 PA |
1204 | tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) |
1205 | { | |
1206 | u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; | |
1207 | struct tun_struct *tun = netdev_priv(dev); | |
1208 | struct tun_pcpu_stats *p; | |
1209 | int i; | |
1210 | ||
1211 | for_each_possible_cpu(i) { | |
1212 | u64 rxpackets, rxbytes, txpackets, txbytes; | |
1213 | unsigned int start; | |
1214 | ||
1215 | p = per_cpu_ptr(tun->pcpu_stats, i); | |
1216 | do { | |
1217 | start = u64_stats_fetch_begin(&p->syncp); | |
1218 | rxpackets = p->rx_packets; | |
1219 | rxbytes = p->rx_bytes; | |
1220 | txpackets = p->tx_packets; | |
1221 | txbytes = p->tx_bytes; | |
1222 | } while (u64_stats_fetch_retry(&p->syncp, start)); | |
1223 | ||
1224 | stats->rx_packets += rxpackets; | |
1225 | stats->rx_bytes += rxbytes; | |
1226 | stats->tx_packets += txpackets; | |
1227 | stats->tx_bytes += txbytes; | |
1228 | ||
1229 | /* u32 counters */ | |
1230 | rx_dropped += p->rx_dropped; | |
1231 | rx_frame_errors += p->rx_frame_errors; | |
1232 | tx_dropped += p->tx_dropped; | |
1233 | } | |
1234 | stats->rx_dropped = rx_dropped; | |
1235 | stats->rx_frame_errors = rx_frame_errors; | |
1236 | stats->tx_dropped = tx_dropped; | |
608b9977 PA |
1237 | } |
1238 | ||
761876c8 JW |
1239 | static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog, |
1240 | struct netlink_ext_ack *extack) | |
1241 | { | |
1242 | struct tun_struct *tun = netdev_priv(dev); | |
1243 | struct bpf_prog *old_prog; | |
1244 | ||
1245 | old_prog = rtnl_dereference(tun->xdp_prog); | |
1246 | rcu_assign_pointer(tun->xdp_prog, prog); | |
1247 | if (old_prog) | |
1248 | bpf_prog_put(old_prog); | |
1249 | ||
1250 | return 0; | |
1251 | } | |
1252 | ||
1253 | static u32 tun_xdp_query(struct net_device *dev) | |
1254 | { | |
1255 | struct tun_struct *tun = netdev_priv(dev); | |
1256 | const struct bpf_prog *xdp_prog; | |
1257 | ||
1258 | xdp_prog = rtnl_dereference(tun->xdp_prog); | |
1259 | if (xdp_prog) | |
1260 | return xdp_prog->aux->id; | |
1261 | ||
1262 | return 0; | |
1263 | } | |
1264 | ||
f4e63525 | 1265 | static int tun_xdp(struct net_device *dev, struct netdev_bpf *xdp) |
761876c8 JW |
1266 | { |
1267 | switch (xdp->command) { | |
1268 | case XDP_SETUP_PROG: | |
1269 | return tun_xdp_set(dev, xdp->prog, xdp->extack); | |
1270 | case XDP_QUERY_PROG: | |
1271 | xdp->prog_id = tun_xdp_query(dev); | |
1272 | xdp->prog_attached = !!xdp->prog_id; | |
1273 | return 0; | |
1274 | default: | |
1275 | return -EINVAL; | |
1276 | } | |
1277 | } | |
1278 | ||
758e43b7 | 1279 | static const struct net_device_ops tun_netdev_ops = { |
c70f1829 | 1280 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1281 | .ndo_open = tun_net_open, |
1282 | .ndo_stop = tun_net_close, | |
00829823 | 1283 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1284 | .ndo_fix_features = tun_net_fix_features, |
c8d68e6b | 1285 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1286 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1287 | .ndo_poll_controller = tun_poll_controller, | |
1288 | #endif | |
eaea34b2 | 1289 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1290 | .ndo_get_stats64 = tun_net_get_stats64, |
758e43b7 SH |
1291 | }; |
1292 | ||
44fa2dbd | 1293 | static int tun_xdp_xmit(struct net_device *dev, struct xdp_frame *frame) |
fc72d1d5 JW |
1294 | { |
1295 | struct tun_struct *tun = netdev_priv(dev); | |
fc72d1d5 JW |
1296 | struct tun_file *tfile; |
1297 | u32 numqueues; | |
1298 | int ret = 0; | |
1299 | ||
fc72d1d5 JW |
1300 | rcu_read_lock(); |
1301 | ||
1302 | numqueues = READ_ONCE(tun->numqueues); | |
1303 | if (!numqueues) { | |
1304 | ret = -ENOSPC; | |
1305 | goto out; | |
1306 | } | |
1307 | ||
1308 | tfile = rcu_dereference(tun->tfiles[smp_processor_id() % | |
1309 | numqueues]); | |
1310 | /* Encode the XDP flag into lowest bit for consumer to differ | |
1311 | * XDP buffer from sk_buff. | |
1312 | */ | |
1ffcbc85 | 1313 | if (ptr_ring_produce(&tfile->tx_ring, tun_xdp_to_ptr(frame))) { |
fc72d1d5 JW |
1314 | this_cpu_inc(tun->pcpu_stats->tx_dropped); |
1315 | ret = -ENOSPC; | |
1316 | } | |
1317 | ||
1318 | out: | |
1319 | rcu_read_unlock(); | |
1320 | return ret; | |
1321 | } | |
1322 | ||
44fa2dbd JDB |
1323 | static int tun_xdp_tx(struct net_device *dev, struct xdp_buff *xdp) |
1324 | { | |
1325 | struct xdp_frame *frame = convert_to_xdp_frame(xdp); | |
1326 | ||
1327 | if (unlikely(!frame)) | |
1328 | return -EOVERFLOW; | |
1329 | ||
1330 | return tun_xdp_xmit(dev, frame); | |
1331 | } | |
1332 | ||
fc72d1d5 JW |
1333 | static void tun_xdp_flush(struct net_device *dev) |
1334 | { | |
1335 | struct tun_struct *tun = netdev_priv(dev); | |
1336 | struct tun_file *tfile; | |
1337 | u32 numqueues; | |
1338 | ||
1339 | rcu_read_lock(); | |
1340 | ||
1341 | numqueues = READ_ONCE(tun->numqueues); | |
1342 | if (!numqueues) | |
1343 | goto out; | |
1344 | ||
1345 | tfile = rcu_dereference(tun->tfiles[smp_processor_id() % | |
1346 | numqueues]); | |
1347 | /* Notify and wake up reader process */ | |
1348 | if (tfile->flags & TUN_FASYNC) | |
1349 | kill_fasync(&tfile->fasync, SIGIO, POLL_IN); | |
1350 | tfile->socket.sk->sk_data_ready(tfile->socket.sk); | |
1351 | ||
1352 | out: | |
1353 | rcu_read_unlock(); | |
1354 | } | |
1355 | ||
758e43b7 | 1356 | static const struct net_device_ops tap_netdev_ops = { |
c70f1829 | 1357 | .ndo_uninit = tun_net_uninit, |
758e43b7 SH |
1358 | .ndo_open = tun_net_open, |
1359 | .ndo_stop = tun_net_close, | |
00829823 | 1360 | .ndo_start_xmit = tun_net_xmit, |
88255375 | 1361 | .ndo_fix_features = tun_net_fix_features, |
afc4b13d | 1362 | .ndo_set_rx_mode = tun_net_mclist, |
758e43b7 SH |
1363 | .ndo_set_mac_address = eth_mac_addr, |
1364 | .ndo_validate_addr = eth_validate_addr, | |
c8d68e6b | 1365 | .ndo_select_queue = tun_select_queue, |
bebd097a NH |
1366 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1367 | .ndo_poll_controller = tun_poll_controller, | |
1368 | #endif | |
5e52796a | 1369 | .ndo_features_check = passthru_features_check, |
eaea34b2 | 1370 | .ndo_set_rx_headroom = tun_set_headroom, |
608b9977 | 1371 | .ndo_get_stats64 = tun_net_get_stats64, |
f4e63525 | 1372 | .ndo_bpf = tun_xdp, |
fc72d1d5 JW |
1373 | .ndo_xdp_xmit = tun_xdp_xmit, |
1374 | .ndo_xdp_flush = tun_xdp_flush, | |
758e43b7 SH |
1375 | }; |
1376 | ||
944a1376 | 1377 | static void tun_flow_init(struct tun_struct *tun) |
96442e42 JW |
1378 | { |
1379 | int i; | |
1380 | ||
96442e42 JW |
1381 | for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) |
1382 | INIT_HLIST_HEAD(&tun->flows[i]); | |
1383 | ||
1384 | tun->ageing_time = TUN_FLOW_EXPIRE; | |
e99e88a9 KC |
1385 | timer_setup(&tun->flow_gc_timer, tun_flow_cleanup, 0); |
1386 | mod_timer(&tun->flow_gc_timer, | |
1387 | round_jiffies_up(jiffies + tun->ageing_time)); | |
96442e42 JW |
1388 | } |
1389 | ||
1390 | static void tun_flow_uninit(struct tun_struct *tun) | |
1391 | { | |
1392 | del_timer_sync(&tun->flow_gc_timer); | |
1393 | tun_flow_flush(tun); | |
96442e42 JW |
1394 | } |
1395 | ||
91572088 JW |
1396 | #define MIN_MTU 68 |
1397 | #define MAX_MTU 65535 | |
1398 | ||
1da177e4 LT |
1399 | /* Initialize net device. */ |
1400 | static void tun_net_init(struct net_device *dev) | |
1401 | { | |
1402 | struct tun_struct *tun = netdev_priv(dev); | |
6aa20a22 | 1403 | |
1da177e4 | 1404 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 | 1405 | case IFF_TUN: |
758e43b7 SH |
1406 | dev->netdev_ops = &tun_netdev_ops; |
1407 | ||
1da177e4 LT |
1408 | /* Point-to-Point TUN Device */ |
1409 | dev->hard_header_len = 0; | |
1410 | dev->addr_len = 0; | |
1411 | dev->mtu = 1500; | |
1412 | ||
1413 | /* Zero header length */ | |
6aa20a22 | 1414 | dev->type = ARPHRD_NONE; |
1da177e4 | 1415 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
1da177e4 LT |
1416 | break; |
1417 | ||
40630b82 | 1418 | case IFF_TAP: |
7a0a9608 | 1419 | dev->netdev_ops = &tap_netdev_ops; |
1da177e4 | 1420 | /* Ethernet TAP Device */ |
1da177e4 | 1421 | ether_setup(dev); |
550fd08c | 1422 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
a676847b | 1423 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
36226a8d | 1424 | |
f2cedb63 | 1425 | eth_hw_addr_random(dev); |
36226a8d | 1426 | |
1da177e4 LT |
1427 | break; |
1428 | } | |
91572088 JW |
1429 | |
1430 | dev->min_mtu = MIN_MTU; | |
1431 | dev->max_mtu = MAX_MTU - dev->hard_header_len; | |
1da177e4 LT |
1432 | } |
1433 | ||
1434 | /* Character device part */ | |
1435 | ||
1436 | /* Poll */ | |
afc9a42b | 1437 | static __poll_t tun_chr_poll(struct file *file, poll_table *wait) |
6aa20a22 | 1438 | { |
b2430de3 | 1439 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1440 | struct tun_struct *tun = tun_get(tfile); |
3c8a9c63 | 1441 | struct sock *sk; |
afc9a42b | 1442 | __poll_t mask = 0; |
1da177e4 LT |
1443 | |
1444 | if (!tun) | |
a9a08845 | 1445 | return EPOLLERR; |
1da177e4 | 1446 | |
54f968d6 | 1447 | sk = tfile->socket.sk; |
3c8a9c63 | 1448 | |
6b8a66ee | 1449 | tun_debug(KERN_INFO, tun, "tun_chr_poll\n"); |
1da177e4 | 1450 | |
9e641bdc | 1451 | poll_wait(file, sk_sleep(sk), wait); |
6aa20a22 | 1452 | |
5990a305 | 1453 | if (!ptr_ring_empty(&tfile->tx_ring)) |
a9a08845 | 1454 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 1455 | |
b20e2d54 HFS |
1456 | if (tun->dev->flags & IFF_UP && |
1457 | (sock_writeable(sk) || | |
1458 | (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) && | |
1459 | sock_writeable(sk)))) | |
a9a08845 | 1460 | mask |= EPOLLOUT | EPOLLWRNORM; |
33dccbb0 | 1461 | |
c70f1829 | 1462 | if (tun->dev->reg_state != NETREG_REGISTERED) |
a9a08845 | 1463 | mask = EPOLLERR; |
c70f1829 | 1464 | |
631ab46b | 1465 | tun_put(tun); |
1da177e4 LT |
1466 | return mask; |
1467 | } | |
1468 | ||
90e33d45 PP |
1469 | static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile, |
1470 | size_t len, | |
1471 | const struct iov_iter *it) | |
1472 | { | |
1473 | struct sk_buff *skb; | |
1474 | size_t linear; | |
1475 | int err; | |
1476 | int i; | |
1477 | ||
1478 | if (it->nr_segs > MAX_SKB_FRAGS + 1) | |
1479 | return ERR_PTR(-ENOMEM); | |
1480 | ||
1481 | local_bh_disable(); | |
1482 | skb = napi_get_frags(&tfile->napi); | |
1483 | local_bh_enable(); | |
1484 | if (!skb) | |
1485 | return ERR_PTR(-ENOMEM); | |
1486 | ||
1487 | linear = iov_iter_single_seg_count(it); | |
1488 | err = __skb_grow(skb, linear); | |
1489 | if (err) | |
1490 | goto free; | |
1491 | ||
1492 | skb->len = len; | |
1493 | skb->data_len = len - linear; | |
1494 | skb->truesize += skb->data_len; | |
1495 | ||
1496 | for (i = 1; i < it->nr_segs; i++) { | |
43a08e0f | 1497 | struct page_frag *pfrag = ¤t->task_frag; |
