?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tun: Fix unicast filter overflow
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / drivers / net / tun.c
blob3ee95593a19cb53132f8f2ba9c28e0e44760e0ed
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 /* Remaining multicast addresses are hashed,
161 * unicast will leave the filter disabled. */
162 memset(filter->mask, 0, sizeof(filter->mask));
163 for (; n < uf.count; n++) {
164 if (!is_multicast_ether_addr(addr[n].u)) {
165 err = 0; /* no filter */
166 goto done;
167 }
168 addr_hash_set(filter->mask, addr[n].u);
169 }
170
171 /* For ALLMULTI just set the mask to all ones.
172 * This overrides the mask populated above. */
173 if ((uf.flags & TUN_FLT_ALLMULTI))
174 memset(filter->mask, ~0, sizeof(filter->mask));
175
176 /* Now enable the filter */
177 wmb();
178 filter->count = nexact;
179
180 /* Return the number of exact filters */
181 err = nexact;
182
183 done:
184 kfree(addr);
185 return err;
186 }
187
188 /* Returns: 0 - drop, !=0 - accept */
189 static int run_filter(struct tap_filter *filter, const struct sk_buff *skb)
190 {
191 /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect
192 * at this point. */
193 struct ethhdr *eh = (struct ethhdr *) skb->data;
194 int i;
195
196 /* Exact match */
197 for (i = 0; i < filter->count; i++)
198 if (!compare_ether_addr(eh->h_dest, filter->addr[i]))
199 return 1;
200
201 /* Inexact match (multicast only) */
202 if (is_multicast_ether_addr(eh->h_dest))
203 return addr_hash_test(filter->mask, eh->h_dest);
204
205 return 0;
206 }
207
208 /*
209 * Checks whether the packet is accepted or not.
210 * Returns: 0 - drop, !=0 - accept
211 */
212 static int check_filter(struct tap_filter *filter, const struct sk_buff *skb)
213 {
214 if (!filter->count)
215 return 1;
216
217 return run_filter(filter, skb);
218 }
219
220 /* Network device part of the driver */
221
222 static unsigned int tun_net_id;
223 struct tun_net {
224 struct list_head dev_list;
225 };
226
227 static const struct ethtool_ops tun_ethtool_ops;
228
229 /* Net device open. */
230 static int tun_net_open(struct net_device *dev)
231 {
232 netif_start_queue(dev);
233 return 0;
234 }
235
236 /* Net device close. */
237 static int tun_net_close(struct net_device *dev)
238 {
239 netif_stop_queue(dev);
240 return 0;
241 }
242
243 /* Net device start xmit */
244 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
245 {
246 struct tun_struct *tun = netdev_priv(dev);
247
248 DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
249
250 /* Drop packet if interface is not attached */
251 if (!tun->attached)
252 goto drop;
253
254 /* Drop if the filter does not like it.
255 * This is a noop if the filter is disabled.
256 * Filter can be enabled only for the TAP devices. */
257 if (!check_filter(&tun->txflt, skb))
258 goto drop;
259
260 if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
261 if (!(tun->flags & TUN_ONE_QUEUE)) {
262 /* Normal queueing mode. */
263 /* Packet scheduler handles dropping of further packets. */
264 netif_stop_queue(dev);
265
266 /* We won't see all dropped packets individually, so overrun
267 * error is more appropriate. */
268 dev->stats.tx_fifo_errors++;
269 } else {
270 /* Single queue mode.
271 * Driver handles dropping of all packets itself. */
272 goto drop;
273 }
274 }
275
276 /* Enqueue packet */
277 skb_queue_tail(&tun->readq, skb);
278 dev->trans_start = jiffies;
279
280 /* Notify and wake up reader process */
281 if (tun->flags & TUN_FASYNC)
282 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
283 wake_up_interruptible(&tun->read_wait);
284 return 0;
285
286 drop:
287 dev->stats.tx_dropped++;
288 kfree_skb(skb);
289 return 0;
290 }
291
292 static void tun_net_mclist(struct net_device *dev)
293 {
294 /*
295 * This callback is supposed to deal with mc filter in
296 * _rx_ path and has nothing to do with the _tx_ path.
297 * In rx path we always accept everything userspace gives us.
