tun: Fix minor race in TUNSETLINK ioctl handling.
[safe/jmp/linux-2.6] / 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  *  Brian Braunstein <linuxkernel@bristyle.com> 2007/03/23
22  *    Fixed hw address handling.  Now net_device.dev_addr is kept consistent
23  *    with tun.dev_addr when the address is set by this module.
24  *
25  *  Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14
26  *    Add TUNSETLINK ioctl to set the link encapsulation
27  *
28  *  Mark Smith <markzzzsmith@yahoo.com.au>
29  *   Use random_ether_addr() for tap MAC address.
30  *
31  *  Harald Roelle <harald.roelle@ifi.lmu.de>  2004/04/20
32  *    Fixes in packet dropping, queue length setting and queue wakeup.
33  *    Increased default tx queue length.
34  *    Added ethtool API.
35  *    Minor cleanups
36  *
37  *  Daniel Podlejski <underley@underley.eu.org>
38  *    Modifications for 2.3.99-pre5 kernel.
39  */
40
41 #define DRV_NAME        "tun"
42 #define DRV_VERSION     "1.6"
43 #define DRV_DESCRIPTION "Universal TUN/TAP device driver"
44 #define DRV_COPYRIGHT   "(C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>"
45
46 #include <linux/module.h>
47 #include <linux/errno.h>
48 #include <linux/kernel.h>
49 #include <linux/major.h>
50 #include <linux/slab.h>
51 #include <linux/poll.h>
52 #include <linux/fcntl.h>
53 #include <linux/init.h>
54 #include <linux/skbuff.h>
55 #include <linux/netdevice.h>
56 #include <linux/etherdevice.h>
57 #include <linux/miscdevice.h>
58 #include <linux/ethtool.h>
59 #include <linux/rtnetlink.h>
60 #include <linux/if.h>
61 #include <linux/if_arp.h>
62 #include <linux/if_ether.h>
63 #include <linux/if_tun.h>
64 #include <linux/crc32.h>
65 #include <linux/nsproxy.h>
66 #include <net/net_namespace.h>
67 #include <net/netns/generic.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 struct tun_struct {
86         struct list_head        list;
87         unsigned long           flags;
88         int                     attached;
89         uid_t                   owner;
90         gid_t                   group;
91
92         wait_queue_head_t       read_wait;
93         struct sk_buff_head     readq;
94
95         struct net_device       *dev;
96
97         struct fasync_struct    *fasync;
98
99         unsigned long if_flags;
100         u8 dev_addr[ETH_ALEN];
101         u32 chr_filter[2];
102         u32 net_filter[2];
103
104 #ifdef TUN_DEBUG
105         int debug;
106 #endif
107 };
108
109 /* Network device part of the driver */
110
111 static unsigned int tun_net_id;
112 struct tun_net {
113         struct list_head dev_list;
114 };
115
116 static const struct ethtool_ops tun_ethtool_ops;
117
118 /* Net device open. */
119 static int tun_net_open(struct net_device *dev)
120 {
121         netif_start_queue(dev);
122         return 0;
123 }
124
125 /* Net device close. */
126 static int tun_net_close(struct net_device *dev)
127 {
128         netif_stop_queue(dev);
129         return 0;
130 }
131
132 /* Net device start xmit */
133 static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
134 {
135         struct tun_struct *tun = netdev_priv(dev);
136
137         DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len);
138
139         /* Drop packet if interface is not attached */
140         if (!tun->attached)
141                 goto drop;
142
143         /* Packet dropping */
144         if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) {
145                 if (!(tun->flags & TUN_ONE_QUEUE)) {
146                         /* Normal queueing mode. */
147                         /* Packet scheduler handles dropping of further packets. */
148                         netif_stop_queue(dev);
149
150                         /* We won't see all dropped packets individually, so overrun
151                          * error is more appropriate. */
152                         dev->stats.tx_fifo_errors++;
153                 } else {
154                         /* Single queue mode.
