projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
[deliverable/linux.git] / drivers / net / tun.c
1 /*
2 * TUN - Universal TUN/TAP device driver.
3 * Copyright (C) 1999-2002 Maxim Krasnyansky <maxk@qualcomm.com>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * $Id: tun.c,v 1.15 2002/03/01 02:44:24 maxk Exp $
16 */
17
18 /*
19 * Changes:
20 *
21 * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
22 * Add TUNSETLINK ioctl to set the link encapsulation
23 *
24 * Mark Smith <markzzzsmith@yahoo.com.au>
25 * Use random_ether_addr() for tap MAC address.
26 *
27 * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20
28 * Fixes in packet dropping, queue length setting and queue wakeup.
29 * Increased default tx queue length.
30 * Added ethtool API.
31 * Minor cleanups
32 *
33 * Daniel Podlejski <underley@underley.eu.org>
34 * Modifications for 2.3.99-pre5 kernel.
35 */
36
37 #define DRV_NAME "tun"
38 #define DRV_VERSION "1.6"
39 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
40 #define DRV_COPYRIGHT "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
41
42 #include <linux/module.h>
43 #include <linux/errno.h>
44 #include <linux/kernel.h>
45 #include <linux/major.h>
46 #include <linux/slab.h>
47 #include <linux/smp_lock.h>
48 #include <linux/poll.h>
49 #include <linux/fcntl.h>
50 #include <linux/init.h>
51 #include <linux/skbuff.h>
52 #include <linux/netdevice.h>
53 #include <linux/etherdevice.h>
54 #include <linux/miscdevice.h>
55 #include <linux/ethtool.h>
56 #include <linux/rtnetlink.h>
57 #include <linux/if.h>
58 #include <linux/if_arp.h>
59 #include <linux/if_ether.h>
60 #include <linux/if_tun.h>
61 #include <linux/crc32.h>
62 #include <linux/nsproxy.h>
63 #include <linux/virtio_net.h>
64 #include <net/net_namespace.h>
65 #include <net/netns/generic.h>
66 #include <net/rtnetlink.h>
67 #include <net/sock.h>
68
69 #include <asm/system.h>
70 #include <asm/uaccess.h>
71
72 /* Uncomment to enable debugging */
73 /* #define TUN_DEBUG 1 */
74
75 #ifdef TUN_DEBUG
76 static int debug;
77
78 #define DBG if(tun->debug)printk
79 #define DBG1 if(debug==2)printk
80 #else
81 #define DBG( a... )
82 #define DBG1( a... )
83 #endif
84
85 #define FLT_EXACT_COUNT 8
86 struct tap_filter {
87 unsigned int count; /* Number of addrs. Zero means disabled */
88 u32 mask[2]; /* Mask of the hashed addrs */
89 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
90 };
91
92 struct tun_file {
93 atomic_t count;
94 struct tun_struct *tun;
95 struct net *net;
96 };
97
98 struct tun_sock;
99
100 struct tun_struct {
101 struct tun_file *tfile;
102 unsigned int flags;
103 uid_t owner;
104 gid_t group;
105
106 struct sk_buff_head readq;
107
108 struct net_device *dev;
109 struct fasync_struct *fasync;
110
111 struct tap_filter txflt;
112 struct sock *sk;
113 struct socket socket;
114
115 #ifdef TUN_DEBUG
116 int debug;
117 #endif
118 };
119
120 struct tun_sock {
121 struct sock sk;
122 struct tun_struct *tun;
123 };
124
125 static inline struct tun_sock *tun_sk(struct sock *sk)
126 {
127 return container_of(sk, struct tun_sock, sk);
128 }
129
130 static int tun_attach(struct tun_struct *tun, struct file *file)
131 {
132 struct tun_file *tfile = file->private_data;
133 const struct cred *cred = current_cred();
134 int err;
135
136 ASSERT_RTNL();
137
138 /* Check permissions */
139 if (((tun->owner != -1 && cred->euid != tun->owner) ||
140 (tun->group != -1 && !in_egroup_p(tun->group))) &&
141 !capable(CAP_NET_ADMIN))
142 return -EPERM;
143
144 netif_tx_lock_bh(tun->dev);
145
146 err = -EINVAL;
147 if (tfile->tun)
148 goto out;
149
150 err = -EBUSY;
151 if (tun->tfile)
152 goto out;
153
154 err = 0;
155 tfile->tun = tun;
156 tun->tfile = tfile;
157 dev_hold(tun->dev);
158 sock_hold(tun->sk);
159 atomic_inc(&tfile->count);
160
161 out:
162 netif_tx_unlock_bh(tun->dev);
163 return err;
164 }
165
166 static void __tun_detach(struct tun_struct *tun)
167 {
168 /* Detach from net device */
169 netif_tx_lock_bh(tun->dev);
170 tun->tfile = NULL;
171 netif_tx_unlock_bh(tun->dev);
172
173 /* Drop read queue */
174 skb_queue_purge(&tun->readq);
175
176 /* Drop the extra count on the net device */
177 dev_put(tun->dev);
178 }
179
180 static void tun_detach(struct tun_struct *tun)
181 {
182 rtnl_lock();
183 __tun_detach(tun);
184 rtnl_unlock();
185 }
186
187 static struct tun_struct *__tun_get(struct tun_file *tfile)
188 {
189 struct tun_struct *tun = NULL;
190
191 if (atomic_inc_not_zero(&tfile->count))
192 tun = tfile->tun;
