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 #include <linux/etherdevice.h> 53 54 #define IP6_VTI_HASH_SIZE_SHIFT 5 55 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT) 56 57 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 58 { 59 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 60 61 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT); 62 } 63 64 static int vti6_dev_init(struct net_device *dev); 65 static void vti6_dev_setup(struct net_device *dev); 66 static struct rtnl_link_ops vti6_link_ops __read_mostly; 67 68 static unsigned int vti6_net_id __read_mostly; 69 struct vti6_net { 70 /* the vti6 tunnel fallback device */ 71 struct net_device *fb_tnl_dev; 72 /* lists for storing tunnels in use */ 73 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE]; 74 struct ip6_tnl __rcu *tnls_wc[1]; 75 struct ip6_tnl __rcu **tnls[2]; 76 }; 77 78 #define for_each_vti6_tunnel_rcu(start) \ 79 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 80 81 /** 82 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses 83 * @net: network namespace 84 * @remote: the address of the tunnel exit-point 85 * @local: the address of the tunnel entry-point 86 * 87 * Return: 88 * tunnel matching given end-points if found, 89 * else fallback tunnel if its device is up, 90 * else %NULL 91 **/ 92 static struct ip6_tnl * 93 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote, 94 const struct in6_addr *local) 95 { 96 unsigned int hash = HASH(remote, local); 97 struct ip6_tnl *t; 98 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 99 struct in6_addr any; 100 101 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 102 if (ipv6_addr_equal(local, &t->parms.laddr) && 103 ipv6_addr_equal(remote, &t->parms.raddr) && 104 (t->dev->flags & IFF_UP)) 105 return t; 106 } 107 108 memset(&any, 0, sizeof(any)); 109 hash = HASH(&any, local); 110 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 111 if (ipv6_addr_equal(local, &t->parms.laddr) && 112 (t->dev->flags & IFF_UP)) 113 return t; 114 } 115 116 hash = HASH(remote, &any); 117 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 118 if (ipv6_addr_equal(remote, &t->parms.raddr) && 119 (t->dev->flags & IFF_UP)) 120 return t; 121 } 122 123 t = rcu_dereference(ip6n->tnls_wc[0]); 124 if (t && (t->dev->flags & IFF_UP)) 125 return t; 126 127 return NULL; 128 } 129 130 /** 131 * vti6_tnl_bucket - get head of list matching given tunnel parameters 132 * @p: parameters containing tunnel end-points 133 * 134 * Description: 135 * vti6_tnl_bucket() returns the head of the list matching the 136 * &struct in6_addr entries laddr and raddr in @p. 137 * 138 * Return: head of IPv6 tunnel list 139 **/ 140 static struct ip6_tnl __rcu ** 141 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 142 { 143 const struct in6_addr *remote = &p->raddr; 144 const struct in6_addr *local = &p->laddr; 145 unsigned int h = 0; 146 int prio = 0; 147 148 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 149 prio = 1; 150 h = HASH(remote, local); 151 } 152 return &ip6n->tnls[prio][h]; 153 } 154 155 static void 156 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 157 { 158 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 159 160 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 161 rcu_assign_pointer(*tp, t); 162 } 163 164 static void 165 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 166 { 167 struct ip6_tnl __rcu **tp; 168 struct ip6_tnl *iter; 169 170 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 171 (iter = rtnl_dereference(*tp)) != NULL; 172 tp = &iter->next) { 173 if (t == iter) { 174 rcu_assign_pointer(*tp, t->next); 175 break; 176 } 177 } 178 } 179 180 static void vti6_dev_free(struct net_device *dev) 181 { 182 free_percpu(dev->tstats); 183 free_netdev(dev); 184 } 185 186 static int vti6_tnl_create2(struct net_device *dev) 187 { 188 struct