155                          * Driver handles dropping of all packets itself. */
156                         goto drop;
157                 }
158         }
159
160         /* Queue packet */
161         skb_queue_tail(&tun->readq, skb);
162         dev->trans_start = jiffies;
163
164         /* Notify and wake up reader process */
165         if (tun->flags & TUN_FASYNC)
166                 kill_fasync(&tun->fasync, SIGIO, POLL_IN);
167         wake_up_interruptible(&tun->read_wait);
168         return 0;
169
170 drop:
171         dev->stats.tx_dropped++;
172         kfree_skb(skb);
173         return 0;
174 }
175
176 /** Add the specified Ethernet address to this multicast filter. */
177 static void
178 add_multi(u32* filter, const u8* addr)
179 {
180         int bit_nr = ether_crc(ETH_ALEN, addr) >> 26;
181         filter[bit_nr >> 5] |= 1 << (bit_nr & 31);
182 }
183
184 /** Remove the specified Ethernet addres from this multicast filter. */
185 static void
186 del_multi(u32* filter, const u8* addr)
187 {
188         int bit_nr = ether_crc(ETH_ALEN, addr) >> 26;
189         filter[bit_nr >> 5] &= ~(1 << (bit_nr & 31));
190 }
191
192 /** Update the list of multicast groups to which the network device belongs.
193  * This list is used to filter packets being sent from the character device to
194  * the network device. */
195 static void
196 tun_net_mclist(struct net_device *dev)
197 {
198         struct tun_struct *tun = netdev_priv(dev);
199         const struct dev_mc_list *mclist;
200         int i;
201         DECLARE_MAC_BUF(mac);
202         DBG(KERN_DEBUG "%s: tun_net_mclist: mc_count %d\n",
203                         dev->name, dev->mc_count);
204         memset(tun->chr_filter, 0, sizeof tun->chr_filter);
205         for (i = 0, mclist = dev->mc_list; i < dev->mc_count && mclist != NULL;
206                         i++, mclist = mclist->next) {
207                 add_multi(tun->net_filter, mclist->dmi_addr);
208                 DBG(KERN_DEBUG "%s: tun_net_mclist: %s\n",
209                     dev->name, print_mac(mac, mclist->dmi_addr));
210         }
211 }
212
213 #define MIN_MTU 68
214 #define MAX_MTU 65535
215
216 static int
217 tun_net_change_mtu(struct net_device *dev, int new_mtu)
218 {
219         if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
220                 return -EINVAL;
221         dev->mtu = new_mtu;
222         return 0;
223 }
224
225 /* Initialize net device. */
226 static void tun_net_init(struct net_device *dev)
227 {
228         struct tun_struct *tun = netdev_priv(dev);
229
230         switch (tun->flags & TUN_TYPE_MASK) {
231         case TUN_TUN_DEV:
232                 /* Point-to-Point TUN Device */
233                 dev->hard_header_len = 0;
234                 dev->addr_len = 0;
235                 dev->mtu = 1500;
236                 dev->change_mtu = tun_net_change_mtu;
237
238                 /* Zero header length */
239                 dev->type = ARPHRD_NONE;
240                 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
241                 dev->tx_queue_len = TUN_READQ_SIZE;  /* We prefer our own queue length */
242                 break;
243
244         case TUN_TAP_DEV:
245                 /* Ethernet TAP Device */
246                 dev->set_multicast_list = tun_net_mclist;
247
248                 ether_setup(dev);
249                 dev->change_mtu = tun_net_change_mtu;
250
251                 /* random address already created for us by tun_set_iff, use it */
252                 memcpy(dev->dev_addr, tun->dev_addr, min(sizeof(tun->dev_addr), sizeof(dev->dev_addr)) );
253
254                 dev->tx_queue_len = TUN_READQ_SIZE;  /* We prefer our own queue length */
255                 break;
256         }
257 }
258
259 /* Character device part */
260
261 /* Poll */
262 static unsigned int tun_chr_poll(struct file *file, poll_table * wait)
263 {
264         struct tun_struct *tun = file->private_data;
265         unsigned int mask = POLLOUT | POLLWRNORM;
266
267         if (!tun)
268                 return -EBADFD;
269
270         DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name);
271
272         poll_wait(file, &tun->read_wait, wait);
273
274         if (!skb_queue_empty(&tun->readq))