90e33d45 | 1498 | size_t fragsz = it->iov[i].iov_len; |
90e33d45 PP |
1499 | |
1500 | if (fragsz == 0 || fragsz > PAGE_SIZE) { | |
1501 | err = -EINVAL; | |
1502 | goto free; | |
1503 | } | |
1504 | ||
43a08e0f | 1505 | if (!skb_page_frag_refill(fragsz, pfrag, GFP_KERNEL)) { |
90e33d45 PP |
1506 | err = -ENOMEM; |
1507 | goto free; | |
1508 | } | |
1509 | ||
43a08e0f ED |
1510 | skb_fill_page_desc(skb, i - 1, pfrag->page, |
1511 | pfrag->offset, fragsz); | |
1512 | page_ref_inc(pfrag->page); | |
1513 | pfrag->offset += fragsz; | |
90e33d45 PP |
1514 | } |
1515 | ||
1516 | return skb; | |
1517 | free: | |
1518 | /* frees skb and all frags allocated with napi_alloc_frag() */ | |
1519 | napi_free_frags(&tfile->napi); | |
1520 | return ERR_PTR(err); | |
1521 | } | |
1522 | ||
f42157cb RR |
1523 | /* prepad is the amount to reserve at front. len is length after that. |
1524 | * linear is a hint as to how much to copy (usually headers). */ | |
54f968d6 | 1525 | static struct sk_buff *tun_alloc_skb(struct tun_file *tfile, |
6f7c156c | 1526 | size_t prepad, size_t len, |
1527 | size_t linear, int noblock) | |
f42157cb | 1528 | { |
54f968d6 | 1529 | struct sock *sk = tfile->socket.sk; |
f42157cb | 1530 | struct sk_buff *skb; |
33dccbb0 | 1531 | int err; |
f42157cb RR |
1532 | |
1533 | /* Under a page? Don't bother with paged skb. */ | |
0eca93bc | 1534 | if (prepad + len < PAGE_SIZE || !linear) |
33dccbb0 | 1535 | linear = len; |
f42157cb | 1536 | |
33dccbb0 | 1537 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
28d64271 | 1538 | &err, 0); |
f42157cb | 1539 | if (!skb) |
33dccbb0 | 1540 | return ERR_PTR(err); |
f42157cb RR |
1541 | |
1542 | skb_reserve(skb, prepad); | |
1543 | skb_put(skb, linear); | |
33dccbb0 HX |
1544 | skb->data_len = len - linear; |
1545 | skb->len += len - linear; | |
f42157cb RR |
1546 | |
1547 | return skb; | |
1548 | } | |
1549 | ||
5503fcec JW |
1550 | static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile, |
1551 | struct sk_buff *skb, int more) | |
1552 | { | |
1553 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; | |
1554 | struct sk_buff_head process_queue; | |
1555 | u32 rx_batched = tun->rx_batched; | |
1556 | bool rcv = false; | |
1557 | ||
1558 | if (!rx_batched || (!more && skb_queue_empty(queue))) { | |
1559 | local_bh_disable(); | |
1560 | netif_receive_skb(skb); | |
1561 | local_bh_enable(); | |
1562 | return; | |
1563 | } | |
1564 | ||
1565 | spin_lock(&queue->lock); | |
1566 | if (!more || skb_queue_len(queue) == rx_batched) { | |
1567 | __skb_queue_head_init(&process_queue); | |
1568 | skb_queue_splice_tail_init(queue, &process_queue); | |
1569 | rcv = true; | |
1570 | } else { | |
1571 | __skb_queue_tail(queue, skb); | |
1572 | } | |
1573 | spin_unlock(&queue->lock); | |
1574 | ||
1575 | if (rcv) { | |
1576 | struct sk_buff *nskb; | |
1577 | ||
1578 | local_bh_disable(); | |
1579 | while ((nskb = __skb_dequeue(&process_queue))) | |
1580 | netif_receive_skb(nskb); | |
1581 | netif_receive_skb(skb); | |
1582 | local_bh_enable(); | |
1583 | } | |
1584 | } | |
1585 | ||
66ccbc9c JW |
1586 | static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile, |
1587 | int len, int noblock, bool zerocopy) | |
1588 | { | |
1589 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) | |
1590 | return false; | |
1591 | ||
1592 | if (tfile->socket.sk->sk_sndbuf != INT_MAX) | |
1593 | return false; | |
1594 | ||
1595 | if (!noblock) | |
1596 | return false; | |
1597 | ||
1598 | if (zerocopy) | |
1599 | return false; | |
1600 | ||
1601 | if (SKB_DATA_ALIGN(len + TUN_RX_PAD) + | |
1602 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE) | |
1603 | return false; | |
1604 | ||
1605 | return true; | |
1606 | } | |
1607 | ||
761876c8 JW |
1608 | static struct sk_buff *tun_build_skb(struct tun_struct *tun, |
1609 | struct tun_file *tfile, | |
66ccbc9c | 1610 | struct iov_iter *from, |
761876c8 | 1611 | struct virtio_net_hdr *hdr, |
1cfe6e93 | 1612 | int len, int *skb_xdp) |
66ccbc9c | 1613 | { |
0bbd7dad | 1614 | struct page_frag *alloc_frag = ¤t->task_frag; |
66ccbc9c | 1615 | struct sk_buff *skb; |
761876c8 | 1616 | struct bpf_prog *xdp_prog; |
7df13219 | 1617 | int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)); |
761876c8 | 1618 | unsigned int delta = 0; |
66ccbc9c JW |
1619 | char *buf; |
1620 | size_t copied; | |
7df13219 JW |
1621 | int err, pad = TUN_RX_PAD; |
1622 | ||
1623 | rcu_read_lock(); | |
1624 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1625 | if (xdp_prog) | |
1626 | pad += TUN_HEADROOM; | |
1627 | buflen += SKB_DATA_ALIGN(len + pad); | |
1628 | rcu_read_unlock(); | |
66ccbc9c | 1629 | |
63b9ab65 | 1630 | alloc_frag->offset = ALIGN((u64)alloc_frag->offset, SMP_CACHE_BYTES); |
66ccbc9c JW |
1631 | if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL))) |
1632 | return ERR_PTR(-ENOMEM); | |
1633 | ||
1634 | buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset; | |
1635 | copied = copy_page_from_iter(alloc_frag->page, | |
7df13219 | 1636 | alloc_frag->offset + pad, |
66ccbc9c JW |
1637 | len, from); |
1638 | if (copied != len) | |
1639 | return ERR_PTR(-EFAULT); | |
1640 | ||
7df13219 JW |
1641 | /* There's a small window that XDP may be set after the check |
1642 | * of xdp_prog above, this should be rare and for simplicity | |
1643 | * we do XDP on skb in case the headroom is not enough. | |
1644 | */ | |
1645 | if (hdr->gso_type || !xdp_prog) | |
1cfe6e93 | 1646 | *skb_xdp = 1; |
761876c8 | 1647 | else |
1cfe6e93 | 1648 | *skb_xdp = 0; |
761876c8 | 1649 | |
23e43f07 | 1650 | preempt_disable(); |
761876c8 JW |
1651 | rcu_read_lock(); |
1652 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1cfe6e93 | 1653 | if (xdp_prog && !*skb_xdp) { |
761876c8 JW |
1654 | struct xdp_buff xdp; |
1655 | void *orig_data; | |
1656 | u32 act; | |
1657 | ||
1658 | xdp.data_hard_start = buf; | |
7df13219 | 1659 | xdp.data = buf + pad; |
de8f3a83 | 1660 | xdp_set_data_meta_invalid(&xdp); |
761876c8 | 1661 | xdp.data_end = xdp.data + len; |
8bf5c4ee | 1662 | xdp.rxq = &tfile->xdp_rxq; |
761876c8 JW |
1663 | orig_data = xdp.data; |
1664 | act = bpf_prog_run_xdp(xdp_prog, &xdp); | |
1665 | ||
1666 | switch (act) { | |
1667 | case XDP_REDIRECT: | |
1668 | get_page(alloc_frag->page); | |
1669 | alloc_frag->offset += buflen; | |
1670 | err = xdp_do_redirect(tun->dev, &xdp, xdp_prog); | |
1bb4f2e8 | 1671 | xdp_do_flush_map(); |
761876c8 JW |
1672 | if (err) |
1673 | goto err_redirect; | |
654d5738 | 1674 | rcu_read_unlock(); |
23e43f07 | 1675 | preempt_enable(); |
761876c8 JW |
1676 | return NULL; |
1677 | case XDP_TX: | |
59655a5b JW |
1678 | get_page(alloc_frag->page); |
1679 | alloc_frag->offset += buflen; | |
44fa2dbd | 1680 | if (tun_xdp_tx(tun->dev, &xdp)) |
59655a5b JW |
1681 | goto err_redirect; |
1682 | tun_xdp_flush(tun->dev); | |
1683 | rcu_read_unlock(); | |
1684 | preempt_enable(); | |
1685 | return NULL; | |
761876c8 JW |
1686 | case XDP_PASS: |
1687 | delta = orig_data - xdp.data; | |
8fb58f1e | 1688 | len = xdp.data_end - xdp.data; |
761876c8 JW |
1689 | break; |
1690 | default: | |
1691 | bpf_warn_invalid_xdp_action(act); | |
1692 | /* fall through */ | |
1693 | case XDP_ABORTED: | |
1694 | trace_xdp_exception(tun->dev, xdp_prog, act); | |
1695 | /* fall through */ | |
1696 | case XDP_DROP: | |
1697 | goto err_xdp; | |
1698 | } | |
1699 | } | |
1700 | ||
66ccbc9c | 1701 | skb = build_skb(buf, buflen); |
761876c8 JW |
1702 | if (!skb) { |
1703 | rcu_read_unlock(); | |
23e43f07 | 1704 | preempt_enable(); |
66ccbc9c | 1705 | return ERR_PTR(-ENOMEM); |
761876c8 | 1706 | } |
66ccbc9c | 1707 | |
7df13219 | 1708 | skb_reserve(skb, pad - delta); |
8fb58f1e | 1709 | skb_put(skb, len); |
66ccbc9c JW |
1710 | get_page(alloc_frag->page); |
1711 | alloc_frag->offset += buflen; | |
1712 | ||
761876c8 | 1713 | rcu_read_unlock(); |
23e43f07 | 1714 | preempt_enable(); |
761876c8 | 1715 | |
66ccbc9c | 1716 | return skb; |
761876c8 JW |
1717 | |
1718 | err_redirect: | |
1719 | put_page(alloc_frag->page); | |
1720 | err_xdp: | |
1721 | rcu_read_unlock(); | |
23e43f07 | 1722 | preempt_enable(); |
761876c8 JW |
1723 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
1724 | return NULL; | |
66ccbc9c JW |
1725 | } |
1726 | ||
1da177e4 | 1727 | /* Get packet from user space buffer */ |
54f968d6 | 1728 | static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, |
f5ff53b4 | 1729 | void *msg_control, struct iov_iter *from, |
5503fcec | 1730 | int noblock, bool more) |
1da177e4 | 1731 | { |
09640e63 | 1732 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
1da177e4 | 1733 | struct sk_buff *skb; |
f5ff53b4 | 1734 | size_t total_len = iov_iter_count(from); |
eaea34b2 | 1735 | size_t len = total_len, align = tun->align, linear; |
f43798c2 | 1736 | struct virtio_net_hdr gso = { 0 }; |
608b9977 | 1737 | struct tun_pcpu_stats *stats; |
96f8d9ec | 1738 | int good_linear; |
0690899b MT |
1739 | int copylen; |
1740 | bool zerocopy = false; | |
1741 | int err; | |
96f84061 | 1742 | u32 rxhash = 0; |
1cfe6e93 | 1743 | int skb_xdp = 1; |
90e33d45 | 1744 | bool frags = tun_napi_frags_enabled(tun); |
1da177e4 | 1745 | |
1bd4978a ED |
1746 | if (!(tun->dev->flags & IFF_UP)) |
1747 | return -EIO; | |
1748 | ||
40630b82 | 1749 | if (!(tun->flags & IFF_NO_PI)) { |
15718ea0 | 1750 | if (len < sizeof(pi)) |
1da177e4 | 1751 | return -EINVAL; |
15718ea0 | 1752 | len -= sizeof(pi); |
1da177e4 | 1753 | |
cbbd26b8 | 1754 | if (!copy_from_iter_full(&pi, sizeof(pi), from)) |
1da177e4 LT |
1755 | return -EFAULT; |
1756 | } | |
1757 | ||
40630b82 | 1758 | if (tun->flags & IFF_VNET_HDR) { |
e1edab87 WB |
1759 | int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1760 | ||
1761 | if (len < vnet_hdr_sz) | |
f43798c2 | 1762 | return -EINVAL; |
e1edab87 | 1763 | len -= vnet_hdr_sz; |
f43798c2 | 1764 | |
cbbd26b8 | 1765 | if (!copy_from_iter_full(&gso, sizeof(gso), from)) |
f43798c2 RR |
1766 | return -EFAULT; |
1767 | ||
4909122f | 1768 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
56f0dcc5 MT |
1769 | tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len)) |
1770 | gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2); | |
4909122f | 1771 | |
56f0dcc5 | 1772 | if (tun16_to_cpu(tun, gso.hdr_len) > len) |
f43798c2 | 1773 | return -EINVAL; |
e1edab87 | 1774 | iov_iter_advance(from, vnet_hdr_sz - sizeof(gso)); |
f43798c2 RR |
1775 | } |
1776 | ||
40630b82 | 1777 | if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) { |
a504b86e | 1778 | align += NET_IP_ALIGN; |
0eca93bc | 1779 | if (unlikely(len < ETH_HLEN || |
56f0dcc5 | 1780 | (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN))) |
