SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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                 if (tun->flags & TUN_NO_PI) {
 317                         switch (skb->data[0] & 0xf0) {
 318                         case 0x40:
 319                                 pi.proto = htons(ETH_P_IP);
 320                                 break;
 321                         case 0x60:
 322                                 pi.proto = htons(ETH_P_IPV6);
 323                                 break;
 324                         default:
 325                                 tun->dev->stats.rx_dropped++;
 326                                 kfree_skb(skb);
 327                                 return -EINVAL;
 328                         }
 329                 }
 330
 331                 skb_reset_mac_header(skb);
 332                 skb->protocol = pi.proto;
 333                 skb->dev = tun->dev;
 334                 break;
 335         case TUN_TAP_DEV:
 336                 skb->protocol = eth_type_trans(skb, tun->dev);
 337                 break;
 338         };
 339
 340         if (tun->flags & TUN_NOCHECKSUM)
 341                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 342
 343         netif_rx_ni(skb);
 344         tun->dev->last_rx = jiffies;
 345
 346         tun->dev->stats.rx_packets++;
 347         tun->dev->stats.rx_bytes += len;
 348
 349         return count;
 350 }
 351
 352 static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv,
 353                               unsigned long count, loff_t pos)
 354 {
 355         struct tun_struct *tun = iocb->ki_filp->private_data;
 356
 357         if (!tun)
 358                 return -EBADFD;
 359
 360         DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count);
 361
 362         return tun_get_user(tun, (struct iovec *) iv, iov_length(iv, count));
 363 }
 364
 365 /* Put packet to the user space buffer */
 366 static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
 367                                        struct sk_buff *skb,
 368                                        struct iovec *iv, int len)
 369 {
 370         struct tun_pi pi = { 0, skb->protocol };
 371         ssize_t total = 0;
 372
 373         if (!(tun->flags & TUN_NO_PI)) {
 374                 if ((len -= sizeof(pi)) < 0)
 375                         return -EINVAL;
 376
 377                 if (len < skb->len) {
 378                         /* Packet will be striped */
 379                         pi.flags |= TUN_PKT_STRIP;
 380                 }
 381
 382                 if (memcpy_toiovec(iv, (void *) &pi, sizeof(pi)))
 383                         return -EFAULT;
 384                 total += sizeof(pi);
 385         }
 386
 387         len = min_t(int, skb->len, len);
 388
 389         skb_copy_datagram_iovec(skb, 0, iv, len);
 390         total += len;
 391
 392         tun->dev->stats.tx_packets++;
 393         tun->dev->stats.tx_bytes += len;
 394
 395         return total;
 396 }
 397
 398 static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv,
 399                             unsigned long count, loff_t pos)
 400 {
 401         struct file *file = iocb->ki_filp;
 402         struct tun_struct *tun = file->private_data;
 403         DECLARE_WAITQUEUE(wait, current);
 404         struct sk_buff *skb;
 405         ssize_t len, ret = 0;
 406         DECLARE_MAC_BUF(mac);
 407
 408         if (!tun)
 409                 return -EBADFD;
 410
 411         DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name);
 412
 413         len = iov_length(iv, count);
 414         if (len < 0)
 415                 return -EINVAL;
 416
 417         add_wait_queue(&tun->read_wait, &wait);
 418         while (len) {
 419                 const u8 ones[ ETH_ALEN] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
 420                 u8 addr[ ETH_ALEN];
 421                 int bit_nr;
 422
 423                 current->state = TASK_INTERRUPTIBLE;
 424
 425                 /* Read frames from the queue */
 426                 if (!(skb=skb_dequeue(&tun->readq))) {
 427                         if (file->f_flags & O_NONBLOCK) {
 428                                 ret = -EAGAIN;
 429                                 break;
 430                         }
 431                         if (signal_pending(current)) {
 432                                 ret = -ERESTARTSYS;
 433                                 break;
 434                         }
 435
 436                         /* Nothing to read, let's sleep */
 437                         schedule();
 438                         continue;
 439                 }
 440                 netif_wake_queue(tun->dev);
 441
 442                 /** Decide whether to accept this packet. This code is designed to
 443                  * behave identically to an Ethernet interface. Accept the packet if
 444                  * - we are promiscuous.
 445                  * - the packet is addressed to us.
 446                  * - the packet is broadcast.
 447                  * - the packet is multicast and
 448                  *   - we are multicast promiscous.
 449                  *   - we belong to the multicast group.
 450                  */
 451                 skb_copy_from_linear_data(skb, addr, min_t(size_t, sizeof addr,
 452                                                                    skb->len));
 453                 bit_nr = ether_crc(sizeof addr, addr) >> 26;
 454                 if ((tun->if_flags & IFF_PROMISC) ||
 455                                 memcmp(addr, tun->dev_addr, sizeof addr) == 0 ||
 456                                 memcmp(addr, ones, sizeof addr) == 0 ||
 457                                 (((addr[0] == 1 && addr[1] == 0 && addr[2] == 0x5e) ||
 458                                   (addr[0] == 0x33 && addr[1] == 0x33)) &&
 459                                  ((tun->if_flags & IFF_ALLMULTI) ||
 460                                   (tun->chr_filter[bit_nr >> 5] & (1 << (bit_nr & 31)))))) {
 461                         DBG(KERN_DEBUG "%s: tun_chr_readv: accepted: %s\n",
 462                                         tun->dev->name, print_mac(mac, addr));
 463                         ret = tun_put_user(tun, skb, (struct iovec *) iv, len);
 464                         kfree_skb(skb);
 465                         break;
 466                 } else {
 467                         DBG(KERN_DEBUG "%s: tun_chr_readv: rejected: %s\n",
 468                                         tun->dev->name, print_mac(mac, addr));
 469                         kfree_skb(skb);
 470                         continue;
 471                 }
 472         }
 473
 474         current->state = TASK_RUNNING;
 475         remove_wait_queue(&tun->read_wait, &wait);
 476
 477         return ret;
 478 }
 479
 480 static void tun_setup(struct net_device *dev)
 481 {
 482         struct tun_struct *tun = netdev_priv(dev);
 483
 484         skb_queue_head_init(&tun->readq);
 485         init_waitqueue_head(&tun->read_wait);
 486
 487         tun->owner = -1;
 488         tun->group = -1;
 489
 490         dev->open = tun_net_open;
 491         dev->hard_start_xmit = tun_net_xmit;
 492         dev->stop = tun_net_close;
 493         dev->ethtool_ops = &tun_ethtool_ops;
 494         dev->destructor = free_netdev;
 495         dev->features |= NETIF_F_NETNS_LOCAL;
 496 }
 497
 498 static struct tun_struct *tun_get_by_name(struct tun_net *tn, const char *name)
 499 {
 500         struct tun_struct *tun;
 501
 502         ASSERT_RTNL();
 503         list_for_each_entry(tun, &tn->dev_list, list) {
 504                 if (!strncmp(tun->dev->name, name, IFNAMSIZ))
 505                     return tun;
 506         }
 507
 508         return NULL;
 509 }
 510
 511 static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
 512 {
 513         struct tun_net *tn;
 514         struct tun_struct *tun;
 515         struct net_device *dev;
 516         int err;
 517
 518         tn = net_generic(net, tun_net_id);
 519         tun = tun_get_by_name(tn, ifr->ifr_name);
 520         if (tun) {
 521                 if (tun->attached)
 522                         return -EBUSY;
 523
 524                 /* Check permissions */
 525                 if (((tun->owner != -1 &&
 526                       current->euid != tun->owner) ||
 527                      (tun->group != -1 &&
 528                       current->egid != tun->group)) &&
 529                      !capable(CAP_NET_ADMIN))
 530                         return -EPERM;
 531         }
 532         else if (__dev_get_by_name(net, ifr->ifr_name))
 533                 return -EINVAL;
 534         else {
 535                 char *name;
 536                 unsigned long flags = 0;
 537
 538                 err = -EINVAL;
 539
 540                 if (!capable(CAP_NET_ADMIN))
 541                         return -EPERM;
 542
 543                 /* Set dev type */
 544                 if (ifr->ifr_flags & IFF_TUN) {
 545                         /* TUN device */
 546                         flags |= TUN_TUN_DEV;
 547                         name = "tun%d";
 548                 } else if (ifr->ifr_flags & IFF_TAP) {
 549                         /* TAP device */
 550                         flags |= TUN_TAP_DEV;
 551                         name = "tap%d";
 552                 } else
 553                         goto failed;
 554
 555                 if (*ifr->ifr_name)
 556                         name = ifr->ifr_name;
 557
 558                 dev = alloc_netdev(sizeof(struct tun_struct), name,
 559                                    tun_setup);
 560                 if (!dev)
 561                         return -ENOMEM;
 562
 563                 dev_net_set(dev, net);
 564                 tun = netdev_priv(dev);
 565                 tun->dev = dev;
 566                 tun->flags = flags;
 567                 /* Be promiscuous by default to maintain previous behaviour. */
 568                 tun->if_flags = IFF_PROMISC;
 569                 /* Generate random Ethernet address. */
 570                 *(__be16 *)tun->dev_addr = htons(0x00FF);
 571                 get_random_bytes(tun->dev_addr + sizeof(u16), 4);
 572                 memset(tun->chr_filter, 0, sizeof tun->chr_filter);
 573
 574                 tun_net_init(dev);
 575
 576                 if (strchr(dev->name, '%')) {
 577                         err = dev_alloc_name(dev, dev->name);
 578                         if (err < 0)
 579                                 goto err_free_dev;
 580                 }
 581
 582                 err = register_netdevice(tun->dev);
 583                 if (err < 0)
 584                         goto err_free_dev;
 585
 586                 list_add(&tun->list, &tn->dev_list);
 587         }
 588
 589         DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name);
 590
 591         if (ifr->ifr_flags & IFF_NO_PI)
 592                 tun->flags |= TUN_NO_PI;
 593         else
 594                 tun->flags &= ~TUN_NO_PI;
 595
 596         if (ifr->ifr_flags & IFF_ONE_QUEUE)
 597                 tun->flags |= TUN_ONE_QUEUE;
 598         else
 599                 tun->flags &= ~TUN_ONE_QUEUE;
 600
 601         file->private_data = tun;
 602         tun->attached = 1;
 603         get_net(dev_net(tun->dev));
 604
 605         strcpy(ifr->ifr_name, tun->dev->name);
 606         return 0;
 607
 608  err_free_dev:
 609         free_netdev(dev);
 610  failed:
 611         return err;
 612 }
 613
 614 /* This is like a cut-down ethtool ops, except done via tun fd so no
 615  * privs required. */
 616 static int set_offload(struct net_device *dev, unsigned long arg)
 617 {
 618         unsigned int old_features, features;
 619
 620         old_features = dev->features;
 621         /* Unset features, set them as we chew on the arg. */
 622         features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST
 623                                     |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6));
 624
 625         if (arg & TUN_F_CSUM) {
 626                 features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST;
 627                 arg &= ~TUN_F_CSUM;
 628
 629                 if (arg & (TUN_F_TSO4|TUN_F_TSO6)) {
 630                         if (arg & TUN_F_TSO_ECN) {
 631                                 features |= NETIF_F_TSO_ECN;
 632                                 arg &= ~TUN_F_TSO_ECN;
 633                         }
 634                         if (arg & TUN_F_TSO4)
 635                                 features |= NETIF_F_TSO;
 636                         if (arg & TUN_F_TSO6)
 637                                 features |= NETIF_F_TSO6;
 638                         arg &= ~(TUN_F_TSO4|TUN_F_TSO6);
 639                 }
 640         }
 641
 642         /* This gives the user a way to test for new features in future by
 643          * trying to set them. */
 644         if (arg)
 645                 return -EINVAL;
 646
 647         dev->features = features;
 648         if (old_features != dev->features)
 649                 netdev_features_change(dev);
 650
 651         return 0;
 652 }
 653
 654 static int tun_chr_ioctl(struct inode *inode, struct file *file,
 655                          unsigned int cmd, unsigned long arg)
 656 {
 657         struct tun_struct *tun = file->private_data;
 658         void __user* argp = (void __user*)arg;
 659         struct ifreq ifr;
 660         DECLARE_MAC_BUF(mac);
 661
 662         if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89)
 663                 if (copy_from_user(&ifr, argp, sizeof ifr))
 664                         return -EFAULT;
 665
 666         if (cmd == TUNSETIFF && !tun) {
 667                 int err;
 668
 669                 ifr.ifr_name[IFNAMSIZ-1] = '\0';
 670
 671                 rtnl_lock();
 672                 err = tun_set_iff(current->nsproxy->net_ns, file, &ifr);
 673                 rtnl_unlock();
 674
 675                 if (err)
 676                         return err;
 677
 678                 if (copy_to_user(argp, &ifr, sizeof(ifr)))
 679                         return -EFAULT;
 680                 return 0;
 681         }
 682
 683         if (cmd == TUNGETFEATURES) {
 684                 /* Currently this just means: "what IFF flags are valid?".
 685                  * This is needed because we never checked for invalid flags on
 686                  * TUNSETIFF. */
 687                 return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE,
 688                                 (unsigned int __user*)argp);
 689         }
 690
 691         if (!tun)
 692                 return -EBADFD;
 693
 694         DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd);
 695
 696         switch (cmd) {
 697         case TUNSETNOCSUM:
 698                 /* Disable/Enable checksum */
 699                 if (arg)
 700                         tun->flags |= TUN_NOCHECKSUM;
 701                 else
 702                         tun->flags &= ~TUN_NOCHECKSUM;
 703
 704                 DBG(KERN_INFO "%s: checksum %s\n",
 705                     tun->dev->name, arg ? "disabled" : "enabled");
 706                 break;
 707
 708         case TUNSETPERSIST:
 709                 /* Disable/Enable persist mode */
 710                 if (arg)
 711                         tun->flags |= TUN_PERSIST;
 712                 else
 713                         tun->flags &= ~TUN_PERSIST;
 714
 715                 DBG(KERN_INFO "%s: persist %s\n",
 716                     tun->dev->name, arg ? "enabled" : "disabled");
 717                 break;
 718
 719         case TUNSETOWNER:
 720                 /* Set owner of the device */
 721                 tun->owner = (uid_t) arg;
 722
 723                 DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner);
 724                 break;
 725
 726         case TUNSETGROUP:
 727                 /* Set group of the device */
 728                 tun->group= (gid_t) arg;
 729
 730                 DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group);
 731                 break;
 732
 733         case TUNSETLINK:
 734         {
 735                 int ret;
 736
 737                 /* Only allow setting the type when the interface is down */
 738                 rtnl_lock();
 739                 if (tun->dev->flags & IFF_UP) {
 740                         DBG(KERN_INFO "%s: Linktype set failed because interface is up\n",
 741                                 tun->dev->name);
 742                         ret = -EBUSY;
 743                 } else {
 744                         tun->dev->type = (int) arg;
 745                         DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type);
 746                         ret = 0;
 747                 }
 748                 rtnl_unlock();
 749                 return ret;
 750         }
 751
 752 #ifdef TUN_DEBUG
 753         case TUNSETDEBUG:
 754                 tun->debug = arg;
 755                 break;
 756 #endif
 757
 758         case TUNSETOFFLOAD:
 759         {
 760                 int ret;
 761                 rtnl_lock();
 762                 ret = set_offload(tun->dev, arg);
 763                 rtnl_unlock();
 764                 return ret;
 765         }
 766
 767         case SIOCGIFFLAGS:
 768                 ifr.ifr_flags = tun->if_flags;
 769                 if (copy_to_user( argp, &ifr, sizeof ifr))
 770                         return -EFAULT;
 771                 return 0;
 772
 773         case SIOCSIFFLAGS:
 774                 /** Set the character device's interface flags. Currently only
 775                  * IFF_PROMISC and IFF_ALLMULTI are used. */
 776                 tun->if_flags = ifr.ifr_flags;
 777                 DBG(KERN_INFO "%s: interface flags 0x%lx\n",
 778                                 tun->dev->name, tun->if_flags);
 779                 return 0;
 780
 781         case SIOCGIFHWADDR:
 782                 /* Note: the actual net device's address may be different */
 783                 memcpy(ifr.ifr_hwaddr.sa_data, tun->dev_addr,
 784                                 min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr));
 785                 if (copy_to_user( argp, &ifr, sizeof ifr))
 786                         return -EFAULT;
 787                 return 0;
 788
 789         case SIOCSIFHWADDR:
 790         {
 791                 /* try to set the actual net device's hw address */
 792                 int ret;
 793
 794                 rtnl_lock();
 795                 ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr);
 796                 rtnl_unlock();
 797
 798                 if (ret == 0) {
 799                         /** Set the character device's hardware address. This is used when
 800                          * filtering packets being sent from the network device to the character
 801                          * device. */
 802                         memcpy(tun->dev_addr, ifr.ifr_hwaddr.sa_data,
 803                                         min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr));
 804                         DBG(KERN_DEBUG "%s: set hardware address: %x:%x:%x:%x:%x:%x\n",
 805                                         tun->dev->name,
 806                                         tun->dev_addr[0], tun->dev_addr[1], tun->dev_addr[2],
 807                                         tun->dev_addr[3], tun->dev_addr[4], tun->dev_addr[5]);
 808                 }
 809
 810                 return  ret;
 811         }
 812
 813         case SIOCADDMULTI:
 814                 /** Add the specified group to the character device's multicast filter
 815                  * list. */
 816                 rtnl_lock();
 817                 netif_tx_lock_bh(tun->dev);
 818                 add_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data);
 819                 netif_tx_unlock_bh(tun->dev);
 820                 rtnl_unlock();
 821
 822                 DBG(KERN_DEBUG "%s: add multi: %s\n",
 823                     tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data));
 824                 return 0;
 825
 826         case SIOCDELMULTI:
 827                 /** Remove the specified group from the character device's multicast
 828                  * filter list. */
 829                 rtnl_lock();
 830                 netif_tx_lock_bh(tun->dev);
 831                 del_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data);
 832                 netif_tx_unlock_bh(tun->dev);
 833                 rtnl_unlock();
 834
 835                 DBG(KERN_DEBUG "%s: del multi: %s\n",
 836                     tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data));
 837                 return 0;
 838
 839         default:
 840                 return -EINVAL;
 841         };
 842
 843         return 0;
 844 }
 845
 846 static int tun_chr_fasync(int fd, struct file *file, int on)
 847 {
 848         struct tun_struct *tun = file->private_data;
 849         int ret;
 850
 851         if (!tun)
 852                 return -EBADFD;
 853
 854         DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on);
 855
 856         if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0)
 857                 return ret;
 858
 859         if (on) {
 860                 ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0);
 861                 if (ret)
 862                         return ret;
 863                 tun->flags |= TUN_FASYNC;
 864         } else
 865                 tun->flags &= ~TUN_FASYNC;
 866
 867         return 0;
 868 }
 869
 870 static int tun_chr_open(struct inode *inode, struct file * file)
 871 {
 872         DBG1(KERN_INFO "tunX: tun_chr_open\n");
 873         file->private_data = NULL;
 874         return 0;
 875 }
 876
 877 static int tun_chr_close(struct inode *inode, struct file *file)
 878 {
 879         struct tun_struct *tun = file->private_data;
 880
 881         if (!tun)
 882                 return 0;
 883
 884         DBG(KERN_INFO "%s: tun_chr_close\n", tun->dev->name);
 885
 886         tun_chr_fasync(-1, file, 0);
 887
 888         rtnl_lock();
 889
 890         /* Detach from net device */
 891         file->private_data = NULL;
 892         tun->attached = 0;
 893         put_net(dev_net(tun->dev));
 894
 895         /* Drop read queue */
 896         skb_queue_purge(&tun->readq);
 897
 898         if (!(tun->flags & TUN_PERSIST)) {
 899                 list_del(&tun->list);
 900                 unregister_netdevice(tun->dev);
 901         }
 902
 903         rtnl_unlock();
 904
 905         return 0;
 906 }
 907
 908 static const struct file_operations tun_fops = {
 909         .owner  = THIS_MODULE,
 910         .llseek = no_llseek,
 911         .read  = do_sync_read,
 912         .aio_read  = tun_chr_aio_read,
 913         .write = do_sync_write,
 914         .aio_write = tun_chr_aio_write,
 915         .poll   = tun_chr_poll,
 916         .ioctl  = tun_chr_ioctl,
 917         .open   = tun_chr_open,
 918         .release = tun_chr_close,
 919         .fasync = tun_chr_fasync
 920 };
 921
 922 static struct miscdevice tun_miscdev = {
 923         .minor = TUN_MINOR,
 924         .name = "tun",
 925         .fops = &tun_fops,
 926 };
 927
 928 /* ethtool interface */
 929
 930 static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
 931 {
 932         cmd->supported          = 0;
 933         cmd->advertising        = 0;
 934         cmd->speed              = SPEED_10;
 935         cmd->duplex             = DUPLEX_FULL;
 936         cmd->port               = PORT_TP;
 937         cmd->phy_address        = 0;
 938         cmd->transceiver        = XCVR_INTERNAL;
 939         cmd->autoneg            = AUTONEG_DISABLE;
 940         cmd->maxtxpkt           = 0;
 941         cmd->maxrxpkt           = 0;
 942         return 0;
 943 }
 944
 945 static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
 946 {
 947         struct tun_struct *tun = netdev_priv(dev);
 948
 949         strcpy(info->driver, DRV_NAME);
 950         strcpy(info->version, DRV_VERSION);
 951         strcpy(info->fw_version, "N/A");
 952
 953         switch (tun->flags & TUN_TYPE_MASK) {
 954         case TUN_TUN_DEV:
 955                 strcpy(info->bus_info, "tun");
 956                 break;
 957         case TUN_TAP_DEV:
 958                 strcpy(info->bus_info, "tap");
 959                 break;
 960         }
 961 }
 962
 963 static u32 tun_get_msglevel(struct net_device *dev)
 964 {
 965 #ifdef TUN_DEBUG
 966         struct tun_struct *tun = netdev_priv(dev);
 967         return tun->debug;
 968 #else
 969         return -EOPNOTSUPP;
 970 #endif
 971 }
 972
 973 static void tun_set_msglevel(struct net_device *dev, u32 value)
 974 {
 975 #ifdef TUN_DEBUG
 976         struct tun_struct *tun = netdev_priv(dev);
 977         tun->debug = value;
 978 #endif
 979 }
 980
 981 static u32 tun_get_link(struct net_device *dev)
 982 {
 983         struct tun_struct *tun = netdev_priv(dev);
 984         return tun->attached;
 985 }
 986
 987 static u32 tun_get_rx_csum(struct net_device *dev)
 988 {
 989         struct tun_struct *tun = netdev_priv(dev);
 990         return (tun->flags & TUN_NOCHECKSUM) == 0;
 991 }
 992
 993 static int tun_set_rx_csum(struct net_device *dev, u32 data)
 994 {
 995         struct tun_struct *tun = netdev_priv(dev);
 996         if (data)
 997                 tun->flags &= ~TUN_NOCHECKSUM;
 998         else
 999                 tun->flags |= TUN_NOCHECKSUM;
1000         return 0;
1001 }
1002
1003 static const struct ethtool_ops tun_ethtool_ops = {
1004         .get_settings   = tun_get_settings,
1005         .get_drvinfo    = tun_get_drvinfo,
1006         .get_msglevel   = tun_get_msglevel,
1007         .set_msglevel   = tun_set_msglevel,
1008         .get_link       = tun_get_link,
1009         .get_rx_csum    = tun_get_rx_csum,
1010         .set_rx_csum    = tun_set_rx_csum
1011 };
1012
1013 static int tun_init_net(struct net *net)
1014 {
1015         struct tun_net *tn;
1016
1017         tn = kmalloc(sizeof(*tn), GFP_KERNEL);
1018         if (tn == NULL)
1019                 return -ENOMEM;
1020
1021         INIT_LIST_HEAD(&tn->dev_list);
1022
1023         if (net_assign_generic(net, tun_net_id, tn)) {
1024                 kfree(tn);
1025                 return -ENOMEM;
1026         }
1027
1028         return 0;
1029 }
1030
1031 static void tun_exit_net(struct net *net)
1032 {
1033         struct tun_net *tn;
1034         struct tun_struct *tun, *nxt;
1035
1036         tn = net_generic(net, tun_net_id);
1037
1038         rtnl_lock();
1039         list_for_each_entry_safe(tun, nxt, &tn->dev_list, list) {
1040                 DBG(KERN_INFO "%s cleaned up\n", tun->dev->name);
1041                 unregister_netdevice(tun->dev);
1042         }
1043         rtnl_unlock();
1044
1045         kfree(tn);
1046 }
1047
1048 static struct pernet_operations tun_net_ops = {
1049         .init = tun_init_net,
1050         .exit = tun_exit_net,
1051 };
1052
1053 static int __init tun_init(void)
1054 {
1055         int ret = 0;
1056
1057         printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
1058         printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT);
1059
1060         ret = register_pernet_gen_device(&tun_net_id, &tun_net_ops);
1061         if (ret) {
1062                 printk(KERN_ERR "tun: Can't register pernet ops\n");
1063                 goto err_pernet;
1064         }
1065
1066         ret = misc_register(&tun_miscdev);
1067         if (ret) {
1068                 printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR);
1069                 goto err_misc;
1070         }
1071         return 0;
1072
1073 err_misc:
1074         unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1075 err_pernet:
1076         return ret;
1077 }
1078
1079 static void tun_cleanup(void)
1080 {
1081         misc_deregister(&tun_miscdev);
1082         unregister_pernet_gen_device(tun_net_id, &tun_net_ops);
1083 }
1084
1085 module_init(tun_init);
1086 module_exit(tun_cleanup);
1087 MODULE_DESCRIPTION(DRV_DESCRIPTION);
1088 MODULE_AUTHOR(DRV_COPYRIGHT);
1089 MODULE_LICENSE("GPL");
1090 MODULE_ALIAS_MISCDEV(TUN_MINOR);