ip6_tnl *t = netdev_priv(dev); 189 struct net *net = dev_net(dev); 190 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 191 int err; 192 193 dev->rtnl_link_ops = &vti6_link_ops; 194 err = register_netdevice(dev); 195 if (err < 0) 196 goto out; 197 198 strcpy(t->parms.name, dev->name); 199 200 dev_hold(dev); 201 vti6_tnl_link(ip6n, t); 202 203 return 0; 204 205 out: 206 return err; 207 } 208 209 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 210 { 211 struct net_device *dev; 212 struct ip6_tnl *t; 213 char name[IFNAMSIZ]; 214 int err; 215 216 if (p->name[0]) 217 strlcpy(name, p->name, IFNAMSIZ); 218 else 219 sprintf(name, "ip6_vti%%d"); 220 221 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 222 if (!dev) 223 goto failed; 224 225 dev_net_set(dev, net); 226 227 t = netdev_priv(dev); 228 t->parms = *p; 229 t->net = dev_net(dev); 230 231 err = vti6_tnl_create2(dev); 232 if (err < 0) 233 goto failed_free; 234 235 return t; 236 237 failed_free: 238 vti6_dev_free(dev); 239 failed: 240 return NULL; 241 } 242 243 /** 244 * vti6_locate - find or create tunnel matching given parameters 245 * @net: network namespace 246 * @p: tunnel parameters 247 * @create: != 0 if allowed to create new tunnel if no match found 248 * 249 * Description: 250 * vti6_locate() first tries to locate an existing tunnel 251 * based on @parms. If this is unsuccessful, but @create is set a new 252 * tunnel device is created and registered for use. 253 * 254 * Return: 255 * matching tunnel or NULL 256 **/ 257 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 258 int create) 259 { 260 const struct in6_addr *remote = &p->raddr; 261 const struct in6_addr *local = &p->laddr; 262 struct ip6_tnl __rcu **tp; 263 struct ip6_tnl *t; 264 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 265 266 for (tp = vti6_tnl_bucket(ip6n, p); 267 (t = rtnl_dereference(*tp)) != NULL; 268 tp = &t->next) { 269 if (ipv6_addr_equal(local, &t->parms.laddr) && 270 ipv6_addr_equal(remote, &t->parms.raddr)) { 271 if (create) 272 return NULL; 273 274 return t; 275 } 276 } 277 if (!create) 278 return NULL; 279 return vti6_tnl_create(net, p); 280 } 281 282 /** 283 * vti6_dev_uninit - tunnel device uninitializer 284 * @dev: the device to be destroyed 285 * 286 * Description: 287 * vti6_dev_uninit() removes tunnel from its list 288 **/ 289 static void vti6_dev_uninit(struct net_device *dev) 290 { 291 struct ip6_tnl *t = netdev_priv(dev); 292 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 293 294 if (dev == ip6n->fb_tnl_dev) 295 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 296 else 297 vti6_tnl_unlink(ip6n, t); 298 dev_put(dev); 299 } 300 301 static int vti6_rcv(struct sk_buff *skb) 302 { 303 struct ip6_tnl *t; 304 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 305 306 rcu_read_lock(); 307 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 308 if (t) { 309 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 310 rcu_read_unlock(); 311 goto discard; 312 } 313 314 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 315 rcu_read_unlock(); 316 return 0; 317 } 318 319 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 320 t->dev->stats.rx_dropped++; 321 rcu_read_unlock(); 322 goto discard; 323 } 324 325 rcu_read_unlock(); 326 327 return xfrm6_rcv_tnl(skb, t); 328 } 329 rcu_read_unlock(); 330 return -EINVAL; 331 discard: 332 kfree_skb(skb); 333 return 0; 334 } 335 336 static int vti6_rcv_cb(struct sk_buff *skb, int err) 337 { 338 unsigned short family; 339 struct net_device *dev; 340 struct pcpu_sw_netstats *tstats; 341 struct xfrm_state *x; 342 struct xfrm_mode *inner_mode; 343 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 344 u32 orig_mark = skb->mark; 345 int ret; 346 347 if (!t) 348 return 1; 349 350 dev = t->dev; 351 352 if (err) { 353 dev->stats.rx_errors++; 354 dev->stats.rx_dropped++; 355 356 return 0; 