1 /* 2 * IPv6 virtual tunneling interface 3 * 4 * Copyright (C) 2013 secunet Security Networks AG 5 * 6 * Author: 7 * Steffen Klassert <steffen.klassert@secunet.com> 8 * 9 * Based on: 10 * net/ipv6/ip6_tunnel.c 11 * 12 * This program is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU General Public License 14 * as published by the Free Software Foundation; either version 15 * 2 of the License, or (at your option) any later version. 16 */ 17 18 #include <linux/module.h> 19 #include <linux/capability.h> 20 #include <linux/errno.h> 21 #include <linux/types.h> 22 #include <linux/sockios.h> 23 #include <linux/icmp.h> 24 #include <linux/if.h> 25 #include <linux/in.h> 26 #include <linux/ip.h> 27 #include <linux/net.h> 28 #include <linux/in6.h> 29 #include <linux/netdevice.h> 30 #include <linux/if_arp.h> 31 #include <linux/icmpv6.h> 32 #include <linux/init.h> 33 #include <linux/route.h> 34 #include <linux/rtnetlink.h> 35 #include <linux/netfilter_ipv6.h> 36 #include <linux/slab.h> 37 #include <linux/hash.h> 38 39 #include <linux/uaccess.h> 40 #include <linux/atomic.h> 41 42 #include <net/icmp.h> 43 #include <net/ip.h> 44 #include <net/ip_tunnels.h> 45 #include <net/ipv6.h> 46 #include <net/ip6_route.h> 47 #include <net/addrconf.h> 48 #include <net/ip6_tunnel.h> 49 #include <net/xfrm.h> 50 #include <net/net_namespace.h> 51 #include <net/netns/generic.h> 52 53 #define HASH_SIZE_SHIFT 5 54 #define HASH_SIZE (1 << HASH_SIZE_SHIFT) 55 56 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 57 { 58 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 59 60 return hash_32(hash, HASH_SIZE_SHIFT); 61 } 62 63 static int vti6_dev_init(struct net_device *dev); 64 static void vti6_dev_setup(struct net_device *dev); 65 static struct rtnl_link_ops vti6_link_ops __read_mostly; 66 67 static int vti6_net_id __read_mostly; 68 struct vti6_net { 69 /* the vti6 tunnel fallback device */ 70 struct net_device *fb_tnl_dev; 71 /* lists for storing tunnels in use */ 72 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; 73 struct ip6_tnl __rcu *tnls_wc[1]; 74 struct ip6_tnl __rcu **tnls[2]; 75 }; 76 77 #define for_each_vti6_tunnel_rcu(start) \ 78 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 79 80 /** 81 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses 82 * @net: network namespace 83 * @remote: the address of the tunnel exit-point 84 * @local: the address of the tunnel entry-point 85 * 86 * Return: 87 * tunnel matching given end-points if found, 88 * else fallback tunnel if its device is up, 89 * else %NULL 90 **/ 91 static struct ip6_tnl * 92 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote, 93 const struct in6_addr *local) 94 { 95 unsigned int hash = HASH(remote, local); 96 struct ip6_tnl *t; 97 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 98 struct in6_addr any; 99 100 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 101 if (ipv6_addr_equal(local, &t->parms.laddr) && 102 ipv6_addr_equal(remote, &t->parms.raddr) && 103 (t->dev->flags & IFF_UP)) 104 return t; 105 } 106 107 memset(&any, 0, sizeof(any)); 108 hash = HASH(&any, local); 109 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 110 if (ipv6_addr_equal(local, &t->parms.laddr) && 111 (t->dev->flags & IFF_UP)) 112 return t; 113 } 114 115 hash = HASH(remote, &any); 116 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 117 if (ipv6_addr_equal(remote, &t->parms.raddr) && 118 (t->dev->flags & IFF_UP)) 119 return t; 120 } 121 122 t = rcu_dereference(ip6n->tnls_wc[0]); 123 if (t && (t->dev->flags & IFF_UP)) 124 return t; 125 126 return NULL; 127 } 128 129 /** 130 * vti6_tnl_bucket - get head of list matching given tunnel parameters 131 * @p: parameters containing tunnel end-points 132 * 133 * Description: 134 * vti6_tnl_bucket() returns the head of the list matching the 135 * &struct in6_addr