275                 mask |= POLLIN | POLLRDNORM;
276
277         return mask;
278 }
279
280 /* Get packet from user space buffer */
281 static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, size_t count)
282 {
283         struct tun_pi pi = { 0, __constant_htons(ETH_P_IP) };
284         struct sk_buff *skb;
285         size_t len = count, align = 0;
286
287         if (!(tun->flags & TUN_NO_PI)) {
288                 if ((len -= sizeof(pi)) > count)
289                         return -EINVAL;
290
291                 if(memcpy_fromiovec((void *)&pi, iv, sizeof(pi)))
292                         return -EFAULT;
293         }
294
295         if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) {
296                 align = NET_IP_ALIGN;
297                 if (unlikely(len < ETH_HLEN))
298                         return -EINVAL;
299         }
300
301         if (!(skb = alloc_skb(len + align, GFP_KERNEL))) {
302                 tun->dev->stats.rx_dropped++;
303                 return -ENOMEM;
304         }
305
306         if (align)
307                 skb_reserve(skb, align);
308         if (memcpy_fromiovec(skb_put(skb, len), iv, len)) {
309                 tun->dev->stats.rx_dropped++;
310                 kfree_skb(skb);
311                 return -EFAULT;
312         }
313
314         switch (tun->flags & TUN_TYPE_MASK) {
315         case TUN_TUN_DEV:
316                 skb_reset_mac_header(skb);
317                 skb->protocol = pi.proto;
318                 skb->dev = tun->dev;
319                 break;
320         case TUN_TAP_DEV:
321                 skb->protocol = eth_type_trans(skb, tun->dev);
322                 break;
323         };
324
325         if (tun->flags & TUN_NOCHECKSUM)
326                 skb->ip_summed = CHECKSUM_UNNECESSARY;
327
328         netif_rx_ni(skb);
329         tun->dev->last_rx = jiffies;
330
331         tun->dev->stats.rx_packets++;
332         tun->dev->stats.rx_bytes += len;
333
334         return count;
335 }
336
337 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
338                               unsigned long count, loff_t pos)
339 {
340         struct tun_struct *tun = iocb->ki_filp->private_data;
341
342         if (!tun)
343                 return -EBADFD;
344
345         DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
346
347         return tun_get_user(tun, (struct iovec *) iv, iov_length(iv, count));
348 }
349
350 /* Put packet to the user space buffer */
351 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
352                                        struct sk_buff *skb,
353                                        struct iovec *iv, int len)
354 {
355         struct tun_pi pi = { 0, skb->protocol };
356         ssize_t total = 0;
357
358         if (!(tun->flags & TUN_NO_PI)) {
359                 if ((len -= sizeof(pi)) < 0)
360                         return -EINVAL;
361
362                 if (len < skb->len) {
363                         /* Packet will be striped */
364                         pi.flags |= TUN_PKT_STRIP;
365                 }
366
367                 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
368                         return -EFAULT;
369                 total += sizeof(pi);
370         }
371
372         len = min_t(int, skb->len, len);
373
374         skb_copy_datagram_iovec(skb, 0, iv, len);
375         total += len;
376
377         tun->dev->stats.tx_packets++;
378         tun->dev->stats.tx_bytes += len;
379
380         return total;
381 }
382
383 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
384                             unsigned long count, loff_t pos)
385 {
386         struct file *file = iocb->ki_filp;
387         struct tun_struct *tun = file->private_data;
388         DECLARE_WAITQUEUE(wait, current);
389         struct sk_buff *skb;
390         ssize_t len, ret = 0;
391         DECLARE_MAC_BUF(mac);
392
393         if (!tun)
394                 return -EBADFD;
395
396         DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
397
398         len = iov_length(iv, count);
399         if (len < 0)
400                 return -EINVAL;
401
402         add_wait_queue(&tun->read_wait, &wait);
403         while (len) {