357 } 358 359 x = xfrm_input_state(skb); 360 361 inner_mode = x->inner_mode; 362 363 if (x->sel.family == AF_UNSPEC) { 364 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 365 if (inner_mode == NULL) { 366 XFRM_INC_STATS(dev_net(skb->dev), 367 LINUX_MIB_XFRMINSTATEMODEERROR); 368 return -EINVAL; 369 } 370 } 371 372 family = inner_mode->afinfo->family; 373 374 skb->mark = be32_to_cpu(t->parms.i_key); 375 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 376 skb->mark = orig_mark; 377 378 if (!ret) 379 return -EPERM; 380 381 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 382 skb->dev = dev; 383 384 tstats = this_cpu_ptr(dev->tstats); 385 u64_stats_update_begin(&tstats->syncp); 386 tstats->rx_packets++; 387 tstats->rx_bytes += skb->len; 388 u64_stats_update_end(&tstats->syncp); 389 390 return 0; 391 } 392 393 /** 394 * vti6_addr_conflict - compare packet addresses to tunnel's own 395 * @t: the outgoing tunnel device 396 * @hdr: IPv6 header from the incoming packet 397 * 398 * Description: 399 * Avoid trivial tunneling loop by checking that tunnel exit-point 400 * doesn't match source of incoming packet. 401 * 402 * Return: 403 * 1 if conflict, 404 * 0 else 405 **/ 406 static inline bool 407 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 408 { 409 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 410 } 411 412 static bool vti6_state_check(const struct xfrm_state *x, 413 const struct in6_addr *dst, 414 const struct in6_addr *src) 415 { 416 xfrm_address_t *daddr = (xfrm_address_t *)dst; 417 xfrm_address_t *saddr = (xfrm_address_t *)src; 418 419 /* if there is no transform then this tunnel is not functional. 420 * Or if the xfrm is not mode tunnel. 421 */ 422 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 423 x->props.family != AF_INET6) 424 return false; 425 426 if (ipv6_addr_any(dst)) 427 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 428 429 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 430 return false; 431 432 return true; 433 } 434 435 /** 436 * vti6_xmit - send a packet 437 * @skb: the outgoing socket buffer 438 * @dev: the outgoing tunnel device 439 * @fl: the flow informations for the xfrm_lookup 440 **/ 441 static int 442 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 443 { 444 struct ip6_tnl *t = netdev_priv(dev); 445 struct net_device_stats *stats = &t->dev->stats; 446 struct dst_entry *dst = skb_dst(skb); 447 struct net_device *tdev; 448 struct xfrm_state *x; 449 int err = -1; 450 int mtu; 451 452 if (!dst) 453 goto tx_err_link_failure; 454 455 dst_hold(dst); 456 dst = xfrm_lookup(t->net, dst, fl, NULL, 0); 457 if (IS_ERR(dst)) { 458 err = PTR_ERR(dst); 459 dst = NULL; 460 goto tx_err_link_failure; 461 } 462 463 x = dst->xfrm; 464 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 465 goto tx_err_link_failure; 466 467 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 468 (const struct in6_addr *)&x->id.daddr)) 469 goto tx_err_link_failure; 470 471 tdev = dst->dev; 472 473 if (tdev == dev) { 474 stats->collisions++; 475 net_warn_ratelimited("%s: Local routing loop detected!\n", 476 t->parms.name); 477 goto tx_err_dst_release; 478 } 479 480 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 481 skb_dst_set(skb, dst); 482 skb->dev = skb_dst(skb)->dev; 483 484 mtu = dst_mtu(dst); 485 if (!skb->ignore_df && skb->len > mtu) { 486 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu); 487 488 if (skb->protocol == htons(ETH_P_IPV6)) 489 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 490 else 491 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 492 htonl(mtu)); 493 494 return -EMSGSIZE; 495 } 496 497 err = dst_output(t->net, skb->sk, skb); 498 if (net_xmit_eval(err) == 0) { 499 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 500 501 u64_stats_update_begin(&tstats->syncp); 502 tstats->tx_bytes += skb->len; 503 