tun.c 85.6 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-or-later
L
Linus Torvalds 已提交
2 3 4 5 6 7 8 9 10 11
/*
 *  TUN - Universal TUN/TAP device driver.
 *  Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
 *
 *  $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
 */

/*
 *  Changes:
 *
12 13 14
 *  Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
 *    Add TUNSETLINK ioctl to set the link encapsulation
 *
L
Linus Torvalds 已提交
15
 *  Mark Smith <markzzzsmith@yahoo.com.au>
J
Joe Perches 已提交
16
 *    Use eth_random_addr() for tap MAC address.
L
Linus Torvalds 已提交
17 18 19 20 21 22 23 24 25 26 27
 *
 *  Harald Roelle <harald.roelle@ifi.lmu.de>  2004/04/20
 *    Fixes in packet dropping, queue length setting and queue wakeup.
 *    Increased default tx queue length.
 *    Added ethtool API.
 *    Minor cleanups
 *
 *  Daniel Podlejski <underley@underley.eu.org>
 *    Modifications for 2.3.99-pre5 kernel.
 */

28 29
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

L
Linus Torvalds 已提交
30 31 32 33 34 35 36 37
#define DRV_NAME	"tun"
#define DRV_VERSION	"1.6"
#define DRV_DESCRIPTION	"Universal TUN/TAP device driver"
#define DRV_COPYRIGHT	"(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"

#include <linux/module.h>
#include <linux/errno.h>
#include <linux/kernel.h>
38
#include <linux/sched/signal.h>
L
Linus Torvalds 已提交
39 40 41 42 43 44 45 46 47 48 49
#include <linux/major.h>
#include <linux/slab.h>
#include <linux/poll.h>
#include <linux/fcntl.h>
#include <linux/init.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/miscdevice.h>
#include <linux/ethtool.h>
#include <linux/rtnetlink.h>
50
#include <linux/compat.h>
L
Linus Torvalds 已提交
51 52 53 54
#include <linux/if.h>
#include <linux/if_arp.h>
#include <linux/if_ether.h>
#include <linux/if_tun.h>
J
Jason Wang 已提交
55
#include <linux/if_vlan.h>
L
Linus Torvalds 已提交
56
#include <linux/crc32.h>
57
#include <linux/nsproxy.h>
58
#include <linux/virtio_net.h>
M
Michael S. Tsirkin 已提交
59
#include <linux/rcupdate.h>
60
#include <net/net_namespace.h>
61
#include <net/netns/generic.h>
62
#include <net/rtnetlink.h>
63
#include <net/sock.h>
64
#include <net/xdp.h>
65
#include <linux/seq_file.h>
H
Herbert Xu 已提交
66
#include <linux/uio.h>
J
Jason Wang 已提交
67
#include <linux/skb_array.h>
J
Jason Wang 已提交
68 69
#include <linux/bpf.h>
#include <linux/bpf_trace.h>
70
#include <linux/mutex.h>
L
Linus Torvalds 已提交
71

72
#include <linux/uaccess.h>
73
#include <linux/proc_fs.h>
L
Linus Torvalds 已提交
74

75 76 77
static void tun_default_link_ksettings(struct net_device *dev,
				       struct ethtool_link_ksettings *cmd);

78
#define TUN_RX_PAD (NET_IP_ALIGN + NET_SKB_PAD)
79

80 81 82 83 84 85
/* TUN device flags */

/* IFF_ATTACH_QUEUE is never stored in device flags,
 * overload it to mean fasync when stored there.
 */
#define TUN_FASYNC	IFF_ATTACH_QUEUE
M
Michael S. Tsirkin 已提交
86 87
/* High bits in flags field are unused. */
#define TUN_VNET_LE     0x80000000
88
#define TUN_VNET_BE     0x40000000
89 90

#define TUN_FEATURES (IFF_NO_PI | IFF_ONE_QUEUE | IFF_VNET_HDR | \
91 92
		      IFF_MULTI_QUEUE | IFF_NAPI | IFF_NAPI_FRAGS)

93 94
#define GOODCOPY_LEN 128

95 96 97 98 99 100 101
#define FLT_EXACT_COUNT 8
struct tap_filter {
	unsigned int    count;    /* Number of addrs. Zero means disabled */
	u32             mask[2];  /* Mask of the hashed addrs */
	unsigned char	addr[FLT_EXACT_COUNT][ETH_ALEN];
};

102 103 104
/* MAX_TAP_QUEUES 256 is chosen to allow rx/tx queues to be equal
 * to max number of VCPUs in guest. */
#define MAX_TAP_QUEUES 256
105
#define MAX_TAP_FLOWS  4096
J
Jason Wang 已提交
106

J
Jason Wang 已提交
107 108
#define TUN_FLOW_EXPIRE (3 * HZ)

109
struct tun_pcpu_stats {
E
Eric Dumazet 已提交
110 111 112 113
	u64_stats_t rx_packets;
	u64_stats_t rx_bytes;
	u64_stats_t tx_packets;
	u64_stats_t tx_bytes;
114 115 116 117 118 119
	struct u64_stats_sync syncp;
	u32 rx_dropped;
	u32 tx_dropped;
	u32 rx_frame_errors;
};

J
Jason Wang 已提交
120
/* A tun_file connects an open character device to a tuntap netdevice. It
S
stephen hemminger 已提交
121
 * also contains all socket related structures (except sock_fprog and tap_filter)
J
Jason Wang 已提交
122 123
 * to serve as one transmit queue for tuntap device. The sock_fprog and
 * tap_filter were kept in tun_struct since they were used for filtering for the
R
Rami Rosen 已提交
124
 * netdevice not for a specific queue (at least I didn't see the requirement for
J
Jason Wang 已提交
125
 * this).
126 127
 *
 * RCU usage:
R
Rami Rosen 已提交
128
 * The tun_file and tun_struct are loosely coupled, the pointer from one to the
129
 * other can only be read while rcu_read_lock or rtnl_lock is held.
J
Jason Wang 已提交
130
 */
E
Eric W. Biederman 已提交
131
struct tun_file {
J
Jason Wang 已提交
132 133
	struct sock sk;
	struct socket socket;
134
	struct tun_struct __rcu *tun;
J
Jason Wang 已提交
135 136 137
	struct fasync_struct *fasync;
	/* only used for fasnyc */
	unsigned int flags;
138 139 140 141
	union {
		u16 queue_index;
		unsigned int ifindex;
	};
142
	struct napi_struct napi;
143
	bool napi_enabled;
E
Eric Dumazet 已提交
144
	bool napi_frags_enabled;
145
	struct mutex napi_mutex;	/* Protects access to the above napi */
J
Jason Wang 已提交
146 147
	struct list_head next;
	struct tun_struct *detached;
148
	struct ptr_ring tx_ring;
J
Jesper Dangaard Brouer 已提交
149
	struct xdp_rxq_info xdp_rxq;
E
Eric W. Biederman 已提交
150 151
};

152 153 154 155 156
struct tun_page {
	struct page *page;
	int count;
};

J
Jason Wang 已提交
157 158 159 160 161 162
struct tun_flow_entry {
	struct hlist_node hash_link;
	struct rcu_head rcu;
	struct tun_struct *tun;

	u32 rxhash;
163
	u32 rps_rxhash;
J
Jason Wang 已提交
164
	int queue_index;
165
	unsigned long updated ____cacheline_aligned_in_smp;
J
Jason Wang 已提交
166 167 168
};

#define TUN_NUM_FLOW_ENTRIES 1024
L
Li RongQing 已提交
169
#define TUN_MASK_FLOW_ENTRIES (TUN_NUM_FLOW_ENTRIES - 1)
J
Jason Wang 已提交
170

171
struct tun_prog {
172 173 174 175
	struct rcu_head rcu;
	struct bpf_prog *prog;
};

J
Jason Wang 已提交
176
/* Since the socket were moved to tun_file, to preserve the behavior of persist
R
Rami Rosen 已提交
177
 * device, socket filter, sndbuf and vnet header size were restore when the
J
Jason Wang 已提交
178 179
 * file were attached to a persist device.
 */
180
struct tun_struct {
J
Jason Wang 已提交
181 182
	struct tun_file __rcu	*tfiles[MAX_TAP_QUEUES];
	unsigned int            numqueues;
183
	unsigned int 		flags;
184 185
	kuid_t			owner;
	kgid_t			group;
186 187

	struct net_device	*dev;
188
	netdev_features_t	set_features;
189
#define TUN_USER_FEATURES (NETIF_F_HW_CSUM|NETIF_F_TSO_ECN|NETIF_F_TSO| \
190
			  NETIF_F_TSO6)
191

192
	int			align;
193
	int			vnet_hdr_sz;
J
Jason Wang 已提交
194 195 196 197 198
	int			sndbuf;
	struct tap_filter	txflt;
	struct sock_fprog	fprog;
	/* protected by rtnl lock */
	bool			filter_attached;
199
	u32			msg_enable;
J
Jason Wang 已提交
200 201 202 203
	spinlock_t lock;
	struct hlist_head flows[TUN_NUM_FLOW_ENTRIES];
	struct timer_list flow_gc_timer;
	unsigned long ageing_time;
J
Jason Wang 已提交
204 205
	unsigned int numdisabled;
	struct list_head disabled;
206
	void *security;
207
	u32 flow_count;
J
Jason Wang 已提交
208
	u32 rx_batched;
209
	struct tun_pcpu_stats __percpu *pcpu_stats;
J
Jason Wang 已提交
210
	struct bpf_prog __rcu *xdp_prog;
211
	struct tun_prog __rcu *steering_prog;
212
	struct tun_prog __rcu *filter_prog;
213
	struct ethtool_link_ksettings link_ksettings;
214
};
L
Linus Torvalds 已提交
215

216 217 218
struct veth {
	__be16 h_vlan_proto;
	__be16 h_vlan_TCI;
219
};
L
Linus Torvalds 已提交
220

221
bool tun_is_xdp_frame(void *ptr)
J
Jason Wang 已提交
222 223 224
{
	return (unsigned long)ptr & TUN_XDP_FLAG;
}
225
EXPORT_SYMBOL(tun_is_xdp_frame);
J
Jason Wang 已提交
226 227 228 229 230 231 232 233 234 235 236 237 238

void *tun_xdp_to_ptr(void *ptr)
{
	return (void *)((unsigned long)ptr | TUN_XDP_FLAG);
}
EXPORT_SYMBOL(tun_xdp_to_ptr);

void *tun_ptr_to_xdp(void *ptr)
{
	return (void *)((unsigned long)ptr & ~TUN_XDP_FLAG);
}
EXPORT_SYMBOL(tun_ptr_to_xdp);

239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
static int tun_napi_receive(struct napi_struct *napi, int budget)
{
	struct tun_file *tfile = container_of(napi, struct tun_file, napi);
	struct sk_buff_head *queue = &tfile->sk.sk_write_queue;
	struct sk_buff_head process_queue;
	struct sk_buff *skb;
	int received = 0;

	__skb_queue_head_init(&process_queue);

	spin_lock(&queue->lock);
	skb_queue_splice_tail_init(queue, &process_queue);
	spin_unlock(&queue->lock);

	while (received < budget && (skb = __skb_dequeue(&process_queue))) {
		napi_gro_receive(napi, skb);
		++received;
	}

	if (!skb_queue_empty(&process_queue)) {
		spin_lock(&queue->lock);
		skb_queue_splice(&process_queue, queue);
		spin_unlock(&queue->lock);
	}

	return received;
}

static int tun_napi_poll(struct napi_struct *napi, int budget)
{
	unsigned int received;

	received = tun_napi_receive(napi, budget);

	if (received < budget)
		napi_complete_done(napi, received);

	return received;
}

static void tun_napi_init(struct tun_struct *tun, struct tun_file *tfile,
E
Eric Dumazet 已提交
280
			  bool napi_en, bool napi_frags)
281
{
282
	tfile->napi_enabled = napi_en;
E
Eric Dumazet 已提交
283
	tfile->napi_frags_enabled = napi_en && napi_frags;
284
	if (napi_en) {
285 286
		netif_tx_napi_add(tun->dev, &tfile->napi, tun_napi_poll,
				  NAPI_POLL_WEIGHT);
287 288 289 290
		napi_enable(&tfile->napi);
	}
}

E
Eric Dumazet 已提交
291
static void tun_napi_disable(struct tun_file *tfile)
292
{
293
	if (tfile->napi_enabled)
294 295 296
		napi_disable(&tfile->napi);
}

E
Eric Dumazet 已提交
297
static void tun_napi_del(struct tun_file *tfile)
298
{
299
	if (tfile->napi_enabled)
300 301 302
		netif_napi_del(&tfile->napi);
}

E
Eric Dumazet 已提交
303
static bool tun_napi_frags_enabled(const struct tun_file *tfile)
304
{
E
Eric Dumazet 已提交
305
	return tfile->napi_frags_enabled;
306 307
}

308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
#ifdef CONFIG_TUN_VNET_CROSS_LE
static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
{
	return tun->flags & TUN_VNET_BE ? false :
		virtio_legacy_is_little_endian();
}

static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
{
	int be = !!(tun->flags & TUN_VNET_BE);

	if (put_user(be, argp))
		return -EFAULT;

	return 0;
}

static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
{
	int be;

	if (get_user(be, argp))
		return -EFAULT;

	if (be)
		tun->flags |= TUN_VNET_BE;
	else
		tun->flags &= ~TUN_VNET_BE;

	return 0;
}
#else
static inline bool tun_legacy_is_little_endian(struct tun_struct *tun)
{
	return virtio_legacy_is_little_endian();
}

static long tun_get_vnet_be(struct tun_struct *tun, int __user *argp)
{
	return -EINVAL;
}

static long tun_set_vnet_be(struct tun_struct *tun, int __user *argp)
{
	return -EINVAL;
}
#endif /* CONFIG_TUN_VNET_CROSS_LE */

356 357
static inline bool tun_is_little_endian(struct tun_struct *tun)
{
358
	return tun->flags & TUN_VNET_LE ||
359
		tun_legacy_is_little_endian(tun);
360 361
}

362 363
static inline u16 tun16_to_cpu(struct tun_struct *tun, __virtio16 val)
{
364
	return __virtio16_to_cpu(tun_is_little_endian(tun), val);
365 366 367 368
}

static inline __virtio16 cpu_to_tun16(struct tun_struct *tun, u16 val)
{
369
	return __cpu_to_virtio16(tun_is_little_endian(tun), val);
370 371
}

J
Jason Wang 已提交
372 373
static inline u32 tun_hashfn(u32 rxhash)
{
L
Li RongQing 已提交
374
	return rxhash & TUN_MASK_FLOW_ENTRIES;
J
Jason Wang 已提交
375 376 377 378 379 380
}

static struct tun_flow_entry *tun_flow_find(struct hlist_head *head, u32 rxhash)
{
	struct tun_flow_entry *e;

381
	hlist_for_each_entry_rcu(e, head, hash_link) {
J
Jason Wang 已提交
382 383 384 385 386 387 388 389 390 391
		if (e->rxhash == rxhash)
			return e;
	}
	return NULL;
}

static struct tun_flow_entry *tun_flow_create(struct tun_struct *tun,
					      struct hlist_head *head,
					      u32 rxhash, u16 queue_index)
{
392 393
	struct tun_flow_entry *e = kmalloc(sizeof(*e), GFP_ATOMIC);

J
Jason Wang 已提交
394
	if (e) {
395 396 397
		netif_info(tun, tx_queued, tun->dev,
			   "create flow: hash %u index %u\n",
			   rxhash, queue_index);
J
Jason Wang 已提交
398 399
		e->updated = jiffies;
		e->rxhash = rxhash;
400
		e->rps_rxhash = 0;
J
Jason Wang 已提交
401 402 403
		e->queue_index = queue_index;
		e->tun = tun;
		hlist_add_head_rcu(&e->hash_link, head);
404
		++tun->flow_count;
J
Jason Wang 已提交
405 406 407 408 409 410
	}
	return e;
}

static void tun_flow_delete(struct tun_struct *tun, struct tun_flow_entry *e)
{
411 412
	netif_info(tun, tx_queued, tun->dev, "delete flow: hash %u index %u\n",
		   e->rxhash, e->queue_index);
J
Jason Wang 已提交
413
	hlist_del_rcu(&e->hash_link);
414
	kfree_rcu(e, rcu);
415
	--tun->flow_count;
J
Jason Wang 已提交
416 417 418 419 420 421 422 423 424
}

static void tun_flow_flush(struct tun_struct *tun)
{
	int i;

	spin_lock_bh(&tun->lock);
	for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
		struct tun_flow_entry *e;
425
		struct hlist_node *n;
J
Jason Wang 已提交
426

427
		hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link)
J
Jason Wang 已提交
428 429 430 431 432 433 434 435 436 437 438 439
			tun_flow_delete(tun, e);
	}
	spin_unlock_bh(&tun->lock);
}

static void tun_flow_delete_by_queue(struct tun_struct *tun, u16 queue_index)
{
	int i;

	spin_lock_bh(&tun->lock);
	for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
		struct tun_flow_entry *e;
440
		struct hlist_node *n;
J
Jason Wang 已提交
441

442
		hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
J
Jason Wang 已提交
443 444 445 446 447 448 449
			if (e->queue_index == queue_index)
				tun_flow_delete(tun, e);
		}
	}
	spin_unlock_bh(&tun->lock);
}

450
static void tun_flow_cleanup(struct timer_list *t)
J
Jason Wang 已提交
451
{
452
	struct tun_struct *tun = from_timer(tun, t, flow_gc_timer);
J
Jason Wang 已提交
453 454 455 456 457
	unsigned long delay = tun->ageing_time;
	unsigned long next_timer = jiffies + delay;
	unsigned long count = 0;
	int i;

458
	spin_lock(&tun->lock);
J
Jason Wang 已提交
459 460
	for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++) {
		struct tun_flow_entry *e;
461
		struct hlist_node *n;
J
Jason Wang 已提交
462

463
		hlist_for_each_entry_safe(e, n, &tun->flows[i], hash_link) {
J
Jason Wang 已提交
464
			unsigned long this_timer;
465

J
Jason Wang 已提交
466
			this_timer = e->updated + delay;
467
			if (time_before_eq(this_timer, jiffies)) {
J
Jason Wang 已提交
468
				tun_flow_delete(tun, e);
469 470 471 472
				continue;
			}
			count++;
			if (time_before(this_timer, next_timer))
J
Jason Wang 已提交
473 474 475 476 477 478
				next_timer = this_timer;
		}
	}

	if (count)
		mod_timer(&tun->flow_gc_timer, round_jiffies_up(next_timer));
479
	spin_unlock(&tun->lock);
J
Jason Wang 已提交
480 481
}

482
static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
483
			    struct tun_file *tfile)
J
Jason Wang 已提交
484 485 486 487
{
	struct hlist_head *head;
	struct tun_flow_entry *e;
	unsigned long delay = tun->ageing_time;
488
	u16 queue_index = tfile->queue_index;
J
Jason Wang 已提交
489