404                 const u8 ones[ ETH_ALEN] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
405                 u8 addr[ ETH_ALEN];
406                 int bit_nr;
407
408                 current->state = TASK_INTERRUPTIBLE;
409
410                 /* Read frames from the queue */
411                 if (!(skb=skb_dequeue(&tun->readq))) {
412                         if (file->f_flags & O_NONBLOCK) {
413                                 ret = -EAGAIN;
414                                 break;
415                         }
416                         if (signal_pending(current)) {
417                                 ret = -ERESTARTSYS;
418                                 break;
419                         }
420
421                         /* Nothing to read, let's sleep */
422                         schedule();
423                         continue;
424                 }
425                 netif_wake_queue(tun->dev);
426
427                 /** Decide whether to accept this packet. This code is designed to
428                  * behave identically to an Ethernet interface. Accept the packet if
429                  * - we are promiscuous.
430                  * - the packet is addressed to us.
431                  * - the packet is broadcast.
432                  * - the packet is multicast and
433                  *   - we are multicast promiscous.
434                  *   - we belong to the multicast group.
435                  */
436                 skb_copy_from_linear_data(skb, addr, min_t(size_t, sizeof addr,
437                                                                    skb->len));
438                 bit_nr = ether_crc(sizeof addr, addr) >> 26;
439                 if ((tun->if_flags & IFF_PROMISC) ||
440                                 memcmp(addr, tun->dev_addr, sizeof addr) == 0 ||
441                                 memcmp(addr, ones, sizeof addr) == 0 ||
442                                 (((addr[0] == 1 && addr[1] == 0 && addr[2] == 0x5e) ||
443                                   (addr[0] == 0x33 && addr[1] == 0x33)) &&
444                                  ((tun->if_flags & IFF_ALLMULTI) ||
445                                   (tun->chr_filter[bit_nr >> 5] & (1 << (bit_nr & 31)))))) {
446                         DBG(KERN_DEBUG "%s: tun_chr_readv: accepted: %s\n",
447                                         tun->dev->name, print_mac(mac, addr));
448                         ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
449                         kfree_skb(skb);
450                         break;
451                 } else {
452                         DBG(KERN_DEBUG "%s: tun_chr_readv: rejected: %s\n",
453                                         tun->dev->name, print_mac(mac, addr));
454                         kfree_skb(skb);
455                         continue;
456                 }
457         }
458
459         current->state = TASK_RUNNING;
460         remove_wait_queue(&tun->read_wait, &wait);
461
462         return ret;
463 }
464
465 static void tun_setup(struct net_device *dev)
466 {
467         struct tun_struct *tun = netdev_priv(dev);
468
469         skb_queue_head_init(&tun->readq);
470         init_waitqueue_head(&tun->read_wait);
471
472         tun->owner = -1;
473         tun->group = -1;
474
475         dev->open = tun_net_open;
476         dev->hard_start_xmit = tun_net_xmit;
477         dev->stop = tun_net_close;
478         dev->ethtool_ops = &tun_ethtool_ops;
479         dev->destructor = free_netdev;
480         dev->features |= NETIF_F_NETNS_LOCAL;
481 }
482
483 static struct tun_struct *tun_get_by_name(struct tun_net *tn, const char *name)
484 {
485         struct tun_struct *tun;
486
487         ASSERT_RTNL();
488         list_for_each_entry(tun, &tn->dev_list, list) {
489                 if (!strncmp(tun->dev->name, name, IFNAMSIZ))
490                     return tun;
491         }
492
493         return NULL;
494 }
495
496 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
497 {
498         struct tun_net *tn;
499         struct tun_struct *tun;
500         struct net_device *dev;
501         int err;
502
503         tn = net_generic(net, tun_net_id);
504         tun = tun_get_by_name(tn, ifr->ifr_name);