entries laddr and raddr in @p. 136 * 137 * Return: head of IPv6 tunnel list 138 **/ 139 static struct ip6_tnl __rcu ** 140 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 141 { 142 const struct in6_addr *remote = &p->raddr; 143 const struct in6_addr *local = &p->laddr; 144 unsigned int h = 0; 145 int prio = 0; 146 147 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 148 prio = 1; 149 h = HASH(remote, local); 150 } 151 return &ip6n->tnls[prio][h]; 152 } 153 154 static void 155 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 156 { 157 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 158 159 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 160 rcu_assign_pointer(*tp, t); 161 } 162 163 static void 164 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 165 { 166 struct ip6_tnl __rcu **tp; 167 struct ip6_tnl *iter; 168 169 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 170 (iter = rtnl_dereference(*tp)) != NULL; 171 tp = &iter->next) { 172 if (t == iter) { 173 rcu_assign_pointer(*tp, t->next); 174 break; 175 } 176 } 177 } 178 179 static void vti6_dev_free(struct net_device *dev) 180 { 181 free_percpu(dev->tstats); 182 free_netdev(dev); 183 } 184 185 static int vti6_tnl_create2(struct net_device *dev) 186 { 187 struct ip6_tnl *t = netdev_priv(dev); 188 struct net *net = dev_net(dev); 189 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 190 int err; 191 192 err = register_netdevice(dev); 193 if (err < 0) 194 goto out; 195 196 strcpy(t->parms.name, dev->name); 197 dev->rtnl_link_ops = &vti6_link_ops; 198 199 dev_hold(dev); 200 vti6_tnl_link(ip6n, t); 201 202 return 0; 203 204 out: 205 return err; 206 } 207 208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 209 { 210 struct net_device *dev; 211 struct ip6_tnl *t; 212 char name[IFNAMSIZ]; 213 int err; 214 215 if (p->name[0]) 216 strlcpy(name, p->name, IFNAMSIZ); 217 else 218 sprintf(name, "ip6_vti%%d"); 219 220 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 221 if (!dev) 222 goto failed; 223 224 dev_net_set(dev, net); 225 226 t = netdev_priv(dev); 227 t->parms = *p; 228 t->net = dev_net(dev); 229 230 err = vti6_tnl_create2(dev); 231 if (err < 0) 232 goto failed_free; 233 234 return t; 235 236 failed_free: 237 vti6_dev_free(dev); 238 failed: 239 return NULL; 240 } 241 242 /** 243 * vti6_locate - find or create tunnel matching given parameters 244 * @net: network namespace 245 * @p: tunnel parameters 246 * @create: != 0 if allowed to create new tunnel if no match found 247 * 248 * Description: 249 * vti6_locate() first tries to locate an existing tunnel 250 * based on @parms. If this is unsuccessful, but @create is set a new 251 * tunnel device is created and registered for use. 252 * 253 * Return: 254 * matching tunnel or NULL 255 **/ 256 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 257 int create) 258 { 259 const struct in6_addr *remote = &p->raddr; 260 const struct in6_addr *local = &p->laddr; 261 struct ip6_tnl __rcu **tp; 262 struct ip6_tnl *t; 263 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 264 265 for (tp = vti6_tnl_bucket(ip6n, p); 266 (t = rtnl_dereference(*tp)) != NULL; 267 tp = &t->next) { 268 if (ipv6_addr_equal(local, &t->parms.laddr) && 269 ipv6_addr_equal(remote, &t->parms.raddr)) { 270 if (create) 271 return NULL; 272 273 return t; 274 } 275 } 276 if (!create) 277 return NULL; 278 return vti6_tnl_create(net, p); 279 } 280 281 /** 282 * vti6_dev_uninit - tunnel device uninitializer 283 * @dev: the device to be destroyed 284 * 285 * Description: 286 * vti6_dev_uninit() removes tunnel from its list 287 **/ 288 static void vti6_dev_uninit(struct net_device *dev) 289 { 290 struct ip6_tnl *t = netdev_priv(dev); 291 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 292 293 if (dev == ip6n->fb_tnl_dev) 294 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 