1 /* 2 * IPv6 tunneling device 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Ville Nuorvala <vnuorval@tcs.hut.fi> 7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org> 8 * 9 * Based on: 10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c 11 * 12 * RFC 2473 13 * 14 * This program is free software; you can redistribute it and/or 15 * modify it under the terms of the GNU General Public License 16 * as published by the Free Software Foundation; either version 17 * 2 of the License, or (at your option) any later version. 18 * 19 */ 20 21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 22 23 #include <linux/module.h> 24 #include <linux/capability.h> 25 #include <linux/errno.h> 26 #include <linux/types.h> 27 #include <linux/sockios.h> 28 #include <linux/icmp.h> 29 #include <linux/if.h> 30 #include <linux/in.h> 31 #include <linux/ip.h> 32 #include <linux/net.h> 33 #include <linux/in6.h> 34 #include <linux/netdevice.h> 35 #include <linux/if_arp.h> 36 #include <linux/icmpv6.h> 37 #include <linux/init.h> 38 #include <linux/route.h> 39 #include <linux/rtnetlink.h> 40 #include <linux/netfilter_ipv6.h> 41 #include <linux/slab.h> 42 #include <linux/hash.h> 43 #include <linux/etherdevice.h> 44 45 #include <linux/uaccess.h> 46 #include <linux/atomic.h> 47 48 #include <net/icmp.h> 49 #include <net/ip.h> 50 #include <net/ip_tunnels.h> 51 #include <net/ipv6.h> 52 #include <net/ip6_route.h> 53 #include <net/addrconf.h> 54 #include <net/ip6_tunnel.h> 55 #include <net/xfrm.h> 56 #include <net/dsfield.h> 57 #include <net/inet_ecn.h> 58 #include <net/net_namespace.h> 59 #include <net/netns/generic.h> 60 #include <net/dst_metadata.h> 61 62 MODULE_AUTHOR("Ville Nuorvala"); 63 MODULE_DESCRIPTION("IPv6 tunneling device"); 64 MODULE_LICENSE("GPL"); 65 MODULE_ALIAS_RTNL_LINK("ip6tnl"); 66 MODULE_ALIAS_NETDEV("ip6tnl0"); 67 68 #define IP6_TUNNEL_HASH_SIZE_SHIFT 5 69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT) 70 71 static bool log_ecn_error = true; 72 module_param(log_ecn_error, bool, 0644); 73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 74 75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 76 { 77 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 78 79 return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT); 80 } 81 82 static int ip6_tnl_dev_init(struct net_device *dev); 83 static void ip6_tnl_dev_setup(struct net_device *dev); 84 static struct rtnl_link_ops ip6_link_ops __read_mostly; 85 86 static unsigned int ip6_tnl_net_id __read_mostly; 87 struct ip6_tnl_net { 88 /* the IPv6 tunnel fallback device */ 89 struct net_device *fb_tnl_dev; 90 /* lists for storing tunnels in use */ 91 struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE]; 92 struct ip6_tnl __rcu *tnls_wc[1]; 93 struct ip6_tnl __rcu **tnls[2]; 94 struct ip6_tnl __rcu *collect_md_tun; 95 }; 96 97 static struct net_device_stats *ip6_get_stats(struct net_device *dev) 98 { 99 struct pcpu_sw_netstats tmp, sum = { 0 }; 100 int i; 101 102 for_each_possible_cpu(i) { 103 unsigned int start; 104 const struct pcpu_sw_netstats *tstats = 105 per_cpu_ptr(dev->tstats, i); 106 107 do { 108 start = u64_stats_fetch_begin_irq(&tstats->syncp); 109 tmp.rx_packets = tstats->rx_packets; 110 tmp.rx_bytes = tstats->rx_bytes; 111 tmp.tx_packets = tstats->tx_packets; 112 tmp.tx_bytes = tstats->tx_bytes; 113 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start)); 114 115 sum.rx_packets += tmp.rx_packets; 116 sum.rx_bytes += tmp.rx_bytes; 117 sum.tx_packets += tmp.tx_packets; 118 sum.tx_bytes += tmp.tx_bytes; 119 } 120 dev->stats.rx_packets = sum.rx_packets; 121 dev->stats.rx_bytes = sum.rx_bytes; 122 dev->stats.tx_packets = sum.tx_packets; 123 dev->stats.tx_bytes = sum.tx_bytes; 124 return &dev->stats; 125 } 126 127 /** 128 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 129 * @remote: the address of the tunnel exit-point 130 * @local: the address of the tunnel entry-point 131 * 132 * Return: 133 * tunnel matching given end-points if found, 134 * else fallback tunnel if its device is up, 135 * else %NULL 136 **/ 137 138 #define for_each_ip6_tunnel_rcu(start) \ 139 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 140 141 static struct ip6_tnl * 142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local) 143 { 144 unsigned int hash = HASH(remote, local); 145 struct ip6_tnl *t; 146 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 147 struct in6_addr any; 148 149 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 150 if (ipv6_addr_equal(local, &t->parms.laddr) && 151 ipv6_addr_equal(remote, &t->parms.raddr) && 152 (t->dev->flags & IFF_UP)) 153 return t; 154 } 155 156 memset(&any, 0, sizeof(any)); 157 hash = HASH(&any, local); 158 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 159 if (ipv6_addr_equal(local, &t->parms.laddr) && 160 ipv6_addr_any(&t->parms.raddr) && 161 (t->dev->flags & IFF_UP)) 162 return t; 163 } 164 165 hash = HASH(remote, &any); 166 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 167 if (ipv6_addr_equal(remote, &t->parms.raddr) && 168 ipv6_addr_any(&t->parms.laddr) && 169 (t->dev->flags & IFF_UP)) 170 return t; 171 } 172 173 t = rcu_dereference(ip6n->collect_md_tun); 174 if (t) 175 return t; 176 177 t = rcu_dereference(ip6n->tnls_wc[0]); 178 if (t && (t->dev->flags & IFF_UP)) 179 return t; 180 181 return NULL; 182 } 183 184 /** 185 * ip6_tnl_bucket - get head of list matching given tunnel parameters 186 * @p: parameters containing tunnel end-points 187 * 188 * Description: 189 * ip6_tnl_bucket() returns the head of the list matching the 190 * &struct in6_addr entries laddr and raddr in @p. 191 * 192 * Return: head of IPv6 tunnel list 193 **/ 194 195 static struct ip6_tnl __rcu ** 196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 197 { 198 const struct in6_addr *remote = &p->raddr; 199 const struct in6_addr *local = &p->laddr; 200 unsigned int h = 0; 201 int prio = 0; 202 203 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 204 prio = 1; 205 h = HASH(remote, local); 206 } 207 return &ip6n->tnls[prio][h]; 208 } 209 210 /** 211 * ip6_tnl_link - add tunnel to hash table 212 * @t: tunnel to be added 213 **/ 214 215 static void 216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 217 { 218 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 219 220 if (t->parms.collect_md) 221 rcu_assign_pointer(ip6n->collect_md_tun, t); 222 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 223 rcu_assign_pointer(*tp, t); 224 } 225 226 /** 227 * ip6_tnl_unlink - remove tunnel from hash table 228 * @t: tunnel to be removed 229 **/ 230 231 static void 232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 233 { 234 struct ip6_tnl __rcu **tp; 235 struct ip6_tnl *iter; 236 237 if (t->parms.collect_md) 238 rcu_assign_pointer(ip6n->collect_md_tun, NULL); 239 240 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 241 (iter = rtnl_dereference(*tp)) != NULL; 242 tp = &iter->next) { 243 if (t == iter) { 244 rcu_assign_pointer(*tp, t->next); 245 break; 246 } 247 } 248 } 249 250 static void ip6_dev_free(struct net_device *dev) 251 { 252 struct ip6_tnl *t = netdev_priv(dev); 253 254 gro_cells_destroy(&t->gro_cells); 255 dst_cache_destroy(&t->dst_cache); 256 free_percpu(dev->tstats); 257 free_netdev(dev); 258 } 259 260 static int ip6_tnl_create2(struct net_device *dev) 261 { 262 struct ip6_tnl *t = netdev_priv(dev); 263 struct net *net = dev_net(dev); 264 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 265 int err; 266 267 t = netdev_priv(dev); 268 269 dev->rtnl_link_ops = &ip6_link_ops; 270 err = register_netdevice(dev); 271 if (err < 0) 272 goto out; 273 274 strcpy(t->parms.name, dev->name); 275 276 dev_hold(dev); 277 ip6_tnl_link(ip6n, t); 278 return 0; 279 280 out: 281 return err; 282 } 283 284 /** 285 * ip6_tnl_create - create a new tunnel 286 * @p: tunnel parameters 287 * @pt: pointer to new tunnel 288 * 289 * Description: 290 * Create tunnel matching given parameters. 