e01bf1c8 RR |
1781 | return -EINVAL; |
1782 | } | |
6aa20a22 | 1783 | |
96f8d9ec JW |
1784 | good_linear = SKB_MAX_HEAD(align); |
1785 | ||
88529176 | 1786 | if (msg_control) { |
f5ff53b4 AV |
1787 | struct iov_iter i = *from; |
1788 | ||
88529176 JW |
1789 | /* There are 256 bytes to be copied in skb, so there is |
1790 | * enough room for skb expand head in case it is used. | |
0690899b MT |
1791 | * The rest of the buffer is mapped from userspace. |
1792 | */ | |
56f0dcc5 | 1793 | copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN; |
96f8d9ec JW |
1794 | if (copylen > good_linear) |
1795 | copylen = good_linear; | |
3dd5c330 | 1796 | linear = copylen; |
f5ff53b4 AV |
1797 | iov_iter_advance(&i, copylen); |
1798 | if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS) | |
88529176 JW |
1799 | zerocopy = true; |
1800 | } | |
1801 | ||
90e33d45 | 1802 | if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) { |
1cfe6e93 JW |
1803 | /* For the packet that is not easy to be processed |
1804 | * (e.g gso or jumbo packet), we will do it at after | |
1805 | * skb was created with generic XDP routine. | |
1806 | */ | |
1807 | skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp); | |
66ccbc9c | 1808 | if (IS_ERR(skb)) { |
608b9977 | 1809 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
66ccbc9c JW |
1810 | return PTR_ERR(skb); |
1811 | } | |
761876c8 JW |
1812 | if (!skb) |
1813 | return total_len; | |
66ccbc9c JW |
1814 | } else { |
1815 | if (!zerocopy) { | |
1816 | copylen = len; | |
1817 | if (tun16_to_cpu(tun, gso.hdr_len) > good_linear) | |
1818 | linear = good_linear; | |
1819 | else | |
1820 | linear = tun16_to_cpu(tun, gso.hdr_len); | |
1821 | } | |
1da177e4 | 1822 | |
90e33d45 PP |
1823 | if (frags) { |
1824 | mutex_lock(&tfile->napi_mutex); | |
1825 | skb = tun_napi_alloc_frags(tfile, copylen, from); | |
1826 | /* tun_napi_alloc_frags() enforces a layout for the skb. | |
1827 | * If zerocopy is enabled, then this layout will be | |
1828 | * overwritten by zerocopy_sg_from_iter(). | |
1829 | */ | |
1830 | zerocopy = false; | |
1831 | } else { | |
1832 | skb = tun_alloc_skb(tfile, align, copylen, linear, | |
1833 | noblock); | |
1834 | } | |
1835 | ||
66ccbc9c JW |
1836 | if (IS_ERR(skb)) { |
1837 | if (PTR_ERR(skb) != -EAGAIN) | |
1838 | this_cpu_inc(tun->pcpu_stats->rx_dropped); | |
90e33d45 PP |
1839 | if (frags) |
1840 | mutex_unlock(&tfile->napi_mutex); | |
66ccbc9c JW |
1841 | return PTR_ERR(skb); |
1842 | } | |
0690899b | 1843 | |
66ccbc9c JW |
1844 | if (zerocopy) |
1845 | err = zerocopy_sg_from_iter(skb, from); | |
1846 | else | |
1847 | err = skb_copy_datagram_from_iter(skb, 0, from, len); | |
1848 | ||
1849 | if (err) { | |
1850 | this_cpu_inc(tun->pcpu_stats->rx_dropped); | |
1851 | kfree_skb(skb); | |
90e33d45 PP |
1852 | if (frags) { |
1853 | tfile->napi.skb = NULL; | |
1854 | mutex_unlock(&tfile->napi_mutex); | |
1855 | } | |
1856 | ||
66ccbc9c JW |
1857 | return -EFAULT; |
1858 | } | |
8f22757e | 1859 | } |
1da177e4 | 1860 | |
3e9e40e7 | 1861 | if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) { |
df10db98 PA |
1862 | this_cpu_inc(tun->pcpu_stats->rx_frame_errors); |
1863 | kfree_skb(skb); | |
90e33d45 PP |
1864 | if (frags) { |
1865 | tfile->napi.skb = NULL; | |
1866 | mutex_unlock(&tfile->napi_mutex); | |
1867 | } | |
1868 | ||
df10db98 PA |
1869 | return -EINVAL; |
1870 | } | |
1871 | ||
1da177e4 | 1872 | switch (tun->flags & TUN_TYPE_MASK) { |
40630b82 MT |
1873 | case IFF_TUN: |
1874 | if (tun->flags & IFF_NO_PI) { | |
2580c4c1 AP |
1875 | u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0; |
1876 | ||
1877 | switch (ip_version) { | |
1878 | case 4: | |
f09f7ee2 AWC |
1879 | pi.proto = htons(ETH_P_IP); |
1880 | break; | |
2580c4c1 | 1881 | case 6: |
f09f7ee2 AWC |
1882 | pi.proto = htons(ETH_P_IPV6); |
1883 | break; | |
1884 | default: | |
608b9977 | 1885 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
f09f7ee2 AWC |
1886 | kfree_skb(skb); |
1887 | return -EINVAL; | |
1888 | } | |
1889 | } | |
1890 | ||
459a98ed | 1891 | skb_reset_mac_header(skb); |
1da177e4 | 1892 | skb->protocol = pi.proto; |
4c13eb66 | 1893 | skb->dev = tun->dev; |
1da177e4 | 1894 | break; |
40630b82 | 1895 | case IFF_TAP: |
90e33d45 PP |
1896 | if (!frags) |
1897 | skb->protocol = eth_type_trans(skb, tun->dev); | |
1da177e4 | 1898 | break; |
6403eab1 | 1899 | } |
1da177e4 | 1900 | |
0690899b MT |
1901 | /* copy skb_ubuf_info for callback when skb has no error */ |
1902 | if (zerocopy) { | |
1903 | skb_shinfo(skb)->destructor_arg = msg_control; | |
1904 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
c9af6db4 | 1905 | skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; |
af1cc7a2 JW |
1906 | } else if (msg_control) { |
1907 | struct ubuf_info *uarg = msg_control; | |
1908 | uarg->callback(uarg, false); | |
0690899b MT |
1909 | } |
1910 | ||
72f65107 | 1911 | skb_reset_network_header(skb); |
40893fd0 | 1912 | skb_probe_transport_header(skb, 0); |
38502af7 | 1913 | |
1cfe6e93 | 1914 | if (skb_xdp) { |
761876c8 JW |
1915 | struct bpf_prog *xdp_prog; |
1916 | int ret; | |
1917 | ||
1918 | rcu_read_lock(); | |
1919 | xdp_prog = rcu_dereference(tun->xdp_prog); | |
1920 | if (xdp_prog) { | |
1921 | ret = do_xdp_generic(xdp_prog, skb); | |
1922 | if (ret != XDP_PASS) { | |
1923 | rcu_read_unlock(); | |
1924 | return total_len; | |
1925 | } | |
1926 | } | |
1927 | rcu_read_unlock(); | |
1928 | } | |
1929 | ||
cf1a1e07 PA |
1930 | /* Compute the costly rx hash only if needed for flow updates. |
1931 | * We may get a very small possibility of OOO during switching, not | |
1932 | * worth to optimize. | |
1933 | */ | |
1934 | if (!rcu_access_pointer(tun->steering_prog) && tun->numqueues > 1 && | |
1935 | !tfile->detached) | |
96f84061 | 1936 | rxhash = __skb_get_hash_symmetric(skb); |
94317099 | 1937 | |
90e33d45 PP |
1938 | if (frags) { |
1939 | /* Exercise flow dissector code path. */ | |
1940 | u32 headlen = eth_get_headlen(skb->data, skb_headlen(skb)); | |
1941 | ||
010f245b | 1942 | if (unlikely(headlen > skb_headlen(skb))) { |
90e33d45 PP |
1943 | this_cpu_inc(tun->pcpu_stats->rx_dropped); |
1944 | napi_free_frags(&tfile->napi); | |
1945 | mutex_unlock(&tfile->napi_mutex); | |
1946 | WARN_ON(1); | |
1947 | return -ENOMEM; | |
1948 | } | |
1949 | ||
1950 | local_bh_disable(); | |
1951 | napi_gro_frags(&tfile->napi); | |
1952 | local_bh_enable(); | |
1953 | mutex_unlock(&tfile->napi_mutex); | |
aec72f33 | 1954 | } else if (tfile->napi_enabled) { |
94317099 PP |
1955 | struct sk_buff_head *queue = &tfile->sk.sk_write_queue; |
1956 | int queue_len; | |
1957 | ||
1958 | spin_lock_bh(&queue->lock); | |
1959 | __skb_queue_tail(queue, skb); | |
1960 | queue_len = skb_queue_len(queue); | |
1961 | spin_unlock(&queue->lock); | |
1962 | ||
1963 | if (!more || queue_len > NAPI_POLL_WEIGHT) | |
1964 | napi_schedule(&tfile->napi); | |
1965 | ||
1966 | local_bh_enable(); | |
1967 | } else if (!IS_ENABLED(CONFIG_4KSTACKS)) { | |
1968 | tun_rx_batched(tun, tfile, skb, more); | |
1969 | } else { | |
1970 | netif_rx_ni(skb); | |
1971 | } | |
6aa20a22 | 1972 | |
608b9977 PA |
1973 | stats = get_cpu_ptr(tun->pcpu_stats); |
1974 | u64_stats_update_begin(&stats->syncp); | |
1975 | stats->rx_packets++; | |
1976 | stats->rx_bytes += len; | |
1977 | u64_stats_update_end(&stats->syncp); | |
1978 | put_cpu_ptr(stats); | |
1da177e4 | 1979 | |
96f84061 JW |
1980 | if (rxhash) |
1981 | tun_flow_update(tun, rxhash, tfile); | |
1982 | ||
0690899b | 1983 | return total_len; |
6aa20a22 | 1984 | } |
1da177e4 | 1985 | |
f5ff53b4 | 1986 | static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 1987 | { |
33dccbb0 | 1988 | struct file *file = iocb->ki_filp; |
54f968d6 | 1989 | struct tun_file *tfile = file->private_data; |
9484dc74 | 1990 | struct tun_struct *tun = tun_get(tfile); |
631ab46b | 1991 | ssize_t result; |
1da177e4 LT |
1992 | |
1993 | if (!tun) | |
1994 | return -EBADFD; | |
1995 | ||
5503fcec JW |
1996 | result = tun_get_user(tun, tfile, NULL, from, |
1997 | file->f_flags & O_NONBLOCK, false); | |
631ab46b EB |
1998 | |
1999 | tun_put(tun); | |
2000 | return result; | |
1da177e4 LT |
2001 | } |
2002 | ||
fc72d1d5 JW |
2003 | static ssize_t tun_put_user_xdp(struct tun_struct *tun, |
2004 | struct tun_file *tfile, | |
1ffcbc85 | 2005 | struct xdp_frame *xdp_frame, |
fc72d1d5 JW |
2006 | struct iov_iter *iter) |
2007 | { | |
2008 | int vnet_hdr_sz = 0; | |
1ffcbc85 | 2009 | size_t size = xdp_frame->len; |
fc72d1d5 JW |
2010 | struct tun_pcpu_stats *stats; |
2011 | size_t ret; | |
2012 | ||
2013 | if (tun->flags & IFF_VNET_HDR) { | |
2014 | struct virtio_net_hdr gso = { 0 }; | |
2015 | ||
2016 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); | |
2017 | if (unlikely(iov_iter_count(iter) < vnet_hdr_sz)) | |
2018 | return -EINVAL; | |
2019 | if (unlikely(copy_to_iter(&gso, sizeof(gso), iter) != | |
2020 | sizeof(gso))) | |
2021 | return -EFAULT; | |
2022 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
2023 | } | |
2024 | ||
1ffcbc85 | 2025 | ret = copy_to_iter(xdp_frame->data, size, iter) + vnet_hdr_sz; |
fc72d1d5 JW |
2026 | |
2027 | stats = get_cpu_ptr(tun->pcpu_stats); | |
2028 | u64_stats_update_begin(&stats->syncp); | |
2029 | stats->tx_packets++; | |
2030 | stats->tx_bytes += ret; | |
2031 | u64_stats_update_end(&stats->syncp); | |
2032 | put_cpu_ptr(tun->pcpu_stats); | |
2033 | ||
2034 | return ret; | |
2035 | } | |
2036 | ||
1da177e4 | 2037 | /* Put packet to the user space buffer */ |
6f7c156c | 2038 | static ssize_t tun_put_user(struct tun_struct *tun, |
54f968d6 | 2039 | struct tun_file *tfile, |
6f7c156c | 2040 | struct sk_buff *skb, |
e0b46d0e | 2041 | struct iov_iter *iter) |
1da177e4 LT |
2042 | { |
2043 | struct tun_pi pi = { 0, skb->protocol }; | |
608b9977 | 2044 | struct tun_pcpu_stats *stats; |
e0b46d0e | 2045 | ssize_t total; |
8c847d25 | 2046 | int vlan_offset = 0; |
a8f9bfdf | 2047 | int vlan_hlen = 0; |
2eb783c4 | 2048 | int vnet_hdr_sz = 0; |
a8f9bfdf | 2049 | |
df8a39de | 2050 | if (skb_vlan_tag_present(skb)) |
a8f9bfdf | 2051 | vlan_hlen = VLAN_HLEN; |
1da177e4 | 2052 | |
40630b82 | 2053 | if (tun->flags & IFF_VNET_HDR) |
e1edab87 | 2054 | vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz); |
1da177e4 | 2055 | |
e0b46d0e HX |
2056 | total = skb->len + vlan_hlen + vnet_hdr_sz; |
2057 | ||
40630b82 | 2058 | if (!(tun->flags & IFF_NO_PI)) { |
e0b46d0e | 2059 | if (iov_iter_count(iter) < sizeof(pi)) |
1da177e4 LT |
2060 | return -EINVAL; |
2061 | ||
e0b46d0e HX |
2062 | total += sizeof(pi); |
2063 | if (iov_iter_count(iter) < total) { | |
1da177e4 LT |