193
194 return tun;
195 }
196
197 static struct tun_struct *tun_get(struct file *file)
198 {
199 return __tun_get(file->private_data);
200 }
201
202 static void tun_put(struct tun_struct *tun)
203 {
204 struct tun_file *tfile = tun->tfile;
205
206 if (atomic_dec_and_test(&tfile->count))
207 tun_detach(tfile->tun);
208 }
209
210 /* TAP filterting */
211 static void addr_hash_set(u32 *mask, const u8 *addr)
212 {
213 int n = ether_crc(ETH_ALEN, addr) >> 26;
214 mask[n >> 5] |= (1 << (n & 31));
215 }
216
217 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
218 {
219 int n = ether_crc(ETH_ALEN, addr) >> 26;
220 return mask[n >> 5] & (1 << (n & 31));
221 }
222
223 static int update_filter(struct tap_filter *filter, void __user *arg)
224 {
225 struct { u8 u[ETH_ALEN]; } *addr;
226 struct tun_filter uf;
227 int err, alen, n, nexact;
228
229 if (copy_from_user(&uf, arg, sizeof(uf)))
230 return -EFAULT;
231
232 if (!uf.count) {
233 /* Disabled */
234 filter->count = 0;
235 return 0;
236 }
237
238 alen = ETH_ALEN * uf.count;
239 addr = kmalloc(alen, GFP_KERNEL);
240 if (!addr)
241 return -ENOMEM;
242
243 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
244 err = -EFAULT;
245 goto done;
246 }
247
248 /* The filter is updated without holding any locks. Which is
249 * perfectly safe. We disable it first and in the worst
250 * case we'll accept a few undesired packets. */
251 filter->count = 0;
252 wmb();
253
254 /* Use first set of addresses as an exact filter */
255 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
256 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
257
258 nexact = n;
259
260 /* Remaining multicast addresses are hashed,
261 * unicast will leave the filter disabled. */
262 memset(filter->mask, 0, sizeof(filter->mask));
263 for (; n < uf.count; n++) {
264 if (!is_multicast_ether_addr(addr[n].u)) {
265 err = 0; /* no filter */
266 goto done;
267 }
268 addr_hash_set(filter->mask, addr[n].u);
269 }
270
271 /* For ALLMULTI just set the mask to all ones.
272 * This overrides the mask populated above. */
273 if ((uf.flags & TUN_FLT_ALLMULTI))
274 memset(filter->mask, ~0, sizeof(filter->mask));
275
276 /* Now enable the filter */
277 wmb();
278 filter->count = nexact;
279
280 /* Return the number of exact filters */
281 err = nexact;
282
283 done:
284 kfree(addr);
285 return err;
286 }
287
288 /* Returns: 0 - drop, !=0 - accept */
289 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
290 {
291 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
292 * at this point. */
293 struct ethhdr *eh = (struct ethhdr *) skb->data;
294 int i;
295
296 /* Exact match */
297 for (i = 0; i < filter->count; i++)
298 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
299 return 1;
300
301 /* Inexact match (multicast only) */
302 if (is_multicast_ether_addr(eh->h_dest))
303 return addr_hash_test(filter->mask, eh->h_dest);
304
305 return 0;
306 }
307
308 /*
309 * Checks whether the packet is accepted or not.
310 * Returns: 0 - drop, !=0 - accept
311 */
312 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
313 {
314 if (!filter->count)
315 return 1;
316
317 return run_filter(filter, skb);
318 }
319
320 /* Network device part of the driver */
321
322 static const struct ethtool_ops tun_ethtool_ops;
323
324 /* Net device detach from fd. */
325 static void tun_net_uninit(struct net_device *dev)
326 {
327 struct tun_struct *tun = netdev_priv(dev);
328 struct tun_file *tfile = tun->tfile;
329
330 /* Inform the methods they need to stop using the dev.
331 */
332 if (tfile) {
333 wake_up_all(&tun->socket.wait);
334 if (atomic_dec_and_test(&tfile->count))
335 __tun_detach(tun);
336 }
337 }
338
339 static void tun_free_netdev(struct net_device *dev)
340 {
341 struct tun_struct *tun = netdev_priv(dev);
342
343 sock_put(tun->sk);
344 }
345
346 /* Net device open. */
347 static int tun_net_open(struct net_device *dev)
348 {
349 netif_start_queue(dev);
350 return 0;
351 }
352
353 /* Net device close. */
354 static int tun_net_close(struct net_device *dev)
355 {
356 netif_stop_queue(dev);
357 return 0;
358 }
359
360 /* Net device start xmit */
361 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
362 {
363 struct tun_struct *tun = netdev_priv(dev);
364
365 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
366
367 /* Drop packet if interface is not attached */
368 if (!tun->tfile)
369 goto drop;
370
371 /* Drop if the filter does not like it.
372 * This is a noop if the filter is disabled.