291 * 292 * Return: 293 * created tunnel or error pointer 294 **/ 295 296 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 297 { 298 struct net_device *dev; 299 struct ip6_tnl *t; 300 char name[IFNAMSIZ]; 301 int err = -ENOMEM; 302 303 if (p->name[0]) 304 strlcpy(name, p->name, IFNAMSIZ); 305 else 306 sprintf(name, "ip6tnl%%d"); 307 308 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 309 ip6_tnl_dev_setup); 310 if (!dev) 311 goto failed; 312 313 dev_net_set(dev, net); 314 315 t = netdev_priv(dev); 316 t->parms = *p; 317 t->net = dev_net(dev); 318 err = ip6_tnl_create2(dev); 319 if (err < 0) 320 goto failed_free; 321 322 return t; 323 324 failed_free: 325 ip6_dev_free(dev); 326 failed: 327 return ERR_PTR(err); 328 } 329 330 /** 331 * ip6_tnl_locate - find or create tunnel matching given parameters 332 * @p: tunnel parameters 333 * @create: != 0 if allowed to create new tunnel if no match found 334 * 335 * Description: 336 * ip6_tnl_locate() first tries to locate an existing tunnel 337 * based on @parms. If this is unsuccessful, but @create is set a new 338 * tunnel device is created and registered for use. 339 * 340 * Return: 341 * matching tunnel or error pointer 342 **/ 343 344 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 345 struct __ip6_tnl_parm *p, int create) 346 { 347 const struct in6_addr *remote = &p->raddr; 348 const struct in6_addr *local = &p->laddr; 349 struct ip6_tnl __rcu **tp; 350 struct ip6_tnl *t; 351 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 352 353 for (tp = ip6_tnl_bucket(ip6n, p); 354 (t = rtnl_dereference(*tp)) != NULL; 355 tp = &t->next) { 356 if (ipv6_addr_equal(local, &t->parms.laddr) && 357 ipv6_addr_equal(remote, &t->parms.raddr)) { 358 if (create) 359 return ERR_PTR(-EEXIST); 360 361 return t; 362 } 363 } 364 if (!create) 365 return ERR_PTR(-ENODEV); 366 return ip6_tnl_create(net, p); 367 } 368 369 /** 370 * ip6_tnl_dev_uninit - tunnel device uninitializer 371 * @dev: the device to be destroyed 372 * 373 * Description: 374 * ip6_tnl_dev_uninit() removes tunnel from its list 375 **/ 376 377 static void 378 ip6_tnl_dev_uninit(struct net_device *dev) 379 { 380 struct ip6_tnl *t = netdev_priv(dev); 381 struct net *net = t->net; 382 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 383 384 if (dev == ip6n->fb_tnl_dev) 385 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 386 else 387 ip6_tnl_unlink(ip6n, t); 388 dst_cache_reset(&t->dst_cache); 389 dev_put(dev); 390 } 391 392 /** 393 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 394 * @skb: received socket buffer 395 * 396 * Return: 397 * 0 if none was found, 398 * else index to encapsulation limit 399 **/ 400 401 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 402 { 403 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw; 404 unsigned int nhoff = raw - skb->data; 405 unsigned int off = nhoff + sizeof(*ipv6h); 406 u8 next, nexthdr = ipv6h->nexthdr; 407 408 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 409 struct ipv6_opt_hdr *hdr; 410 u16 optlen; 411 412 if (!pskb_may_pull(skb, off + sizeof(*hdr))) 413 break; 414 415 hdr = (struct ipv6_opt_hdr *)(skb->data + off); 416 if (nexthdr == NEXTHDR_FRAGMENT) { 417 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 418 if (frag_hdr->frag_off) 419 break; 420 optlen = 8; 421 } else if (nexthdr == NEXTHDR_AUTH) { 422 optlen = (hdr->hdrlen + 2) << 2; 423 } else { 424 optlen = ipv6_optlen(hdr); 425 } 426 /* cache hdr->nexthdr, since pskb_may_pull() might 427 * invalidate hdr 428 */ 429 next = hdr->nexthdr; 430 if (nexthdr == NEXTHDR_DEST) { 431 u16 i = 2; 432 433 /* Remember : hdr is no longer valid at this point. */ 434 if (!pskb_may_pull(skb, off + optlen)) 435 break; 436 437 while (1) { 438 struct ipv6_tlv_tnl_enc_lim *tel; 439 440 /* No more room for encapsulation limit */ 441 if (i + sizeof(*tel) > optlen) 442 break; 443 444 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i); 445 /* return index of option if found and valid */ 446 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 447 tel->length == 1) 448 return i + off - nhoff; 449 /* else jump to next option */ 450 if (tel->type) 451 i += tel->length + 2; 452 else 453 i++; 454 } 455 } 456 nexthdr = next; 457 off += optlen; 458 } 459 return 0; 460 } 461 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 462 463 /** 464 * ip6_tnl_err - tunnel error handler 465 * 466 * Description: 467 * ip6_tnl_err() should handle errors in the tunnel according 468 * to the specifications in RFC 2473. 469 **/ 470 471 static int 472 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 473 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 474 { 475 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data; 476 struct ip6_tnl *t; 477 int rel_msg = 0; 478 u8 rel_type = ICMPV6_DEST_UNREACH; 479 u8 rel_code = ICMPV6_ADDR_UNREACH; 480 u8 tproto; 481 __u32 rel_info = 0; 482 __u16 len; 483 int err = -ENOENT; 484 485 /* If the packet doesn't contain the original IPv6 header we are 486 in trouble since we might need the source address for further 487 processing of the error. */ 488 489 rcu_read_lock(); 490 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr); 491 if (!t) 492 goto out; 493 494 tproto = ACCESS_ONCE(t->parms.proto); 495 if (tproto != ipproto && tproto != 0) 496 goto out; 497 498 err = 0; 499 500 switch (*type) { 501 __u32 teli; 502 struct ipv6_tlv_tnl_enc_lim *tel; 503 __u32 mtu; 504 case ICMPV6_DEST_UNREACH: 505 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 506 t->parms.name); 507 rel_msg = 1; 508 break; 509 case ICMPV6_TIME_EXCEED: 510 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 511 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 512 t->parms.name); 513 rel_msg = 1; 514 } 515 break; 516 case ICMPV6_PARAMPROB: 517 teli = 0; 518 if ((*code) == ICMPV6_HDR_FIELD) 519 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 520 521 if (teli && teli == *info - 2) { 522 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 523 if (tel->encap_limit == 0) { 524 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 525 t->parms.name); 526 rel_msg = 1; 527 } 528 } else { 529 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 530 t->parms.name); 531 } 532 break; 533 case ICMPV6_PKT_TOOBIG: 534 mtu = *info - offset; 535 if (mtu < IPV6_MIN_MTU) 536 mtu = IPV6_MIN_MTU; 537 t->dev->mtu = mtu; 538 539 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len); 540 if (len > mtu) { 541 rel_type = ICMPV6_PKT_TOOBIG; 542 rel_code = 0; 543 rel_info = mtu; 544 rel_msg = 1; 545 } 546 break; 547 } 548 549 *type = rel_type; 550 *code = rel_code; 551 *info = rel_info; 552 *msg = rel_msg; 553 554 out: 555 rcu_read_unlock(); 556 return err; 557 } 558 559 static int 560 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 561 u8 type, u8 code, int offset, __be32 info) 562 { 563 int rel_msg = 0; 564 u8 rel_type = type; 565 u8 rel_code = code; 566 __u32 rel_info = ntohl(info); 567 int err; 568 struct sk_buff *skb2; 569 const struct iphdr *eiph; 570 struct rtable *rt; 571 struct flowi4 fl4; 572 573 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 574 &rel_msg, &rel_info, offset); 575 if (err < 0) 576 return err; 577 578 if (rel_msg == 0) 579 return 0; 580 581 switch (rel_type) { 582 case ICMPV6_DEST_UNREACH: 583 if (rel_code != ICMPV6_ADDR_UNREACH) 584 return 0; 585 rel_type = ICMP_DEST_UNREACH; 586 rel_code = ICMP_HOST_UNREACH; 587 break; 588 case ICMPV6_PKT_TOOBIG: 589 if (rel_code != 0) 590 return 0; 591 rel_type = ICMP_DEST_UNREACH; 592 rel_code = ICMP_FRAG_NEEDED; 593 break; 594 case NDISC_REDIRECT: 595 rel_type = ICMP_REDIRECT; 596 rel_code = ICMP_REDIR_HOST; 597 default: 598 return 0; 599 } 600 601 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 602 return 0; 603 604 skb2 = skb_clone(skb, GFP_ATOMIC); 605 if (!skb2) 606 return 0; 607 608 skb_dst_drop(skb2); 609 610 skb_pull(skb2, offset); 611 skb_reset_network_header(skb2); 612 eiph = ip_hdr(skb2); 613 614 /* Try to guess incoming interface */ 615 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 616 eiph->saddr, 0, 617 0, 0, 618 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 619 if (IS_ERR(rt)) 620 goto out; 621 622 skb2->dev = rt->dst.dev; 623 624 /* route "incoming" packet */ 625 if (rt->rt_flags & RTCF_LOCAL) { 626 ip_rt_put(rt); 627 rt = NULL; 628 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 629 eiph->daddr, eiph->saddr, 630 0, 0, 631 IPPROTO_IPIP, 632 RT_TOS(eiph->tos), 0); 633 if (IS_ERR(rt) || 634 rt->dst.dev->type != ARPHRD_TUNNEL) { 635 if (!IS_ERR(rt)) 636 ip_rt_put(rt); 637 goto out; 638 } 639 skb_dst_set(skb2, &rt->dst); 640 } else { 641 ip_rt_put(rt); 642 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 643 skb2->dev) || 644 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 645 goto out; 646 } 647 648 /* change mtu on this route */ 649 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 650 if (rel_info > dst_mtu(skb_dst(skb2))) 651 goto out; 652 653 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info); 654 } 655 if (rel_type == ICMP_REDIRECT) 656 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2); 657 658 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 659 660 out: 661 kfree_skb(skb2); 662 return 0; 663 } 664 665 static int 666 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 667 u8 type, u8 code, int offset, __be32 info) 668 { 669 int rel_msg = 0; 670 u8 rel_type = type; 671 u8 rel_code = code; 672 __u32 rel_info = ntohl(info); 673 int err; 674 675 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 676 &rel_msg, &rel_info, offset); 677 if (err < 0) 678 return err; 679 680 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 681 struct rt6_info *rt; 682 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 683 684 if (!skb2) 685 return 0; 686 687 skb_dst_drop(skb2); 688 skb_pull(skb2, offset); 689 skb_reset_network_header(skb2); 690 691 /* Try to guess incoming interface */ 692 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 693 NULL, 0, 0); 694 695 if (rt && rt->dst.dev) 696 skb2->dev = rt->dst.dev; 697 698 icmpv6_send(skb2, rel_type, rel_code, rel_info); 699 700 ip6_rt_put(rt); 701 702 kfree_skb(skb2); 703 } 704 705 return 0; 706 } 707 708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 709 const struct ipv6hdr *ipv6h, 710 struct sk_buff *skb) 711 { 712 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 713 714 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 715 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 716 717 return IP6_ECN_decapsulate(ipv6h, skb); 718 } 719 720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 721 const struct ipv6hdr *ipv6h, 722 struct sk_buff *skb) 723 { 724 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 725 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 726 727 return IP6_ECN_decapsulate(ipv6h, skb); 728 } 729 730 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 731 const struct in6_addr *laddr, 732 const struct in6_addr *raddr) 733 { 734 struct __ip6_tnl_parm *p = &t->parms; 735 int ltype = ipv6_addr_type(laddr); 736 int rtype = ipv6_addr_type(raddr); 737 __u32 flags = 0; 738 739 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 740 flags = IP6_TNL_F_CAP_PER_PACKET; 741 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 742 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 743 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 744 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 745 if (ltype&IPV6_ADDR_UNICAST) 746 flags |= IP6_TNL_F_CAP_XMIT; 747 if (rtype&IPV6_ADDR_UNICAST) 748 flags |= IP6_TNL_F_CAP_RCV; 749 } 750 return flags; 751 } 752 EXPORT_SYMBOL(ip6_tnl_get_cap); 753 754 /* called with rcu_read_lock() */ 755 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 756 const struct in6_addr *laddr, 757 const struct in6_addr *raddr) 758 { 759 struct __ip6_tnl_parm *p = &t->parms; 760 int ret = 0; 761 struct net *net = t->net; 762 763 if ((p->flags & IP6_TNL_F_CAP_RCV) || 764 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 765 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 766 struct net_device *ldev = NULL; 767 768 if (p->link) 769 ldev = dev_get_by_index_rcu(net, p->link); 770 771 if ((ipv6_addr_is_multicast(laddr) || 772 likely(ipv6_chk_addr(net, laddr, ldev, 0))) && 773 likely(!ipv6_chk_addr(net, raddr, NULL, 0))) 774 ret = 1; 775 } 776 return ret; 777 } 778 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 779 780 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb, 781 const struct tnl_ptk_info *tpi, 782 struct metadata_dst *tun_dst, 783 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 784 const struct ipv6hdr *ipv6h, 785 struct sk_buff *skb), 786 bool log_ecn_err) 787 { 788 struct pcpu_sw_netstats *tstats; 789 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 790 int err; 791 792 if ((!(tpi->flags & TUNNEL_CSUM) && 793 (tunnel->parms.i_flags & TUNNEL_CSUM)) || 794 ((tpi->flags & TUNNEL_CSUM) && 795 !(tunnel->parms.i_flags & TUNNEL_CSUM))) { 796 tunnel->dev->stats.rx_crc_errors++; 797 tunnel->dev->stats.rx_errors++; 798 goto drop; 799 } 800 801 if (tunnel->parms.i_flags & TUNNEL_SEQ) { 802 if (!(tpi->flags & TUNNEL_SEQ) || 803 (tunnel->i_seqno && 804 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) { 805 tunnel->dev->stats.rx_fifo_errors++; 806 tunnel->dev->stats.rx_errors++; 807 goto drop; 808 } 809 tunnel->i_seqno = ntohl(tpi->seq) + 1; 810 } 811 812 skb->protocol = tpi->proto; 813 814 /* Warning: All skb pointers will be invalidated! */ 815 if (tunnel->dev->type == ARPHRD_ETHER) { 816 if (!pskb_may_pull(skb, ETH_HLEN)) { 817 tunnel->dev->stats.rx_length_errors++; 818 tunnel->dev->stats.rx_errors++; 819 goto drop; 820 } 821 822 ipv6h = ipv6_hdr(skb); 823 skb->protocol = eth_type_trans(skb, tunnel->dev); 824 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 825 } else { 826 skb->dev = tunnel->dev; 827 } 828 829 skb_reset_network_header(skb); 830 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 831 832 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 833 834 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb); 835 if (unlikely(err)) { 836 if (log_ecn_err) 837 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n", 838 &ipv6h->saddr, 839 ipv6_get_dsfield(ipv6h)); 840 if (err > 1) { 841 ++tunnel->dev->stats.rx_frame_errors; 842 ++tunnel->dev->stats.rx_errors; 843 goto drop; 844 } 845 } 846 847 tstats = this_cpu_ptr(tunnel->dev->tstats); 848 u64_stats_update_begin(&tstats->syncp); 849 tstats->rx_packets++; 850 tstats->rx_bytes += skb->len; 851 u64_stats_update_end(&tstats->syncp); 852 853 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev))); 854 855 if (tun_dst) 856 skb_dst_set(skb, (struct dst_entry *)tun_dst); 857 858 gro_cells_receive(&tunnel->gro_cells, skb); 859 return 0; 860 861 drop: 862 kfree_skb(skb); 863 return 0; 864 } 865 866 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb, 867 const struct tnl_ptk_info *tpi, 868 struct metadata_dst *tun_dst, 869 bool log_ecn_err) 870 { 871 return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate, 872 log_ecn_err); 873 } 874 EXPORT_SYMBOL(ip6_tnl_rcv); 875 876 static const struct tnl_ptk_info tpi_v6 = { 877 /* no tunnel info required for ipxip6. */ 878 .proto = htons(ETH_P_IPV6), 879 }; 880 881 static const struct tnl_ptk_info tpi_v4 = { 882 /* no tunnel info required for ipxip6. */ 883 .proto = htons(ETH_P_IP), 884 }; 885 886 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto, 887 const struct tnl_ptk_info *tpi, 888 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 889 const struct ipv6hdr *ipv6h, 890 struct sk_buff *skb)) 891 { 892 struct