490
	head = &tun->flows[tun_hashfn(rxhash)];
J
Jason Wang 已提交
491 492 493 494 495 496

	rcu_read_lock();

	e = tun_flow_find(head, rxhash);
	if (likely(e)) {
		/* TODO: keep queueing to old queue until it's empty? */
497 498
		if (READ_ONCE(e->queue_index) != queue_index)
			WRITE_ONCE(e->queue_index, queue_index);
499 500
		if (e->updated != jiffies)
			e->updated = jiffies;
501
		sock_rps_record_flow_hash(e->rps_rxhash);
J
Jason Wang 已提交
502 503
	} else {
		spin_lock_bh(&tun->lock);
504 505
		if (!tun_flow_find(head, rxhash) &&
		    tun->flow_count < MAX_TAP_FLOWS)
J
Jason Wang 已提交
506 507 508 509 510 511 512 513 514 515 516
			tun_flow_create(tun, head, rxhash, queue_index);

		if (!timer_pending(&tun->flow_gc_timer))
			mod_timer(&tun->flow_gc_timer,
				  round_jiffies_up(jiffies + delay));
		spin_unlock_bh(&tun->lock);
	}

	rcu_read_unlock();
}

517
/* Save the hash received in the stack receive path and update the
518 519 520 521
 * flow_hash table accordingly.
 */
static inline void tun_flow_save_rps_rxhash(struct tun_flow_entry *e, u32 hash)
{
522
	if (unlikely(e->rps_rxhash != hash))
523 524 525
		e->rps_rxhash = hash;
}

526
/* We try to identify a flow through its rxhash. The reason that
S
stephen hemminger 已提交
527
 * we do not check rxq no. is because some cards(e.g 82599), chooses
J
Jason Wang 已提交
528 529
 * the rxq based on the txq where the last packet of the flow comes. As
 * the userspace application move between processors, we may get a
530
 * different rxq no. here.
J
Jason Wang 已提交
531
 */
532
static u16 tun_automq_select_queue(struct tun_struct *tun, struct sk_buff *skb)
J
Jason Wang 已提交
533
{
J
Jason Wang 已提交
534
	struct tun_flow_entry *e;
J
Jason Wang 已提交
535 536 537
	u32 txq = 0;
	u32 numqueues = 0;

538
	numqueues = READ_ONCE(tun->numqueues);
J
Jason Wang 已提交
539

J
Jason Wang 已提交
540
	txq = __skb_get_hash_symmetric(skb);
541 542 543 544 545 546 547
	e = tun_flow_find(&tun->flows[tun_hashfn(txq)], txq);
	if (e) {
		tun_flow_save_rps_rxhash(e, txq);
		txq = e->queue_index;
	} else {
		/* use multiply and shift instead of expensive divide */
		txq = ((u64)txq * numqueues) >> 32;
J
Jason Wang 已提交
548 549 550 551 552
	}

	return txq;
}

553 554
static u16 tun_ebpf_select_queue(struct tun_struct *tun, struct sk_buff *skb)
{
555
	struct tun_prog *prog;
556
	u32 numqueues;
557 558
	u16 ret = 0;

559 560 561 562
	numqueues = READ_ONCE(tun->numqueues);
	if (!numqueues)
		return 0;

563 564 565 566
	prog = rcu_dereference(tun->steering_prog);
	if (prog)
		ret = bpf_prog_run_clear_cb(prog->prog, skb);

567
	return ret % numqueues;
568 569 570
}

static u16 tun_select_queue(struct net_device *dev, struct sk_buff *skb,
571
			    struct net_device *sb_dev)
572 573 574 575 576 577 578 579 580 581 582 583 584 585
{
	struct tun_struct *tun = netdev_priv(dev);
	u16 ret;

	rcu_read_lock();
	if (rcu_dereference(tun->steering_prog))
		ret = tun_ebpf_select_queue(tun, skb);
	else
		ret = tun_automq_select_queue(tun, skb);
	rcu_read_unlock();

	return ret;
}

586 587 588
static inline bool tun_not_capable(struct tun_struct *tun)
{
	const struct cred *cred = current_cred();
589
	struct net *net = dev_net(tun->dev);
590 591 592

	return ((uid_valid(tun->owner) && !uid_eq(cred->euid, tun->owner)) ||
		  (gid_valid(tun->group) && !in_egroup_p(tun->group))) &&
593
		!ns_capable(net->user_ns, CAP_NET_ADMIN);
594 595
}

J
Jason Wang 已提交
596 597 598 599 600 601
static void tun_set_real_num_queues(struct tun_struct *tun)
{
	netif_set_real_num_tx_queues(tun->dev, tun->numqueues);
	netif_set_real_num_rx_queues(tun->dev, tun->numqueues);
}

J
Jason Wang 已提交
602 603 604 605 606 607 608
static void tun_disable_queue(struct tun_struct *tun, struct tun_file *tfile)
{
	tfile->detached = tun;
	list_add_tail(&tfile->next, &tun->disabled);
	++tun->numdisabled;
}

J
Jason Wang 已提交
609
static struct tun_struct *tun_enable_queue(struct tun_file *tfile)
J
Jason Wang 已提交
610 611 612 613 614 615 616 617 618
{
	struct tun_struct *tun = tfile->detached;

	tfile->detached = NULL;
	list_del_init(&tfile->next);
	--tun->numdisabled;
	return tun;
}

619
void tun_ptr_free(void *ptr)
J
Jason Wang 已提交
620 621 622
{
	if (!ptr)
		return;
623 624
	if (tun_is_xdp_frame(ptr)) {
		struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr);
J
Jason Wang 已提交
625

626
		xdp_return_frame(xdpf);
J
Jason Wang 已提交
627 628 629 630
	} else {
		__skb_array_destroy_skb(ptr);
	}
}
631
EXPORT_SYMBOL_GPL(tun_ptr_free);
J
Jason Wang 已提交
632

633 634
static void tun_queue_purge(struct tun_file *tfile)
{
J
Jason Wang 已提交
635
	void *ptr;
J
Jason Wang 已提交
636

J
Jason Wang 已提交
637 638
	while ((ptr = ptr_ring_consume(&tfile->tx_ring)) != NULL)
		tun_ptr_free(ptr);
J
Jason Wang 已提交
639

J
Jason Wang 已提交
640
	skb_queue_purge(&tfile->sk.sk_write_queue);
641 642 643
	skb_queue_purge(&tfile->sk.sk_error_queue);
}

J
Jason Wang 已提交
644 645 646 647 648
static void __tun_detach(struct tun_file *tfile, bool clean)
{
	struct tun_file *ntfile;
	struct tun_struct *tun;

649 650
	tun = rtnl_dereference(tfile->tun);

651
	if (tun && clean) {
E
Eric Dumazet 已提交
652 653
		tun_napi_disable(tfile);
		tun_napi_del(tfile);
654 655
	}

656
	if (tun && !tfile->detached) {
J
Jason Wang 已提交
657 658 659 660 661
		u16 index = tfile->queue_index;
		BUG_ON(index >= tun->numqueues);

		rcu_assign_pointer(tun->tfiles[index],
				   tun->tfiles[tun->numqueues - 1]);
662
		ntfile = rtnl_dereference(tun->tfiles[index]);
J
Jason Wang 已提交
663
		ntfile->queue_index = index;
664 665
		rcu_assign_pointer(tun->tfiles[tun->numqueues - 1],
				   NULL);
J
Jason Wang 已提交
666 667

		--tun->numqueues;
668
		if (clean) {
669
			RCU_INIT_POINTER(tfile->tun, NULL);
J
Jason Wang 已提交
670
			sock_put(&tfile->sk);
671
		} else
J
Jason Wang 已提交
672
			tun_disable_queue(tun, tfile);
J
Jason Wang 已提交
673 674

		synchronize_net();
J
Jason Wang 已提交
675
		tun_flow_delete_by_queue(tun, tun->numqueues + 1);
J
Jason Wang 已提交
676
		/* Drop read queue */
677
		tun_queue_purge(tfile);
J
Jason Wang 已提交
678
		tun_set_real_num_queues(tun);
J
Jason Wang 已提交
679
	} else if (tfile->detached && clean) {
J
Jason Wang 已提交
680
		tun = tun_enable_queue(tfile);
J
Jason Wang 已提交
681 682
		sock_put(&tfile->sk);
	}
J
Jason Wang 已提交
683 684

	if (clean) {
685 686 687
		if (tun && tun->numqueues == 0 && tun->numdisabled == 0) {
			netif_carrier_off(tun->dev);

M
Michael S. Tsirkin 已提交
688
			if (!(tun->flags & IFF_PERSIST) &&
689
			    tun->dev->reg_state == NETREG_REGISTERED)
J
Jason Wang 已提交
690
				unregister_netdevice(tun->dev);
691
		}
692 693
		if (tun)
			xdp_rxq_info_unreg(&tfile->xdp_rxq);
694
		ptr_ring_cleanup(&tfile->tx_ring, tun_ptr_free);
695
		sock_put(&tfile->sk);
J
Jason Wang 已提交
696 697 698 699 700
	}
}

static void tun_detach(struct tun_file *tfile, bool clean)
{
701 702 703
	struct tun_struct *tun;
	struct net_device *dev;

J
Jason Wang 已提交
704
	rtnl_lock();
705 706
	tun = rtnl_dereference(tfile->tun);
	dev = tun ? tun->dev : NULL;
J
Jason Wang 已提交
707
	__tun_detach(tfile, clean);
708 709
	if (dev)
		netdev_state_change(dev);
J
Jason Wang 已提交
710 711 712 713 714 715
	rtnl_unlock();
}

static void tun_detach_all(struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);
J
Jason Wang 已提交
716
	struct tun_file *tfile, *tmp;
J
Jason Wang 已提交
717 718 719
	int i, n = tun->numqueues;

	for (i = 0; i < n; i++) {
720
		tfile = rtnl_dereference(tun->tfiles[i]);
J
Jason Wang 已提交
721
		BUG_ON(!tfile);
E
Eric Dumazet 已提交
722
		tun_napi_disable(tfile);
723
		tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN;
724
		tfile->socket.sk->sk_data_ready(tfile->socket.sk);
725
		RCU_INIT_POINTER(tfile->tun, NULL);
J
Jason Wang 已提交
726 727
		--tun->numqueues;
	}
728
	list_for_each_entry(tfile, &tun->disabled, next) {
729
		tfile->socket.sk->sk_shutdown = RCV_SHUTDOWN;
730
		tfile->socket.sk->sk_data_ready(tfile->socket.sk);
731
		RCU_INIT_POINTER(tfile->tun, NULL);
732
	}
J
Jason Wang 已提交
733 734 735 736
	BUG_ON(tun->numqueues != 0);

	synchronize_net();
	for (i = 0; i < n; i++) {
737
		tfile = rtnl_dereference(tun->tfiles[i]);
E
Eric Dumazet 已提交
738
		tun_napi_del(tfile);
J
Jason Wang 已提交
739
		/* Drop read queue */
740
		tun_queue_purge(tfile);
741
		xdp_rxq_info_unreg(&tfile->xdp_rxq);
J
Jason Wang 已提交
742 743
		sock_put(&tfile->sk);
	}
J
Jason Wang 已提交
744 745
	list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
		tun_enable_queue(tfile);
746
		tun_queue_purge(tfile);
747
		xdp_rxq_info_unreg(&tfile->xdp_rxq);
J
Jason Wang 已提交
748 749 750
		sock_put(&tfile->sk);
	}
	BUG_ON(tun->numdisabled != 0);
J
Jason Wang 已提交
751

M
Michael S. Tsirkin 已提交
752
	if (tun->flags & IFF_PERSIST)
J
Jason Wang 已提交
753
		module_put(THIS_MODULE);
J
Jason Wang 已提交
754 755
}

756
static int tun_attach(struct tun_struct *tun, struct file *file,
757 758
		      bool skip_filter, bool napi, bool napi_frags,
		      bool publish_tun)
759
{
E
Eric W. Biederman 已提交
760
	struct tun_file *tfile = file->private_data;
J
Jason Wang 已提交
761
	struct net_device *dev = tun->dev;
762
	int err;
763

764 765 766 767
	err = security_tun_dev_attach(tfile->socket.sk, tun->security);
	if (err < 0)
		goto out;

768
	err = -EINVAL;
769
	if (rtnl_dereference(tfile->tun) && !tfile->detached)
770 771 772
		goto out;

	err = -EBUSY;
M
Michael S. Tsirkin 已提交
773
	if (!(tun->flags & IFF_MULTI_QUEUE) && tun->numqueues == 1)
J
Jason Wang 已提交
774 775 776
		goto out;

	err = -E2BIG;
J
Jason Wang 已提交
777 778
	if (!tfile->detached &&
	    tun->numqueues + tun->numdisabled == MAX_TAP_QUEUES)
779 780 781
		goto out;

	err = 0;
J
Jason Wang 已提交
782

S
stephen hemminger 已提交
783
	/* Re-attach the filter to persist device */
784
	if (!skip_filter && (tun->filter_attached == true)) {
785 786 787
		lock_sock(tfile->socket.sk);
		err = sk_attach_filter(&tun->fprog, tfile->socket.sk);
		release_sock(tfile->socket.sk);
J
Jason Wang 已提交
788 789 790
		if (!err)
			goto out;
	}
J
Jason Wang 已提交
791 792

	if (!tfile->detached &&
793 794
	    ptr_ring_resize(&tfile->tx_ring, dev->tx_queue_len,
			    GFP_KERNEL, tun_ptr_free)) {
J
Jason Wang 已提交
795 796 797 798
		err = -ENOMEM;
		goto out;
	}

J
Jason Wang 已提交
799
	tfile->queue_index = tun->numqueues;
800
	tfile->socket.sk->sk_shutdown &= ~RCV_SHUTDOWN;
J
Jesper Dangaard Brouer 已提交
801 802 803 804 805 806 807 808 809 810 811 812 813

	if (tfile->detached) {
		/* Re-attach detached tfile, updating XDP queue_index */
		WARN_ON(!xdp_rxq_info_is_reg(&tfile->xdp_rxq));

		if (tfile->xdp_rxq.queue_index    != tfile->queue_index)
			tfile->xdp_rxq.queue_index = tfile->queue_index;
	} else {
		/* Setup XDP RX-queue info, for new tfile getting attached */
		err = xdp_rxq_info_reg(&tfile->xdp_rxq,
				       tun->dev, tfile->queue_index);
		if (err < 0)
			goto out;
814 815 816 817 818 819
		err = xdp_rxq_info_reg_mem_model(&tfile->xdp_rxq,
						 MEM_TYPE_PAGE_SHARED, NULL);
		if (err < 0) {
			xdp_rxq_info_unreg(&tfile->xdp_rxq);
			goto out;
		}
J
Jesper Dangaard Brouer 已提交
820 821 822
		err = 0;
	}

823
	if (tfile->detached) {
J
Jason Wang 已提交
824
		tun_enable_queue(tfile);
825
	} else {
J
Jason Wang 已提交
826
		sock_hold(&tfile->sk);
E
Eric Dumazet 已提交
827
		tun_napi_init(tun, tfile, napi, napi_frags);
828
	}
J
Jason Wang 已提交
829

J
Jason Wang 已提交
830 831 832
	if (rtnl_dereference(tun->xdp_prog))
		sock_set_flag(&tfile->sk, SOCK_XDP);

J
Jason Wang 已提交
833 834 835 836
	/* device is allowed to go away first, so no need to hold extra
	 * refcnt.
	 */

837 838 839 840
	/* Publish tfile->tun and tun->tfiles only after we've fully
	 * initialized tfile; otherwise we risk using half-initialized
	 * object.
	 */
841 842
	if (publish_tun)
		rcu_assign_pointer(tfile->tun, tun);
843 844
	rcu_assign_pointer(tun->tfiles[tun->numqueues], tfile);
	tun->numqueues++;
845
	tun_set_real_num_queues(tun);
J
Jason Wang 已提交
846 847
out:
	return err;
E
Eric W. Biederman 已提交
848 849
}

850
static struct tun_struct *tun_get(struct tun_file *tfile)
E
Eric W. Biederman 已提交
851
{
852
	struct tun_struct *tun;
853

854 855 856 857 858
	rcu_read_lock();
	tun = rcu_dereference(tfile->tun);
	if (tun)
		dev_hold(tun->dev);
	rcu_read_unlock();
859 860

	return tun;
E
Eric W. Biederman 已提交
861 862 863 864
}

static void tun_put(struct tun_struct *tun)
{
865
	dev_put(tun->dev);
E
Eric W. Biederman 已提交
866 867
}

868
/* TAP filtering */
869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896
static void addr_hash_set(u32 *mask, const u8 *addr)
{
	int n = ether_crc(ETH_ALEN, addr) >> 26;
	mask[n >> 5] |= (1 << (n & 31));
}

static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
{
	int n = ether_crc(ETH_ALEN, addr) >> 26;
	return mask[n >> 5] & (1 << (n & 31));
}

static int update_filter(struct tap_filter *filter, void __user *arg)
{
	struct { u8 u[ETH_ALEN]; } *addr;
	struct tun_filter uf;
	int err, alen, n, nexact;

	if (copy_from_user(&uf, arg, sizeof(uf)))
		return -EFAULT;

	if (!uf.count) {
		/* Disabled */
		filter->count = 0;
		return 0;
	}

	alen = ETH_ALEN * uf.count;
897 898 899
	addr = memdup_user(arg + sizeof(uf), alen);
	if (IS_ERR(addr))
		return PTR_ERR(addr);
900 901 902 903 904 905 906 907 908 909 910 911 912

	/* The filter is updated without holding any locks. Which is
	 * perfectly safe. We disable it first and in the worst
	 * case we'll accept a few undesired packets. */
	filter->count = 0;
	wmb();

	/* Use first set of addresses as an exact filter */
	for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
		memcpy(filter->addr[n], addr[n].u, ETH_ALEN);

	nexact = n;

913 914
	/* Remaining multicast addresses are hashed,
	 * unicast will leave the filter disabled. */
915
	memset(filter->mask, 0, sizeof(filter->mask));
916 917 918
	for (; n < uf.count; n++) {
		if (!is_multicast_ether_addr(addr[n].u)) {
			err = 0; /* no filter */
919
			goto free_addr;
920
		}
921
		addr_hash_set(filter->mask, addr[n].u);
922
	}
923 924 925 926 927 928 929 930 931 932 933 934

	/* For ALLMULTI just set the mask to all ones.
	 * This overrides the mask populated above. */
	if ((uf.flags & TUN_FLT_ALLMULTI))
		memset(filter->mask, ~0, sizeof(filter->mask));

	/* Now enable the filter */
	wmb();
	filter->count = nexact;

	/* Return the number of exact filters */
	err = nexact;
935
free_addr:
936 937 938 939 940 941 942 943 944 945 946 947 948 949
	kfree(addr);
	return err;
}

/* Returns: 0 - drop, !=0 - accept */
static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
{
	/* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
	 * at this point. */
	struct ethhdr *eh = (struct ethhdr *) skb->data;
	int i;

	/* Exact match */
	for (i = 0; i < filter->count; i++)
950
		if (ether_addr_equal(eh->h_dest, filter->addr[i]))
951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971
			return 1;