298 */
299 return;
300 }
301
302 #define MIN_MTU 68
303 #define MAX_MTU 65535
304
305 static int
306 tun_net_change_mtu(struct net_device *dev, int new_mtu)
307 {
308 if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
309 return -EINVAL;
310 dev->mtu = new_mtu;
311 return 0;
312 }
313
314 /* Initialize net device. */
315 static void tun_net_init(struct net_device *dev)
316 {
317 struct tun_struct *tun = netdev_priv(dev);
318
319 switch (tun->flags & TUN_TYPE_MASK) {
320 case TUN_TUN_DEV:
321 /* Point-to-Point TUN Device */
322 dev->hard_header_len = 0;
323 dev->addr_len = 0;
324 dev->mtu = 1500;
325 dev->change_mtu = tun_net_change_mtu;
326
327 /* Zero header length */
328 dev->type = ARPHRD_NONE;
329 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
330 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
331 break;
332
333 case TUN_TAP_DEV:
334 /* Ethernet TAP Device */
335 ether_setup(dev);
336 dev->change_mtu = tun_net_change_mtu;
337 dev->set_multicast_list = tun_net_mclist;
338
339 random_ether_addr(dev->dev_addr);
340
341 dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */
342 break;
343 }
344 }
345
346 /* Character device part */
347
348 /* Poll */
349 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
350 {
351 struct tun_struct *tun = file->private_data;
352 unsigned int mask = POLLOUT | POLLWRNORM;
353
354 if (!tun)
355 return -EBADFD;
356
357 DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
358
359 poll_wait(file, &tun->read_wait, wait);
360
361 if (!skb_queue_empty(&tun->readq))
362 mask |= POLLIN | POLLRDNORM;
363
364 return mask;
365 }
366
367 /* prepad is the amount to reserve at front. len is length after that.
368 * linear is a hint as to how much to copy (usually headers). */
369 static struct sk_buff *tun_alloc_skb(size_t prepad, size_t len, size_t linear,
370 gfp_t gfp)
371 {
372 struct sk_buff *skb;
373 unsigned int i;
374
375 skb = alloc_skb(prepad + len, gfp|__GFP_NOWARN);
376 if (skb) {
377 skb_reserve(skb, prepad);
378 skb_put(skb, len);
379 return skb;
380 }
381
382 /* Under a page? Don't bother with paged skb. */
383 if (prepad + len < PAGE_SIZE)
384 return NULL;
385
386 /* Start with a normal skb, and add pages. */
387 skb = alloc_skb(prepad + linear, gfp);
388 if (!skb)
389 return NULL;
390
391 skb_reserve(skb, prepad);
392 skb_put(skb, linear);
393
394 len -= linear;
395
396 for (i = 0; i < MAX_SKB_FRAGS; i++) {
397 skb_frag_t *f = &skb_shinfo(skb)->frags[i];
398
399 f->page = alloc_page(gfp|__GFP_ZERO);
400 if (!f->page)
401 break;
402
403 f->page_offset = 0;
404 f->size = PAGE_SIZE;
405
406 skb->data_len += PAGE_SIZE;
407 skb->len += PAGE_SIZE;
408 skb->truesize += PAGE_SIZE;
409 skb_shinfo(skb)->nr_frags++;
410
411 if (len < PAGE_SIZE) {
412 len = 0;
413 break;
414 }
415 len -= PAGE_SIZE;
416 }
417
418 /* Too large, or alloc fail? */
419 if (unlikely(len)) {
420 kfree_skb(skb);
421 skb = NULL;
422 }
423
424 return skb;
425 }
426
427 /* Get packet from user space buffer */
428 static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, size_t count)
429 {
430 struct tun_pi pi = { 0, __constant_htons(ETH_P_IP) };
431 struct sk_buff *skb;
432 size_t len = count, align = 0;
433 struct virtio_net_hdr gso = { 0 };
434
435 if (!(tun->flags & TUN_NO_PI)) {
436 if ((len -= sizeof(pi)) > count)
437 return -EINVAL;
438
439 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
440 return -EFAULT;
441 }
442
443 if (tun->flags & TUN_VNET_HDR) {