ip6_tnl *t; 893 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 894 struct metadata_dst *tun_dst = NULL; 895 int ret = -1; 896 897 rcu_read_lock(); 898 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 899 900 if (t) { 901 u8 tproto = ACCESS_ONCE(t->parms.proto); 902 903 if (tproto != ipproto && tproto != 0) 904 goto drop; 905 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 906 goto drop; 907 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) 908 goto drop; 909 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 910 goto drop; 911 if (t->parms.collect_md) { 912 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0); 913 if (!tun_dst) 914 return 0; 915 } 916 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 917 log_ecn_error); 918 } 919 920 rcu_read_unlock(); 921 922 return ret; 923 924 drop: 925 rcu_read_unlock(); 926 kfree_skb(skb); 927 return 0; 928 } 929 930 static int ip4ip6_rcv(struct sk_buff *skb) 931 { 932 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4, 933 ip4ip6_dscp_ecn_decapsulate); 934 } 935 936 static int ip6ip6_rcv(struct sk_buff *skb) 937 { 938 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6, 939 ip6ip6_dscp_ecn_decapsulate); 940 } 941 942 struct ipv6_tel_txoption { 943 struct ipv6_txoptions ops; 944 __u8 dst_opt[8]; 945 }; 946 947 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 948 { 949 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 950 951 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 952 opt->dst_opt[3] = 1; 953 opt->dst_opt[4] = encap_limit; 954 opt->dst_opt[5] = IPV6_TLV_PADN; 955 opt->dst_opt[6] = 1; 956 957 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt; 958 opt->ops.opt_nflen = 8; 959 } 960 961 /** 962 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 963 * @t: the outgoing tunnel device 964 * @hdr: IPv6 header from the incoming packet 965 * 966 * Description: 967 * Avoid trivial tunneling loop by checking that tunnel exit-point 968 * doesn't match source of incoming packet. 969 * 970 * Return: 971 * 1 if conflict, 972 * 0 else 973 **/ 974 975 static inline bool 976 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 977 { 978 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 979 } 980 981 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, 982 const struct in6_addr *laddr, 983 const struct in6_addr *raddr) 984 { 985 struct __ip6_tnl_parm *p = &t->parms; 986 int ret = 0; 987 struct net *net = t->net; 988 989 if ((p->flags & IP6_TNL_F_CAP_XMIT) || 990 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 991 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) { 992 struct net_device *ldev = NULL; 993 994 rcu_read_lock(); 995 if (p->link) 996 ldev = dev_get_by_index_rcu(net, p->link); 997 998 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0))) 999 pr_warn("%s xmit: Local address not yet configured!\n", 1000 p->name); 1001 else if (!ipv6_addr_is_multicast(raddr) && 1002 unlikely(ipv6_chk_addr(net, raddr, NULL, 0))) 1003 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 1004 p->name); 1005 else 1006 ret = 1; 1007 rcu_read_unlock(); 1008 } 1009 return ret; 1010 } 1011 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 1012 1013 /** 1014 * ip6_tnl_xmit - encapsulate packet and send 1015 * @skb: the outgoing socket buffer 1016 * @dev: the outgoing tunnel device 1017 * @dsfield: dscp code for outer header 1018 * @fl6: flow of tunneled packet 1019 * @encap_limit: encapsulation limit 1020 * @pmtu: Path MTU is stored if packet is too big 1021 * @proto: next header value 1022 * 1023 * Description: 1024 * Build new header and do some sanity checks on the packet before sending 1025 * it. 1026 * 1027 * Return: 1028 * 0 on success 1029 * -1 fail 1030 * %-EMSGSIZE message too big. return mtu in this case. 1031 **/ 1032 1033 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield, 1034 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, 1035 __u8 proto) 1036 { 1037 struct ip6_tnl *t = netdev_priv(dev); 1038 struct net *net = t->net; 1039 struct net_device_stats *stats = &t->dev->stats; 1040 struct ipv6hdr *ipv6h; 1041 struct ipv6_tel_txoption opt; 1042 struct dst_entry *dst = NULL, *ndst = NULL; 1043 struct net_device *tdev; 1044 int mtu; 1045 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen; 1046 unsigned int max_headroom = psh_hlen; 1047 bool use_cache = false; 1048 u8 hop_limit; 1049 int err = -1; 1050 1051 if (t->parms.collect_md) { 1052 hop_limit = skb_tunnel_info(skb)->key.ttl; 1053 goto route_lookup; 1054 } else { 1055 hop_limit = t->parms.hop_limit; 1056 } 1057 1058 /* NBMA tunnel */ 1059 if (ipv6_addr_any(&t->parms.raddr)) { 1060 if (skb->protocol == htons(ETH_P_IPV6)) { 1061 struct in6_addr *addr6; 1062 struct neighbour *neigh; 1063 int addr_type; 1064 1065 if (!skb_dst(skb)) 1066 goto tx_err_link_failure; 1067 1068 neigh = dst_neigh_lookup(skb_dst(skb), 1069 &ipv6_hdr(skb)->daddr); 1070 if (!neigh) 1071 goto tx_err_link_failure; 1072 1073 addr6 = (struct in6_addr *)&neigh->primary_key; 1074 addr_type = ipv6_addr_type(addr6); 1075 1076 if (addr_type == IPV6_ADDR_ANY) 1077 addr6 = &ipv6_hdr(skb)->daddr; 1078 1079 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr)); 1080 neigh_release(neigh); 1081 } 1082 } else if (!(t->parms.flags & 1083 (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) { 1084 /* enable the cache only only if the routing decision does 1085 * not depend on the current inner header value 1086 */ 1087 use_cache = true; 1088 } 1089 1090 if (use_cache) 1091 dst = dst_cache_get(&t->dst_cache); 1092 1093 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1094 goto tx_err_link_failure; 1095 1096 if (!dst) { 1097 route_lookup: 1098 /* add dsfield to flowlabel for route lookup */ 1099 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1100 1101 dst = ip6_route_output(net, NULL, fl6); 1102 1103 if (dst->error) 1104 goto tx_err_link_failure; 1105 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1106 if (IS_ERR(dst)) { 1107 err = PTR_ERR(dst); 1108 dst = NULL; 1109 goto tx_err_link_failure; 1110 } 1111 if (t->parms.collect_md && 1112 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1113 &fl6->daddr, 0, &fl6->saddr)) 1114 goto tx_err_link_failure; 1115 ndst = dst; 1116 } 1117 1118 tdev = dst->dev; 1119 1120 if (tdev == dev) { 1121 stats->collisions++; 1122 net_warn_ratelimited("%s: Local routing loop detected!\n", 1123 t->parms.name); 1124 goto tx_err_dst_release; 1125 } 1126 mtu = dst_mtu(dst) - psh_hlen - t->tun_hlen; 1127 if (encap_limit >= 0) { 1128 max_headroom += 8; 1129 mtu -= 8; 1130 } 1131 if (mtu < IPV6_MIN_MTU) 1132 mtu = IPV6_MIN_MTU; 1133 if (skb_dst(skb) && !t->parms.collect_md) 1134 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 1135 if (skb->len - t->tun_hlen > mtu && !skb_is_gso(skb)) { 1136 *pmtu = mtu; 1137 err = -EMSGSIZE; 1138 goto tx_err_dst_release; 1139 } 1140 1141 if (t->err_count > 0) { 1142 if (time_before(jiffies, 1143 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1144 t->err_count--; 1145 1146 dst_link_failure(skb); 1147 } else { 1148 t->err_count = 0; 1149 } 1150 } 1151 1152 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1153 1154 /* 1155 * Okay, now see if we can stuff it in the buffer as-is. 1156 */ 1157 max_headroom += LL_RESERVED_SPACE(tdev); 1158 1159 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1160 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1161 struct sk_buff *new_skb; 1162 1163 new_skb = skb_realloc_headroom(skb, max_headroom); 1164 if (!new_skb) 1165 goto tx_err_dst_release; 1166 1167 if (skb->sk) 1168 skb_set_owner_w(new_skb, skb->sk); 1169 consume_skb(skb); 1170 skb = new_skb; 1171 } 1172 1173 if (t->parms.collect_md) { 1174 if (t->encap.type != TUNNEL_ENCAP_NONE) 1175 goto tx_err_dst_release; 1176 } else { 1177 if (use_cache && ndst) 1178 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1179 } 1180 skb_dst_set(skb, dst); 1181 1182 if (encap_limit >= 0) { 1183 init_tel_txopt(&opt, encap_limit); 1184 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1185 } 1186 1187 /* Calculate max headroom for all the headers and adjust 1188 * needed_headroom if necessary. 