295 else 296 vti6_tnl_unlink(ip6n, t); 297 dev_put(dev); 298 } 299 300 static int vti6_rcv(struct sk_buff *skb) 301 { 302 struct ip6_tnl *t; 303 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 304 305 rcu_read_lock(); 306 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 307 if (t) { 308 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 309 rcu_read_unlock(); 310 goto discard; 311 } 312 313 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 314 rcu_read_unlock(); 315 return 0; 316 } 317 318 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 319 t->dev->stats.rx_dropped++; 320 rcu_read_unlock(); 321 goto discard; 322 } 323 324 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 325 skb->mark = be32_to_cpu(t->parms.i_key); 326 327 rcu_read_unlock(); 328 329 return xfrm6_rcv(skb); 330 } 331 rcu_read_unlock(); 332 return -EINVAL; 333 discard: 334 kfree_skb(skb); 335 return 0; 336 } 337 338 static int vti6_rcv_cb(struct sk_buff *skb, int err) 339 { 340 unsigned short family; 341 struct net_device *dev; 342 struct pcpu_sw_netstats *tstats; 343 struct xfrm_state *x; 344 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 345 346 if (!t) 347 return 1; 348 349 dev = t->dev; 350 351 if (err) { 352 dev->stats.rx_errors++; 353 dev->stats.rx_dropped++; 354 355 return 0; 356 } 357 358 x = xfrm_input_state(skb); 359 family = x->inner_mode->afinfo->family; 360 361 if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family)) 362 return -EPERM; 363 364 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 365 skb->dev = dev; 366 367 tstats = this_cpu_ptr(dev->tstats); 368 u64_stats_update_begin(&tstats->syncp); 369 tstats->rx_packets++; 370 tstats->rx_bytes += skb->len; 371 u64_stats_update_end(&tstats->syncp); 372 373 return 0; 374 } 375 376 /** 377 * vti6_addr_conflict - compare packet addresses to tunnel's own 378 * @t: the outgoing tunnel device 379 * @hdr: IPv6 header from the incoming packet 380 * 381 * Description: 382 * Avoid trivial tunneling loop by checking that tunnel exit-point 383 * doesn't match source of incoming packet. 384 * 385 * Return: 386 * 1 if conflict, 387 * 0 else 388 **/ 389 static inline bool 390 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 391 { 392 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 393 } 394 395 static bool vti6_state_check(const struct xfrm_state *x, 396 const struct in6_addr *dst, 397 const struct in6_addr *src) 398 { 399 xfrm_address_t *daddr = (xfrm_address_t *)dst; 400 xfrm_address_t *saddr = (xfrm_address_t *)src; 401 402 /* if there is no transform then this tunnel is not functional. 403 * Or if the xfrm is not mode tunnel. 404 */ 405 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 406 x->props.family != AF_INET6) 407 return false; 408 409 if (ipv6_addr_any(dst)) 410 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 411 412 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 413 return false; 414 415 return true; 416 } 417 418 /** 419 * vti6_xmit - send a packet 420 * @skb: the outgoing socket buffer 421 * @dev: the outgoing tunnel device 422 * @fl: the flow informations for the xfrm_lookup 423 **/ 424 static int 425 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 426 { 427 struct ip6_tnl *t = netdev_priv(dev); 428 struct net_device_stats *stats = &t->dev->stats; 429 struct dst_entry *dst = skb_dst(skb); 430 struct net_device *tdev; 431 struct xfrm_state *x; 432 int err = -1; 433 434 if (!dst) 435 goto tx_err_link_failure; 436 437 dst_hold(dst); 438 dst = xfrm_lookup(t->net, dst, fl, NULL, 0); 439 if (IS_ERR(dst)) { 440 err = PTR_ERR(dst); 441 dst = NULL; 442 goto tx_err_link_failure; 443 } 444 445 x = dst->xfrm; 446 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 447 goto tx_err_link_failure; 448 449 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 450 (const struct in6_addr *)&x->id.daddr)) 451 goto tx_err_link_failure; 452 453 tdev = dst->dev; 454 455 if (tdev == dev) { 456 stats->collisions++; 457 net_warn_ratelimited("%s: Local routing loop detected!