505         if (tun) {
506                 if (tun->attached)
507                         return -EBUSY;
508
509                 /* Check permissions */
510                 if (((tun->owner != -1 &&
511                       current->euid != tun->owner) ||
512                      (tun->group != -1 &&
513                       current->egid != tun->group)) &&
514                      !capable(CAP_NET_ADMIN))
515                         return -EPERM;
516         }
517         else if (__dev_get_by_name(net, ifr->ifr_name))
518                 return -EINVAL;
519         else {
520                 char *name;
521                 unsigned long flags = 0;
522
523                 err = -EINVAL;
524
525                 if (!capable(CAP_NET_ADMIN))
526                         return -EPERM;
527
528                 /* Set dev type */
529                 if (ifr->ifr_flags & IFF_TUN) {
530                         /* TUN device */
531                         flags |= TUN_TUN_DEV;
532                         name = "tun%d";
533                 } else if (ifr->ifr_flags & IFF_TAP) {
534                         /* TAP device */
535                         flags |= TUN_TAP_DEV;
536                         name = "tap%d";
537                 } else
538                         goto failed;
539
540                 if (*ifr->ifr_name)
541                         name = ifr->ifr_name;
542
543                 dev = alloc_netdev(sizeof(struct tun_struct), name,
544                                    tun_setup);
545                 if (!dev)
546                         return -ENOMEM;
547
548                 dev_net_set(dev, net);
549                 tun = netdev_priv(dev);
550                 tun->dev = dev;
551                 tun->flags = flags;
552                 /* Be promiscuous by default to maintain previous behaviour. */
553                 tun->if_flags = IFF_PROMISC;
554                 /* Generate random Ethernet address. */
555                 *(__be16 *)tun->dev_addr = htons(0x00FF);
556                 get_random_bytes(tun->dev_addr + sizeof(u16), 4);
557                 memset(tun->chr_filter, 0, sizeof tun->chr_filter);
558
559                 tun_net_init(dev);
560
561                 if (strchr(dev->name, '%')) {
562                         err = dev_alloc_name(dev, dev->name);
563                         if (err < 0)
564                                 goto err_free_dev;
565                 }
566
567                 err = register_netdevice(tun->dev);
568                 if (err < 0)
569                         goto err_free_dev;
570
571                 list_add(&tun->list, &tn->dev_list);
572         }
573
574         DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
575
576         if (ifr->ifr_flags & IFF_NO_PI)
577                 tun->flags |= TUN_NO_PI;
578         else
579                 tun->flags &= ~TUN_NO_PI;
580
581         if (ifr->ifr_flags & IFF_ONE_QUEUE)
582                 tun->flags |= TUN_ONE_QUEUE;
583         else
584                 tun->flags &= ~TUN_ONE_QUEUE;
585
586         file->private_data = tun;
587         tun->attached = 1;
588         get_net(dev_net(tun->dev));
589
590         strcpy(ifr->ifr_name, tun->dev->name);
591         return 0;
592
593  err_free_dev:
594         free_netdev(dev);
595  failed:
596         return err;
597 }
598
599 static int tun_chr_ioctl(struct inode *inode, struct file *file,
600                          unsigned int cmd, unsigned long arg)
601 {
602         struct tun_struct *tun = file->private_data;
603         void __user* argp = (void __user*)arg;
604         struct ifreq ifr;
605         DECLARE_MAC_BUF(mac);
606
607         if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
608                 if (copy_from_user(&ifr, argp, sizeof ifr))
609                         return -EFAULT;
610
611         if (cmd == TUNSETIFF && !tun) {
612                 int err;
613
614                 ifr.ifr_name[IFNAMSIZ-1] = '\0';
615
616                 rtnl_lock();
617                 err = tun_set_iff(current->nsproxy->net_ns, file, &ifr);
618                 rtnl_unlock();
619
620                 if (err)
621                         return err;
622
623                 if (copy_to_user(argp, &ifr, sizeof(ifr)))