tstats->tx_packets++; 504 u64_stats_update_end(&tstats->syncp); 505 } else { 506 stats->tx_errors++; 507 stats->tx_aborted_errors++; 508 } 509 510 return 0; 511 tx_err_link_failure: 512 stats->tx_carrier_errors++; 513 dst_link_failure(skb); 514 tx_err_dst_release: 515 dst_release(dst); 516 return err; 517 } 518 519 static netdev_tx_t 520 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 521 { 522 struct ip6_tnl *t = netdev_priv(dev); 523 struct net_device_stats *stats = &t->dev->stats; 524 struct ipv6hdr *ipv6h; 525 struct flowi fl; 526 int ret; 527 528 memset(&fl, 0, sizeof(fl)); 529 530 switch (skb->protocol) { 531 case htons(ETH_P_IPV6): 532 ipv6h = ipv6_hdr(skb); 533 534 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 535 vti6_addr_conflict(t, ipv6h)) 536 goto tx_err; 537 538 xfrm_decode_session(skb, &fl, AF_INET6); 539 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 540 break; 541 case htons(ETH_P_IP): 542 xfrm_decode_session(skb, &fl, AF_INET); 543 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 544 break; 545 default: 546 goto tx_err; 547 } 548 549 /* override mark with tunnel output key */ 550 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 551 552 ret = vti6_xmit(skb, dev, &fl); 553 if (ret < 0) 554 goto tx_err; 555 556 return NETDEV_TX_OK; 557 558 tx_err: 559 stats->tx_errors++; 560 stats->tx_dropped++; 561 kfree_skb(skb); 562 return NETDEV_TX_OK; 563 } 564 565 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 566 u8 type, u8 code, int offset, __be32 info) 567 { 568 __be32 spi; 569 __u32 mark; 570 struct xfrm_state *x; 571 struct ip6_tnl *t; 572 struct ip_esp_hdr *esph; 573 struct ip_auth_hdr *ah; 574 struct ip_comp_hdr *ipch; 575 struct net *net = dev_net(skb->dev); 576 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 577 int protocol = iph->nexthdr; 578 579 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 580 if (!t) 581 return -1; 582 583 mark = be32_to_cpu(t->parms.o_key); 584 585 switch (protocol) { 586 case IPPROTO_ESP: 587 esph = (struct ip_esp_hdr *)(skb->data + offset); 588 spi = esph->spi; 589 break; 590 case IPPROTO_AH: 591 ah = (struct ip_auth_hdr *)(skb->data + offset); 592 spi = ah->spi; 593 break; 594 case IPPROTO_COMP: 595 ipch = (struct ip_comp_hdr *)(skb->data + offset); 596 spi = htonl(ntohs(ipch->cpi)); 597 break; 598 default: 599 return 0; 600 } 601 602 if (type != ICMPV6_PKT_TOOBIG && 603 type != NDISC_REDIRECT) 604 return 0; 605 606 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 607 spi, protocol, AF_INET6); 608 if (!x) 609 return 0; 610 611 if (type == NDISC_REDIRECT) 612 ip6_redirect(skb, net, skb->dev->ifindex, 0, 613 sock_net_uid(net, NULL)); 614 else 615 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 616 xfrm_state_put(x); 617 618 return 0; 619 } 620 621 static void vti6_link_config(struct ip6_tnl *t) 622 { 623 struct net_device *dev = t->dev; 624 struct __ip6_tnl_parm *p = &t->parms; 625 626 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 627 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 628 629 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 630 IP6_TNL_F_CAP_PER_PACKET); 631 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 632 633 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 634 dev->flags |= IFF_POINTOPOINT; 635 else 636 dev->flags &= ~IFF_POINTOPOINT; 637 } 638 639 /** 640 * vti6_tnl_change - update the tunnel parameters 641 * @t: tunnel to be changed 642 * @p: tunnel configuration parameters 643 * 644 * Description: 645 * vti6_tnl_change() updates the tunnel parameters 646 **/ 647 static int 648 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 649 { 650 t->parms.laddr = p->laddr; 651 t->parms.raddr = p->raddr; 652 t->parms.link = p->link; 653 t->parms.i_key = p->i_key; 654 t->parms.o_key = p->o_key; 655 t->parms.proto = p->proto; 656 dst_cache_reset(&t->dst_cache); 