\n", 458 t->parms.name); 459 goto tx_err_dst_release; 460 } 461 462 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 463 skb_dst_set(skb, dst); 464 skb->dev = skb_dst(skb)->dev; 465 466 err = dst_output(skb); 467 if (net_xmit_eval(err) == 0) { 468 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 469 470 u64_stats_update_begin(&tstats->syncp); 471 tstats->tx_bytes += skb->len; 472 tstats->tx_packets++; 473 u64_stats_update_end(&tstats->syncp); 474 } else { 475 stats->tx_errors++; 476 stats->tx_aborted_errors++; 477 } 478 479 return 0; 480 tx_err_link_failure: 481 stats->tx_carrier_errors++; 482 dst_link_failure(skb); 483 tx_err_dst_release: 484 dst_release(dst); 485 return err; 486 } 487 488 static netdev_tx_t 489 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 490 { 491 struct ip6_tnl *t = netdev_priv(dev); 492 struct net_device_stats *stats = &t->dev->stats; 493 struct ipv6hdr *ipv6h; 494 struct flowi fl; 495 int ret; 496 497 memset(&fl, 0, sizeof(fl)); 498 skb->mark = be32_to_cpu(t->parms.o_key); 499 500 switch (skb->protocol) { 501 case htons(ETH_P_IPV6): 502 ipv6h = ipv6_hdr(skb); 503 504 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 505 vti6_addr_conflict(t, ipv6h)) 506 goto tx_err; 507 508 xfrm_decode_session(skb, &fl, AF_INET6); 509 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 510 break; 511 case htons(ETH_P_IP): 512 xfrm_decode_session(skb, &fl, AF_INET); 513 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 514 break; 515 default: 516 goto tx_err; 517 } 518 519 ret = vti6_xmit(skb, dev, &fl); 520 if (ret < 0) 521 goto tx_err; 522 523 return NETDEV_TX_OK; 524 525 tx_err: 526 stats->tx_errors++; 527 stats->tx_dropped++; 528 kfree_skb(skb); 529 return NETDEV_TX_OK; 530 } 531 532 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 533 u8 type, u8 code, int offset, __be32 info) 534 { 535 __be32 spi; 536 __u32 mark; 537 struct xfrm_state *x; 538 struct ip6_tnl *t; 539 struct ip_esp_hdr *esph; 540 struct ip_auth_hdr *ah; 541 struct ip_comp_hdr *ipch; 542 struct net *net = dev_net(skb->dev); 543 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 544 int protocol = iph->nexthdr; 545 546 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 547 if (!t) 548 return -1; 549 550 mark = be32_to_cpu(t->parms.o_key); 551 552 switch (protocol) { 553 case IPPROTO_ESP: 554 esph = (struct ip_esp_hdr *)(skb->data + offset); 555 spi = esph->spi; 556 break; 557 case IPPROTO_AH: 558 ah = (struct ip_auth_hdr *)(skb->data + offset); 559 spi = ah->spi; 560 break; 561 case IPPROTO_COMP: 562 ipch = (struct ip_comp_hdr *)(skb->data + offset); 563 spi = htonl(ntohs(ipch->cpi)); 564 break; 565 default: 566 return 0; 567 } 568 569 if (type != ICMPV6_PKT_TOOBIG && 570 type != NDISC_REDIRECT) 571 return 0; 572 573 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 574 spi, protocol, AF_INET6); 575 if (!x) 576 return 0; 577 578 if (type == NDISC_REDIRECT) 579 ip6_redirect(skb, net, skb->dev->ifindex, 0); 580 else 581 ip6_update_pmtu(skb, net, info, 0, 0); 582 xfrm_state_put(x); 583 584 return 0; 585 } 586 587 static void vti6_link_config(struct ip6_tnl *t) 588 { 589 struct net_device *dev = t->dev; 590 struct __ip6_tnl_parm *p = &t->parms; 591 592 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 593 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 594 595 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 596 IP6_TNL_F_CAP_PER_PACKET); 597 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 598 599 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 600 dev->flags |= IFF_POINTOPOINT; 601 else 602 dev->flags &= ~IFF_POINTOPOINT; 603 } 604 605 /** 606 * vti6_tnl_change - update the tunnel parameters 607 * @t: tunnel to be changed 608 * @p: tunnel configuration parameters 609 * 610 * Description: 611 * vti6_tnl_change() updates the tunnel parameters 612 **/ 613 static int 614 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 615 { 616 t->parms.laddr = p->laddr; 617 t->parms.raddr = p->raddr; 618 t->parms.link = p->link; 619 t->parms.i_key = p->i_key; 620 t->parms.o_key = p->o_key; 621 t->parms.proto = p->proto; 622 ip6_tnl_dst_reset(t); 623 vti6_link_config(t); 624 return 0; 625 } 626 627 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 628 { 629 struct net *net = dev_net(t->dev); 630 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 631 int err; 632 633 vti6_tnl_unlink(ip6n, t); 634 synchronize_net(); 635 err = vti6_tnl_change(t, p); 636 vti6_tnl_link(ip6n, t); 637 netdev_state_change(t->dev); 638 return err; 639 } 640 641 static void 642 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 643 { 644 p->laddr = u->laddr; 645 p->raddr = u->raddr; 646 p->link = u->link; 647 p->i_key = u->i_key; 648 p->o_key = u->o_key; 649 p->proto = u->proto; 650 651 memcpy(p->name, u->name, sizeof(u->name)); 652 } 653 654 static void 655 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 656 { 657 u->laddr = p->laddr; 658 u->raddr = p->raddr; 659 u->link = p->link; 660 u->i_key = p->i_key; 661 u->o_key = p->o_key; 662 u->proto = p->proto; 663 664 memcpy(u->name, p->name, sizeof(u->name)); 665 } 666 667 /** 668 * vti6_tnl_ioctl - configure vti6 tunnels from userspace 669 * @dev: virtual device associated with tunnel 670 * @ifr: parameters passed from userspace 671 * @cmd: command to be performed 672 * 673 * Description: 674 * vti6_ioctl() is used for managing vti6 tunnels 675 * from userspace. 676 * 677 * The possible commands are the following: 678 * %SIOCGETTUNNEL: get tunnel parameters for device 679 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 680 * %SIOCCHGTUNNEL: change tunnel parameters to those given 681 * %SIOCDELTUNNEL: delete tunnel 682 * 683 * The fallback device "ip6_vti0", created during module 684 * initialization, can be used for creating other tunnel devices. 685 * 686 * Return: 687 * 0 on success, 688 * %-EFAULT if unable to copy data to or from userspace, 689 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 690 * %-EINVAL if passed tunnel parameters are invalid, 691 * %-EEXIST if changing a tunnel's parameters would cause a conflict 692 * %-ENODEV if attempting to change or delete a nonexisting device 693 **/ 694 static int 695 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 696 { 697 int err = 0; 698 struct ip6_tnl_parm2 p; 699 struct __ip6_tnl_parm p1; 700 struct ip6_tnl *t = NULL; 701 struct net *net = dev_net(dev); 702 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 703 704 switch (cmd) { 705 case SIOCGETTUNNEL: 706 if (dev == ip6n->fb_tnl_dev) { 707 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 708 err = -EFAULT; 709 break; 710 } 711 vti6_parm_from_user(&p1, &p); 712 t = vti6_locate(net, &p1, 0); 713 } else { 714 memset(&p, 0, sizeof(p)); 715 } 716 if (!t) 717 t = netdev_priv(dev); 718 vti6_parm_to_user(&p, &t->parms); 719 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 720 err = -EFAULT; 721 break; 722 case SIOCADDTUNNEL: 723 case SIOCCHGTUNNEL: 724 err = -EPERM; 725 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 726 break; 727 err = -EFAULT; 728 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 729 break; 730 err = -EINVAL; 731 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 732 break; 733 vti6_parm_from_user(&p1, &p); 734 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 735 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 736 if (t) { 737 if (t->dev != dev) { 738 err = -EEXIST; 739 break; 740 } 741 } else 742 t = netdev_priv(dev); 743 744 err = vti6_update(t, &p1); 745 } 746 if (t) { 747 err = 0; 748 vti6_parm_to_user(&p, &t->parms); 749 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 750 err = -EFAULT; 751 752 } else 753 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 754 break; 755 case SIOCDELTUNNEL: 756 err = -EPERM; 757 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 758 break; 759 760 if (dev == ip6n->fb_tnl_dev) { 761 err = -EFAULT; 762 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 763 break; 764 err = -ENOENT; 765 vti6_parm_from_user(&p1, &p); 766 t = vti6_locate(net, &p1, 0); 767 if (!t) 768 break; 769 err = -EPERM; 770 if (t->dev == ip6n->fb_tnl_dev) 771 break; 772 dev = t->dev; 773 } 774 err = 0; 775 unregister_netdevice(dev); 776 break; 777 default: 778 err = -EINVAL; 779 } 780 return err; 781 } 782 783 /** 784 * vti6_tnl_change_mtu - change mtu manually for tunnel device 785 * @dev: virtual device associated with tunnel 786 * @new_mtu: the new mtu 787 * 788 * Return: 789 * 0 on success, 790 * %-EINVAL if mtu too small 791 **/ 792 static int vti6_change_mtu(struct net_device *dev, int new_mtu) 793 { 794 if (new_mtu < IPV6_MIN_MTU) 795 return -EINVAL; 796 797 dev->mtu = new_mtu; 798 return 0; 799 } 800 801 static const struct net_device_ops vti6_netdev_ops = { 802 .ndo_init = vti6_dev_init, 803 .ndo_uninit = vti6_dev_uninit, 804 .ndo_start_xmit = vti6_tnl_xmit, 805 .ndo_do_ioctl = vti6_ioctl, 806 .ndo_change_mtu = vti6_change_mtu, 807 .ndo_get_stats64 = ip_tunnel_get_stats64, 808 .ndo_get_iflink = ip6_tnl_get_iflink, 809 }; 810 811 /** 812 * vti6_dev_setup - setup virtual tunnel device 813 * @dev: virtual device associated with tunnel 814 * 815 * Description: 816 * Initialize function pointers and device parameters 817 **/ 818 static void vti6_dev_setup(struct net_device *dev) 819 { 820 dev->netdev_ops = &vti6_netdev_ops; 821 dev->destructor = vti6_dev_free; 822 823 dev->type = ARPHRD_TUNNEL6; 824 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr); 825 dev->mtu = ETH_DATA_LEN; 826 dev->flags |= IFF_NOARP; 827 dev->addr_len = sizeof(struct in6_addr); 828 netif_keep_dst(dev); 829 } 830 831 /** 832 * vti6_dev_init_gen - general initializer for all tunnel devices 833 * @dev: virtual device associated with tunnel 834 **/ 835 static inline int vti6_dev_init_gen(struct net_device *dev) 836 { 837 struct ip6_tnl *t = netdev_priv(dev); 838 839 t->dev = dev; 840 t->net = dev_net(dev); 841 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 842 if (!dev->tstats) 843 return -ENOMEM; 844 return 0; 845 } 846 847 /** 848 * vti6_dev_init - initializer for all non fallback tunnel devices 849 * @dev: virtual device associated with tunnel 850 **/ 851 static int vti6_dev_init(struct net_device *dev) 852 { 853 struct ip6_tnl *t = netdev_priv(dev); 854 int err = vti6_dev_init_gen(dev); 855 856 if (err) 857 return err; 858 vti6_link_config(t); 859 return 0; 860 } 861 862 /** 863 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 864 * @dev: fallback device 865 * 866 * Return: 0 867 **/ 868 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 869 { 870 struct ip6_tnl *t = netdev_priv(dev); 871 struct net *net = dev_net(dev); 872 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 873 874 t->parms.proto = IPPROTO_IPV6; 875 dev_hold(dev); 876 877 rcu_assign_pointer(ip6n->tnls_wc[0], t); 878 return 0; 879 } 880 881 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[]) 882 { 883 return 0; 884 } 885 886 static void vti6_netlink_parms(struct nlattr *data[], 887 struct __ip6_tnl_parm *parms) 888 { 889 memset(parms, 0, sizeof(*parms)); 890 891 if (!data) 892 return; 893 894 if (data[IFLA_VTI_LINK]) 895 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 896 897 if (data[IFLA_VTI_LOCAL]) 898 