2064 | /* Packet will be striped */ |
2065 | pi.flags |= TUN_PKT_STRIP; | |
2066 | } | |
6aa20a22 | 2067 | |
e0b46d0e | 2068 | if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi)) |
1da177e4 | 2069 | return -EFAULT; |
6aa20a22 | 2070 | } |
1da177e4 | 2071 | |
2eb783c4 | 2072 | if (vnet_hdr_sz) { |
9403cd7c | 2073 | struct virtio_net_hdr gso; |
34166093 | 2074 | |
e0b46d0e | 2075 | if (iov_iter_count(iter) < vnet_hdr_sz) |
f43798c2 RR |
2076 | return -EINVAL; |
2077 | ||
3e9e40e7 | 2078 | if (virtio_net_hdr_from_skb(skb, &gso, |
6391a448 | 2079 | tun_is_little_endian(tun), true)) { |
f43798c2 | 2080 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
34166093 MR |
2081 | pr_err("unexpected GSO type: " |
2082 | "0x%x, gso_size %d, hdr_len %d\n", | |
2083 | sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size), | |
2084 | tun16_to_cpu(tun, gso.hdr_len)); | |
2085 | print_hex_dump(KERN_ERR, "tun: ", | |
2086 | DUMP_PREFIX_NONE, | |
2087 | 16, 1, skb->head, | |
2088 | min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true); | |
2089 | WARN_ON_ONCE(1); | |
2090 | return -EINVAL; | |
2091 | } | |
f43798c2 | 2092 | |
e0b46d0e | 2093 | if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso)) |
f43798c2 | 2094 | return -EFAULT; |
8c847d25 JW |
2095 | |
2096 | iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso)); | |
f43798c2 RR |
2097 | } |
2098 | ||
a8f9bfdf | 2099 | if (vlan_hlen) { |
e0b46d0e | 2100 | int ret; |
aff3d70a | 2101 | struct veth veth; |
6680ec68 JW |
2102 | |
2103 | veth.h_vlan_proto = skb->vlan_proto; | |
df8a39de | 2104 | veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb)); |
6680ec68 JW |
2105 | |
2106 | vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto); | |
6680ec68 | 2107 | |
e0b46d0e HX |
2108 | ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset); |
2109 | if (ret || !iov_iter_count(iter)) | |
6680ec68 JW |
2110 | goto done; |
2111 | ||
e0b46d0e HX |
2112 | ret = copy_to_iter(&veth, sizeof(veth), iter); |
2113 | if (ret != sizeof(veth) || !iov_iter_count(iter)) | |
6680ec68 JW |
2114 | goto done; |
2115 | } | |
1da177e4 | 2116 | |
e0b46d0e | 2117 | skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); |
1da177e4 | 2118 | |
6680ec68 | 2119 | done: |
608b9977 PA |
2120 | /* caller is in process context, */ |
2121 | stats = get_cpu_ptr(tun->pcpu_stats); | |
2122 | u64_stats_update_begin(&stats->syncp); | |
2123 | stats->tx_packets++; | |
2124 | stats->tx_bytes += skb->len + vlan_hlen; | |
2125 | u64_stats_update_end(&stats->syncp); | |
2126 | put_cpu_ptr(tun->pcpu_stats); | |
1da177e4 LT |
2127 | |
2128 | return total; | |
2129 | } | |
2130 | ||
fc72d1d5 | 2131 | static void *tun_ring_recv(struct tun_file *tfile, int noblock, int *err) |
1576d986 JW |
2132 | { |
2133 | DECLARE_WAITQUEUE(wait, current); | |
fc72d1d5 | 2134 | void *ptr = NULL; |
f48cc6b2 | 2135 | int error = 0; |
1576d986 | 2136 | |
fc72d1d5 JW |
2137 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2138 | if (ptr) | |
1576d986 JW |
2139 | goto out; |
2140 | if (noblock) { | |
f48cc6b2 | 2141 | error = -EAGAIN; |
1576d986 JW |
2142 | goto out; |
2143 | } | |
2144 | ||
2145 | add_wait_queue(&tfile->wq.wait, &wait); | |
2146 | current->state = TASK_INTERRUPTIBLE; | |
2147 | ||
2148 | while (1) { | |
fc72d1d5 JW |
2149 | ptr = ptr_ring_consume(&tfile->tx_ring); |
2150 | if (ptr) | |
1576d986 JW |
2151 | break; |
2152 | if (signal_pending(current)) { | |
f48cc6b2 | 2153 | error = -ERESTARTSYS; |
1576d986 JW |
2154 | break; |
2155 | } | |
2156 | if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) { | |
f48cc6b2 | 2157 | error = -EFAULT; |
1576d986 JW |
2158 | break; |
2159 | } | |
2160 | ||
2161 | schedule(); | |
2162 | } | |
2163 | ||
2164 | current->state = TASK_RUNNING; | |
2165 | remove_wait_queue(&tfile->wq.wait, &wait); | |
2166 | ||
2167 | out: | |
f48cc6b2 | 2168 | *err = error; |
fc72d1d5 | 2169 | return ptr; |
1576d986 JW |
2170 | } |
2171 | ||
54f968d6 | 2172 | static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile, |
9b067034 | 2173 | struct iov_iter *to, |
fc72d1d5 | 2174 | int noblock, void *ptr) |
1da177e4 | 2175 | { |
9b067034 | 2176 | ssize_t ret; |
1576d986 | 2177 | int err; |
1da177e4 | 2178 | |
3872baf6 | 2179 | tun_debug(KERN_INFO, tun, "tun_do_read\n"); |
1da177e4 | 2180 | |
c33ee15b | 2181 | if (!iov_iter_count(to)) { |
fc72d1d5 | 2182 | tun_ptr_free(ptr); |
9b067034 | 2183 | return 0; |
c33ee15b | 2184 | } |
1da177e4 | 2185 | |
fc72d1d5 | 2186 | if (!ptr) { |
ac77cfd4 | 2187 | /* Read frames from ring */ |
fc72d1d5 JW |
2188 | ptr = tun_ring_recv(tfile, noblock, &err); |
2189 | if (!ptr) | |
ac77cfd4 JW |
2190 | return err; |
2191 | } | |
e0b46d0e | 2192 | |
1ffcbc85 JDB |
2193 | if (tun_is_xdp_frame(ptr)) { |
2194 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2195 | |
1ffcbc85 | 2196 | ret = tun_put_user_xdp(tun, tfile, xdpf, to); |
03993094 | 2197 | xdp_return_frame(xdpf); |
fc72d1d5 JW |
2198 | } else { |
2199 | struct sk_buff *skb = ptr; | |
2200 | ||
2201 | ret = tun_put_user(tun, tfile, skb, to); | |
2202 | if (unlikely(ret < 0)) | |
2203 | kfree_skb(skb); | |
2204 | else | |
2205 | consume_skb(skb); | |
2206 | } | |
1da177e4 | 2207 | |
05c2828c MT |
2208 | return ret; |
2209 | } | |
2210 | ||
9b067034 | 2211 | static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to) |
05c2828c MT |
2212 | { |
2213 | struct file *file = iocb->ki_filp; | |
2214 | struct tun_file *tfile = file->private_data; | |
9484dc74 | 2215 | struct tun_struct *tun = tun_get(tfile); |
9b067034 | 2216 | ssize_t len = iov_iter_count(to), ret; |
05c2828c MT |
2217 | |
2218 | if (!tun) | |
2219 | return -EBADFD; | |
ac77cfd4 | 2220 | ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK, NULL); |
42404c09 | 2221 | ret = min_t(ssize_t, ret, len); |
d0b7da8a ZYW |
2222 | if (ret > 0) |
2223 | iocb->ki_pos = ret; | |
631ab46b | 2224 | tun_put(tun); |
1da177e4 LT |
2225 | return ret; |
2226 | } | |
2227 | ||
cd5681d7 | 2228 | static void tun_prog_free(struct rcu_head *rcu) |
96f84061 | 2229 | { |
cd5681d7 | 2230 | struct tun_prog *prog = container_of(rcu, struct tun_prog, rcu); |
96f84061 JW |
2231 | |
2232 | bpf_prog_destroy(prog->prog); | |
2233 | kfree(prog); | |
2234 | } | |
2235 | ||
9d6474e4 JW |
2236 | static int __tun_set_ebpf(struct tun_struct *tun, |
2237 | struct tun_prog __rcu **prog_p, | |
cd5681d7 | 2238 | struct bpf_prog *prog) |
96f84061 | 2239 | { |
cd5681d7 | 2240 | struct tun_prog *old, *new = NULL; |
96f84061 JW |
2241 | |
2242 | if (prog) { | |
2243 | new = kmalloc(sizeof(*new), GFP_KERNEL); | |
2244 | if (!new) | |
2245 | return -ENOMEM; | |
2246 | new->prog = prog; | |
2247 | } | |
2248 | ||
124da8f6 | 2249 | spin_lock_bh(&tun->lock); |
cd5681d7 | 2250 | old = rcu_dereference_protected(*prog_p, |
124da8f6 | 2251 | lockdep_is_held(&tun->lock)); |
cd5681d7 | 2252 | rcu_assign_pointer(*prog_p, new); |
124da8f6 | 2253 | spin_unlock_bh(&tun->lock); |
96f84061 JW |
2254 | |
2255 | if (old) | |
cd5681d7 | 2256 | call_rcu(&old->rcu, tun_prog_free); |
96f84061 JW |
2257 | |
2258 | return 0; | |
2259 | } | |
2260 | ||
96442e42 JW |
2261 | static void tun_free_netdev(struct net_device *dev) |
2262 | { | |
2263 | struct tun_struct *tun = netdev_priv(dev); | |
2264 | ||
4008e97f | 2265 | BUG_ON(!(list_empty(&tun->disabled))); |
608b9977 | 2266 | free_percpu(tun->pcpu_stats); |
96442e42 | 2267 | tun_flow_uninit(tun); |
5dbbaf2d | 2268 | security_tun_dev_free_security(tun->security); |
cd5681d7 | 2269 | __tun_set_ebpf(tun, &tun->steering_prog, NULL); |
aff3d70a | 2270 | __tun_set_ebpf(tun, &tun->filter_prog, NULL); |
96442e42 JW |
2271 | } |
2272 | ||
1da177e4 LT |
2273 | static void tun_setup(struct net_device *dev) |
2274 | { | |
2275 | struct tun_struct *tun = netdev_priv(dev); | |
2276 | ||
0625c883 EB |
2277 | tun->owner = INVALID_UID; |
2278 | tun->group = INVALID_GID; | |
1da177e4 | 2279 | |
1da177e4 | 2280 | dev->ethtool_ops = &tun_ethtool_ops; |
cf124db5 DM |
2281 | dev->needs_free_netdev = true; |
2282 | dev->priv_destructor = tun_free_netdev; | |
016adb72 JW |
2283 | /* We prefer our own queue length */ |
2284 | dev->tx_queue_len = TUN_READQ_SIZE; | |
1da177e4 LT |
2285 | } |
2286 | ||
f019a7a5 EB |
2287 | /* Trivial set of netlink ops to allow deleting tun or tap |
2288 | * device with netlink. | |
2289 | */ | |
a8b8a889 MS |
2290 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[], |
2291 | struct netlink_ext_ack *extack) | |
f019a7a5 EB |
2292 | { |
2293 | return -EINVAL; | |
2294 | } | |
2295 | ||
1ec010e7 SD |
2296 | static size_t tun_get_size(const struct net_device *dev) |
2297 | { | |
2298 | BUILD_BUG_ON(sizeof(u32) != sizeof(uid_t)); | |
2299 | BUILD_BUG_ON(sizeof(u32) != sizeof(gid_t)); | |
2300 | ||
2301 | return nla_total_size(sizeof(uid_t)) + /* OWNER */ | |
2302 | nla_total_size(sizeof(gid_t)) + /* GROUP */ | |
2303 | nla_total_size(sizeof(u8)) + /* TYPE */ | |
2304 | nla_total_size(sizeof(u8)) + /* PI */ | |
2305 | nla_total_size(sizeof(u8)) + /* VNET_HDR */ | |
2306 | nla_total_size(sizeof(u8)) + /* PERSIST */ | |
2307 | nla_total_size(sizeof(u8)) + /* MULTI_QUEUE */ | |
2308 | nla_total_size(sizeof(u32)) + /* NUM_QUEUES */ | |
2309 | nla_total_size(sizeof(u32)) + /* NUM_DISABLED_QUEUES */ | |
2310 | 0; | |
2311 | } | |
2312 | ||
2313 | static int tun_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
2314 | { | |
2315 | struct tun_struct *tun = netdev_priv(dev); | |
2316 | ||
2317 | if (nla_put_u8(skb, IFLA_TUN_TYPE, tun->flags & TUN_TYPE_MASK)) | |
2318 | goto nla_put_failure; | |
2319 | if (uid_valid(tun->owner) && | |
2320 | nla_put_u32(skb, IFLA_TUN_OWNER, | |
2321 | from_kuid_munged(current_user_ns(), tun->owner))) | |
2322 | goto nla_put_failure; | |
2323 | if (gid_valid(tun->group) && | |
2324 | nla_put_u32(skb, IFLA_TUN_GROUP, | |
2325 | from_kgid_munged(current_user_ns(), tun->group))) | |
2326 | goto nla_put_failure; | |
2327 | if (nla_put_u8(skb, IFLA_TUN_PI, !(tun->flags & IFF_NO_PI))) | |
2328 | goto nla_put_failure; | |
2329 | if (nla_put_u8(skb, IFLA_TUN_VNET_HDR, !!(tun->flags & IFF_VNET_HDR))) | |
2330 | goto nla_put_failure; | |
2331 | if (nla_put_u8(skb, IFLA_TUN_PERSIST, !!(tun->flags & IFF_PERSIST))) | |
2332 | goto nla_put_failure; | |
2333 | if (nla_put_u8(skb, IFLA_TUN_MULTI_QUEUE, | |
2334 | !!(tun->flags & IFF_MULTI_QUEUE))) | |
2335 | goto nla_put_failure; | |
2336 | if (tun->flags & IFF_MULTI_QUEUE) { | |
2337 | if (nla_put_u32(skb, IFLA_TUN_NUM_QUEUES, tun->numqueues)) | |
2338 | goto nla_put_failure; | |
2339 | if (nla_put_u32(skb, IFLA_TUN_NUM_DISABLED_QUEUES, | |
2340 | tun->numdisabled)) | |