444 if ((len -= sizeof(gso)) > count)
445 return -EINVAL;
446
447 if (memcpy_fromiovec((void *)&gso, iv, sizeof(gso)))
448 return -EFAULT;
449
450 if (gso.hdr_len > len)
451 return -EINVAL;
452 }
453
454 if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
455 align = NET_IP_ALIGN;
456 if (unlikely(len < ETH_HLEN))
457 return -EINVAL;
458 }
459
460 if (!(skb = tun_alloc_skb(align, len, gso.hdr_len, GFP_KERNEL))) {
461 tun->dev->stats.rx_dropped++;
462 return -ENOMEM;
463 }
464
465 if (skb_copy_datagram_from_iovec(skb, 0, iv, len)) {
466 tun->dev->stats.rx_dropped++;
467 kfree_skb(skb);
468 return -EFAULT;
469 }
470
471 if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) {
472 if (!skb_partial_csum_set(skb, gso.csum_start,
473 gso.csum_offset)) {
474 tun->dev->stats.rx_frame_errors++;
475 kfree_skb(skb);
476 return -EINVAL;
477 }
478 } else if (tun->flags & TUN_NOCHECKSUM)
479 skb->ip_summed = CHECKSUM_UNNECESSARY;
480
481 switch (tun->flags & TUN_TYPE_MASK) {
482 case TUN_TUN_DEV:
483 if (tun->flags & TUN_NO_PI) {
484 switch (skb->data[0] & 0xf0) {
485 case 0x40:
486 pi.proto = htons(ETH_P_IP);
487 break;
488 case 0x60:
489 pi.proto = htons(ETH_P_IPV6);
490 break;
491 default:
492 tun->dev->stats.rx_dropped++;
493 kfree_skb(skb);
494 return -EINVAL;
495 }
496 }
497
498 skb_reset_mac_header(skb);
499 skb->protocol = pi.proto;
500 skb->dev = tun->dev;
501 break;
502 case TUN_TAP_DEV:
503 skb->protocol = eth_type_trans(skb, tun->dev);
504 break;
505 };
506
507 if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) {
508 pr_debug("GSO!\n");
509 switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) {
510 case VIRTIO_NET_HDR_GSO_TCPV4:
511 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
512 break;
513 case VIRTIO_NET_HDR_GSO_TCPV6:
514 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
515 break;
516 default:
517 tun->dev->stats.rx_frame_errors++;
518 kfree_skb(skb);
519 return -EINVAL;
520 }
521
522 if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN)
523 skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN;
524
525 skb_shinfo(skb)->gso_size = gso.gso_size;
526 if (skb_shinfo(skb)->gso_size == 0) {
527 tun->dev->stats.rx_frame_errors++;
528 kfree_skb(skb);
529 return -EINVAL;
530 }
531
532 /* Header must be checked, and gso_segs computed. */
533 skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY;
534 skb_shinfo(skb)->gso_segs = 0;
535 }
536
537 netif_rx_ni(skb);
538 tun->dev->last_rx = jiffies;
539
540 tun->dev->stats.rx_packets++;
541 tun->dev->stats.rx_bytes += len;
542
543 return count;
544 }
545
546 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
547 unsigned long count, loff_t pos)
548 {
549 struct tun_struct *tun = iocb->ki_filp->private_data;
550
551 if (!tun)
552 return -EBADFD;
553
554 DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
555
556 return tun_get_user(tun, (struct iovec *) iv, iov_length(iv, count));
557 }
558
559 /* Put packet to the user space buffer */
560 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
561 struct sk_buff *skb,
562 struct iovec *iv, int len)
563 {
564 struct tun_pi pi = { 0, skb->protocol };
565 ssize_t total = 0;
566
567 if (!(tun->flags & TUN_NO_PI)) {
568 if ((len -= sizeof(pi)) < 0)
569 return -EINVAL;
570
571 if (len < skb->len) {
572 /* Packet will be striped */
573 pi.flags |= TUN_PKT_STRIP;
574 }
575
576 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