	/* Inexact match (multicast only) */
	if (is_multicast_ether_addr(eh->h_dest))
		return addr_hash_test(filter->mask, eh->h_dest);

	return 0;
}

/*
 * Checks whether the packet is accepted or not.
 * Returns: 0 - drop, !=0 - accept
 */
static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
{
	if (!filter->count)
		return 1;

	return run_filter(filter, skb);
}

L
Linus Torvalds 已提交
972 973
/* Network device part of the driver */

974
static const struct ethtool_ops tun_ethtool_ops;
L
Linus Torvalds 已提交
975

976 977 978
/* Net device detach from fd. */
static void tun_net_uninit(struct net_device *dev)
{
J
Jason Wang 已提交
979
	tun_detach_all(dev);
980 981
}

L
Linus Torvalds 已提交
982 983 984
/* Net device open. */
static int tun_net_open(struct net_device *dev)
{
J
Jason Wang 已提交
985
	netif_tx_start_all_queues(dev);
986

L
Linus Torvalds 已提交
987 988 989 990 991 992
	return 0;
}

/* Net device close. */
static int tun_net_close(struct net_device *dev)
{
J
Jason Wang 已提交
993
	netif_tx_stop_all_queues(dev);
L
Linus Torvalds 已提交
994 995 996 997
	return 0;
}

/* Net device start xmit */
998
static void tun_automq_xmit(struct tun_struct *tun, struct sk_buff *skb)
L
Linus Torvalds 已提交
999
{
1000
#ifdef CONFIG_RPS
1001
	if (tun->numqueues == 1 && static_branch_unlikely(&rps_needed)) {
1002 1003 1004
		/* Select queue was not called for the skbuff, so we extract the
		 * RPS hash and save it into the flow_table here.
		 */
1005
		struct tun_flow_entry *e;
1006 1007
		__u32 rxhash;

J
Jason Wang 已提交
1008
		rxhash = __skb_get_hash_symmetric(skb);
1009 1010 1011
		e = tun_flow_find(&tun->flows[tun_hashfn(rxhash)], rxhash);
		if (e)
			tun_flow_save_rps_rxhash(e, rxhash);
1012
	}
1013
#endif
1014 1015
}

1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
static unsigned int run_ebpf_filter(struct tun_struct *tun,
				    struct sk_buff *skb,
				    int len)
{
	struct tun_prog *prog = rcu_dereference(tun->filter_prog);

	if (prog)
		len = bpf_prog_run_clear_cb(prog->prog, skb);

	return len;
}

1028 1029 1030 1031 1032 1033
/* Net device start xmit */
static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);
	int txq = skb->queue_mapping;
	struct tun_file *tfile;
1034
	int len = skb->len;
1035 1036 1037 1038 1039

	rcu_read_lock();
	tfile = rcu_dereference(tun->tfiles[txq]);

	/* Drop packet if interface is not attached */
1040
	if (!tfile)
1041 1042 1043 1044
		goto drop;

	if (!rcu_dereference(tun->steering_prog))
		tun_automq_xmit(tun, skb);
1045

1046
	netif_info(tun, tx_queued, tun->dev, "%s %d\n", __func__, skb->len);
1047

1048 1049 1050 1051 1052 1053
	/* Drop if the filter does not like it.
	 * This is a noop if the filter is disabled.
	 * Filter can be enabled only for the TAP devices. */
	if (!check_filter(&tun->txflt, skb))
		goto drop;

J
Jason Wang 已提交
1054 1055
	if (tfile->socket.sk->sk_filter &&
	    sk_filter(tfile->socket.sk, skb))
M
Michael S. Tsirkin 已提交
1056 1057
		goto drop;

1058
	len = run_ebpf_filter(tun, skb, len);
B
Bjørn Mork 已提交
1059
	if (len == 0 || pskb_trim(skb, len))
1060 1061
		goto drop;

W
Willem de Bruijn 已提交
1062
	if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC)))
1063 1064
		goto drop;

1065
	skb_tx_timestamp(skb);
1066

M
Michael S. Tsirkin 已提交
1067
	/* Orphan the skb - required as we might hang on to it
1068 1069
	 * for indefinite time.
	 */
M
Michael S. Tsirkin 已提交
1070 1071
	skb_orphan(skb);

1072
	nf_reset_ct(skb);
1073

1074
	if (ptr_ring_produce(&tfile->tx_ring, skb))
J
Jason Wang 已提交
1075
		goto drop;
L
Linus Torvalds 已提交
1076 1077

	/* Notify and wake up reader process */
J
Jason Wang 已提交
1078 1079
	if (tfile->flags & TUN_FASYNC)
		kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
1080
	tfile->socket.sk->sk_data_ready(tfile->socket.sk);
1081 1082

	rcu_read_unlock();
1083
	return NETDEV_TX_OK;
L
Linus Torvalds 已提交
1084 1085

drop:
1086
	this_cpu_inc(tun->pcpu_stats->tx_dropped);
1087
	skb_tx_error(skb);
L
Linus Torvalds 已提交
1088
	kfree_skb(skb);
1089
	rcu_read_unlock();
1090
	return NET_XMIT_DROP;
L
Linus Torvalds 已提交
1091 1092
}

1093
static void tun_net_mclist(struct net_device *dev)
L
Linus Torvalds 已提交
1094
{
1095 1096 1097 1098 1099
	/*
	 * This callback is supposed to deal with mc filter in
	 * _rx_ path and has nothing to do with the _tx_ path.
	 * In rx path we always accept everything userspace gives us.
	 */
L
Linus Torvalds 已提交
1100 1101
}

1102 1103
static netdev_features_t tun_net_fix_features(struct net_device *dev,
	netdev_features_t features)
1104 1105 1106 1107 1108
{
	struct tun_struct *tun = netdev_priv(dev);

	return (features & tun->set_features) | (features & ~TUN_USER_FEATURES);
}
1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119

static void tun_set_headroom(struct net_device *dev, int new_hr)
{
	struct tun_struct *tun = netdev_priv(dev);

	if (new_hr < NET_SKB_PAD)
		new_hr = NET_SKB_PAD;

	tun->align = new_hr;
}

1120
static void
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats)
{
	u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0;
	struct tun_struct *tun = netdev_priv(dev);
	struct tun_pcpu_stats *p;
	int i;

	for_each_possible_cpu(i) {
		u64 rxpackets, rxbytes, txpackets, txbytes;
		unsigned int start;

		p = per_cpu_ptr(tun->pcpu_stats, i);
		do {
			start = u64_stats_fetch_begin(&p->syncp);
E
Eric Dumazet 已提交
1135 1136 1137 1138
			rxpackets	= u64_stats_read(&p->rx_packets);
			rxbytes		= u64_stats_read(&p->rx_bytes);
			txpackets	= u64_stats_read(&p->tx_packets);
			txbytes		= u64_stats_read(&p->tx_bytes);
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
		} while (u64_stats_fetch_retry(&p->syncp, start));

		stats->rx_packets	+= rxpackets;
		stats->rx_bytes		+= rxbytes;
		stats->tx_packets	+= txpackets;
		stats->tx_bytes		+= txbytes;

		/* u32 counters */
		rx_dropped	+= p->rx_dropped;
		rx_frame_errors	+= p->rx_frame_errors;
		tx_dropped	+= p->tx_dropped;
	}
	stats->rx_dropped  = rx_dropped;
	stats->rx_frame_errors = rx_frame_errors;
	stats->tx_dropped = tx_dropped;
}

J
Jason Wang 已提交
1156 1157 1158 1159
static int tun_xdp_set(struct net_device *dev, struct bpf_prog *prog,
		       struct netlink_ext_ack *extack)
{
	struct tun_struct *tun = netdev_priv(dev);
J
Jason Wang 已提交
1160
	struct tun_file *tfile;
J
Jason Wang 已提交
1161
	struct bpf_prog *old_prog;
J
Jason Wang 已提交
1162
	int i;
J
Jason Wang 已提交
1163 1164 1165 1166 1167 1168

	old_prog = rtnl_dereference(tun->xdp_prog);
	rcu_assign_pointer(tun->xdp_prog, prog);
	if (old_prog)
		bpf_prog_put(old_prog);

J
Jason Wang 已提交
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
	for (i = 0; i < tun->numqueues; i++) {
		tfile = rtnl_dereference(tun->tfiles[i]);
		if (prog)
			sock_set_flag(&tfile->sk, SOCK_XDP);
		else
			sock_reset_flag(&tfile->sk, SOCK_XDP);
	}
	list_for_each_entry(tfile, &tun->disabled, next) {
		if (prog)
			sock_set_flag(&tfile->sk, SOCK_XDP);
		else
			sock_reset_flag(&tfile->sk, SOCK_XDP);
	}

J
Jason Wang 已提交
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
	return 0;
}

static u32 tun_xdp_query(struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);
	const struct bpf_prog *xdp_prog;

	xdp_prog = rtnl_dereference(tun->xdp_prog);
	if (xdp_prog)
		return xdp_prog->aux->id;

	return 0;
}

1198
static int tun_xdp(struct net_device *dev, struct netdev_bpf *xdp)
J
Jason Wang 已提交
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
{
	switch (xdp->command) {
	case XDP_SETUP_PROG:
		return tun_xdp_set(dev, xdp->prog, xdp->extack);
	case XDP_QUERY_PROG:
		xdp->prog_id = tun_xdp_query(dev);
		return 0;
	default:
		return -EINVAL;
	}
}

N
Nicolas Dichtel 已提交
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
static int tun_net_change_carrier(struct net_device *dev, bool new_carrier)
{
	if (new_carrier) {
		struct tun_struct *tun = netdev_priv(dev);

		if (!tun->numqueues)
			return -EPERM;

		netif_carrier_on(dev);
	} else {
		netif_carrier_off(dev);
	}
	return 0;
}

S
Stephen Hemminger 已提交
1226
static const struct net_device_ops tun_netdev_ops = {
1227
	.ndo_uninit		= tun_net_uninit,
S
Stephen Hemminger 已提交
1228 1229
	.ndo_open		= tun_net_open,
	.ndo_stop		= tun_net_close,
1230
	.ndo_start_xmit		= tun_net_xmit,
1231
	.ndo_fix_features	= tun_net_fix_features,
J
Jason Wang 已提交
1232
	.ndo_select_queue	= tun_select_queue,
1233
	.ndo_set_rx_headroom	= tun_set_headroom,
1234
	.ndo_get_stats64	= tun_net_get_stats64,
N
Nicolas Dichtel 已提交
1235
	.ndo_change_carrier	= tun_net_change_carrier,
S
Stephen Hemminger 已提交
1236 1237
};

1238 1239 1240 1241 1242 1243 1244 1245
static void __tun_xdp_flush_tfile(struct tun_file *tfile)
{
	/* Notify and wake up reader process */
	if (tfile->flags & TUN_FASYNC)
		kill_fasync(&tfile->fasync, SIGIO, POLL_IN);
	tfile->socket.sk->sk_data_ready(tfile->socket.sk);
}

1246 1247
static int tun_xdp_xmit(struct net_device *dev, int n,
			struct xdp_frame **frames, u32 flags)
J
Jason Wang 已提交
1248 1249 1250 1251
{
	struct tun_struct *tun = netdev_priv(dev);
	struct tun_file *tfile;
	u32 numqueues;
1252 1253 1254
	int drops = 0;
	int cnt = n;
	int i;
J
Jason Wang 已提交
1255

1256
	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
1257 1258
		return -EINVAL;

J
Jason Wang 已提交
1259 1260
	rcu_read_lock();

1261
resample:
J
Jason Wang 已提交
1262 1263
	numqueues = READ_ONCE(tun->numqueues);
	if (!numqueues) {
1264 1265
		rcu_read_unlock();
		return -ENXIO; /* Caller will free/return all frames */
J
Jason Wang 已提交
1266 1267 1268 1269
	}

	tfile = rcu_dereference(tun->tfiles[smp_processor_id() %
					    numqueues]);
1270 1271
	if (unlikely(!tfile))
		goto resample;
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285

	spin_lock(&tfile->tx_ring.producer_lock);
	for (i = 0; i < n; i++) {
		struct xdp_frame *xdp = frames[i];
		/* Encode the XDP flag into lowest bit for consumer to differ
		 * XDP buffer from sk_buff.
		 */
		void *frame = tun_xdp_to_ptr(xdp);

		if (__ptr_ring_produce(&tfile->tx_ring, frame)) {
			this_cpu_inc(tun->pcpu_stats->tx_dropped);
			xdp_return_frame_rx_napi(xdp);
			drops++;
		}
J
Jason Wang 已提交
1286
	}
1287
	spin_unlock(&tfile->tx_ring.producer_lock);
J
Jason Wang 已提交
1288

1289 1290 1291
	if (flags & XDP_XMIT_FLUSH)
		__tun_xdp_flush_tfile(tfile);

J
Jason Wang 已提交
1292
	rcu_read_unlock();
1293
	return cnt - drops;
J
Jason Wang 已提交
1294 1295
}

1296 1297 1298 1299 1300 1301 1302
static int tun_xdp_tx(struct net_device *dev, struct xdp_buff *xdp)
{
	struct xdp_frame *frame = convert_to_xdp_frame(xdp);

	if (unlikely(!frame))
		return -EOVERFLOW;

1303
	return tun_xdp_xmit(dev, 1, &frame, XDP_XMIT_FLUSH);
J
Jason Wang 已提交
1304 1305
}

S
Stephen Hemminger 已提交
1306
static const struct net_device_ops tap_netdev_ops = {
1307
	.ndo_uninit		= tun_net_uninit,
S
Stephen Hemminger 已提交
1308 1309
	.ndo_open		= tun_net_open,
	.ndo_stop		= tun_net_close,
1310
	.ndo_start_xmit		= tun_net_xmit,
1311
	.ndo_fix_features	= tun_net_fix_features,
1312
	.ndo_set_rx_mode	= tun_net_mclist,
S
Stephen Hemminger 已提交
1313 1314
	.ndo_set_mac_address	= eth_mac_addr,
	.ndo_validate_addr	= eth_validate_addr,
J
Jason Wang 已提交
1315
	.ndo_select_queue	= tun_select_queue,
1316
	.ndo_features_check	= passthru_features_check,
1317
	.ndo_set_rx_headroom	= tun_set_headroom,
1318
	.ndo_get_stats64	= tun_net_get_stats64,
1319
	.ndo_bpf		= tun_xdp,
J
Jason Wang 已提交
1320
	.ndo_xdp_xmit		= tun_xdp_xmit,
N
Nicolas Dichtel 已提交
1321
	.ndo_change_carrier	= tun_net_change_carrier,
S
Stephen Hemminger 已提交
1322 1323
};

1324
static void tun_flow_init(struct tun_struct *tun)
J
Jason Wang 已提交
1325 1326 1327 1328 1329 1330 1331
{
	int i;

	for (i = 0; i < TUN_NUM_FLOW_ENTRIES; i++)
		INIT_HLIST_HEAD(&tun->flows[i]);

	tun->ageing_time = TUN_FLOW_EXPIRE;
1332 1333 1334
	timer_setup(&tun->flow_gc_timer, tun_flow_cleanup, 0);
	mod_timer(&tun->flow_gc_timer,
		  round_jiffies_up(jiffies + tun->ageing_time));
J
Jason Wang 已提交
1335 1336 1337 1338 1339 1340 1341 1342
}

static void tun_flow_uninit(struct tun_struct *tun)
{
	del_timer_sync(&tun->flow_gc_timer);
	tun_flow_flush(tun);
}

1343 1344 1345
#define MIN_MTU 68
#define MAX_MTU 65535

L
Linus Torvalds 已提交
1346 1347 1348 1349
/* Initialize net device. */
static void tun_net_init(struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);
1350

L
Linus Torvalds 已提交
1351
	switch (tun->flags & TUN_TYPE_MASK) {
M
Michael S. Tsirkin 已提交
1352
	case IFF_TUN:
S
Stephen Hemminger 已提交
1353 1354
		dev->netdev_ops = &tun_netdev_ops;

L
Linus Torvalds 已提交
1355 1356 1357 1358 1359 1360
		/* Point-to-Point TUN Device */
		dev->hard_header_len = 0;
		dev->addr_len = 0;
		dev->mtu = 1500;

		/* Zero header length */
1361
		dev->type = ARPHRD_NONE;
L
Linus Torvalds 已提交
1362 1363 1364
		dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
		break;

M
Michael S. Tsirkin 已提交
1365
	case IFF_TAP:
K
Kusanagi Kouichi 已提交
1366
		dev->netdev_ops = &tap_netdev_ops;
L
Linus Torvalds 已提交
1367 1368
		/* Ethernet TAP Device */
		ether_setup(dev);
1369
		dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1370
		dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1371

1372
		eth_hw_addr_random(dev);
1373

L
Linus Torvalds 已提交
1374 1375
		break;
	}
1376 1377 1378

	dev->min_mtu = MIN_MTU;
	dev->max_mtu = MAX_MTU - dev->hard_header_len;
L
Linus Torvalds 已提交
1379 1380
}

1381 1382 1383 1384 1385 1386 1387
static bool tun_sock_writeable(struct tun_struct *tun, struct tun_file *tfile)
{
	struct sock *sk = tfile->socket.sk;

	return (tun->dev->flags & IFF_UP) && sock_writeable(sk);
}

L
Linus Torvalds 已提交
1388 1389 1390
/* Character device part */

/* Poll */
1391
static __poll_t tun_chr_poll(struct file *file, poll_table *wait)
1392
{
1393
	struct tun_file *tfile = file->private_data;
1394
	struct tun_struct *tun = tun_get(tfile);
1395
	struct sock *sk;
1396
	__poll_t mask = 0;
L
Linus Torvalds 已提交
1397 1398

	if (!tun)
1399
		return EPOLLERR;
L
Linus Torvalds 已提交
1400

J
Jason Wang 已提交
1401
	sk = tfile->socket.sk;
1402

1403
	poll_wait(file, sk_sleep(sk), wait);
1404

1405
	if (!ptr_ring_empty(&tfile->tx_ring))
1406
		mask |= EPOLLIN | EPOLLRDNORM;
L
Linus Torvalds 已提交
1407

1408 1409 1410 1411 1412 1413 1414 1415
	/* Make sure SOCKWQ_ASYNC_NOSPACE is set if not writable to
	 * guarantee EPOLLOUT to be raised by either here or
	 * tun_sock_write_space(). Then process could get notification
	 * after it writes to a down device and meets -EIO.
	 */
	if (tun_sock_writeable(tun, tfile) ||
	    (!test_and_set_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
	     tun_sock_writeable(tun, tfile)))
1416
		mask |= EPOLLOUT | EPOLLWRNORM;
1417

1418
	if (tun->dev->reg_state != NETREG_REGISTERED)
1419
		mask = EPOLLERR;
1420

E
Eric W. Biederman 已提交
1421
	tun_put(tun);
L
Linus Torvalds 已提交
1422 1423 1424
	return mask;
}

