| 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/poll.h> |
| 48 | #include <linux/fcntl.h> |
| 49 | #include <linux/init.h> |
| 50 | #include <linux/skbuff.h> |
| 51 | #include <linux/netdevice.h> |
| 52 | #include <linux/etherdevice.h> |
| 53 | #include <linux/miscdevice.h> |
| 54 | #include <linux/ethtool.h> |
| 55 | #include <linux/rtnetlink.h> |
| 56 | #include <linux/if.h> |
| 57 | #include <linux/if_arp.h> |
| 58 | #include <linux/if_ether.h> |
| 59 | #include <linux/if_tun.h> |
| 60 | #include <linux/crc32.h> |
| 61 | #include <linux/nsproxy.h> |
| 62 | #include <linux/virtio_net.h> |
| 63 | #include <net/net_namespace.h> |
| 64 | #include <net/netns/generic.h> |
| 65 | #include <net/rtnetlink.h> |
| 66 | #include <net/sock.h> |
| 67 | |
| 68 | #include <asm/system.h> |
| 69 | #include <asm/uaccess.h> |
| 70 | |
| 71 | /* Uncomment to enable debugging */ |
| 72 | /* #define TUN_DEBUG 1 */ |
| 73 | |
| 74 | #ifdef TUN_DEBUG |
| 75 | static int debug; |
| 76 | |
| 77 | #define DBG if(tun->debug)printk |
| 78 | #define DBG1 if(debug==2)printk |
| 79 | #else |
| 80 | #define DBG( a... ) |
| 81 | #define DBG1( a... ) |
| 82 | #endif |
| 83 | |
| 84 | #define FLT_EXACT_COUNT 8 |
| 85 | struct tap_filter { |
| 86 | unsigned int count; /* Number of addrs. Zero means disabled */ |
| 87 | u32 mask[2]; /* Mask of the hashed addrs */ |
| 88 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; |
| 89 | }; |
| 90 | |
| 91 | struct tun_file { |
| 92 | atomic_t count; |
| 93 | struct tun_struct *tun; |
| 94 | struct net *net; |
| 95 | }; |
| 96 | |
| 97 | struct tun_sock; |
| 98 | |
| 99 | struct tun_struct { |
| 100 | struct tun_file *tfile; |
| 101 | unsigned int flags; |
| 102 | uid_t owner; |
| 103 | gid_t group; |
| 104 | |
| 105 | struct net_device *dev; |
| 106 | struct fasync_struct *fasync; |
| 107 | |
| 108 | struct tap_filter txflt; |
| 109 | struct socket socket; |
| 110 | |
| 111 | #ifdef TUN_DEBUG |
| 112 | int debug; |
| 113 | #endif |
| 114 | }; |
| 115 | |
| 116 | struct tun_sock { |
| 117 | struct sock sk; |
| 118 | struct tun_struct *tun; |
| 119 | }; |
| 120 | |
| 121 | static inline struct tun_sock *tun_sk(struct sock *sk) |
| 122 | { |
| 123 | return container_of(sk, struct tun_sock, sk); |
| 124 | } |
| 125 | |
| 126 | static int tun_attach(struct tun_struct *tun, struct file *file) |
| 127 | { |
| 128 | struct tun_file *tfile = file->private_data; |
| 129 | int err; |
| 130 | |
| 131 | ASSERT_RTNL(); |
| 132 | |
| 133 | netif_tx_lock_bh(tun->dev); |
| 134 | |
| 135 | err = -EINVAL; |
| 136 | if (tfile->tun) |
| 137 | goto out; |
| 138 | |
| 139 | err = -EBUSY; |
| 140 | if (tun->tfile) |
| 141 | goto out; |
| 142 | |
| 143 | err = 0; |
| 144 | tfile->tun = tun; |
| 145 | tun->tfile = tfile; |
| 146 | dev_hold(tun->dev); |
| 147 | sock_hold(tun->socket.sk); |
| 148 | atomic_inc(&tfile->count); |
| 149 | |
| 150 | out: |
| 151 | netif_tx_unlock_bh(tun->dev); |
| 152 | return err; |
| 153 | } |
| 154 | |
| 155 | static void __tun_detach(struct tun_struct *tun) |
| 156 | { |
| 157 | /* Detach from net device */ |
| 158 | netif_tx_lock_bh(tun->dev); |
| 159 | tun->tfile = NULL; |
| 160 | netif_tx_unlock_bh(tun->dev); |
| 161 | |
| 162 | /* Drop read queue */ |
| 163 | skb_queue_purge(&tun->socket.sk->sk_receive_queue); |
| 164 | |
| 165 | /* Drop the extra count on the net device */ |
| 166 | dev_put(tun->dev); |
| 167 | } |
| 168 | |
| 169 | static void tun_detach(struct tun_struct *tun) |
| 170 | { |
| 171 | rtnl_lock(); |
| 172 | __tun_detach(tun); |
| 173 | rtnl_unlock(); |
| 174 | } |
| 175 | |
| 176 | static struct tun_struct *__tun_get(struct tun_file *tfile) |
| 177 | { |
| 178 | struct tun_struct *tun = NULL; |
| 179 | |
| 180 | if (atomic_inc_not_zero(&tfile->count)) |
| 181 | tun = tfile->tun; |
| 182 | |
| 183 | return tun; |
| 184 | } |
| 185 | |
| 186 | static struct tun_struct *tun_get(struct file *file) |
| 187 | { |
| 188 | return __tun_get(file->private_data); |
| 189 | } |
| 190 | |
| 191 | static void tun_put(struct tun_struct *tun) |
| 192 | { |
| 193 | struct tun_file *tfile = tun->tfile; |
| 194 | |
| 195 | if (atomic_dec_and_test(&tfile->count)) |
| 196 | tun_detach(tfile->tun); |
| 197 | } |
| 198 | |
| 199 | /* TAP filterting */ |
| 200 | static void addr_hash_set(u32 *mask, const u8 *addr) |
| 201 | { |
| 202 | int n = ether_crc(ETH_ALEN, addr) >> 26; |
| 203 | mask[n >> 5] |= (1 << (n & 31)); |
| 204 | } |
| 205 | |
| 206 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) |
| 207 | { |
| 208 | int n = ether_crc(ETH_ALEN, addr) >> 26; |
| 209 | return mask[n >> 5] & (1 << (n & 31)); |
| 210 | } |
| 211 | |
| 212 | static int update_filter(struct tap_filter *filter, void __user *arg) |
| 213 | { |
| 214 | struct { u8 u[ETH_ALEN]; } *addr; |
| 215 | struct tun_filter uf; |
| 216 | int err, alen, n, nexact; |
| 217 | |
| 218 | if (copy_from_user(&uf, arg, sizeof(uf))) |
| 219 | return -EFAULT; |
| 220 | |
| 221 | if (!uf.count) { |
| 222 | /* Disabled */ |
| 223 | filter->count = 0; |
| 224 | return 0; |
| 225 | } |
| 226 | |
| 227 | alen = ETH_ALEN * uf.count; |
| 228 | addr = kmalloc(alen, GFP_KERNEL); |
| 229 | if (!addr) |
| 230 | return -ENOMEM; |
| 231 | |
| 232 | if (copy_from_user(addr, arg + sizeof(uf), alen)) { |
| 233 | err = -EFAULT; |
| 234 | goto done; |
| 235 | } |
| 236 | |
| 237 | /* The filter is updated without holding any locks. Which is |
| 238 | * perfectly safe. We disable it first and in the worst |
| 239 | * case we'll accept a few undesired packets. */ |
| 240 | filter->count = 0; |
| 241 | wmb(); |
| 242 | |
| 243 | /* Use first set of addresses as an exact filter */ |
| 244 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) |
| 245 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); |
| 246 | |
| 247 | nexact = n; |
| 248 | |
| 249 | /* Remaining multicast addresses are hashed, |
| 250 | * unicast will leave the filter disabled. */ |
| 251 | memset(filter->mask, 0, sizeof(filter->mask)); |
| 252 | for (; n < uf.count; n++) { |
| 253 | if (!is_multicast_ether_addr(addr[n].u)) { |
| 254 | err = 0; /* no filter */ |
| 255 | goto done; |
| 256 | } |
| 257 | addr_hash_set(filter->mask, addr[n].u); |
| 258 | } |
| 259 | |
| 260 | /* For ALLMULTI just set the mask to all ones. |
| 261 | * This overrides the mask populated above. */ |
| 262 | if ((uf.flags & TUN_FLT_ALLMULTI)) |
| 263 | memset(filter->mask, ~0, sizeof(filter->mask)); |
| 264 | |
| 265 | /* Now enable the filter */ |
| 266 | wmb(); |
| 267 | filter->count = nexact; |
| 268 | |
| 269 | /* Return the number of exact filters */ |
| 270 | err = nexact; |
| 271 | |
| 272 | done: |
| 273 | kfree(addr); |
| 274 | return err; |
| 275 | } |
| 276 | |
| 277 | /* Returns: 0 - drop, !=0 - accept */ |
| 278 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) |
| 279 | { |
| 280 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect |
| 281 | * at this point. */ |
| 282 | struct ethhdr *eh = (struct ethhdr *) skb->data; |
| 283 | int i; |
| 284 | |
| 285 | /* Exact match */ |
| 286 | for (i = 0; i < filter->count; i++) |
| 287 | if (!compare_ether_addr(eh->h_dest, filter->addr[i])) |
| 288 | return 1; |
| 289 | |
| 290 | /* Inexact match (multicast only) */ |
| 291 | if (is_multicast_ether_addr(eh->h_dest)) |
| 292 | return addr_hash_test(filter->mask, eh->h_dest); |
| 293 | |
| 294 | return 0; |
| 295 | } |
| 296 | |
| 297 | /* |
| 298 | * Checks whether the packet is accepted or not. |
| 299 | * Returns: 0 - drop, !=0 - accept |
| 300 | */ |
| 301 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) |
| 302 | { |
| 303 | if (!filter->count) |
| 304 | return 1; |
| 305 | |
| 306 | return run_filter(filter, skb); |
| 307 | } |
| 308 | |
| 309 | /* Network device part of the driver */ |
| 310 | |
| 311 | static const struct ethtool_ops tun_ethtool_ops; |
| 312 | |
| 313 | /* Net device detach from fd. */ |
| 314 | static void tun_net_uninit(struct net_device *dev) |
| 315 | { |
| 316 | struct tun_struct *tun = netdev_priv(dev); |
| 317 | struct tun_file *tfile = tun->tfile; |
| 318 | |
| 319 | /* Inform the methods they need to stop using the dev. |
| 320 | */ |
| 321 | if (tfile) { |
| 322 | wake_up_all(&tun->socket.wait); |
| 323 | if (atomic_dec_and_test(&tfile->count)) |
| 324 | __tun_detach(tun); |
| 325 | } |
| 326 | } |
| 327 | |
| 328 | static void tun_free_netdev(struct net_device *dev) |
| 329 | { |
| 330 | struct tun_struct *tun = netdev_priv(dev); |
| 331 | |
| 332 | sock_put(tun->socket.sk); |
| 333 | } |
| 334 | |
| 335 | /* Net device open. */ |
| 336 | static int tun_net_open(struct net_device *dev) |
| 337 | { |
| 338 | netif_start_queue(dev); |
| 339 | return 0; |
| 340 | } |
| 341 | |
| 342 | /* Net device close. */ |
| 343 | static int tun_net_close(struct net_device *dev) |
| 344 | { |
| 345 | netif_stop_queue(dev); |
| 346 | return 0; |
| 347 | } |
| 348 | |
| 349 | /* Net device start xmit */ |
| 350 | static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) |
| 351 | { |
| 352 | struct tun_struct *tun = netdev_priv(dev); |
| 353 | |
| 354 | DBG(KERN_INFO "%s: tun_net_xmit %d\n", tun->dev->name, skb->len); |
| 355 | |
| 356 | /* Drop packet if interface is not attached */ |
| 357 | if (!tun->tfile) |
| 358 | goto drop; |
| 359 | |
| 360 | /* Drop if the filter does not like it. |
| 361 | * This is a noop if the filter is disabled. |
| 362 | * Filter can be enabled only for the TAP devices. */ |
| 363 | if (!check_filter(&tun->txflt, skb)) |
| 364 | goto drop; |
| 365 | |
| 366 | if (skb_queue_len(&tun->socket.sk->sk_receive_queue) >= dev->tx_queue_len) { |
| 367 | if (!(tun->flags & TUN_ONE_QUEUE)) { |
| 368 | /* Normal queueing mode. */ |
| 369 | /* Packet scheduler handles dropping of further packets. */ |
| 370 | netif_stop_queue(dev); |
| 371 | |
| 372 | /* We won't see all dropped packets individually, so overrun |
| 373 | * error is more appropriate. */ |
| 374 | dev->stats.tx_fifo_errors++; |
| 375 | } else { |
| 376 | /* Single queue mode. |
| 377 | * Driver handles dropping of all packets itself. */ |
| 378 | goto drop; |
| 379 | } |
| 380 | } |
| 381 | |
| 382 | /* Enqueue packet */ |
| 383 | skb_queue_tail(&tun->socket.sk->sk_receive_queue, skb); |
| 384 | dev->trans_start = jiffies; |
| 385 | |
| 386 | /* Notify and wake up reader process */ |
| 387 | if (tun->flags & TUN_FASYNC) |
| 388 | kill_fasync(&tun->fasync, SIGIO, POLL_IN); |
| 389 | wake_up_interruptible(&tun->socket.wait); |
| 390 | return NETDEV_TX_OK; |
| 391 | |
| 392 | drop: |
| 393 | dev->stats.tx_dropped++; |
| 394 | kfree_skb(skb); |
| 395 | return NETDEV_TX_OK; |
| 396 | } |
| 397 | |
| 398 | static void tun_net_mclist(struct net_device *dev) |
| 399 | { |
| 400 | /* |
| 401 | * This callback is supposed to deal with mc filter in |
| 402 | * _rx_ path and has nothing to do with the _tx_ path. |
| 403 | * In rx path we always accept everything userspace gives us. |
| 404 | */ |
| 405 | return; |
| 406 | } |
| 407 | |
| 408 | #define MIN_MTU 68 |
| 409 | #define MAX_MTU 65535 |
| 410 | |
| 411 | static int |
| 412 | tun_net_change_mtu(struct net_device *dev, int new_mtu) |
| 413 | { |
| 414 | if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU) |
| 415 | return -EINVAL; |
| 416 | dev->mtu = new_mtu; |
| 417 | return 0; |
| 418 | } |
| 419 | |
| 420 | static const struct net_device_ops tun_netdev_ops = { |
| 421 | .ndo_uninit = tun_net_uninit, |
| 422 | .ndo_open = tun_net_open, |
| 423 | .ndo_stop = tun_net_close, |
| 424 | .ndo_start_xmit = tun_net_xmit, |
| 425 | .ndo_change_mtu = tun_net_change_mtu, |
| 426 | }; |
| 427 | |
| 428 | static const struct net_device_ops tap_netdev_ops = { |
| 429 | .ndo_uninit = tun_net_uninit, |
| 430 | .ndo_open = tun_net_open, |
| 431 | .ndo_stop = tun_net_close, |
| 432 | .ndo_start_xmit = tun_net_xmit, |
| 433 | .ndo_change_mtu = tun_net_change_mtu, |
| 434 | .ndo_set_multicast_list = tun_net_mclist, |
| 435 | .ndo_set_mac_address = eth_mac_addr, |
| 436 | .ndo_validate_addr = eth_validate_addr, |
| 437 | }; |
| 438 | |
| 439 | /* Initialize net device. */ |
| 440 | static void tun_net_init(struct net_device *dev) |
| 441 | { |
| 442 | struct tun_struct *tun = netdev_priv(dev); |
| 443 | |
| 444 | switch (tun->flags & TUN_TYPE_MASK) { |
| 445 | case TUN_TUN_DEV: |
| 446 | dev->netdev_ops = &tun_netdev_ops; |
| 447 | |
| 448 | /* Point-to-Point TUN Device */ |
| 449 | dev->hard_header_len = 0; |
| 450 | dev->addr_len = 0; |
| 451 | dev->mtu = 1500; |
| 452 | |
| 453 | /* Zero header length */ |
| 454 | dev->type = ARPHRD_NONE; |
| 455 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
| 456 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ |
| 457 | break; |
| 458 | |
| 459 | case TUN_TAP_DEV: |
| 460 | dev->netdev_ops = &tap_netdev_ops; |
| 461 | /* Ethernet TAP Device */ |
| 462 | ether_setup(dev); |
| 463 | |
| 464 | random_ether_addr(dev->dev_addr); |
| 465 | |
| 466 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ |
| 467 | break; |
| 468 | } |
| 469 | } |
| 470 | |
| 471 | /* Character device part */ |
| 472 | |
| 473 | /* Poll */ |
| 474 | static unsigned int tun_chr_poll(struct file *file, poll_table * wait) |
| 475 | { |
| 476 | struct tun_file *tfile = file->private_data; |
| 477 | struct tun_struct *tun = __tun_get(tfile); |
| 478 | struct sock *sk; |
| 479 | unsigned int mask = 0; |
| 480 | |
| 481 | if (!tun) |
| 482 | return POLLERR; |
| 483 | |
| 484 | sk = tun->socket.sk; |
| 485 | |
| 486 | DBG(KERN_INFO "%s: tun_chr_poll\n", tun->dev->name); |
| 487 | |
| 488 | poll_wait(file, &tun->socket.wait, wait); |
| 489 | |
| 490 | if (!skb_queue_empty(&sk->sk_receive_queue)) |
| 491 | mask |= POLLIN | POLLRDNORM; |
| 492 | |
| 493 | if (sock_writeable(sk) || |
| 494 | (!test_and_set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags) && |
| 495 | sock_writeable(sk))) |
| 496 | mask |= POLLOUT | POLLWRNORM; |
| 497 | |
| 498 | if (tun->dev->reg_state != NETREG_REGISTERED) |
| 499 | mask = POLLERR; |
| 500 | |
| 501 | tun_put(tun); |
| 502 | return mask; |
| 503 | } |
| 504 | |
| 505 | /* prepad is the amount to reserve at front. len is length after that. |
| 506 | * linear is a hint as to how much to copy (usually headers). */ |
| 507 | static inline struct sk_buff *tun_alloc_skb(struct tun_struct *tun, |
| 508 | size_t prepad, size_t len, |
| 509 | size_t linear, int noblock) |
| 510 | { |
| 511 | struct sock *sk = tun->socket.sk; |
| 512 | struct sk_buff *skb; |
| 513 | int err; |
| 514 | |
| 515 | /* Under a page? Don't bother with paged skb. */ |
| 516 | if (prepad + len < PAGE_SIZE || !linear) |
| 517 | linear = len; |
| 518 | |
| 519 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
| 520 | &err); |
| 521 | if (!skb) |
| 522 | return ERR_PTR(err); |
| 523 | |
| 524 | skb_reserve(skb, prepad); |
| 525 | skb_put(skb, linear); |
| 526 | skb->data_len = len - linear; |
| 527 | skb->len += len - linear; |
| 528 | |
| 529 | return skb; |
| 530 | } |
| 531 | |
| 532 | /* Get packet from user space buffer */ |
| 533 | static __inline__ ssize_t tun_get_user(struct tun_struct *tun, |
| 534 | const struct iovec *iv, size_t count, |
| 535 | int noblock) |
| 536 | { |
| 537 | struct tun_pi pi = { 0, cpu_to_be16(ETH_P_IP) }; |
| 538 | struct sk_buff *skb; |
| 539 | size_t len = count, align = 0; |
| 540 | struct virtio_net_hdr gso = { 0 }; |
| 541 | int offset = 0; |
| 542 | |
| 543 | if (!(tun->flags & TUN_NO_PI)) { |
| 544 | if ((len -= sizeof(pi)) > count) |
| 545 | return -EINVAL; |
| 546 | |
| 547 | if (memcpy_fromiovecend((void *)&pi, iv, 0, sizeof(pi))) |
| 548 | return -EFAULT; |
| 549 | offset += sizeof(pi); |
| 550 | } |
| 551 | |
| 552 | if (tun->flags & TUN_VNET_HDR) { |
| 553 | if ((len -= sizeof(gso)) > count) |
| 554 | return -EINVAL; |
| 555 | |
| 556 | if (memcpy_fromiovecend((void *)&gso, iv, offset, sizeof(gso))) |
| 557 | return -EFAULT; |
| 558 | |
| 559 | if ((gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
| 560 | gso.csum_start + gso.csum_offset + 2 > gso.hdr_len) |
| 561 | gso.hdr_len = gso.csum_start + gso.csum_offset + 2; |
| 562 | |
| 563 | if (gso.hdr_len > len) |
| 564 | return -EINVAL; |
| 565 | offset += sizeof(gso); |
| 566 | } |
| 567 | |
| 568 | if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) { |
| 569 | align = NET_IP_ALIGN; |
| 570 | if (unlikely(len < ETH_HLEN || |
| 571 | (gso.hdr_len && gso.hdr_len < ETH_HLEN))) |
| 572 | return -EINVAL; |
| 573 | } |
| 574 | |
| 575 | skb = tun_alloc_skb(tun, align, len, gso.hdr_len, noblock); |
| 576 | if (IS_ERR(skb)) { |
| 577 | if (PTR_ERR(skb) != -EAGAIN) |
| 578 | tun->dev->stats.rx_dropped++; |
| 579 | return PTR_ERR(skb); |
| 580 | } |
| 581 | |
| 582 | if (skb_copy_datagram_from_iovec(skb, 0, iv, offset, len)) { |
| 583 | tun->dev->stats.rx_dropped++; |
| 584 | kfree_skb(skb); |
| 585 | return -EFAULT; |
| 586 | } |
| 587 | |
| 588 | if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { |
| 589 | if (!skb_partial_csum_set(skb, gso.csum_start, |
| 590 | gso.csum_offset)) { |
| 591 | tun->dev->stats.rx_frame_errors++; |
| 592 | kfree_skb(skb); |
| 593 | return -EINVAL; |
| 594 | } |
| 595 | } else if (tun->flags & TUN_NOCHECKSUM) |
| 596 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
| 597 | |
| 598 | switch (tun->flags & TUN_TYPE_MASK) { |
| 599 | case TUN_TUN_DEV: |
| 600 | if (tun->flags & TUN_NO_PI) { |
| 601 | switch (skb->data[0] & 0xf0) { |
| 602 | case 0x40: |
| 603 | pi.proto = htons(ETH_P_IP); |
| 604 | break; |
| 605 | case 0x60: |
| 606 | pi.proto = htons(ETH_P_IPV6); |
| 607 | break; |
| 608 | default: |
| 609 | tun->dev->stats.rx_dropped++; |
| 610 | kfree_skb(skb); |
| 611 | return -EINVAL; |
| 612 | } |
| 613 | } |
| 614 | |
| 615 | skb_reset_mac_header(skb); |
| 616 | skb->protocol = pi.proto; |
| 617 | skb->dev = tun->dev; |
| 618 | break; |
| 619 | case TUN_TAP_DEV: |
| 620 | skb->protocol = eth_type_trans(skb, tun->dev); |
| 621 | break; |
| 622 | }; |
| 623 | |
| 624 | if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) { |
| 625 | pr_debug("GSO!\n"); |
| 626 | switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { |
| 627 | case VIRTIO_NET_HDR_GSO_TCPV4: |
| 628 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; |
| 629 | break; |
| 630 | case VIRTIO_NET_HDR_GSO_TCPV6: |
| 631 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; |
| 632 | break; |
| 633 | case VIRTIO_NET_HDR_GSO_UDP: |
| 634 | skb_shinfo(skb)->gso_type = SKB_GSO_UDP; |
| 635 | break; |
| 636 | default: |
| 637 | tun->dev->stats.rx_frame_errors++; |
| 638 | kfree_skb(skb); |
| 639 | return -EINVAL; |
| 640 | } |
| 641 | |
| 642 | if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN) |
| 643 | skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; |
| 644 | |
| 645 | skb_shinfo(skb)->gso_size = gso.gso_size; |
| 646 | if (skb_shinfo(skb)->gso_size == 0) { |
| 647 | tun->dev->stats.rx_frame_errors++; |
| 648 | kfree_skb(skb); |
| 649 | return -EINVAL; |
| 650 | } |
| 651 | |
| 652 | /* Header must be checked, and gso_segs computed. */ |
| 653 | skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY; |
| 654 | skb_shinfo(skb)->gso_segs = 0; |
| 655 | } |
| 656 | |
| 657 | netif_rx_ni(skb); |
| 658 | |
| 659 | tun->dev->stats.rx_packets++; |
| 660 | tun->dev->stats.rx_bytes += len; |
| 661 | |
| 662 | return count; |
| 663 | } |
| 664 | |
| 665 | static ssize_t tun_chr_aio_write(struct kiocb *iocb, const struct iovec *iv, |
| 666 | unsigned long count, loff_t pos) |
| 667 | { |
| 668 | struct file *file = iocb->ki_filp; |
| 669 | struct tun_struct *tun = tun_get(file); |
| 670 | ssize_t result; |
| 671 | |
| 672 | if (!tun) |
| 673 | return -EBADFD; |
| 674 | |
| 675 | DBG(KERN_INFO "%s: tun_chr_write %ld\n", tun->dev->name, count); |
| 676 | |
| 677 | result = tun_get_user(tun, iv, iov_length(iv, count), |
| 678 | file->f_flags & O_NONBLOCK); |
| 679 | |
| 680 | tun_put(tun); |
| 681 | return result; |
| 682 | } |
| 683 | |
| 684 | /* Put packet to the user space buffer */ |
| 685 | static __inline__ ssize_t tun_put_user(struct tun_struct *tun, |
| 686 | struct sk_buff *skb, |
| 687 | const struct iovec *iv, int len) |
| 688 | { |
| 689 | struct tun_pi pi = { 0, skb->protocol }; |
| 690 | ssize_t total = 0; |
| 691 | |
| 692 | if (!(tun->flags & TUN_NO_PI)) { |
| 693 | if ((len -= sizeof(pi)) < 0) |
| 694 | return -EINVAL; |
| 695 | |
| 696 | if (len < skb->len) { |
| 697 | /* Packet will be striped */ |
| 698 | pi.flags |= TUN_PKT_STRIP; |
| 699 | } |
| 700 | |
| 701 | if (memcpy_toiovecend(iv, (void *) &pi, 0, sizeof(pi))) |
| 702 | return -EFAULT; |
| 703 | total += sizeof(pi); |
| 704 | } |
| 705 | |
| 706 | if (tun->flags & TUN_VNET_HDR) { |
| 707 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ |
| 708 | if ((len -= sizeof(gso)) < 0) |
| 709 | return -EINVAL; |
| 710 | |
| 711 | if (skb_is_gso(skb)) { |
| 712 | struct skb_shared_info *sinfo = skb_shinfo(skb); |
| 713 | |
| 714 | /* This is a hint as to how much should be linear. */ |
| 715 | gso.hdr_len = skb_headlen(skb); |
| 716 | gso.gso_size = sinfo->gso_size; |
| 717 | if (sinfo->gso_type & SKB_GSO_TCPV4) |
| 718 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4; |
| 719 | else if (sinfo->gso_type & SKB_GSO_TCPV6) |
| 720 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6; |
| 721 | else if (sinfo->gso_type & SKB_GSO_UDP) |
| 722 | gso.gso_type = VIRTIO_NET_HDR_GSO_UDP; |
| 723 | else |
| 724 | BUG(); |
| 725 | if (sinfo->gso_type & SKB_GSO_TCP_ECN) |
| 726 | gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN; |
| 727 | } else |
| 728 | gso.gso_type = VIRTIO_NET_HDR_GSO_NONE; |
| 729 | |
| 730 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
| 731 | gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; |
| 732 | gso.csum_start = skb->csum_start - skb_headroom(skb); |
| 733 | gso.csum_offset = skb->csum_offset; |
| 734 | } /* else everything is zero */ |
| 735 | |
| 736 | if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total, |
| 737 | sizeof(gso)))) |
| 738 | return -EFAULT; |
| 739 | total += sizeof(gso); |
| 740 | } |
| 741 | |
| 742 | len = min_t(int, skb->len, len); |
| 743 | |
| 744 | skb_copy_datagram_const_iovec(skb, 0, iv, total, len); |
| 745 | total += len; |
| 746 | |
| 747 | tun->dev->stats.tx_packets++; |
| 748 | tun->dev->stats.tx_bytes += len; |
| 749 | |
| 750 | return total; |
| 751 | } |
| 752 | |
| 753 | static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv, |
| 754 | unsigned long count, loff_t pos) |
| 755 | { |
| 756 | struct file *file = iocb->ki_filp; |
| 757 | struct tun_file *tfile = file->private_data; |
| 758 | struct tun_struct *tun = __tun_get(tfile); |
| 759 | DECLARE_WAITQUEUE(wait, current); |
| 760 | struct sk_buff *skb; |
| 761 | ssize_t len, ret = 0; |
| 762 | |
| 763 | if (!tun) |
| 764 | return -EBADFD; |
| 765 | |
| 766 | DBG(KERN_INFO "%s: tun_chr_read\n", tun->dev->name); |
| 767 | |
| 768 | len = iov_length(iv, count); |
| 769 | if (len < 0) { |
| 770 | ret = -EINVAL; |
| 771 | goto out; |
| 772 | } |
| 773 | |
| 774 | add_wait_queue(&tun->socket.wait, &wait); |
| 775 | while (len) { |
| 776 | current->state = TASK_INTERRUPTIBLE; |
| 777 | |
| 778 | /* Read frames from the queue */ |
| 779 | if (!(skb=skb_dequeue(&tun->socket.sk->sk_receive_queue))) { |
| 780 | if (file->f_flags & O_NONBLOCK) { |
| 781 | ret = -EAGAIN; |
| 782 | break; |
| 783 | } |
| 784 | if (signal_pending(current)) { |
| 785 | ret = -ERESTARTSYS; |
| 786 | break; |
| 787 | } |
| 788 | if (tun->dev->reg_state != NETREG_REGISTERED) { |
| 789 | ret = -EIO; |
| 790 | break; |
| 791 | } |
| 792 | |
| 793 | /* Nothing to read, let's sleep */ |
| 794 | schedule(); |
| 795 | continue; |
| 796 | } |
| 797 | netif_wake_queue(tun->dev); |
| 798 | |
| 799 | ret = tun_put_user(tun, skb, iv, len); |
| 800 | kfree_skb(skb); |
| 801 | break; |
| 802 | } |
| 803 | |
| 804 | current->state = TASK_RUNNING; |
| 805 | remove_wait_queue(&tun->socket.wait, &wait); |
| 806 | |
| 807 | out: |
| 808 | tun_put(tun); |
| 809 | return ret; |
| 810 | } |
| 811 | |
| 812 | static void tun_setup(struct net_device *dev) |
| 813 | { |
| 814 | struct tun_struct *tun = netdev_priv(dev); |
| 815 | |
| 816 | tun->owner = -1; |
| 817 | tun->group = -1; |
| 818 | |
| 819 | dev->ethtool_ops = &tun_ethtool_ops; |
| 820 | dev->destructor = tun_free_netdev; |
| 821 | } |
| 822 | |
| 823 | /* Trivial set of netlink ops to allow deleting tun or tap |
| 824 | * device with netlink. |
| 825 | */ |
| 826 | static int tun_validate(struct nlattr *tb[], struct nlattr *data[]) |
| 827 | { |
| 828 | return -EINVAL; |
| 829 | } |
| 830 | |
| 831 | static struct rtnl_link_ops tun_link_ops __read_mostly = { |
| 832 | .kind = DRV_NAME, |
| 833 | .priv_size = sizeof(struct tun_struct), |
| 834 | .setup = tun_setup, |
| 835 | .validate = tun_validate, |
| 836 | }; |
| 837 | |
| 838 | static void tun_sock_write_space(struct sock *sk) |
| 839 | { |
| 840 | struct tun_struct *tun; |
| 841 | |
| 842 | if (!sock_writeable(sk)) |
| 843 | return; |
| 844 | |
| 845 | if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags)) |
| 846 | return; |
| 847 | |
| 848 | if (sk->sk_sleep && waitqueue_active(sk->sk_sleep)) |
| 849 | wake_up_interruptible_sync(sk->sk_sleep); |
| 850 | |
| 851 | tun = container_of(sk, struct tun_sock, sk)->tun; |
| 852 | kill_fasync(&tun->fasync, SIGIO, POLL_OUT); |
| 853 | } |
| 854 | |
| 855 | static void tun_sock_destruct(struct sock *sk) |
| 856 | { |
| 857 | free_netdev(container_of(sk, struct tun_sock, sk)->tun->dev); |
| 858 | } |
| 859 | |
| 860 | static struct proto tun_proto = { |
| 861 | .name = "tun", |
| 862 | .owner = THIS_MODULE, |
| 863 | .obj_size = sizeof(struct tun_sock), |
| 864 | }; |
| 865 | |
| 866 | static int tun_flags(struct tun_struct *tun) |
| 867 | { |
| 868 | int flags = 0; |
| 869 | |
| 870 | if (tun->flags & TUN_TUN_DEV) |
| 871 | flags |= IFF_TUN; |
| 872 | else |
| 873 | flags |= IFF_TAP; |
| 874 | |
| 875 | if (tun->flags & TUN_NO_PI) |
| 876 | flags |= IFF_NO_PI; |
| 877 | |
| 878 | if (tun->flags & TUN_ONE_QUEUE) |
| 879 | flags |= IFF_ONE_QUEUE; |
| 880 | |
| 881 | if (tun->flags & TUN_VNET_HDR) |
| 882 | flags |= IFF_VNET_HDR; |
| 883 | |
| 884 | return flags; |
| 885 | } |
| 886 | |
| 887 | static ssize_t tun_show_flags(struct device *dev, struct device_attribute *attr, |
| 888 | char *buf) |
| 889 | { |
| 890 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); |
| 891 | return sprintf(buf, "0x%x\n", tun_flags(tun)); |
| 892 | } |
| 893 | |
| 894 | static ssize_t tun_show_owner(struct device *dev, struct device_attribute *attr, |
| 895 | char *buf) |
| 896 | { |
| 897 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); |
| 898 | return sprintf(buf, "%d\n", tun->owner); |
| 899 | } |
| 900 | |
| 901 | static ssize_t tun_show_group(struct device *dev, struct device_attribute *attr, |
| 902 | char *buf) |
| 903 | { |
| 904 | struct tun_struct *tun = netdev_priv(to_net_dev(dev)); |
| 905 | return sprintf(buf, "%d\n", tun->group); |
| 906 | } |
| 907 | |
| 908 | static DEVICE_ATTR(tun_flags, 0444, tun_show_flags, NULL); |
| 909 | static DEVICE_ATTR(owner, 0444, tun_show_owner, NULL); |
| 910 | static DEVICE_ATTR(group, 0444, tun_show_group, NULL); |
| 911 | |
| 912 | static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) |
| 913 | { |
| 914 | struct sock *sk; |
| 915 | struct tun_struct *tun; |
| 916 | struct net_device *dev; |
| 917 | int err; |
| 918 | |
| 919 | dev = __dev_get_by_name(net, ifr->ifr_name); |
| 920 | if (dev) { |
| 921 | const struct cred *cred = current_cred(); |
| 922 | |
| 923 | if (ifr->ifr_flags & IFF_TUN_EXCL) |
| 924 | return -EBUSY; |
| 925 | if ((ifr->ifr_flags & IFF_TUN) && dev->netdev_ops == &tun_netdev_ops) |
| 926 | tun = netdev_priv(dev); |
| 927 | else if ((ifr->ifr_flags & IFF_TAP) && dev->netdev_ops == &tap_netdev_ops) |
| 928 | tun = netdev_priv(dev); |
| 929 | else |
| 930 | return -EINVAL; |
| 931 | |
| 932 | if (((tun->owner != -1 && cred->euid != tun->owner) || |
| 933 | (tun->group != -1 && !in_egroup_p(tun->group))) && |
| 934 | !capable(CAP_NET_ADMIN)) |
| 935 | return -EPERM; |
| 936 | err = security_tun_dev_attach(tun->socket.sk); |
| 937 | if (err < 0) |
| 938 | return err; |
| 939 | |
| 940 | err = tun_attach(tun, file); |
| 941 | if (err < 0) |
| 942 | return err; |
| 943 | } |
| 944 | else { |
| 945 | char *name; |
| 946 | unsigned long flags = 0; |
| 947 | |
| 948 | if (!capable(CAP_NET_ADMIN)) |
| 949 | return -EPERM; |
| 950 | err = security_tun_dev_create(); |
| 951 | if (err < 0) |
| 952 | return err; |
| 953 | |
| 954 | /* Set dev type */ |
| 955 | if (ifr->ifr_flags & IFF_TUN) { |
| 956 | /* TUN device */ |
| 957 | flags |= TUN_TUN_DEV; |
| 958 | name = "tun%d"; |
| 959 | } else if (ifr->ifr_flags & IFF_TAP) { |
| 960 | /* TAP device */ |
| 961 | flags |= TUN_TAP_DEV; |
| 962 | name = "tap%d"; |
| 963 | } else |
| 964 | return -EINVAL; |
| 965 | |
| 966 | if (*ifr->ifr_name) |
| 967 | name = ifr->ifr_name; |
| 968 | |
| 969 | dev = alloc_netdev(sizeof(struct tun_struct), name, |
| 970 | tun_setup); |
| 971 | if (!dev) |
| 972 | return -ENOMEM; |
| 973 | |
| 974 | dev_net_set(dev, net); |
| 975 | dev->rtnl_link_ops = &tun_link_ops; |
| 976 | |
| 977 | tun = netdev_priv(dev); |
| 978 | tun->dev = dev; |
| 979 | tun->flags = flags; |
| 980 | tun->txflt.count = 0; |
| 981 | |
| 982 | err = -ENOMEM; |
| 983 | sk = sk_alloc(net, AF_UNSPEC, GFP_KERNEL, &tun_proto); |
| 984 | if (!sk) |
| 985 | goto err_free_dev; |
| 986 | |
| 987 | init_waitqueue_head(&tun->socket.wait); |
| 988 | sock_init_data(&tun->socket, sk); |
| 989 | sk->sk_write_space = tun_sock_write_space; |
| 990 | sk->sk_sndbuf = INT_MAX; |
| 991 | |
| 992 | container_of(sk, struct tun_sock, sk)->tun = tun; |
| 993 | |
| 994 | security_tun_dev_post_create(sk); |
| 995 | |
| 996 | tun_net_init(dev); |
| 997 | |
| 998 | if (strchr(dev->name, '%')) { |
| 999 | err = dev_alloc_name(dev, dev->name); |
| 1000 | if (err < 0) |
| 1001 | goto err_free_sk; |
| 1002 | } |
| 1003 | |
| 1004 | err = register_netdevice(tun->dev); |
| 1005 | if (err < 0) |
| 1006 | goto err_free_sk; |
| 1007 | |
| 1008 | if (device_create_file(&tun->dev->dev, &dev_attr_tun_flags) || |
| 1009 | device_create_file(&tun->dev->dev, &dev_attr_owner) || |
| 1010 | device_create_file(&tun->dev->dev, &dev_attr_group)) |
| 1011 | printk(KERN_ERR "Failed to create tun sysfs files\n"); |
| 1012 | |
| 1013 | sk->sk_destruct = tun_sock_destruct; |
| 1014 | |
| 1015 | err = tun_attach(tun, file); |
| 1016 | if (err < 0) |
| 1017 | goto failed; |
| 1018 | } |
| 1019 | |
| 1020 | DBG(KERN_INFO "%s: tun_set_iff\n", tun->dev->name); |
| 1021 | |
| 1022 | if (ifr->ifr_flags & IFF_NO_PI) |
| 1023 | tun->flags |= TUN_NO_PI; |
| 1024 | else |
| 1025 | tun->flags &= ~TUN_NO_PI; |
| 1026 | |
| 1027 | if (ifr->ifr_flags & IFF_ONE_QUEUE) |
| 1028 | tun->flags |= TUN_ONE_QUEUE; |
| 1029 | else |
| 1030 | tun->flags &= ~TUN_ONE_QUEUE; |
| 1031 | |
| 1032 | if (ifr->ifr_flags & IFF_VNET_HDR) |
| 1033 | tun->flags |= TUN_VNET_HDR; |
| 1034 | else |
| 1035 | tun->flags &= ~TUN_VNET_HDR; |
| 1036 | |
| 1037 | /* Make sure persistent devices do not get stuck in |
| 1038 | * xoff state. |
| 1039 | */ |
| 1040 | if (netif_running(tun->dev)) |
| 1041 | netif_wake_queue(tun->dev); |
| 1042 | |
| 1043 | strcpy(ifr->ifr_name, tun->dev->name); |
| 1044 | return 0; |
| 1045 | |
| 1046 | err_free_sk: |
| 1047 | sock_put(sk); |
| 1048 | err_free_dev: |
| 1049 | free_netdev(dev); |
| 1050 | failed: |
| 1051 | return err; |
| 1052 | } |
| 1053 | |
| 1054 | static int tun_get_iff(struct net *net, struct tun_struct *tun, |
| 1055 | struct ifreq *ifr) |
| 1056 | { |
| 1057 | DBG(KERN_INFO "%s: tun_get_iff\n", tun->dev->name); |
| 1058 | |
| 1059 | strcpy(ifr->ifr_name, tun->dev->name); |
| 1060 | |
| 1061 | ifr->ifr_flags = tun_flags(tun); |
| 1062 | |
| 1063 | return 0; |
| 1064 | } |
| 1065 | |
| 1066 | /* This is like a cut-down ethtool ops, except done via tun fd so no |
| 1067 | * privs required. */ |
| 1068 | static int set_offload(struct net_device *dev, unsigned long arg) |
| 1069 | { |
| 1070 | unsigned int old_features, features; |
| 1071 | |
| 1072 | old_features = dev->features; |
| 1073 | /* Unset features, set them as we chew on the arg. */ |
| 1074 | features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST |
| 1075 | |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6 |
| 1076 | |NETIF_F_UFO)); |
| 1077 | |
| 1078 | if (arg & TUN_F_CSUM) { |
| 1079 | features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST; |
| 1080 | arg &= ~TUN_F_CSUM; |
| 1081 | |
| 1082 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { |
| 1083 | if (arg & TUN_F_TSO_ECN) { |
| 1084 | features |= NETIF_F_TSO_ECN; |
| 1085 | arg &= ~TUN_F_TSO_ECN; |
| 1086 | } |
| 1087 | if (arg & TUN_F_TSO4) |
| 1088 | features |= NETIF_F_TSO; |
| 1089 | if (arg & TUN_F_TSO6) |
| 1090 | features |= NETIF_F_TSO6; |
| 1091 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); |
| 1092 | } |
| 1093 | |
| 1094 | if (arg & TUN_F_UFO) { |
| 1095 | features |= NETIF_F_UFO; |
| 1096 | arg &= ~TUN_F_UFO; |
| 1097 | } |
| 1098 | } |
| 1099 | |
| 1100 | /* This gives the user a way to test for new features in future by |
| 1101 | * trying to set them. */ |
| 1102 | if (arg) |
| 1103 | return -EINVAL; |
| 1104 | |
| 1105 | dev->features = features; |
| 1106 | if (old_features != dev->features) |
| 1107 | netdev_features_change(dev); |
| 1108 | |
| 1109 | return 0; |
| 1110 | } |
| 1111 | |
| 1112 | static long tun_chr_ioctl(struct file *file, unsigned int cmd, |
| 1113 | unsigned long arg) |
| 1114 | { |
| 1115 | struct tun_file *tfile = file->private_data; |
| 1116 | struct tun_struct *tun; |
| 1117 | void __user* argp = (void __user*)arg; |
| 1118 | struct ifreq ifr; |
| 1119 | int sndbuf; |
| 1120 | int ret; |
| 1121 | |
| 1122 | if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89) |
| 1123 | if (copy_from_user(&ifr, argp, sizeof ifr)) |
| 1124 | return -EFAULT; |
| 1125 | |
| 1126 | if (cmd == TUNGETFEATURES) { |
| 1127 | /* Currently this just means: "what IFF flags are valid?". |
| 1128 | * This is needed because we never checked for invalid flags on |
| 1129 | * TUNSETIFF. */ |
| 1130 | return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE | |
| 1131 | IFF_VNET_HDR, |
| 1132 | (unsigned int __user*)argp); |
| 1133 | } |
| 1134 | |
| 1135 | rtnl_lock(); |
| 1136 | |
| 1137 | tun = __tun_get(tfile); |
| 1138 | if (cmd == TUNSETIFF && !tun) { |
| 1139 | ifr.ifr_name[IFNAMSIZ-1] = '\0'; |
| 1140 | |
| 1141 | ret = tun_set_iff(tfile->net, file, &ifr); |
| 1142 | |
| 1143 | if (ret) |
| 1144 | goto unlock; |
| 1145 | |
| 1146 | if (copy_to_user(argp, &ifr, sizeof(ifr))) |
| 1147 | ret = -EFAULT; |
| 1148 | goto unlock; |
| 1149 | } |
| 1150 | |
| 1151 | ret = -EBADFD; |
| 1152 | if (!tun) |
| 1153 | goto unlock; |
| 1154 | |
| 1155 | DBG(KERN_INFO "%s: tun_chr_ioctl cmd %d\n", tun->dev->name, cmd); |
| 1156 | |
| 1157 | ret = 0; |
| 1158 | switch (cmd) { |
| 1159 | case TUNGETIFF: |
| 1160 | ret = tun_get_iff(current->nsproxy->net_ns, tun, &ifr); |
| 1161 | if (ret) |
| 1162 | break; |
| 1163 | |
| 1164 | if (copy_to_user(argp, &ifr, sizeof(ifr))) |
| 1165 | ret = -EFAULT; |
| 1166 | break; |
| 1167 | |
| 1168 | case TUNSETNOCSUM: |
| 1169 | /* Disable/Enable checksum */ |
| 1170 | if (arg) |
| 1171 | tun->flags |= TUN_NOCHECKSUM; |
| 1172 | else |
| 1173 | tun->flags &= ~TUN_NOCHECKSUM; |
| 1174 | |
| 1175 | DBG(KERN_INFO "%s: checksum %s\n", |
| 1176 | tun->dev->name, arg ? "disabled" : "enabled"); |
| 1177 | break; |
| 1178 | |
| 1179 | case TUNSETPERSIST: |
| 1180 | /* Disable/Enable persist mode */ |
| 1181 | if (arg) |
| 1182 | tun->flags |= TUN_PERSIST; |
| 1183 | else |
| 1184 | tun->flags &= ~TUN_PERSIST; |
| 1185 | |
| 1186 | DBG(KERN_INFO "%s: persist %s\n", |
| 1187 | tun->dev->name, arg ? "enabled" : "disabled"); |
| 1188 | break; |
| 1189 | |
| 1190 | case TUNSETOWNER: |
| 1191 | /* Set owner of the device */ |
| 1192 | tun->owner = (uid_t) arg; |
| 1193 | |
| 1194 | DBG(KERN_INFO "%s: owner set to %d\n", tun->dev->name, tun->owner); |
| 1195 | break; |
| 1196 | |
| 1197 | case TUNSETGROUP: |
| 1198 | /* Set group of the device */ |
| 1199 | tun->group= (gid_t) arg; |
| 1200 | |
| 1201 | DBG(KERN_INFO "%s: group set to %d\n", tun->dev->name, tun->group); |
| 1202 | break; |
| 1203 | |
| 1204 | case TUNSETLINK: |
| 1205 | /* Only allow setting the type when the interface is down */ |
| 1206 | if (tun->dev->flags & IFF_UP) { |
| 1207 | DBG(KERN_INFO "%s: Linktype set failed because interface is up\n", |
| 1208 | tun->dev->name); |
| 1209 | ret = -EBUSY; |
| 1210 | } else { |
| 1211 | tun->dev->type = (int) arg; |
| 1212 | DBG(KERN_INFO "%s: linktype set to %d\n", tun->dev->name, tun->dev->type); |
| 1213 | ret = 0; |
| 1214 | } |
| 1215 | break; |
| 1216 | |
| 1217 | #ifdef TUN_DEBUG |
| 1218 | case TUNSETDEBUG: |
| 1219 | tun->debug = arg; |
| 1220 | break; |
| 1221 | #endif |
| 1222 | case TUNSETOFFLOAD: |
| 1223 | ret = set_offload(tun->dev, arg); |
| 1224 | break; |
| 1225 | |
| 1226 | case TUNSETTXFILTER: |
| 1227 | /* Can be set only for TAPs */ |
| 1228 | ret = -EINVAL; |
| 1229 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) |
| 1230 | break; |
| 1231 | ret = update_filter(&tun->txflt, (void __user *)arg); |
| 1232 | break; |
| 1233 | |
| 1234 | case SIOCGIFHWADDR: |
| 1235 | /* Get hw addres */ |
| 1236 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
| 1237 | ifr.ifr_hwaddr.sa_family = tun->dev->type; |
| 1238 | if (copy_to_user(argp, &ifr, sizeof ifr)) |
| 1239 | ret = -EFAULT; |
| 1240 | break; |
| 1241 | |
| 1242 | case SIOCSIFHWADDR: |
| 1243 | /* Set hw address */ |
| 1244 | DBG(KERN_DEBUG "%s: set hw address: %pM\n", |
| 1245 | tun->dev->name, ifr.ifr_hwaddr.sa_data); |
| 1246 | |
| 1247 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
| 1248 | break; |
| 1249 | |
| 1250 | case TUNGETSNDBUF: |
| 1251 | sndbuf = tun->socket.sk->sk_sndbuf; |
| 1252 | if (copy_to_user(argp, &sndbuf, sizeof(sndbuf))) |
| 1253 | ret = -EFAULT; |
| 1254 | break; |
| 1255 | |
| 1256 | case TUNSETSNDBUF: |
| 1257 | if (copy_from_user(&sndbuf, argp, sizeof(sndbuf))) { |
| 1258 | ret = -EFAULT; |
| 1259 | break; |
| 1260 | } |
| 1261 | |
| 1262 | tun->socket.sk->sk_sndbuf = sndbuf; |
| 1263 | break; |
| 1264 | |
| 1265 | default: |
| 1266 | ret = -EINVAL; |
| 1267 | break; |
| 1268 | }; |
| 1269 | |
| 1270 | unlock: |
| 1271 | rtnl_unlock(); |
| 1272 | if (tun) |
| 1273 | tun_put(tun); |
| 1274 | return ret; |
| 1275 | } |
| 1276 | |
| 1277 | static int tun_chr_fasync(int fd, struct file *file, int on) |
| 1278 | { |
| 1279 | struct tun_struct *tun = tun_get(file); |
| 1280 | int ret; |
| 1281 | |
| 1282 | if (!tun) |
| 1283 | return -EBADFD; |
| 1284 | |
| 1285 | DBG(KERN_INFO "%s: tun_chr_fasync %d\n", tun->dev->name, on); |
| 1286 | |
| 1287 | if ((ret = fasync_helper(fd, file, on, &tun->fasync)) < 0) |
| 1288 | goto out; |
| 1289 | |
| 1290 | if (on) { |
| 1291 | ret = __f_setown(file, task_pid(current), PIDTYPE_PID, 0); |
| 1292 | if (ret) |
| 1293 | goto out; |
| 1294 | tun->flags |= TUN_FASYNC; |
| 1295 | } else |
| 1296 | tun->flags &= ~TUN_FASYNC; |
| 1297 | ret = 0; |
| 1298 | out: |
| 1299 | tun_put(tun); |
| 1300 | return ret; |
| 1301 | } |
| 1302 | |
| 1303 | static int tun_chr_open(struct inode *inode, struct file * file) |
| 1304 | { |
| 1305 | struct tun_file *tfile; |
| 1306 | |
| 1307 | DBG1(KERN_INFO "tunX: tun_chr_open\n"); |
| 1308 | |
| 1309 | tfile = kmalloc(sizeof(*tfile), GFP_KERNEL); |
| 1310 | if (!tfile) |
| 1311 | return -ENOMEM; |
| 1312 | atomic_set(&tfile->count, 0); |