2341 | goto nla_put_failure; | |
2342 | } | |
2343 | ||
2344 | return 0; | |
2345 | ||
2346 | nla_put_failure: | |
2347 | return -EMSGSIZE; | |
2348 | } | |
2349 | ||
f019a7a5 EB |
2350 | static struct rtnl_link_ops tun_link_ops __read_mostly = { |
2351 | .kind = DRV_NAME, | |
2352 | .priv_size = sizeof(struct tun_struct), | |
2353 | .setup = tun_setup, | |
2354 | .validate = tun_validate, | |
1ec010e7 SD |
2355 | .get_size = tun_get_size, |
2356 | .fill_info = tun_fill_info, | |
f019a7a5 EB |
2357 | }; |
2358 | ||
33dccbb0 HX |
2359 | static void tun_sock_write_space(struct sock *sk) |
2360 | { | |
54f968d6 | 2361 | struct tun_file *tfile; |
43815482 | 2362 | wait_queue_head_t *wqueue; |
33dccbb0 HX |
2363 | |
2364 | if (!sock_writeable(sk)) | |
2365 | return; | |
2366 | ||
9cd3e072 | 2367 | if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
33dccbb0 HX |
2368 | return; |
2369 | ||
43815482 ED |
2370 | wqueue = sk_sleep(sk); |
2371 | if (wqueue && waitqueue_active(wqueue)) | |
a9a08845 LT |
2372 | wake_up_interruptible_sync_poll(wqueue, EPOLLOUT | |
2373 | EPOLLWRNORM | EPOLLWRBAND); | |
c722c625 | 2374 | |
54f968d6 JW |
2375 | tfile = container_of(sk, struct tun_file, sk); |
2376 | kill_fasync(&tfile->fasync, SIGIO, POLL_OUT); | |
33dccbb0 HX |
2377 | } |
2378 | ||
1b784140 | 2379 | static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len) |
05c2828c | 2380 | { |
54f968d6 JW |
2381 | int ret; |
2382 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
9484dc74 | 2383 | struct tun_struct *tun = tun_get(tfile); |
54f968d6 JW |
2384 | |
2385 | if (!tun) | |
2386 | return -EBADFD; | |
f5ff53b4 | 2387 | |
c0371da6 | 2388 | ret = tun_get_user(tun, tfile, m->msg_control, &m->msg_iter, |
5503fcec JW |
2389 | m->msg_flags & MSG_DONTWAIT, |
2390 | m->msg_flags & MSG_MORE); | |
54f968d6 JW |
2391 | tun_put(tun); |
2392 | return ret; | |
05c2828c MT |
2393 | } |
2394 | ||
1b784140 | 2395 | static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len, |
05c2828c MT |
2396 | int flags) |
2397 | { | |
54f968d6 | 2398 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); |
9484dc74 | 2399 | struct tun_struct *tun = tun_get(tfile); |
fc72d1d5 | 2400 | void *ptr = m->msg_control; |
05c2828c | 2401 | int ret; |
54f968d6 | 2402 | |
c33ee15b WX |
2403 | if (!tun) { |
2404 | ret = -EBADFD; | |
fc72d1d5 | 2405 | goto out_free; |
c33ee15b | 2406 | } |
54f968d6 | 2407 | |
eda29772 | 2408 | if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) { |
3811ae76 | 2409 | ret = -EINVAL; |
c33ee15b | 2410 | goto out_put_tun; |
3811ae76 | 2411 | } |
eda29772 RC |
2412 | if (flags & MSG_ERRQUEUE) { |
2413 | ret = sock_recv_errqueue(sock->sk, m, total_len, | |
2414 | SOL_PACKET, TUN_TX_TIMESTAMP); | |
2415 | goto out; | |
2416 | } | |
fc72d1d5 | 2417 | ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, ptr); |
87897931 | 2418 | if (ret > (ssize_t)total_len) { |
42404c09 DM |
2419 | m->msg_flags |= MSG_TRUNC; |
2420 | ret = flags & MSG_TRUNC ? ret : total_len; | |
2421 | } | |
3811ae76 | 2422 | out: |
54f968d6 | 2423 | tun_put(tun); |
05c2828c | 2424 | return ret; |
c33ee15b WX |
2425 | |
2426 | out_put_tun: | |
2427 | tun_put(tun); | |
fc72d1d5 JW |
2428 | out_free: |
2429 | tun_ptr_free(ptr); | |
c33ee15b | 2430 | return ret; |
05c2828c MT |
2431 | } |
2432 | ||
fc72d1d5 JW |
2433 | static int tun_ptr_peek_len(void *ptr) |
2434 | { | |
2435 | if (likely(ptr)) { | |
1ffcbc85 JDB |
2436 | if (tun_is_xdp_frame(ptr)) { |
2437 | struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr); | |
fc72d1d5 | 2438 | |
1ffcbc85 | 2439 | return xdpf->len; |
fc72d1d5 JW |
2440 | } |
2441 | return __skb_array_len_with_tag(ptr); | |
2442 | } else { | |
2443 | return 0; | |
2444 | } | |
2445 | } | |
2446 | ||
1576d986 JW |
2447 | static int tun_peek_len(struct socket *sock) |
2448 | { | |
2449 | struct tun_file *tfile = container_of(sock, struct tun_file, socket); | |
2450 | struct tun_struct *tun; | |
2451 | int ret = 0; | |
2452 | ||
9484dc74 | 2453 | tun = tun_get(tfile); |
1576d986 JW |
2454 | if (!tun) |
2455 | return 0; | |
2456 | ||
fc72d1d5 | 2457 | ret = PTR_RING_PEEK_CALL(&tfile->tx_ring, tun_ptr_peek_len); |
1576d986 JW |
2458 | tun_put(tun); |
2459 | ||
2460 | return ret; | |
2461 | } | |
2462 | ||
05c2828c MT |
2463 | /* Ops structure to mimic raw sockets with tun */ |
2464 | static const struct proto_ops tun_socket_ops = { | |
1576d986 | 2465 | .peek_len = tun_peek_len, |
05c2828c MT |
2466 | .sendmsg = tun_sendmsg, |
2467 | .recvmsg = tun_recvmsg, | |
2468 | }; | |
2469 | ||
33dccbb0 HX |
2470 | static struct proto tun_proto = { |
2471 | .name = "tun", | |
2472 | .owner = THIS_MODULE, | |
54f968d6 | 2473 | .obj_size = sizeof(struct tun_file), |
33dccbb0 | 2474 | }; |
f019a7a5 | 2475 | |
980c9e8c DW |
2476 | static int tun_flags(struct tun_struct *tun) |
2477 | { | |
031f5e03 | 2478 | return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP); |
980c9e8c DW |
2479 | } |
2480 | ||
2481 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, | |
2482 | char *buf) | |
2483 | { | |
2484 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
2485 | return sprintf(buf, "0x%x\n", tun_flags(tun)); | |
2486 | } | |
2487 | ||
2488 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, | |
2489 | char *buf) | |
2490 | { | |
2491 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2492 | return uid_valid(tun->owner)? |
2493 | sprintf(buf, "%u\n", | |
2494 | from_kuid_munged(current_user_ns(), tun->owner)): | |
2495 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2496 | } |
2497 | ||
2498 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, | |
2499 | char *buf) | |
2500 | { | |
2501 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); | |
0625c883 EB |
2502 | return gid_valid(tun->group) ? |
2503 | sprintf(buf, "%u\n", | |
2504 | from_kgid_munged(current_user_ns(), tun->group)): | |
2505 | sprintf(buf, "-1\n"); | |
980c9e8c DW |
2506 | } |
2507 | ||
2508 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); | |
2509 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); | |
2510 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); | |
2511 | ||
c4d33e24 TI |
2512 | static struct attribute *tun_dev_attrs[] = { |
2513 | &dev_attr_tun_flags.attr, | |
2514 | &dev_attr_owner.attr, | |
2515 | &dev_attr_group.attr, | |
2516 | NULL | |
2517 | }; | |
2518 | ||
2519 | static const struct attribute_group tun_attr_group = { | |
2520 | .attrs = tun_dev_attrs | |
2521 | }; | |
2522 | ||
d647a591 | 2523 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
1da177e4 LT |
2524 | { |
2525 | struct tun_struct *tun; | |
54f968d6 | 2526 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
2527 | struct net_device *dev; |
2528 | int err; | |
2529 | ||
7c0c3b1a JW |
2530 | if (tfile->detached) |
2531 | return -EINVAL; | |
2532 | ||
90e33d45 PP |
2533 | if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) { |
2534 | if (!capable(CAP_NET_ADMIN)) | |
2535 | return -EPERM; | |
2536 | ||
2537 | if (!(ifr->ifr_flags & IFF_NAPI) || | |
2538 | (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP) | |
2539 | return -EINVAL; | |
2540 | } | |
2541 | ||
74a3e5a7 EB |
2542 | dev = __dev_get_by_name(net, ifr->ifr_name); |
2543 | if (dev) { | |
f85ba780 DW |
2544 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
2545 | return -EBUSY; | |
74a3e5a7 EB |
2546 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
2547 | tun = netdev_priv(dev); | |
2548 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) | |
2549 | tun = netdev_priv(dev); | |
2550 | else | |
2551 | return -EINVAL; | |
2552 | ||
8e6d91ae | 2553 | if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) != |
40630b82 | 2554 | !!(tun->flags & IFF_MULTI_QUEUE)) |
8e6d91ae JW |
2555 | return -EINVAL; |
2556 | ||
cde8b15f | 2557 | if (tun_not_capable(tun)) |
2b980dbd | 2558 | return -EPERM; |
5dbbaf2d | 2559 | err = security_tun_dev_open(tun->security); |
2b980dbd PM |
2560 | if (err < 0) |
2561 | return err; | |
2562 | ||
94317099 PP |
2563 | err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER, |
2564 | ifr->ifr_flags & IFF_NAPI); | |
a7385ba2 EB |
2565 | if (err < 0) |
2566 | return err; | |
4008e97f | 2567 | |
40630b82 | 2568 | if (tun->flags & IFF_MULTI_QUEUE && |
e8dbad66 JW |
2569 | (tun->numqueues + tun->numdisabled > 1)) { |
2570 | /* One or more queue has already been attached, no need | |
2571 | * to initialize the device again. | |
2572 | */ | |
83c1f36f | 2573 | netdev_state_change(dev); |
e8dbad66 JW |
2574 | return 0; |
2575 | } | |
9fffc5c6 SD |
2576 | |
2577 | tun->flags = (tun->flags & ~TUN_FEATURES) | | |
2578 | (ifr->ifr_flags & TUN_FEATURES); | |
83c1f36f SD |
2579 | |
2580 | netdev_state_change(dev); | |
2581 | } else { | |
1da177e4 LT |
2582 | char *name; |
2583 | unsigned long flags = 0; | |
edfb6a14 JW |
2584 | int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ? |
2585 | MAX_TAP_QUEUES : 1; | |
1da177e4 | 2586 | |
c260b772 | 2587 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
ca6bb5d7 | 2588 | return -EPERM; |
2b980dbd PM |
2589 | err = security_tun_dev_create(); |
2590 | if (err < 0) | |
2591 | return err; | |
ca6bb5d7 | 2592 | |
1da177e4 LT |
2593 | /* Set dev type */ |
2594 | if (ifr->ifr_flags & IFF_TUN) { | |
2595 | /* TUN device */ | |
40630b82 | 2596 | flags |= IFF_TUN; |
1da177e4 LT |
2597 | name = "tun%d"; |
2598 | } else if (ifr->ifr_flags & IFF_TAP) { | |
2599 | /* TAP device */ | |
40630b82 | 2600 | flags |= IFF_TAP; |
1da177e4 | 2601 | name = "tap%d"; |
6aa20a22 | 2602 | } else |
36989b90 | 2603 | return -EINVAL; |
6aa20a22 | 2604 | |
1da177e4 LT |
2605 | if (*ifr->ifr_name) |
2606 | name = ifr->ifr_name; | |
2607 | ||
c8d68e6b | 2608 | dev = alloc_netdev_mqs(sizeof(struct tun_struct), name, |
c835a677 TG |
2609 | NET_NAME_UNKNOWN, tun_setup, queues, |
2610 | queues); | |
edfb6a14 | 2611 | |
1da177e4 LT |
2612 | if (!dev) |
2613 | return -ENOMEM; | |
0ad646c8 | 2614 | err = dev_get_valid_name(net, dev, name); |
5c25f65f | 2615 | if (err < 0) |
0ad646c8 | 2616 | goto err_free_dev; |
1da177e4 | 2617 | |
fc54c658 | 2618 | dev_net_set(dev, net); |
f019a7a5 | 2619 | dev->rtnl_link_ops = &tun_link_ops; |
fb7589a1 | 2620 | dev->ifindex = tfile->ifindex; |
c4d33e24 | 2621 | dev->sysfs_groups[0] = &tun_attr_group; |
758e43b7 | 2622 | |
1da177e4 LT |
2623 | tun = netdev_priv(dev); |
2624 | tun->dev = dev; | |
2625 | tun->flags = flags; | |
f271b2cc | 2626 | tun->txflt.count = 0; |
d9d52b51 | 2627 | tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr); |
33dccbb0 | 2628 | |
eaea34b2 | 2629 | tun->align = NET_SKB_PAD; |
54f968d6 JW |
2630 | tun->filter_attached = false; |