624                         return -EFAULT;
625                 return 0;
626         }
627
628         if (!tun)
629                 return -EBADFD;
630
631         DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
632
633         switch (cmd) {
634         case TUNSETNOCSUM:
635                 /* Disable/Enable checksum */
636                 if (arg)
637                         tun->flags |= TUN_NOCHECKSUM;
638                 else
639                         tun->flags &= ~TUN_NOCHECKSUM;
640
641                 DBG(KERN_INFO "%s: checksum %s\n",
642                     tun->dev->name, arg ? "disabled" : "enabled");
643                 break;
644
645         case TUNSETPERSIST:
646                 /* Disable/Enable persist mode */
647                 if (arg)
648                         tun->flags |= TUN_PERSIST;
649                 else
650                         tun->flags &= ~TUN_PERSIST;
651
652                 DBG(KERN_INFO "%s: persist %s\n",
653                     tun->dev->name, arg ? "enabled" : "disabled");
654                 break;
655
656         case TUNSETOWNER:
657                 /* Set owner of the device */
658                 tun->owner = (uid_t) arg;
659
660                 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
661                 break;
662
663         case TUNSETGROUP:
664                 /* Set group of the device */
665                 tun->group= (gid_t) arg;
666
667                 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
668                 break;
669
670         case TUNSETLINK:
671         {
672                 int ret;
673
674                 /* Only allow setting the type when the interface is down */
675                 rtnl_lock();
676                 if (tun->dev->flags & IFF_UP) {
677                         DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
678                                 tun->dev->name);
679                         ret = -EBUSY;
680                 } else {
681                         tun->dev->type = (int) arg;
682                         DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
683                         ret = 0;
684                 }
685                 rtnl_unlock();
686                 return ret;
687         }
688
689 #ifdef TUN_DEBUG
690         case TUNSETDEBUG:
691                 tun->debug = arg;
692                 break;
693 #endif
694
695         case SIOCGIFFLAGS:
696                 ifr.ifr_flags = tun->if_flags;
697                 if (copy_to_user( argp, &ifr, sizeof ifr))
698                         return -EFAULT;
699                 return 0;
700
701         case SIOCSIFFLAGS:
702                 /** Set the character device's interface flags. Currently only
703                  * IFF_PROMISC and IFF_ALLMULTI are used. */
704                 tun->if_flags = ifr.ifr_flags;
705                 DBG(KERN_INFO "%s: interface flags 0x%lx\n",
706                                 tun->dev->name, tun->if_flags);
707                 return 0;
708
709         case SIOCGIFHWADDR:
710                 /* Note: the actual net device's address may be different */
711                 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev_addr,
712                                 min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr));
713                 if (copy_to_user( argp, &ifr, sizeof ifr))
714                         return -EFAULT;
715                 return 0;
716
717         case SIOCSIFHWADDR:
718         {
719                 /* try to set the actual net device's hw address */
720                 int ret;
721
722                 rtnl_lock();
723                 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
724                 rtnl_unlock();
725
726                 if (ret == 0) {
727                         /** Set the character device's hardware address. This is used when
728                          * filtering packets being sent from the network device to the character
729                          * device. */
730                         memcpy(tun->dev_addr, ifr.ifr_hwaddr.sa_data,
731                                         min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr));