1189 */ 1190 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1191 + dst->header_len + t->hlen; 1192 if (max_headroom > dev->needed_headroom) 1193 dev->needed_headroom = max_headroom; 1194 1195 err = ip6_tnl_encap(skb, t, &proto, fl6); 1196 if (err) 1197 return err; 1198 1199 skb_push(skb, sizeof(struct ipv6hdr)); 1200 skb_reset_network_header(skb); 1201 ipv6h = ipv6_hdr(skb); 1202 ip6_flow_hdr(ipv6h, dsfield, 1203 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1204 ipv6h->hop_limit = hop_limit; 1205 ipv6h->nexthdr = proto; 1206 ipv6h->saddr = fl6->saddr; 1207 ipv6h->daddr = fl6->daddr; 1208 ip6tunnel_xmit(NULL, skb, dev); 1209 return 0; 1210 tx_err_link_failure: 1211 stats->tx_carrier_errors++; 1212 dst_link_failure(skb); 1213 tx_err_dst_release: 1214 dst_release(dst); 1215 return err; 1216 } 1217 EXPORT_SYMBOL(ip6_tnl_xmit); 1218 1219 static inline int 1220 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1221 { 1222 struct ip6_tnl *t = netdev_priv(dev); 1223 const struct iphdr *iph = ip_hdr(skb); 1224 int encap_limit = -1; 1225 struct flowi6 fl6; 1226 __u8 dsfield; 1227 __u32 mtu; 1228 u8 tproto; 1229 int err; 1230 1231 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1232 1233 tproto = ACCESS_ONCE(t->parms.proto); 1234 if (tproto != IPPROTO_IPIP && tproto != 0) 1235 return -1; 1236 1237 if (t->parms.collect_md) { 1238 struct ip_tunnel_info *tun_info; 1239 const struct ip_tunnel_key *key; 1240 1241 tun_info = skb_tunnel_info(skb); 1242 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1243 ip_tunnel_info_af(tun_info) != AF_INET6)) 1244 return -1; 1245 key = &tun_info->key; 1246 memset(&fl6, 0, sizeof(fl6)); 1247 fl6.flowi6_proto = IPPROTO_IPIP; 1248 fl6.daddr = key->u.ipv6.dst; 1249 fl6.flowlabel = key->label; 1250 dsfield = ip6_tclass(key->label); 1251 } else { 1252 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1253 encap_limit = t->parms.encap_limit; 1254 1255 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1256 fl6.flowi6_proto = IPPROTO_IPIP; 1257 1258 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1259 dsfield = ipv4_get_dsfield(iph); 1260 else 1261 dsfield = ip6_tclass(t->parms.flowinfo); 1262 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1263 fl6.flowi6_mark = skb->mark; 1264 else 1265 fl6.flowi6_mark = t->parms.fwmark; 1266 } 1267 1268 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1269 1270 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1271 return -1; 1272 1273 dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph)); 1274 1275 skb_set_inner_ipproto(skb, IPPROTO_IPIP); 1276 1277 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1278 IPPROTO_IPIP); 1279 if (err != 0) { 1280 /* XXX: send ICMP error even if DF is not set. */ 1281 if (err == -EMSGSIZE) 1282 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1283 htonl(mtu)); 1284 return -1; 1285 } 1286 1287 return 0; 1288 } 1289 1290 static inline int 1291 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1292 { 1293 struct ip6_tnl *t = netdev_priv(dev); 1294 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1295 int encap_limit = -1; 1296 __u16 offset; 1297 struct flowi6 fl6; 1298 __u8 dsfield; 1299 __u32 mtu; 1300 u8 tproto; 1301 int err; 1302 1303 tproto = ACCESS_ONCE(t->parms.proto); 1304 if ((tproto != IPPROTO_IPV6 && tproto != 0) || 1305 ip6_tnl_addr_conflict(t, ipv6h)) 1306 return -1; 1307 1308 if (t->parms.collect_md) { 1309 struct ip_tunnel_info *tun_info; 1310 const struct ip_tunnel_key *key; 1311 1312 tun_info = skb_tunnel_info(skb); 1313 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1314 ip_tunnel_info_af(tun_info) != AF_INET6)) 1315 return -1; 1316 key = &tun_info->key; 1317 memset(&fl6, 0, sizeof(fl6)); 1318 fl6.flowi6_proto = IPPROTO_IPV6; 1319 fl6.daddr = key->u.ipv6.dst; 1320 fl6.flowlabel = key->label; 1321 dsfield = ip6_tclass(key->label); 1322 } else { 1323 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 1324 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 1325 ipv6h = ipv6_hdr(skb); 1326 if (offset > 0) { 1327 struct ipv6_tlv_tnl_enc_lim *tel; 1328 1329 tel = (void *)&skb_network_header(skb)[offset]; 1330 if (tel->encap_limit == 0) { 1331 icmpv6_send(skb, ICMPV6_PARAMPROB, 1332 ICMPV6_HDR_FIELD, offset + 2); 1333 return -1; 1334 } 1335 encap_limit = tel->encap_limit - 1; 1336 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 1337 encap_limit = t->parms.encap_limit; 1338 } 1339 1340 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1341 fl6.flowi6_proto = IPPROTO_IPV6; 1342 1343 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1344 dsfield = ipv6_get_dsfield(ipv6h); 1345 else 1346 dsfield = ip6_tclass(t->parms.flowinfo); 1347 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1348 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1349 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1350 fl6.flowi6_mark = skb->mark; 1351 else 1352 fl6.flowi6_mark = t->parms.fwmark; 1353 } 1354 1355 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1356 1357 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1358 return -1; 1359 1360 dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h)); 1361 1362 skb_set_inner_ipproto(skb, IPPROTO_IPV6); 1363 1364 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1365 IPPROTO_IPV6); 1366 if (err != 0) { 1367 if (err == -EMSGSIZE) 1368 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1369 return -1; 1370 } 1371 1372 return 0; 1373 } 1374 1375 static netdev_tx_t 1376 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1377 { 1378 struct ip6_tnl *t = netdev_priv(dev); 1379 struct net_device_stats *stats = &t->dev->stats; 1380 int ret; 1381 1382 switch (skb->protocol) { 1383 case htons(ETH_P_IP): 1384 ret = ip4ip6_tnl_xmit(skb, dev); 1385 break; 1386 case htons(ETH_P_IPV6): 1387 ret = ip6ip6_tnl_xmit(skb, dev); 1388 break; 1389 default: 1390 goto tx_err; 1391 } 1392 1393 if (ret < 0) 1394 goto tx_err; 1395 1396 return NETDEV_TX_OK; 1397 1398 tx_err: 1399 stats->tx_errors++; 1400 stats->tx_dropped++; 1401 kfree_skb(skb); 1402 return NETDEV_TX_OK; 1403 } 1404 1405 static void ip6_tnl_link_config(struct ip6_tnl *t) 1406 { 1407 struct net_device *dev = t->dev; 1408 struct __ip6_tnl_parm *p = &t->parms; 1409 struct flowi6 *fl6 = &t->fl.u.ip6; 1410 int t_hlen; 1411 1412 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1413 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1414 1415 /* Set up flowi template */ 1416 fl6->saddr = p->laddr; 1417 fl6->daddr = p->raddr; 1418 fl6->flowi6_oif = p->link; 1419 fl6->flowlabel = 0; 1420 1421 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1422 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1423 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1424 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1425 1426 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1427 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1428 1429 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1430 dev->flags |= IFF_POINTOPOINT; 1431 else 1432 dev->flags &= ~IFF_POINTOPOINT; 1433 1434 t->tun_hlen = 0; 1435 t->hlen = t->encap_hlen + t->tun_hlen; 1436 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1437 1438 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1439 int strict = (ipv6_addr_type(&p->raddr) & 1440 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1441 1442 struct rt6_info *rt = rt6_lookup(t->net, 1443 &p->raddr, &p->laddr, 1444 p->link, strict); 1445 1446 if (!rt) 1447 return; 1448 1449 if (rt->dst.dev) { 1450 dev->hard_header_len = rt->dst.dev->hard_header_len + 1451 t_hlen; 1452 1453 dev->mtu = rt->dst.dev->mtu - t_hlen; 1454 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1455 dev->mtu -= 8; 1456 1457 if (dev->mtu < IPV6_MIN_MTU) 1458 dev->mtu = IPV6_MIN_MTU; 1459 } 1460 ip6_rt_put(rt); 1461 } 1462 } 1463 1464 /** 1465 * ip6_tnl_change - update the tunnel parameters 1466 * @t: tunnel to be changed 1467 * @p: tunnel configuration parameters 1468 * 1469 * Description: 1470 * ip6_tnl_change() updates the tunnel parameters 1471 **/ 1472 1473 static int 1474 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1475 { 1476 t->parms.laddr = p->laddr; 1477 t->parms.raddr = p->raddr; 1478 t->parms.flags = p->flags; 1479 t->parms.hop_limit = p->hop_limit; 1480 t->parms.encap_limit = p->encap_limit; 1481 t->parms.flowinfo = p->flowinfo; 1482 t->parms.link = p->link; 1483 t->parms.proto = p->proto; 1484 t->parms.fwmark = p->fwmark; 1485 dst_cache_reset(&t->dst_cache); 1486 ip6_tnl_link_config(t); 1487 return 0; 1488 } 1489 1490 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1491 { 1492 struct net *net = t->net; 1493 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1494 int err; 1495 1496 ip6_tnl_unlink(ip6n, t); 1497 synchronize_net(); 1498 err = ip6_tnl_change(t, p); 1499 ip6_tnl_link(ip6n, t); 1500 netdev_state_change(t->dev); 1501 return err; 1502 } 1503 1504 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1505 { 1506 /* for default tnl0 device allow to change only the proto */ 1507 t->parms.proto = p->proto; 1508 netdev_state_change(t->dev); 1509 return 0; 1510 } 1511 1512 static void 1513 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1514 { 1515 p->laddr = u->laddr; 1516 p->raddr = u->raddr; 1517 p->flags = u->flags; 1518 p->hop_limit = u->hop_limit; 1519 p->encap_limit = u->encap_limit; 1520 p->flowinfo = u->flowinfo; 1521 p->link = u->link; 1522 p->proto = u->proto; 1523 memcpy(p->name, u->name, sizeof(u->name)); 1524 } 1525 1526 static void 1527 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1528 { 1529 u->laddr = p->laddr; 1530 u->raddr = p->raddr; 1531 u->flags = p->flags; 1532 u->hop_limit = p->hop_limit; 1533 u->encap_limit = p->encap_limit; 1534 u->flowinfo = p->flowinfo; 1535 u->link = p->link; 1536 u->proto = p->proto; 1537 memcpy(u->name, p->name, sizeof(u->name)); 1538 } 1539 1540 /** 1541 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1542 * @dev: virtual device associated with tunnel 1543 * @ifr: parameters passed from userspace 1544 * @cmd: command to be performed 1545 * 1546 * Description: 1547 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1548 * from userspace. 1549 * 1550 * The possible commands are the following: 1551 * %SIOCGETTUNNEL: get tunnel parameters for device 1552 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1553 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1554 * %SIOCDELTUNNEL: delete tunnel 1555 * 1556 * The fallback device "ip6tnl0", created during module 1557 * initialization, can be used for creating other tunnel devices. 1558 * 1559 * Return: 1560 * 0 on success, 1561 * %-EFAULT if unable to copy data to or from userspace, 1562 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1563 * %-EINVAL if passed tunnel parameters are invalid, 1564 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1565 * %-ENODEV if attempting to change or delete a nonexisting device 1566 **/ 1567 1568 static int 1569 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1570 { 1571 int err = 0; 1572 struct ip6_tnl_parm p; 1573 struct __ip6_tnl_parm p1; 1574 struct ip6_tnl *t = netdev_priv(dev); 1575 struct net *net = t->net; 1576 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1577 1578 memset(&p1, 0, sizeof(p1)); 1579 1580 switch (cmd) { 1581 case SIOCGETTUNNEL: 1582 if (dev == ip6n->fb_tnl_dev) { 1583 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1584 err = -EFAULT; 1585 break; 1586 } 1587 ip6_tnl_parm_from_user(&p1, &p); 1588 t = ip6_tnl_locate(net, &p1, 0); 1589 if (IS_ERR(t)) 1590 t = netdev_priv(dev); 1591 } else { 1592 memset(&p, 0, sizeof(p)); 1593 } 1594 ip6_tnl_parm_to_user(&p, &t->parms); 1595 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) { 1596 err = -EFAULT; 1597 } 1598 break; 1599 case SIOCADDTUNNEL: 1600 case SIOCCHGTUNNEL: 1601 err = -EPERM; 1602 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1603 break; 1604 err = -EFAULT; 1605 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1606 break; 1607 err = -EINVAL; 1608 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1609 p.proto != 0) 1610 break; 1611 ip6_tnl_parm_from_user(&p1, &p); 1612 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1613 if (cmd == SIOCCHGTUNNEL) { 1614 if (!IS_ERR(t)) { 1615 if (t->dev != dev) { 1616 err = -EEXIST; 1617 break; 1618 } 1619 } else 1620 t = netdev_priv(dev); 1621 if (dev == ip6n->fb_tnl_dev) 1622 err = ip6_tnl0_update(t, &p1); 1623 else 1624 err = ip6_tnl_update(t, &p1); 1625 } 1626 if (!IS_ERR(t)) { 1627 err = 0; 1628 ip6_tnl_parm_to_user(&p, &t->parms); 1629 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1630 err = -EFAULT; 1631 1632 } else { 1633 err = PTR_ERR(t); 1634 } 1635 break; 1636 case SIOCDELTUNNEL: 1637 err = -EPERM; 1638 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1639 break; 1640 1641 if (dev == ip6n->fb_tnl_dev) { 1642 err = -EFAULT; 1643 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1644 break; 1645 err = -ENOENT; 1646 ip6_tnl_parm_from_user(&p1, &p); 1647 t = ip6_tnl_locate(net, &p1, 0); 1648 if (IS_ERR(t)) 1649 break; 1650 err = -EPERM; 1651 if (t->dev == ip6n->fb_tnl_dev) 1652 break; 1653 dev = t->dev; 1654 } 1655 err = 0; 1656 unregister_netdevice(dev); 1657 break; 1658 default: 1659 err = -EINVAL; 1660 } 1661 return err; 1662 } 1663 1664 /** 1665 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1666 * @dev: virtual device associated with tunnel 1667 * @new_mtu: the new mtu 1668 * 1669 * Return: 1670 * 0 on success, 1671 * %-EINVAL if mtu too small 1672 **/ 1673 1674 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1675 { 1676 struct ip6_tnl *tnl = netdev_priv(dev); 1677 1678 if (tnl->parms.proto == IPPROTO_IPIP) { 1679 if (new_mtu < ETH_MIN_MTU) 1680 return -EINVAL; 1681 } else { 1682 if (new_mtu < IPV6_MIN_MTU) 1683 return -EINVAL; 1684 } 1685 if (new_mtu > 0xFFF8 - dev->hard_header_len) 1686 return -EINVAL; 1687 dev->mtu = new_mtu; 1688 return 0; 1689 } 1690 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1691 1692 int ip6_tnl_get_iflink(const struct net_device *dev) 1693 { 1694 struct ip6_tnl *t = netdev_priv(dev); 1695 1696 return t->parms.link; 1697 } 1698 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1699 1700 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1701 unsigned int num) 1702 { 1703 if (num >= MAX_IPTUN_ENCAP_OPS) 1704 return -ERANGE; 1705 1706 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1707 &ip6tun_encaps[num], 1708 NULL, ops) ? 0 : -1; 1709 } 1710 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1711 1712 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1713 unsigned int num) 1714 { 1715 int ret; 1716 1717 if (num >= MAX_IPTUN_ENCAP_OPS) 1718 return -ERANGE; 1719 1720 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1721 &ip6tun_encaps[num], 1722 ops, NULL) == ops) ? 