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 899 900 if (data[IFLA_VTI_REMOTE]) 901 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 902 903 if (data[IFLA_VTI_IKEY]) 904 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 905 906 if (data[IFLA_VTI_OKEY]) 907 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 908 } 909 910 static int vti6_newlink(struct net *src_net, struct net_device *dev, 911 struct nlattr *tb[], struct nlattr *data[]) 912 { 913 struct net *net = dev_net(dev); 914 struct ip6_tnl *nt; 915 916 nt = netdev_priv(dev); 917 vti6_netlink_parms(data, &nt->parms); 918 919 nt->parms.proto = IPPROTO_IPV6; 920 921 if (vti6_locate(net, &nt->parms, 0)) 922 return -EEXIST; 923 924 return vti6_tnl_create2(dev); 925 } 926 927 static void vti6_dellink(struct net_device *dev, struct list_head *head) 928 { 929 struct net *net = dev_net(dev); 930 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 931 932 if (dev != ip6n->fb_tnl_dev) 933 unregister_netdevice_queue(dev, head); 934 } 935 936 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 937 struct nlattr *data[]) 938 { 939 struct ip6_tnl *t; 940 struct __ip6_tnl_parm p; 941 struct net *net = dev_net(dev); 942 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 943 944 if (dev == ip6n->fb_tnl_dev) 945 return -EINVAL; 946 947 vti6_netlink_parms(data, &p); 948 949 t = vti6_locate(net, &p, 0); 950 951 if (t) { 952 if (t->dev != dev) 953 return -EEXIST; 954 } else 955 t = netdev_priv(dev); 956 957 return vti6_update(t, &p); 958 } 959 960 static size_t vti6_get_size(const struct net_device *dev) 961 { 962 return 963 /* IFLA_VTI_LINK */ 964 nla_total_size(4) + 965 /* IFLA_VTI_LOCAL */ 966 nla_total_size(sizeof(struct in6_addr)) + 967 /* IFLA_VTI_REMOTE */ 968 nla_total_size(sizeof(struct in6_addr)) + 969 /* IFLA_VTI_IKEY */ 970 nla_total_size(4) + 971 /* IFLA_VTI_OKEY */ 972 nla_total_size(4) + 973 0; 974 } 975 976 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 977 { 978 struct ip6_tnl *tunnel = netdev_priv(dev); 979 struct __ip6_tnl_parm *parm = &tunnel->parms; 980 981 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 982 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 983 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 984 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 985 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key)) 986 goto nla_put_failure; 987 return 0; 988 989 nla_put_failure: 990 return -EMSGSIZE; 991 } 992 993 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 994 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 995 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 996 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 997 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 998 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 999 }; 1000 1001 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1002 .kind = "vti6", 1003 .maxtype = IFLA_VTI_MAX, 1004 .policy = vti6_policy, 1005 .priv_size = sizeof(struct ip6_tnl), 1006 .setup = vti6_dev_setup, 1007 .validate = vti6_validate, 1008 .newlink = vti6_newlink, 1009 .dellink = vti6_dellink, 1010 .changelink = vti6_changelink, 1011 .get_size = vti6_get_size, 1012 .fill_info = vti6_fill_info, 1013 .get_link_net = ip6_tnl_get_link_net, 1014 }; 1015 1016 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n) 1017 { 1018 int h; 1019 struct ip6_tnl *t; 1020 LIST_HEAD(list); 1021 1022 for (h = 0; h < HASH_SIZE; h++) { 1023 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1024 while (t) { 1025 unregister_netdevice_queue(t->dev, &list); 1026 t = rtnl_dereference(t->next); 1027 } 1028 } 1029 1030 t = rtnl_dereference(ip6n->tnls_wc[0]); 1031 unregister_netdevice_queue(t->dev, &list); 1032 unregister_netdevice_many(&list); 1033 } 1034 1035 static int __net_init vti6_init_net(struct net *net) 1036 { 1037 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1038 struct ip6_tnl *t = NULL; 1039 int err; 1040 1041 ip6n->tnls[0] = ip6n->tnls_wc; 1042 ip6n->tnls[1] = ip6n->tnls_r_l; 1043 1044 err = -ENOMEM; 1045 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1046 NET_NAME_UNKNOWN, vti6_dev_setup); 1047 1048 if (!ip6n->fb_tnl_dev) 1049 goto err_alloc_dev; 1050 dev_net_set(ip6n->fb_tnl_dev, net); 1051 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1052 1053 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1054 if (err < 0) 1055 goto err_register; 1056 1057 err = register_netdev(ip6n->fb_tnl_dev); 1058 if (err < 0) 1059 goto err_register; 1060 1061 t = netdev_priv(ip6n->fb_tnl_dev); 1062 1063 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1064 return 0; 1065 1066 err_register: 1067 vti6_dev_free(ip6n->fb_tnl_dev); 1068 err_alloc_dev: 1069 return err; 1070 } 1071 1072 static void __net_exit vti6_exit_net(struct net *net) 1073 { 1074 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1075 1076 rtnl_lock(); 1077 vti6_destroy_tunnels(ip6n); 1078 rtnl_unlock(); 1079 } 1080 1081 static struct pernet_operations vti6_net_ops = { 1082 .init = vti6_init_net, 1083 .exit = vti6_exit_net, 1084 .id = &vti6_net_id, 1085 .size = sizeof(struct vti6_net), 1086 }; 1087 1088 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1089 .handler = vti6_rcv, 1090 .cb_handler = vti6_rcv_cb, 1091 .err_handler = vti6_err, 1092 .priority = 100, 1093 }; 1094 1095 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1096 .handler = vti6_rcv, 1097 .cb_handler = vti6_rcv_cb, 1098 .err_handler = vti6_err, 1099 .priority = 100, 1100 }; 1101 1102 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1103 .handler = vti6_rcv, 1104 .cb_handler = vti6_rcv_cb, 1105 .err_handler = vti6_err, 1106 .priority = 100, 1107 }; 1108 1109 /** 1110 * vti6_tunnel_init - register protocol and reserve needed resources 1111 * 1112 * Return: 0 on success 1113 **/ 1114 static int __init vti6_tunnel_init(void) 1115 { 1116 const char *msg; 1117 int err; 1118 1119 msg = "tunnel device"; 1120 err = register_pernet_device(&vti6_net_ops); 1121 if (err < 0) 1122 goto pernet_dev_failed; 1123 1124 msg = "tunnel protocols"; 1125 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1126 if (err < 0) 1127 goto xfrm_proto_esp_failed; 1128 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1129 if (err < 0) 1130 goto xfrm_proto_ah_failed; 1131 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1132 if (err < 0) 1133 goto xfrm_proto_comp_failed; 1134 1135 msg = "netlink interface"; 1136 err = rtnl_link_register(&vti6_link_ops); 1137 if (err < 0) 1138 goto rtnl_link_failed; 1139 1140 return 0; 1141 1142 rtnl_link_failed: 1143 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1144 xfrm_proto_comp_failed: 1145 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1146 xfrm_proto_ah_failed: 1147 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1148 xfrm_proto_esp_failed: 1149 unregister_pernet_device(&vti6_net_ops); 1150 pernet_dev_failed: 1151 pr_err("vti6 init: failed to register %s\n", msg); 1152 return err; 1153 } 1154 1155 /** 1156 * vti6_tunnel_cleanup - free resources and unregister protocol 1157 **/ 1158 static void __exit vti6_tunnel_cleanup(void) 1159 { 1160 rtnl_link_unregister(&vti6_link_ops); 1161 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1162 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1163 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1164 unregister_pernet_device(&vti6_net_ops); 1165 } 1166 1167 module_init(vti6_tunnel_init); 1168 module_exit(vti6_tunnel_cleanup); 1169 MODULE_LICENSE("GPL"); 1170 MODULE_ALIAS_RTNL_LINK("vti6"); 1171 MODULE_ALIAS_NETDEV("ip6_vti0"); 1172 MODULE_AUTHOR("Steffen Klassert"); 1173 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1174