657 vti6_link_config(t); 658 return 0; 659 } 660 661 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 662 { 663 struct net *net = dev_net(t->dev); 664 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 665 int err; 666 667 vti6_tnl_unlink(ip6n, t); 668 synchronize_net(); 669 err = vti6_tnl_change(t, p); 670 vti6_tnl_link(ip6n, t); 671 netdev_state_change(t->dev); 672 return err; 673 } 674 675 static void 676 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 677 { 678 p->laddr = u->laddr; 679 p->raddr = u->raddr; 680 p->link = u->link; 681 p->i_key = u->i_key; 682 p->o_key = u->o_key; 683 p->proto = u->proto; 684 685 memcpy(p->name, u->name, sizeof(u->name)); 686 } 687 688 static void 689 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 690 { 691 u->laddr = p->laddr; 692 u->raddr = p->raddr; 693 u->link = p->link; 694 u->i_key = p->i_key; 695 u->o_key = p->o_key; 696 u->proto = p->proto; 697 698 memcpy(u->name, p->name, sizeof(u->name)); 699 } 700 701 /** 702 * vti6_tnl_ioctl - configure vti6 tunnels from userspace 703 * @dev: virtual device associated with tunnel 704 * @ifr: parameters passed from userspace 705 * @cmd: command to be performed 706 * 707 * Description: 708 * vti6_ioctl() is used for managing vti6 tunnels 709 * from userspace. 710 * 711 * The possible commands are the following: 712 * %SIOCGETTUNNEL: get tunnel parameters for device 713 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 714 * %SIOCCHGTUNNEL: change tunnel parameters to those given 715 * %SIOCDELTUNNEL: delete tunnel 716 * 717 * The fallback device "ip6_vti0", created during module 718 * initialization, can be used for creating other tunnel devices. 719 * 720 * Return: 721 * 0 on success, 722 * %-EFAULT if unable to copy data to or from userspace, 723 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 724 * %-EINVAL if passed tunnel parameters are invalid, 725 * %-EEXIST if changing a tunnel's parameters would cause a conflict 726 * %-ENODEV if attempting to change or delete a nonexisting device 727 **/ 728 static int 729 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 730 { 731 int err = 0; 732 struct ip6_tnl_parm2 p; 733 struct __ip6_tnl_parm p1; 734 struct ip6_tnl *t = NULL; 735 struct net *net = dev_net(dev); 736 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 737 738 switch (cmd) { 739 case SIOCGETTUNNEL: 740 if (dev == ip6n->fb_tnl_dev) { 741 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 742 err = -EFAULT; 743 break; 744 } 745 vti6_parm_from_user(&p1, &p); 746 t = vti6_locate(net, &p1, 0); 747 } else { 748 memset(&p, 0, sizeof(p)); 749 } 750 if (!t) 751 t = netdev_priv(dev); 752 vti6_parm_to_user(&p, &t->parms); 753 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 754 err = -EFAULT; 755 break; 756 case SIOCADDTUNNEL: 757 case SIOCCHGTUNNEL: 758 err = -EPERM; 759 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 760 break; 761 err = -EFAULT; 762 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 763 break; 764 err = -EINVAL; 765 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 766 break; 767 vti6_parm_from_user(&p1, &p); 768 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 769 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 770 if (t) { 771 if (t->dev != dev) { 772 err = -EEXIST; 773 break; 774 } 775 } else 776 t = netdev_priv(dev); 777 778 err = vti6_update(t, &p1); 779 } 780 if (t) { 781 err = 0; 782 vti6_parm_to_user(&p, &t->parms); 783 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 784 err = -EFAULT; 785 786 } else 787 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 788 break; 789 case SIOCDELTUNNEL: 790 err = -EPERM; 791 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 792 break; 793 794 if (dev == ip6n->fb_tnl_dev) { 795 err = -EFAULT; 796 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 797 break; 798 err = -ENOENT; 799 vti6_parm_from_user(&p1, &p); 800 t = vti6_locate(net, &p1, 0); 801 if (!t) 802 break; 803 err = -EPERM; 804 if (t->dev == ip6n->fb_tnl_dev) 805 break; 806 dev = t->dev; 807 } 808 err = 0; 809 unregister_netdevice(dev); 810 break; 811 default: 812 err = -EINVAL; 813 } 814 return err; 815 } 816 817 static const struct net_device_ops vti6_netdev_ops = { 818 .ndo_init = vti6_dev_init, 819 .ndo_uninit = vti6_dev_uninit, 820 .ndo_start_xmit = vti6_tnl_xmit, 821 .ndo_do_ioctl = vti6_ioctl, 822 .ndo_get_stats64 = ip_tunnel_get_stats64, 823 .ndo_get_iflink = ip6_tnl_get_iflink, 824 }; 825 826 /** 827 * vti6_dev_setup - setup virtual tunnel device 828 * @dev: virtual device associated with tunnel 829 * 830 * Description: 831 * Initialize function pointers and device parameters 832 **/ 833 static void vti6_dev_setup(struct net_device *dev) 834 { 835 dev->netdev_ops = &vti6_netdev_ops; 836 dev->destructor = vti6_dev_free; 837 838 dev->type = ARPHRD_TUNNEL6; 839 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr); 840 dev->mtu = ETH_DATA_LEN; 841 dev->min_mtu = IPV6_MIN_MTU; 842 dev->max_mtu = IP_MAX_MTU; 843 dev->flags |= IFF_NOARP; 844 dev->addr_len = sizeof(struct in6_addr); 845 netif_keep_dst(dev); 846 /* This perm addr will be used as interface identifier by IPv6 */ 847 dev->addr_assign_type = NET_ADDR_RANDOM; 848 eth_random_addr(dev->perm_addr); 849 } 850 851 /** 852 * vti6_dev_init_gen - general initializer for all tunnel devices 853 * @dev: virtual device associated with tunnel 854 **/ 855 static inline int vti6_dev_init_gen(struct net_device *dev) 856 { 857 struct ip6_tnl *t = netdev_priv(dev); 858 859 t->dev = dev; 860 t->net = dev_net(dev); 861 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 862 if (!dev->tstats) 863 return -ENOMEM; 864 return 0; 865 } 866 867 /** 868 * vti6_dev_init - initializer for all non fallback tunnel devices 869 * @dev: virtual device associated with tunnel 870 **/ 871 static int vti6_dev_init(struct net_device *dev) 872 { 873 struct ip6_tnl *t = netdev_priv(dev); 874 int err = vti6_dev_init_gen(dev); 875 876 if (err) 877 return err; 878 vti6_link_config(t); 879 return 0; 880 } 881 882 /** 883 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 884 * @dev: fallback device 885 * 886 * Return: 0 887 **/ 888 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 889 { 890 struct ip6_tnl *t = netdev_priv(dev); 891 struct net *net = dev_net(dev); 892 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 893 894 t->parms.proto = IPPROTO_IPV6; 895 dev_hold(dev); 896 897 rcu_assign_pointer(ip6n->tnls_wc[0], t); 898 return 0; 899 } 900 901 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[]) 902 { 903 return 0; 904 } 905 906 static void vti6_netlink_parms(struct nlattr *data[], 907 struct __ip6_tnl_parm *parms) 908 { 909 memset(parms, 0, sizeof(*parms)); 910 911 if (!data) 912 return; 913 914 if (data[IFLA_VTI_LINK]) 915 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 916 917 if (data[IFLA_VTI_LOCAL]) 918 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 919 920 if (data[IFLA_VTI_REMOTE]) 921 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 922 923 if (data[IFLA_VTI_IKEY]) 924 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 925 926 if (data[IFLA_VTI_OKEY]) 927 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 928 } 929 930 static int vti6_newlink(struct net *src_net, struct net_device *dev, 931 struct nlattr *tb[], struct nlattr *data[]) 932 { 933 struct net *net = dev_net(dev); 934 struct ip6_tnl *nt; 935 936 nt = netdev_priv(dev); 937 vti6_netlink_parms(data, &nt->parms); 938 939 nt->parms.proto = IPPROTO_IPV6; 940 941 if (vti6_locate(net, &nt->parms, 0)) 942 return -EEXIST; 943 944 return vti6_tnl_create2(dev); 945 } 946 947 static void vti6_dellink(struct net_device *dev, struct list_head *head) 948 { 949 struct net *net = dev_net(dev); 950 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 951 952 if (dev != ip6n->fb_tnl_dev) 953 unregister_netdevice_queue(dev, head); 954 } 955 956 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 957 struct nlattr *data[]) 958 { 959 struct ip6_tnl *t; 960 struct __ip6_tnl_parm p; 961 struct net *net = dev_net(dev); 962 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 963 964 if (dev == ip6n->fb_tnl_dev) 965 return -EINVAL; 966 967 vti6_netlink_parms(data, &p); 968 969 t = vti6_locate(net, &p, 0); 970 971 if (t) { 972 if (t->dev != dev) 973 return -EEXIST; 974 } else 975 t = netdev_priv(dev); 976 977 return vti6_update(t, &p); 978 } 979 980 static size_t vti6_get_size(const struct net_device *dev) 981 { 982 return 983 /* IFLA_VTI_LINK */ 984 nla_total_size(4) + 985 /* IFLA_VTI_LOCAL */ 986 nla_total_size(sizeof(struct in6_addr)) + 987 /* IFLA_VTI_REMOTE */ 988 nla_total_size(sizeof(struct in6_addr)) + 989 /* IFLA_VTI_IKEY */ 990 nla_total_size(4) + 991 /* IFLA_VTI_OKEY */ 992 nla_total_size(4) + 993 0; 994 } 995 996 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 997 { 998 struct ip6_tnl *tunnel = netdev_priv(dev); 999 struct __ip6_tnl_parm *parm = &tunnel->parms; 1000 1001 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1002 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1003 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1004 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1005 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key)) 1006 goto nla_put_failure; 1007 return 0; 1008 1009 nla_put_failure: 1010 return -EMSGSIZE; 1011 } 1012 1013 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1014 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1015 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1016 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1017 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1018 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1019 }; 1020 1021 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1022 .kind = "vti6", 1023 .maxtype = IFLA_VTI_MAX, 1024 .policy = vti6_policy, 1025 .priv_size = sizeof(struct ip6_tnl), 1026 .setup = vti6_dev_setup, 1027 .validate = vti6_validate, 1028 .newlink = vti6_newlink, 1029 .dellink = vti6_dellink, 1030 .changelink = vti6_changelink, 1031 .get_size = vti6_get_size, 1032 .fill_info = vti6_fill_info, 1033 .get_link_net = ip6_tnl_get_link_net, 1034 }; 1035 1036 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n) 1037 { 1038 int h; 1039 struct ip6_tnl *t; 1040 LIST_HEAD(list); 1041 1042 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1043 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1044 while (t) { 1045 unregister_netdevice_queue(t->dev, &list); 1046 t = rtnl_dereference(t->next); 1047 } 1048 } 1049 1050 t = rtnl_dereference(ip6n->tnls_wc[0]); 1051 unregister_netdevice_queue(t->dev, &list); 1052 unregister_netdevice_many(&list); 1053 } 1054 1055 static int __net_init vti6_init_net(struct net *net) 1056 { 1057 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1058 struct ip6_tnl *t = NULL; 1059 int err; 1060 1061 ip6n->tnls[0] = ip6n->tnls_wc; 1062 ip6n->tnls[1] = ip6n->tnls_r_l; 1063 1064 err = -ENOMEM; 1065 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1066 NET_NAME_UNKNOWN, vti6_dev_setup); 1067 1068 if (!ip6n->fb_tnl_dev) 1069 goto err_alloc_dev; 1070 dev_net_set(ip6n->fb_tnl_dev, net); 1071 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1072 1073 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1074 if (err < 0) 1075 goto err_register; 1076 1077 err = register_netdev(ip6n->fb_tnl_dev); 1078 if (err < 0) 1079 goto err_register; 1080 1081 t = netdev_priv(ip6n->fb_tnl_dev); 1082 1083 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1084 return 0; 1085 1086 err_register: 1087 vti6_dev_free(ip6n->fb_tnl_dev); 1088 