577 return -EFAULT;
578 total += sizeof(pi);
579 }
580
581 if (tun->flags & TUN_VNET_HDR) {
582 struct virtio_net_hdr gso = { 0 }; /* no info leak */
583 if ((len -= sizeof(gso)) < 0)
584 return -EINVAL;
585
586 if (skb_is_gso(skb)) {
587 struct skb_shared_info *sinfo = skb_shinfo(skb);
588
589 /* This is a hint as to how much should be linear. */
590 gso.hdr_len = skb_headlen(skb);
591 gso.gso_size = sinfo->gso_size;
592 if (sinfo->gso_type & SKB_GSO_TCPV4)
593 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4;
594 else if (sinfo->gso_type & SKB_GSO_TCPV6)
595 gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6;
596 else
597 BUG();
598 if (sinfo->gso_type & SKB_GSO_TCP_ECN)
599 gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN;
600 } else
601 gso.gso_type = VIRTIO_NET_HDR_GSO_NONE;
602
603 if (skb->ip_summed == CHECKSUM_PARTIAL) {
604 gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
605 gso.csum_start = skb->csum_start - skb_headroom(skb);
606 gso.csum_offset = skb->csum_offset;
607 } /* else everything is zero */
608
609 if (unlikely(memcpy_toiovec(iv, (void *)&gso, sizeof(gso))))
610 return -EFAULT;
611 total += sizeof(gso);
612 }
613
614 len = min_t(int, skb->len, len);
615
616 skb_copy_datagram_iovec(skb, 0, iv, len);
617 total += len;
618
619 tun->dev->stats.tx_packets++;
620 tun->dev->stats.tx_bytes += len;
621
622 return total;
623 }
624
625 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
626 unsigned long count, loff_t pos)
627 {
628 struct file *file = iocb->ki_filp;
629 struct tun_struct *tun = file->private_data;
630 DECLARE_WAITQUEUE(wait, current);
631 struct sk_buff *skb;
632 ssize_t len, ret = 0;
633
634 if (!tun)
635 return -EBADFD;
636
637 DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
638
639 len = iov_length(iv, count);
640 if (len < 0)
641 return -EINVAL;
642
643 add_wait_queue(&tun->read_wait, &wait);
644 while (len) {
645 current->state = TASK_INTERRUPTIBLE;
646
647 /* Read frames from the queue */
648 if (!(skb=skb_dequeue(&tun->readq))) {
649 if (file->f_flags & O_NONBLOCK) {
650 ret = -EAGAIN;
651 break;
652 }
653 if (signal_pending(current)) {
654 ret = -ERESTARTSYS;
655 break;
656 }
657
658 /* Nothing to read, let's sleep */
659 schedule();
660 continue;
661 }
662 netif_wake_queue(tun->dev);
663
664 ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
665 kfree_skb(skb);
666 break;
667 }
668
669 current->state = TASK_RUNNING;
670 remove_wait_queue(&tun->read_wait, &wait);
671
672 return ret;
673 }
674
675 static void tun_setup(struct net_device *dev)
676 {
677 struct tun_struct *tun = netdev_priv(dev);
678
679 skb_queue_head_init(&tun->readq);
680 init_waitqueue_head(&tun->read_wait);
681
682 tun->owner = -1;
683 tun->group = -1;
684
685 dev->open = tun_net_open;
686 dev->hard_start_xmit = tun_net_xmit;
687 dev->stop = tun_net_close;
688 dev->ethtool_ops = &tun_ethtool_ops;
689 dev->destructor = free_netdev;
690 dev->features |= NETIF_F_NETNS_LOCAL;
691 }
692
693 static struct tun_struct *tun_get_by_name(struct tun_net *tn, const char *name)
694 {
695 struct tun_struct *tun;
696
697 ASSERT_RTNL();
698 list_for_each_entry(tun, &tn->dev_list, list) {
699 if (!strncmp(tun->dev->name, name, IFNAMSIZ))
700 return tun;
701 }
702
703 return NULL;
704 }
705
706 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
707 {
708 struct tun_net *tn;
709 struct tun_struct *tun;
710 struct net_device *dev;
711 int err;
712
