?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
iwlwifi: iwl_poll_{direct_}bit cleanup
[linux-2.6/verdex.git] / drivers / net / tun.c
blobfd0b11ea5562fe12cd7b2e517470010bae77f0bd
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
67 #include <asm/system.h>
68 #include <asm/uaccess.h>
69
70 /* Uncomment to enable debugging */
71 /* #define TUN_DEBUG 1 */
72
73 #ifdef TUN_DEBUG
74 static int debug;
75
76 #define DBG if(tun->debug)printk
77 #define DBG1 if(debug==2)printk
78 #else
79 #define DBG( a... )
80 #define DBG1( a... )
81 #endif
82
83 #define FLT_EXACT_COUNT 8
84 struct tap_filter {
85 unsigned int count; /* Number of addrs. Zero means disabled */
86 u32 mask[2]; /* Mask of the hashed addrs */
87 unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN];
88 };
89
90 struct tun_struct {
91 struct list_head list;
92 unsigned int flags;
93 int attached;
94 uid_t owner;
95 gid_t group;
96
97 wait_queue_head_t read_wait;
98 struct sk_buff_head readq;
99
100 struct net_device *dev;
101 struct fasync_struct *fasync;
102
103 struct tap_filter txflt;
104
105 #ifdef TUN_DEBUG
106 int debug;
107 #endif
108 };
109
110 /* TAP filterting */
111 static void addr_hash_set(u32 *mask, const u8 *addr)
112 {
113 int n = ether_crc(ETH_ALEN, addr) >> 26;
114 mask[n >> 5] |= (1 << (n & 31));
115 }
116
117 static unsigned int addr_hash_test(const u32 *mask, const u8 *addr)
118 {
119 int n = ether_crc(ETH_ALEN, addr) >> 26;
120 return mask[n >> 5] & (1 << (n & 31));
121 }
122
123 static int update_filter(struct tap_filter *filter, void __user *arg)
124 {
125 struct { u8 u[ETH_ALEN]; } *addr;
126 struct tun_filter uf;
127 int err, alen, n, nexact;
128
129 if (copy_from_user(&uf, arg, sizeof(uf)))
130 return -EFAULT;
131
132 if (!uf.count) {
133 /* Disabled */
134 filter->count = 0;
135 return 0;
136 }
137
138 alen = ETH_ALEN * uf.count;
139 addr = kmalloc(alen, GFP_KERNEL);
140 if (!addr)
141 return -ENOMEM;
142
143 if (copy_from_user(addr, arg + sizeof(uf), alen)) {
144 err = -EFAULT;
145 goto done;
146 }
147
148 /* The filter is updated without holding any locks. Which is
149 * perfectly safe. We disable it first and in the worst
150 * case we'll accept a few undesired packets. */
151 filter->count = 0;
152 wmb();
153
154 /* Use first set of addresses as an exact filter */
155 for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++)
156 memcpy(filter->addr[n], addr[n].u, ETH_ALEN);
157
158 nexact = n;
159
160 /* The rest is hashed */
161 memset(filter->mask, 0, sizeof(filter->mask));
162 for (; n < uf.count; n++)
163 addr_hash_set(filter->mask, addr[n].u);
164
165 /* For ALLMULTI just set the mask to all ones.
166 * This overrides the mask populated above. */
167 if ((uf.flags & TUN_FLT_ALLMULTI))
168 memset(filter->mask, ~0, sizeof(filter->mask));
169
170 /* Now enable the filter */
171 wmb();
172 filter->count = nexact;
173
174 /* Return the number of exact filters */
175 err = nexact;
176
177 done:
178 kfree(addr);
179 return err;
180 }
181
182 /* Returns: 0 - drop, !=0 - accept */
183 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
184 {
185 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
186 * at this point. */
187 struct ethhdr *eh = (struct ethhdr *) skb->data;
188 int i;
189
190 /* Exact match */
191 for (i = 0; i < filter->count; i++)
192 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
193 return 1;
194
195 /* Inexact match (multicast only) */
196 if (is_multicast_ether_addr(eh->h_dest))
197 return addr_hash_test(filter->mask, eh->h_dest);
198
199 return 0;
200 }
201
202 /*
203 * Checks whether the packet is accepted or not.
204 * Returns: 0 - drop, !=0 - accept
205 */
206 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
207 {
208 if (!filter->count)
209 return 1;
210
211 return run_filter(filter, skb);
212 }
213
214 /* Network device part of the driver */
215
216 static unsigned int tun_net_id;
217 struct tun_net {
218 struct list_head dev_list;
219 };
220
221 static const struct ethtool_ops tun_ethtool_ops;
222
223 /* Net device open. */
224 static int tun_net_open(struct net_device *dev)
225 {
226 netif_start_queue(dev);
227 return 0;
228 }
229
230 /* Net device close. */
231 static int tun_net_close(struct net_device *dev)
232 {
233 netif_stop_queue(dev);
234 return 0;
235 }
236
237 /* Net device start xmit */
238 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
239 {
240 struct tun_struct *tun = netdev_priv(dev);
241
242 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
243
244 /* Drop packet if interface is not attached */
245 if (!tun->attached)
246 goto drop;
247
248 /* Drop if the filter does not like it.
249 * This is a noop if the filter is disabled.
250 * Filter can be enabled only for the TAP devices. */
251 if (!check_filter(&tun->txflt, skb))
252 goto drop;
253
254 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
255 if (!(tun->flags & TUN_ONE_QUEUE)) {
256 /* Normal queueing mode. */
257 /* Packet scheduler handles dropping of further packets. */
258 netif_stop_queue(dev);
259
260 /* We won't see all dropped packets individually, so overrun
261 * error is more appropriate. */
262 dev->stats.tx_fifo_errors++;
263 } else {
264 /* Single queue mode.
265 * Driver handles dropping of all packets itself. */
266 goto drop;
267 }
268 }
269
270 /* Enqueue packet */
271 skb_queue_tail(&tun->readq, skb);
272 dev->trans_start = jiffies;
273
274 /* Notify and wake up reader process */
275 if (tun->flags & TUN_FASYNC)
276 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
277 wake_up_interruptible(&tun->read_wait);
278 return 0;
279
280 drop:
281 dev->stats.tx_dropped++;
282 kfree_skb(skb);
283 return 0;
284 }
285
286 static void tun_net_mclist(struct net_device *dev)
287 {
288 /*
289 * This callback is supposed to deal with mc filter in
290 * _rx_ path and has nothing to do with the _tx_ path.
291 * In rx path we always accept everything userspace gives us.
292 */
293 return;
294 }
295
296 #define MIN_MTU 68
297 #define MAX_MTU 65535
298
299 static int
300 tun_net_change_mtu(struct net_device *dev, int new_mtu)
301 {
302 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
303 return -EINVAL;
304 dev->mtu = new_mtu;
305 return 0;
306 }
307
308 static const struct net_device_ops tun_netdev_ops = {
309 .ndo_open = tun_net_open,
310 .ndo_stop = tun_net_close,
311 .ndo_start_xmit = tun_net_xmit,
312 .ndo_change_mtu = tun_net_change_mtu,
313 };
314
315 static const struct net_device_ops tap_netdev_ops = {
316 .ndo_open = tun_net_open,
317 .ndo_stop = tun_net_close,
318 .ndo_start_xmit = tun_net_xmit,
319 .ndo_change_mtu = tun_net_change_mtu,
320 .ndo_set_multicast_list = tun_net_mclist,
321 .ndo_set_mac_address = eth_mac_addr,
322 .ndo_validate_addr = eth_validate_addr,
323 };
324
325 /* Initialize net device. */
326 static void tun_net_init(struct net_device *dev)
327 {
328 struct tun_struct *tun = netdev_priv(dev);
329
330 switch (tun->flags & TUN_TYPE_MASK) {
331 case TUN_TUN_DEV:
332 dev->netdev_ops = &tun_netdev_ops;
333
334 /* Point-to-Point TUN Device */
335 dev->hard_header_len = 0;
336 dev->addr_len = 0;
337 dev->mtu = 1500;
338
339 /* Zero header length */
340 dev->type = ARPHRD_NONE;
341 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
342 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
343 break;
344
345 case TUN_TAP_DEV:
346 dev->netdev_ops = &tun_netdev_ops;
347 /* Ethernet TAP Device */
348 ether_setup(dev);
349
350 random_ether_addr(dev->dev_addr);
351
352 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
353 break;
354 }
355 }
356
357 /* Character device part */
358
359 /* Poll */
360 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
361 {
362 struct tun_struct *tun = file->private_data;
363 unsigned int mask = POLLOUT | POLLWRNORM;
364
365 if (!tun)
366 return -EBADFD;
367
368 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
369
370 poll_wait(file, &tun->read_wait, wait);
371
372 if (!skb_queue_empty(&tun->readq))
373 mask |= POLLIN | POLLRDNORM;
374
375 return mask;
376 }
377
378 /* prepad is the amount to reserve at front. len is length after that.