err_alloc_dev: 1089 return err; 1090 } 1091 1092 static void __net_exit vti6_exit_net(struct net *net) 1093 { 1094 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1095 1096 rtnl_lock(); 1097 vti6_destroy_tunnels(ip6n); 1098 rtnl_unlock(); 1099 } 1100 1101 static struct pernet_operations vti6_net_ops = { 1102 .init = vti6_init_net, 1103 .exit = vti6_exit_net, 1104 .id = &vti6_net_id, 1105 .size = sizeof(struct vti6_net), 1106 }; 1107 1108 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1109 .handler = vti6_rcv, 1110 .cb_handler = vti6_rcv_cb, 1111 .err_handler = vti6_err, 1112 .priority = 100, 1113 }; 1114 1115 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1116 .handler = vti6_rcv, 1117 .cb_handler = vti6_rcv_cb, 1118 .err_handler = vti6_err, 1119 .priority = 100, 1120 }; 1121 1122 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1123 .handler = vti6_rcv, 1124 .cb_handler = vti6_rcv_cb, 1125 .err_handler = vti6_err, 1126 .priority = 100, 1127 }; 1128 1129 static bool is_vti6_tunnel(const struct net_device *dev) 1130 { 1131 return dev->netdev_ops == &vti6_netdev_ops; 1132 } 1133 1134 static int vti6_device_event(struct notifier_block *unused, 1135 unsigned long event, void *ptr) 1136 { 1137 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1138 struct ip6_tnl *t = netdev_priv(dev); 1139 1140 if (!is_vti6_tunnel(dev)) 1141 return NOTIFY_DONE; 1142 1143 switch (event) { 1144 case NETDEV_DOWN: 1145 if (!net_eq(t->net, dev_net(dev))) 1146 xfrm_garbage_collect(t->net); 1147 break; 1148 } 1149 return NOTIFY_DONE; 1150 } 1151 1152 static struct notifier_block vti6_notifier_block __read_mostly = { 1153 .notifier_call = vti6_device_event, 1154 }; 1155 1156 /** 1157 * vti6_tunnel_init - register protocol and reserve needed resources 1158 * 1159 * Return: 0 on success 1160 **/ 1161 static int __init vti6_tunnel_init(void) 1162 { 1163 const char *msg; 1164 int err; 1165 1166 register_netdevice_notifier(&vti6_notifier_block); 1167 1168 msg = "tunnel device"; 1169 err = register_pernet_device(&vti6_net_ops); 1170 if (err < 0) 1171 goto pernet_dev_failed; 1172 1173 msg = "tunnel protocols"; 1174 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1175 if (err < 0) 1176 goto xfrm_proto_esp_failed; 1177 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1178 if (err < 0) 1179 goto xfrm_proto_ah_failed; 1180 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1181 if (err < 0) 1182 goto xfrm_proto_comp_failed; 1183 1184 msg = "netlink interface"; 1185 err = rtnl_link_register(&vti6_link_ops); 1186 if (err < 0) 1187 goto rtnl_link_failed; 1188 1189 return 0; 1190 1191 rtnl_link_failed: 1192 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1193 xfrm_proto_comp_failed: 1194 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1195 xfrm_proto_ah_failed: 1196 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1197 xfrm_proto_esp_failed: 1198 unregister_pernet_device(&vti6_net_ops); 1199 pernet_dev_failed: 1200 unregister_netdevice_notifier(&vti6_notifier_block); 1201 pr_err("vti6 init: failed to register %s\n", msg); 1202 return err; 1203 } 1204 1205 /** 1206 * vti6_tunnel_cleanup - free resources and unregister protocol 1207 **/ 1208 static void __exit vti6_tunnel_cleanup(void) 1209 { 1210 rtnl_link_unregister(&vti6_link_ops); 1211 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1212 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1213 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1214 unregister_pernet_device(&vti6_net_ops); 1215 unregister_netdevice_notifier(&vti6_notifier_block); 1216 } 1217 1218 module_init(vti6_tunnel_init); 1219 module_exit(vti6_tunnel_cleanup); 1220 MODULE_LICENSE("GPL"); 1221 MODULE_ALIAS_RTNL_LINK("vti6"); 1222 MODULE_ALIAS_NETDEV("ip6_vti0"); 1223 MODULE_AUTHOR("Steffen Klassert"); 1224 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1225