2631 | tun->sndbuf = tfile->socket.sk->sk_sndbuf; | |
5503fcec | 2632 | tun->rx_batched = 0; |
96f84061 | 2633 | RCU_INIT_POINTER(tun->steering_prog, NULL); |
33dccbb0 | 2634 | |
608b9977 PA |
2635 | tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); |
2636 | if (!tun->pcpu_stats) { | |
2637 | err = -ENOMEM; | |
2638 | goto err_free_dev; | |
2639 | } | |
2640 | ||
96442e42 JW |
2641 | spin_lock_init(&tun->lock); |
2642 | ||
5dbbaf2d PM |
2643 | err = security_tun_dev_alloc_security(&tun->security); |
2644 | if (err < 0) | |
608b9977 | 2645 | goto err_free_stat; |
2b980dbd | 2646 | |
1da177e4 | 2647 | tun_net_init(dev); |
944a1376 | 2648 | tun_flow_init(tun); |
96442e42 | 2649 | |
88255375 | 2650 | dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST | |
6680ec68 JW |
2651 | TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX | |
2652 | NETIF_F_HW_VLAN_STAG_TX; | |
2a2bbf17 | 2653 | dev->features = dev->hw_features | NETIF_F_LLTX; |
6671b224 FLVC |
2654 | dev->vlan_features = dev->features & |
2655 | ~(NETIF_F_HW_VLAN_CTAG_TX | | |
2656 | NETIF_F_HW_VLAN_STAG_TX); | |
88255375 | 2657 | |
9fffc5c6 SD |
2658 | tun->flags = (tun->flags & ~TUN_FEATURES) | |
2659 | (ifr->ifr_flags & TUN_FEATURES); | |
2660 | ||
4008e97f | 2661 | INIT_LIST_HEAD(&tun->disabled); |
94317099 | 2662 | err = tun_attach(tun, file, false, ifr->ifr_flags & IFF_NAPI); |
eb0fb363 | 2663 | if (err < 0) |
662ca437 | 2664 | goto err_free_flow; |
eb0fb363 | 2665 | |
1da177e4 LT |
2666 | err = register_netdevice(tun->dev); |
2667 | if (err < 0) | |
662ca437 | 2668 | goto err_detach; |
1da177e4 LT |
2669 | } |
2670 | ||
af668b3c MT |
2671 | netif_carrier_on(tun->dev); |
2672 | ||
6b8a66ee | 2673 | tun_debug(KERN_INFO, tun, "tun_set_iff\n"); |
1da177e4 | 2674 | |
e35259a9 MK |
2675 | /* Make sure persistent devices do not get stuck in |
2676 | * xoff state. | |
2677 | */ | |
2678 | if (netif_running(tun->dev)) | |
c8d68e6b | 2679 | netif_tx_wake_all_queues(tun->dev); |
e35259a9 | 2680 | |
1da177e4 LT |
2681 | strcpy(ifr->ifr_name, tun->dev->name); |
2682 | return 0; | |
2683 | ||
662ca437 JW |
2684 | err_detach: |
2685 | tun_detach_all(dev); | |
ff244c6b ED |
2686 | /* register_netdevice() already called tun_free_netdev() */ |
2687 | goto err_free_dev; | |
2688 | ||
662ca437 JW |
2689 | err_free_flow: |
2690 | tun_flow_uninit(tun); | |
2691 | security_tun_dev_free_security(tun->security); | |
608b9977 PA |
2692 | err_free_stat: |
2693 | free_percpu(tun->pcpu_stats); | |
662ca437 | 2694 | err_free_dev: |
1da177e4 | 2695 | free_netdev(dev); |
1da177e4 LT |
2696 | return err; |
2697 | } | |
2698 | ||
9ce99cf6 | 2699 | static void tun_get_iff(struct net *net, struct tun_struct *tun, |
876bfd4d | 2700 | struct ifreq *ifr) |
e3b99556 | 2701 | { |
6b8a66ee | 2702 | tun_debug(KERN_INFO, tun, "tun_get_iff\n"); |
e3b99556 MM |
2703 | |
2704 | strcpy(ifr->ifr_name, tun->dev->name); | |
2705 | ||
980c9e8c | 2706 | ifr->ifr_flags = tun_flags(tun); |
e3b99556 | 2707 | |
e3b99556 MM |
2708 | } |
2709 | ||
5228ddc9 RR |
2710 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
2711 | * privs required. */ | |
88255375 | 2712 | static int set_offload(struct tun_struct *tun, unsigned long arg) |
5228ddc9 | 2713 | { |
c8f44aff | 2714 | netdev_features_t features = 0; |
5228ddc9 RR |
2715 | |
2716 | if (arg & TUN_F_CSUM) { | |
88255375 | 2717 | features |= NETIF_F_HW_CSUM; |
5228ddc9 RR |
2718 | arg &= ~TUN_F_CSUM; |
2719 | ||
2720 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | |
2721 | if (arg & TUN_F_TSO_ECN) { | |
2722 | features |= NETIF_F_TSO_ECN; | |
2723 | arg &= ~TUN_F_TSO_ECN; | |
2724 | } | |
2725 | if (arg & TUN_F_TSO4) | |
2726 | features |= NETIF_F_TSO; | |
2727 | if (arg & TUN_F_TSO6) | |
2728 | features |= NETIF_F_TSO6; | |
2729 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | |
2730 | } | |
0c19f846 WB |
2731 | |
2732 | arg &= ~TUN_F_UFO; | |
5228ddc9 RR |
2733 | } |
2734 | ||
2735 | /* This gives the user a way to test for new features in future by | |
2736 | * trying to set them. */ | |
2737 | if (arg) | |
2738 | return -EINVAL; | |
2739 | ||
88255375 | 2740 | tun->set_features = features; |
09050957 YI |
2741 | tun->dev->wanted_features &= ~TUN_USER_FEATURES; |
2742 | tun->dev->wanted_features |= features; | |
88255375 | 2743 | netdev_update_features(tun->dev); |
5228ddc9 RR |
2744 | |
2745 | return 0; | |
2746 | } | |
2747 | ||
c8d68e6b JW |
2748 | static void tun_detach_filter(struct tun_struct *tun, int n) |
2749 | { | |
2750 | int i; | |
2751 | struct tun_file *tfile; | |
2752 | ||
2753 | for (i = 0; i < n; i++) { | |
b8deabd3 | 2754 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2755 | lock_sock(tfile->socket.sk); |
2756 | sk_detach_filter(tfile->socket.sk); | |
2757 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2758 | } |
2759 | ||
2760 | tun->filter_attached = false; | |
2761 | } | |
2762 | ||
2763 | static int tun_attach_filter(struct tun_struct *tun) | |
2764 | { | |
2765 | int i, ret = 0; | |
2766 | struct tun_file *tfile; | |
2767 | ||
2768 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2769 | tfile = rtnl_dereference(tun->tfiles[i]); |
8ced425e HFS |
2770 | lock_sock(tfile->socket.sk); |
2771 | ret = sk_attach_filter(&tun->fprog, tfile->socket.sk); | |
2772 | release_sock(tfile->socket.sk); | |
c8d68e6b JW |
2773 | if (ret) { |
2774 | tun_detach_filter(tun, i); | |
2775 | return ret; | |
2776 | } | |
2777 | } | |
2778 | ||
2779 | tun->filter_attached = true; | |
2780 | return ret; | |
2781 | } | |
2782 | ||
2783 | static void tun_set_sndbuf(struct tun_struct *tun) | |
2784 | { | |
2785 | struct tun_file *tfile; | |
2786 | int i; | |
2787 | ||
2788 | for (i = 0; i < tun->numqueues; i++) { | |
b8deabd3 | 2789 | tfile = rtnl_dereference(tun->tfiles[i]); |
c8d68e6b JW |
2790 | tfile->socket.sk->sk_sndbuf = tun->sndbuf; |
2791 | } | |
2792 | } | |
2793 | ||
cde8b15f JW |
2794 | static int tun_set_queue(struct file *file, struct ifreq *ifr) |
2795 | { | |
2796 | struct tun_file *tfile = file->private_data; | |
2797 | struct tun_struct *tun; | |
cde8b15f JW |
2798 | int ret = 0; |
2799 | ||
2800 | rtnl_lock(); | |
2801 | ||
2802 | if (ifr->ifr_flags & IFF_ATTACH_QUEUE) { | |
4008e97f | 2803 | tun = tfile->detached; |
5dbbaf2d | 2804 | if (!tun) { |
cde8b15f | 2805 | ret = -EINVAL; |
5dbbaf2d PM |
2806 | goto unlock; |
2807 | } | |
2808 | ret = security_tun_dev_attach_queue(tun->security); | |
2809 | if (ret < 0) | |
2810 | goto unlock; | |
94317099 | 2811 | ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI); |
4008e97f | 2812 | } else if (ifr->ifr_flags & IFF_DETACH_QUEUE) { |
b8deabd3 | 2813 | tun = rtnl_dereference(tfile->tun); |
40630b82 | 2814 | if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached) |
4008e97f JW |
2815 | ret = -EINVAL; |
2816 | else | |
2817 | __tun_detach(tfile, false); | |
2818 | } else | |
cde8b15f JW |
2819 | ret = -EINVAL; |
2820 | ||
83c1f36f SD |
2821 | if (ret >= 0) |
2822 | netdev_state_change(tun->dev); | |
2823 | ||
5dbbaf2d | 2824 | unlock: |
cde8b15f JW |
2825 | rtnl_unlock(); |
2826 | return ret; | |
2827 | } | |
2828 | ||
cd5681d7 JW |
2829 | static int tun_set_ebpf(struct tun_struct *tun, struct tun_prog **prog_p, |
2830 | void __user *data) | |
96f84061 JW |
2831 | { |
2832 | struct bpf_prog *prog; | |
2833 | int fd; | |
2834 | ||
2835 | if (copy_from_user(&fd, data, sizeof(fd))) | |
2836 | return -EFAULT; | |
2837 | ||
2838 | if (fd == -1) { | |
2839 | prog = NULL; | |
2840 | } else { | |
2841 | prog = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); | |
2842 | if (IS_ERR(prog)) | |
2843 | return PTR_ERR(prog); | |
2844 | } | |
2845 | ||
cd5681d7 | 2846 | return __tun_set_ebpf(tun, prog_p, prog); |
96f84061 JW |
2847 | } |
2848 | ||
50857e2a AB |
2849 | static long __tun_chr_ioctl(struct file *file, unsigned int cmd, |
2850 | unsigned long arg, int ifreq_len) | |
1da177e4 | 2851 | { |
36b50bab | 2852 | struct tun_file *tfile = file->private_data; |
631ab46b | 2853 | struct tun_struct *tun; |
1da177e4 LT |
2854 | void __user* argp = (void __user*)arg; |
2855 | struct ifreq ifr; | |
f2780d6d | 2856 | struct net *net; |
0625c883 EB |
2857 | kuid_t owner; |
2858 | kgid_t group; | |
33dccbb0 | 2859 | int sndbuf; |
d9d52b51 | 2860 | int vnet_hdr_sz; |
fb7589a1 | 2861 | unsigned int ifindex; |
1cf8e410 | 2862 | int le; |
f271b2cc | 2863 | int ret; |
83c1f36f | 2864 | bool do_notify = false; |
1da177e4 | 2865 | |
f2780d6d KT |
2866 | if (cmd == TUNSETIFF || cmd == TUNSETQUEUE || |
2867 | (_IOC_TYPE(cmd) == SOCK_IOC_TYPE && cmd != SIOCGSKNS)) { | |
50857e2a | 2868 | if (copy_from_user(&ifr, argp, ifreq_len)) |
1da177e4 | 2869 | return -EFAULT; |
8bbb1813 | 2870 | } else { |
a117dacd | 2871 | memset(&ifr, 0, sizeof(ifr)); |
8bbb1813 | 2872 | } |
631ab46b EB |
2873 | if (cmd == TUNGETFEATURES) { |
2874 | /* Currently this just means: "what IFF flags are valid?". | |
2875 | * This is needed because we never checked for invalid flags on | |
031f5e03 MT |
2876 | * TUNSETIFF. |
2877 | */ | |
2878 | return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES, | |
631ab46b | 2879 | (unsigned int __user*)argp); |
cde8b15f JW |
2880 | } else if (cmd == TUNSETQUEUE) |
2881 | return tun_set_queue(file, &ifr); | |
631ab46b | 2882 | |
c8d68e6b | 2883 | ret = 0; |
876bfd4d HX |
2884 | rtnl_lock(); |
2885 | ||
9484dc74 | 2886 | tun = tun_get(tfile); |
f2780d6d | 2887 | net = sock_net(&tfile->sk); |
0f16bc13 GF |
2888 | if (cmd == TUNSETIFF) { |
2889 | ret = -EEXIST; | |
2890 | if (tun) | |
2891 | goto unlock; | |
2892 | ||
1da177e4 LT |
2893 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
2894 | ||
f2780d6d | 2895 | ret = tun_set_iff(net, file, &ifr); |
1da177e4 | 2896 | |
876bfd4d HX |
2897 | if (ret) |
2898 | goto unlock; | |
1da177e4 | 2899 | |
50857e2a | 2900 | if (copy_to_user(argp, &ifr, ifreq_len)) |
876bfd4d HX |
2901 | ret = -EFAULT; |
2902 | goto unlock; | |
1da177e4 | 2903 | } |
fb7589a1 PE |
2904 | if (cmd == TUNSETIFINDEX) { |
2905 | ret = -EPERM; | |
2906 | if (tun) | |
2907 | goto unlock; | |
2908 | ||
2909 | ret = -EFAULT; | |
2910 | if (copy_from_user(&ifindex, argp, sizeof(ifindex))) | |
2911 | goto unlock; | |
2912 | ||
2913 | ret = 0; | |
2914 | tfile->ifindex = ifindex; | |
2915 | goto unlock; | |
2916 | } | |
f2780d6d KT |
2917 | if (cmd == SIOCGSKNS) { |
2918 | ret = -EPERM; | |
2919 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
2920 | goto unlock; | |
2921 | ||
2922 | ret = open_related_ns(&net->ns, get_net_ns); | |
2923 | goto unlock; | |
2924 | } | |
1da177e4 | 2925 | |
876bfd4d | 2926 | ret = -EBADFD; |
1da177e4 | 2927 | if (!tun) |