732                         DBG(KERN_DEBUG "%s: set hardware address: %x:%x:%x:%x:%x:%x\n",
733                                         tun->dev->name,
734                                         tun->dev_addr[0], tun->dev_addr[1], tun->dev_addr[2],
735                                         tun->dev_addr[3], tun->dev_addr[4], tun->dev_addr[5]);
736                 }
737
738                 return  ret;
739         }
740
741         case SIOCADDMULTI:
742                 /** Add the specified group to the character device's multicast filter
743                  * list. */
744                 add_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data);
745                 DBG(KERN_DEBUG "%s: add multi: %s\n",
746                     tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data));
747                 return 0;
748
749         case SIOCDELMULTI:
750                 /** Remove the specified group from the character device's multicast
751                  * filter list. */
752                 del_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data);
753                 DBG(KERN_DEBUG "%s: del multi: %s\n",
754                     tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data));
755                 return 0;
756
757         default:
758                 return -EINVAL;
759         };
760
761         return 0;
762 }
763
764 static int tun_chr_fasync(int fd, struct file *file, int on)
765 {
766         struct tun_struct *tun = file->private_data;
767         int ret;
768
769         if (!tun)
770                 return -EBADFD;
771
772         DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
773
774         if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
775                 return ret;
776
777         if (on) {
778                 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
779                 if (ret)
780                         return ret;
781                 tun->flags |= TUN_FASYNC;
782         } else
783                 tun->flags &= ~TUN_FASYNC;
784
785         return 0;
786 }
787
788 static int tun_chr_open(struct inode *inode, struct file * file)
789 {
790         DBG1(KERN_INFO "tunX: tun_chr_open\n");
791         file->private_data = NULL;
792         return 0;
793 }
794
795 static int tun_chr_close(struct inode *inode, struct file *file)
796 {
797         struct tun_struct *tun = file->private_data;
798
799         if (!tun)
800                 return 0;
801
802         DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
803
804         tun_chr_fasync(-1, file, 0);
805
806         rtnl_lock();
807
808         /* Detach from net device */
809         file->private_data = NULL;
810         tun->attached = 0;
811         put_net(dev_net(tun->dev));
812
813         /* Drop read queue */
814         skb_queue_purge(&tun->readq);
815
816         if (!(tun->flags & TUN_PERSIST)) {
817                 list_del(&tun->list);
818                 unregister_netdevice(tun->dev);
819         }
820
821         rtnl_unlock();
822
823         return 0;
824 }
825
826 static const struct file_operations tun_fops = {
827         .owner  = THIS_MODULE,
828         .llseek = no_llseek,
829         .read  = do_sync_read,
830         .aio_read  = tun_chr_aio_read,
831         .write = do_sync_write,
832         .aio_write = tun_chr_aio_write,
833         .poll   = tun_chr_poll,
834         .ioctl  = tun_chr_ioctl,
835         .open   = tun_chr_open,
836         .release = tun_chr_close,
837         .fasync = tun_chr_fasync
838 };
839
840 static struct miscdevice tun_miscdev = {
841         .minor = TUN_MINOR,
842         .name = "tun",
843         .fops = &tun_fops,
844 };
845
846 /* ethtool interface */
847
848 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
849 {
850         cmd->supported          = 0;
851         cmd->advertising        = 0;
852         cmd->speed              = SPEED_10;
853         cmd->duplex             = DUPLEX_FULL;
854         cmd->port               = PORT_TP;
855         cmd->phy_address        = 0;
856         cmd->transceiver        = XCVR_INTERNAL;
857         cmd->autoneg            = AUTONEG_DISABLE;
858         cmd->maxtxpkt           = 0;
859         cmd->maxrxpkt           = 0;
860         return 0;
861 }
862
863 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
864 {
865         struct tun_struct *tun = netdev_priv(dev);
866
867         strcpy(info->driver, DRV_NAME);
868         strcpy(info->version, DRV_VERSION);
869         strcpy(info->fw_version, "N/A");
870
871         switch (tun->flags & TUN_TYPE_MASK) {
872         case TUN_TUN_DEV:
873                 strcpy(info->bus_info, "tun");
874                 break;
875         case TUN_TAP_DEV:
876                 strcpy(info->bus_info, "tap");
877                 break;
878         }
879 }
880
881 static u32 tun_get_msglevel(struct net_device *dev)
882 {
883 #ifdef TUN_DEBUG
884         struct tun_struct *tun = netdev_priv(dev);
885         return tun->debug;
886 #else
887         return -EOPNOTSUPP;
888 #endif
889 }
890
891 static void tun_set_msglevel(struct net_device *dev, u32 value)
892 {
893 #ifdef TUN_DEBUG
894         struct tun_struct *tun = netdev_priv(dev);
895         tun->debug = value;
896 #endif
897 }
898
899 static u32 tun_get_link(struct net_device *dev)
900 {
901         struct tun_struct *tun = netdev_priv(dev);
902         return tun->attached;
903 }
904
905 static u32 tun_get_rx_csum(struct net_device *dev)
906 {
907         struct tun_struct *tun = netdev_priv(dev);
908         return (tun->flags & TUN_NOCHECKSUM) == 0;
909 }
910
911 static int tun_set_rx_csum(struct net_device *dev, u32 data)
912 {
913         struct tun_struct *tun = netdev_priv(dev);
914         if (data)
915                 tun->flags &= ~TUN_NOCHECKSUM;
916         else
917                 tun->flags |= TUN_NOCHECKSUM;
918         return 0;
919 }
920
921 static const struct ethtool_ops tun_ethtool_ops = {
922         .get_settings   = tun_get_settings,
923         .get_drvinfo    = tun_get_drvinfo,
924         .get_msglevel   = tun_get_msglevel,
925         .set_msglevel   = tun_set_msglevel,
926         .get_link       = tun_get_link,
927         .get_rx_csum    = tun_get_rx_csum,
928         .set_rx_csum    = tun_set_rx_csum
929 };
930
931 static int tun_init_net(struct net *net)
932 {
933         struct tun_net *tn;
934
935         tn = kmalloc(sizeof(*tn), GFP_KERNEL);
936         if (tn == NULL)
937                 return -ENOMEM;
938
939         INIT_LIST_HEAD(&tn->dev_list);
940
941         if (net_assign_generic(net, tun_net_id, tn)) {
942                 kfree(tn);
943                 return -ENOMEM;
944         }
945
946         return 0;
947 }
948
949 static void tun_exit_net(struct net *net)
950 {
951         struct tun_net *tn;
952         struct tun_struct *tun, *nxt;
953
954         tn = net_generic(net, tun_net_id);
955
956         rtnl_lock();
957         list_for_each_entry_safe(tun, nxt, &tn->dev_list, list) {
958                 DBG(KERN_INFO "%s cleaned up\n", tun->dev->name);
959                 unregister_netdevice(tun->dev);
960         }
961         rtnl_unlock();
962
963         kfree(tn);
964 }
965
966 static struct pernet_operations tun_net_ops = {
967         .init = tun_init_net,
968         .exit = tun_exit_net,
969 };
970
971 static int __init tun_init(void)
972 {
973         int ret = 0;
974
975         printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
976         printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
977
978         ret = register_pernet_gen_device(&tun_net_id, &tun_net_ops);
979         if (ret) {
980                 printk(KERN_ERR "tun: Can't register pernet ops\n");
981                 goto err_pernet;
982         }
983
984         ret = misc_register(&tun_miscdev);
985         if (ret) {
986                 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
987                 goto err_misc;
988         }
989         return 0;
990
991 err_misc:
992         unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
993 err_pernet:
994         return ret;
995 }
996
997 static void tun_cleanup(void)
998 {
999         misc_deregister(&tun_miscdev);
1000         unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1001 }
1002
1003 module_init(tun_init);
1004 module_exit(tun_cleanup);
1005 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1006 MODULE_AUTHOR(DRV_COPYRIGHT);
1007 MODULE_LICENSE("GPL");
1008 MODULE_ALIAS_MISCDEV(TUN_MINOR);