713 tn = net_generic(net, tun_net_id);
714 tun = tun_get_by_name(tn, ifr->ifr_name);
715 if (tun) {
716 if (tun->attached)
717 return -EBUSY;
718
719 /* Check permissions */
720 if (((tun->owner != -1 &&
721 current->euid != tun->owner) ||
722 (tun->group != -1 &&
723 current->egid != tun->group)) &&
724 !capable(CAP_NET_ADMIN))
725 return -EPERM;
726 }
727 else if (__dev_get_by_name(net, ifr->ifr_name))
728 return -EINVAL;
729 else {
730 char *name;
731 unsigned long flags = 0;
732
733 err = -EINVAL;
734
735 if (!capable(CAP_NET_ADMIN))
736 return -EPERM;
737
738 /* Set dev type */
739 if (ifr->ifr_flags & IFF_TUN) {
740 /* TUN device */
741 flags |= TUN_TUN_DEV;
742 name = "tun%d";
743 } else if (ifr->ifr_flags & IFF_TAP) {
744 /* TAP device */
745 flags |= TUN_TAP_DEV;
746 name = "tap%d";
747 } else
748 goto failed;
749
750 if (*ifr->ifr_name)
751 name = ifr->ifr_name;
752
753 dev = alloc_netdev(sizeof(struct tun_struct), name,
754 tun_setup);
755 if (!dev)
756 return -ENOMEM;
757
758 dev_net_set(dev, net);
759 tun = netdev_priv(dev);
760 tun->dev = dev;
761 tun->flags = flags;
762 tun->txflt.count = 0;
763
764 tun_net_init(dev);
765
766 if (strchr(dev->name, '%')) {
767 err = dev_alloc_name(dev, dev->name);
768 if (err < 0)
769 goto err_free_dev;
770 }
771
772 err = register_netdevice(tun->dev);
773 if (err < 0)
774 goto err_free_dev;
775
776 list_add(&tun->list, &tn->dev_list);
777 }
778
779 DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
780
781 if (ifr->ifr_flags & IFF_NO_PI)
782 tun->flags |= TUN_NO_PI;
783 else
784 tun->flags &= ~TUN_NO_PI;
785
786 if (ifr->ifr_flags & IFF_ONE_QUEUE)
787 tun->flags |= TUN_ONE_QUEUE;
788 else
789 tun->flags &= ~TUN_ONE_QUEUE;
790
791 if (ifr->ifr_flags & IFF_VNET_HDR)
792 tun->flags |= TUN_VNET_HDR;
793 else
794 tun->flags &= ~TUN_VNET_HDR;
795
796 file->private_data = tun;
797 tun->attached = 1;
798 get_net(dev_net(tun->dev));
799
800 /* Make sure persistent devices do not get stuck in
801 * xoff state.
802 */
803 if (netif_running(tun->dev))
804 netif_wake_queue(tun->dev);
805
806 strcpy(ifr->ifr_name, tun->dev->name);
807 return 0;
808
809 err_free_dev:
810 free_netdev(dev);
811 failed:
812 return err;
813 }
814
815 static int tun_get_iff(struct net *net, struct file *file, struct ifreq *ifr)
816 {
817 struct tun_struct *tun = file->private_data;
818
819 if (!tun)
820 return -EBADFD;
821
822 DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name);
823
824 strcpy(ifr->ifr_name, tun->dev->name);
825
826 ifr->ifr_flags = 0;
827
828 if (ifr->ifr_flags & TUN_TUN_DEV)
829 ifr->ifr_flags |= IFF_TUN;
830 else
831 ifr->ifr_flags |= IFF_TAP;
832
833 if (tun->flags & TUN_NO_PI)
834 ifr->ifr_flags |= IFF_NO_PI;
835
836 if (tun->flags & TUN_ONE_QUEUE)
837 ifr->ifr_flags |= IFF_ONE_QUEUE;
838
839 if (tun->flags & TUN_VNET_HDR)
840 ifr->ifr_flags |= IFF_VNET_HDR;
841
842 return 0;
843 }
844
845 /* This is like a cut-down ethtool ops, except done via tun fd so no
846 * privs required. */
847 static int set_offload(struct net_device *dev, unsigned long arg)
848 {
849 unsigned int old_features, features;
850
851 old_features = dev->features;
852 /* Unset features, set them as we chew on the arg. */
853 features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
854 |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
855
856 if (arg & TUN_F_CSUM) {
857 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
858 arg &= ~TUN_F_CSUM;
859