373 * Filter can be enabled only for the TAP devices. */
374 if (!check_filter(&tun->txflt, skb))
375 goto drop;
376
377 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
378 if (!(tun->flags & TUN_ONE_QUEUE)) {
379 /* Normal queueing mode. */
380 /* Packet scheduler handles dropping of further packets. */
381 netif_stop_queue(dev);
382
383 /* We won't see all dropped packets individually, so overrun
384 * error is more appropriate. */
385 dev->stats.tx_fifo_errors++;
386 } else {
387 /* Single queue mode.
388 * Driver handles dropping of all packets itself. */
389 goto drop;
390 }
391 }
392
393 /* Enqueue packet */
394 skb_queue_tail(&tun->readq, skb);
395 dev->trans_start = jiffies;
396
397 /* Notify and wake up reader process */
398 if (tun->flags & TUN_FASYNC)
399 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
400 wake_up_interruptible(&tun->socket.wait);
401 return 0;
402
403 drop:
404 dev->stats.tx_dropped++;
405 kfree_skb(skb);
406 return 0;
407 }
408
409 static void tun_net_mclist(struct net_device *dev)
410 {
411 /*
412 * This callback is supposed to deal with mc filter in
413 * _rx_ path and has nothing to do with the _tx_ path.
414 * In rx path we always accept everything userspace gives us.
415 */
416 return;
417 }
418
419 #define MIN_MTU 68
420 #define MAX_MTU 65535
421
422 static int
423 tun_net_change_mtu(struct net_device *dev, int new_mtu)
424 {
425 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
426 return -EINVAL;
427 dev->mtu = new_mtu;
428 return 0;
429 }
430
431 static const struct net_device_ops tun_netdev_ops = {
432 .ndo_uninit = tun_net_uninit,
433 .ndo_open = tun_net_open,
434 .ndo_stop = tun_net_close,
435 .ndo_start_xmit = tun_net_xmit,
436 .ndo_change_mtu = tun_net_change_mtu,
437 };
438
439 static const struct net_device_ops tap_netdev_ops = {
440 .ndo_uninit = tun_net_uninit,
441 .ndo_open = tun_net_open,
442 .ndo_stop = tun_net_close,
443 .ndo_start_xmit = tun_net_xmit,
444 .ndo_change_mtu = tun_net_change_mtu,
445 .ndo_set_multicast_list = tun_net_mclist,
446 .ndo_set_mac_address = eth_mac_addr,
447 .ndo_validate_addr = eth_validate_addr,
448 };
449
450 /* Initialize net device. */
451 static void tun_net_init(struct net_device *dev)
452 {
453 struct tun_struct *tun = netdev_priv(dev);
454
455 switch (tun->flags & TUN_TYPE_MASK) {
456 case TUN_TUN_DEV:
457 dev->netdev_ops = &tun_netdev_ops;
458
459 /* Point-to-Point TUN Device */
460 dev->hard_header_len = 0;
461 dev->addr_len = 0;
462 dev->mtu = 1500;
463
464 /* Zero header length */
465 dev->type = ARPHRD_NONE;
466 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
467 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
468 break;
469
470 case TUN_TAP_DEV:
471 dev->netdev_ops = &tap_netdev_ops;
472 /* Ethernet TAP Device */
473 ether_setup(dev);
474
475 random_ether_addr(dev->dev_addr);
476
477 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
478 break;
479 }
480 }
481
482 /* Character device part */
483
484 /* Poll */
485 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
486 {
487 struct tun_file *tfile = file->private_data;
488 struct tun_struct *tun = __tun_get(tfile);
489 struct sock *sk = tun->sk;
490 unsigned int mask = 0;
491
492 if (!tun)
493 return POLLERR;
494
495 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
496
497 poll_wait(file, &tun->socket.wait, wait);
498
499 if (!skb_queue_empty(&tun->readq))
500 mask |= POLLIN | POLLRDNORM;
501
502 if (sock_writeable(sk) ||
503 (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) &&
504 sock_writeable(sk)))
505 mask |= POLLOUT | POLLWRNORM;
506
507 if (tun->dev->reg_state != NETREG_REGISTERED)
508 mask = POLLERR;
509
510 tun_put(tun);
511 return mask;
512 }
513
514 /* prepad is the amount to reserve at front. len is length after that.
515 * linear is a hint as to how much to copy (usually headers). */
516 static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun,
517 size_t prepad, size_t len,
518 size_t linear, int noblock)
519 {
520 struct sock *sk = tun->sk;
521 struct sk_buff *skb;
522 int err;
523
524 /* Under a page? Don't bother with paged skb. */
525 if (prepad + len < PAGE_SIZE || !linear)
526 linear = len;
527
528 skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock,
529 &err);
530 if (!skb)
531 return ERR_PTR(err);
532
533 skb_reserve(skb, prepad);
534 skb_put(skb, linear);
535 skb->data_len = len - linear;
536 skb->len += len - linear;
537
538 return skb;
539 }
540
541 /* Get packet from user space buffer */
542 static __inline__ ssize_t tun_get_user(struct tun_struct *tun,
543 const struct iovec *iv, size_t count,
544 int noblock)
545 {
546 struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) };
547 struct sk_buff *skb;
548 size_t len = count, align = 0;
549 struct virtio_net_hdr gso = { 0 };
550 int offset = 0;
551
552 if (!(tun->flags & TUN_NO_PI)) {
553 if ((len -= sizeof(pi)) > count)
554 return -EINVAL;
555
556 if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi)))
557 return -EFAULT;
558 offset += sizeof(pi);
559 }
560
561 if (tun->flags & TUN_VNET_HDR) {
562 if ((len -= sizeof(gso)) > count)
563 return -EINVAL;
564
565 if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso)))
566 return -EFAULT;
567
568 if (gso.hdr_len > len)
569 return -EINVAL;
570 offset += sizeof(pi);
571 }
572
573 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
574 align = NET_IP_ALIGN;
575 if (unlikely(len < ETH_HLEN ||
576 (gso.hdr_len && gso.hdr_len < ETH_HLEN)))
577 return -EINVAL;
578 }
579
580 skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock);
581 if (IS_ERR(skb)) {
582 if (PTR_ERR(skb) != -EAGAIN)
583 tun->dev->stats.rx_dropped++;
584 return PTR_ERR(skb);
585 }
586
587 if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) {
588 tun->dev->stats.rx_dropped++;
589 kfree_skb(skb);
590 return -EFAULT;
591 }
592
593 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
594 if (!skb_partial_csum_set(skb, gso.csum_start,
595 gso.csum_offset)) {
596 tun->dev->stats.rx_frame_errors++;
597 kfree_skb(skb);
598 return -EINVAL;
599 }
600 } else if (tun->flags & TUN_NOCHECKSUM)
601 skb->ip_summed = CHECKSUM_UNNECESSARY;
602
603 switch (tun->flags & TUN_TYPE_MASK) {
604 case TUN_TUN_DEV:
605 if (tun->flags & TUN_NO_PI) {
606 switch (skb->data[0] & 0xf0) {
607 case 0x40:
608 pi.proto = htons(ETH_P_IP);
609 break;
610 case 0x60:
611 pi.proto = htons(ETH_P_IPV6);
612 break;
613 default:
614 tun->dev->stats.rx_dropped++;
615 kfree_skb(skb);
616 return -EINVAL;
617 }
618 }
619
620 skb_reset_mac_header(skb);
621 skb->protocol = pi.proto;
622 skb->dev = tun->dev;
623 break;
624 case TUN_TAP_DEV:
625 skb->protocol = eth_type_trans(skb, tun->dev);
626 break;
627 };
628
629 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
630 pr_debug("GSO!\n");
631 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
632 case VIRTIO_NET_HDR_GSO_TCPV4:
633 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
634 break;
635 case VIRTIO_NET_HDR_GSO_TCPV6:
636 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
637 break;
638 default:
639 tun->dev->stats.rx_frame_errors++;
640 kfree_skb(skb);
641 return -EINVAL;
642 }
643
644 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
645 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
646
647 skb_shinfo(skb)->gso_size = gso.gso_size;
648 if (skb_shinfo(skb)->gso_size == 0) {
649 tun->dev->stats.rx_frame_errors++;
650 kfree_skb(skb);
651 return -EINVAL;
652 }
653
654 /* Header must be checked, and gso_segs computed. */
655 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
656 skb_shinfo(skb)->gso_segs = 0;
657 }
658
659 netif_rx_ni(skb);
660
661 tun->dev->stats.rx_packets++;
662 tun->dev->stats.rx_bytes += len;
663
664 return count;
665 }
666
667 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
668 unsigned long count, loff_t pos)
669 {
670 struct file *file = iocb->ki_filp;
671 struct tun_struct *tun = tun_get(file);
672 ssize_t result;
673
674 if (!tun)
675 return -EBADFD;
676
677 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
678
679 result = tun_get_user(tun, iv, iov_length(iv, count),
680 file->f_flags & O_NONBLOCK);
681
682 tun_put(tun);
683 return result;
684 }
685
686 /* Put packet to the user space buffer */
687 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
688 struct sk_buff *skb,
689 const struct iovec *iv, int len)
690 {
691 struct tun_pi pi = { 0, skb->protocol };
692 ssize_t total = 0;
693
694 if (!(tun->flags & TUN_NO_PI)) {
695 if ((len -= sizeof(pi)) < 0)
696 return -EINVAL;
697
698 if (len < skb->len) {
699 /* Packet will be striped */
700 pi.flags |= TUN_PKT_STRIP;
701 }
702
703 if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi)))
704 return -EFAULT;
705 total += sizeof(pi);
706 }
707
708 if (tun->flags & TUN_VNET_HDR) {
709 struct virtio_net_hdr gso = { 0 }; /* no info leak */
710 if ((len -= sizeof(gso)) < 0)
711 return -EINVAL;
712
713 if (skb_is_gso(skb)) {
714 struct skb_shared_info *sinfo = skb_shinfo(skb);
715
716 /* This is a hint as to how much should be linear. */
717 gso.hdr_len = skb_headlen(skb);
718 gso.gso_size = sinfo->gso_size;
719 if (sinfo->gso_type & SKB_GSO_TCPV4)
720 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
721 else if (sinfo->gso_type & SKB_GSO_TCPV6)
722 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
723 else
724 BUG();
725 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
726 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
727 } else
728 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
729
730 if (skb->ip_summed == CHECKSUM_PARTIAL) {
731 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
732 gso.csum_start = skb->csum_start - skb_headroom(skb);
733 gso.csum_offset = skb->csum_offset;
734 } /* else everything is zero */
735
736 if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
737 sizeof(gso))))
738 return -EFAULT;
739 total += sizeof(gso);
740 }
741
742 len = min_t(int, skb->len, len);
743
744 skb_copy_datagram_const_iovec(skb, 0, iv, total, len);
745 total += len;
746
747 tun->dev->stats.tx_packets++;
748 tun->dev->stats.tx_bytes += len;
749
750 return total;
751 }
752
753 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
754 unsigned long count, loff_t pos)
755 {
756 struct file *file = iocb->ki_filp;
757 struct tun_file *tfile = file->private_data;
758 struct tun_struct *tun = __tun_get(tfile);
759 DECLARE_WAITQUEUE(wait, current);
760 struct sk_buff *skb;
761 ssize_t len, ret = 0;
762
763 if (!tun)
764 return -EBADFD;
765
766 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
767
768 len = iov_length(iv, count);
769 if (len < 0) {
770 ret = -EINVAL;
771 goto out;
772 }
773
774 add_wait_queue(&tun->socket.wait, &wait);
775 while (len) {
776 current->state = TASK_INTERRUPTIBLE;
777
778 /* Read frames from the queue */
779 if (!(skb=skb_dequeue(&tun->readq))) {
780 if (file->f_flags & O_NONBLOCK) {
781 ret = -EAGAIN;
782 break;
783 }
784 if (signal_pending(current)) {
785 ret = -ERESTARTSYS;
786 break;
787 }
788 if (tun->dev->reg_state != NETREG_REGISTERED) {
789 ret = -EIO;
790 break;
791 }
792
793 /* Nothing to read, let's sleep */
794 schedule();
795 continue;
796 }
797 netif_wake_queue(tun->dev);
798
799 ret = tun_put_user(tun, skb, iv, len);
800 kfree_skb(skb);
801 break;
802 }
803
804 current->state = TASK_RUNNING;
805 remove_wait_queue(&tun->socket.wait, &wait);
806
807 out:
808 tun_put(tun);
809 return ret;
810 }
811
812 static void tun_setup(struct net_device *dev)
813 {
814 struct tun_struct *tun = netdev_priv(dev);
815
816 skb_queue_head_init(&tun->readq);
817
818 tun->owner = -1;
819 tun->group = -1;
820
821 dev->ethtool_ops = &tun_ethtool_ops;
822 dev->destructor = tun_free_netdev;
823 }
824
825 /* Trivial set of netlink ops to allow deleting tun or tap
826 * device with netlink.
827 */
828 static int tun_validate(struct nlattr *tb[], struct nlattr *data[])
829 {
830 return -EINVAL;
831 }
832
833 static struct rtnl_link_ops tun_link_ops __read_mostly = {
834 .kind = DRV_NAME,
835 .priv_size = sizeof(struct tun_struct),
836 .setup = tun_setup,
837 .validate = tun_validate,
838 };
839
840 static void tun_sock_write_space(struct sock *sk)
841 {
842 struct tun_struct *tun;
843
844 if (!sock_writeable(sk))
845 return;
846
847 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
848 wake_up_interruptible_sync(sk->sk_sleep);
849
850 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags))
851 return;
852
853 tun = container_of(sk, struct tun_sock, sk)->tun;
854 kill_fasync(&tun->fasync, SIGIO, POLL_OUT);
855 }
856
857 static void tun_sock_destruct(struct sock *sk)
858 {
859 free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev);
860 }
861
862 static struct proto tun_proto = {
863 .name = "tun",
864 .owner = THIS_MODULE,
865 .obj_size = sizeof(struct tun_sock),
866 };
867
868 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
869 {
870 struct sock *sk;
871 struct tun_struct *tun;
872 struct net_device *dev;
873 int err;
874
875 dev = __dev_get_by_name(net, ifr->ifr_name);
876 if (dev) {
877 if (ifr->ifr_flags & IFF_TUN_EXCL)
878 return -EBUSY;
879 if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops)
880 tun = netdev_priv(dev);
881 else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops)
882 tun = netdev_priv(dev);
883 else
884 return -EINVAL;
885
886 err = tun_attach(tun, file);
887 if (err < 0)
888 return err;
889 }
890 else {
891 char *name;
892 unsigned long flags = 0;
893
894 err = -EINVAL;
895
896 if (!capable(CAP_NET_ADMIN))
897 return -EPERM;
898
899 /* Set dev type */
900 if (ifr->ifr_flags & IFF_TUN) {
901 /* TUN device */
902 flags |= TUN_TUN_DEV;
903 name = "tun%d";
904 } else if (ifr->ifr_flags & IFF_TAP) {
905 /* TAP device */
906 flags |= TUN_TAP_DEV;
907 name = "tap%d";
908 } else
909 goto failed;
910
911 if (*ifr->ifr_name)
912 name = ifr->ifr_name;
913
914 dev = alloc_netdev(sizeof(struct tun_struct), name,
915 tun_setup);
916 if (!dev)
917 return -ENOMEM;
918
919 dev_net_set(dev, net);
920 dev->rtnl_link_ops = &tun_link_ops;
921
922 tun = netdev_priv(dev);
923 tun->dev = dev;
924 tun->flags = flags;
925 tun->txflt.count = 0;
926
927 err = -ENOMEM;
928 sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto);
929 if (!sk)
930 goto err_free_dev;
931
932 init_waitqueue_head(&tun->socket.wait);
933 sock_init_data(&tun->socket, sk);
934 sk->sk_write_space = tun_sock_write_space;
935 sk->sk_sndbuf = INT_MAX;
936
937 tun->sk = sk;
938 container_of(sk, struct tun_sock, sk)->tun = tun;
939
940 tun_net_init(dev);
941
942 if (strchr(dev->name, '%')) {
943 err = dev_alloc_name(dev, dev->name);
944 if (err < 0)
945 goto err_free_sk;
946 }
947
948 err = -EINVAL;
949 err = register_netdevice(tun->dev);
950 if (err < 0)
951 goto err_free_sk;
952
953 sk->sk_destruct = tun_sock_destruct;
954
955 err = tun_attach(tun, file);
956 if (err < 0)
957 goto failed;
958 }
959
960 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
961
962 if (ifr->ifr_flags & IFF_NO_PI)
963 tun->flags |= TUN_NO_PI;
964 else
965 tun->flags &= ~TUN_NO_PI;
966
967 if (ifr->ifr_flags & IFF_ONE_QUEUE)
968 tun->flags |= TUN_ONE_QUEUE;
969 else
970 tun->flags &= ~TUN_ONE_QUEUE;
971
972 if (ifr->ifr_flags & IFF_VNET_HDR)
973 tun->flags |= TUN_VNET_HDR;
974 else
975 tun->flags &= ~TUN_VNET_HDR;
976
977 /* Make sure persistent devices do not get stuck in
978 * xoff state.
979 */
980 if (netif_running(tun->dev))
981 netif_wake_queue(tun->dev);
982
983 strcpy(ifr->ifr_name, tun->dev->name);
984 return 0;
985
986 err_free_sk:
987 sock_put(sk);
988 err_free_dev:
989 free_netdev(dev);
990 failed:
991 return err;
992 }
993
994 static int tun_get_iff(struct net *net, struct file *file, struct ifreq *ifr)
995 {
996 struct tun_struct *tun = tun_get(file);
997
998 if (!tun)
999 return -EBADFD;
1000
1001 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
1002
1003 strcpy(ifr->ifr_name, tun->dev->name);
1004
1005 ifr->ifr_flags = 0;
1006
1007 if (ifr->ifr_flags & TUN_TUN_DEV)
1008 ifr->ifr_flags |= IFF_TUN;
1009 else
1010 ifr->ifr_flags |= IFF_TAP;
1011
1012 if (tun->flags & TUN_NO_PI)
1013 ifr->ifr_flags |= IFF_NO_PI;
1014
1015 if (tun->flags & TUN_ONE_QUEUE)
1016 ifr->ifr_flags |= IFF_ONE_QUEUE;
1017
1018 if (tun->flags & TUN_VNET_HDR)
1019 ifr->ifr_flags |= IFF_VNET_HDR;
1020
1021 tun_put(tun);
1022 return 0;
1023 }
1024
1025 /* This is like a cut-down ethtool ops, except done via tun fd so no
1026 * privs required. */
1027 static int set_offload(struct net_device *dev, unsigned long arg)
1028 {
1029 unsigned int old_features, features;
1030
1031 old_features = dev->features;
1032 /* Unset features, set them as we chew on the arg. */
1033 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
1034 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
1035
1036 if (arg & TUN_F_CSUM) {
1037 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
1038 arg &= ~TUN_F_CSUM;
1039
1040 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
1041 if (arg & TUN_F_TSO_ECN) {
1042 features |= NETIF_F_TSO_ECN;
1043 arg &= ~TUN_F_TSO_ECN;
1044 }
1045 if (arg & TUN_F_TSO4)
1046 features |= NETIF_F_TSO;
1047 if (arg & TUN_F_TSO6)
1048 features |= NETIF_F_TSO6;
1049 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
1050 }
1051 }
1052
1053 /* This gives the user a way to test for new features in future by
1054 * trying to set them. */
1055 if (arg)
1056 return -EINVAL;
1057
1058 dev->features = features;
1059 if (old_features != dev->features)
1060 netdev_features_change(dev);
1061
1062 return 0;
1063 }
1064
1065 static int tun_chr_ioctl(struct inode *inode, struct file *file,
1066 unsigned int cmd, unsigned long arg)
1067 {
1068 struct tun_file *tfile = file->private_data;
1069 struct tun_struct *tun;
1070 void __user* argp = (void __user*)arg;
1071 struct ifreq ifr;
1072 int sndbuf;
1073 int ret;
1074
1075 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
1076 if (copy_from_user(&ifr, argp, sizeof ifr))
1077 return -EFAULT;
1078
1079 if (cmd == TUNGETFEATURES) {
1080 /* Currently this just means: "what IFF flags are valid?".
1081 * This is needed because we never checked for invalid flags on
1082 * TUNSETIFF. */
1083 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
1084 IFF_VNET_HDR,
1085 (unsigned int __user*)argp);
1086 }
1087
1088 tun = __tun_get(tfile);
1089 if (cmd == TUNSETIFF && !tun) {
1090 int err;
1091
1092 ifr.ifr_name[IFNAMSIZ-1] = '\0';
1093
1094 rtnl_lock();
1095 err = tun_set_iff(tfile->net, file, &ifr);
1096 rtnl_unlock();
1097
1098 if (err)
1099 return err;
1100
1101 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1102 return -EFAULT;
1103 return 0;
1104 }
1105
1106
1107 if (!tun)
1108 return -EBADFD;
1109
1110 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
1111
1112 ret = 0;
1113 switch (cmd) {
1114 case TUNGETIFF:
1115 ret = tun_get_iff(current->nsproxy->net_ns, file, &ifr);
1116 if (ret)
1117 break;
1118
1119 if (copy_to_user(argp, &ifr, sizeof(ifr)))
1120 ret = -EFAULT;
1121 break;
1122
1123 case TUNSETNOCSUM:
1124 /* Disable/Enable checksum */
1125 if (arg)
1126 tun->flags |= TUN_NOCHECKSUM;
1127 else
1128 tun->flags &= ~TUN_NOCHECKSUM;
1129
1130 DBG(KERN_INFO "%s: checksum %s\n",
1131 tun->dev->name, arg ? "disabled" : "enabled");
1132 break;
1133
1134 case TUNSETPERSIST:
1135 /* Disable/Enable persist mode */
1136 if (arg)
1137 tun->flags |= TUN_PERSIST;
1138 else
1139 tun->flags &= ~TUN_PERSIST;
1140
1141 DBG(KERN_INFO "%s: persist %s\n",
1142 tun->dev->name, arg ? "enabled" : "disabled");
1143 break;
1144
1145 case TUNSETOWNER:
1146 /* Set owner of the device */
1147 tun->owner = (uid_t) arg;
1148
1149 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
1150 break;
1151
1152 case TUNSETGROUP:
1153 /* Set group of the device */
1154 tun->group= (gid_t) arg;
1155
1156 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
1157 break;
1158
1159 case TUNSETLINK:
1160 /* Only allow setting the type when the interface is down */
1161 rtnl_lock();
1162 if (tun->dev->flags & IFF_UP) {
1163 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
1164 tun->dev->name);
1165 ret = -EBUSY;
1166 } else {
1167 tun->dev->type = (int) arg;
1168 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
1169 ret = 0;
1170 }
1171 rtnl_unlock();
1172 break;
1173
1174 #ifdef TUN_DEBUG
1175 case TUNSETDEBUG:
1176 tun->debug = arg;
1177 break;
1178 #endif
1179 case TUNSETOFFLOAD:
1180 rtnl_lock();
1181 ret = set_offload(tun->dev, arg);
1182 rtnl_unlock();
1183 break;
1184
1185 case TUNSETTXFILTER:
1186 /* Can be set only for TAPs */
1187 ret = -EINVAL;
1188 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1189 break;
1190 rtnl_lock();
1191 ret = update_filter(&tun->txflt, (void __user *)arg);
1192 rtnl_unlock();
1193 break;
1194
1195 case SIOCGIFHWADDR:
1196 /* Get hw addres */
1197 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1198 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1199 if (copy_to_user(argp, &ifr, sizeof ifr))
1200 ret = -EFAULT;
1201 break;
1202
1203 case SIOCSIFHWADDR:
1204 /* Set hw address */
1205 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1206 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1207
1208 rtnl_lock();
1209 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1210 rtnl_unlock();
1211 break;
1212
1213 case TUNGETSNDBUF:
1214 sndbuf = tun->sk->sk_sndbuf;
1215 if (copy_to_user(argp, &sndbuf, sizeof(sndbuf)))
1216 ret = -EFAULT;
1217 break;
1218
1219 case TUNSETSNDBUF:
1220 if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) {
1221 ret = -EFAULT;
1222 break;
1223 }
1224
1225 tun->sk->sk_sndbuf = sndbuf;
1226 break;
1227
1228 default:
1229 ret = -EINVAL;
1230 break;
1231 };
1232
1233 tun_put(tun);
1234 return ret;
1235 }
1236
1237 static int tun_chr_fasync(int fd, struct file *file, int on)
1238 {
1239 struct tun_struct *tun = tun_get(file);
1240 int ret;
1241
1242 if (!tun)
1243 return -EBADFD;
1244
1245 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1246
1247 lock_kernel();
1248 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1249 goto out;
1250
1251 if (on) {
1252 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1253 if (ret)
1254 goto out;
1255 tun->flags |= TUN_FASYNC;
1256 } else
1257 tun->flags &= ~TUN_FASYNC;
1258 ret = 0;
1259 out:
1260 unlock_kernel();
1261 tun_put(tun);
1262 return ret;
1263 }
1264
1265 static int tun_chr_open(struct inode *inode, struct file * file)
1266 {
1267 struct tun_file *tfile;
1268 cycle_kernel_lock();
1269 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1270
1271 tfile = kmalloc(sizeof(*tfile), GFP_KERNEL);
1272 if (!tfile)
1273 return -ENOMEM;
1274 atomic_set(&tfile->count, 0);
1275 tfile->tun = NULL;
1276 tfile->net = get_net(current->nsproxy->net_ns);
1277 file->private_data = tfile;
1278 return 0;
1279 }
1280
1281 static int tun_chr_close(struct inode *inode, struct file *file)
1282 {
1283 struct tun_file *tfile = file->private_data;
1284 struct tun_struct *tun = __tun_get(tfile);
1285
1286
1287 if (tun) {
1288 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
1289
1290 rtnl_lock();
1291 __tun_detach(tun);
1292
1293 /* If desireable, unregister the netdevice. */
1294 if (!(tun->flags & TUN_PERSIST))
1295 unregister_netdevice(tun->dev);
1296
1297 rtnl_unlock();
1298 }
1299
1300 tun = tfile->tun;
1301 if (tun)
1302 sock_put(tun->sk);
1303
1304 put_net(tfile->net);
1305 kfree(tfile);
1306
1307 return 0;
1308 }
1309
1310 static const struct file_operations tun_fops = {
1311 .owner = THIS_MODULE,
1312 .llseek = no_llseek,
1313 .read = do_sync_read,
1314 .aio_read = tun_chr_aio_read,
1315 .write = do_sync_write,
1316 .aio_write = tun_chr_aio_write,
1317 .poll = tun_chr_poll,
1318 .ioctl = tun_chr_ioctl,
1319 .open = tun_chr_open,
1320 .release = tun_chr_close,
1321 .fasync = tun_chr_fasync
1322 };
1323
1324 static struct miscdevice tun_miscdev = {
1325 .minor = TUN_MINOR,
1326 .name = "tun",
1327 .fops = &tun_fops,
1328 };
1329
1330 /* ethtool interface */
1331
1332 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1333 {
1334 cmd->supported = 0;
1335 cmd->advertising = 0;
1336 cmd->speed = SPEED_10;
1337 cmd->duplex = DUPLEX_FULL;
1338 cmd->port = PORT_TP;
1339 cmd->phy_address = 0;
1340 cmd->transceiver = XCVR_INTERNAL;
1341 cmd->autoneg = AUTONEG_DISABLE;
1342 cmd->maxtxpkt = 0;
1343 cmd->maxrxpkt = 0;
1344 return 0;
1345 }
1346
1347 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1348 {
1349 struct tun_struct *tun = netdev_priv(dev);
1350
1351 strcpy(info->driver, DRV_NAME);
1352 strcpy(info->version, DRV_VERSION);
1353 strcpy(info->fw_version, "N/A");
1354
1355 switch (tun->flags & TUN_TYPE_MASK) {
1356 case TUN_TUN_DEV:
1357 strcpy(info->bus_info, "tun");
1358 break;
1359 case TUN_TAP_DEV:
1360 strcpy(info->bus_info, "tap");
1361 break;
1362 }
1363 }
1364
1365 static u32 tun_get_msglevel(struct net_device *dev)
1366 {
1367 #ifdef TUN_DEBUG
1368 struct tun_struct *tun = netdev_priv(dev);
1369 return tun->debug;
1370 #else
1371 return -EOPNOTSUPP;
1372 #endif
1373 }
1374
1375 static void tun_set_msglevel(struct net_device *dev, u32 value)
1376 {
1377 #ifdef TUN_DEBUG
1378 struct tun_struct *tun = netdev_priv(dev);
1379 tun->debug = value;
1380 #endif
1381 }
1382
1383 static u32 tun_get_link(struct net_device *dev)
1384 {
1385 struct tun_struct *tun = netdev_priv(dev);
1386 return !!tun->tfile;
1387 }
1388
1389 static u32 tun_get_rx_csum(struct net_device *dev)
1390 {
1391 struct tun_struct *tun = netdev_priv(dev);
1392 return (tun->flags & TUN_NOCHECKSUM) == 0;
1393 }
1394
1395 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1396 {
1397 struct tun_struct *tun = netdev_priv(dev);
1398 if (data)
1399 tun->flags &= ~TUN_NOCHECKSUM;
1400 else
1401 tun->flags |= TUN_NOCHECKSUM;
1402 return 0;
1403 }
1404
1405 static const struct ethtool_ops tun_ethtool_ops = {
1406 .get_settings = tun_get_settings,
1407 .get_drvinfo = tun_get_drvinfo,
1408 .get_msglevel = tun_get_msglevel,
1409 .set_msglevel = tun_set_msglevel,
1410 .get_link = tun_get_link,
1411 .get_rx_csum = tun_get_rx_csum,
1412 .set_rx_csum = tun_set_rx_csum
1413 };
1414
1415
1416 static int __init tun_init(void)
1417 {
1418 int ret = 0;
1419
1420 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1421 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1422
1423 ret = rtnl_link_register(&tun_link_ops);
1424 if (ret) {
1425 printk(KERN_ERR "tun: Can't register link_ops\n");
1426 goto err_linkops;
1427 }
1428
1429 ret = misc_register(&tun_miscdev);
1430 if (ret) {
1431 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1432 goto err_misc;
1433 }
1434 return 0;
1435 err_misc:
1436 rtnl_link_unregister(&tun_link_ops);
1437 err_linkops:
1438 return ret;
1439 }
1440
1441 static void tun_cleanup(void)
1442 {
1443 misc_deregister(&tun_miscdev);
1444 rtnl_link_unregister(&tun_link_ops);
1445 }
1446
1447 module_init(tun_init);
1448 module_exit(tun_cleanup);
1449 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1450 MODULE_AUTHOR(DRV_COPYRIGHT);
1451 MODULE_LICENSE("GPL");
1452 MODULE_ALIAS_MISCDEV(TUN_MINOR);
Linux kernel - Deliverables
This page took 0.085834 seconds and 6 git commands to generate.