| 1313 | tfile->tun = NULL; |
| 1314 | tfile->net = get_net(current->nsproxy->net_ns); |
| 1315 | file->private_data = tfile; |
| 1316 | return 0; |
| 1317 | } |
| 1318 | |
| 1319 | static int tun_chr_close(struct inode *inode, struct file *file) |
| 1320 | { |
| 1321 | struct tun_file *tfile = file->private_data; |
| 1322 | struct tun_struct *tun; |
| 1323 | |
| 1324 | tun = __tun_get(tfile); |
| 1325 | if (tun) { |
| 1326 | struct net_device *dev = tun->dev; |
| 1327 | |
| 1328 | DBG(KERN_INFO "%s: tun_chr_close\n", dev->name); |
| 1329 | |
| 1330 | __tun_detach(tun); |
| 1331 | |
| 1332 | /* If desireable, unregister the netdevice. */ |
| 1333 | if (!(tun->flags & TUN_PERSIST)) { |
| 1334 | rtnl_lock(); |
| 1335 | if (dev->reg_state == NETREG_REGISTERED) |
| 1336 | unregister_netdevice(dev); |
| 1337 | rtnl_unlock(); |
| 1338 | } |
| 1339 | } |
| 1340 | |
| 1341 | tun = tfile->tun; |
| 1342 | if (tun) |
| 1343 | sock_put(tun->socket.sk); |
| 1344 | |
| 1345 | put_net(tfile->net); |
| 1346 | kfree(tfile); |
| 1347 | |
| 1348 | return 0; |
| 1349 | } |
| 1350 | |
| 1351 | static const struct file_operations tun_fops = { |
| 1352 | .owner = THIS_MODULE, |
| 1353 | .llseek = no_llseek, |
| 1354 | .read = do_sync_read, |
| 1355 | .aio_read = tun_chr_aio_read, |
| 1356 | .write = do_sync_write, |
| 1357 | .aio_write = tun_chr_aio_write, |
| 1358 | .poll = tun_chr_poll, |
| 1359 | .unlocked_ioctl = tun_chr_ioctl, |
| 1360 | .open = tun_chr_open, |
| 1361 | .release = tun_chr_close, |
| 1362 | .fasync = tun_chr_fasync |
| 1363 | }; |
| 1364 | |
| 1365 | static struct miscdevice tun_miscdev = { |
| 1366 | .minor = TUN_MINOR, |
| 1367 | .name = "tun", |
| 1368 | .nodename = "net/tun", |
| 1369 | .fops = &tun_fops, |
| 1370 | }; |
| 1371 | |
| 1372 | /* ethtool interface */ |
| 1373 | |
| 1374 | static int tun_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) |
| 1375 | { |
| 1376 | cmd->supported = 0; |
| 1377 | cmd->advertising = 0; |
| 1378 | cmd->speed = SPEED_10; |
| 1379 | cmd->duplex = DUPLEX_FULL; |
| 1380 | cmd->port = PORT_TP; |
| 1381 | cmd->phy_address = 0; |
| 1382 | cmd->transceiver = XCVR_INTERNAL; |
| 1383 | cmd->autoneg = AUTONEG_DISABLE; |
| 1384 | cmd->maxtxpkt = 0; |
| 1385 | cmd->maxrxpkt = 0; |
| 1386 | return 0; |
| 1387 | } |
| 1388 | |
| 1389 | static void tun_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) |
| 1390 | { |
| 1391 | struct tun_struct *tun = netdev_priv(dev); |
| 1392 | |
| 1393 | strcpy(info->driver, DRV_NAME); |
| 1394 | strcpy(info->version, DRV_VERSION); |
| 1395 | strcpy(info->fw_version, "N/A"); |
| 1396 | |
| 1397 | switch (tun->flags & TUN_TYPE_MASK) { |
| 1398 | case TUN_TUN_DEV: |
| 1399 | strcpy(info->bus_info, "tun"); |
| 1400 | break; |
| 1401 | case TUN_TAP_DEV: |
| 1402 | strcpy(info->bus_info, "tap"); |
| 1403 | break; |
| 1404 | } |
| 1405 | } |
| 1406 | |
| 1407 | static u32 tun_get_msglevel(struct net_device *dev) |
| 1408 | { |
| 1409 | #ifdef TUN_DEBUG |
| 1410 | struct tun_struct *tun = netdev_priv(dev); |
| 1411 | return tun->debug; |
| 1412 | #else |
| 1413 | return -EOPNOTSUPP; |
| 1414 | #endif |
| 1415 | } |
| 1416 | |
| 1417 | static void tun_set_msglevel(struct net_device *dev, u32 value) |
| 1418 | { |
| 1419 | #ifdef TUN_DEBUG |
| 1420 | struct tun_struct *tun = netdev_priv(dev); |
| 1421 | tun->debug = value; |
| 1422 | #endif |
| 1423 | } |
| 1424 | |
| 1425 | static u32 tun_get_link(struct net_device *dev) |
| 1426 | { |
| 1427 | struct tun_struct *tun = netdev_priv(dev); |
| 1428 | return !!tun->tfile; |
| 1429 | } |
| 1430 | |
| 1431 | static u32 tun_get_rx_csum(struct net_device *dev) |
| 1432 | { |
| 1433 | struct tun_struct *tun = netdev_priv(dev); |
| 1434 | return (tun->flags & TUN_NOCHECKSUM) == 0; |
| 1435 | } |
| 1436 | |
| 1437 | static int tun_set_rx_csum(struct net_device *dev, u32 data) |
| 1438 | { |
| 1439 | struct tun_struct *tun = netdev_priv(dev); |
| 1440 | if (data) |
| 1441 | tun->flags &= ~TUN_NOCHECKSUM; |
| 1442 | else |
| 1443 | tun->flags |= TUN_NOCHECKSUM; |
| 1444 | return 0; |
| 1445 | } |
| 1446 | |
| 1447 | static const struct ethtool_ops tun_ethtool_ops = { |
| 1448 | .get_settings = tun_get_settings, |
| 1449 | .get_drvinfo = tun_get_drvinfo, |
| 1450 | .get_msglevel = tun_get_msglevel, |
| 1451 | .set_msglevel = tun_set_msglevel, |
| 1452 | .get_link = tun_get_link, |
| 1453 | .get_rx_csum = tun_get_rx_csum, |
| 1454 | .set_rx_csum = tun_set_rx_csum |
| 1455 | }; |
| 1456 | |
| 1457 | |
| 1458 | static int __init tun_init(void) |
| 1459 | { |
| 1460 | int ret = 0; |
| 1461 | |
| 1462 | printk(KERN_INFO "tun: %s, %s\n", DRV_DESCRIPTION, DRV_VERSION); |
| 1463 | printk(KERN_INFO "tun: %s\n", DRV_COPYRIGHT); |
| 1464 | |
| 1465 | ret = rtnl_link_register(&tun_link_ops); |
| 1466 | if (ret) { |
| 1467 | printk(KERN_ERR "tun: Can't register link_ops\n"); |
| 1468 | goto err_linkops; |
| 1469 | } |
| 1470 | |
| 1471 | ret = misc_register(&tun_miscdev); |
| 1472 | if (ret) { |
| 1473 | printk(KERN_ERR "tun: Can't register misc device %d\n", TUN_MINOR); |
| 1474 | goto err_misc; |
| 1475 | } |
| 1476 | return 0; |
| 1477 | err_misc: |
| 1478 | rtnl_link_unregister(&tun_link_ops); |
| 1479 | err_linkops: |
| 1480 | return ret; |
| 1481 | } |
| 1482 | |
| 1483 | static void tun_cleanup(void) |
| 1484 | { |
| 1485 | misc_deregister(&tun_miscdev); |
| 1486 | rtnl_link_unregister(&tun_link_ops); |
| 1487 | } |
| 1488 | |
| 1489 | module_init(tun_init); |
| 1490 | module_exit(tun_cleanup); |
| 1491 | MODULE_DESCRIPTION(DRV_DESCRIPTION); |
| 1492 | MODULE_AUTHOR(DRV_COPYRIGHT); |
| 1493 | MODULE_LICENSE("GPL"); |
| 1494 | MODULE_ALIAS_MISCDEV(TUN_MINOR); |