379 * linear is a hint as to how much to copy (usually headers). */
380 static struct sk_buff *tun_alloc_skb(size_t prepad, size_t len, size_t linear,
381 gfp_t gfp)
382 {
383 struct sk_buff *skb;
384 unsigned int i;
385
386 skb = alloc_skb(prepad + len, gfp|__GFP_NOWARN);
387 if (skb) {
388 skb_reserve(skb, prepad);
389 skb_put(skb, len);
390 return skb;
391 }
392
393 /* Under a page? Don't bother with paged skb. */
394 if (prepad + len < PAGE_SIZE)
395 return NULL;
396
397 /* Start with a normal skb, and add pages. */
398 skb = alloc_skb(prepad + linear, gfp);
399 if (!skb)
400 return NULL;
401
402 skb_reserve(skb, prepad);
403 skb_put(skb, linear);
404
405 len -= linear;
406
407 for (i = 0; i < MAX_SKB_FRAGS; i++) {
408 skb_frag_t *f = &skb_shinfo(skb)->frags[i];
409
410 f->page = alloc_page(gfp|__GFP_ZERO);
411 if (!f->page)
412 break;
413
414 f->page_offset = 0;
415 f->size = PAGE_SIZE;
416
417 skb->data_len += PAGE_SIZE;
418 skb->len += PAGE_SIZE;
419 skb->truesize += PAGE_SIZE;
420 skb_shinfo(skb)->nr_frags++;
421
422 if (len < PAGE_SIZE) {
423 len = 0;
424 break;
425 }
426 len -= PAGE_SIZE;
427 }
428
429 /* Too large, or alloc fail? */
430 if (unlikely(len)) {
431 kfree_skb(skb);
432 skb = NULL;
433 }
434
435 return skb;
436 }
437
438 /* Get packet from user space buffer */
439 static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, size_t count)
440 {
441 struct tun_pi pi = { 0, __constant_htons(ETH_P_IP) };
442 struct sk_buff *skb;
443 size_t len = count, align = 0;
444 struct virtio_net_hdr gso = { 0 };
445
446 if (!(tun->flags & TUN_NO_PI)) {
447 if ((len -= sizeof(pi)) > count)
448 return -EINVAL;
449
450 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
451 return -EFAULT;
452 }
453
454 if (tun->flags & TUN_VNET_HDR) {
455 if ((len -= sizeof(gso)) > count)
456 return -EINVAL;
457
458 if (memcpy_fromiovec((void *)&gso, iv, sizeof(gso)))
459 return -EFAULT;
460
461 if (gso.hdr_len > len)
462 return -EINVAL;
463 }
464
465 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
466 align = NET_IP_ALIGN;
467 if (unlikely(len < ETH_HLEN))
468 return -EINVAL;
469 }
470
471 if (!(skb = tun_alloc_skb(align, len, gso.hdr_len, GFP_KERNEL))) {
472 tun->dev->stats.rx_dropped++;
473 return -ENOMEM;
474 }
475
476 if (skb_copy_datagram_from_iovec(skb, 0, iv, len)) {
477 tun->dev->stats.rx_dropped++;
478 kfree_skb(skb);
479 return -EFAULT;
480 }
481
482 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
483 if (!skb_partial_csum_set(skb, gso.csum_start,
484 gso.csum_offset)) {
485 tun->dev->stats.rx_frame_errors++;
486 kfree_skb(skb);
487 return -EINVAL;
488 }
489 } else if (tun->flags & TUN_NOCHECKSUM)
490 skb->ip_summed = CHECKSUM_UNNECESSARY;
491
492 switch (tun->flags & TUN_TYPE_MASK) {
493 case TUN_TUN_DEV:
494 if (tun->flags & TUN_NO_PI) {
495 switch (skb->data[0] & 0xf0) {
496 case 0x40:
497 pi.proto = htons(ETH_P_IP);
498 break;
499 case 0x60:
500 pi.proto = htons(ETH_P_IPV6);
501 break;
502 default:
503 tun->dev->stats.rx_dropped++;
504 kfree_skb(skb);
505 return -EINVAL;
506 }
507 }
508
509 skb_reset_mac_header(skb);
510 skb->protocol = pi.proto;
511 skb->dev = tun->dev;
512 break;
513 case TUN_TAP_DEV:
514 skb->protocol = eth_type_trans(skb, tun->dev);
515 break;
516 };
517
518 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
519 pr_debug("GSO!\n");
520 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
521 case VIRTIO_NET_HDR_GSO_TCPV4:
522 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
523 break;
524 case VIRTIO_NET_HDR_GSO_TCPV6:
525 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
526 break;
527 default:
528 tun->dev->stats.rx_frame_errors++;
529 kfree_skb(skb);
530 return -EINVAL;
531 }
532
533 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
534 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
535
536 skb_shinfo(skb)->gso_size = gso.gso_size;
537 if (skb_shinfo(skb)->gso_size == 0) {
538 tun->dev->stats.rx_frame_errors++;
539 kfree_skb(skb);
540 return -EINVAL;
541 }
542
543 /* Header must be checked, and gso_segs computed. */
544 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
545 skb_shinfo(skb)->gso_segs = 0;
546 }
547
548 netif_rx_ni(skb);
549
550 tun->dev->stats.rx_packets++;
551 tun->dev->stats.rx_bytes += len;
552
553 return count;
554 }
555
556 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
557 unsigned long count, loff_t pos)
558 {
559 struct tun_struct *tun = iocb->ki_filp->private_data;
560
561 if (!tun)
562 return -EBADFD;
563
564 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
565
566 return tun_get_user(tun, (struct iovec *) iv, iov_length(iv, count));
567 }
568
569 /* Put packet to the user space buffer */
570 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
571 struct sk_buff *skb,
572 struct iovec *iv, int len)
573 {
574 struct tun_pi pi = { 0, skb->protocol };
575 ssize_t total = 0;
576
577 if (!(tun->flags & TUN_NO_PI)) {
578 if ((len -= sizeof(pi)) < 0)
579 return -EINVAL;
580
581 if (len < skb->len) {
582 /* Packet will be striped */
583 pi.flags |= TUN_PKT_STRIP;
584 }
585
586 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
587 return -EFAULT;
588 total += sizeof(pi);
589 }
590
591 if (tun->flags & TUN_VNET_HDR) {
592 struct virtio_net_hdr gso = { 0 }; /* no info leak */
593 if ((len -= sizeof(gso)) < 0)
594 return -EINVAL;
595
596 if (skb_is_gso(skb)) {
597 struct skb_shared_info *sinfo = skb_shinfo(skb);
598
599 /* This is a hint as to how much should be linear. */
600 gso.hdr_len = skb_headlen(skb);
601 gso.gso_size = sinfo->gso_size;
602 if (sinfo->gso_type & SKB_GSO_TCPV4)
603 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
604 else if (sinfo->gso_type & SKB_GSO_TCPV6)
605 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
606 else
607 BUG();
608 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
609 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
610 } else
611 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
612
613 if (skb->ip_summed == CHECKSUM_PARTIAL) {
614 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
615 gso.csum_start = skb->csum_start - skb_headroom(skb);
616 gso.csum_offset = skb->csum_offset;
617 } /* else everything is zero */
618
619 if (unlikely(memcpy_toiovec(iv, (void *)&gso, sizeof(gso))))
620 return -EFAULT;
621 total += sizeof(gso);
622 }
623
624 len = min_t(int, skb->len, len);
625
626 skb_copy_datagram_iovec(skb, 0, iv, len);
627 total += len;
628
629 tun->dev->stats.tx_packets++;
630 tun->dev->stats.tx_bytes += len;
631
632 return total;
633 }
634
635 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
636 unsigned long count, loff_t pos)
637 {
638 struct file *file = iocb->ki_filp;
639 struct tun_struct *tun = file->private_data;
640 DECLARE_WAITQUEUE(wait, current);
641 struct sk_buff *skb;
642 ssize_t len, ret = 0;
643
644 if (!tun)
645 return -EBADFD;
646
647 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
648
649 len = iov_length(iv, count);
650 if (len < 0)
651 return -EINVAL;