860 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
861 if (arg & TUN_F_TSO_ECN) {
862 features |= NETIF_F_TSO_ECN;
863 arg &= ~TUN_F_TSO_ECN;
864 }
865 if (arg & TUN_F_TSO4)
866 features |= NETIF_F_TSO;
867 if (arg & TUN_F_TSO6)
868 features |= NETIF_F_TSO6;
869 arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
870 }
871 }
872
873 /* This gives the user a way to test for new features in future by
874 * trying to set them. */
875 if (arg)
876 return -EINVAL;
877
878 dev->features = features;
879 if (old_features != dev->features)
880 netdev_features_change(dev);
881
882 return 0;
883 }
884
885 static int tun_chr_ioctl(struct inode *inode, struct file *file,
886 unsigned int cmd, unsigned long arg)
887 {
888 struct tun_struct *tun = file->private_data;
889 void __user* argp = (void __user*)arg;
890 struct ifreq ifr;
891 int ret;
892 DECLARE_MAC_BUF(mac);
893
894 if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
895 if (copy_from_user(&ifr, argp, sizeof ifr))
896 return -EFAULT;
897
898 if (cmd == TUNSETIFF && !tun) {
899 int err;
900
901 ifr.ifr_name[IFNAMSIZ-1] = '\0';
902
903 rtnl_lock();
904 err = tun_set_iff(current->nsproxy->net_ns, file, &ifr);
905 rtnl_unlock();
906
907 if (err)
908 return err;
909
910 if (copy_to_user(argp, &ifr, sizeof(ifr)))
911 return -EFAULT;
912 return 0;
913 }
914
915 if (cmd == TUNGETFEATURES) {
916 /* Currently this just means: "what IFF flags are valid?".
917 * This is needed because we never checked for invalid flags on
918 * TUNSETIFF. */
919 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
920 IFF_VNET_HDR,
921 (unsigned int __user*)argp);
922 }
923
924 if (!tun)
925 return -EBADFD;
926
927 DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
928
929 switch (cmd) {
930 case TUNGETIFF:
931 ret = tun_get_iff(current->nsproxy->net_ns, file, &ifr);
932 if (ret)
933 return ret;
934
935 if (copy_to_user(argp, &ifr, sizeof(ifr)))
936 return -EFAULT;
937 break;
938
939 case TUNSETNOCSUM:
940 /* Disable/Enable checksum */
941 if (arg)
942 tun->flags |= TUN_NOCHECKSUM;
943 else
944 tun->flags &= ~TUN_NOCHECKSUM;
945
946 DBG(KERN_INFO "%s: checksum %s\n",
947 tun->dev->name, arg ? "disabled" : "enabled");
948 break;
949
950 case TUNSETPERSIST:
951 /* Disable/Enable persist mode */
952 if (arg)
953 tun->flags |= TUN_PERSIST;
954 else
955 tun->flags &= ~TUN_PERSIST;
956
957 DBG(KERN_INFO "%s: persist %s\n",
958 tun->dev->name, arg ? "enabled" : "disabled");
959 break;
960
961 case TUNSETOWNER:
962 /* Set owner of the device */
963 tun->owner = (uid_t) arg;
964
965 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
966 break;
967
968 case TUNSETGROUP:
969 /* Set group of the device */
970 tun->group= (gid_t) arg;
971
972 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
973 break;
974
975 case TUNSETLINK:
976 /* Only allow setting the type when the interface is down */
977 rtnl_lock();
978 if (tun->dev->flags & IFF_UP) {
979 DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
980 tun->dev->name);
981 ret = -EBUSY;
982 } else {
983 tun->dev->type = (int) arg;
984 DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
985 ret = 0;
986 }
987 rtnl_unlock();
988 return ret;
989
990 #ifdef TUN_DEBUG
991 case TUNSETDEBUG:
992 tun->debug = arg;
993 break;
994 #endif
995 case TUNSETOFFLOAD:
996 rtnl_lock();
997 ret = set_offload(tun->dev, arg);
998 rtnl_unlock();
999 return ret;
1000
1001 case TUNSETTXFILTER:
1002 /* Can be set only for TAPs */