876bfd4d | 2928 | goto unlock; |
1da177e4 | 2929 | |
1e588338 | 2930 | tun_debug(KERN_INFO, tun, "tun_chr_ioctl cmd %u\n", cmd); |
1da177e4 | 2931 | |
631ab46b | 2932 | ret = 0; |
1da177e4 | 2933 | switch (cmd) { |
e3b99556 | 2934 | case TUNGETIFF: |
9ce99cf6 | 2935 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
e3b99556 | 2936 | |
3d407a80 PE |
2937 | if (tfile->detached) |
2938 | ifr.ifr_flags |= IFF_DETACH_QUEUE; | |
849c9b6f PE |
2939 | if (!tfile->socket.sk->sk_filter) |
2940 | ifr.ifr_flags |= IFF_NOFILTER; | |
3d407a80 | 2941 | |
50857e2a | 2942 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b | 2943 | ret = -EFAULT; |
e3b99556 MM |
2944 | break; |
2945 | ||
1da177e4 LT |
2946 | case TUNSETNOCSUM: |
2947 | /* Disable/Enable checksum */ | |
1da177e4 | 2948 | |
88255375 MM |
2949 | /* [unimplemented] */ |
2950 | tun_debug(KERN_INFO, tun, "ignored: set checksum %s\n", | |
6b8a66ee | 2951 | arg ? "disabled" : "enabled"); |
1da177e4 LT |
2952 | break; |
2953 | ||
2954 | case TUNSETPERSIST: | |
54f968d6 JW |
2955 | /* Disable/Enable persist mode. Keep an extra reference to the |
2956 | * module to prevent the module being unprobed. | |
2957 | */ | |
40630b82 MT |
2958 | if (arg && !(tun->flags & IFF_PERSIST)) { |
2959 | tun->flags |= IFF_PERSIST; | |
54f968d6 | 2960 | __module_get(THIS_MODULE); |
83c1f36f | 2961 | do_notify = true; |
dd38bd85 | 2962 | } |
40630b82 MT |
2963 | if (!arg && (tun->flags & IFF_PERSIST)) { |
2964 | tun->flags &= ~IFF_PERSIST; | |
54f968d6 | 2965 | module_put(THIS_MODULE); |
83c1f36f | 2966 | do_notify = true; |
54f968d6 | 2967 | } |
1da177e4 | 2968 | |
6b8a66ee JP |
2969 | tun_debug(KERN_INFO, tun, "persist %s\n", |
2970 | arg ? "enabled" : "disabled"); | |
1da177e4 LT |
2971 | break; |
2972 | ||
2973 | case TUNSETOWNER: | |
2974 | /* Set owner of the device */ | |
0625c883 EB |
2975 | owner = make_kuid(current_user_ns(), arg); |
2976 | if (!uid_valid(owner)) { | |
2977 | ret = -EINVAL; | |
2978 | break; | |
2979 | } | |
2980 | tun->owner = owner; | |
83c1f36f | 2981 | do_notify = true; |
1e588338 | 2982 | tun_debug(KERN_INFO, tun, "owner set to %u\n", |
0625c883 | 2983 | from_kuid(&init_user_ns, tun->owner)); |
1da177e4 LT |
2984 | break; |
2985 | ||
8c644623 GG |
2986 | case TUNSETGROUP: |
2987 | /* Set group of the device */ | |
0625c883 EB |
2988 | group = make_kgid(current_user_ns(), arg); |
2989 | if (!gid_valid(group)) { | |
2990 | ret = -EINVAL; | |
2991 | break; | |
2992 | } | |
2993 | tun->group = group; | |
83c1f36f | 2994 | do_notify = true; |
1e588338 | 2995 | tun_debug(KERN_INFO, tun, "group set to %u\n", |
0625c883 | 2996 | from_kgid(&init_user_ns, tun->group)); |
8c644623 GG |
2997 | break; |
2998 | ||
ff4cc3ac MK |
2999 | case TUNSETLINK: |
3000 | /* Only allow setting the type when the interface is down */ | |
3001 | if (tun->dev->flags & IFF_UP) { | |
6b8a66ee JP |
3002 | tun_debug(KERN_INFO, tun, |
3003 | "Linktype set failed because interface is up\n"); | |
48abfe05 | 3004 | ret = -EBUSY; |
ff4cc3ac MK |
3005 | } else { |
3006 | tun->dev->type = (int) arg; | |
6b8a66ee JP |
3007 | tun_debug(KERN_INFO, tun, "linktype set to %d\n", |
3008 | tun->dev->type); | |
48abfe05 | 3009 | ret = 0; |
ff4cc3ac | 3010 | } |
631ab46b | 3011 | break; |
ff4cc3ac | 3012 | |
1da177e4 LT |
3013 | #ifdef TUN_DEBUG |
3014 | case TUNSETDEBUG: | |
3015 | tun->debug = arg; | |
3016 | break; | |
3017 | #endif | |
5228ddc9 | 3018 | case TUNSETOFFLOAD: |
88255375 | 3019 | ret = set_offload(tun, arg); |
631ab46b | 3020 | break; |
5228ddc9 | 3021 | |
f271b2cc MK |
3022 | case TUNSETTXFILTER: |
3023 | /* Can be set only for TAPs */ | |
631ab46b | 3024 | ret = -EINVAL; |
40630b82 | 3025 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
631ab46b | 3026 | break; |
c0e5a8c2 | 3027 | ret = update_filter(&tun->txflt, (void __user *)arg); |
631ab46b | 3028 | break; |
1da177e4 LT |
3029 | |
3030 | case SIOCGIFHWADDR: | |
b595076a | 3031 | /* Get hw address */ |
f271b2cc MK |
3032 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
3033 | ifr.ifr_hwaddr.sa_family = tun->dev->type; | |
50857e2a | 3034 | if (copy_to_user(argp, &ifr, ifreq_len)) |
631ab46b EB |
3035 | ret = -EFAULT; |
3036 | break; | |
1da177e4 LT |
3037 | |
3038 | case SIOCSIFHWADDR: | |
f271b2cc | 3039 | /* Set hw address */ |
6b8a66ee JP |
3040 | tun_debug(KERN_DEBUG, tun, "set hw address: %pM\n", |
3041 | ifr.ifr_hwaddr.sa_data); | |
40102371 | 3042 | |
40102371 | 3043 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
631ab46b | 3044 | break; |
33dccbb0 HX |
3045 | |
3046 | case TUNGETSNDBUF: | |
54f968d6 | 3047 | sndbuf = tfile->socket.sk->sk_sndbuf; |
33dccbb0 HX |
3048 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
3049 | ret = -EFAULT; | |
3050 | break; | |
3051 | ||
3052 | case TUNSETSNDBUF: | |
3053 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { | |
3054 | ret = -EFAULT; | |
3055 | break; | |
3056 | } | |
93161922 CG |
3057 | if (sndbuf <= 0) { |
3058 | ret = -EINVAL; | |
3059 | break; | |
3060 | } | |
33dccbb0 | 3061 | |
c8d68e6b JW |
3062 | tun->sndbuf = sndbuf; |
3063 | tun_set_sndbuf(tun); | |
33dccbb0 HX |
3064 | break; |
3065 | ||
d9d52b51 MT |
3066 | case TUNGETVNETHDRSZ: |
3067 | vnet_hdr_sz = tun->vnet_hdr_sz; | |
3068 | if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz))) | |
3069 | ret = -EFAULT; | |
3070 | break; | |
3071 | ||
3072 | case TUNSETVNETHDRSZ: | |
3073 | if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) { | |
3074 | ret = -EFAULT; | |
3075 | break; | |
3076 | } | |
3077 | if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) { | |
3078 | ret = -EINVAL; | |
3079 | break; | |
3080 | } | |
3081 | ||
3082 | tun->vnet_hdr_sz = vnet_hdr_sz; | |
3083 | break; | |
3084 | ||
1cf8e410 MT |
3085 | case TUNGETVNETLE: |
3086 | le = !!(tun->flags & TUN_VNET_LE); | |
3087 | if (put_user(le, (int __user *)argp)) | |
3088 | ret = -EFAULT; | |
3089 | break; | |
3090 | ||
3091 | case TUNSETVNETLE: | |
3092 | if (get_user(le, (int __user *)argp)) { | |
3093 | ret = -EFAULT; | |
3094 | break; | |
3095 | } | |
3096 | if (le) | |
3097 | tun->flags |= TUN_VNET_LE; | |
3098 | else | |
3099 | tun->flags &= ~TUN_VNET_LE; | |
3100 | break; | |
3101 | ||
8b8e658b GK |
3102 | case TUNGETVNETBE: |
3103 | ret = tun_get_vnet_be(tun, argp); | |
3104 | break; | |
3105 | ||
3106 | case TUNSETVNETBE: | |
3107 | ret = tun_set_vnet_be(tun, argp); | |
3108 | break; | |
3109 | ||
99405162 MT |
3110 | case TUNATTACHFILTER: |
3111 | /* Can be set only for TAPs */ | |
3112 | ret = -EINVAL; | |
40630b82 | 3113 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 MT |
3114 | break; |
3115 | ret = -EFAULT; | |
54f968d6 | 3116 | if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog))) |
99405162 MT |
3117 | break; |
3118 | ||
c8d68e6b | 3119 | ret = tun_attach_filter(tun); |
99405162 MT |
3120 | break; |
3121 | ||
3122 | case TUNDETACHFILTER: | |
3123 | /* Can be set only for TAPs */ | |
3124 | ret = -EINVAL; | |
40630b82 | 3125 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
99405162 | 3126 | break; |
c8d68e6b JW |
3127 | ret = 0; |
3128 | tun_detach_filter(tun, tun->numqueues); | |
99405162 MT |
3129 | break; |
3130 | ||
76975e9c PE |
3131 | case TUNGETFILTER: |
3132 | ret = -EINVAL; | |
40630b82 | 3133 | if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP) |
76975e9c PE |
3134 | break; |
3135 | ret = -EFAULT; | |
3136 | if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog))) | |
3137 | break; | |
3138 | ret = 0; | |
3139 | break; | |
3140 | ||
96f84061 | 3141 | case TUNSETSTEERINGEBPF: |
cd5681d7 | 3142 | ret = tun_set_ebpf(tun, &tun->steering_prog, argp); |
96f84061 JW |
3143 | break; |
3144 | ||
aff3d70a JW |
3145 | case TUNSETFILTEREBPF: |
3146 | ret = tun_set_ebpf(tun, &tun->filter_prog, argp); | |
3147 | break; | |
3148 | ||
1da177e4 | 3149 | default: |
631ab46b EB |
3150 | ret = -EINVAL; |
3151 | break; | |
ee289b64 | 3152 | } |
1da177e4 | 3153 | |
83c1f36f SD |
3154 | if (do_notify) |
3155 | netdev_state_change(tun->dev); | |
3156 | ||
876bfd4d HX |
3157 | unlock: |
3158 | rtnl_unlock(); | |
3159 | if (tun) | |
3160 | tun_put(tun); | |
631ab46b | 3161 | return ret; |
1da177e4 LT |
3162 | } |
3163 | ||
50857e2a AB |
3164 | static long tun_chr_ioctl(struct file *file, |
3165 | unsigned int cmd, unsigned long arg) | |
3166 | { | |
3167 | return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq)); | |
3168 | } | |
3169 | ||
3170 | #ifdef CONFIG_COMPAT | |
3171 | static long tun_chr_compat_ioctl(struct file *file, | |
3172 | unsigned int cmd, unsigned long arg) | |
3173 | { | |
3174 | switch (cmd) { | |
3175 | case TUNSETIFF: | |
3176 | case TUNGETIFF: | |
3177 | case TUNSETTXFILTER: | |
3178 | case TUNGETSNDBUF: | |
3179 | case TUNSETSNDBUF: | |
3180 | case SIOCGIFHWADDR: | |
3181 | case SIOCSIFHWADDR: | |
3182 | arg = (unsigned long)compat_ptr(arg); | |
3183 | break; | |
3184 | default: | |
3185 | arg = (compat_ulong_t)arg; | |
3186 | break; | |
3187 | } | |
3188 | ||
3189 | /* | |
3190 | * compat_ifreq is shorter than ifreq, so we must not access beyond | |
3191 | * the end of that structure. All fields that are used in this | |
3192 | * driver are compatible though, we don't need to convert the | |
3193 | * contents. | |
3194 | */ | |
3195 | return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq)); | |
3196 | } | |
3197 | #endif /* CONFIG_COMPAT */ | |
3198 | ||
1da177e4 LT |
3199 | static int tun_chr_fasync(int fd, struct file *file, int on) |
3200 | { | |
54f968d6 | 3201 | struct tun_file *tfile = file->private_data; |
1da177e4 LT |
3202 | int ret; |
3203 | ||
54f968d6 | 3204 | if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0) |
9d319522 | 3205 | goto out; |
6aa20a22 | 3206 | |
1da177e4 | 3207 | if (on) { |
e0b93edd | 3208 | __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
54f968d6 | 3209 | tfile->flags |= TUN_FASYNC; |
6aa20a22 | 3210 | } else |
54f968d6 | 3211 | tfile->flags &= ~TUN_FASYNC; |
9d319522 JC |
3212 | ret = 0; |
3213 | out: | |
9d319522 | 3214 | return ret; |
1da177e4 LT |
3215 | } |
3216 | ||
3217 | static int tun_chr_open(struct inode *inode, struct file * file) | |
3218 | { | |
140e807d | 3219 | struct net *net = current->nsproxy->net_ns; |
631ab46b | 3220 | struct tun_file *tfile; |
deed49fb | 3221 | |
6b8a66ee | 3222 | DBG1(KERN_INFO, "tunX: tun_chr_open\n"); |
631ab46b | 3223 | |
140e807d | 3224 | tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL, |
11aa9c28 | 3225 | &tun_proto, 0); |
631ab46b EB |
3226 | if (!tfile) |
3227 | return -ENOMEM; | |
c956674b | 3228 | RCU_INIT_POINTER(tfile->tun, NULL); |
54f968d6 | 3229 | tfile->flags = 0; |
fb7589a1 | 3230 | tfile->ifindex = 0; |
54f968d6 | 3231 | |
54f968d6 | 3232 | init_waitqueue_head(&tfile->wq.wait); |
9e641bdc | 3233 | RCU_INIT_POINTER(tfile->socket.wq, &tfile->wq); |
54f968d6 JW |
3234 | |
3235 | tfile->socket.file = file; | |
3236 | tfile->socket.ops = &tun_socket_ops; | |
3237 | ||
3238 | sock_init_data(&tfile->socket, &tfile->sk); | |
54f968d6 JW |
3239 | |
3240 | tfile->sk.sk_write_space = tun_sock_write_space; | |
3241 | tfile->sk.sk_sndbuf = INT_MAX; | |
3242 | ||
631ab46b | 3243 | file->private_data = tfile; |
4008e97f | 3244 | INIT_LIST_HEAD(&tfile->next); |
54f968d6 | 3245 | |
19a6afb2 JW |
3246 | sock_set_flag(&tfile->sk, SOCK_ZEROCOPY); |
3247 | ||
8565d26b | 3248 | memset(&tfile->tx_ring, 0, sizeof(tfile->tx_ring)); |
4df0bfc7 | 3249 | |
1da177e4 LT |
3250 | return 0; |
3251 | } | |
3252 | ||
3253 | static int tun_chr_close(struct inode *inode, struct file *file) | |
3254 | { | |
631ab46b | 3255 | struct tun_file *tfile = file->private_data; |
1da177e4 | 3256 | |
c8d68e6b | 3257 | tun_detach(tfile, true); |
1da177e4 LT |
3258 | |
3259 | return 0; | |
3260 | } | |
3261 | ||
93e14b6d | 3262 | #ifdef CONFIG_PROC_FS |
9484dc74 | 3263 | static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file) |
93e14b6d | 3264 | { |
9484dc74 | 3265 | struct tun_file *tfile = file->private_data; |
93e14b6d MY |
3266 | struct tun_struct *tun; |
3267 | struct ifreq ifr; | |
3268 | ||
3269 | memset(&ifr, 0, sizeof(ifr)); | |
3270 | ||
3271 | rtnl_lock(); | |
9484dc74 | 3272 | tun = tun_get(tfile); |
93e14b6d MY |
3273 | if (tun) |
3274 | tun_get_iff(current->nsproxy->net_ns, tun, &ifr); | |
3275 | rtnl_unlock(); | |
3276 | ||
3277 | if (tun) | |
3278 | tun_put(tun); | |
3279 | ||
a3816ab0 | 3280 | seq_printf(m, "iff:\t%s\n", ifr.ifr_name); |
93e14b6d MY |
3281 | } |
3282 | #endif | |
3283 | ||
d54b1fdb | 3284 | static const struct file_operations tun_fops = { |
6aa20a22 | 3285 | .owner = THIS_MODULE, |
1da177e4 | 3286 | .llseek = no_llseek, |
9b067034 | 3287 | .read_iter = tun_chr_read_iter, |
f5ff53b4 | 3288 | .write_iter = tun_chr_write_iter, |
1da177e4 | 3289 | .poll = tun_chr_poll, |
50857e2a AB |
3290 | .unlocked_ioctl = tun_chr_ioctl, |
3291 | #ifdef CONFIG_COMPAT | |
3292 | .compat_ioctl = tun_chr_compat_ioctl, | |
3293 | #endif | |
1da177e4 LT |
3294 | .open = tun_chr_open, |
3295 | .release = tun_chr_close, | |
93e14b6d MY |
3296 | .fasync = tun_chr_fasync, |
3297 | #ifdef CONFIG_PROC_FS | |
3298 | .show_fdinfo = tun_chr_show_fdinfo, | |
3299 | #endif | |
1da177e4 LT |
3300 | }; |
3301 | ||
3302 | static struct miscdevice tun_miscdev = { | |
3303 | .minor = TUN_MINOR, | |
3304 | .name = "tun", | |
e454cea2 | 3305 | .nodename = "net/tun", |
1da177e4 | 3306 | .fops = &tun_fops, |
1da177e4 LT |
3307 | }; |
3308 | ||
3309 | /* ethtool interface */ | |
3310 | ||
29ccc49d PR |
3311 | static int tun_get_link_ksettings(struct net_device *dev, |
3312 | struct ethtool_link_ksettings *cmd) | |
3313 | { | |
3314 | ethtool_link_ksettings_zero_link_mode(cmd, supported); | |
3315 | ethtool_link_ksettings_zero_link_mode(cmd, advertising); | |
3316 | cmd->base.speed = SPEED_10; | |
3317 | cmd->base.duplex = DUPLEX_FULL; | |
3318 | cmd->base.port = PORT_TP; | |
3319 | cmd->base.phy_address = 0; | |
3320 | cmd->base.autoneg = AUTONEG_DISABLE; | |
1da177e4 LT |
3321 | return 0; |
3322 | } | |
3323 | ||
3324 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
3325 | { | |
3326 | struct tun_struct *tun = netdev_priv(dev); | |
3327 | ||
33a5ba14 RJ |
3328 | strlcpy(info->driver, DRV_NAME, sizeof(info->driver)); |
3329 | strlcpy(info->version, DRV_VERSION, sizeof(info->version)); | |
1da177e4 LT |
3330 | |
3331 | switch (tun->flags & TUN_TYPE_MASK) { | |
40630b82 | 3332 | case IFF_TUN: |
33a5ba14 | 3333 | strlcpy(info->bus_info, "tun", sizeof(info->bus_info)); |
1da177e4 | 3334 | break; |
40630b82 | 3335 | case IFF_TAP: |
33a5ba14 | 3336 | strlcpy(info->bus_info, "tap", sizeof(info->bus_info)); |
1da177e4 LT |
3337 | break; |
3338 | } | |
3339 | } | |
3340 | ||
3341 | static u32 tun_get_msglevel(struct net_device *dev) | |
3342 | { | |
3343 | #ifdef TUN_DEBUG | |
3344 | struct tun_struct *tun = netdev_priv(dev); | |
3345 | return tun->debug; | |
3346 | #else | |
3347 | return -EOPNOTSUPP; | |
3348 | #endif | |
3349 | } | |
3350 | ||
3351 | static void tun_set_msglevel(struct net_device *dev, u32 value) | |
3352 | { | |
3353 | #ifdef TUN_DEBUG | |
3354 | struct tun_struct *tun = netdev_priv(dev); | |
3355 | tun->debug = value; | |
3356 | #endif | |
3357 | } | |
3358 | ||
5503fcec JW |
3359 | static int tun_get_coalesce(struct net_device *dev, |
3360 | struct ethtool_coalesce *ec) | |
3361 | { | |
3362 | struct tun_struct *tun = netdev_priv(dev); | |
3363 | ||
3364 | ec->rx_max_coalesced_frames = tun->rx_batched; | |
3365 | ||
3366 | return 0; | |
3367 | } | |
3368 | ||
3369 | static int tun_set_coalesce(struct net_device *dev, | |
3370 | struct ethtool_coalesce *ec) | |
3371 | { | |
3372 | struct tun_struct *tun = netdev_priv(dev); | |
3373 | ||
3374 | if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT) | |
3375 | tun->rx_batched = NAPI_POLL_WEIGHT; | |
3376 | else | |
3377 | tun->rx_batched = ec->rx_max_coalesced_frames; | |
3378 | ||
3379 | return 0; | |
3380 | } | |
3381 | ||
7282d491 | 3382 | static const struct ethtool_ops tun_ethtool_ops = { |
1da177e4 LT |
3383 | .get_drvinfo = tun_get_drvinfo, |
3384 | .get_msglevel = tun_get_msglevel, | |
3385 | .set_msglevel = tun_set_msglevel, | |
bee31369 | 3386 | .get_link = ethtool_op_get_link, |
eda29772 | 3387 | .get_ts_info = ethtool_op_get_ts_info, |
5503fcec JW |
3388 | .get_coalesce = tun_get_coalesce, |
3389 | .set_coalesce = tun_set_coalesce, | |
29ccc49d | 3390 | .get_link_ksettings = tun_get_link_ksettings, |
1da177e4 LT |
3391 | }; |
3392 | ||
1576d986 JW |
3393 | static int tun_queue_resize(struct tun_struct *tun) |
3394 | { | |
3395 | struct net_device *dev = tun->dev; | |
3396 | struct tun_file *tfile; | |
5990a305 | 3397 | struct ptr_ring **rings; |
1576d986 JW |
3398 | int n = tun->numqueues + tun->numdisabled; |
3399 | int ret, i; | |
3400 | ||
5990a305 JW |
3401 | rings = kmalloc_array(n, sizeof(*rings), GFP_KERNEL); |
3402 | if (!rings) | |
1576d986 JW |
3403 | return -ENOMEM; |
3404 | ||
3405 | for (i = 0; i < tun->numqueues; i++) { | |
3406 | tfile = rtnl_dereference(tun->tfiles[i]); | |
5990a305 | 3407 | rings[i] = &tfile->tx_ring; |
1576d986 JW |
3408 | } |
3409 | list_for_each_entry(tfile, &tun->disabled, next) | |
5990a305 | 3410 | rings[i++] = &tfile->tx_ring; |
1576d986 | 3411 | |
5990a305 JW |
3412 | ret = ptr_ring_resize_multiple(rings, n, |
3413 | dev->tx_queue_len, GFP_KERNEL, | |
fc72d1d5 | 3414 | tun_ptr_free); |
1576d986 | 3415 | |
5990a305 | 3416 | kfree(rings); |
1576d986 JW |
3417 | return ret; |
3418 | } | |
3419 | ||
3420 | static int tun_device_event(struct notifier_block *unused, | |
3421 | unsigned long event, void *ptr) | |
3422 | { | |
3423 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
3424 | struct tun_struct *tun = netdev_priv(dev); | |
3425 | ||
86dfb4ac CG |
3426 | if (dev->rtnl_link_ops != &tun_link_ops) |
3427 | return NOTIFY_DONE; | |
3428 | ||
1576d986 JW |
3429 | switch (event) { |
3430 | case NETDEV_CHANGE_TX_QUEUE_LEN: | |
3431 | if (tun_queue_resize(tun)) | |
3432 | return NOTIFY_BAD; | |
3433 | break; | |
3434 | default: | |
3435 | break; | |
3436 | } | |
3437 | ||
3438 | return NOTIFY_DONE; | |
3439 | } | |
3440 | ||
3441 | static struct notifier_block tun_notifier_block __read_mostly = { | |
3442 | .notifier_call = tun_device_event, | |
3443 | }; | |
79d17604 | 3444 | |
1da177e4 LT |
3445 | static int __init tun_init(void) |
3446 | { | |
3447 | int ret = 0; | |
3448 | ||
6b8a66ee | 3449 | pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
1da177e4 | 3450 | |
f019a7a5 | 3451 | ret = rtnl_link_register(&tun_link_ops); |
79d17604 | 3452 | if (ret) { |
6b8a66ee | 3453 | pr_err("Can't register link_ops\n"); |
f019a7a5 | 3454 | goto err_linkops; |
79d17604 PE |
3455 | } |
3456 | ||
1da177e4 | 3457 | ret = misc_register(&tun_miscdev); |
79d17604 | 3458 | if (ret) { |
6b8a66ee | 3459 | pr_err("Can't register misc device %d\n", TUN_MINOR); |
79d17604 PE |
3460 | goto err_misc; |
3461 | } | |
1576d986 | 3462 | |
5edfbd3c TZ |
3463 | ret = register_netdevice_notifier(&tun_notifier_block); |
3464 | if (ret) { | |
3465 | pr_err("Can't register netdevice notifier\n"); | |
3466 | goto err_notifier; | |
3467 | } | |
3468 | ||
f019a7a5 | 3469 | return 0; |
5edfbd3c TZ |
3470 | |
3471 | err_notifier: | |
3472 | misc_deregister(&tun_miscdev); | |
79d17604 | 3473 | err_misc: |
f019a7a5 EB |
3474 | rtnl_link_unregister(&tun_link_ops); |
3475 | err_linkops: | |
1da177e4 LT |
3476 | return ret; |
3477 | } | |
3478 | ||
3479 | static void tun_cleanup(void) | |
3480 | { | |
6aa20a22 | 3481 | misc_deregister(&tun_miscdev); |
f019a7a5 | 3482 | rtnl_link_unregister(&tun_link_ops); |
1576d986 | 3483 | unregister_netdevice_notifier(&tun_notifier_block); |
1da177e4 LT |
3484 | } |
3485 | ||
05c2828c MT |
3486 | /* Get an underlying socket object from tun file. Returns error unless file is |
3487 | * attached to a device. The returned object works like a packet socket, it | |
3488 | * can be used for sock_sendmsg/sock_recvmsg. The caller is responsible for | |
3489 | * holding a reference to the file for as long as the socket is in use. */ | |
3490 | struct socket *tun_get_socket(struct file *file) | |
3491 | { | |
6e914fc7 | 3492 | struct tun_file *tfile; |
05c2828c MT |
3493 | if (file->f_op != &tun_fops) |
3494 | return ERR_PTR(-EINVAL); | |
6e914fc7 JW |
3495 | tfile = file->private_data; |
3496 | if (!tfile) | |
05c2828c | 3497 | return ERR_PTR(-EBADFD); |
54f968d6 | 3498 | return &tfile->socket; |
05c2828c MT |
3499 | } |
3500 | EXPORT_SYMBOL_GPL(tun_get_socket); | |
3501 | ||
5990a305 | 3502 | struct ptr_ring *tun_get_tx_ring(struct file *file) |
83339c6b JW |
3503 | { |
3504 | struct tun_file *tfile; | |
3505 | ||
3506 | if (file->f_op != &tun_fops) | |
3507 | return ERR_PTR(-EINVAL); | |
3508 | tfile = file->private_data; | |
3509 | if (!tfile) | |
3510 | return ERR_PTR(-EBADFD); | |
5990a305 | 3511 | return &tfile->tx_ring; |
83339c6b | 3512 | } |
5990a305 | 3513 | EXPORT_SYMBOL_GPL(tun_get_tx_ring); |
83339c6b | 3514 | |
1da177e4 LT |
3515 | module_init(tun_init); |
3516 | module_exit(tun_cleanup); | |
3517 | MODULE_DESCRIPTION(DRV_DESCRIPTION); | |
3518 | MODULE_AUTHOR(DRV_COPYRIGHT); | |
3519 | MODULE_LICENSE("GPL"); | |
3520 | MODULE_ALIAS_MISCDEV(TUN_MINOR); | |
578454ff | 3521 | MODULE_ALIAS("devname:net/tun"); |