652
653 add_wait_queue(&tun->read_wait, &wait);
654 while (len) {
655 current->state = TASK_INTERRUPTIBLE;
656
657 /* Read frames from the queue */
658 if (!(skb=skb_dequeue(&tun->readq))) {
659 if (file->f_flags & O_NONBLOCK) {
660 ret = -EAGAIN;
661 break;
662 }
663 if (signal_pending(current)) {
664 ret = -ERESTARTSYS;
665 break;
666 }
667
668 /* Nothing to read, let's sleep */
669 schedule();
670 continue;
671 }
672 netif_wake_queue(tun->dev);
673
674 ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
675 kfree_skb(skb);
676 break;
677 }
678
679 current->state = TASK_RUNNING;
680 remove_wait_queue(&tun->read_wait, &wait);
681
682 return ret;
683 }
684
685 static void tun_setup(struct net_device *dev)
686 {
687 struct tun_struct *tun = netdev_priv(dev);
688
689 skb_queue_head_init(&tun->readq);
690 init_waitqueue_head(&tun->read_wait);
691
692 tun->owner = -1;
693 tun->group = -1;
694
695 dev->ethtool_ops = &tun_ethtool_ops;
696 dev->destructor = free_netdev;
697 dev->features |= NETIF_F_NETNS_LOCAL;
698 }
699
700 static struct tun_struct *tun_get_by_name(struct tun_net *tn, const char *name)
701 {
702 struct tun_struct *tun;
703
704 ASSERT_RTNL();
705 list_for_each_entry(tun, &tn->dev_list, list) {
706 if (!strncmp(tun->dev->name, name, IFNAMSIZ))
707 return tun;
708 }
709
710 return NULL;
711 }
712
713 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
714 {
715 struct tun_net *tn;
716 struct tun_struct *tun;
717 struct net_device *dev;
718 int err;
719
720 tn = net_generic(net, tun_net_id);
721 tun = tun_get_by_name(tn, ifr->ifr_name);
722 if (tun) {
723 if (tun->attached)
724 return -EBUSY;
725
726 /* Check permissions */
727 if (((tun->owner != -1 &&
728 current->euid != tun->owner) ||
729 (tun->group != -1 &&
730 current->egid != tun->group)) &&
731 !capable(CAP_NET_ADMIN))
732 return -EPERM;
733 }
734 else if (__dev_get_by_name(net, ifr->ifr_name))
735 return -EINVAL;
736 else {
737 char *name;
738 unsigned long flags = 0;
739
740 err = -EINVAL;
741
742 if (!capable(CAP_NET_ADMIN))
743 return -EPERM;
744
745 /* Set dev type */
746 if (ifr->ifr_flags & IFF_TUN) {
747 /* TUN device */
748 flags |= TUN_TUN_DEV;
749 name = "tun%d";
750 } else if (ifr->ifr_flags & IFF_TAP) {
751 /* TAP device */
752 flags |= TUN_TAP_DEV;
753 name = "tap%d";
754 } else
755 goto failed;
756
757 if (*ifr->ifr_name)
758 name = ifr->ifr_name;
759
760 dev = alloc_netdev(sizeof(struct tun_struct), name,
761 tun_setup);
762 if (!dev)
763 return -ENOMEM;
764
765 dev_net_set(dev, net);
766
767 tun = netdev_priv(dev);
768 tun->dev = dev;
769 tun->flags = flags;
770 tun->txflt.count = 0;
771
772 tun_net_init(dev);
773
774 if (strchr(dev->name, '%')) {
775 err = dev_alloc_name(dev, dev->name);
776 if (err < 0)
777 goto err_free_dev;
778 }
779
780 err = register_netdevice(tun->dev);
781 if (err < 0)
782 goto err_free_dev;
783
784 list_add(&tun->list, &tn->dev_list);
785 }
786
787 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
788
789 if (ifr->ifr_flags & IFF_NO_PI)
790 tun->flags |= TUN_NO_PI;
791 else
792 tun->flags &= ~TUN_NO_PI;
793
794 if (ifr->ifr_flags & IFF_ONE_QUEUE)
795 tun->flags |= TUN_ONE_QUEUE;
796 else
797 tun->flags &= ~TUN_ONE_QUEUE;
798
799 if (ifr->ifr_flags & IFF_VNET_HDR)
800 tun->flags |= TUN_VNET_HDR;
801 else
802 tun->flags &= ~TUN_VNET_HDR;
803
804 file->private_data = tun;
805 tun->attached = 1;
806 get_net(dev_net(tun->dev));
807
808 /* Make sure persistent devices do not get stuck in
809 * xoff state.
810 */
811 if (netif_running(tun->dev))
812 netif_wake_queue(tun->dev);
813
814 strcpy(ifr->ifr_name, tun->dev->name);
815 return 0;
816
817 err_free_dev:
818 free_netdev(dev);
819 failed:
820 return err;
821 }
822
823 static int tun_get_iff(struct net *net, struct file *file, struct ifreq *ifr)
824 {
825 struct tun_struct *tun = file->private_data;
826
827 if (!tun)
828 return -EBADFD;
829
830 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
831
832 strcpy(ifr->ifr_name, tun->dev->name);
833
834 ifr->ifr_flags = 0;
835
836 if (ifr->ifr_flags & TUN_TUN_DEV)
837 ifr->ifr_flags |= IFF_TUN;
838 else
839 ifr->ifr_flags |= IFF_TAP;
840
841 if (tun->flags & TUN_NO_PI)
842 ifr->ifr_flags |= IFF_NO_PI;
843
844 if (tun->flags & TUN_ONE_QUEUE)
845 ifr->ifr_flags |= IFF_ONE_QUEUE;
846
847 if (tun->flags & TUN_VNET_HDR)
848 ifr->ifr_flags |= IFF_VNET_HDR;
849
850 return 0;
851 }
852
853 /* This is like a cut-down ethtool ops, except done via tun fd so no
854 * privs required. */
855 static int set_offload(struct net_device *dev, unsigned long arg)
856 {
857 unsigned int old_features, features;
858
859 old_features = dev->features;
860 /* Unset features, set them as we chew on the arg. */
861 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
862 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
863
864 if (arg & TUN_F_CSUM) {
865 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
866 arg &= ~TUN_F_CSUM;
867
868 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
869 if (arg & TUN_F_TSO_ECN) {
870 features |= NETIF_F_TSO_ECN;
871 arg &= ~TUN_F_TSO_ECN;
872 }
873 if (arg & TUN_F_TSO4)
874 features |= NETIF_F_TSO;
875 if (arg & TUN_F_TSO6)
876 features |= NETIF_F_TSO6;
877 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
878 }
879 }
880
881 /* This gives the user a way to test for new features in future by
882 * trying to set them. */
883 if (arg)
884 return -EINVAL;
885
886 dev->features = features;
887 if (old_features != dev->features)
888 netdev_features_change(dev);
889
890 return 0;
891 }
892
893 static int tun_chr_ioctl(struct inode *inode, struct file *file,
894 unsigned int cmd, unsigned long arg)
895 {
896 struct tun_struct *tun = file->private_data;
897 void __user* argp = (void __user*)arg;
898 struct ifreq ifr;
899 int ret;
900
901 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
902 if (copy_from_user(&ifr, argp, sizeof ifr))
903 return -EFAULT;
904
905 if (cmd == TUNSETIFF && !tun) {
906 int err;
907
908 ifr.ifr_name[IFNAMSIZ-1] = '\0';
909
910 rtnl_lock();
911 err = tun_set_iff(current->nsproxy->net_ns, file, &ifr);
912 rtnl_unlock();
913
914 if (err)
915 return err;
916
917 if (copy_to_user(argp, &ifr, sizeof(ifr)))
918 return -EFAULT;
919 return 0;
920 }
921
922 if (cmd == TUNGETFEATURES) {
923 /* Currently this just means: "what IFF flags are valid?".
924 * This is needed because we never checked for invalid flags on
925 * TUNSETIFF. */
926 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
927 IFF_VNET_HDR,
928 (unsigned int __user*)argp);
929 }
930
931 if (!tun)
932 return -EBADFD;
933
934 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
935
936 switch (cmd) {
937 case TUNGETIFF:
938 ret = tun_get_iff(current->nsproxy->net_ns, file, &ifr);
939 if (ret)
940 return ret;
941
942 if (copy_to_user(argp, &ifr, sizeof(ifr)))
943 return -EFAULT;
944 break;
945
946 case TUNSETNOCSUM:
947 /* Disable/Enable checksum */
948 if (arg)
949 tun->flags |= TUN_NOCHECKSUM;
950 else
951 tun->flags &= ~TUN_NOCHECKSUM;
952
953 DBG(KERN_INFO "%s: checksum %s\n",
954 tun->dev->name, arg ? "disabled" : "enabled");
955 break;
956
957 case TUNSETPERSIST:
958 /* Disable/Enable persist mode */
959 if (arg)
960 tun->flags |= TUN_PERSIST;
961 else
962 tun->flags &= ~TUN_PERSIST;
963
964 DBG(KERN_INFO "%s: persist %s\n",
965 tun->dev->name, arg ? "enabled" : "disabled");
966 break;
967
968 case TUNSETOWNER:
969 /* Set owner of the device */
970 tun->owner = (uid_t) arg;
971
972 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
973 break;
974
975 case TUNSETGROUP:
976 /* Set group of the device */
977 tun->group= (gid_t) arg;
978
979 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
980 break;
981
982 case TUNSETLINK:
983 /* Only allow setting the type when the interface is down */
984 rtnl_lock();
985 if (tun->dev->flags & IFF_UP) {
986 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
987 tun->dev->name);
988 ret = -EBUSY;
989 } else {
990 tun->dev->type = (int) arg;
991 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
992 ret = 0;
993 }
994 rtnl_unlock();
995 return ret;
996
997 #ifdef TUN_DEBUG
998 case TUNSETDEBUG:
999 tun->debug = arg;
1000 break;
1001 #endif
1002 case TUNSETOFFLOAD:
1003 rtnl_lock();
1004 ret = set_offload(tun->dev, arg);
1005 rtnl_unlock();
1006 return ret;
1007
1008 case TUNSETTXFILTER:
1009 /* Can be set only for TAPs */
1010 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1011 return -EINVAL;
1012 rtnl_lock();
1013 ret = update_filter(&tun->txflt, (void __user *)arg);
1014 rtnl_unlock();
1015 return ret;
1016
1017 case SIOCGIFHWADDR:
1018 /* Get hw addres */
1019 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1020 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1021 if (copy_to_user(argp, &ifr, sizeof ifr))
1022 return -EFAULT;
1023 return 0;
1024
1025 case SIOCSIFHWADDR:
1026 /* Set hw address */
1027 DBG(KERN_DEBUG "%s: set hw address: %pM\n",
1028 tun->dev->name, ifr.ifr_hwaddr.sa_data);
1029
1030 rtnl_lock();
1031 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1032 rtnl_unlock();
1033 return ret;
1034
1035 default:
1036 return -EINVAL;
1037 };
1038
1039 return 0;
1040 }
1041
1042 static int tun_chr_fasync(int fd, struct file *file, int on)
1043 {
1044 struct tun_struct *tun = file->private_data;
1045 int ret;
1046
1047 if (!tun)
1048 return -EBADFD;
1049
1050 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1051
1052 lock_kernel();
1053 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1054 goto out;
1055
1056 if (on) {
1057 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1058 if (ret)
1059 goto out;
1060 tun->flags |= TUN_FASYNC;
1061 } else
1062 tun->flags &= ~TUN_FASYNC;
1063 ret = 0;
1064 out:
1065 unlock_kernel();
1066 return ret;
1067 }
1068
1069 static int tun_chr_open(struct inode *inode, struct file * file)
1070 {
1071 cycle_kernel_lock();
1072 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1073 file->private_data = NULL;
1074 return 0;
1075 }
1076
1077 static int tun_chr_close(struct inode *inode, struct file *file)
1078 {
1079 struct tun_struct *tun = file->private_data;
1080
1081 if (!tun)
1082 return 0;
1083
1084 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
1085
1086 rtnl_lock();
1087
1088 /* Detach from net device */
1089 file->private_data = NULL;
1090 tun->attached = 0;
1091 put_net(dev_net(tun->dev));
1092
1093 /* Drop read queue */
1094 skb_queue_purge(&tun->readq);
1095
1096 if (!(tun->flags & TUN_PERSIST)) {
1097 list_del(&tun->list);
1098 unregister_netdevice(tun->dev);
1099 }
1100
1101 rtnl_unlock();
1102
1103 return 0;
1104 }
1105
1106 static const struct file_operations tun_fops = {
1107 .owner = THIS_MODULE,
1108 .llseek = no_llseek,
1109 .read = do_sync_read,
1110 .aio_read = tun_chr_aio_read,
1111 .write = do_sync_write,
1112 .aio_write = tun_chr_aio_write,
1113 .poll = tun_chr_poll,
1114 .ioctl = tun_chr_ioctl,
1115 .open = tun_chr_open,
1116 .release = tun_chr_close,
1117 .fasync = tun_chr_fasync
1118 };
1119
1120 static struct miscdevice tun_miscdev = {
1121 .minor = TUN_MINOR,
1122 .name = "tun",
1123 .fops = &tun_fops,
1124 };
1125
1126 /* ethtool interface */
1127
1128 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1129 {
1130 cmd->supported = 0;
1131 cmd->advertising = 0;
1132 cmd->speed = SPEED_10;
1133 cmd->duplex = DUPLEX_FULL;
1134 cmd->port = PORT_TP;
1135 cmd->phy_address = 0;
1136 cmd->transceiver = XCVR_INTERNAL;
1137 cmd->autoneg = AUTONEG_DISABLE;
1138 cmd->maxtxpkt = 0;
1139 cmd->maxrxpkt = 0;
1140 return 0;
1141 }
1142
1143 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1144 {
1145 struct tun_struct *tun = netdev_priv(dev);
1146
1147 strcpy(info->driver, DRV_NAME);
1148 strcpy(info->version, DRV_VERSION);
1149 strcpy(info->fw_version, "N/A");
1150
1151 switch (tun->flags & TUN_TYPE_MASK) {
1152 case TUN_TUN_DEV:
1153 strcpy(info->bus_info, "tun");
1154 break;
1155 case TUN_TAP_DEV:
1156 strcpy(info->bus_info, "tap");
1157 break;
1158 }
1159 }
1160
1161 static u32 tun_get_msglevel(struct net_device *dev)
1162 {
1163 #ifdef TUN_DEBUG
1164 struct tun_struct *tun = netdev_priv(dev);
1165 return tun->debug;
1166 #else
1167 return -EOPNOTSUPP;
1168 #endif
1169 }
1170
1171 static void tun_set_msglevel(struct net_device *dev, u32 value)
1172 {
1173 #ifdef TUN_DEBUG
1174 struct tun_struct *tun = netdev_priv(dev);
1175 tun->debug = value;
1176 #endif
1177 }
1178
1179 static u32 tun_get_link(struct net_device *dev)
1180 {
1181 struct tun_struct *tun = netdev_priv(dev);
1182 return tun->attached;
1183 }
1184
1185 static u32 tun_get_rx_csum(struct net_device *dev)
1186 {
1187 struct tun_struct *tun = netdev_priv(dev);
1188 return (tun->flags & TUN_NOCHECKSUM) == 0;
1189 }
1190
1191 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1192 {
1193 struct tun_struct *tun = netdev_priv(dev);
1194 if (data)
1195 tun->flags &= ~TUN_NOCHECKSUM;
1196 else
1197 tun->flags |= TUN_NOCHECKSUM;
1198 return 0;
1199 }
1200
1201 static const struct ethtool_ops tun_ethtool_ops = {
1202 .get_settings = tun_get_settings,
1203 .get_drvinfo = tun_get_drvinfo,
1204 .get_msglevel = tun_get_msglevel,
1205 .set_msglevel = tun_set_msglevel,
1206 .get_link = tun_get_link,
1207 .get_rx_csum = tun_get_rx_csum,
1208 .set_rx_csum = tun_set_rx_csum
1209 };
1210
1211 static int tun_init_net(struct net *net)
1212 {
1213 struct tun_net *tn;
1214
1215 tn = kmalloc(sizeof(*tn), GFP_KERNEL);
1216 if (tn == NULL)
1217 return -ENOMEM;
1218
1219 INIT_LIST_HEAD(&tn->dev_list);
1220
1221 if (net_assign_generic(net, tun_net_id, tn)) {
1222 kfree(tn);
1223 return -ENOMEM;
1224 }
1225
1226 return 0;
1227 }
1228
1229 static void tun_exit_net(struct net *net)
1230 {
1231 struct tun_net *tn;
1232 struct tun_struct *tun, *nxt;
1233
1234 tn = net_generic(net, tun_net_id);
1235
1236 rtnl_lock();
1237 list_for_each_entry_safe(tun, nxt, &tn->dev_list, list) {
1238 DBG(KERN_INFO "%s cleaned up\n", tun->dev->name);
1239 unregister_netdevice(tun->dev);
1240 }
1241 rtnl_unlock();
1242
1243 kfree(tn);
1244 }
1245
1246 static struct pernet_operations tun_net_ops = {
1247 .init = tun_init_net,
1248 .exit = tun_exit_net,
1249 };
1250
1251 static int __init tun_init(void)
1252 {
1253 int ret = 0;
1254
1255 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1256 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1257
1258 ret = register_pernet_gen_device(&tun_net_id, &tun_net_ops);
1259 if (ret) {
1260 printk(KERN_ERR "tun: Can't register pernet ops\n");
1261 goto err_pernet;
1262 }
1263
1264 ret = misc_register(&tun_miscdev);
1265 if (ret) {
1266 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1267 goto err_misc;
1268 }
1269 return 0;
1270
1271 err_misc:
1272 unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1273 err_pernet:
1274 return ret;
1275 }
1276
1277 static void tun_cleanup(void)
1278 {
1279 misc_deregister(&tun_miscdev);
1280 unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1281 }
1282
1283 module_init(tun_init);
1284 module_exit(tun_cleanup);
1285 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1286 MODULE_AUTHOR(DRV_COPYRIGHT);
1287 MODULE_LICENSE("GPL");
1288 MODULE_ALIAS_MISCDEV(TUN_MINOR);
Verdex Pro support