1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453
static struct sk_buff *tun_napi_alloc_frags(struct tun_file *tfile,
					    size_t len,
					    const struct iov_iter *it)
{
	struct sk_buff *skb;
	size_t linear;
	int err;
	int i;

	if (it->nr_segs > MAX_SKB_FRAGS + 1)
		return ERR_PTR(-ENOMEM);

	local_bh_disable();
	skb = napi_get_frags(&tfile->napi);
	local_bh_enable();
	if (!skb)
		return ERR_PTR(-ENOMEM);

	linear = iov_iter_single_seg_count(it);
	err = __skb_grow(skb, linear);
	if (err)
		goto free;

	skb->len = len;
	skb->data_len = len - linear;
	skb->truesize += skb->data_len;

	for (i = 1; i < it->nr_segs; i++) {
		size_t fragsz = it->iov[i].iov_len;
1454 1455
		struct page *page;
		void *frag;
1456 1457 1458 1459 1460

		if (fragsz == 0 || fragsz > PAGE_SIZE) {
			err = -EINVAL;
			goto free;
		}
1461 1462
		frag = netdev_alloc_frag(fragsz);
		if (!frag) {
1463 1464 1465
			err = -ENOMEM;
			goto free;
		}
1466 1467 1468
		page = virt_to_head_page(frag);
		skb_fill_page_desc(skb, i - 1, page,
				   frag - page_address(page), fragsz);
1469 1470 1471 1472 1473 1474 1475 1476 1477
	}

	return skb;
free:
	/* frees skb and all frags allocated with napi_alloc_frag() */
	napi_free_frags(&tfile->napi);
	return ERR_PTR(err);
}

1478 1479
/* prepad is the amount to reserve at front.  len is length after that.
 * linear is a hint as to how much to copy (usually headers). */
J
Jason Wang 已提交
1480
static struct sk_buff *tun_alloc_skb(struct tun_file *tfile,
1481 1482
				     size_t prepad, size_t len,
				     size_t linear, int noblock)
1483
{
J
Jason Wang 已提交
1484
	struct sock *sk = tfile->socket.sk;
1485
	struct sk_buff *skb;
1486
	int err;
1487 1488

	/* Under a page?  Don't bother with paged skb. */
H
Herbert Xu 已提交
1489
	if (prepad + len < PAGE_SIZE || !linear)
1490
		linear = len;
1491

1492
	skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
1493
				   &err, 0);
1494
	if (!skb)
1495
		return ERR_PTR(err);
1496 1497 1498

	skb_reserve(skb, prepad);
	skb_put(skb, linear);
1499 1500
	skb->data_len = len - linear;
	skb->len += len - linear;
1501 1502 1503 1504

	return skb;
}

J
Jason Wang 已提交
1505 1506 1507 1508 1509 1510 1511 1512 1513 1514
static void tun_rx_batched(struct tun_struct *tun, struct tun_file *tfile,
			   struct sk_buff *skb, int more)
{
	struct sk_buff_head *queue = &tfile->sk.sk_write_queue;
	struct sk_buff_head process_queue;
	u32 rx_batched = tun->rx_batched;
	bool rcv = false;

	if (!rx_batched || (!more && skb_queue_empty(queue))) {
		local_bh_disable();
M
Matthew Cover 已提交
1515
		skb_record_rx_queue(skb, tfile->queue_index);
J
Jason Wang 已提交
1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
		netif_receive_skb(skb);
		local_bh_enable();
		return;
	}

	spin_lock(&queue->lock);
	if (!more || skb_queue_len(queue) == rx_batched) {
		__skb_queue_head_init(&process_queue);
		skb_queue_splice_tail_init(queue, &process_queue);
		rcv = true;
	} else {
		__skb_queue_tail(queue, skb);
	}
	spin_unlock(&queue->lock);

	if (rcv) {
		struct sk_buff *nskb;

		local_bh_disable();
M
Matthew Cover 已提交
1535 1536
		while ((nskb = __skb_dequeue(&process_queue))) {
			skb_record_rx_queue(nskb, tfile->queue_index);
J
Jason Wang 已提交
1537
			netif_receive_skb(nskb);
M
Matthew Cover 已提交
1538 1539
		}
		skb_record_rx_queue(skb, tfile->queue_index);
J
Jason Wang 已提交
1540 1541 1542 1543 1544
		netif_receive_skb(skb);
		local_bh_enable();
	}
}

1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
static bool tun_can_build_skb(struct tun_struct *tun, struct tun_file *tfile,
			      int len, int noblock, bool zerocopy)
{
	if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
		return false;

	if (tfile->socket.sk->sk_sndbuf != INT_MAX)
		return false;

	if (!noblock)
		return false;

	if (zerocopy)
		return false;

	if (SKB_DATA_ALIGN(len + TUN_RX_PAD) +
	    SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) > PAGE_SIZE)
		return false;

	return true;
}

1567 1568
static struct sk_buff *__tun_build_skb(struct tun_file *tfile,
				       struct page_frag *alloc_frag, char *buf,
J
Jason Wang 已提交
1569
				       int buflen, int len, int pad)
1570 1571 1572 1573 1574 1575
{
	struct sk_buff *skb = build_skb(buf, buflen);

	if (!skb)
		return ERR_PTR(-ENOMEM);

J
Jason Wang 已提交
1576
	skb_reserve(skb, pad);
1577
	skb_put(skb, len);
1578
	skb_set_owner_w(skb, tfile->socket.sk);
1579 1580 1581 1582 1583 1584 1585

	get_page(alloc_frag->page);
	alloc_frag->offset += buflen;

	return skb;
}

J
Jason Wang 已提交
1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
static int tun_xdp_act(struct tun_struct *tun, struct bpf_prog *xdp_prog,
		       struct xdp_buff *xdp, u32 act)
{
	int err;

	switch (act) {
	case XDP_REDIRECT:
		err = xdp_do_redirect(tun->dev, xdp, xdp_prog);
		if (err)
			return err;
		break;
	case XDP_TX:
		err = tun_xdp_tx(tun->dev, xdp);
		if (err < 0)
			return err;
		break;
	case XDP_PASS:
		break;
	default:
		bpf_warn_invalid_xdp_action(act);
		/* fall through */
	case XDP_ABORTED:
		trace_xdp_exception(tun->dev, xdp_prog, act);
		/* fall through */
	case XDP_DROP:
		this_cpu_inc(tun->pcpu_stats->rx_dropped);
		break;
	}

	return act;
}

J
Jason Wang 已提交
1618 1619
static struct sk_buff *tun_build_skb(struct tun_struct *tun,
				     struct tun_file *tfile,
1620
				     struct iov_iter *from,
J
Jason Wang 已提交
1621
				     struct virtio_net_hdr *hdr,
J
Jason Wang 已提交
1622
				     int len, int *skb_xdp)
1623
{
1624
	struct page_frag *alloc_frag = &current->task_frag;
J
Jason Wang 已提交
1625
	struct bpf_prog *xdp_prog;
1626
	int buflen = SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
1627 1628
	char *buf;
	size_t copied;
J
Jason Wang 已提交
1629 1630
	int pad = TUN_RX_PAD;
	int err = 0;
1631 1632 1633 1634

	rcu_read_lock();
	xdp_prog = rcu_dereference(tun->xdp_prog);
	if (xdp_prog)
1635
		pad += XDP_PACKET_HEADROOM;
1636 1637
	buflen += SKB_DATA_ALIGN(len + pad);
	rcu_read_unlock();
1638

1639
	alloc_frag->offset = ALIGN((u64)alloc_frag->offset, SMP_CACHE_BYTES);
1640 1641 1642 1643 1644
	if (unlikely(!skb_page_frag_refill(buflen, alloc_frag, GFP_KERNEL)))
		return ERR_PTR(-ENOMEM);

	buf = (char *)page_address(alloc_frag->page) + alloc_frag->offset;
	copied = copy_page_from_iter(alloc_frag->page,
1645
				     alloc_frag->offset + pad,
1646 1647 1648 1649
				     len, from);
	if (copied != len)
		return ERR_PTR(-EFAULT);

1650 1651 1652 1653
	/* There's a small window that XDP may be set after the check
	 * of xdp_prog above, this should be rare and for simplicity
	 * we do XDP on skb in case the headroom is not enough.
	 */
1654
	if (hdr->gso_type || !xdp_prog) {
J
Jason Wang 已提交
1655
		*skb_xdp = 1;
1656 1657
		return __tun_build_skb(tfile, alloc_frag, buf, buflen, len,
				       pad);
1658 1659 1660
	}

	*skb_xdp = 0;
J
Jason Wang 已提交
1661

1662
	local_bh_disable();
J
Jason Wang 已提交
1663 1664
	rcu_read_lock();
	xdp_prog = rcu_dereference(tun->xdp_prog);
J
Jason Wang 已提交
1665
	if (xdp_prog) {
J
Jason Wang 已提交
1666 1667 1668 1669
		struct xdp_buff xdp;
		u32 act;

		xdp.data_hard_start = buf;
1670
		xdp.data = buf + pad;
1671
		xdp_set_data_meta_invalid(&xdp);
J
Jason Wang 已提交
1672
		xdp.data_end = xdp.data + len;
J
Jesper Dangaard Brouer 已提交
1673
		xdp.rxq = &tfile->xdp_rxq;
J
Jesper Dangaard Brouer 已提交
1674
		xdp.frame_sz = buflen;
J
Jason Wang 已提交
1675

J
Jason Wang 已提交
1676 1677
		act = bpf_prog_run_xdp(xdp_prog, &xdp);
		if (act == XDP_REDIRECT || act == XDP_TX) {
J
Jason Wang 已提交
1678 1679
			get_page(alloc_frag->page);
			alloc_frag->offset += buflen;
J
Jason Wang 已提交
1680
		}
J
Jason Wang 已提交
1681
		err = tun_xdp_act(tun, xdp_prog, &xdp, act);
1682 1683 1684 1685 1686 1687
		if (err < 0) {
			if (act == XDP_REDIRECT || act == XDP_TX)
				put_page(alloc_frag->page);
			goto out;
		}

1688
		if (err == XDP_REDIRECT)
1689
			xdp_do_flush();
J
Jason Wang 已提交
1690 1691 1692 1693 1694
		if (err != XDP_PASS)
			goto out;

		pad = xdp.data - xdp.data_hard_start;
		len = xdp.data_end - xdp.data;
J
Jason Wang 已提交
1695
	}
1696 1697
	rcu_read_unlock();
	local_bh_enable();
J
Jason Wang 已提交
1698

1699
	return __tun_build_skb(tfile, alloc_frag, buf, buflen, len, pad);
J
Jason Wang 已提交
1700

1701
out:
J
Jason Wang 已提交
1702
	rcu_read_unlock();
1703
	local_bh_enable();
J
Jason Wang 已提交
1704
	return NULL;
1705 1706
}

L
Linus Torvalds 已提交
1707
/* Get packet from user space buffer */
J
Jason Wang 已提交
1708
static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile,
1709
			    void *msg_control, struct iov_iter *from,
J
Jason Wang 已提交
1710
			    int noblock, bool more)
L
Linus Torvalds 已提交
1711
{
1712
	struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
L
Linus Torvalds 已提交
1713
	struct sk_buff *skb;
1714
	size_t total_len = iov_iter_count(from);
1715
	size_t len = total_len, align = tun->align, linear;
1716
	struct virtio_net_hdr gso = { 0 };
1717
	struct tun_pcpu_stats *stats;
1718
	int good_linear;
1719 1720 1721
	int copylen;
	bool zerocopy = false;
	int err;
1722
	u32 rxhash = 0;
J
Jason Wang 已提交
1723
	int skb_xdp = 1;
E
Eric Dumazet 已提交
1724
	bool frags = tun_napi_frags_enabled(tfile);
L
Linus Torvalds 已提交
1725

M
Michael S. Tsirkin 已提交
1726
	if (!(tun->flags & IFF_NO_PI)) {
1727
		if (len < sizeof(pi))
L
Linus Torvalds 已提交
1728
			return -EINVAL;
1729
		len -= sizeof(pi);
L
Linus Torvalds 已提交
1730

1731
		if (!copy_from_iter_full(&pi, sizeof(pi), from))
L
Linus Torvalds 已提交
1732 1733 1734
			return -EFAULT;
	}

M
Michael S. Tsirkin 已提交
1735
	if (tun->flags & IFF_VNET_HDR) {
W
Willem de Bruijn 已提交
1736 1737 1738
		int vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz);

		if (len < vnet_hdr_sz)
1739
			return -EINVAL;
W
Willem de Bruijn 已提交
1740
		len -= vnet_hdr_sz;
1741

1742
		if (!copy_from_iter_full(&gso, sizeof(gso), from))
1743 1744
			return -EFAULT;

1745
		if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) &&
1746 1747
		    tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2 > tun16_to_cpu(tun, gso.hdr_len))
			gso.hdr_len = cpu_to_tun16(tun, tun16_to_cpu(tun, gso.csum_start) + tun16_to_cpu(tun, gso.csum_offset) + 2);
1748

1749
		if (tun16_to_cpu(tun, gso.hdr_len) > len)
1750
			return -EINVAL;
W
Willem de Bruijn 已提交
1751
		iov_iter_advance(from, vnet_hdr_sz - sizeof(gso));
1752 1753
	}

M
Michael S. Tsirkin 已提交
1754
	if ((tun->flags & TUN_TYPE_MASK) == IFF_TAP) {
1755
		align += NET_IP_ALIGN;
H
Herbert Xu 已提交
1756
		if (unlikely(len < ETH_HLEN ||
1757
			     (gso.hdr_len && tun16_to_cpu(tun, gso.hdr_len) < ETH_HLEN)))
1758 1759
			return -EINVAL;
	}
1760

1761 1762
	good_linear = SKB_MAX_HEAD(align);

1763
	if (msg_control) {
1764 1765
		struct iov_iter i = *from;

1766 1767
		/* There are 256 bytes to be copied in skb, so there is
		 * enough room for skb expand head in case it is used.
1768 1769
		 * The rest of the buffer is mapped from userspace.
		 */
1770
		copylen = gso.hdr_len ? tun16_to_cpu(tun, gso.hdr_len) : GOODCOPY_LEN;
1771 1772
		if (copylen > good_linear)
			copylen = good_linear;
1773
		linear = copylen;
1774 1775
		iov_iter_advance(&i, copylen);
		if (iov_iter_npages(&i, INT_MAX) <= MAX_SKB_FRAGS)
1776 1777 1778
			zerocopy = true;
	}

1779
	if (!frags && tun_can_build_skb(tun, tfile, len, noblock, zerocopy)) {
J
Jason Wang 已提交
1780 1781 1782 1783 1784
		/* For the packet that is not easy to be processed
		 * (e.g gso or jumbo packet), we will do it at after
		 * skb was created with generic XDP routine.
		 */
		skb = tun_build_skb(tun, tfile, from, &gso, len, &skb_xdp);
1785
		if (IS_ERR(skb)) {
1786
			this_cpu_inc(tun->pcpu_stats->rx_dropped);
1787 1788
			return PTR_ERR(skb);
		}
J
Jason Wang 已提交
1789 1790
		if (!skb)
			return total_len;
1791 1792 1793 1794 1795 1796 1797 1798
	} else {
		if (!zerocopy) {
			copylen = len;
			if (tun16_to_cpu(tun, gso.hdr_len) > good_linear)
				linear = good_linear;
			else
				linear = tun16_to_cpu(tun, gso.hdr_len);
		}
L
Linus Torvalds 已提交
1799

1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
		if (frags) {
			mutex_lock(&tfile->napi_mutex);
			skb = tun_napi_alloc_frags(tfile, copylen, from);
			/* tun_napi_alloc_frags() enforces a layout for the skb.
			 * If zerocopy is enabled, then this layout will be
			 * overwritten by zerocopy_sg_from_iter().
			 */
			zerocopy = false;
		} else {
			skb = tun_alloc_skb(tfile, align, copylen, linear,
					    noblock);
		}

1813 1814 1815
		if (IS_ERR(skb)) {
			if (PTR_ERR(skb) != -EAGAIN)
				this_cpu_inc(tun->pcpu_stats->rx_dropped);
1816 1817
			if (frags)
				mutex_unlock(&tfile->napi_mutex);
1818 1819
			return PTR_ERR(skb);
		}
1820

1821 1822 1823 1824 1825 1826
		if (zerocopy)
			err = zerocopy_sg_from_iter(skb, from);
		else
			err = skb_copy_datagram_from_iter(skb, 0, from, len);

		if (err) {
E
Eric Dumazet 已提交
1827 1828
			err = -EFAULT;
drop:
1829 1830
			this_cpu_inc(tun->pcpu_stats->rx_dropped);
			kfree_skb(skb);
1831 1832 1833 1834 1835
			if (frags) {
				tfile->napi.skb = NULL;
				mutex_unlock(&tfile->napi_mutex);
			}

E
Eric Dumazet 已提交
1836
			return err;
1837
		}
D
Dave Jones 已提交
1838
	}
L
Linus Torvalds 已提交
1839

1840
	if (virtio_net_hdr_to_skb(skb, &gso, tun_is_little_endian(tun))) {
1841 1842
		this_cpu_inc(tun->pcpu_stats->rx_frame_errors);
		kfree_skb(skb);
1843 1844 1845 1846 1847
		if (frags) {
			tfile->napi.skb = NULL;
			mutex_unlock(&tfile->napi_mutex);
		}

1848 1849 1850
		return -EINVAL;
	}

L
Linus Torvalds 已提交
1851
	switch (tun->flags & TUN_TYPE_MASK) {
M
Michael S. Tsirkin 已提交
1852 1853
	case IFF_TUN:
		if (tun->flags & IFF_NO_PI) {
1854 1855 1856 1857
			u8 ip_version = skb->len ? (skb->data[0] >> 4) : 0;

			switch (ip_version) {
			case 4:
1858 1859
				pi.proto = htons(ETH_P_IP);
				break;
1860
			case 6:
1861 1862 1863
				pi.proto = htons(ETH_P_IPV6);
				break;
			default:
1864
				this_cpu_inc(tun->pcpu_stats->rx_dropped);
1865 1866 1867 1868 1869
				kfree_skb(skb);
				return -EINVAL;
			}
		}

1870
		skb_reset_mac_header(skb);
L
Linus Torvalds 已提交
1871
		skb->protocol = pi.proto;
1872
		skb->dev = tun->dev;
L
Linus Torvalds 已提交
1873
		break;
M
Michael S. Tsirkin 已提交
1874
	case IFF_TAP:
1875 1876 1877 1878 1879
		if (frags && !pskb_may_pull(skb, ETH_HLEN)) {
			err = -ENOMEM;
			goto drop;
		}
		skb->protocol = eth_type_trans(skb, tun->dev);
L
Linus Torvalds 已提交
1880
		break;
1881
	}
L
Linus Torvalds 已提交
1882

1883 1884 1885 1886
	/* copy skb_ubuf_info for callback when skb has no error */
	if (zerocopy) {
		skb_shinfo(skb)->destructor_arg = msg_control;
		skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
1887
		skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG;
1888 1889 1890
	} else if (msg_control) {
		struct ubuf_info *uarg = msg_control;
		uarg->callback(uarg, false);
1891 1892
	}

1893
	skb_reset_network_header(skb);
1894
	skb_probe_transport_header(skb);
1895
	skb_record_rx_queue(skb, tfile->queue_index);
1896