0 : -1; 1723 1724 synchronize_net(); 1725 1726 return ret; 1727 } 1728 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1729 1730 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1731 struct ip_tunnel_encap *ipencap) 1732 { 1733 int hlen; 1734 1735 memset(&t->encap, 0, sizeof(t->encap)); 1736 1737 hlen = ip6_encap_hlen(ipencap); 1738 if (hlen < 0) 1739 return hlen; 1740 1741 t->encap.type = ipencap->type; 1742 t->encap.sport = ipencap->sport; 1743 t->encap.dport = ipencap->dport; 1744 t->encap.flags = ipencap->flags; 1745 1746 t->encap_hlen = hlen; 1747 t->hlen = t->encap_hlen + t->tun_hlen; 1748 1749 return 0; 1750 } 1751 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1752 1753 static const struct net_device_ops ip6_tnl_netdev_ops = { 1754 .ndo_init = ip6_tnl_dev_init, 1755 .ndo_uninit = ip6_tnl_dev_uninit, 1756 .ndo_start_xmit = ip6_tnl_start_xmit, 1757 .ndo_do_ioctl = ip6_tnl_ioctl, 1758 .ndo_change_mtu = ip6_tnl_change_mtu, 1759 .ndo_get_stats = ip6_get_stats, 1760 .ndo_get_iflink = ip6_tnl_get_iflink, 1761 }; 1762 1763 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1764 NETIF_F_FRAGLIST | \ 1765 NETIF_F_HIGHDMA | \ 1766 NETIF_F_GSO_SOFTWARE | \ 1767 NETIF_F_HW_CSUM) 1768 1769 /** 1770 * ip6_tnl_dev_setup - setup virtual tunnel device 1771 * @dev: virtual device associated with tunnel 1772 * 1773 * Description: 1774 * Initialize function pointers and device parameters 1775 **/ 1776 1777 static void ip6_tnl_dev_setup(struct net_device *dev) 1778 { 1779 dev->netdev_ops = &ip6_tnl_netdev_ops; 1780 dev->destructor = ip6_dev_free; 1781 1782 dev->type = ARPHRD_TUNNEL6; 1783 dev->flags |= IFF_NOARP; 1784 dev->addr_len = sizeof(struct in6_addr); 1785 dev->features |= NETIF_F_LLTX; 1786 netif_keep_dst(dev); 1787 1788 dev->features |= IPXIPX_FEATURES; 1789 dev->hw_features |= IPXIPX_FEATURES; 1790 1791 /* This perm addr will be used as interface identifier by IPv6 */ 1792 dev->addr_assign_type = NET_ADDR_RANDOM; 1793 eth_random_addr(dev->perm_addr); 1794 } 1795 1796 1797 /** 1798 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1799 * @dev: virtual device associated with tunnel 1800 **/ 1801 1802 static inline int 1803 ip6_tnl_dev_init_gen(struct net_device *dev) 1804 { 1805 struct ip6_tnl *t = netdev_priv(dev); 1806 int ret; 1807 int t_hlen; 1808 1809 t->dev = dev; 1810 t->net = dev_net(dev); 1811 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1812 if (!dev->tstats) 1813 return -ENOMEM; 1814 1815 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1816 if (ret) 1817 goto free_stats; 1818 1819 ret = gro_cells_init(&t->gro_cells, dev); 1820 if (ret) 1821 goto destroy_dst; 1822 1823 t->tun_hlen = 0; 1824 t->hlen = t->encap_hlen + t->tun_hlen; 1825 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1826 1827 dev->type = ARPHRD_TUNNEL6; 1828 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1829 dev->mtu = ETH_DATA_LEN - t_hlen; 1830 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1831 dev->mtu -= 8; 1832 dev->min_mtu = ETH_MIN_MTU; 1833 dev->max_mtu = 0xFFF8 - dev->hard_header_len; 1834 1835 return 0; 1836 1837 destroy_dst: 1838 dst_cache_destroy(&t->dst_cache); 1839 free_stats: 1840 free_percpu(dev->tstats); 1841 dev->tstats = NULL; 1842 1843 return ret; 1844 } 1845 1846 /** 1847 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1848 * @dev: virtual device associated with tunnel 1849 **/ 1850 1851 static int ip6_tnl_dev_init(struct net_device *dev) 1852 { 1853 struct ip6_tnl *t = netdev_priv(dev); 1854 int err = ip6_tnl_dev_init_gen(dev); 1855 1856 if (err) 1857 return err; 1858 ip6_tnl_link_config(t); 1859 if (t->parms.collect_md) { 1860 dev->features |= NETIF_F_NETNS_LOCAL; 1861 netif_keep_dst(dev); 1862 } 1863 return 0; 1864 } 1865 1866 /** 1867 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1868 * @dev: fallback device 1869 * 1870 * Return: 0 1871 **/ 1872 1873 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1874 { 1875 struct ip6_tnl *t = netdev_priv(dev); 1876 struct net *net = dev_net(dev); 1877 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1878 1879 t->parms.proto = IPPROTO_IPV6; 1880 dev_hold(dev); 1881 1882 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1883 return 0; 1884 } 1885 1886 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[]) 1887 { 1888 u8 proto; 1889 1890 if (!data || !data[IFLA_IPTUN_PROTO]) 1891 return 0; 1892 1893 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1894 if (proto != IPPROTO_IPV6 && 1895 proto != IPPROTO_IPIP && 1896 proto != 0) 1897 return -EINVAL; 1898 1899 return 0; 1900 } 1901 1902 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1903 struct __ip6_tnl_parm *parms) 1904 { 1905 memset(parms, 0, sizeof(*parms)); 1906 1907 if (!data) 1908 return; 1909 1910 if (data[IFLA_IPTUN_LINK]) 1911 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1912 1913 if (data[IFLA_IPTUN_LOCAL]) 1914 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1915 1916 if (data[IFLA_IPTUN_REMOTE]) 1917 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1918 1919 if (data[IFLA_IPTUN_TTL]) 1920 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1921 1922 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1923 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1924 1925 if (data[IFLA_IPTUN_FLOWINFO]) 1926 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1927 1928 if (data[IFLA_IPTUN_FLAGS]) 1929 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1930 1931 if (data[IFLA_IPTUN_PROTO]) 1932 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1933 1934 if (data[IFLA_IPTUN_COLLECT_METADATA]) 1935 parms->collect_md = true; 1936 1937 if (data[IFLA_IPTUN_FWMARK]) 1938 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 1939 } 1940 1941 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[], 1942 struct ip_tunnel_encap *ipencap) 1943 { 1944 bool ret = false; 1945 1946 memset(ipencap, 0, sizeof(*ipencap)); 1947 1948 if (!data) 1949 return ret; 1950 1951 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 1952 ret = true; 1953 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 1954 } 1955 1956 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 1957 ret = true; 1958 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 1959 } 1960 1961 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 1962 ret = true; 1963 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 1964 } 1965 1966 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 1967 ret = true; 1968 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 1969 } 1970 1971 return ret; 1972 } 1973 1974 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 1975 struct nlattr *tb[], struct nlattr *data[]) 1976 { 1977 struct net *net = dev_net(dev); 1978 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1979 struct ip6_tnl *nt, *t; 1980 struct ip_tunnel_encap ipencap; 1981 1982 nt = netdev_priv(dev); 1983 1984 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 1985 int err = ip6_tnl_encap_setup(nt, &ipencap); 1986 1987 if (err < 0) 1988 return err; 1989 } 1990 1991 ip6_tnl_netlink_parms(data, &nt->parms); 1992 1993 if (nt->parms.collect_md) { 1994 if (rtnl_dereference(ip6n->collect_md_tun)) 1995 return -EEXIST; 1996 } else { 1997 t = ip6_tnl_locate(net, &nt->parms, 0); 1998 if (!IS_ERR(t)) 1999 return -EEXIST; 2000 } 2001 2002 return ip6_tnl_create2(dev); 2003 } 2004 2005 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2006 struct nlattr *data[]) 2007 { 2008 struct ip6_tnl *t = netdev_priv(dev); 2009 struct __ip6_tnl_parm p; 2010 struct net *net = t->net; 2011 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2012 struct ip_tunnel_encap ipencap; 2013 2014 if (dev == ip6n->fb_tnl_dev) 2015 return -EINVAL; 2016 2017 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2018 int err = ip6_tnl_encap_setup(t, &ipencap); 2019 2020 if (err < 0) 2021 return err; 2022 } 2023 ip6_tnl_netlink_parms(data, &p); 2024 if (p.collect_md) 2025 return -EINVAL; 2026 2027 t = ip6_tnl_locate(net, &p, 0); 2028 if (!IS_ERR(t)) { 2029 if (t->dev != dev) 2030 return -EEXIST; 2031 } else 2032 t = netdev_priv(dev); 2033 2034 return ip6_tnl_update(t, &p); 2035 } 2036 2037 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2038 { 2039 struct net *net = dev_net(dev); 2040 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2041 2042 if (dev != ip6n->fb_tnl_dev) 2043 unregister_netdevice_queue(dev, head); 2044 } 2045 2046 static size_t ip6_tnl_get_size(const struct net_device *dev) 2047 { 2048 return 2049 /* IFLA_IPTUN_LINK */ 2050 nla_total_size(4) + 2051 /* IFLA_IPTUN_LOCAL */ 2052 nla_total_size(sizeof(struct in6_addr)) + 2053 /* IFLA_IPTUN_REMOTE */ 2054 nla_total_size(sizeof(struct in6_addr)) + 2055 /* IFLA_IPTUN_TTL */ 2056 nla_total_size(1) + 2057 /* IFLA_IPTUN_ENCAP_LIMIT */ 2058 nla_total_size(1) + 2059 /* IFLA_IPTUN_FLOWINFO */ 2060 nla_total_size(4) + 2061 /* IFLA_IPTUN_FLAGS */ 2062 nla_total_size(4) + 2063 /* IFLA_IPTUN_PROTO */ 2064 nla_total_size(1) + 2065 /* IFLA_IPTUN_ENCAP_TYPE */ 2066 nla_total_size(2) + 2067 /* IFLA_IPTUN_ENCAP_FLAGS */ 2068 nla_total_size(2) + 2069 /* IFLA_IPTUN_ENCAP_SPORT */ 2070 nla_total_size(2) + 2071 /* IFLA_IPTUN_ENCAP_DPORT */ 2072 nla_total_size(2) + 2073 /* IFLA_IPTUN_COLLECT_METADATA */ 2074 nla_total_size(0) + 2075 /* IFLA_IPTUN_FWMARK */ 2076 nla_total_size(4) + 2077 0; 2078 } 2079 2080 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2081 { 2082 struct ip6_tnl *tunnel = netdev_priv(dev); 2083 struct __ip6_tnl_parm *parm = &tunnel->parms; 2084 2085 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2086 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2087 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2088 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2089 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2090 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2091 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2092 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2093 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2094 goto nla_put_failure; 2095 2096 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2097 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2098 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2099 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2100 goto nla_put_failure; 2101 2102 if (parm->collect_md) 2103 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2104 goto nla_put_failure; 2105 2106 return 0; 2107 2108 nla_put_failure: 2109 return -EMSGSIZE; 2110 } 2111 2112 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2113 { 2114 struct ip6_tnl *tunnel = netdev_priv(dev); 2115 2116 return tunnel->net; 2117 } 2118 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2119 2120 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2121 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2122 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2123 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2124 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2125 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2126 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2127 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2128 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2129 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2130 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2131 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2132 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2133 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2134 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2135 }; 2136 2137 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2138 .kind = "ip6tnl", 2139 .maxtype = IFLA_IPTUN_MAX, 2140 .policy = ip6_tnl_policy, 2141 .priv_size = sizeof(struct ip6_tnl), 2142 .setup = ip6_tnl_dev_setup, 2143 .validate = ip6_tnl_validate, 2144 .newlink = ip6_tnl_newlink, 2145 .changelink = ip6_tnl_changelink, 2146 .dellink = ip6_tnl_dellink, 2147 .get_size = ip6_tnl_get_size, 2148 .fill_info = ip6_tnl_fill_info, 2149 .get_link_net = ip6_tnl_get_link_net, 2150 }; 2151 2152 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2153 .handler = ip4ip6_rcv, 2154 .err_handler = ip4ip6_err, 2155 .priority = 1, 2156 }; 2157 2158 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2159 .handler = ip6ip6_rcv, 2160 .err_handler = ip6ip6_err, 2161 .priority = 1, 2162 }; 2163 2164 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net) 2165 { 2166 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2167 struct net_device *dev, *aux; 2168 int h; 2169 struct ip6_tnl *t; 2170 LIST_HEAD(list); 2171 2172 for_each_netdev_safe(net, dev, aux) 2173 if (dev->rtnl_link_ops == &ip6_link_ops) 2174 unregister_netdevice_queue(dev, &list); 2175 2176 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2177 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2178 while (t) { 2179 /* If dev is in the same netns, it has already 2180 * been added to the list by the previous loop. 2181 */ 2182 if (!net_eq(dev_net(t->dev), net)) 2183 unregister_netdevice_queue(t->dev, &list); 2184 t = rtnl_dereference(t->next); 2185 } 2186 } 2187 2188 unregister_netdevice_many(&list); 2189 } 2190 2191 static int __net_init ip6_tnl_init_net(struct net *net) 2192 { 2193 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2194 struct ip6_tnl *t = NULL; 2195 int err; 2196 2197 ip6n->tnls[0] = ip6n->tnls_wc; 2198 ip6n->tnls[1] = ip6n->tnls_r_l; 2199 2200 err = -ENOMEM; 2201 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2202 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2203 2204 if (!ip6n->fb_tnl_dev) 2205 goto err_alloc_dev; 2206 dev_net_set(ip6n->fb_tnl_dev, net); 2207 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2208 /* FB netdevice is special: we have one, and only one per netns. 2209 * Allowing to move it to another netns is clearly unsafe. 2210 */ 2211 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 2212 2213 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2214 if (err < 0) 2215 goto err_register; 2216 2217 err = register_netdev(ip6n->fb_tnl_dev); 2218 if (err < 0) 2219 goto err_register; 2220 2221 t = netdev_priv(ip6n->fb_tnl_dev); 2222 2223 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2224 return 0; 2225 2226 err_register: 2227 ip6_dev_free(ip6n->fb_tnl_dev); 2228 err_alloc_dev: 2229 return err; 2230 } 2231 2232 static void __net_exit ip6_tnl_exit_net(struct net *net) 2233 { 2234 rtnl_lock(); 2235 ip6_tnl_destroy_tunnels(net); 2236 rtnl_unlock(); 2237 } 2238 2239 static struct pernet_operations ip6_tnl_net_ops = { 2240 .init = ip6_tnl_init_net, 2241 .exit = ip6_tnl_exit_net, 2242 .id = &ip6_tnl_net_id, 2243 .size = sizeof(struct ip6_tnl_net), 2244 }; 2245 2246 /** 2247 * ip6_tunnel_init - register protocol and reserve needed resources 2248 * 2249 * Return: 0 on success 2250 **/ 2251 2252 static int __init ip6_tunnel_init(void) 2253 { 2254 int err; 2255 2256 err = register_pernet_device(&ip6_tnl_net_ops); 2257 if (err < 0) 2258 goto out_pernet; 2259 2260 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2261 if (err < 0) { 2262 pr_err("%s: can't register ip4ip6\n", __func__); 2263 goto out_ip4ip6; 2264 } 2265 2266 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2267 if (err < 0) { 2268 pr_err("%s: can't register ip6ip6\n", __func__); 2269 goto out_ip6ip6; 2270 } 2271 err = rtnl_link_register(&ip6_link_ops); 2272 if (err < 0) 2273 goto rtnl_link_failed; 2274 2275 return 0; 2276 2277 rtnl_link_failed: 2278 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2279 out_ip6ip6: 2280 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2281 out_ip4ip6: 2282 unregister_pernet_device(&ip6_tnl_net_ops); 2283 out_pernet: 2284 return err; 2285 } 2286 2287 /** 2288 * ip6_tunnel_cleanup - free resources and unregister protocol 2289 **/ 2290 2291 static void __exit ip6_tunnel_cleanup(void) 2292 { 2293 rtnl_link_unregister(&ip6_link_ops); 2294 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2295 pr_info("%s: can't deregister ip4ip6\n", __func__); 2296 2297 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2298 pr_info("%s: can't deregister ip6ip6\n", __func__); 2299 2300 unregister_pernet_device(&ip6_tnl_net_ops); 2301 } 2302 2303 module_init(ip6_tunnel_init); 2304 module_exit(ip6_tunnel_cleanup); 2305