1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 tunneling device 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Ville Nuorvala <vnuorval@tcs.hut.fi> 8 * Yasuyuki Kozakai <kozakai@linux-ipv6.org> 9 * 10 * Based on: 11 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c 12 * 13 * RFC 2473 14 */ 15 16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 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 #include <linux/etherdevice.h> 39 40 #include <linux/uaccess.h> 41 #include <linux/atomic.h> 42 43 #include <net/icmp.h> 44 #include <net/ip.h> 45 #include <net/ip_tunnels.h> 46 #include <net/ipv6.h> 47 #include <net/ip6_route.h> 48 #include <net/addrconf.h> 49 #include <net/ip6_tunnel.h> 50 #include <net/xfrm.h> 51 #include <net/dsfield.h> 52 #include <net/inet_ecn.h> 53 #include <net/net_namespace.h> 54 #include <net/netns/generic.h> 55 #include <net/dst_metadata.h> 56 57 MODULE_AUTHOR("Ville Nuorvala"); 58 MODULE_DESCRIPTION("IPv6 tunneling device"); 59 MODULE_LICENSE("GPL"); 60 MODULE_ALIAS_RTNL_LINK("ip6tnl"); 61 MODULE_ALIAS_NETDEV("ip6tnl0"); 62 63 #define IP6_TUNNEL_HASH_SIZE_SHIFT 5 64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT) 65 66 static bool log_ecn_error = true; 67 module_param(log_ecn_error, bool, 0644); 68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 69 70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 71 { 72 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 73 74 return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT); 75 } 76 77 static int ip6_tnl_dev_init(struct net_device *dev); 78 static void ip6_tnl_dev_setup(struct net_device *dev); 79 static struct rtnl_link_ops ip6_link_ops __read_mostly; 80 81 static unsigned int ip6_tnl_net_id __read_mostly; 82 struct ip6_tnl_net { 83 /* the IPv6 tunnel fallback device */ 84 struct net_device *fb_tnl_dev; 85 /* lists for storing tunnels in use */ 86 struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE]; 87 struct ip6_tnl __rcu *tnls_wc[1]; 88 struct ip6_tnl __rcu **tnls[2]; 89 struct ip6_tnl __rcu *collect_md_tun; 90 }; 91 92 static inline int ip6_tnl_mpls_supported(void) 93 { 94 return IS_ENABLED(CONFIG_MPLS); 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 * @link: ifindex of underlying interface 130 * @remote: the address of the tunnel exit-point 131 * @local: the address of the tunnel entry-point 132 * 133 * Return: 134 * tunnel matching given end-points if found, 135 * else fallback tunnel if its device is up, 136 * else %NULL 137 **/ 138 139 #define for_each_ip6_tunnel_rcu(start) \ 140 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 141 142 static struct ip6_tnl * 143 ip6_tnl_lookup(struct net *net, int link, 144 const struct in6_addr *remote, const struct in6_addr *local) 145 { 146 unsigned int hash = HASH(remote, local); 147 struct ip6_tnl *t, *cand = NULL; 148 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 149 struct in6_addr any; 150 151 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 152 if (!ipv6_addr_equal(local, &t->parms.laddr) || 153 !ipv6_addr_equal(remote, &t->parms.raddr) || 154 !(t->dev->flags & IFF_UP)) 155 continue; 156 157 if (link == t->parms.link) 158 return t; 159 else 160 cand = t; 161 } 162 163 memset(&any, 0, sizeof(any)); 164 hash = HASH(&any, local); 165 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 166 if (!ipv6_addr_equal(local, &t->parms.laddr) || 167 !ipv6_addr_any(&t->parms.raddr) || 168 !(t->dev->flags & IFF_UP)) 169 continue; 170 171 if (link == t->parms.link) 172 return t; 173 else if (!cand) 174 cand = t; 175 } 176 177 hash = HASH(remote, &any); 178 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 179 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 180 !ipv6_addr_any(&t->parms.laddr) || 181 !(t->dev->flags & IFF_UP)) 182 continue; 183 184 if (link == t->parms.link) 185 return t; 186 else if (!cand) 187 cand = t; 188 } 189 190 if (cand) 191 return cand; 192 193 t = rcu_dereference(ip6n->collect_md_tun); 194 if (t && t->dev->flags & IFF_UP) 195 return t; 196 197 t = rcu_dereference(ip6n->tnls_wc[0]); 198 if (t && (t->dev->flags & IFF_UP)) 199 return t; 200 201 return NULL; 202 } 203 204 /** 205 * ip6_tnl_bucket - get head of list matching given tunnel parameters 206 * @p: parameters containing tunnel end-points 207 * 208 * Description: 209 * ip6_tnl_bucket() returns the head of the list matching the 210 * &struct in6_addr entries laddr and raddr in @p. 211 * 212 * Return: head of IPv6 tunnel list 213 **/ 214 215 static struct ip6_tnl __rcu ** 216 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 217 { 218 const struct in6_addr *remote = &p->raddr; 219 const struct in6_addr *local = &p->laddr; 220 unsigned int h = 0; 221 int prio = 0; 222 223 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 224 prio = 1; 225 h = HASH(remote, local); 226 } 227 return &ip6n->tnls[prio][h]; 228 } 229 230 /** 231 * ip6_tnl_link - add tunnel to hash table 232 * @t: tunnel to be added 233 **/ 234 235 static void 236 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 237 { 238 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 239 240 if (t->parms.collect_md) 241 rcu_assign_pointer(ip6n->collect_md_tun, t); 242 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 243 rcu_assign_pointer(*tp, t); 244 } 245 246 /** 247 * ip6_tnl_unlink - remove tunnel from hash table 248 * @t: tunnel to be removed 249 **/ 250 251 static void 252 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 253 { 254 struct ip6_tnl __rcu **tp; 255 struct ip6_tnl *iter; 256 257 if (t->parms.collect_md) 258 rcu_assign_pointer(ip6n->collect_md_tun, NULL); 259 260 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 261 (iter = rtnl_dereference(*tp)) != NULL; 262 tp = &iter->next) { 263 if (t == iter) { 264 rcu_assign_pointer(*tp, t->next); 265 break; 266 } 267 } 268 } 269 270 static void ip6_dev_free(struct net_device *dev) 271 { 272 struct ip6_tnl *t = netdev_priv(dev); 273 274 gro_cells_destroy(&t->gro_cells); 275 dst_cache_destroy(&t->dst_cache); 276 free_percpu(dev->tstats); 277 } 278 279 static int ip6_tnl_create2(struct net_device *dev) 280 { 281 struct ip6_tnl *t = netdev_priv(dev); 282 struct net *net = dev_net(dev); 283 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 284 int err; 285 286 t = netdev_priv(dev); 287 288 dev->rtnl_link_ops = &ip6_link_ops; 289 err = register_netdevice(dev); 290 if (err < 0) 291 goto out; 292 293 strcpy(t->parms.name, dev->name); 294 295 dev_hold(dev); 296 ip6_tnl_link(ip6n, t); 297 return 0; 298 299 out: 300 return err; 301 } 302 303 /** 304 * ip6_tnl_create - create a new tunnel 305 * @p: tunnel parameters 306 * @pt: pointer to new tunnel 307 * 308 * Description: 309 * Create tunnel matching given parameters. 310 * 311 * Return: 312 * created tunnel or error pointer 313 **/ 314 315 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 316 { 317 struct net_device *dev; 318 struct ip6_tnl *t; 319 char name[IFNAMSIZ]; 320 int err = -E2BIG; 321 322 if (p->name[0]) { 323 if (!dev_valid_name(p->name)) 324 goto failed; 325 strlcpy(name, p->name, IFNAMSIZ); 326 } else { 327 sprintf(name, "ip6tnl%%d"); 328 } 329 err = -ENOMEM; 330 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 331 ip6_tnl_dev_setup); 332 if (!dev) 333 goto failed; 334 335 dev_net_set(dev, net); 336 337 t = netdev_priv(dev); 338 t->parms = *p; 339 t->net = dev_net(dev); 340 err = ip6_tnl_create2(dev); 341 if (err < 0) 342 goto failed_free; 343 344 return t; 345 346 failed_free: 347 free_netdev(dev); 348 failed: 349 return ERR_PTR(err); 350 } 351 352 /** 353 * ip6_tnl_locate - find or create tunnel matching given parameters 354 * @p: tunnel parameters 355 * @create: != 0 if allowed to create new tunnel if no match found 356 * 357 * Description: 358 * ip6_tnl_locate() first tries to locate an existing tunnel 359 * based on @parms. If this is unsuccessful, but @create is set a new 360 * tunnel device is created and registered for use. 361 * 362 * Return: 363 * matching tunnel or error pointer 364 **/ 365 366 