J
Jason Wang 已提交
1897
	if (skb_xdp) {
J
Jason Wang 已提交
1898 1899 1900
		struct bpf_prog *xdp_prog;
		int ret;

1901
		local_bh_disable();
J
Jason Wang 已提交
1902 1903 1904 1905 1906 1907
		rcu_read_lock();
		xdp_prog = rcu_dereference(tun->xdp_prog);
		if (xdp_prog) {
			ret = do_xdp_generic(xdp_prog, skb);
			if (ret != XDP_PASS) {
				rcu_read_unlock();
1908
				local_bh_enable();
1909 1910 1911 1912
				if (frags) {
					tfile->napi.skb = NULL;
					mutex_unlock(&tfile->napi_mutex);
				}
J
Jason Wang 已提交
1913 1914 1915 1916
				return total_len;
			}
		}
		rcu_read_unlock();
1917
		local_bh_enable();
J
Jason Wang 已提交
1918 1919
	}

1920 1921 1922 1923 1924 1925
	/* Compute the costly rx hash only if needed for flow updates.
	 * We may get a very small possibility of OOO during switching, not
	 * worth to optimize.
	 */
	if (!rcu_access_pointer(tun->steering_prog) && tun->numqueues > 1 &&
	    !tfile->detached)
1926
		rxhash = __skb_get_hash_symmetric(skb);
1927

E
Eric Dumazet 已提交
1928 1929 1930
	rcu_read_lock();
	if (unlikely(!(tun->dev->flags & IFF_UP))) {
		err = -EIO;
1931
		rcu_read_unlock();
E
Eric Dumazet 已提交
1932 1933 1934
		goto drop;
	}

1935
	if (frags) {
1936 1937
		u32 headlen;

1938
		/* Exercise flow dissector code path. */
1939 1940 1941
		skb_push(skb, ETH_HLEN);
		headlen = eth_get_headlen(tun->dev, skb->data,
					  skb_headlen(skb));
1942

1943
		if (unlikely(headlen > skb_headlen(skb))) {
1944 1945
			this_cpu_inc(tun->pcpu_stats->rx_dropped);
			napi_free_frags(&tfile->napi);
E
Eric Dumazet 已提交
1946
			rcu_read_unlock();
1947 1948 1949 1950 1951 1952 1953 1954 1955
			mutex_unlock(&tfile->napi_mutex);
			WARN_ON(1);
			return -ENOMEM;
		}

		local_bh_disable();
		napi_gro_frags(&tfile->napi);
		local_bh_enable();
		mutex_unlock(&tfile->napi_mutex);
1956
	} else if (tfile->napi_enabled) {
1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973
		struct sk_buff_head *queue = &tfile->sk.sk_write_queue;
		int queue_len;

		spin_lock_bh(&queue->lock);
		__skb_queue_tail(queue, skb);
		queue_len = skb_queue_len(queue);
		spin_unlock(&queue->lock);

		if (!more || queue_len > NAPI_POLL_WEIGHT)
			napi_schedule(&tfile->napi);

		local_bh_enable();
	} else if (!IS_ENABLED(CONFIG_4KSTACKS)) {
		tun_rx_batched(tun, tfile, skb, more);
	} else {
		netif_rx_ni(skb);
	}
E
Eric Dumazet 已提交
1974
	rcu_read_unlock();
1975

1976 1977
	stats = get_cpu_ptr(tun->pcpu_stats);
	u64_stats_update_begin(&stats->syncp);
E
Eric Dumazet 已提交
1978 1979
	u64_stats_inc(&stats->rx_packets);
	u64_stats_add(&stats->rx_bytes, len);
1980 1981
	u64_stats_update_end(&stats->syncp);
	put_cpu_ptr(stats);
L
Linus Torvalds 已提交
1982

1983 1984 1985
	if (rxhash)
		tun_flow_update(tun, rxhash, tfile);

1986
	return total_len;
1987
}
L
Linus Torvalds 已提交
1988

1989
static ssize_t tun_chr_write_iter(struct kiocb *iocb, struct iov_iter *from)
L
Linus Torvalds 已提交
1990
{
1991
	struct file *file = iocb->ki_filp;
J
Jason Wang 已提交
1992
	struct tun_file *tfile = file->private_data;
1993
	struct tun_struct *tun = tun_get(tfile);
E
Eric W. Biederman 已提交
1994
	ssize_t result;
L
Linus Torvalds 已提交
1995 1996 1997 1998

	if (!tun)
		return -EBADFD;

J
Jason Wang 已提交
1999 2000
	result = tun_get_user(tun, tfile, NULL, from,
			      file->f_flags & O_NONBLOCK, false);
E
Eric W. Biederman 已提交
2001 2002 2003

	tun_put(tun);
	return result;
L
Linus Torvalds 已提交
2004 2005
}

J
Jason Wang 已提交
2006 2007
static ssize_t tun_put_user_xdp(struct tun_struct *tun,
				struct tun_file *tfile,
2008
				struct xdp_frame *xdp_frame,
J
Jason Wang 已提交
2009 2010 2011
				struct iov_iter *iter)
{
	int vnet_hdr_sz = 0;
2012
	size_t size = xdp_frame->len;
J
Jason Wang 已提交
2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027
	struct tun_pcpu_stats *stats;
	size_t ret;

	if (tun->flags & IFF_VNET_HDR) {
		struct virtio_net_hdr gso = { 0 };

		vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz);
		if (unlikely(iov_iter_count(iter) < vnet_hdr_sz))
			return -EINVAL;
		if (unlikely(copy_to_iter(&gso, sizeof(gso), iter) !=
			     sizeof(gso)))
			return -EFAULT;
		iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso));
	}

2028
	ret = copy_to_iter(xdp_frame->data, size, iter) + vnet_hdr_sz;
J
Jason Wang 已提交
2029 2030 2031

	stats = get_cpu_ptr(tun->pcpu_stats);
	u64_stats_update_begin(&stats->syncp);
E
Eric Dumazet 已提交
2032 2033
	u64_stats_inc(&stats->tx_packets);
	u64_stats_add(&stats->tx_bytes, ret);
J
Jason Wang 已提交
2034 2035 2036 2037 2038 2039
	u64_stats_update_end(&stats->syncp);
	put_cpu_ptr(tun->pcpu_stats);

	return ret;
}

L
Linus Torvalds 已提交
2040
/* Put packet to the user space buffer */
2041
static ssize_t tun_put_user(struct tun_struct *tun,
J
Jason Wang 已提交
2042
			    struct tun_file *tfile,
2043
			    struct sk_buff *skb,
H
Herbert Xu 已提交
2044
			    struct iov_iter *iter)
L
Linus Torvalds 已提交
2045 2046
{
	struct tun_pi pi = { 0, skb->protocol };
2047
	struct tun_pcpu_stats *stats;
H
Herbert Xu 已提交
2048
	ssize_t total;
2049
	int vlan_offset = 0;
2050
	int vlan_hlen = 0;
H
Herbert Xu 已提交
2051
	int vnet_hdr_sz = 0;
2052

2053
	if (skb_vlan_tag_present(skb))
2054
		vlan_hlen = VLAN_HLEN;
L
Linus Torvalds 已提交
2055

M
Michael S. Tsirkin 已提交
2056
	if (tun->flags & IFF_VNET_HDR)
W
Willem de Bruijn 已提交
2057
		vnet_hdr_sz = READ_ONCE(tun->vnet_hdr_sz);
L
Linus Torvalds 已提交
2058

H
Herbert Xu 已提交
2059 2060
	total = skb->len + vlan_hlen + vnet_hdr_sz;

M
Michael S. Tsirkin 已提交
2061
	if (!(tun->flags & IFF_NO_PI)) {
H
Herbert Xu 已提交
2062
		if (iov_iter_count(iter) < sizeof(pi))
L
Linus Torvalds 已提交
2063 2064
			return -EINVAL;

H
Herbert Xu 已提交
2065 2066
		total += sizeof(pi);
		if (iov_iter_count(iter) < total) {
L
Linus Torvalds 已提交
2067 2068 2069
			/* Packet will be striped */
			pi.flags |= TUN_PKT_STRIP;
		}
2070

H
Herbert Xu 已提交
2071
		if (copy_to_iter(&pi, sizeof(pi), iter) != sizeof(pi))
L
Linus Torvalds 已提交
2072
			return -EFAULT;
2073
	}
L
Linus Torvalds 已提交
2074

H
Herbert Xu 已提交
2075
	if (vnet_hdr_sz) {
2076
		struct virtio_net_hdr gso;
2077

H
Herbert Xu 已提交
2078
		if (iov_iter_count(iter) < vnet_hdr_sz)
2079 2080
			return -EINVAL;

2081
		if (virtio_net_hdr_from_skb(skb, &gso,
2082 2083
					    tun_is_little_endian(tun), true,
					    vlan_hlen)) {
2084
			struct skb_shared_info *sinfo = skb_shinfo(skb);
2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095
			pr_err("unexpected GSO type: "
			       "0x%x, gso_size %d, hdr_len %d\n",
			       sinfo->gso_type, tun16_to_cpu(tun, gso.gso_size),
			       tun16_to_cpu(tun, gso.hdr_len));
			print_hex_dump(KERN_ERR, "tun: ",
				       DUMP_PREFIX_NONE,
				       16, 1, skb->head,
				       min((int)tun16_to_cpu(tun, gso.hdr_len), 64), true);
			WARN_ON_ONCE(1);
			return -EINVAL;
		}
2096

H
Herbert Xu 已提交
2097
		if (copy_to_iter(&gso, sizeof(gso), iter) != sizeof(gso))
2098
			return -EFAULT;
2099 2100

		iov_iter_advance(iter, vnet_hdr_sz - sizeof(gso));
2101 2102
	}

2103
	if (vlan_hlen) {
H
Herbert Xu 已提交
2104
		int ret;
2105
		struct veth veth;
J
Jason Wang 已提交
2106 2107

		veth.h_vlan_proto = skb->vlan_proto;
2108
		veth.h_vlan_TCI = htons(skb_vlan_tag_get(skb));
J
Jason Wang 已提交
2109 2110 2111

		vlan_offset = offsetof(struct vlan_ethhdr, h_vlan_proto);

H
Herbert Xu 已提交
2112 2113
		ret = skb_copy_datagram_iter(skb, 0, iter, vlan_offset);
		if (ret || !iov_iter_count(iter))
J
Jason Wang 已提交
2114 2115
			goto done;

H
Herbert Xu 已提交
2116 2117
		ret = copy_to_iter(&veth, sizeof(veth), iter);
		if (ret != sizeof(veth) || !iov_iter_count(iter))
J
Jason Wang 已提交
2118 2119
			goto done;
	}
L
Linus Torvalds 已提交
2120

H
Herbert Xu 已提交
2121
	skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset);
L
Linus Torvalds 已提交
2122

J
Jason Wang 已提交
2123
done:
2124 2125 2126
	/* caller is in process context, */
	stats = get_cpu_ptr(tun->pcpu_stats);
	u64_stats_update_begin(&stats->syncp);
E
Eric Dumazet 已提交
2127 2128
	u64_stats_inc(&stats->tx_packets);
	u64_stats_add(&stats->tx_bytes, skb->len + vlan_hlen);
2129 2130
	u64_stats_update_end(&stats->syncp);
	put_cpu_ptr(tun->pcpu_stats);
L
Linus Torvalds 已提交
2131 2132 2133 2134

	return total;
}

J
Jason Wang 已提交
2135
static void *tun_ring_recv(struct tun_file *tfile, int noblock, int *err)
J
Jason Wang 已提交
2136 2137
{
	DECLARE_WAITQUEUE(wait, current);
J
Jason Wang 已提交
2138
	void *ptr = NULL;
J
Jason Wang 已提交
2139
	int error = 0;
J
Jason Wang 已提交
2140

J
Jason Wang 已提交
2141 2142
	ptr = ptr_ring_consume(&tfile->tx_ring);
	if (ptr)
J
Jason Wang 已提交
2143 2144
		goto out;
	if (noblock) {
J
Jason Wang 已提交
2145
		error = -EAGAIN;
J
Jason Wang 已提交
2146 2147 2148
		goto out;
	}

2149
	add_wait_queue(&tfile->socket.wq.wait, &wait);
J
Jason Wang 已提交
2150 2151

	while (1) {
T
Timur Celik 已提交
2152
		set_current_state(TASK_INTERRUPTIBLE);
J
Jason Wang 已提交
2153 2154
		ptr = ptr_ring_consume(&tfile->tx_ring);
		if (ptr)
J
Jason Wang 已提交
2155 2156
			break;
		if (signal_pending(current)) {
J
Jason Wang 已提交
2157
			error = -ERESTARTSYS;
J
Jason Wang 已提交
2158 2159 2160
			break;
		}
		if (tfile->socket.sk->sk_shutdown & RCV_SHUTDOWN) {
J
Jason Wang 已提交
2161
			error = -EFAULT;
J
Jason Wang 已提交
2162 2163 2164 2165 2166 2167
			break;
		}

		schedule();
	}

2168
	__set_current_state(TASK_RUNNING);
2169
	remove_wait_queue(&tfile->socket.wq.wait, &wait);
J
Jason Wang 已提交
2170 2171

out:
J
Jason Wang 已提交
2172
	*err = error;
J
Jason Wang 已提交
2173
	return ptr;
J
Jason Wang 已提交
2174 2175
}

J
Jason Wang 已提交
2176
static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
2177
			   struct iov_iter *to,
J
Jason Wang 已提交
2178
			   int noblock, void *ptr)
L
Linus Torvalds 已提交
2179
{
2180
	ssize_t ret;
J
Jason Wang 已提交
2181
	int err;
L
Linus Torvalds 已提交
2182

W
Wei Xu 已提交
2183
	if (!iov_iter_count(to)) {
J
Jason Wang 已提交
2184
		tun_ptr_free(ptr);
2185
		return 0;
W
Wei Xu 已提交
2186
	}
L
Linus Torvalds 已提交
2187

J
Jason Wang 已提交
2188
	if (!ptr) {
2189
		/* Read frames from ring */
J
Jason Wang 已提交
2190 2191
		ptr = tun_ring_recv(tfile, noblock, &err);
		if (!ptr)
2192 2193
			return err;
	}
H
Herbert Xu 已提交
2194

2195 2196
	if (tun_is_xdp_frame(ptr)) {
		struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr);
J
Jason Wang 已提交
2197

2198
		ret = tun_put_user_xdp(tun, tfile, xdpf, to);
2199
		xdp_return_frame(xdpf);
J
Jason Wang 已提交
2200 2201 2202 2203 2204 2205 2206 2207 2208
	} else {
		struct sk_buff *skb = ptr;

		ret = tun_put_user(tun, tfile, skb, to);
		if (unlikely(ret < 0))
			kfree_skb(skb);
		else
			consume_skb(skb);
	}
L
Linus Torvalds 已提交
2209

2210 2211 2212
	return ret;
}

2213
static ssize_t tun_chr_read_iter(struct kiocb *iocb, struct iov_iter *to)
2214 2215 2216
{
	struct file *file = iocb->ki_filp;
	struct tun_file *tfile = file->private_data;
2217
	struct tun_struct *tun = tun_get(tfile);
2218
	ssize_t len = iov_iter_count(to), ret;
2219 2220 2221

	if (!tun)
		return -EBADFD;
2222
	ret = tun_do_read(tun, tfile, to, file->f_flags & O_NONBLOCK, NULL);
2223
	ret = min_t(ssize_t, ret, len);
Z
Zhi Yong Wu 已提交
2224 2225
	if (ret > 0)
		iocb->ki_pos = ret;
E
Eric W. Biederman 已提交
2226
	tun_put(tun);
L
Linus Torvalds 已提交
2227 2228 2229
	return ret;
}

2230
static void tun_prog_free(struct rcu_head *rcu)
2231
{
2232
	struct tun_prog *prog = container_of(rcu, struct tun_prog, rcu);
2233 2234 2235 2236 2237

	bpf_prog_destroy(prog->prog);
	kfree(prog);
}

J
Jason Wang 已提交
2238 2239
static int __tun_set_ebpf(struct tun_struct *tun,
			  struct tun_prog __rcu **prog_p,
2240
			  struct bpf_prog *prog)
2241
{
2242
	struct tun_prog *old, *new = NULL;
2243 2244 2245 2246 2247 2248 2249 2250

	if (prog) {
		new = kmalloc(sizeof(*new), GFP_KERNEL);
		if (!new)
			return -ENOMEM;
		new->prog = prog;
	}

2251
	spin_lock_bh(&tun->lock);
2252
	old = rcu_dereference_protected(*prog_p,
2253
					lockdep_is_held(&tun->lock));
2254
	rcu_assign_pointer(*prog_p, new);
2255
	spin_unlock_bh(&tun->lock);
2256 2257

	if (old)
2258
		call_rcu(&old->rcu, tun_prog_free);
2259 2260 2261 2262

	return 0;
}

J
Jason Wang 已提交
2263 2264 2265 2266
static void tun_free_netdev(struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);

J
Jason Wang 已提交
2267
	BUG_ON(!(list_empty(&tun->disabled)));
E
Eric Dumazet 已提交
2268

2269
	free_percpu(tun->pcpu_stats);
E
Eric Dumazet 已提交
2270 2271 2272 2273 2274
	/* We clear pcpu_stats so that tun_set_iff() can tell if
	 * tun_free_netdev() has been called from register_netdevice().
	 */
	tun->pcpu_stats = NULL;

J
Jason Wang 已提交
2275
	tun_flow_uninit(tun);
2276
	security_tun_dev_free_security(tun->security);
2277
	__tun_set_ebpf(tun, &tun->steering_prog, NULL);
2278
	__tun_set_ebpf(tun, &tun->filter_prog, NULL);
J
Jason Wang 已提交
2279 2280
}

L
Linus Torvalds 已提交
2281 2282 2283 2284
static void tun_setup(struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);

2285 2286
	tun->owner = INVALID_UID;
	tun->group = INVALID_GID;
2287
	tun_default_link_ksettings(dev, &tun->link_ksettings);
L
Linus Torvalds 已提交
2288 2289

	dev->ethtool_ops = &tun_ethtool_ops;
2290 2291
	dev->needs_free_netdev = true;
	dev->priv_destructor = tun_free_netdev;
J
Jason Wang 已提交
2292 2293
	/* We prefer our own queue length */
	dev->tx_queue_len = TUN_READQ_SIZE;
L
Linus Torvalds 已提交
2294 2295
}

2296 2297 2298
/* Trivial set of netlink ops to allow deleting tun or tap
 * device with netlink.
 */
2299 2300
static int tun_validate(struct nlattr *tb[], struct nlattr *data[],
			struct netlink_ext_ack *extack)
2301
{
2302 2303 2304
	NL_SET_ERR_MSG(extack,
		       "tun/tap creation via rtnetlink is not supported.");
	return -EOPNOTSUPP;
2305 2306
}