1003 if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV)
1004 return -EINVAL;
1005 rtnl_lock();
1006 ret = update_filter(&tun->txflt, (void __user *)arg);
1007 rtnl_unlock();
1008 return ret;
1009
1010 case SIOCGIFHWADDR:
1011 /* Get hw addres */
1012 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN);
1013 ifr.ifr_hwaddr.sa_family = tun->dev->type;
1014 if (copy_to_user(argp, &ifr, sizeof ifr))
1015 return -EFAULT;
1016 return 0;
1017
1018 case SIOCSIFHWADDR:
1019 /* Set hw address */
1020 DBG(KERN_DEBUG "%s: set hw address: %s\n",
1021 tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data));
1022
1023 rtnl_lock();
1024 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
1025 rtnl_unlock();
1026 return ret;
1027
1028 default:
1029 return -EINVAL;
1030 };
1031
1032 return 0;
1033 }
1034
1035 static int tun_chr_fasync(int fd, struct file *file, int on)
1036 {
1037 struct tun_struct *tun = file->private_data;
1038 int ret;
1039
1040 if (!tun)
1041 return -EBADFD;
1042
1043 DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
1044
1045 lock_kernel();
1046 if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
1047 goto out;
1048
1049 if (on) {
1050 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
1051 if (ret)
1052 goto out;
1053 tun->flags |= TUN_FASYNC;
1054 } else
1055 tun->flags &= ~TUN_FASYNC;
1056 ret = 0;
1057 out:
1058 unlock_kernel();
1059 return ret;
1060 }
1061
1062 static int tun_chr_open(struct inode *inode, struct file * file)
1063 {
1064 cycle_kernel_lock();
1065 DBG1(KERN_INFO "tunX: tun_chr_open\n");
1066 file->private_data = NULL;
1067 return 0;
1068 }
1069
1070 static int tun_chr_close(struct inode *inode, struct file *file)
1071 {
1072 struct tun_struct *tun = file->private_data;
1073
1074 if (!tun)
1075 return 0;
1076
1077 DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
1078
1079 rtnl_lock();
1080
1081 /* Detach from net device */
1082 file->private_data = NULL;
1083 tun->attached = 0;
1084 put_net(dev_net(tun->dev));
1085
1086 /* Drop read queue */
1087 skb_queue_purge(&tun->readq);
1088
1089 if (!(tun->flags & TUN_PERSIST)) {
1090 list_del(&tun->list);
1091 unregister_netdevice(tun->dev);
1092 }
1093
1094 rtnl_unlock();
1095
1096 return 0;
1097 }
1098
1099 static const struct file_operations tun_fops = {
1100 .owner = THIS_MODULE,
1101 .llseek = no_llseek,
1102 .read = do_sync_read,
1103 .aio_read = tun_chr_aio_read,
1104 .write = do_sync_write,
1105 .aio_write = tun_chr_aio_write,
1106 .poll = tun_chr_poll,
1107 .ioctl = tun_chr_ioctl,
1108 .open = tun_chr_open,
1109 .release = tun_chr_close,
1110 .fasync = tun_chr_fasync
1111 };
1112
1113 static struct miscdevice tun_miscdev = {
1114 .minor = TUN_MINOR,
1115 .name = "tun",
1116 .fops = &tun_fops,
1117 };
1118
1119 /* ethtool interface */
1120
1121 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
1122 {
1123 cmd->supported = 0;
1124 cmd->advertising = 0;
1125 cmd->speed = SPEED_10;
1126 cmd->duplex = DUPLEX_FULL;
1127 cmd->port = PORT_TP;
1128 cmd->phy_address = 0;
1129 cmd->transceiver = XCVR_INTERNAL;
1130 cmd->autoneg = AUTONEG_DISABLE;
1131 cmd->maxtxpkt = 0;
1132 cmd->maxrxpkt = 0;
1133 return 0;
1134 }
1135
1136 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
1137 {
1138 struct tun_struct *tun = netdev_priv(dev);
1139
1140 strcpy(info->driver, DRV_NAME);
1141 strcpy(info->version, DRV_VERSION);
1142 strcpy(info->fw_version, "N/A");
1143