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 367 struct __ip6_tnl_parm *p, int create) 368 { 369 const struct in6_addr *remote = &p->raddr; 370 const struct in6_addr *local = &p->laddr; 371 struct ip6_tnl __rcu **tp; 372 struct ip6_tnl *t; 373 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 374 375 for (tp = ip6_tnl_bucket(ip6n, p); 376 (t = rtnl_dereference(*tp)) != NULL; 377 tp = &t->next) { 378 if (ipv6_addr_equal(local, &t->parms.laddr) && 379 ipv6_addr_equal(remote, &t->parms.raddr) && 380 p->link == t->parms.link) { 381 if (create) 382 return ERR_PTR(-EEXIST); 383 384 return t; 385 } 386 } 387 if (!create) 388 return ERR_PTR(-ENODEV); 389 return ip6_tnl_create(net, p); 390 } 391 392 /** 393 * ip6_tnl_dev_uninit - tunnel device uninitializer 394 * @dev: the device to be destroyed 395 * 396 * Description: 397 * ip6_tnl_dev_uninit() removes tunnel from its list 398 **/ 399 400 static void 401 ip6_tnl_dev_uninit(struct net_device *dev) 402 { 403 struct ip6_tnl *t = netdev_priv(dev); 404 struct net *net = t->net; 405 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 406 407 if (dev == ip6n->fb_tnl_dev) 408 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 409 else 410 ip6_tnl_unlink(ip6n, t); 411 dst_cache_reset(&t->dst_cache); 412 dev_put(dev); 413 } 414 415 /** 416 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 417 * @skb: received socket buffer 418 * 419 * Return: 420 * 0 if none was found, 421 * else index to encapsulation limit 422 **/ 423 424 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 425 { 426 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw; 427 unsigned int nhoff = raw - skb->data; 428 unsigned int off = nhoff + sizeof(*ipv6h); 429 u8 next, nexthdr = ipv6h->nexthdr; 430 431 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 432 struct ipv6_opt_hdr *hdr; 433 u16 optlen; 434 435 if (!pskb_may_pull(skb, off + sizeof(*hdr))) 436 break; 437 438 hdr = (struct ipv6_opt_hdr *)(skb->data + off); 439 if (nexthdr == NEXTHDR_FRAGMENT) { 440 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 441 if (frag_hdr->frag_off) 442 break; 443 optlen = 8; 444 } else if (nexthdr == NEXTHDR_AUTH) { 445 optlen = ipv6_authlen(hdr); 446 } else { 447 optlen = ipv6_optlen(hdr); 448 } 449 /* cache hdr->nexthdr, since pskb_may_pull() might 450 * invalidate hdr 451 */ 452 next = hdr->nexthdr; 453 if (nexthdr == NEXTHDR_DEST) { 454 u16 i = 2; 455 456 /* Remember : hdr is no longer valid at this point. */ 457 if (!pskb_may_pull(skb, off + optlen)) 458 break; 459 460 while (1) { 461 struct ipv6_tlv_tnl_enc_lim *tel; 462 463 /* No more room for encapsulation limit */ 464 if (i + sizeof(*tel) > optlen) 465 break; 466 467 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i); 468 /* return index of option if found and valid */ 469 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 470 tel->length == 1) 471 return i + off - nhoff; 472 /* else jump to next option */ 473 if (tel->type) 474 i += tel->length + 2; 475 else 476 i++; 477 } 478 } 479 nexthdr = next; 480 off += optlen; 481 } 482 return 0; 483 } 484 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 485 486 /** 487 * ip6_tnl_err - tunnel error handler 488 * 489 * Description: 490 * ip6_tnl_err() should handle errors in the tunnel according 491 * to the specifications in RFC 2473. 492 **/ 493 494 static int 495 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 496 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 497 { 498 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data; 499 struct net *net = dev_net(skb->dev); 500 u8 rel_type = ICMPV6_DEST_UNREACH; 501 u8 rel_code = ICMPV6_ADDR_UNREACH; 502 __u32 rel_info = 0; 503 struct ip6_tnl *t; 504 int err = -ENOENT; 505 int rel_msg = 0; 506 u8 tproto; 507 __u16 len; 508 509 /* If the packet doesn't contain the original IPv6 header we are 510 in trouble since we might need the source address for further 511 processing of the error. */ 512 513 rcu_read_lock(); 514 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr); 515 if (!t) 516 goto out; 517 518 tproto = READ_ONCE(t->parms.proto); 519 if (tproto != ipproto && tproto != 0) 520 goto out; 521 522 err = 0; 523 524 switch (*type) { 525 case ICMPV6_DEST_UNREACH: 526 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 527 t->parms.name); 528 rel_msg = 1; 529 break; 530 case ICMPV6_TIME_EXCEED: 531 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 532 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 533 t->parms.name); 534 rel_msg = 1; 535 } 536 break; 537 case ICMPV6_PARAMPROB: { 538 struct ipv6_tlv_tnl_enc_lim *tel; 539 __u32 teli; 540 541 teli = 0; 542 if ((*code) == ICMPV6_HDR_FIELD) 543 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 544 545 if (teli && teli == *info - 2) { 546 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 547 if (tel->encap_limit == 0) { 548 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 549 t->parms.name); 550 rel_msg = 1; 551 } 552 } else { 553 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 554 t->parms.name); 555 } 556 break; 557 } 558 case ICMPV6_PKT_TOOBIG: { 559 __u32 mtu; 560 561 ip6_update_pmtu(skb, net, htonl(*info), 0, 0, 562 sock_net_uid(net, NULL)); 563 mtu = *info - offset; 564 if (mtu < IPV6_MIN_MTU) 565 mtu = IPV6_MIN_MTU; 566 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len); 567 if (len > mtu) { 568 rel_type = ICMPV6_PKT_TOOBIG; 569 rel_code = 0; 570 rel_info = mtu; 571 rel_msg = 1; 572 } 573 break; 574 } 575 case NDISC_REDIRECT: 576 ip6_redirect(skb, net, skb->dev->ifindex, 0, 577 sock_net_uid(net, NULL)); 578 break; 579 } 580 581 *type = rel_type; 582 *code = rel_code; 583 *info = rel_info; 584 *msg = rel_msg; 585 586 out: 587 rcu_read_unlock(); 588 return err; 589 } 590 591 static int 592 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 593 u8 type, u8 code, int offset, __be32 info) 594 { 595 __u32 rel_info = ntohl(info); 596 const struct iphdr *eiph; 597 struct sk_buff *skb2; 598 int err, rel_msg = 0; 599 u8 rel_type = type; 600 u8 rel_code = code; 601 struct rtable *rt; 602 struct flowi4 fl4; 603 604 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 605 &rel_msg, &rel_info, offset); 606 if (err < 0) 607 return err; 608 609 if (rel_msg == 0) 610 return 0; 611 612 switch (rel_type) { 613 case ICMPV6_DEST_UNREACH: 614 if (rel_code != ICMPV6_ADDR_UNREACH) 615 return 0; 616 rel_type = ICMP_DEST_UNREACH; 617 rel_code = ICMP_HOST_UNREACH; 618 break; 619 case ICMPV6_PKT_TOOBIG: 620 if (rel_code != 0) 621 return 0; 622 rel_type = ICMP_DEST_UNREACH; 623 rel_code = ICMP_FRAG_NEEDED; 624 break; 625 default: 626 return 0; 627 } 628 629 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 630 return 0; 631 632 skb2 = skb_clone(skb, GFP_ATOMIC); 633 if (!skb2) 634 return 0; 635 636 skb_dst_drop(skb2); 637 638 skb_pull(skb2, offset); 639 skb_reset_network_header(skb2); 640 eiph = ip_hdr(skb2); 641 642 /* Try to guess incoming interface */ 643 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr, 644 0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 645 if (IS_ERR(rt)) 646 goto out; 647 648 skb2->dev = rt->dst.dev; 649 ip_rt_put(rt); 650 651 /* route "incoming" packet */ 652 if (rt->rt_flags & RTCF_LOCAL) { 653 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 654 eiph->daddr, eiph->saddr, 0, 0, 655 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 656 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) { 657 if (!IS_ERR(rt)) 658 ip_rt_put(rt); 659 goto out; 660 } 661 skb_dst_set(skb2, &rt->dst); 662 } else { 663 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 664 skb2->dev) || 665 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6) 666 goto out; 667 } 668 669 /* change mtu on this route */ 670 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 671 if (rel_info > dst_mtu(skb_dst(skb2))) 672 goto out; 673 674 skb_dst_update_pmtu_no_confirm(skb2, rel_info); 675 } 676 677 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 678 679 out: 680 kfree_skb(skb2); 681 return 0; 682 } 683 684 static int 685 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 686 u8 type, u8 code, int offset, __be32 info) 687 { 688 __u32 rel_info = ntohl(info); 689 int err, rel_msg = 0; 690 u8 rel_type = type; 691 u8 rel_code = code; 692 693 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 694 &rel_msg, &rel_info, offset); 695 if (err < 0) 696 return err; 697 698 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 699 struct rt6_info *rt; 700 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 701 702 if (!skb2) 703 return 0; 704 705 skb_dst_drop(skb2); 706 skb_pull(skb2, offset); 707 skb_reset_network_header(skb2); 708 709 /* Try to guess incoming interface */ 710 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 711 NULL, 0, skb2, 0); 712 713 if (rt && rt->dst.dev) 714 skb2->dev = rt->dst.dev; 715 716 icmpv6_send(skb2, rel_type, rel_code, rel_info); 717 718 ip6_rt_put(rt); 719 720 kfree_skb(skb2); 721 } 722 723 return 0; 724 } 725 726 static int 727 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 728 u8 type, u8 code, int offset, __be32 info) 729 { 730 __u32 rel_info = ntohl(info); 731 int err, rel_msg = 0; 732 u8 rel_type = type; 733 u8 rel_code = code; 734 735 err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code, 736 &rel_msg, &rel_info, offset); 737 return err; 738 } 739 740 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 741 const struct ipv6hdr *ipv6h, 742 struct sk_buff *skb) 743 { 744 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 745 746 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 747 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 748 749 return IP6_ECN_decapsulate(ipv6h, skb); 750 } 751 752 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 753 const struct ipv6hdr *ipv6h, 754 struct sk_buff *skb) 755 { 756 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 757 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 758 759 return IP6_ECN_decapsulate(ipv6h, skb); 760 } 761 762 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 763 const struct ipv6hdr *ipv6h, 764 struct sk_buff *skb) 765 { 766 /* ECN is not supported in AF_MPLS */ 767 return 0; 768 } 769 770 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 771 const struct in6_addr *laddr, 772 const struct in6_addr *raddr) 773 { 774 struct __ip6_tnl_parm *p = &t->parms; 775 int ltype = ipv6_addr_type(laddr); 776 int rtype = ipv6_addr_type(raddr); 777 __u32 flags = 0; 778 779 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 780 flags = IP6_TNL_F_CAP_PER_PACKET; 781 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 782 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 783 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 784 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 785 if (ltype&IPV6_ADDR_UNICAST) 786 flags |= IP6_TNL_F_CAP_XMIT; 787 if (rtype&IPV6_ADDR_UNICAST) 788 flags |= IP6_TNL_F_CAP_RCV; 789 } 790 return flags; 791 } 792 EXPORT_SYMBOL(ip6_tnl_get_cap); 793 794 /* called with rcu_read_lock() */ 795 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 796 const struct in6_addr *laddr, 797 const struct in6_addr *raddr) 798 { 799 struct __ip6_tnl_parm *p = &t->parms; 800 int ret = 0; 801 struct net *net = t->net; 802 803 if ((p->flags & IP6_TNL_F_CAP_RCV) || 804 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 805 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 806 struct net_device *ldev = NULL; 807 808 if (p->link) 809 ldev = dev_get_by_index_rcu(net, p->link); 810 811 if ((ipv6_addr_is_multicast(laddr) || 812 likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false, 813 0, IFA_F_TENTATIVE))) && 814 ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) || 815 likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true, 816 0, IFA_F_TENTATIVE)))) 817 ret = 1; 818 } 819 return ret; 820 } 821 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 822 823 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb, 824 const struct tnl_ptk_info *tpi, 825 struct metadata_dst *tun_dst, 826 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 827 const struct ipv6hdr *ipv6h, 828 struct sk_buff *skb), 829 bool log_ecn_err) 830 { 831 struct pcpu_sw_netstats *tstats; 832 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 833 int err; 834 835 if ((!(tpi->flags & TUNNEL_CSUM) && 836 (tunnel->parms.i_flags & TUNNEL_CSUM)) || 837 ((tpi->flags & TUNNEL_CSUM) && 838 !(tunnel->parms.i_flags & TUNNEL_CSUM))) { 839 tunnel->dev->stats.rx_crc_errors++; 840 tunnel->dev->stats.rx_errors++; 841 goto drop; 842 } 843 844 if (tunnel->parms.i_flags & TUNNEL_SEQ) { 845 if (!(tpi->flags & TUNNEL_SEQ) || 846 (tunnel->i_seqno && 847 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) { 848 tunnel->dev->stats.rx_fifo_errors++; 849 tunnel->dev->stats.rx_errors++; 850 goto drop; 851 } 852 tunnel->i_seqno = ntohl(tpi->seq) + 1; 853 } 854 855 skb->protocol = tpi->proto; 856 857 /* Warning: All skb pointers will be invalidated! */ 858 if (tunnel->dev->type == ARPHRD_ETHER) { 859 if (!pskb_may_pull(skb, ETH_HLEN)) { 860 tunnel->dev->stats.rx_length_errors++; 861 tunnel->dev->stats.rx_errors++; 862 goto drop; 863 } 864 865 ipv6h = ipv6_hdr(skb); 866 skb->protocol = eth_type_trans(skb, tunnel->dev); 867 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 868 } else { 869 skb->dev = tunnel->dev; 870 } 871 872 skb_reset_network_header(skb); 873 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 874 875 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 876 877 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb); 878 if (unlikely(err)) { 879 if (log_ecn_err) 880 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n", 881 &ipv6h->saddr, 882 ipv6_get_dsfield(ipv6h)); 883 if (err > 1) { 884 ++tunnel->dev->stats.rx_frame_errors; 885 ++tunnel->dev->stats.rx_errors; 886 goto drop; 887 } 888 } 889 890 tstats = this_cpu_ptr(tunnel->dev->tstats); 891 u64_stats_update_begin(&tstats->syncp); 892 tstats->rx_packets++; 893 tstats->rx_bytes += skb->len; 894 u64_stats_update_end(&tstats->syncp); 895 896 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev))); 897 898 if (tun_dst) 899 skb_dst_set(skb, (struct dst_entry *)tun_dst); 900 901 gro_cells_receive(&tunnel->gro_cells, skb); 902 return 0; 903 904 drop: 905 if (tun_dst) 906 dst_release((struct dst_entry *)tun_dst); 907 kfree_skb(skb); 908 return 0; 909 } 910 911 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb, 912 const struct tnl_ptk_info *tpi, 913 struct metadata_dst *tun_dst, 914 bool log_ecn_err) 915 { 916 return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate, 917 log_ecn_err); 918 } 919 EXPORT_SYMBOL(ip6_tnl_rcv); 920 921 static const struct tnl_ptk_info tpi_v6 = { 922 /* no tunnel info required for ipxip6. */ 923 .proto = htons(ETH_P_IPV6), 924 }; 925 926 static const struct tnl_ptk_info tpi_v4 = { 927 /* no tunnel info required for ipxip6. */ 928 .proto = htons(ETH_P_IP), 929 }; 930 931 static const struct tnl_ptk_info tpi_mpls = { 932 /* no tunnel info required for mplsip6. */ 933 .proto = htons(ETH_P_MPLS_UC), 934 }; 935 936 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto, 937 const struct tnl_ptk_info *tpi, 938 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 939 const struct ipv6hdr *ipv6h, 940 struct sk_buff *skb)) 941 { 942 struct ip6_tnl *t; 943 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 944 struct metadata_dst *tun_dst = NULL; 945 int ret = -1; 946 947 rcu_read_lock(); 948 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr); 949 950 if (t) { 951 u8 tproto = READ_ONCE(t->parms.proto); 952 953 if (tproto != ipproto && tproto != 0) 954 goto drop; 955 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 956 goto drop; 957 ipv6h = ipv6_hdr(skb); 958 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) 959 goto drop; 960 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 961 goto drop; 962 if (t->parms.collect_md) { 963 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0); 964 if (!tun_dst) 965 goto drop; 966 } 967 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 968 log_ecn_error); 969 } 970 971 rcu_read_unlock(); 972 973 return ret; 974 975 drop: 976 rcu_read_unlock(); 977 kfree_skb(skb); 978 return 0; 979 } 980 981 static int ip4ip6_rcv(struct sk_buff *skb) 982 { 983 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4, 984 ip4ip6_dscp_ecn_decapsulate); 985 } 986 987 static int ip6ip6_rcv(struct sk_buff *skb) 988 { 989 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6, 990 ip6ip6_dscp_ecn_decapsulate); 991 } 992 993 static int mplsip6_rcv(struct sk_buff *skb) 994 { 995 return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls, 996 mplsip6_dscp_ecn_decapsulate); 997 } 998 999 struct ipv6_tel_txoption { 1000 struct ipv6_txoptions ops; 1001 __u8 dst_opt[8]; 1002 }; 1003 1004 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 1005 { 1006 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 1007 1008 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 1009 opt->dst_opt[3] = 1; 1010 opt->dst_opt[4] = encap_limit; 1011 opt->dst_opt[5] = IPV6_TLV_PADN; 1012 opt->dst_opt[6] = 1; 1013 1014 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt; 1015 opt->ops.opt_nflen = 8; 1016 } 1017 1018 /** 1019 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 1020 * @t: the outgoing tunnel device 1021 * @hdr: IPv6 header from the incoming packet 1022 * 1023 * Description: 1024 * Avoid trivial tunneling loop by checking that tunnel exit-point 1025 * doesn't match source of incoming packet. 1026 * 1027 * Return: 1028 * 1 if conflict, 1029 * 0 else 1030 **/ 1031 1032 static inline bool 1033 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 1034 { 1035 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 1036 } 1037 1038 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, 1039 const struct in6_addr *laddr, 1040 const struct in6_addr *raddr) 1041 { 1042 struct __ip6_tnl_parm *p = &t->parms; 1043 int ret = 0; 1044 struct net *net = t->net; 1045 1046 if (t->parms.collect_md) 1047 return 1; 1048 1049 if ((p->flags & IP6_TNL_F_CAP_XMIT) || 1050 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 1051 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) { 1052 struct net_device *ldev = NULL; 1053 1054 rcu_read_lock(); 1055 if (p->link) 1056 ldev = dev_get_by_index_rcu(net, p->link); 1057 1058 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false, 1059 0, IFA_F_TENTATIVE))) 1060 pr_warn("%s xmit: Local address not yet configured!\n", 1061 p->name); 1062 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) && 1063 !ipv6_addr_is_multicast(raddr) && 1064 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev, 1065 true, 0, IFA_F_TENTATIVE))) 1066 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 1067 p->name); 1068 else 1069 ret = 1; 1070 rcu_read_unlock(); 1071 } 1072 return ret; 1073 } 1074 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 1075 1076 /** 1077 * ip6_tnl_xmit - encapsulate packet and send 1078 * @skb: the outgoing socket buffer 1079 * @dev: the outgoing tunnel device 1080 * @dsfield: dscp code for outer header 1081 * @fl6: flow of tunneled packet 1082 * @encap_limit: encapsulation limit 1083 * @pmtu: Path MTU is stored if packet is too big 1084 * @proto: next header value 1085 * 1086 * Description: 1087 * Build new header and do some sanity checks on the packet before sending 1088 * it. 1089 * 1090 * Return: 1091 * 0 on success 1092 * -1 fail 1093 * %-EMSGSIZE message too big. return mtu in this case. 1094 **/ 1095 1096 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield, 1097 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, 1098 __u8 proto) 1099 { 1100 struct ip6_tnl *t = netdev_priv(dev); 1101 struct net *net = t->net; 1102 struct net_device_stats *stats = &t->dev->stats; 1103 struct ipv6hdr *ipv6h; 1104 struct ipv6_tel_txoption opt; 1105 struct dst_entry *dst = NULL, *ndst = NULL; 1106 struct net_device *tdev; 1107 int mtu; 1108 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0; 1109 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen; 1110 unsigned int max_headroom = psh_hlen; 1111 bool use_cache = false; 1112 u8 hop_limit; 1113 int err = -1; 1114 1115 if (t->parms.collect_md) { 1116 hop_limit = skb_tunnel_info(skb)->key.ttl; 1117 goto route_lookup; 1118 } else { 1119 hop_limit = t->parms.hop_limit; 1120 } 1121 1122 /* NBMA tunnel */ 1123 if (ipv6_addr_any(&t->parms.raddr)) { 1124 if (skb->protocol == htons(ETH_P_IPV6)) { 1125 struct in6_addr *addr6; 1126 struct neighbour *neigh; 1127 int addr_type; 1128 1129 if (!skb_dst(skb)) 1130 goto tx_err_link_failure; 1131 1132 neigh = dst_neigh_lookup(skb_dst(skb), 1133 &ipv6_hdr(skb)->daddr); 1134 if (!neigh) 1135 goto tx_err_link_failure; 1136 1137 addr6 = (struct in6_addr *)&neigh->primary_key; 1138 addr_type = ipv6_addr_type(addr6); 1139 1140 if (addr_type == IPV6_ADDR_ANY) 1141 addr6 = &ipv6_hdr(skb)->daddr; 1142 1143 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr)); 1144 neigh_release(neigh); 1145 } 1146 } else if (t->parms.proto != 0 && !(t->parms.flags & 1147 (IP6_TNL_F_USE_ORIG_TCLASS | 1148 IP6_TNL_F_USE_ORIG_FWMARK))) { 1149 /* enable the cache only if neither the outer protocol nor the 1150 * routing decision depends on the current inner header value 1151 */ 1152 use_cache = true; 1153 } 1154 1155 if (use_cache) 1156 dst = dst_cache_get(&t->dst_cache); 1157 1158 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1159 goto tx_err_link_failure; 1160 1161 if (!dst) { 1162 route_lookup: 1163 /* add dsfield to flowlabel for route lookup */ 1164 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1165 1166 dst = ip6_route_output(net, NULL, fl6); 1167 1168 if (dst->error) 1169 goto tx_err_link_failure; 1170 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1171 if (IS_ERR(dst)) { 1172 err = PTR_ERR(dst); 1173 dst = NULL; 1174 goto tx_err_link_failure; 1175 } 1176 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) && 1177 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1178 &fl6->daddr, 0, &fl6->saddr)) 1179 goto tx_err_link_failure; 1180 ndst = dst; 1181 } 1182 1183 tdev = dst->dev; 1184 1185 if (tdev == dev) { 1186 stats->collisions++; 1187 net_warn_ratelimited("%s: Local routing loop detected!\n", 1188 t->parms.name); 1189 goto tx_err_dst_release; 1190 } 1191 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen; 1192 if (encap_limit >= 0) { 1193 max_headroom += 8; 1194 mtu -= 8; 1195 } 1196 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ? 1197 IPV6_MIN_MTU : IPV4_MIN_MTU); 1198 1199 skb_dst_update_pmtu_no_confirm(skb, mtu); 1200 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) { 1201 *pmtu = mtu; 1202 err = -EMSGSIZE; 1203 goto tx_err_dst_release; 1204 } 1205 1206 if (t->err_count > 0) { 1207 if (time_before(jiffies, 1208 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1209 t->err_count--; 1210 1211 dst_link_failure(skb); 1212 } else { 1213 t->err_count = 0; 1214 } 1215 } 1216 1217 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1218 1219 /* 1220 * Okay, now see if we can stuff it in the buffer as-is. 1221 */ 1222 max_headroom += LL_RESERVED_SPACE(tdev); 1223 1224 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1225 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1226 struct sk_buff *new_skb; 1227 1228 new_skb = skb_realloc_headroom(skb, max_headroom); 1229 if (!new_skb) 1230 goto tx_err_dst_release; 1231 1232 if (skb->sk) 1233 skb_set_owner_w(new_skb, skb->sk); 1234 consume_skb(skb); 1235 skb = new_skb; 1236 } 1237 1238 if (t->parms.collect_md) { 1239 if (t->encap.type != TUNNEL_ENCAP_NONE) 1240 goto tx_err_dst_release; 1241 } else { 1242 if (use_cache && ndst) 1243 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1244 } 1245 skb_dst_set(skb, dst); 1246 1247 if (hop_limit == 0) { 1248 if (skb->protocol == htons(ETH_P_IP)) 1249 hop_limit = ip_hdr(skb)->ttl; 1250 else if (skb->protocol == htons(ETH_P_IPV6)) 1251 hop_limit = ipv6_hdr(skb)->hop_limit; 1252 else 1253 hop_limit = ip6_dst_hoplimit(dst); 1254 } 1255 1256 /* Calculate max headroom for all the headers and adjust 1257 * needed_headroom if necessary. 1258 */ 1259 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1260 + dst->header_len + t->hlen; 1261 if (max_headroom > dev->needed_headroom) 1262 dev->needed_headroom = max_headroom; 1263 1264 err = ip6_tnl_encap(skb, t, &proto, fl6); 1265 if (err) 1266 return err; 1267 1268 if (encap_limit >= 0) { 1269 init_tel_txopt(&opt, encap_limit); 1270 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1271 } 1272 1273 skb_set_inner_ipproto(skb, proto); 1274 1275 skb_push(skb, sizeof(struct ipv6hdr)); 1276 skb_reset_network_header(skb); 1277 ipv6h = ipv6_hdr(skb); 1278 ip6_flow_hdr(ipv6h, dsfield, 1279 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1280 ipv6h->hop_limit = hop_limit; 1281 ipv6h->nexthdr = proto; 1282 ipv6h->saddr = fl6->saddr; 1283 ipv6h->daddr = fl6->daddr; 1284 ip6tunnel_xmit(NULL, skb, dev); 1285 return 0; 1286 tx_err_link_failure: 1287 stats->tx_carrier_errors++; 1288 dst_link_failure(skb); 1289 tx_err_dst_release: 1290 dst_release(dst); 1291 return err; 1292 } 1293 EXPORT_SYMBOL(ip6_tnl_xmit); 1294 1295 static inline int 1296 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, 1297 u8 protocol) 1298 { 1299 struct ip6_tnl *t = netdev_priv(dev); 1300 struct ipv6hdr *ipv6h; 1301 const struct iphdr *iph; 1302 int encap_limit = -1; 1303 __u16 offset; 1304 struct flowi6 fl6; 1305 __u8 dsfield, orig_dsfield; 1306 __u32 mtu; 1307 u8 tproto; 1308 int err; 1309 1310 tproto = READ_ONCE(t->parms.proto); 1311 if (tproto != protocol && tproto != 0) 1312 return -1; 1313 1314 if (t->parms.collect_md) { 1315 struct ip_tunnel_info *tun_info; 1316 const struct ip_tunnel_key *key; 1317 1318 tun_info = skb_tunnel_info(skb); 1319 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1320 ip_tunnel_info_af(tun_info) != AF_INET6)) 1321 return -1; 1322 key = &tun_info->key; 1323 memset(&fl6, 0, sizeof(fl6)); 1324 fl6.flowi6_proto = protocol; 1325 fl6.saddr = key->u.ipv6.src; 1326 fl6.daddr = key->u.ipv6.dst; 1327 fl6.flowlabel = key->label; 1328 dsfield = key->tos; 1329 switch (protocol) { 1330 case IPPROTO_IPIP: 1331 iph = ip_hdr(skb); 1332 orig_dsfield = ipv4_get_dsfield(iph); 1333 break; 1334 case IPPROTO_IPV6: 1335 ipv6h = ipv6_hdr(skb); 1336 orig_dsfield = ipv6_get_dsfield(ipv6h); 1337 break; 1338 default: 1339 orig_dsfield = dsfield; 1340 break; 1341 } 1342 } else { 1343 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1344 encap_limit = t->parms.encap_limit; 1345 if (protocol == IPPROTO_IPV6) { 1346 offset = ip6_tnl_parse_tlv_enc_lim(skb, 1347 skb_network_header(skb)); 1348 /* ip6_tnl_parse_tlv_enc_lim() might have 1349 * reallocated skb->head 1350 */ 1351 if (offset > 0) { 1352 struct ipv6_tlv_tnl_enc_lim *tel; 1353 1354 tel = (void *)&skb_network_header(skb)[offset]; 1355 if (tel->encap_limit == 0) { 1356 icmpv6_send(skb, ICMPV6_PARAMPROB, 1357 ICMPV6_HDR_FIELD, offset + 2); 1358 return -1; 1359 } 1360 encap_limit = tel->encap_limit - 1; 1361 } 1362 } 1363 1364 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1365 fl6.flowi6_proto = protocol; 1366 1367 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1368 fl6.flowi6_mark = skb->mark; 1369 else 1370 fl6.flowi6_mark = t->parms.fwmark; 1371 switch (protocol) { 1372 case IPPROTO_IPIP: 1373 iph = ip_hdr(skb); 1374 orig_dsfield = ipv4_get_dsfield(iph); 1375 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1376 dsfield = orig_dsfield; 1377 else 1378 dsfield = ip6_tclass(t->parms.flowinfo); 1379 break; 1380 case IPPROTO_IPV6: 1381 ipv6h = ipv6_hdr(skb); 1382 orig_dsfield = ipv6_get_dsfield(ipv6h); 1383 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1384 dsfield = orig_dsfield; 1385 else 1386 dsfield = ip6_tclass(t->parms.flowinfo); 1387 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1388 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1389 break; 1390 default: 1391 orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo); 1392 break; 1393 } 1394 } 1395 1396 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1397 dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield); 1398 1399 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1400 return -1; 1401 1402 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1403 protocol); 1404 if (err != 0) { 1405 /* XXX: send ICMP error even if DF is not set. */ 1406 if (err == -EMSGSIZE) 1407 switch (protocol) { 1408 case IPPROTO_IPIP: 1409 icmp_send(skb, ICMP_DEST_UNREACH, 1410 ICMP_FRAG_NEEDED, htonl(mtu)); 1411 break; 1412 case IPPROTO_IPV6: 1413 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1414 break; 1415 default: 1416 break; 1417 } 1418 return -1; 1419 } 1420 1421 return 0; 1422 } 1423 1424 static netdev_tx_t 1425 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1426 { 1427 struct ip6_tnl *t = netdev_priv(dev); 1428 struct net_device_stats *stats = &t->dev->stats; 1429 u8 ipproto; 1430 int ret; 1431 1432 if (!pskb_inet_may_pull(skb)) 1433 goto tx_err; 1434 1435 switch (skb->protocol) { 1436 case htons(ETH_P_IP): 1437 ipproto = IPPROTO_IPIP; 1438 break; 1439 case htons(ETH_P_IPV6): 1440 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb))) 1441 goto tx_err; 1442 ipproto = IPPROTO_IPV6; 1443 break; 1444 case htons(ETH_P_MPLS_UC): 1445 ipproto = IPPROTO_MPLS; 1446 break; 1447 default: 1448 goto tx_err; 1449 } 1450 1451 ret = ipxip6_tnl_xmit(skb, dev, ipproto); 1452 if (ret < 0) 1453 goto tx_err; 1454 1455 return NETDEV_TX_OK; 1456 1457 tx_err: 1458 stats->tx_errors++; 1459 stats->tx_dropped++; 1460 kfree_skb(skb); 1461 return NETDEV_TX_OK; 1462 } 1463 1464 static void ip6_tnl_link_config(struct ip6_tnl *t) 1465 { 1466 struct net_device *dev = t->dev; 1467 struct net_device *tdev = NULL; 1468 struct __ip6_tnl_parm *p = &t->parms; 1469 struct flowi6 *fl6 = &t->fl.u.ip6; 1470 unsigned int mtu; 1471 int t_hlen; 1472 1473 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1474 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1475 1476 /* Set up flowi template */ 1477 fl6->saddr = p->laddr; 1478 fl6->daddr = p->raddr; 1479 fl6->flowi6_oif = p->link; 1480 fl6->flowlabel = 0; 1481 1482 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1483 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1484 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1485 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1486 1487 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1488 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1489 1490 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1491 dev->flags |= IFF_POINTOPOINT; 1492 else 1493 dev->flags &= ~IFF_POINTOPOINT; 1494 1495 t->tun_hlen = 0; 1496 t->hlen = t->encap_hlen + t->tun_hlen; 1497 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1498 1499 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1500 int strict = (ipv6_addr_type(&p->raddr) & 1501 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1502 1503 struct rt6_info *rt = rt6_lookup(t->net, 1504 &p->raddr, &p->laddr, 1505 p->link, NULL, strict); 1506 if (rt) { 1507 tdev = rt->dst.dev; 1508 ip6_rt_put(rt); 1509 } 1510 1511 if (!tdev && p->link) 1512 tdev = __dev_get_by_index(t->net, p->link); 1513 1514 if (tdev) { 1515 dev->hard_header_len = tdev->hard_header_len + t_hlen; 1516 mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU); 1517 1518 dev->mtu = mtu - t_hlen; 1519 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1520 dev->mtu -= 8; 1521 1522 if (dev->mtu < IPV6_MIN_MTU) 1523 dev->mtu = IPV6_MIN_MTU; 1524 } 1525 } 1526 } 1527 1528 /** 1529 * ip6_tnl_change - update the tunnel parameters 1530 * @t: tunnel to be changed 1531 * @p: tunnel configuration parameters 1532 * 1533 * Description: 1534 * ip6_tnl_change() updates the tunnel parameters 1535 **/ 1536 1537 static int 1538 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1539 { 1540 t->parms.laddr = p->laddr; 1541 t->parms.raddr = p->raddr; 1542 t->parms.flags = p->flags; 1543 t->parms.hop_limit = p->hop_limit; 1544 t->parms.encap_limit = p->encap_limit; 1545 t->parms.flowinfo = p->flowinfo; 1546 t->parms.link = p->link; 1547 t->parms.proto = p->proto; 1548 t->parms.fwmark = p->fwmark; 1549 dst_cache_reset(&t->dst_cache); 1550 ip6_tnl_link_config(t); 1551 return 0; 1552 } 1553 1554 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1555 { 1556 struct net *net = t->net; 1557 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1558 int err; 1559 1560 ip6_tnl_unlink(ip6n, t); 1561 synchronize_net(); 1562 err = ip6_tnl_change(t, p); 1563 ip6_tnl_link(ip6n, t); 1564 netdev_state_change(t->dev); 1565 return err; 1566 } 1567 1568 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1569 { 1570 /* for default tnl0 device allow to change only the proto */ 1571 t->parms.proto = p->proto; 1572 netdev_state_change(t->dev); 1573 return 0; 1574 } 1575 1576 static void 1577 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1578 { 1579 p->laddr = u->laddr; 1580 p->raddr = u->raddr; 1581 p->flags = u->flags; 1582 p->hop_limit = u->hop_limit; 1583 p->encap_limit = u->encap_limit; 1584 p->flowinfo = u->flowinfo; 1585 p->link = u->link; 1586 p->proto = u->proto; 1587 memcpy(p->name, u->name, sizeof(u->name)); 1588 } 1589 1590 static void 1591 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1592 { 1593 u->laddr = p->laddr; 1594 u->raddr = p->raddr; 1595 u->flags = p->flags; 1596 u->hop_limit = p->hop_limit; 1597 u->encap_limit = p->encap_limit; 1598 u->flowinfo = p->flowinfo; 1599 u->link = p->link; 1600 u->proto = p->proto; 1601 memcpy(u->name, p->name, sizeof(u->name)); 1602 } 1603 1604 /** 1605 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1606 * @dev: virtual device associated with tunnel 1607 * @ifr: parameters passed from userspace 1608 * @cmd: command to be performed 1609 * 1610 * Description: 1611 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1612 * from userspace. 1613 * 1614 * The possible commands are the following: 1615 * %SIOCGETTUNNEL: get tunnel parameters for device 1616 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1617 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1618 * %SIOCDELTUNNEL: delete tunnel 1619 * 1620 * The fallback device "ip6tnl0", created during module 1621 * initialization, can be used for creating other tunnel devices. 1622 * 1623 * Return: 1624 * 0 on success, 1625 * %-EFAULT if unable to copy data to or from userspace, 1626 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1627 * %-EINVAL if passed tunnel parameters are invalid, 1628 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1629 * %-ENODEV if attempting to change or delete a nonexisting device 1630 **/ 1631 1632 static int 1633 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1634 { 1635 int err = 0; 1636 struct ip6_tnl_parm p; 1637 struct __ip6_tnl_parm p1; 1638 struct ip6_tnl *t = netdev_priv(dev); 1639 struct net *net = t->net; 1640 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1641 1642 memset(&p1, 0, sizeof(p1)); 1643 1644 switch (cmd) { 1645 case SIOCGETTUNNEL: 1646 if (dev == ip6n->fb_tnl_dev) { 1647 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1648 err = -EFAULT; 1649 break; 1650 } 1651 ip6_tnl_parm_from_user(&p1, &p); 1652 t = ip6_tnl_locate(net, &p1, 0); 1653 if (IS_ERR(t)) 1654 t = netdev_priv(dev); 1655 } else { 1656 memset(&p, 0, sizeof(p)); 1657 } 1658 ip6_tnl_parm_to_user(&p, &t->parms); 1659 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) { 1660 err = -EFAULT; 1661 } 1662 break; 1663 case SIOCADDTUNNEL: 1664 case SIOCCHGTUNNEL: 1665 err = -EPERM; 1666 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1667 break; 1668 err = -EFAULT; 1669 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1670 break; 1671 err = -EINVAL; 1672 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1673 p.proto != 0) 1674 break; 1675 ip6_tnl_parm_from_user(&p1, &p); 1676 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1677 if (cmd == SIOCCHGTUNNEL) { 1678 if (!IS_ERR(t)) { 1679 if (t->dev != dev) { 1680 err = -EEXIST; 1681 break; 1682 } 1683 } else 1684 t = netdev_priv(dev); 1685 if (dev == ip6n->fb_tnl_dev) 1686 err = ip6_tnl0_update(t, &p1); 1687 else 1688 err = ip6_tnl_update(t, &p1); 1689 } 1690 if (!IS_ERR(t)) { 1691 err = 0; 1692 ip6_tnl_parm_to_user(&p, &t->parms); 1693 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1694 err = -EFAULT; 1695 1696 } else { 1697 err = PTR_ERR(t); 1698 } 1699 break; 1700 case SIOCDELTUNNEL: 1701 err = -EPERM; 1702 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1703 break; 1704 1705 if (dev == ip6n->fb_tnl_dev) { 1706 err = -EFAULT; 1707 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1708 break; 1709 err = -ENOENT; 1710 ip6_tnl_parm_from_user(&p1, &p); 1711 t = ip6_tnl_locate(net, &p1, 0); 1712 if (IS_ERR(t)) 1713 break; 1714 err = -EPERM; 1715 if (t->dev == ip6n->fb_tnl_dev) 1716 break; 1717 dev = t->dev; 1718 } 1719 err = 0; 1720 unregister_netdevice(dev); 1721 break; 1722 default: 1723 err = -EINVAL; 1724 } 1725 return err; 1726 } 1727 1728 /** 1729 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1730 * @dev: virtual device associated with tunnel 1731 * @new_mtu: the new mtu 1732 * 1733 * Return: 1734 * 0 on success, 1735 * %-EINVAL if mtu too small 1736 **/ 1737 1738 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1739 { 1740 struct ip6_tnl *tnl = netdev_priv(dev); 1741 1742 if (tnl->parms.proto == IPPROTO_IPV6) { 1743 if (new_mtu < IPV6_MIN_MTU) 1744 return -EINVAL; 1745 } else { 1746 if (new_mtu < ETH_MIN_MTU) 1747 return -EINVAL; 1748 } 1749 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) { 1750 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len) 1751 return -EINVAL; 1752 } else { 1753 if (new_mtu > IP_MAX_MTU - dev->hard_header_len) 1754 return -EINVAL; 1755 } 1756 dev->mtu = new_mtu; 1757 return 0; 1758 } 1759 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1760 1761 int ip6_tnl_get_iflink(const struct net_device *dev) 1762 { 1763 struct ip6_tnl *t = netdev_priv(dev); 1764 1765 return t->parms.link; 1766 } 1767 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1768 1769 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1770 unsigned int num) 1771 { 1772 if (num >= MAX_IPTUN_ENCAP_OPS) 1773 return -ERANGE; 1774 1775 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1776 &ip6tun_encaps[num], 1777 NULL, ops) ? 0 : -1; 1778 } 1779 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1780 1781 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1782 unsigned int num) 1783 { 1784 int ret; 1785 1786 if (num >= MAX_IPTUN_ENCAP_OPS) 1787 return -ERANGE; 1788 1789 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1790 &ip6tun_encaps[num], 1791 ops, NULL) == ops) ? 0 : -1; 1792 1793 synchronize_net(); 1794 1795 return ret; 1796 } 1797 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1798 1799 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1800 struct ip_tunnel_encap *ipencap) 1801 { 1802 int hlen; 1803 1804 memset(&t->encap, 0, sizeof(t->encap)); 1805 1806 hlen = ip6_encap_hlen(ipencap); 1807 if (hlen < 0) 1808 return hlen; 1809 1810 t->encap.type = ipencap->type; 1811 t->encap.sport = ipencap->sport; 1812 t->encap.dport = ipencap->dport; 1813 t->encap.flags = ipencap->flags; 1814 1815 t->encap_hlen = hlen; 1816 t->hlen = t->encap_hlen + t->tun_hlen; 1817 1818 return 0; 1819 } 1820 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1821 1822 static const struct net_device_ops ip6_tnl_netdev_ops = { 1823 .ndo_init = ip6_tnl_dev_init, 1824 .ndo_uninit = ip6_tnl_dev_uninit, 1825 .ndo_start_xmit = ip6_tnl_start_xmit, 1826 .ndo_do_ioctl = ip6_tnl_ioctl, 1827 .ndo_change_mtu = ip6_tnl_change_mtu, 1828 .ndo_get_stats = ip6_get_stats, 1829 .ndo_get_iflink = ip6_tnl_get_iflink, 1830 }; 1831 1832 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1833 NETIF_F_FRAGLIST | \ 1834 NETIF_F_HIGHDMA | \ 1835 NETIF_F_GSO_SOFTWARE | \ 1836 NETIF_F_HW_CSUM) 1837 1838 /** 1839 * ip6_tnl_dev_setup - setup virtual tunnel device 1840 * @dev: virtual device associated with tunnel 1841 * 1842 * Description: 1843 * Initialize function pointers and device parameters 1844 **/ 1845 1846 static void ip6_tnl_dev_setup(struct net_device *dev) 1847 { 1848 dev->netdev_ops = &ip6_tnl_netdev_ops; 1849 dev->needs_free_netdev = true; 1850 dev->priv_destructor = ip6_dev_free; 1851 1852 dev->type = ARPHRD_TUNNEL6; 1853 dev->flags |= IFF_NOARP; 1854 dev->addr_len = sizeof(struct in6_addr); 1855 dev->features |= NETIF_F_LLTX; 1856 netif_keep_dst(dev); 1857 1858 dev->features |= IPXIPX_FEATURES; 1859 dev->hw_features |= IPXIPX_FEATURES; 1860 1861 /* This perm addr will be used as interface identifier by IPv6 */ 1862 dev->addr_assign_type = NET_ADDR_RANDOM; 1863 eth_random_addr(dev->perm_addr); 1864 } 1865 1866 1867 /** 1868 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1869 * @dev: virtual device associated with tunnel 1870 **/ 1871 1872 static inline int 1873 ip6_tnl_dev_init_gen(struct net_device *dev) 1874 { 1875 struct ip6_tnl *t = netdev_priv(dev); 1876 int ret; 1877 int t_hlen; 1878 1879 t->dev = dev; 1880 t->net = dev_net(dev); 1881 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1882 if (!dev->tstats) 1883 return -ENOMEM; 1884 1885 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1886 if (ret) 1887 goto free_stats; 1888 1889 ret = gro_cells_init(&t->gro_cells, dev); 1890 if (ret) 1891 goto destroy_dst; 1892 1893 t->tun_hlen = 0; 1894 t->hlen = t->encap_hlen + t->tun_hlen; 1895 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1896 1897 dev->type = ARPHRD_TUNNEL6; 1898 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1899 dev->mtu = ETH_DATA_LEN - t_hlen; 1900 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1901 dev->mtu -= 8; 1902 dev->min_mtu = ETH_MIN_MTU; 1903 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len; 1904 1905 return 0; 1906 1907 destroy_dst: 1908 dst_cache_destroy(&t->dst_cache); 1909 free_stats: 1910 free_percpu(dev->tstats); 1911 dev->tstats = NULL; 1912 1913 return ret; 1914 } 1915 1916 /** 1917 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1918 * @dev: virtual device associated with tunnel 1919 **/ 1920 1921 static int ip6_tnl_dev_init(struct net_device *dev) 1922 { 1923 struct ip6_tnl *t = netdev_priv(dev); 1924 int err = ip6_tnl_dev_init_gen(dev); 1925 1926 if (err) 1927 return err; 1928 ip6_tnl_link_config(t); 1929 if (t->parms.collect_md) 1930 netif_keep_dst(dev); 1931 return 0; 1932 } 1933 1934 /** 1935 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1936 * @dev: fallback device 1937 * 1938 * Return: 0 1939 **/ 1940 1941 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1942 { 1943 struct ip6_tnl *t = netdev_priv(dev); 1944 struct net *net = dev_net(dev); 1945 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1946 1947 t->parms.proto = IPPROTO_IPV6; 1948 dev_hold(dev); 1949 1950 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1951 return 0; 1952 } 1953 1954 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[], 1955 struct netlink_ext_ack *extack) 1956 { 1957 u8 proto; 1958 1959 if (!data || !data[IFLA_IPTUN_PROTO]) 1960 return 0; 1961 1962 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1963 if (proto != IPPROTO_IPV6 && 1964 proto != IPPROTO_IPIP && 1965 proto != 0) 1966 return -EINVAL; 1967 1968 return 0; 1969 } 1970 1971 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1972 struct __ip6_tnl_parm *parms) 1973 { 1974 memset(parms, 0, sizeof(*parms)); 1975 1976 if (!data) 1977 return; 1978 1979 if (data[IFLA_IPTUN_LINK]) 1980 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1981 1982 if (data[IFLA_IPTUN_LOCAL]) 1983 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1984 1985 if (data[IFLA_IPTUN_REMOTE]) 1986 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1987 1988 if (data[IFLA_IPTUN_TTL]) 1989 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1990 1991 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1992 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1993 1994 if (data[IFLA_IPTUN_FLOWINFO]) 1995 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1996 1997 if (data[IFLA_IPTUN_FLAGS]) 1998 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1999 2000 if (data[IFLA_IPTUN_PROTO]) 2001 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 2002 2003 if (data[IFLA_IPTUN_COLLECT_METADATA]) 2004 parms->collect_md = true; 2005 2006 if (data[IFLA_IPTUN_FWMARK]) 2007 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 2008 } 2009 2010 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[], 2011 struct ip_tunnel_encap *ipencap) 2012 { 2013 bool ret = false; 2014 2015 memset(ipencap, 0, sizeof(*ipencap)); 2016 2017 if (!data) 2018 return ret; 2019 2020 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 2021 ret = true; 2022 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 2023 } 2024 2025 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 2026 ret = true; 2027 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 2028 } 2029 2030 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 2031 ret = true; 2032 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 2033 } 2034 2035 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 2036 ret = true; 2037 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 2038 } 2039 2040 return ret; 2041 } 2042 2043 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 2044 struct nlattr *tb[], struct nlattr *data[], 2045 struct netlink_ext_ack *extack) 2046 { 2047 struct net *net = dev_net(dev); 2048 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2049 struct ip_tunnel_encap ipencap; 2050 struct ip6_tnl *nt, *t; 2051 int err; 2052 2053 nt = netdev_priv(dev); 2054 2055 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2056 err = ip6_tnl_encap_setup(nt, &ipencap); 2057 if (err < 0) 2058 return err; 2059 } 2060 2061 ip6_tnl_netlink_parms(data, &nt->parms); 2062 2063 if (nt->parms.collect_md) { 2064 if (rtnl_dereference(ip6n->collect_md_tun)) 2065 return -EEXIST; 2066 } else { 2067 t = ip6_tnl_locate(net, &nt->parms, 0); 2068 if (!IS_ERR(t)) 2069 return -EEXIST; 2070 } 2071 2072 err = ip6_tnl_create2(dev); 2073 if (!err && tb[IFLA_MTU]) 2074 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2075 2076 return err; 2077 } 2078 2079 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2080 struct nlattr *data[], 2081 struct netlink_ext_ack *extack) 2082 { 2083 struct ip6_tnl *t = netdev_priv(dev); 2084 struct __ip6_tnl_parm p; 2085 struct net *net = t->net; 2086 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2087 struct ip_tunnel_encap ipencap; 2088 2089 if (dev == ip6n->fb_tnl_dev) 2090 return -EINVAL; 2091 2092 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2093 int err = ip6_tnl_encap_setup(t, &ipencap); 2094 2095 if (err < 0) 2096 return err; 2097 } 2098 ip6_tnl_netlink_parms(data, &p); 2099 if (p.collect_md) 2100 return -EINVAL; 2101 2102 t = ip6_tnl_locate(net, &p, 0); 2103 if (!IS_ERR(t)) { 2104 if (t->dev != dev) 2105 return -EEXIST; 2106 } else 2107 t = netdev_priv(dev); 2108 2109 return ip6_tnl_update(t, &p); 2110 } 2111 2112 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2113 { 2114 struct net *net = dev_net(dev); 2115 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2116 2117 if (dev != ip6n->fb_tnl_dev) 2118 unregister_netdevice_queue(dev, head); 2119 } 2120 2121 static size_t ip6_tnl_get_size(const struct net_device *dev) 2122 { 2123 return 2124 /* IFLA_IPTUN_LINK */ 2125 nla_total_size(4) + 2126 /* IFLA_IPTUN_LOCAL */ 2127 nla_total_size(sizeof(struct in6_addr)) + 2128 /* IFLA_IPTUN_REMOTE */ 2129 nla_total_size(sizeof(struct in6_addr)) + 2130 /* IFLA_IPTUN_TTL */ 2131 nla_total_size(1) + 2132 /* IFLA_IPTUN_ENCAP_LIMIT */ 2133 nla_total_size(1) + 2134 /* IFLA_IPTUN_FLOWINFO */ 2135 nla_total_size(4) + 2136 /* IFLA_IPTUN_FLAGS */ 2137 nla_total_size(4) + 2138 /* IFLA_IPTUN_PROTO */ 2139 nla_total_size(1) + 2140 /* IFLA_IPTUN_ENCAP_TYPE */ 2141 nla_total_size(2) + 2142 /* IFLA_IPTUN_ENCAP_FLAGS */ 2143 nla_total_size(2) + 2144 /* IFLA_IPTUN_ENCAP_SPORT */ 2145 nla_total_size(2) + 2146 /* IFLA_IPTUN_ENCAP_DPORT */ 2147 nla_total_size(2) + 2148 /* IFLA_IPTUN_COLLECT_METADATA */ 2149 nla_total_size(0) + 2150 /* IFLA_IPTUN_FWMARK */ 2151 nla_total_size(4) + 2152 0; 2153 } 2154 2155 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2156 { 2157 struct ip6_tnl *tunnel = netdev_priv(dev); 2158 struct __ip6_tnl_parm *parm = &tunnel->parms; 2159 2160 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2161 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2162 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2163 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2164 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2165 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2166 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2167 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2168 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2169 goto nla_put_failure; 2170 2171 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2172 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2173 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2174 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2175 goto nla_put_failure; 2176 2177 if (parm->collect_md) 2178 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2179 goto nla_put_failure; 2180 2181 return 0; 2182 2183 nla_put_failure: 2184 return -EMSGSIZE; 2185 } 2186 2187 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2188 { 2189 struct ip6_tnl *tunnel = netdev_priv(dev); 2190 2191 return tunnel->net; 2192 } 2193 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2194 2195 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2196 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2197 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2198 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2199 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2200 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2201 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2202 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2203 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2204 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2205 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2206 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2207 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2208 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2209 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2210 }; 2211 2212 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2213 .kind = "ip6tnl", 2214 .maxtype = IFLA_IPTUN_MAX, 2215 .policy = ip6_tnl_policy, 2216 .priv_size = sizeof(struct ip6_tnl), 2217 .setup = ip6_tnl_dev_setup, 2218 .validate = ip6_tnl_validate, 2219 .newlink = ip6_tnl_newlink, 2220 .changelink = ip6_tnl_changelink, 2221 .dellink = ip6_tnl_dellink, 2222 .get_size = ip6_tnl_get_size, 2223 .fill_info = ip6_tnl_fill_info, 2224 .get_link_net = ip6_tnl_get_link_net, 2225 }; 2226 2227 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2228 .handler = ip4ip6_rcv, 2229 .err_handler = ip4ip6_err, 2230 .priority = 1, 2231 }; 2232 2233 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2234 .handler = ip6ip6_rcv, 2235 .err_handler = ip6ip6_err, 2236 .priority = 1, 2237 }; 2238 2239 static struct xfrm6_tunnel mplsip6_handler __read_mostly = { 2240 .handler = mplsip6_rcv, 2241 .err_handler = mplsip6_err, 2242 .priority = 1, 2243 }; 2244 2245 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list) 2246 { 2247 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2248 struct net_device *dev, *aux; 2249 int h; 2250 struct ip6_tnl *t; 2251 2252 for_each_netdev_safe(net, dev, aux) 2253 if (dev->rtnl_link_ops == &ip6_link_ops) 2254 unregister_netdevice_queue(dev, list); 2255 2256 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2257 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2258 while (t) { 2259 /* If dev is in the same netns, it has already 2260 * been added to the list by the previous loop. 2261 */ 2262 if (!net_eq(dev_net(t->dev), net)) 2263 unregister_netdevice_queue(t->dev, list); 2264 t = rtnl_dereference(t->next); 2265 } 2266 } 2267 } 2268 2269 static int __net_init ip6_tnl_init_net(struct net *net) 2270 { 2271 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2272 struct ip6_tnl *t = NULL; 2273 int err; 2274 2275 ip6n->tnls[0] = ip6n->tnls_wc; 2276 ip6n->tnls[1] = ip6n->tnls_r_l; 2277 2278 if (!net_has_fallback_tunnels(net)) 2279 return 0; 2280 err = -ENOMEM; 2281 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2282 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2283 2284 if (!ip6n->fb_tnl_dev) 2285 goto err_alloc_dev; 2286 dev_net_set(ip6n->fb_tnl_dev, net); 2287 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2288 /* FB netdevice is special: we have one, and only one per netns. 2289 * Allowing to move it to another netns is clearly unsafe. 2290 */ 2291 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 2292 2293 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2294 if (err < 0) 2295 goto err_register; 2296 2297 err = register_netdev(ip6n->fb_tnl_dev); 2298 if (err < 0) 2299 goto err_register; 2300 2301 t = netdev_priv(ip6n->fb_tnl_dev); 2302 2303 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2304 return 0; 2305 2306 err_register: 2307 free_netdev(ip6n->fb_tnl_dev); 2308 err_alloc_dev: 2309 return err; 2310 } 2311 2312 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list) 2313 { 2314 struct net *net; 2315 LIST_HEAD(list); 2316 2317 rtnl_lock(); 2318 list_for_each_entry(net, net_list, exit_list) 2319 ip6_tnl_destroy_tunnels(net, &list); 2320 unregister_netdevice_many(&list); 2321 rtnl_unlock(); 2322 } 2323 2324 static struct pernet_operations ip6_tnl_net_ops = { 2325 .init = ip6_tnl_init_net, 2326 .exit_batch = ip6_tnl_exit_batch_net, 2327 .id = &ip6_tnl_net_id, 2328 .size = sizeof(struct ip6_tnl_net), 2329 }; 2330 2331 /** 2332 * ip6_tunnel_init - register protocol and reserve needed resources 2333 * 2334 * Return: 0 on success 2335 **/ 2336 2337 static int __init ip6_tunnel_init(void) 2338 { 2339 int err; 2340 2341 if (!ipv6_mod_enabled()) 2342 return -EOPNOTSUPP; 2343 2344 err = register_pernet_device(&ip6_tnl_net_ops); 2345 if (err < 0) 2346 goto out_pernet; 2347 2348 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2349 if (err < 0) { 2350 pr_err("%s: can't register ip4ip6\n", __func__); 2351 goto out_ip4ip6; 2352 } 2353 2354 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2355 if (err < 0) { 2356 pr_err("%s: can't register ip6ip6\n", __func__); 2357 goto out_ip6ip6; 2358 } 2359 2360 if (ip6_tnl_mpls_supported()) { 2361 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS); 2362 if (err < 0) { 2363 pr_err("%s: can't register mplsip6\n", __func__); 2364 goto out_mplsip6; 2365 } 2366 } 2367 2368 err = rtnl_link_register(&ip6_link_ops); 2369 if (err < 0) 2370 goto rtnl_link_failed; 2371 2372 return 0; 2373 2374 rtnl_link_failed: 2375 if (ip6_tnl_mpls_supported()) 2376 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS); 2377 out_mplsip6: 2378 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2379 out_ip6ip6: 2380 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2381 out_ip4ip6: 2382 unregister_pernet_device(&ip6_tnl_net_ops); 2383 out_pernet: 2384 return err; 2385 } 2386 2387 /** 2388 * ip6_tunnel_cleanup - free resources and unregister protocol 2389 **/ 2390 2391 static void __exit ip6_tunnel_cleanup(void) 2392 { 2393 rtnl_link_unregister(&ip6_link_ops); 2394 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2395 pr_info("%s: can't deregister ip4ip6\n", __func__); 2396 2397 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2398 pr_info("%s: can't deregister ip6ip6\n", __func__); 2399 2400 if (ip6_tnl_mpls_supported() && 2401 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS)) 2402 pr_info("%s: can't deregister mplsip6\n", __func__); 2403 unregister_pernet_device(&ip6_tnl_net_ops); 2404 } 2405 2406 module_init(ip6_tunnel_init); 2407 module_exit(ip6_tunnel_cleanup); 2408