2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
static size_t tun_get_size(const struct net_device *dev)
{
	BUILD_BUG_ON(sizeof(u32) != sizeof(uid_t));
	BUILD_BUG_ON(sizeof(u32) != sizeof(gid_t));

	return nla_total_size(sizeof(uid_t)) + /* OWNER */
	       nla_total_size(sizeof(gid_t)) + /* GROUP */
	       nla_total_size(sizeof(u8)) + /* TYPE */
	       nla_total_size(sizeof(u8)) + /* PI */
	       nla_total_size(sizeof(u8)) + /* VNET_HDR */
	       nla_total_size(sizeof(u8)) + /* PERSIST */
	       nla_total_size(sizeof(u8)) + /* MULTI_QUEUE */
	       nla_total_size(sizeof(u32)) + /* NUM_QUEUES */
	       nla_total_size(sizeof(u32)) + /* NUM_DISABLED_QUEUES */
	       0;
}

static int tun_fill_info(struct sk_buff *skb, const struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);

	if (nla_put_u8(skb, IFLA_TUN_TYPE, tun->flags & TUN_TYPE_MASK))
		goto nla_put_failure;
	if (uid_valid(tun->owner) &&
	    nla_put_u32(skb, IFLA_TUN_OWNER,
			from_kuid_munged(current_user_ns(), tun->owner)))
		goto nla_put_failure;
	if (gid_valid(tun->group) &&
	    nla_put_u32(skb, IFLA_TUN_GROUP,
			from_kgid_munged(current_user_ns(), tun->group)))
		goto nla_put_failure;
	if (nla_put_u8(skb, IFLA_TUN_PI, !(tun->flags & IFF_NO_PI)))
		goto nla_put_failure;
	if (nla_put_u8(skb, IFLA_TUN_VNET_HDR, !!(tun->flags & IFF_VNET_HDR)))
		goto nla_put_failure;
	if (nla_put_u8(skb, IFLA_TUN_PERSIST, !!(tun->flags & IFF_PERSIST)))
		goto nla_put_failure;
	if (nla_put_u8(skb, IFLA_TUN_MULTI_QUEUE,
		       !!(tun->flags & IFF_MULTI_QUEUE)))
		goto nla_put_failure;
	if (tun->flags & IFF_MULTI_QUEUE) {
		if (nla_put_u32(skb, IFLA_TUN_NUM_QUEUES, tun->numqueues))
			goto nla_put_failure;
		if (nla_put_u32(skb, IFLA_TUN_NUM_DISABLED_QUEUES,
				tun->numdisabled))
			goto nla_put_failure;
	}

	return 0;

nla_put_failure:
	return -EMSGSIZE;
}

2361 2362 2363 2364 2365
static struct rtnl_link_ops tun_link_ops __read_mostly = {
	.kind		= DRV_NAME,
	.priv_size	= sizeof(struct tun_struct),
	.setup		= tun_setup,
	.validate	= tun_validate,
2366 2367
	.get_size       = tun_get_size,
	.fill_info      = tun_fill_info,
2368 2369
};

2370 2371
static void tun_sock_write_space(struct sock *sk)
{
J
Jason Wang 已提交
2372
	struct tun_file *tfile;
2373
	wait_queue_head_t *wqueue;
2374 2375 2376 2377

	if (!sock_writeable(sk))
		return;

2378
	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &sk->sk_socket->flags))
2379 2380
		return;

2381 2382
	wqueue = sk_sleep(sk);
	if (wqueue && waitqueue_active(wqueue))
2383 2384
		wake_up_interruptible_sync_poll(wqueue, EPOLLOUT |
						EPOLLWRNORM | EPOLLWRBAND);
H
Herbert Xu 已提交
2385

J
Jason Wang 已提交
2386 2387
	tfile = container_of(sk, struct tun_file, sk);
	kill_fasync(&tfile->fasync, SIGIO, POLL_OUT);
2388 2389
}

2390 2391 2392 2393 2394 2395
static void tun_put_page(struct tun_page *tpage)
{
	if (tpage->page)
		__page_frag_cache_drain(tpage->page, tpage->count);
}

2396 2397
static int tun_xdp_one(struct tun_struct *tun,
		       struct tun_file *tfile,
2398 2399
		       struct xdp_buff *xdp, int *flush,
		       struct tun_page *tpage)
2400
{
2401
	unsigned int datasize = xdp->data_end - xdp->data;
2402 2403 2404 2405 2406 2407 2408 2409 2410
	struct tun_xdp_hdr *hdr = xdp->data_hard_start;
	struct virtio_net_hdr *gso = &hdr->gso;
	struct tun_pcpu_stats *stats;
	struct bpf_prog *xdp_prog;
	struct sk_buff *skb = NULL;
	u32 rxhash = 0, act;
	int buflen = hdr->buflen;
	int err = 0;
	bool skb_xdp = false;
2411
	struct page *page;
2412 2413 2414 2415 2416 2417 2418 2419 2420

	xdp_prog = rcu_dereference(tun->xdp_prog);
	if (xdp_prog) {
		if (gso->gso_type) {
			skb_xdp = true;
			goto build;
		}
		xdp_set_data_meta_invalid(xdp);
		xdp->rxq = &tfile->xdp_rxq;
J
Jesper Dangaard Brouer 已提交
2421
		xdp->frame_sz = buflen;
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438

		act = bpf_prog_run_xdp(xdp_prog, xdp);
		err = tun_xdp_act(tun, xdp_prog, xdp, act);
		if (err < 0) {
			put_page(virt_to_head_page(xdp->data));
			return err;
		}

		switch (err) {
		case XDP_REDIRECT:
			*flush = true;
			/* fall through */
		case XDP_TX:
			return 0;
		case XDP_PASS:
			break;
		default:
2439 2440 2441 2442 2443 2444 2445 2446
			page = virt_to_head_page(xdp->data);
			if (tpage->page == page) {
				++tpage->count;
			} else {
				tun_put_page(tpage);
				tpage->page = page;
				tpage->count = 1;
			}
2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469
			return 0;
		}
	}

build:
	skb = build_skb(xdp->data_hard_start, buflen);
	if (!skb) {
		err = -ENOMEM;
		goto out;
	}

	skb_reserve(skb, xdp->data - xdp->data_hard_start);
	skb_put(skb, xdp->data_end - xdp->data);

	if (virtio_net_hdr_to_skb(skb, gso, tun_is_little_endian(tun))) {
		this_cpu_inc(tun->pcpu_stats->rx_frame_errors);
		kfree_skb(skb);
		err = -EINVAL;
		goto out;
	}

	skb->protocol = eth_type_trans(skb, tun->dev);
	skb_reset_network_header(skb);
2470
	skb_probe_transport_header(skb);
2471
	skb_record_rx_queue(skb, tfile->queue_index);
2472 2473 2474 2475 2476 2477 2478

	if (skb_xdp) {
		err = do_xdp_generic(xdp_prog, skb);
		if (err != XDP_PASS)
			goto out;
	}

2479 2480
	if (!rcu_dereference(tun->steering_prog) && tun->numqueues > 1 &&
	    !tfile->detached)
2481 2482 2483 2484
		rxhash = __skb_get_hash_symmetric(skb);

	netif_receive_skb(skb);

2485 2486 2487 2488
	/* No need for get_cpu_ptr() here since this function is
	 * always called with bh disabled
	 */
	stats = this_cpu_ptr(tun->pcpu_stats);
2489
	u64_stats_update_begin(&stats->syncp);
E
Eric Dumazet 已提交
2490 2491
	u64_stats_inc(&stats->rx_packets);
	u64_stats_add(&stats->rx_bytes, datasize);
2492 2493 2494 2495 2496 2497 2498 2499 2500
	u64_stats_update_end(&stats->syncp);

	if (rxhash)
		tun_flow_update(tun, rxhash, tfile);

out:
	return err;
}

2501
static int tun_sendmsg(struct socket *sock, struct msghdr *m, size_t total_len)
2502
{
2503
	int ret, i;
J
Jason Wang 已提交
2504
	struct tun_file *tfile = container_of(sock, struct tun_file, socket);
2505
	struct tun_struct *tun = tun_get(tfile);
2506
	struct tun_msg_ctl *ctl = m->msg_control;
2507
	struct xdp_buff *xdp;
J
Jason Wang 已提交
2508 2509 2510

	if (!tun)
		return -EBADFD;
2511

2512
	if (ctl && (ctl->type == TUN_MSG_PTR)) {
2513
		struct tun_page tpage;
2514 2515 2516
		int n = ctl->num;
		int flush = 0;

2517 2518
		memset(&tpage, 0, sizeof(tpage));

2519 2520 2521 2522 2523
		local_bh_disable();
		rcu_read_lock();

		for (i = 0; i < n; i++) {
			xdp = &((struct xdp_buff *)ctl->ptr)[i];
2524
			tun_xdp_one(tun, tfile, xdp, &flush, &tpage);
2525 2526 2527
		}

		if (flush)
2528
			xdp_do_flush();
2529 2530 2531 2532

		rcu_read_unlock();
		local_bh_enable();

2533 2534
		tun_put_page(&tpage);

2535 2536 2537
		ret = total_len;
		goto out;
	}
2538 2539

	ret = tun_get_user(tun, tfile, ctl ? ctl->ptr : NULL, &m->msg_iter,
J
Jason Wang 已提交
2540 2541
			   m->msg_flags & MSG_DONTWAIT,
			   m->msg_flags & MSG_MORE);
2542
out:
J
Jason Wang 已提交
2543 2544
	tun_put(tun);
	return ret;
2545 2546
}

2547
static int tun_recvmsg(struct socket *sock, struct msghdr *m, size_t total_len,
2548 2549
		       int flags)
{
J
Jason Wang 已提交
2550
	struct tun_file *tfile = container_of(sock, struct tun_file, socket);
2551
	struct tun_struct *tun = tun_get(tfile);
J
Jason Wang 已提交
2552
	void *ptr = m->msg_control;
2553
	int ret;
J
Jason Wang 已提交
2554

W
Wei Xu 已提交
2555 2556
	if (!tun) {
		ret = -EBADFD;
J
Jason Wang 已提交
2557
		goto out_free;
W
Wei Xu 已提交
2558
	}
J
Jason Wang 已提交
2559

2560
	if (flags & ~(MSG_DONTWAIT|MSG_TRUNC|MSG_ERRQUEUE)) {
2561
		ret = -EINVAL;
W
Wei Xu 已提交
2562
		goto out_put_tun;
2563
	}
2564 2565 2566 2567 2568
	if (flags & MSG_ERRQUEUE) {
		ret = sock_recv_errqueue(sock->sk, m, total_len,
					 SOL_PACKET, TUN_TX_TIMESTAMP);
		goto out;
	}
J
Jason Wang 已提交
2569
	ret = tun_do_read(tun, tfile, &m->msg_iter, flags & MSG_DONTWAIT, ptr);
2570
	if (ret > (ssize_t)total_len) {
2571 2572 2573
		m->msg_flags |= MSG_TRUNC;
		ret = flags & MSG_TRUNC ? ret : total_len;
	}
2574
out:
J
Jason Wang 已提交
2575
	tun_put(tun);
2576
	return ret;
W
Wei Xu 已提交
2577 2578 2579

out_put_tun:
	tun_put(tun);
J
Jason Wang 已提交
2580 2581
out_free:
	tun_ptr_free(ptr);
W
Wei Xu 已提交
2582
	return ret;
2583 2584
}

J
Jason Wang 已提交
2585 2586 2587
static int tun_ptr_peek_len(void *ptr)
{
	if (likely(ptr)) {
2588 2589
		if (tun_is_xdp_frame(ptr)) {
			struct xdp_frame *xdpf = tun_ptr_to_xdp(ptr);
J
Jason Wang 已提交
2590

2591
			return xdpf->len;
J
Jason Wang 已提交
2592 2593 2594 2595 2596 2597 2598
		}
		return __skb_array_len_with_tag(ptr);
	} else {
		return 0;
	}
}

J
Jason Wang 已提交
2599 2600 2601 2602 2603 2604
static int tun_peek_len(struct socket *sock)
{
	struct tun_file *tfile = container_of(sock, struct tun_file, socket);
	struct tun_struct *tun;
	int ret = 0;

2605
	tun = tun_get(tfile);
J
Jason Wang 已提交
2606 2607 2608
	if (!tun)
		return 0;

J
Jason Wang 已提交
2609
	ret = PTR_RING_PEEK_CALL(&tfile->tx_ring, tun_ptr_peek_len);
J
Jason Wang 已提交
2610 2611 2612 2613 2614
	tun_put(tun);

	return ret;
}

2615 2616
/* Ops structure to mimic raw sockets with tun */
static const struct proto_ops tun_socket_ops = {
J
Jason Wang 已提交
2617
	.peek_len = tun_peek_len,
2618 2619 2620 2621
	.sendmsg = tun_sendmsg,
	.recvmsg = tun_recvmsg,
};

2622 2623 2624
static struct proto tun_proto = {
	.name		= "tun",
	.owner		= THIS_MODULE,
J
Jason Wang 已提交
2625
	.obj_size	= sizeof(struct tun_file),
2626
};
2627

2628 2629
static int tun_flags(struct tun_struct *tun)
{
2630
	return tun->flags & (TUN_FEATURES | IFF_PERSIST | IFF_TUN | IFF_TAP);
2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643
}

static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr,
			      char *buf)
{
	struct tun_struct *tun = netdev_priv(to_net_dev(dev));
	return sprintf(buf, "0x%x\n", tun_flags(tun));
}

static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr,
			      char *buf)
{
	struct tun_struct *tun = netdev_priv(to_net_dev(dev));
2644 2645 2646 2647
	return uid_valid(tun->owner)?
		sprintf(buf, "%u\n",
			from_kuid_munged(current_user_ns(), tun->owner)):
		sprintf(buf, "-1\n");
2648 2649 2650 2651 2652 2653
}

static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr,
			      char *buf)
{
	struct tun_struct *tun = netdev_priv(to_net_dev(dev));
2654 2655 2656 2657
	return gid_valid(tun->group) ?
		sprintf(buf, "%u\n",
			from_kgid_munged(current_user_ns(), tun->group)):
		sprintf(buf, "-1\n");
2658 2659 2660 2661 2662 2663
}

static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL);
static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL);
static DEVICE_ATTR(group, 0444, tun_show_group, NULL);

2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674
static struct attribute *tun_dev_attrs[] = {
	&dev_attr_tun_flags.attr,
	&dev_attr_owner.attr,
	&dev_attr_group.attr,
	NULL
};

static const struct attribute_group tun_attr_group = {
	.attrs = tun_dev_attrs
};

2675
static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
L
Linus Torvalds 已提交
2676 2677
{
	struct tun_struct *tun;
J
Jason Wang 已提交
2678
	struct tun_file *tfile = file->private_data;
L
Linus Torvalds 已提交
2679 2680 2681
	struct net_device *dev;
	int err;

2682 2683 2684
	if (tfile->detached)
		return -EINVAL;

2685 2686 2687 2688 2689 2690 2691 2692 2693
	if ((ifr->ifr_flags & IFF_NAPI_FRAGS)) {
		if (!capable(CAP_NET_ADMIN))
			return -EPERM;

		if (!(ifr->ifr_flags & IFF_NAPI) ||
		    (ifr->ifr_flags & TUN_TYPE_MASK) != IFF_TAP)
			return -EINVAL;
	}

2694 2695
	dev = __dev_get_by_name(net, ifr->ifr_name);
	if (dev) {
2696 2697
		if (ifr->ifr_flags & IFF_TUN_EXCL)
			return -EBUSY;
2698 2699 2700 2701 2702 2703 2704
		if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
			tun = netdev_priv(dev);
		else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
			tun = netdev_priv(dev);
		else
			return -EINVAL;

2705
		if (!!(ifr->ifr_flags & IFF_MULTI_QUEUE) !=
M
Michael S. Tsirkin 已提交
2706
		    !!(tun->flags & IFF_MULTI_QUEUE))
2707 2708
			return -EINVAL;

2709
		if (tun_not_capable(tun))
P
Paul Moore 已提交
2710
			return -EPERM;
2711
		err = security_tun_dev_open(tun->security);
P
Paul Moore 已提交
2712 2713 2714
		if (err < 0)
			return err;

2715
		err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER,
E
Eric Dumazet 已提交
2716
				 ifr->ifr_flags & IFF_NAPI,
2717
				 ifr->ifr_flags & IFF_NAPI_FRAGS, true);
2718 2719
		if (err < 0)
			return err;
J
Jason Wang 已提交
2720

M
Michael S. Tsirkin 已提交
2721
		if (tun->flags & IFF_MULTI_QUEUE &&
2722 2723 2724 2725
		    (tun->numqueues + tun->numdisabled > 1)) {
			/* One or more queue has already been attached, no need
			 * to initialize the device again.
			 */
2726
			netdev_state_change(dev);
2727 2728
			return 0;
		}
2729 2730 2731

		tun->flags = (tun->flags & ~TUN_FEATURES) |
			      (ifr->ifr_flags & TUN_FEATURES);
2732 2733 2734

		netdev_state_change(dev);
	} else {
L
Linus Torvalds 已提交
2735 2736
		char *name;
		unsigned long flags = 0;
2737 2738
		int queues = ifr->ifr_flags & IFF_MULTI_QUEUE ?
			     MAX_TAP_QUEUES : 1;
L
Linus Torvalds 已提交
2739

2740
		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2741
			return -EPERM;
P
Paul Moore 已提交
2742 2743 2744
		err = security_tun_dev_create();
		if (err < 0)
			return err;
2745

L
Linus Torvalds 已提交
2746 2747 2748
		/* Set dev type */
		if (ifr->ifr_flags & IFF_TUN) {
			/* TUN device */
M
Michael S. Tsirkin 已提交
2749
			flags |= IFF_TUN;
L
Linus Torvalds 已提交
2750 2751 2752
			name = "tun%d";
		} else if (ifr->ifr_flags & IFF_TAP) {
			/* TAP device */
M
Michael S. Tsirkin 已提交
2753
			flags |= IFF_TAP;
L
Linus Torvalds 已提交
2754
			name = "tap%d";
2755
		} else
2756
			return -EINVAL;
2757

L
Linus Torvalds 已提交
2758 2759 2760
		if (*ifr->ifr_name)
			name = ifr->ifr_name;

J
Jason Wang 已提交
2761
		dev = alloc_netdev_mqs(sizeof(struct tun_struct), name,
2762 2763
				       NET_NAME_UNKNOWN, tun_setup, queues,
				       queues);
2764

L
Linus Torvalds 已提交
2765 2766 2767
		if (!dev)
			return -ENOMEM;

2768
		dev_net_set(dev, net);
2769
		dev->rtnl_link_ops = &tun_link_ops;
2770
		dev->ifindex = tfile->ifindex;
2771
		dev->sysfs_groups[0] = &tun_attr_group;
S
Stephen Hemminger 已提交
2772

L
Linus Torvalds 已提交
2773 2774 2775
		tun = netdev_priv(dev);
		tun->dev = dev;
		tun->flags = flags;
2776
		tun->txflt.count = 0;
2777
		tun->vnet_hdr_sz = sizeof(struct virtio_net_hdr);
2778

2779
		tun->align = NET_SKB_PAD;
J
Jason Wang 已提交
2780 2781
		tun->filter_attached = false;
		tun->sndbuf = tfile->socket.sk->sk_sndbuf;
J
Jason Wang 已提交
2782
		tun->rx_batched = 0;
2783
		RCU_INIT_POINTER(tun->steering_prog, NULL);
2784

2785 2786 2787 2788 2789 2790
		tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats);
		if (!tun->pcpu_stats) {
			err = -ENOMEM;
			goto err_free_dev;
		}

J
Jason Wang 已提交
2791 2792
		spin_lock_init(&tun->lock);

2793 2794
		err = security_tun_dev_alloc_security(&tun->security);
		if (err < 0)
2795
			goto err_free_stat;
P
Paul Moore 已提交
2796

L
Linus Torvalds 已提交
2797
		tun_net_init(dev);
2798
		tun_flow_init(tun);
J
Jason Wang 已提交
2799

2800
		dev->hw_features = NETIF_F_SG | NETIF_F_FRAGLIST |
J
Jason Wang 已提交
2801 2802
				   TUN_USER_FEATURES | NETIF_F_HW_VLAN_CTAG_TX |
				   NETIF_F_HW_VLAN_STAG_TX;
2803
		dev->features = dev->hw_features | NETIF_F_LLTX;
2804 2805 2806
		dev->vlan_features = dev->features &
				     ~(NETIF_F_HW_VLAN_CTAG_TX |
				       NETIF_F_HW_VLAN_STAG_TX);
2807

2808 2809 2810
		tun->flags = (tun->flags & ~TUN_FEATURES) |
			      (ifr->ifr_flags & TUN_FEATURES);

J
Jason Wang 已提交
2811
		INIT_LIST_HEAD(&tun->disabled);
E
Eric Dumazet 已提交
2812
		err = tun_attach(tun, file, false, ifr->ifr_flags & IFF_NAPI,
2813
				 ifr->ifr_flags & IFF_NAPI_FRAGS, false);
2814
		if (err < 0)
2815
			goto err_free_flow;
2816

L
Linus Torvalds 已提交
2817 2818
		err = register_netdevice(tun->dev);
		if (err < 0)
2819
			goto err_detach;
2820 2821 2822 2823
		/* free_netdev() won't check refcnt, to aovid race
		 * with dev_put() we need publish tun after registration.
		 */
		rcu_assign_pointer(tfile->tun, tun);
L
Linus Torvalds 已提交
2824 2825
	}

2826 2827
	netif_carrier_on(tun->dev);

2828 2829 2830 2831
	/* Make sure persistent devices do not get stuck in
	 * xoff state.
	 */
	if (netif_running(tun->dev))
J
Jason Wang 已提交
2832
		netif_tx_wake_all_queues(tun->dev);
2833

L
Linus Torvalds 已提交
2834 2835 2836
	strcpy(ifr->ifr_name, tun->dev->name);
	return 0;

2837 2838
err_detach:
	tun_detach_all(dev);
E
Eric Dumazet 已提交
2839 2840 2841 2842 2843 2844
	/* We are here because register_netdevice() has failed.
	 * If register_netdevice() already called tun_free_netdev()
	 * while dealing with the error, tun->pcpu_stats has been cleared.
	 */
	if (!tun->pcpu_stats)
		goto err_free_dev;
2845

2846 2847 2848
err_free_flow:
	tun_flow_uninit(tun);
	security_tun_dev_free_security(tun->security);
2849 2850
err_free_stat:
	free_percpu(tun->pcpu_stats);
2851
err_free_dev:
L
Linus Torvalds 已提交
2852 2853 2854 2855
	free_netdev(dev);
	return err;
}

2856
static void tun_get_iff(struct tun_struct *tun, struct ifreq *ifr)
2857 2858 2859
{
	strcpy(ifr->ifr_name, tun->dev->name);

2860
	ifr->ifr_flags = tun_flags(tun);
2861 2862 2863

}

2864 2865
/* This is like a cut-down ethtool ops, except done via tun fd so no
 * privs required. */
2866
static int set_offload(struct tun_struct *tun, unsigned long arg)
2867
{
2868
	netdev_features_t features = 0;
2869 2870

	if (arg & TUN_F_CSUM) {
2871
		features |= NETIF_F_HW_CSUM;
2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884
		arg &= ~TUN_F_CSUM;

		if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
			if (arg & TUN_F_TSO_ECN) {
				features |= NETIF_F_TSO_ECN;
				arg &= ~TUN_F_TSO_ECN;
			}
			if (arg & TUN_F_TSO4)
				features |= NETIF_F_TSO;
			if (arg & TUN_F_TSO6)
				features |= NETIF_F_TSO6;
			arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
		}
2885 2886

		arg &= ~TUN_F_UFO;
2887 2888 2889 2890 2891 2892 2893
	}

	/* This gives the user a way to test for new features in future by
	 * trying to set them. */
	if (arg)
		return -EINVAL;

2894
	tun->set_features = features;
2895 2896
	tun->dev->wanted_features &= ~TUN_USER_FEATURES;
	tun->dev->wanted_features |= features;
2897
	netdev_update_features(tun->dev);
2898 2899 2900 2901

	return 0;
}

J
Jason Wang 已提交
2902 2903 2904 2905 2906 2907
static void tun_detach_filter(struct tun_struct *tun, int n)
{
	int i;
	struct tun_file *tfile;

	for (i = 0; i < n; i++) {
2908
		tfile = rtnl_dereference(tun->tfiles[i]);
2909 2910 2911
		lock_sock(tfile->socket.sk);
		sk_detach_filter(tfile->socket.sk);
		release_sock(tfile->socket.sk);
J
Jason Wang 已提交
2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922
	}

	tun->filter_attached = false;
}

static int tun_attach_filter(struct tun_struct *tun)
{
	int i, ret = 0;
	struct tun_file *tfile;

	for (i = 0; i < tun->numqueues; i++) {
2923
		tfile = rtnl_dereference(tun->tfiles[i]);
2924 2925 2926
		lock_sock(tfile->socket.sk);
		ret = sk_attach_filter(&tun->fprog, tfile->socket.sk);
		release_sock(tfile->socket.sk);
J
Jason Wang 已提交
2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942
		if (ret) {
			tun_detach_filter(tun, i);
			return ret;
		}
	}

	tun->filter_attached = true;
	return ret;
}

static void tun_set_sndbuf(struct tun_struct *tun)
{
	struct tun_file *tfile;
	int i;

	for (i = 0; i < tun->numqueues; i++) {
2943
		tfile = rtnl_dereference(tun->tfiles[i]);
J
Jason Wang 已提交
2944 2945 2946 2947
		tfile->socket.sk->sk_sndbuf = tun->sndbuf;
	}
}

2948 2949 2950 2951 2952 2953 2954 2955 2956
static int tun_set_queue(struct file *file, struct ifreq *ifr)
{
	struct tun_file *tfile = file->private_data;
	struct tun_struct *tun;
	int ret = 0;

	rtnl_lock();

	if (ifr->ifr_flags & IFF_ATTACH_QUEUE) {
J
Jason Wang 已提交
2957
		tun = tfile->detached;
2958
		if (!tun) {
2959
			ret = -EINVAL;
2960 2961 2962 2963 2964
			goto unlock;
		}
		ret = security_tun_dev_attach_queue(tun->security);
		if (ret < 0)
			goto unlock;
E
Eric Dumazet 已提交
2965
		ret = tun_attach(tun, file, false, tun->flags & IFF_NAPI,
2966
				 tun->flags & IFF_NAPI_FRAGS, true);
J
Jason Wang 已提交
2967
	} else if (ifr->ifr_flags & IFF_DETACH_QUEUE) {
2968
		tun = rtnl_dereference(tfile->tun);
M
Michael S. Tsirkin 已提交
2969
		if (!tun || !(tun->flags & IFF_MULTI_QUEUE) || tfile->detached)
J
Jason Wang 已提交
2970 2971 2972 2973
			ret = -EINVAL;
		else
			__tun_detach(tfile, false);
	} else
2974 2975
		ret = -EINVAL;

2976 2977 2978
	if (ret >= 0)
		netdev_state_change(tun->dev);

2979
unlock:
2980 2981 2982 2983
	rtnl_unlock();
	return ret;
}

2984 2985
static int tun_set_ebpf(struct tun_struct *tun, struct tun_prog **prog_p,
			void __user *data)
2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000
{
	struct bpf_prog *prog;
	int fd;

	if (copy_from_user(&fd, data, sizeof(fd)))
		return -EFAULT;

	if (fd == -1) {
		prog = NULL;
	} else {
		prog = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER);
		if (IS_ERR(prog))
			return PTR_ERR(prog);
	}

3001
	return __tun_set_ebpf(tun, prog_p, prog);
3002 3003
}

3004 3005
static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
			    unsigned long arg, int ifreq_len)
L
Linus Torvalds 已提交
3006
{
E
Eric W. Biederman 已提交
3007
	struct tun_file *tfile = file->private_data;
3008
	struct net *net = sock_net(&tfile->sk);
E
Eric W. Biederman 已提交
3009
	struct tun_struct *tun;
L
Linus Torvalds 已提交
3010
	void __user* argp = (void __user*)arg;
N
Nicolas Dichtel 已提交
3011
	unsigned int ifindex, carrier;
L
Linus Torvalds 已提交
3012
	struct ifreq ifr;
3013 3014
	kuid_t owner;
	kgid_t group;
3015
	int sndbuf;
3016
	int vnet_hdr_sz;
M
Michael S. Tsirkin 已提交
3017
	int le;
3018
	int ret;
3019
	bool do_notify = false;
L
Linus Torvalds 已提交
3020

3021 3022
	if (cmd == TUNSETIFF || cmd == TUNSETQUEUE ||
	    (_IOC_TYPE(cmd) == SOCK_IOC_TYPE && cmd != SIOCGSKNS)) {
3023
		if (copy_from_user(&ifr, argp, ifreq_len))
L
Linus Torvalds 已提交
3024
			return -EFAULT;
D
David S. Miller 已提交
3025
	} else {
3026
		memset(&ifr, 0, sizeof(ifr));
D
David S. Miller 已提交
3027
	}
E
Eric W. Biederman 已提交
3028 3029 3030
	if (cmd == TUNGETFEATURES) {
		/* Currently this just means: "what IFF flags are valid?".
		 * This is needed because we never checked for invalid flags on
3031 3032 3033
		 * TUNSETIFF.
		 */
		return put_user(IFF_TUN | IFF_TAP | TUN_FEATURES,
E
Eric W. Biederman 已提交
3034
				(unsigned int __user*)argp);
3035
	} else if (cmd == TUNSETQUEUE) {
3036
		return tun_set_queue(file, &ifr);
3037 3038 3039 3040 3041
	} else if (cmd == SIOCGSKNS) {
		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
			return -EPERM;
		return open_related_ns(&net->ns, get_net_ns);
	}
E
Eric W. Biederman 已提交
3042

J
Jason Wang 已提交
3043
	ret = 0;
3044 3045
	rtnl_lock();

3046
	tun = tun_get(tfile);
3047 3048 3049 3050 3051
	if (cmd == TUNSETIFF) {
		ret = -EEXIST;
		if (tun)
			goto unlock;

L
Linus Torvalds 已提交
3052 3053
		ifr.ifr_name[IFNAMSIZ-1] = '\0';

3054
		ret = tun_set_iff(net, file, &ifr);
L
Linus Torvalds 已提交
3055

3056 3057
		if (ret)
			goto unlock;
L
Linus Torvalds 已提交
3058

3059
		if (copy_to_user(argp, &ifr, ifreq_len))
3060 3061
			ret = -EFAULT;
		goto unlock;
L
Linus Torvalds 已提交
3062
	}
3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075
	if (cmd == TUNSETIFINDEX) {
		ret = -EPERM;
		if (tun)
			goto unlock;

		ret = -EFAULT;
		if (copy_from_user(&ifindex, argp, sizeof(ifindex)))
			goto unlock;

		ret = 0;
		tfile->ifindex = ifindex;
		goto unlock;
	}
L
Linus Torvalds 已提交
3076

3077
	ret = -EBADFD;
L
Linus Torvalds 已提交
3078
	if (!tun)
3079
		goto unlock;
L
Linus Torvalds 已提交
3080

3081
	netif_info(tun, drv, tun->dev, "tun_chr_ioctl cmd %u\n", cmd);
L
Linus Torvalds 已提交
3082

3083
	net = dev_net(tun->dev);
E
Eric W. Biederman 已提交
3084
	ret = 0;
L
Linus Torvalds 已提交
3085
	switch (cmd) {
3086
	case TUNGETIFF:
3087
		tun_get_iff(tun, &ifr);
3088

3089 3090
		if (tfile->detached)
			ifr.ifr_flags |= IFF_DETACH_QUEUE;
3091 3092
		if (!tfile->socket.sk->sk_filter)
			ifr.ifr_flags |= IFF_NOFILTER;
3093

3094
		if (copy_to_user(argp, &ifr, ifreq_len))
E
Eric W. Biederman 已提交
3095
			ret = -EFAULT;
3096 3097
		break;

L
Linus Torvalds 已提交
3098 3099 3100
	case TUNSETNOCSUM:
		/* Disable/Enable checksum */

3101
		/* [unimplemented] */
3102 3103
		netif_info(tun, drv, tun->dev, "ignored: set checksum %s\n",
			   arg ? "disabled" : "enabled");
L
Linus Torvalds 已提交
3104 3105 3106
		break;

	case TUNSETPERSIST:
J
Jason Wang 已提交
3107 3108 3109
		/* Disable/Enable persist mode. Keep an extra reference to the
		 * module to prevent the module being unprobed.
		 */
M
Michael S. Tsirkin 已提交
3110 3111
		if (arg && !(tun->flags & IFF_PERSIST)) {
			tun->flags |= IFF_PERSIST;
J
Jason Wang 已提交
3112
			__module_get(THIS_MODULE);
3113
			do_notify = true;
J
Jason Wang 已提交
3114
		}
M
Michael S. Tsirkin 已提交
3115 3116
		if (!arg && (tun->flags & IFF_PERSIST)) {
			tun->flags &= ~IFF_PERSIST;
J
Jason Wang 已提交
3117
			module_put(THIS_MODULE);
3118
			do_notify = true;
J
Jason Wang 已提交
3119
		}
L
Linus Torvalds 已提交
3120

3121 3122
		netif_info(tun, drv, tun->dev, "persist %s\n",
			   arg ? "enabled" : "disabled");
L
Linus Torvalds 已提交
3123 3124 3125 3126
		break;

	case TUNSETOWNER:
		/* Set owner of the device */
3127 3128 3129 3130 3131 3132
		owner = make_kuid(current_user_ns(), arg);
		if (!uid_valid(owner)) {
			ret = -EINVAL;
			break;
		}
		tun->owner = owner;
3133
		do_notify = true;
3134 3135
		netif_info(tun, drv, tun->dev, "owner set to %u\n",
			   from_kuid(&init_user_ns, tun->owner));
L
Linus Torvalds 已提交
3136 3137
		break;

3138 3139
	case TUNSETGROUP:
		/* Set group of the device */
3140 3141 3142 3143 3144 3145
		group = make_kgid(current_user_ns(), arg);
		if (!gid_valid(group)) {
			ret = -EINVAL;
			break;
		}
		tun->group = group;
3146
		do_notify = true;
3147 3148
		netif_info(tun, drv, tun->dev, "group set to %u\n",
			   from_kgid(&init_user_ns, tun->group));
3149 3150
		break;

3151 3152 3153
	case TUNSETLINK:
		/* Only allow setting the type when the interface is down */
		if (tun->dev->flags & IFF_UP) {
3154 3155
			netif_info(tun, drv, tun->dev,
				   "Linktype set failed because interface is up\n");
3156
			ret = -EBUSY;
3157 3158
		} else {
			tun->dev->type = (int) arg;
3159 3160
			netif_info(tun, drv, tun->dev, "linktype set to %d\n",
				   tun->dev->type);
3161
			ret = 0;
3162
		}
E
Eric W. Biederman 已提交
3163
		break;
3164

L
Linus Torvalds 已提交
3165
	case TUNSETDEBUG:
3166
		tun->msg_enable = (u32)arg;
L
Linus Torvalds 已提交
3167
		break;
3168

3169
	case TUNSETOFFLOAD:
3170
		ret = set_offload(tun, arg);
E
Eric W. Biederman 已提交
3171
		break;
3172

3173 3174
	case TUNSETTXFILTER:
		/* Can be set only for TAPs */
E
Eric W. Biederman 已提交
3175
		ret = -EINVAL;
M
Michael S. Tsirkin 已提交
3176
		if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
E
Eric W. Biederman 已提交
3177
			break;
H
Harvey Harrison 已提交
3178
		ret = update_filter(&tun->txflt, (void __user *)arg);
E
Eric W. Biederman 已提交
3179
		break;
L
Linus Torvalds 已提交
3180 3181

	case SIOCGIFHWADDR:
3182
		/* Get hw address */
3183 3184
		memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
		ifr.ifr_hwaddr.sa_family = tun->dev->type;
3185
		if (copy_to_user(argp, &ifr, ifreq_len))
E
Eric W. Biederman 已提交
3186 3187
			ret = -EFAULT;
		break;
L
Linus Torvalds 已提交
3188 3189

	case SIOCSIFHWADDR:
3190
		/* Set hw address */
3191
		ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr, NULL);
E
Eric W. Biederman 已提交
3192
		break;
3193 3194

	case TUNGETSNDBUF:
J
Jason Wang 已提交
3195
		sndbuf = tfile->socket.sk->sk_sndbuf;
3196 3197 3198 3199 3200 3201 3202 3203 3204
		if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
			ret = -EFAULT;
		break;

	case TUNSETSNDBUF:
		if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
			ret = -EFAULT;
			break;
		}
3205 3206 3207 3208
		if (sndbuf <= 0) {
			ret = -EINVAL;
			break;
		}
3209

J
Jason Wang 已提交
3210 3211
		tun->sndbuf = sndbuf;
		tun_set_sndbuf(tun);
3212 3213
		break;

3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232
	case TUNGETVNETHDRSZ:
		vnet_hdr_sz = tun->vnet_hdr_sz;
		if (copy_to_user(argp, &vnet_hdr_sz, sizeof(vnet_hdr_sz)))
			ret = -EFAULT;
		break;

	case TUNSETVNETHDRSZ:
		if (copy_from_user(&vnet_hdr_sz, argp, sizeof(vnet_hdr_sz))) {
			ret = -EFAULT;
			break;
		}
		if (vnet_hdr_sz < (int)sizeof(struct virtio_net_hdr)) {
			ret = -EINVAL;
			break;
		}

		tun->vnet_hdr_sz = vnet_hdr_sz;
		break;

M
Michael S. Tsirkin 已提交
3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249
	case TUNGETVNETLE:
		le = !!(tun->flags & TUN_VNET_LE);
		if (put_user(le, (int __user *)argp))
			ret = -EFAULT;
		break;

	case TUNSETVNETLE:
		if (get_user(le, (int __user *)argp)) {
			ret = -EFAULT;
			break;
		}
		if (le)
			tun->flags |= TUN_VNET_LE;
		else
			tun->flags &= ~TUN_VNET_LE;
		break;

3250 3251 3252 3253 3254 3255 3256 3257
	case TUNGETVNETBE:
		ret = tun_get_vnet_be(tun, argp);
		break;

	case TUNSETVNETBE:
		ret = tun_set_vnet_be(tun, argp);
		break;

M
Michael S. Tsirkin 已提交
3258 3259 3260
	case TUNATTACHFILTER:
		/* Can be set only for TAPs */
		ret = -EINVAL;
M
Michael S. Tsirkin 已提交
3261
		if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
M
Michael S. Tsirkin 已提交
3262 3263
			break;
		ret = -EFAULT;
J
Jason Wang 已提交
3264
		if (copy_from_user(&tun->fprog, argp, sizeof(tun->fprog)))
M
Michael S. Tsirkin 已提交
3265 3266
			break;

J
Jason Wang 已提交
3267
		ret = tun_attach_filter(tun);
M
Michael S. Tsirkin 已提交
3268 3269 3270 3271 3272
		break;

	case TUNDETACHFILTER:
		/* Can be set only for TAPs */
		ret = -EINVAL;
M
Michael S. Tsirkin 已提交
3273
		if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
M
Michael S. Tsirkin 已提交
3274
			break;
J
Jason Wang 已提交
3275 3276
		ret = 0;
		tun_detach_filter(tun, tun->numqueues);
M
Michael S. Tsirkin 已提交
3277 3278
		break;

P
Pavel Emelyanov 已提交
3279 3280
	case TUNGETFILTER:
		ret = -EINVAL;
M
Michael S. Tsirkin 已提交
3281
		if ((tun->flags & TUN_TYPE_MASK) != IFF_TAP)
P
Pavel Emelyanov 已提交
3282 3283 3284 3285 3286 3287 3288
			break;
		ret = -EFAULT;
		if (copy_to_user(argp, &tun->fprog, sizeof(tun->fprog)))
			break;
		ret = 0;
		break;

3289
	case TUNSETSTEERINGEBPF:
3290
		ret = tun_set_ebpf(tun, &tun->steering_prog, argp);
3291 3292
		break;

3293 3294 3295 3296
	case TUNSETFILTEREBPF:
		ret = tun_set_ebpf(tun, &tun->filter_prog, argp);
		break;

N
Nicolas Dichtel 已提交
3297 3298 3299 3300 3301 3302 3303 3304
	case TUNSETCARRIER:
		ret = -EFAULT;
		if (copy_from_user(&carrier, argp, sizeof(carrier)))
			goto unlock;

		ret = tun_net_change_carrier(tun->dev, (bool)carrier);
		break;

3305 3306 3307 3308 3309 3310 3311
	case TUNGETDEVNETNS:
		ret = -EPERM;
		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
			goto unlock;
		ret = open_related_ns(&net->ns, get_net_ns);
		break;

L
Linus Torvalds 已提交
3312
	default:
E
Eric W. Biederman 已提交
3313 3314
		ret = -EINVAL;
		break;
3315
	}
L
Linus Torvalds 已提交
3316

3317 3318 3319
	if (do_notify)
		netdev_state_change(tun->dev);

3320 3321 3322 3323
unlock:
	rtnl_unlock();
	if (tun)
		tun_put(tun);
E
Eric W. Biederman 已提交
3324
	return ret;
L
Linus Torvalds 已提交
3325 3326
}

3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361
static long tun_chr_ioctl(struct file *file,
			  unsigned int cmd, unsigned long arg)
{
	return __tun_chr_ioctl(file, cmd, arg, sizeof (struct ifreq));
}

#ifdef CONFIG_COMPAT
static long tun_chr_compat_ioctl(struct file *file,
			 unsigned int cmd, unsigned long arg)
{
	switch (cmd) {
	case TUNSETIFF:
	case TUNGETIFF:
	case TUNSETTXFILTER:
	case TUNGETSNDBUF:
	case TUNSETSNDBUF:
	case SIOCGIFHWADDR:
	case SIOCSIFHWADDR:
		arg = (unsigned long)compat_ptr(arg);
		break;
	default:
		arg = (compat_ulong_t)arg;
		break;
	}

	/*
	 * compat_ifreq is shorter than ifreq, so we must not access beyond
	 * the end of that structure. All fields that are used in this
	 * driver are compatible though, we don't need to convert the
	 * contents.
	 */
	return __tun_chr_ioctl(file, cmd, arg, sizeof(struct compat_ifreq));
}
#endif /* CONFIG_COMPAT */

L
Linus Torvalds 已提交
3362 3363
static int tun_chr_fasync(int fd, struct file *file, int on)
{
J
Jason Wang 已提交
3364
	struct tun_file *tfile = file->private_data;
L
Linus Torvalds 已提交
3365 3366
	int ret;

J
Jason Wang 已提交
3367
	if ((ret = fasync_helper(fd, file, on, &tfile->fasync)) < 0)
J
Jonathan Corbet 已提交
3368
		goto out;
3369

L
Linus Torvalds 已提交
3370
	if (on) {
3371
		__f_setown(file, task_pid(current), PIDTYPE_TGID, 0);
J
Jason Wang 已提交
3372
		tfile->flags |= TUN_FASYNC;
3373
	} else
J
Jason Wang 已提交
3374
		tfile->flags &= ~TUN_FASYNC;
J
Jonathan Corbet 已提交
3375 3376 3377
	ret = 0;
out:
	return ret;
L
Linus Torvalds 已提交
3378 3379 3380 3381
}

static int tun_chr_open(struct inode *inode, struct file * file)
{
3382
	struct net *net = current->nsproxy->net_ns;
E
Eric W. Biederman 已提交
3383
	struct tun_file *tfile;
T
Thomas Gleixner 已提交
3384

3385
	tfile = (struct tun_file *)sk_alloc(net, AF_UNSPEC, GFP_KERNEL,
3386
					    &tun_proto, 0);
E
Eric W. Biederman 已提交
3387 3388
	if (!tfile)
		return -ENOMEM;
3389 3390 3391 3392 3393
	if (ptr_ring_init(&tfile->tx_ring, 0, GFP_KERNEL)) {
		sk_free(&tfile->sk);
		return -ENOMEM;
	}

3394
	mutex_init(&tfile->napi_mutex);
3395
	RCU_INIT_POINTER(tfile->tun, NULL);
J
Jason Wang 已提交
3396
	tfile->flags = 0;
3397
	tfile->ifindex = 0;
J
Jason Wang 已提交
3398

3399
	init_waitqueue_head(&tfile->socket.wq.wait);
J
Jason Wang 已提交
3400 3401 3402 3403 3404 3405 3406 3407 3408

	tfile->socket.file = file;
	tfile->socket.ops = &tun_socket_ops;

	sock_init_data(&tfile->socket, &tfile->sk);

	tfile->sk.sk_write_space = tun_sock_write_space;
	tfile->sk.sk_sndbuf = INT_MAX;

E
Eric W. Biederman 已提交
3409
	file->private_data = tfile;
J
Jason Wang 已提交
3410
	INIT_LIST_HEAD(&tfile->next);
J
Jason Wang 已提交
3411

3412 3413
	sock_set_flag(&tfile->sk, SOCK_ZEROCOPY);

L
Linus Torvalds 已提交
3414 3415 3416 3417 3418
	return 0;
}

static int tun_chr_close(struct inode *inode, struct file *file)
{
E
Eric W. Biederman 已提交
3419
	struct tun_file *tfile = file->private_data;
L
Linus Torvalds 已提交
3420

J
Jason Wang 已提交
3421
	tun_detach(tfile, true);
L
Linus Torvalds 已提交
3422 3423 3424 3425

	return 0;
}

3426
#ifdef CONFIG_PROC_FS
3427
static void tun_chr_show_fdinfo(struct seq_file *m, struct file *file)
3428
{
3429
	struct tun_file *tfile = file->private_data;
3430 3431 3432 3433 3434 3435
	struct tun_struct *tun;
	struct ifreq ifr;

	memset(&ifr, 0, sizeof(ifr));

	rtnl_lock();
3436
	tun = tun_get(tfile);
3437
	if (tun)
3438
		tun_get_iff(tun, &ifr);
3439 3440 3441 3442 3443
	rtnl_unlock();

	if (tun)
		tun_put(tun);

3444
	seq_printf(m, "iff:\t%s\n", ifr.ifr_name);
3445 3446 3447
}
#endif

3448
static const struct file_operations tun_fops = {
3449
	.owner	= THIS_MODULE,
L
Linus Torvalds 已提交
3450
	.llseek = no_llseek,
3451
	.read_iter  = tun_chr_read_iter,
3452
	.write_iter = tun_chr_write_iter,
L
Linus Torvalds 已提交
3453
	.poll	= tun_chr_poll,
3454 3455 3456 3457
	.unlocked_ioctl	= tun_chr_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl = tun_chr_compat_ioctl,
#endif
L
Linus Torvalds 已提交
3458 3459
	.open	= tun_chr_open,
	.release = tun_chr_close,
3460 3461 3462 3463
	.fasync = tun_chr_fasync,
#ifdef CONFIG_PROC_FS
	.show_fdinfo = tun_chr_show_fdinfo,
#endif
L
Linus Torvalds 已提交
3464 3465 3466 3467 3468
};

static struct miscdevice tun_miscdev = {
	.minor = TUN_MINOR,
	.name = "tun",
3469
	.nodename = "net/tun",
L
Linus Torvalds 已提交
3470 3471 3472 3473 3474
	.fops = &tun_fops,
};

/* ethtool interface */

3475 3476
static void tun_default_link_ksettings(struct net_device *dev,
				       struct ethtool_link_ksettings *cmd)
3477 3478 3479 3480 3481 3482 3483 3484
{
	ethtool_link_ksettings_zero_link_mode(cmd, supported);
	ethtool_link_ksettings_zero_link_mode(cmd, advertising);
	cmd->base.speed		= SPEED_10;
	cmd->base.duplex	= DUPLEX_FULL;
	cmd->base.port		= PORT_TP;
	cmd->base.phy_address	= 0;
	cmd->base.autoneg	= AUTONEG_DISABLE;
3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501
}

static int tun_get_link_ksettings(struct net_device *dev,
				  struct ethtool_link_ksettings *cmd)
{
	struct tun_struct *tun = netdev_priv(dev);

	memcpy(cmd, &tun->link_ksettings, sizeof(*cmd));
	return 0;
}

static int tun_set_link_ksettings(struct net_device *dev,
				  const struct ethtool_link_ksettings *cmd)
{
	struct tun_struct *tun = netdev_priv(dev);

	memcpy(&tun->link_ksettings, cmd, sizeof(*cmd));
L
Linus Torvalds 已提交
3502 3503 3504 3505 3506 3507 3508
	return 0;
}

static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
{
	struct tun_struct *tun = netdev_priv(dev);

3509 3510
	strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
	strlcpy(info->version, DRV_VERSION, sizeof(info->version));
L
Linus Torvalds 已提交
3511 3512

	switch (tun->flags & TUN_TYPE_MASK) {
M
Michael S. Tsirkin 已提交
3513
	case IFF_TUN:
3514
		strlcpy(info->bus_info, "tun", sizeof(info->bus_info));
L
Linus Torvalds 已提交
3515
		break;
M
Michael S. Tsirkin 已提交
3516
	case IFF_TAP:
3517
		strlcpy(info->bus_info, "tap", sizeof(info->bus_info));
L
Linus Torvalds 已提交
3518 3519 3520 3521 3522 3523 3524
		break;
	}
}

static u32 tun_get_msglevel(struct net_device *dev)
{
	struct tun_struct *tun = netdev_priv(dev);
3525 3526

	return tun->msg_enable;
L
Linus Torvalds 已提交
3527 3528 3529 3530 3531
}

static void tun_set_msglevel(struct net_device *dev, u32 value)
{
	struct tun_struct *tun = netdev_priv(dev);
3532 3533

	tun->msg_enable = value;
L
Linus Torvalds 已提交
3534 3535
}

J
Jason Wang 已提交
3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558
static int tun_get_coalesce(struct net_device *dev,
			    struct ethtool_coalesce *ec)
{
	struct tun_struct *tun = netdev_priv(dev);

	ec->rx_max_coalesced_frames = tun->rx_batched;

	return 0;
}

static int tun_set_coalesce(struct net_device *dev,
			    struct ethtool_coalesce *ec)
{
	struct tun_struct *tun = netdev_priv(dev);

	if (ec->rx_max_coalesced_frames > NAPI_POLL_WEIGHT)
		tun->rx_batched = NAPI_POLL_WEIGHT;
	else
		tun->rx_batched = ec->rx_max_coalesced_frames;

	return 0;
}

3559
static const struct ethtool_ops tun_ethtool_ops = {
3560
	.supported_coalesce_params = ETHTOOL_COALESCE_RX_MAX_FRAMES,
L
Linus Torvalds 已提交
3561 3562 3563
	.get_drvinfo	= tun_get_drvinfo,
	.get_msglevel	= tun_get_msglevel,
	.set_msglevel	= tun_set_msglevel,
3564
	.get_link	= ethtool_op_get_link,
3565
	.get_ts_info	= ethtool_op_get_ts_info,
J
Jason Wang 已提交
3566 3567
	.get_coalesce   = tun_get_coalesce,
	.set_coalesce   = tun_set_coalesce,
3568
	.get_link_ksettings = tun_get_link_ksettings,
3569
	.set_link_ksettings = tun_set_link_ksettings,
L
Linus Torvalds 已提交
3570 3571
};

J
Jason Wang 已提交
3572 3573 3574 3575
static int tun_queue_resize(struct tun_struct *tun)
{
	struct net_device *dev = tun->dev;
	struct tun_file *tfile;
3576
	struct ptr_ring **rings;
J
Jason Wang 已提交
3577 3578 3579
	int n = tun->numqueues + tun->numdisabled;
	int ret, i;

3580 3581
	rings = kmalloc_array(n, sizeof(*rings), GFP_KERNEL);
	if (!rings)
J
Jason Wang 已提交
3582 3583 3584 3585
		return -ENOMEM;

	for (i = 0; i < tun->numqueues; i++) {
		tfile = rtnl_dereference(tun->tfiles[i]);
3586
		rings[i] = &tfile->tx_ring;
J
Jason Wang 已提交
3587 3588
	}
	list_for_each_entry(tfile, &tun->disabled, next)
3589
		rings[i++] = &tfile->tx_ring;
J
Jason Wang 已提交
3590

3591 3592
	ret = ptr_ring_resize_multiple(rings, n,
				       dev->tx_queue_len, GFP_KERNEL,
J
Jason Wang 已提交
3593
				       tun_ptr_free);
J
Jason Wang 已提交
3594

3595
	kfree(rings);
J
Jason Wang 已提交
3596 3597 3598 3599 3600 3601 3602 3603
	return ret;
}

static int tun_device_event(struct notifier_block *unused,
			    unsigned long event, void *ptr)
{
	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
	struct tun_struct *tun = netdev_priv(dev);
3604
	int i;
J
Jason Wang 已提交
3605

3606 3607 3608
	if (dev->rtnl_link_ops != &tun_link_ops)
		return NOTIFY_DONE;

J
Jason Wang 已提交
3609 3610 3611 3612 3613
	switch (event) {
	case NETDEV_CHANGE_TX_QUEUE_LEN:
		if (tun_queue_resize(tun))
			return NOTIFY_BAD;
		break;
3614 3615 3616 3617 3618 3619 3620 3621
	case NETDEV_UP:
		for (i = 0; i < tun->numqueues; i++) {
			struct tun_file *tfile;

			tfile = rtnl_dereference(tun->tfiles[i]);
			tfile->socket.sk->sk_write_space(tfile->socket.sk);
		}
		break;
J
Jason Wang 已提交
3622 3623 3624 3625 3626 3627 3628 3629 3630 3631
	default:
		break;
	}

	return NOTIFY_DONE;
}

static struct notifier_block tun_notifier_block __read_mostly = {
	.notifier_call	= tun_device_event,
};
3632

L
Linus Torvalds 已提交
3633 3634 3635 3636
static int __init tun_init(void)
{
	int ret = 0;

3637
	pr_info("%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
L
Linus Torvalds 已提交
3638

3639
	ret = rtnl_link_register(&tun_link_ops);
3640
	if (ret) {
3641
		pr_err("Can't register link_ops\n");
3642
		goto err_linkops;
3643 3644
	}

L
Linus Torvalds 已提交
3645
	ret = misc_register(&tun_miscdev);
3646
	if (ret) {
3647
		pr_err("Can't register misc device %d\n", TUN_MINOR);
3648 3649
		goto err_misc;
	}
J
Jason Wang 已提交
3650

3651 3652 3653 3654 3655 3656
	ret = register_netdevice_notifier(&tun_notifier_block);
	if (ret) {
		pr_err("Can't register netdevice notifier\n");
		goto err_notifier;
	}

3657
	return  0;
3658 3659 3660

err_notifier:
	misc_deregister(&tun_miscdev);
3661
err_misc:
3662 3663
	rtnl_link_unregister(&tun_link_ops);
err_linkops:
L
Linus Torvalds 已提交
3664 3665 3666 3667 3668
	return ret;
}

static void tun_cleanup(void)
{
3669
	misc_deregister(&tun_miscdev);
3670
	rtnl_link_unregister(&tun_link_ops);
J
Jason Wang 已提交
3671
	unregister_netdevice_notifier(&tun_notifier_block);
L
Linus Torvalds 已提交
3672 3673
}

3674 3675 3676 3677 3678 3679
/* Get an underlying socket object from tun file.  Returns error unless file is
 * attached to a device.  The returned object works like a packet socket, it
 * can be used for sock_sendmsg/sock_recvmsg.  The caller is responsible for
 * holding a reference to the file for as long as the socket is in use. */
struct socket *tun_get_socket(struct file *file)
{
3680
	struct tun_file *tfile;
3681 3682
	if (file->f_op != &tun_fops)
		return ERR_PTR(-EINVAL);
3683 3684
	tfile = file->private_data;
	if (!tfile)
3685
		return ERR_PTR(-EBADFD);
J
Jason Wang 已提交
3686
	return &tfile->socket;
3687 3688 3689
}
EXPORT_SYMBOL_GPL(tun_get_socket);

3690
struct ptr_ring *tun_get_tx_ring(struct file *file)
J
Jason Wang 已提交
3691 3692 3693 3694 3695 3696 3697 3698
{
	struct tun_file *tfile;

	if (file->f_op != &tun_fops)
		return ERR_PTR(-EINVAL);
	tfile = file->private_data;
	if (!tfile)
		return ERR_PTR(-EBADFD);
3699
	return &tfile->tx_ring;
J
Jason Wang 已提交
3700
}
3701
EXPORT_SYMBOL_GPL(tun_get_tx_ring);
J
Jason Wang 已提交
3702

L
Linus Torvalds 已提交
3703 3704 3705 3706 3707 3708
module_init(tun_init);
module_exit(tun_cleanup);
MODULE_DESCRIPTION(DRV_DESCRIPTION);
MODULE_AUTHOR(DRV_COPYRIGHT);
MODULE_LICENSE("GPL");
MODULE_ALIAS_MISCDEV(TUN_MINOR);
3709
MODULE_ALIAS("devname:net/tun");