1144 switch (tun->flags & TUN_TYPE_MASK) {
1145 case TUN_TUN_DEV:
1146 strcpy(info->bus_info, "tun");
1147 break;
1148 case TUN_TAP_DEV:
1149 strcpy(info->bus_info, "tap");
1150 break;
1151 }
1152 }
1153
1154 static u32 tun_get_msglevel(struct net_device *dev)
1155 {
1156 #ifdef TUN_DEBUG
1157 struct tun_struct *tun = netdev_priv(dev);
1158 return tun->debug;
1159 #else
1160 return -EOPNOTSUPP;
1161 #endif
1162 }
1163
1164 static void tun_set_msglevel(struct net_device *dev, u32 value)
1165 {
1166 #ifdef TUN_DEBUG
1167 struct tun_struct *tun = netdev_priv(dev);
1168 tun->debug = value;
1169 #endif
1170 }
1171
1172 static u32 tun_get_link(struct net_device *dev)
1173 {
1174 struct tun_struct *tun = netdev_priv(dev);
1175 return tun->attached;
1176 }
1177
1178 static u32 tun_get_rx_csum(struct net_device *dev)
1179 {
1180 struct tun_struct *tun = netdev_priv(dev);
1181 return (tun->flags & TUN_NOCHECKSUM) == 0;
1182 }
1183
1184 static int tun_set_rx_csum(struct net_device *dev, u32 data)
1185 {
1186 struct tun_struct *tun = netdev_priv(dev);
1187 if (data)
1188 tun->flags &= ~TUN_NOCHECKSUM;
1189 else
1190 tun->flags |= TUN_NOCHECKSUM;
1191 return 0;
1192 }
1193
1194 static const struct ethtool_ops tun_ethtool_ops = {
1195 .get_settings = tun_get_settings,
1196 .get_drvinfo = tun_get_drvinfo,
1197 .get_msglevel = tun_get_msglevel,
1198 .set_msglevel = tun_set_msglevel,
1199 .get_link = tun_get_link,
1200 .get_rx_csum = tun_get_rx_csum,
1201 .set_rx_csum = tun_set_rx_csum
1202 };
1203
1204 static int tun_init_net(struct net *net)
1205 {
1206 struct tun_net *tn;
1207
1208 tn = kmalloc(sizeof(*tn), GFP_KERNEL);
1209 if (tn == NULL)
1210 return -ENOMEM;
1211
1212 INIT_LIST_HEAD(&tn->dev_list);
1213
1214 if (net_assign_generic(net, tun_net_id, tn)) {
1215 kfree(tn);
1216 return -ENOMEM;
1217 }
1218
1219 return 0;
1220 }
1221
1222 static void tun_exit_net(struct net *net)
1223 {
1224 struct tun_net *tn;
1225 struct tun_struct *tun, *nxt;
1226
1227 tn = net_generic(net, tun_net_id);
1228
1229 rtnl_lock();
1230 list_for_each_entry_safe(tun, nxt, &tn->dev_list, list) {
1231 DBG(KERN_INFO "%s cleaned up\n", tun->dev->name);
1232 unregister_netdevice(tun->dev);
1233 }
1234 rtnl_unlock();
1235
1236 kfree(tn);
1237 }
1238
1239 static struct pernet_operations tun_net_ops = {
1240 .init = tun_init_net,
1241 .exit = tun_exit_net,
1242 };
1243
1244 static int __init tun_init(void)
1245 {
1246 int ret = 0;
1247
1248 printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1249 printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1250
1251 ret = register_pernet_gen_device(&tun_net_id, &tun_net_ops);
1252 if (ret) {
1253 printk(KERN_ERR "tun: Can't register pernet ops\n");
1254 goto err_pernet;
1255 }
1256
1257 ret = misc_register(&tun_miscdev);
1258 if (ret) {
1259 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1260 goto err_misc;
1261 }
1262 return 0;
1263
1264 err_misc:
1265 unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1266 err_pernet:
1267 return ret;
1268 }
1269
1270 static void tun_cleanup(void)
1271 {
1272 misc_deregister(&tun_miscdev);
1273 unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1274 }
1275
1276 module_init(tun_init);
1277 module_exit(tun_cleanup);
1278 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1279 MODULE_AUTHOR(DRV_COPYRIGHT);
1280 MODULE_LICENSE("GPL");
1281 MODULE_ALIAS_MISCDEV(TUN_MINOR);
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree