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 #define for_each_ip6_tunnel_rcu(start) \ 128 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 129 130 /** 131 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 132 * @net: network namespace 133 * @link: ifindex of underlying interface 134 * @remote: the address of the tunnel exit-point 135 * @local: the address of the tunnel entry-point 136 * 137 * Return: 138 * tunnel matching given end-points if found, 139 * else fallback tunnel if its device is up, 140 * else %NULL 141 **/ 142 143 static struct ip6_tnl * 144 ip6_tnl_lookup(struct net *net, int link, 145 const struct in6_addr *remote, const struct in6_addr *local) 146 { 147 unsigned int hash = HASH(remote, local); 148 struct ip6_tnl *t, *cand = NULL; 149 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 150 struct in6_addr any; 151 152 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 153 if (!ipv6_addr_equal(local, &t->parms.laddr) || 154 !ipv6_addr_equal(remote, &t->parms.raddr) || 155 !(t->dev->flags & IFF_UP)) 156 continue; 157 158 if (link == t->parms.link) 159 return t; 160 else 161 cand = t; 162 } 163 164 memset(&any, 0, sizeof(any)); 165 hash = HASH(&any, local); 166 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 167 if (!ipv6_addr_equal(local, &t->parms.laddr) || 168 !ipv6_addr_any(&t->parms.raddr) || 169 !(t->dev->flags & IFF_UP)) 170 continue; 171 172 if (link == t->parms.link) 173 return t; 174 else if (!cand) 175 cand = t; 176 } 177 178 hash = HASH(remote, &any); 179 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 180 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 181 !ipv6_addr_any(&t->parms.laddr) || 182 !(t->dev->flags & IFF_UP)) 183 continue; 184 185 if (link == t->parms.link) 186 return t; 187 else if (!cand) 188 cand = t; 189 } 190 191 if (cand) 192 return cand; 193 194 t = rcu_dereference(ip6n->collect_md_tun); 195 if (t && t->dev->flags & IFF_UP) 196 return t; 197 198 t = rcu_dereference(ip6n->tnls_wc[0]); 199 if (t && (t->dev->flags & IFF_UP)) 200 return t; 201 202 return NULL; 203 } 204 205 /** 206 * ip6_tnl_bucket - get head of list matching given tunnel parameters 207 * @p: parameters containing tunnel end-points 208 * 209 * Description: 210 * ip6_tnl_bucket() returns the head of the list matching the 211 * &struct in6_addr entries laddr and raddr in @p. 212 * 213 * Return: head of IPv6 tunnel list 214 **/ 215 216 static struct ip6_tnl __rcu ** 217 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 218 { 219 const struct in6_addr *remote = &p->raddr; 220 const struct in6_addr *local = &p->laddr; 221 unsigned int h = 0; 222 int prio = 0; 223 224 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 225 prio = 1; 226 h = HASH(remote, local); 227 } 228 return &ip6n->tnls[prio][h]; 229 } 230 231 /** 232 * ip6_tnl_link - add tunnel to hash table 233 * @t: tunnel to be added 234 **/ 235 236 static void 237 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 238 { 239 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 240 241 if (t->parms.collect_md) 242 rcu_assign_pointer(ip6n->collect_md_tun, t); 243 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 244 rcu_assign_pointer(*tp, t); 245 } 246 247 /** 248 * ip6_tnl_unlink - remove tunnel from hash table 249 * @t: tunnel to be removed 250 **/ 251 252 static void 253 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 254 { 255 struct ip6_tnl __rcu **tp; 256 struct ip6_tnl *iter; 257 258 if (t->parms.collect_md) 259 rcu_assign_pointer(ip6n->collect_md_tun, NULL); 260 261 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 262 (iter = rtnl_dereference(*tp)) != NULL; 263 tp = &iter->next) { 264 if (t == iter) { 265 rcu_assign_pointer(*tp, t->next); 266 break; 267 } 268 } 269 } 270 271 static void ip6_dev_free(struct net_device *dev) 272 { 273 struct ip6_tnl *t = netdev_priv(dev); 274 275 gro_cells_destroy(&t->gro_cells); 276 dst_cache_destroy(&t->dst_cache); 277 free_percpu(dev->tstats); 278 } 279 280 static int ip6_tnl_create2(struct net_device *dev) 281 { 282 struct ip6_tnl *t = netdev_priv(dev); 283 struct net *net = dev_net(dev); 284 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 285 int err; 286 287 t = netdev_priv(dev); 288 289 dev->rtnl_link_ops = &ip6_link_ops; 290 err = register_netdevice(dev); 291 if (err < 0) 292 goto out; 293 294 strcpy(t->parms.name, dev->name); 295 296 dev_hold(dev); 297 ip6_tnl_link(ip6n, t); 298 return 0; 299 300 out: 301 return err; 302 } 303 304 /** 305 * ip6_tnl_create - create a new tunnel 306 * @net: network namespace 307 * @p: tunnel parameters 308 * 309 * Description: 310 * Create tunnel matching given parameters. 311 * 312 * Return: 313 * created tunnel or error pointer 314 **/ 315 316 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 317 { 318 struct net_device *dev; 319 struct ip6_tnl *t; 320 char name[IFNAMSIZ]; 321 int err = -E2BIG; 322 323 if (p->name[0]) { 324 if (!dev_valid_name(p->name)) 325 goto failed; 326 strlcpy(name, p->name, IFNAMSIZ); 327 } else { 328 sprintf(name, "ip6tnl%%d"); 329 } 330 err = -ENOMEM; 331 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 332 ip6_tnl_dev_setup); 333 if (!dev) 334 goto failed; 335 336 dev_net_set(dev, net); 337 338 t = netdev_priv(dev); 339 t->parms = *p; 340 t->net = dev_net(dev); 341 err = ip6_tnl_create2(dev); 342 if (err < 0) 343 goto failed_free; 344 345 return t; 346 347 failed_free: 348 free_netdev(dev); 349 failed: 350 return ERR_PTR(err); 351 } 352 353 /** 354 * ip6_tnl_locate - find or create tunnel matching given parameters 355 * @net: network namespace 356 * @p: tunnel parameters 357 * @create: != 0 if allowed to create new tunnel if no match found 358 * 359 * Description: 360 * ip6_tnl_locate() first tries to locate an existing tunnel 361 * based on @parms. If this is unsuccessful, but @create is set a new 362 * tunnel device is created and registered for use. 363 * 364 * Return: 365 * matching tunnel or error pointer 366 **/ 367 368 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 369 struct __ip6_tnl_parm *p, int create) 370 { 371 const struct in6_addr *remote = &p->raddr; 372 const struct in6_addr *local = &p->laddr; 373 struct ip6_tnl __rcu **tp; 374 struct ip6_tnl *t; 375 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 376 377 for (tp = ip6_tnl_bucket(ip6n, p); 378 (t = rtnl_dereference(*tp)) != NULL; 379 tp = &t->next) { 380 if (ipv6_addr_equal(local, &t->parms.laddr) && 381 ipv6_addr_equal(remote, &t->parms.raddr) && 382 p->link == t->parms.link) { 383 if (create) 384 return ERR_PTR(-EEXIST); 385 386 return t; 387 } 388 } 389 if (!create) 390 return ERR_PTR(-ENODEV); 391 return ip6_tnl_create(net, p); 392 } 393 394 /** 395 * ip6_tnl_dev_uninit - tunnel device uninitializer 396 * @dev: the device to be destroyed 397 * 398 * Description: 399 * ip6_tnl_dev_uninit() removes tunnel from its list 400 **/ 401 402 static void 403 ip6_tnl_dev_uninit(struct net_device *dev) 404 { 405 struct ip6_tnl *t = netdev_priv(dev); 406 struct net *net = t->net; 407 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 408 409 if (dev == ip6n->fb_tnl_dev) 410 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 411 else 412 ip6_tnl_unlink(ip6n, t); 413 dst_cache_reset(&t->dst_cache); 414 dev_put(dev); 415 } 416 417 /** 418 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 419 * @skb: received socket buffer 420 * 421 * Return: 422 * 0 if none was found, 423 * else index to encapsulation limit 424 **/ 425 426 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 427 { 428 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw; 429 unsigned int nhoff = raw - skb->data; 430 unsigned int off = nhoff + sizeof(*ipv6h); 431 u8 next, nexthdr = ipv6h->nexthdr; 432 433 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 434 struct ipv6_opt_hdr *hdr; 435 u16 optlen; 436 437 if (!pskb_may_pull(skb, off + sizeof(*hdr))) 438 break; 439 440 hdr = (struct ipv6_opt_hdr *)(skb->data + off); 441 if (nexthdr == NEXTHDR_FRAGMENT) { 442 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 443 if (frag_hdr->frag_off) 444 break; 445 optlen = 8; 446 } else if (nexthdr == NEXTHDR_AUTH) { 447 optlen = ipv6_authlen(hdr); 448 } else { 449 optlen = ipv6_optlen(hdr); 450 } 451 /* cache hdr->nexthdr, since pskb_may_pull() might 452 * invalidate hdr 453 */ 454 next = hdr->nexthdr; 455 if (nexthdr == NEXTHDR_DEST) { 456 u16 i = 2; 457 458 /* Remember : hdr is no longer valid at this point. */ 459 if (!pskb_may_pull(skb, off + optlen)) 460 break; 461 462 while (1) { 463 struct ipv6_tlv_tnl_enc_lim *tel; 464 465 /* No more room for encapsulation limit */ 466 if (i + sizeof(*tel) > optlen) 467 break; 468 469 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i); 470 /* return index of option if found and valid */ 471 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 472 tel->length == 1) 473 return i + off - nhoff; 474 /* else jump to next option */ 475 if (tel->type) 476 i += tel->length + 2; 477 else 478 i++; 479 } 480 } 481 nexthdr = next; 482 off += optlen; 483 } 484 return 0; 485 } 486 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 487 488 /** 489 * ip6_tnl_err - tunnel error handler 490 * 491 * Description: 492 * ip6_tnl_err() should handle errors in the tunnel according 493 * to the specifications in RFC 2473. 494 **/ 495 496 static int 497 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 498 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 499 { 500 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data; 501 struct net *net = dev_net(skb->dev); 502 u8 rel_type = ICMPV6_DEST_UNREACH; 503 u8 rel_code = ICMPV6_ADDR_UNREACH; 504 __u32 rel_info = 0; 505 struct ip6_tnl *t; 506 int err = -ENOENT; 507 int rel_msg = 0; 508 u8 tproto; 509 __u16 len; 510 511 /* If the packet doesn't contain the original IPv6 header we are 512 in trouble since we might need the source address for further 513 processing of the error. */ 514 515 rcu_read_lock(); 516 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr); 517 if (!t) 518 goto out; 519 520 tproto = READ_ONCE(t->parms.proto); 521 if (tproto != ipproto && tproto != 0) 522 goto out; 523 524 err = 0; 525 526 switch (*type) { 527 case ICMPV6_DEST_UNREACH: 528 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 529 t->parms.name); 530 rel_msg = 1; 531 break; 532 case ICMPV6_TIME_EXCEED: 533 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 534 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 535 t->parms.name); 536 rel_msg = 1; 537 } 538 break; 539 case ICMPV6_PARAMPROB: { 540 struct ipv6_tlv_tnl_enc_lim *tel; 541 __u32 teli; 542 543 teli = 0; 544 if ((*code) == ICMPV6_HDR_FIELD) 545 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 546 547 if (teli && teli == *info - 2) { 548 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 549 if (tel->encap_limit == 0) { 550 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 551 t->parms.name); 552 rel_msg = 1; 553 } 554 } else { 555 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 556 t->parms.name); 557 } 558 break; 559 } 560 case ICMPV6_PKT_TOOBIG: { 561 __u32 mtu; 562 563 ip6_update_pmtu(skb, net, htonl(*info), 0, 0, 564 sock_net_uid(net, NULL)); 565 mtu = *info - offset; 566 if (mtu < IPV6_MIN_MTU) 567 mtu = IPV6_MIN_MTU; 568 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len); 569 if (len > mtu) { 570 rel_type = ICMPV6_PKT_TOOBIG; 571 rel_code = 0; 572 rel_info = mtu; 573 rel_msg = 1; 574 } 575 break; 576 } 577 case NDISC_REDIRECT: 578 ip6_redirect(skb, net, skb->dev->ifindex, 0, 579 sock_net_uid(net, NULL)); 580 break; 581 } 582 583 *type = rel_type; 584 *code = rel_code; 585 *info = rel_info; 586 *msg = rel_msg; 587 588 out: 589 rcu_read_unlock(); 590 return err; 591 } 592 593 static int 594 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 595 u8 type, u8 code, int offset, __be32 info) 596 { 597 __u32 rel_info = ntohl(info); 598 const struct iphdr *eiph; 599 struct sk_buff *skb2; 600 int err, rel_msg = 0; 601 u8 rel_type = type; 602 u8 rel_code = code; 603 struct rtable *rt; 604 struct flowi4 fl4; 605 606 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 607 &rel_msg, &rel_info, offset); 608 if (err < 0) 609 return err; 610 611 if (rel_msg == 0) 612 return 0; 613 614 switch (rel_type) { 615 case ICMPV6_DEST_UNREACH: 616 if (rel_code != ICMPV6_ADDR_UNREACH) 617 return 0; 618 rel_type = ICMP_DEST_UNREACH; 619 rel_code = ICMP_HOST_UNREACH; 620 break; 621 case ICMPV6_PKT_TOOBIG: 622 if (rel_code != 0) 623 return 0; 624 rel_type = ICMP_DEST_UNREACH; 625 rel_code = ICMP_FRAG_NEEDED; 626 break; 627 default: 628 return 0; 629 } 630 631 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 632 return 0; 633 634 skb2 = skb_clone(skb, GFP_ATOMIC); 635 if (!skb2) 636 return 0; 637 638 skb_dst_drop(skb2); 639 640 skb_pull(skb2, offset); 641 skb_reset_network_header(skb2); 642 eiph = ip_hdr(skb2); 643 644 /* Try to guess incoming interface */ 645 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr, 646 0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 647 if (IS_ERR(rt)) 648 goto out; 649 650 skb2->dev = rt->dst.dev; 651 ip_rt_put(rt); 652 653 /* route "incoming" packet */ 654 if (rt->rt_flags & RTCF_LOCAL) { 655 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 656 eiph->daddr, eiph->saddr, 0, 0, 657 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 658 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) { 659 if (!IS_ERR(rt)) 660 ip_rt_put(rt); 661 goto out; 662 } 663 skb_dst_set(skb2, &rt->dst); 664 } else { 665 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 666 skb2->dev) || 667 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6) 668 goto out; 669 } 670 671 /* change mtu on this route */ 672 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 673 if (rel_info > dst_mtu(skb_dst(skb2))) 674 goto out; 675 676 skb_dst_update_pmtu_no_confirm(skb2, rel_info); 677 } 678 679 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 680 681 out: 682 kfree_skb(skb2); 683 return 0; 684 } 685 686 static int 687 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 688 u8 type, u8 code, int offset, __be32 info) 689 { 690 __u32 rel_info = ntohl(info); 691 int err, rel_msg = 0; 692 u8 rel_type = type; 693 u8 rel_code = code; 694 695 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 696 &rel_msg, &rel_info, offset); 697 if (err < 0) 698 return err; 699 700 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 701 struct rt6_info *rt; 702 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 703 704 if (!skb2) 705 return 0; 706 707 skb_dst_drop(skb2); 708 skb_pull(skb2, offset); 709 skb_reset_network_header(skb2); 710 711 /* Try to guess incoming interface */ 712 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 713 NULL, 0, skb2, 0); 714 715 if (rt && rt->dst.dev) 716 skb2->dev = rt->dst.dev; 717 718 icmpv6_send(skb2, rel_type, rel_code, rel_info); 719 720 ip6_rt_put(rt); 721 722 kfree_skb(skb2); 723 } 724 725 return 0; 726 } 727 728 static int 729 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 730 u8 type, u8 code, int offset, __be32 info) 731 { 732 __u32 rel_info = ntohl(info); 733 int err, rel_msg = 0; 734 u8 rel_type = type; 735 u8 rel_code = code; 736 737 err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code, 738 &rel_msg, &rel_info, offset); 739 return err; 740 } 741 742 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 743 const struct ipv6hdr *ipv6h, 744 struct sk_buff *skb) 745 { 746 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 747 748 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 749 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 750 751 return IP6_ECN_decapsulate(ipv6h, skb); 752 } 753 754 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 755 const struct ipv6hdr *ipv6h, 756 struct sk_buff *skb) 757 { 758 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 759 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 760 761 return IP6_ECN_decapsulate(ipv6h, skb); 762 } 763 764 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 765 const struct ipv6hdr *ipv6h, 766 struct sk_buff *skb) 767 { 768 /* ECN is not supported in AF_MPLS */ 769 return 0; 770 } 771 772 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 773 const struct in6_addr *laddr, 774 const struct in6_addr *raddr) 775 { 776 struct __ip6_tnl_parm *p = &t->parms; 777 int ltype = ipv6_addr_type(laddr); 778 int rtype = ipv6_addr_type(raddr); 779 __u32 flags = 0; 780 781 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 782 flags = IP6_TNL_F_CAP_PER_PACKET; 783 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 784 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 785 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 786 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 787 if (ltype&IPV6_ADDR_UNICAST) 788 flags |= IP6_TNL_F_CAP_XMIT; 789 if (rtype&IPV6_ADDR_UNICAST) 790 flags |= IP6_TNL_F_CAP_RCV; 791 } 792 return flags; 793 } 794 EXPORT_SYMBOL(ip6_tnl_get_cap); 795 796 /* called with rcu_read_lock() */ 797 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 798 const struct in6_addr *laddr, 799 const struct in6_addr *raddr) 800 { 801 struct __ip6_tnl_parm *p = &t->parms; 802 int ret = 0; 803 struct net *net = t->net; 804 805 if ((p->flags & IP6_TNL_F_CAP_RCV) || 806 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 807 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 808 struct net_device *ldev = NULL; 809 810 if (p->link) 811 ldev = dev_get_by_index_rcu(net, p->link); 812 813 if ((ipv6_addr_is_multicast(laddr) || 814 likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false, 815 0, IFA_F_TENTATIVE))) && 816 ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) || 817 likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true, 818 0, IFA_F_TENTATIVE)))) 819 ret = 1; 820 } 821 return ret; 822 } 823 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 824 825 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb, 826 const struct tnl_ptk_info *tpi, 827 struct metadata_dst *tun_dst, 828 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 829 const struct ipv6hdr *ipv6h, 830 struct sk_buff *skb), 831 bool log_ecn_err) 832 { 833 struct pcpu_sw_netstats *tstats; 834 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 835 int err; 836 837 if ((!(tpi->flags & TUNNEL_CSUM) && 838 (tunnel->parms.i_flags & TUNNEL_CSUM)) || 839 ((tpi->flags & TUNNEL_CSUM) && 840 !(tunnel->parms.i_flags & TUNNEL_CSUM))) { 841 tunnel->dev->stats.rx_crc_errors++; 842 tunnel->dev->stats.rx_errors++; 843 goto drop; 844 } 845 846 if (tunnel->parms.i_flags & TUNNEL_SEQ) { 847 if (!(tpi->flags & TUNNEL_SEQ) || 848 (tunnel->i_seqno && 849 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) { 850 tunnel->dev->stats.rx_fifo_errors++; 851 tunnel->dev->stats.rx_errors++; 852 goto drop; 853 } 854 tunnel->i_seqno = ntohl(tpi->seq) + 1; 855 } 856 857 skb->protocol = tpi->proto; 858 859 /* Warning: All skb pointers will be invalidated! */ 860 if (tunnel->dev->type == ARPHRD_ETHER) { 861 if (!pskb_may_pull(skb, ETH_HLEN)) { 862 tunnel->dev->stats.rx_length_errors++; 863 tunnel->dev->stats.rx_errors++; 864 goto drop; 865 } 866 867 ipv6h = ipv6_hdr(skb); 868 skb->protocol = eth_type_trans(skb, tunnel->dev); 869 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 870 } else { 871 skb->dev = tunnel->dev; 872 } 873 874 skb_reset_network_header(skb); 875 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 876 877 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 878 879 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb); 880 if (unlikely(err)) { 881 if (log_ecn_err) 882 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n", 883 &ipv6h->saddr, 884 ipv6_get_dsfield(ipv6h)); 885 if (err > 1) { 886 ++tunnel->dev->stats.rx_frame_errors; 887 ++tunnel->dev->stats.rx_errors; 888 goto drop; 889 } 890 } 891 892 tstats = this_cpu_ptr(tunnel->dev->tstats); 893 u64_stats_update_begin(&tstats->syncp); 894 tstats->rx_packets++; 895 tstats->rx_bytes += skb->len; 896 u64_stats_update_end(&tstats->syncp); 897 898 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev))); 899 900 if (tun_dst) 901 skb_dst_set(skb, (struct dst_entry *)tun_dst); 902 903 gro_cells_receive(&tunnel->gro_cells, skb); 904 return 0; 905 906 drop: 907 if (tun_dst) 908 dst_release((struct dst_entry *)tun_dst); 909 kfree_skb(skb); 910 return 0; 911 } 912 913 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb, 914 const struct tnl_ptk_info *tpi, 915 struct metadata_dst *tun_dst, 916 bool log_ecn_err) 917 { 918 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 919 const struct ipv6hdr *ipv6h, 920 struct sk_buff *skb); 921 922 dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate; 923 if (tpi->proto == htons(ETH_P_IP)) 924 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate; 925 926 return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 927 log_ecn_err); 928 } 929 EXPORT_SYMBOL(ip6_tnl_rcv); 930 931 static const struct tnl_ptk_info tpi_v6 = { 932 /* no tunnel info required for ipxip6. */ 933 .proto = htons(ETH_P_IPV6), 934 }; 935 936 static const struct tnl_ptk_info tpi_v4 = { 937 /* no tunnel info required for ipxip6. */ 938 .proto = htons(ETH_P_IP), 939 }; 940 941 static const struct tnl_ptk_info tpi_mpls = { 942 /* no tunnel info required for mplsip6. */ 943 .proto = htons(ETH_P_MPLS_UC), 944 }; 945 946 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto, 947 const struct tnl_ptk_info *tpi, 948 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 949 const struct ipv6hdr *ipv6h, 950 struct sk_buff *skb)) 951 { 952 struct ip6_tnl *t; 953 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 954 struct metadata_dst *tun_dst = NULL; 955 int ret = -1; 956 957 rcu_read_lock(); 958 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr); 959 960 if (t) { 961 u8 tproto = READ_ONCE(t->parms.proto); 962 963 if (tproto != ipproto && tproto != 0) 964 goto drop; 965 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 966 goto drop; 967 ipv6h = ipv6_hdr(skb); 968 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) 969 goto drop; 970 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 971 goto drop; 972 if (t->parms.collect_md) { 973 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0); 974 if (!tun_dst) 975 goto drop; 976 } 977 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 978 log_ecn_error); 979 } 980 981 rcu_read_unlock(); 982 983 return ret; 984 985 drop: 986 rcu_read_unlock(); 987 kfree_skb(skb); 988 return 0; 989 } 990 991 static int ip4ip6_rcv(struct sk_buff *skb) 992 { 993 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4, 994 ip4ip6_dscp_ecn_decapsulate); 995 } 996 997 static int ip6ip6_rcv(struct sk_buff *skb) 998 { 999 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6, 1000 ip6ip6_dscp_ecn_decapsulate); 1001 } 1002 1003 static int mplsip6_rcv(struct sk_buff *skb) 1004 { 1005 return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls, 1006 mplsip6_dscp_ecn_decapsulate); 1007 } 1008 1009 struct ipv6_tel_txoption { 1010 struct ipv6_txoptions ops; 1011 __u8 dst_opt[8]; 1012 }; 1013 1014 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 1015 { 1016 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 1017 1018 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 1019 opt->dst_opt[3] = 1; 1020 opt->dst_opt[4] = encap_limit; 1021 opt->dst_opt[5] = IPV6_TLV_PADN; 1022 opt->dst_opt[6] = 1; 1023 1024 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt; 1025 opt->ops.opt_nflen = 8; 1026 } 1027 1028 /** 1029 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 1030 * @t: the outgoing tunnel device 1031 * @hdr: IPv6 header from the incoming packet 1032 * 1033 * Description: 1034 * Avoid trivial tunneling loop by checking that tunnel exit-point 1035 * doesn't match source of incoming packet. 1036 * 1037 * Return: 1038 * 1 if conflict, 1039 * 0 else 1040 **/ 1041 1042 static inline bool 1043 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 1044 { 1045 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 1046 } 1047 1048 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, 1049 const struct in6_addr *laddr, 1050 const struct in6_addr *raddr) 1051 { 1052 struct __ip6_tnl_parm *p = &t->parms; 1053 int ret = 0; 1054 struct net *net = t->net; 1055 1056 if (t->parms.collect_md) 1057 return 1; 1058 1059 if ((p->flags & IP6_TNL_F_CAP_XMIT) || 1060 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 1061 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) { 1062 struct net_device *ldev = NULL; 1063 1064 rcu_read_lock(); 1065 if (p->link) 1066 ldev = dev_get_by_index_rcu(net, p->link); 1067 1068 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false, 1069 0, IFA_F_TENTATIVE))) 1070 pr_warn("%s xmit: Local address not yet configured!\n", 1071 p->name); 1072 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) && 1073 !ipv6_addr_is_multicast(raddr) && 1074 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev, 1075 true, 0, IFA_F_TENTATIVE))) 1076 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 1077 p->name); 1078 else 1079 ret = 1; 1080 rcu_read_unlock(); 1081 } 1082 return ret; 1083 } 1084 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 1085 1086 /** 1087 * ip6_tnl_xmit - encapsulate packet and send 1088 * @skb: the outgoing socket buffer 1089 * @dev: the outgoing tunnel device 1090 * @dsfield: dscp code for outer header 1091 * @fl6: flow of tunneled packet 1092 * @encap_limit: encapsulation limit 1093 * @pmtu: Path MTU is stored if packet is too big 1094 * @proto: next header value 1095 * 1096 * Description: 1097 * Build new header and do some sanity checks on the packet before sending 1098 * it. 1099 * 1100 * Return: 1101 * 0 on success 1102 * -1 fail 1103 * %-EMSGSIZE message too big. return mtu in this case. 1104 **/ 1105 1106 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield, 1107 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, 1108 __u8 proto) 1109 { 1110 struct ip6_tnl *t = netdev_priv(dev); 1111 struct net *net = t->net; 1112 struct net_device_stats *stats = &t->dev->stats; 1113 struct ipv6hdr *ipv6h; 1114 struct ipv6_tel_txoption opt; 1115 struct dst_entry *dst = NULL, *ndst = NULL; 1116 struct net_device *tdev; 1117 int mtu; 1118 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0; 1119 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen; 1120 unsigned int max_headroom = psh_hlen; 1121 bool use_cache = false; 1122 u8 hop_limit; 1123 int err = -1; 1124 1125 if (t->parms.collect_md) { 1126 hop_limit = skb_tunnel_info(skb)->key.ttl; 1127 goto route_lookup; 1128 } else { 1129 hop_limit = t->parms.hop_limit; 1130 } 1131 1132 /* NBMA tunnel */ 1133 if (ipv6_addr_any(&t->parms.raddr)) { 1134 if (skb->protocol == htons(ETH_P_IPV6)) { 1135 struct in6_addr *addr6; 1136 struct neighbour *neigh; 1137 int addr_type; 1138 1139 if (!skb_dst(skb)) 1140 goto tx_err_link_failure; 1141 1142 neigh = dst_neigh_lookup(skb_dst(skb), 1143 &ipv6_hdr(skb)->daddr); 1144 if (!neigh) 1145 goto tx_err_link_failure; 1146 1147 addr6 = (struct in6_addr *)&neigh->primary_key; 1148 addr_type = ipv6_addr_type(addr6); 1149 1150 if (addr_type == IPV6_ADDR_ANY) 1151 addr6 = &ipv6_hdr(skb)->daddr; 1152 1153 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr)); 1154 neigh_release(neigh); 1155 } 1156 } else if (t->parms.proto != 0 && !(t->parms.flags & 1157 (IP6_TNL_F_USE_ORIG_TCLASS | 1158 IP6_TNL_F_USE_ORIG_FWMARK))) { 1159 /* enable the cache only if neither the outer protocol nor the 1160 * routing decision depends on the current inner header value 1161 */ 1162 use_cache = true; 1163 } 1164 1165 if (use_cache) 1166 dst = dst_cache_get(&t->dst_cache); 1167 1168 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1169 goto tx_err_link_failure; 1170 1171 if (!dst) { 1172 route_lookup: 1173 /* add dsfield to flowlabel for route lookup */ 1174 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1175 1176 dst = ip6_route_output(net, NULL, fl6); 1177 1178 if (dst->error) 1179 goto tx_err_link_failure; 1180 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1181 if (IS_ERR(dst)) { 1182 err = PTR_ERR(dst); 1183 dst = NULL; 1184 goto tx_err_link_failure; 1185 } 1186 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) && 1187 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1188 &fl6->daddr, 0, &fl6->saddr)) 1189 goto tx_err_link_failure; 1190 ndst = dst; 1191 } 1192 1193 tdev = dst->dev; 1194 1195 if (tdev == dev) { 1196 stats->collisions++; 1197 net_warn_ratelimited("%s: Local routing loop detected!\n", 1198 t->parms.name); 1199 goto tx_err_dst_release; 1200 } 1201 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen; 1202 if (encap_limit >= 0) { 1203 max_headroom += 8; 1204 mtu -= 8; 1205 } 1206 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ? 1207 IPV6_MIN_MTU : IPV4_MIN_MTU); 1208 1209 skb_dst_update_pmtu_no_confirm(skb, mtu); 1210 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) { 1211 *pmtu = mtu; 1212 err = -EMSGSIZE; 1213 goto tx_err_dst_release; 1214 } 1215 1216 if (t->err_count > 0) { 1217 if (time_before(jiffies, 1218 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1219 t->err_count--; 1220 1221 dst_link_failure(skb); 1222 } else { 1223 t->err_count = 0; 1224 } 1225 } 1226 1227 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1228 1229 /* 1230 * Okay, now see if we can stuff it in the buffer as-is. 1231 */ 1232 max_headroom += LL_RESERVED_SPACE(tdev); 1233 1234 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1235 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1236 struct sk_buff *new_skb; 1237 1238 new_skb = skb_realloc_headroom(skb, max_headroom); 1239 if (!new_skb) 1240 goto tx_err_dst_release; 1241 1242 if (skb->sk) 1243 skb_set_owner_w(new_skb, skb->sk); 1244 consume_skb(skb); 1245 skb = new_skb; 1246 } 1247 1248 if (t->parms.collect_md) { 1249 if (t->encap.type != TUNNEL_ENCAP_NONE) 1250 goto tx_err_dst_release; 1251 } else { 1252 if (use_cache && ndst) 1253 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1254 } 1255 skb_dst_set(skb, dst); 1256 1257 if (hop_limit == 0) { 1258 if (skb->protocol == htons(ETH_P_IP)) 1259 hop_limit = ip_hdr(skb)->ttl; 1260 else if (skb->protocol == htons(ETH_P_IPV6)) 1261 hop_limit = ipv6_hdr(skb)->hop_limit; 1262 else 1263 hop_limit = ip6_dst_hoplimit(dst); 1264 } 1265 1266 /* Calculate max headroom for all the headers and adjust 1267 * needed_headroom if necessary. 1268 */ 1269 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1270 + dst->header_len + t->hlen; 1271 if (max_headroom > dev->needed_headroom) 1272 dev->needed_headroom = max_headroom; 1273 1274 err = ip6_tnl_encap(skb, t, &proto, fl6); 1275 if (err) 1276 return err; 1277 1278 if (encap_limit >= 0) { 1279 init_tel_txopt(&opt, encap_limit); 1280 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1281 } 1282 1283 skb_set_inner_ipproto(skb, proto); 1284 1285 skb_push(skb, sizeof(struct ipv6hdr)); 1286 skb_reset_network_header(skb); 1287 ipv6h = ipv6_hdr(skb); 1288 ip6_flow_hdr(ipv6h, dsfield, 1289 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1290 ipv6h->hop_limit = hop_limit; 1291 ipv6h->nexthdr = proto; 1292 ipv6h->saddr = fl6->saddr; 1293 ipv6h->daddr = fl6->daddr; 1294 ip6tunnel_xmit(NULL, skb, dev); 1295 return 0; 1296 tx_err_link_failure: 1297 stats->tx_carrier_errors++; 1298 dst_link_failure(skb); 1299 tx_err_dst_release: 1300 dst_release(dst); 1301 return err; 1302 } 1303 EXPORT_SYMBOL(ip6_tnl_xmit); 1304 1305 static inline int 1306 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, 1307 u8 protocol) 1308 { 1309 struct ip6_tnl *t = netdev_priv(dev); 1310 struct ipv6hdr *ipv6h; 1311 const struct iphdr *iph; 1312 int encap_limit = -1; 1313 __u16 offset; 1314 struct flowi6 fl6; 1315 __u8 dsfield, orig_dsfield; 1316 __u32 mtu; 1317 u8 tproto; 1318 int err; 1319 1320 tproto = READ_ONCE(t->parms.proto); 1321 if (tproto != protocol && tproto != 0) 1322 return -1; 1323 1324 if (t->parms.collect_md) { 1325 struct ip_tunnel_info *tun_info; 1326 const struct ip_tunnel_key *key; 1327 1328 tun_info = skb_tunnel_info(skb); 1329 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1330 ip_tunnel_info_af(tun_info) != AF_INET6)) 1331 return -1; 1332 key = &tun_info->key; 1333 memset(&fl6, 0, sizeof(fl6)); 1334 fl6.flowi6_proto = protocol; 1335 fl6.saddr = key->u.ipv6.src; 1336 fl6.daddr = key->u.ipv6.dst; 1337 fl6.flowlabel = key->label; 1338 dsfield = key->tos; 1339 switch (protocol) { 1340 case IPPROTO_IPIP: 1341 iph = ip_hdr(skb); 1342 orig_dsfield = ipv4_get_dsfield(iph); 1343 break; 1344 case IPPROTO_IPV6: 1345 ipv6h = ipv6_hdr(skb); 1346 orig_dsfield = ipv6_get_dsfield(ipv6h); 1347 break; 1348 default: 1349 orig_dsfield = dsfield; 1350 break; 1351 } 1352 } else { 1353 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1354 encap_limit = t->parms.encap_limit; 1355 if (protocol == IPPROTO_IPV6) { 1356 offset = ip6_tnl_parse_tlv_enc_lim(skb, 1357 skb_network_header(skb)); 1358 /* ip6_tnl_parse_tlv_enc_lim() might have 1359 * reallocated skb->head 1360 */ 1361 if (offset > 0) { 1362 struct ipv6_tlv_tnl_enc_lim *tel; 1363 1364 tel = (void *)&skb_network_header(skb)[offset]; 1365 if (tel->encap_limit == 0) { 1366 icmpv6_send(skb, ICMPV6_PARAMPROB, 1367 ICMPV6_HDR_FIELD, offset + 2); 1368 return -1; 1369 } 1370 encap_limit = tel->encap_limit - 1; 1371 } 1372 } 1373 1374 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1375 fl6.flowi6_proto = protocol; 1376 1377 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1378 fl6.flowi6_mark = skb->mark; 1379 else 1380 fl6.flowi6_mark = t->parms.fwmark; 1381 switch (protocol) { 1382 case IPPROTO_IPIP: 1383 iph = ip_hdr(skb); 1384 orig_dsfield = ipv4_get_dsfield(iph); 1385 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1386 dsfield = orig_dsfield; 1387 else 1388 dsfield = ip6_tclass(t->parms.flowinfo); 1389 break; 1390 case IPPROTO_IPV6: 1391 ipv6h = ipv6_hdr(skb); 1392 orig_dsfield = ipv6_get_dsfield(ipv6h); 1393 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1394 dsfield = orig_dsfield; 1395 else 1396 dsfield = ip6_tclass(t->parms.flowinfo); 1397 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1398 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1399 break; 1400 default: 1401 orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo); 1402 break; 1403 } 1404 } 1405 1406 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1407 dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield); 1408 1409 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1410 return -1; 1411 1412 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1413 protocol); 1414 if (err != 0) { 1415 /* XXX: send ICMP error even if DF is not set. */ 1416 if (err == -EMSGSIZE) 1417 switch (protocol) { 1418 case IPPROTO_IPIP: 1419 icmp_send(skb, ICMP_DEST_UNREACH, 1420 ICMP_FRAG_NEEDED, htonl(mtu)); 1421 break; 1422 case IPPROTO_IPV6: 1423 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1424 break; 1425 default: 1426 break; 1427 } 1428 return -1; 1429 } 1430 1431 return 0; 1432 } 1433 1434 static netdev_tx_t 1435 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1436 { 1437 struct ip6_tnl *t = netdev_priv(dev); 1438 struct net_device_stats *stats = &t->dev->stats; 1439 u8 ipproto; 1440 int ret; 1441 1442 if (!pskb_inet_may_pull(skb)) 1443 goto tx_err; 1444 1445 switch (skb->protocol) { 1446 case htons(ETH_P_IP): 1447 ipproto = IPPROTO_IPIP; 1448 break; 1449 case htons(ETH_P_IPV6): 1450 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb))) 1451 goto tx_err; 1452 ipproto = IPPROTO_IPV6; 1453 break; 1454 case htons(ETH_P_MPLS_UC): 1455 ipproto = IPPROTO_MPLS; 1456 break; 1457 default: 1458 goto tx_err; 1459 } 1460 1461 ret = ipxip6_tnl_xmit(skb, dev, ipproto); 1462 if (ret < 0) 1463 goto tx_err; 1464 1465 return NETDEV_TX_OK; 1466 1467 tx_err: 1468 stats->tx_errors++; 1469 stats->tx_dropped++; 1470 kfree_skb(skb); 1471 return NETDEV_TX_OK; 1472 } 1473 1474 static void ip6_tnl_link_config(struct ip6_tnl *t) 1475 { 1476 struct net_device *dev = t->dev; 1477 struct net_device *tdev = NULL; 1478 struct __ip6_tnl_parm *p = &t->parms; 1479 struct flowi6 *fl6 = &t->fl.u.ip6; 1480 unsigned int mtu; 1481 int t_hlen; 1482 1483 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1484 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1485 1486 /* Set up flowi template */ 1487 fl6->saddr = p->laddr; 1488 fl6->daddr = p->raddr; 1489 fl6->flowi6_oif = p->link; 1490 fl6->flowlabel = 0; 1491 1492 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1493 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1494 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1495 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1496 1497 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1498 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1499 1500 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1501 dev->flags |= IFF_POINTOPOINT; 1502 else 1503 dev->flags &= ~IFF_POINTOPOINT; 1504 1505 t->tun_hlen = 0; 1506 t->hlen = t->encap_hlen + t->tun_hlen; 1507 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1508 1509 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1510 int strict = (ipv6_addr_type(&p->raddr) & 1511 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1512 1513 struct rt6_info *rt = rt6_lookup(t->net, 1514 &p->raddr, &p->laddr, 1515 p->link, NULL, strict); 1516 if (rt) { 1517 tdev = rt->dst.dev; 1518 ip6_rt_put(rt); 1519 } 1520 1521 if (!tdev && p->link) 1522 tdev = __dev_get_by_index(t->net, p->link); 1523 1524 if (tdev) { 1525 dev->hard_header_len = tdev->hard_header_len + t_hlen; 1526 mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU); 1527 1528 dev->mtu = mtu - t_hlen; 1529 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1530 dev->mtu -= 8; 1531 1532 if (dev->mtu < IPV6_MIN_MTU) 1533 dev->mtu = IPV6_MIN_MTU; 1534 } 1535 } 1536 } 1537 1538 /** 1539 * ip6_tnl_change - update the tunnel parameters 1540 * @t: tunnel to be changed 1541 * @p: tunnel configuration parameters 1542 * 1543 * Description: 1544 * ip6_tnl_change() updates the tunnel parameters 1545 **/ 1546 1547 static int 1548 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1549 { 1550 t->parms.laddr = p->laddr; 1551 t->parms.raddr = p->raddr; 1552 t->parms.flags = p->flags; 1553 t->parms.hop_limit = p->hop_limit; 1554 t->parms.encap_limit = p->encap_limit; 1555 t->parms.flowinfo = p->flowinfo; 1556 t->parms.link = p->link; 1557 t->parms.proto = p->proto; 1558 t->parms.fwmark = p->fwmark; 1559 dst_cache_reset(&t->dst_cache); 1560 ip6_tnl_link_config(t); 1561 return 0; 1562 } 1563 1564 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1565 { 1566 struct net *net = t->net; 1567 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1568 int err; 1569 1570 ip6_tnl_unlink(ip6n, t); 1571 synchronize_net(); 1572 err = ip6_tnl_change(t, p); 1573 ip6_tnl_link(ip6n, t); 1574 netdev_state_change(t->dev); 1575 return err; 1576 } 1577 1578 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1579 { 1580 /* for default tnl0 device allow to change only the proto */ 1581 t->parms.proto = p->proto; 1582 netdev_state_change(t->dev); 1583 return 0; 1584 } 1585 1586 static void 1587 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1588 { 1589 p->laddr = u->laddr; 1590 p->raddr = u->raddr; 1591 p->flags = u->flags; 1592 p->hop_limit = u->hop_limit; 1593 p->encap_limit = u->encap_limit; 1594 p->flowinfo = u->flowinfo; 1595 p->link = u->link; 1596 p->proto = u->proto; 1597 memcpy(p->name, u->name, sizeof(u->name)); 1598 } 1599 1600 static void 1601 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1602 { 1603 u->laddr = p->laddr; 1604 u->raddr = p->raddr; 1605 u->flags = p->flags; 1606 u->hop_limit = p->hop_limit; 1607 u->encap_limit = p->encap_limit; 1608 u->flowinfo = p->flowinfo; 1609 u->link = p->link; 1610 u->proto = p->proto; 1611 memcpy(u->name, p->name, sizeof(u->name)); 1612 } 1613 1614 /** 1615 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1616 * @dev: virtual device associated with tunnel 1617 * @ifr: parameters passed from userspace 1618 * @cmd: command to be performed 1619 * 1620 * Description: 1621 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1622 * from userspace. 1623 * 1624 * The possible commands are the following: 1625 * %SIOCGETTUNNEL: get tunnel parameters for device 1626 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1627 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1628 * %SIOCDELTUNNEL: delete tunnel 1629 * 1630 * The fallback device "ip6tnl0", created during module 1631 * initialization, can be used for creating other tunnel devices. 1632 * 1633 * Return: 1634 * 0 on success, 1635 * %-EFAULT if unable to copy data to or from userspace, 1636 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1637 * %-EINVAL if passed tunnel parameters are invalid, 1638 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1639 * %-ENODEV if attempting to change or delete a nonexisting device 1640 **/ 1641 1642 static int 1643 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1644 { 1645 int err = 0; 1646 struct ip6_tnl_parm p; 1647 struct __ip6_tnl_parm p1; 1648 struct ip6_tnl *t = netdev_priv(dev); 1649 struct net *net = t->net; 1650 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1651 1652 memset(&p1, 0, sizeof(p1)); 1653 1654 switch (cmd) { 1655 case SIOCGETTUNNEL: 1656 if (dev == ip6n->fb_tnl_dev) { 1657 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1658 err = -EFAULT; 1659 break; 1660 } 1661 ip6_tnl_parm_from_user(&p1, &p); 1662 t = ip6_tnl_locate(net, &p1, 0); 1663 if (IS_ERR(t)) 1664 t = netdev_priv(dev); 1665 } else { 1666 memset(&p, 0, sizeof(p)); 1667 } 1668 ip6_tnl_parm_to_user(&p, &t->parms); 1669 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) { 1670 err = -EFAULT; 1671 } 1672 break; 1673 case SIOCADDTUNNEL: 1674 case SIOCCHGTUNNEL: 1675 err = -EPERM; 1676 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1677 break; 1678 err = -EFAULT; 1679 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1680 break; 1681 err = -EINVAL; 1682 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1683 p.proto != 0) 1684 break; 1685 ip6_tnl_parm_from_user(&p1, &p); 1686 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1687 if (cmd == SIOCCHGTUNNEL) { 1688 if (!IS_ERR(t)) { 1689 if (t->dev != dev) { 1690 err = -EEXIST; 1691 break; 1692 } 1693 } else 1694 t = netdev_priv(dev); 1695 if (dev == ip6n->fb_tnl_dev) 1696 err = ip6_tnl0_update(t, &p1); 1697 else 1698 err = ip6_tnl_update(t, &p1); 1699 } 1700 if (!IS_ERR(t)) { 1701 err = 0; 1702 ip6_tnl_parm_to_user(&p, &t->parms); 1703 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1704 err = -EFAULT; 1705 1706 } else { 1707 err = PTR_ERR(t); 1708 } 1709 break; 1710 case SIOCDELTUNNEL: 1711 err = -EPERM; 1712 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1713 break; 1714 1715 if (dev == ip6n->fb_tnl_dev) { 1716 err = -EFAULT; 1717 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1718 break; 1719 err = -ENOENT; 1720 ip6_tnl_parm_from_user(&p1, &p); 1721 t = ip6_tnl_locate(net, &p1, 0); 1722 if (IS_ERR(t)) 1723 break; 1724 err = -EPERM; 1725 if (t->dev == ip6n->fb_tnl_dev) 1726 break; 1727 dev = t->dev; 1728 } 1729 err = 0; 1730 unregister_netdevice(dev); 1731 break; 1732 default: 1733 err = -EINVAL; 1734 } 1735 return err; 1736 } 1737 1738 /** 1739 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1740 * @dev: virtual device associated with tunnel 1741 * @new_mtu: the new mtu 1742 * 1743 * Return: 1744 * 0 on success, 1745 * %-EINVAL if mtu too small 1746 **/ 1747 1748 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1749 { 1750 struct ip6_tnl *tnl = netdev_priv(dev); 1751 1752 if (tnl->parms.proto == IPPROTO_IPV6) { 1753 if (new_mtu < IPV6_MIN_MTU) 1754 return -EINVAL; 1755 } else { 1756 if (new_mtu < ETH_MIN_MTU) 1757 return -EINVAL; 1758 } 1759 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) { 1760 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len) 1761 return -EINVAL; 1762 } else { 1763 if (new_mtu > IP_MAX_MTU - dev->hard_header_len) 1764 return -EINVAL; 1765 } 1766 dev->mtu = new_mtu; 1767 return 0; 1768 } 1769 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1770 1771 int ip6_tnl_get_iflink(const struct net_device *dev) 1772 { 1773 struct ip6_tnl *t = netdev_priv(dev); 1774 1775 return t->parms.link; 1776 } 1777 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1778 1779 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1780 unsigned int num) 1781 { 1782 if (num >= MAX_IPTUN_ENCAP_OPS) 1783 return -ERANGE; 1784 1785 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1786 &ip6tun_encaps[num], 1787 NULL, ops) ? 0 : -1; 1788 } 1789 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1790 1791 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1792 unsigned int num) 1793 { 1794 int ret; 1795 1796 if (num >= MAX_IPTUN_ENCAP_OPS) 1797 return -ERANGE; 1798 1799 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1800 &ip6tun_encaps[num], 1801 ops, NULL) == ops) ? 0 : -1; 1802 1803 synchronize_net(); 1804 1805 return ret; 1806 } 1807 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1808 1809 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1810 struct ip_tunnel_encap *ipencap) 1811 { 1812 int hlen; 1813 1814 memset(&t->encap, 0, sizeof(t->encap)); 1815 1816 hlen = ip6_encap_hlen(ipencap); 1817 if (hlen < 0) 1818 return hlen; 1819 1820 t->encap.type = ipencap->type; 1821 t->encap.sport = ipencap->sport; 1822 t->encap.dport = ipencap->dport; 1823 t->encap.flags = ipencap->flags; 1824 1825 t->encap_hlen = hlen; 1826 t->hlen = t->encap_hlen + t->tun_hlen; 1827 1828 return 0; 1829 } 1830 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1831 1832 static const struct net_device_ops ip6_tnl_netdev_ops = { 1833 .ndo_init = ip6_tnl_dev_init, 1834 .ndo_uninit = ip6_tnl_dev_uninit, 1835 .ndo_start_xmit = ip6_tnl_start_xmit, 1836 .ndo_do_ioctl = ip6_tnl_ioctl, 1837 .ndo_change_mtu = ip6_tnl_change_mtu, 1838 .ndo_get_stats = ip6_get_stats, 1839 .ndo_get_iflink = ip6_tnl_get_iflink, 1840 }; 1841 1842 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1843 NETIF_F_FRAGLIST | \ 1844 NETIF_F_HIGHDMA | \ 1845 NETIF_F_GSO_SOFTWARE | \ 1846 NETIF_F_HW_CSUM) 1847 1848 /** 1849 * ip6_tnl_dev_setup - setup virtual tunnel device 1850 * @dev: virtual device associated with tunnel 1851 * 1852 * Description: 1853 * Initialize function pointers and device parameters 1854 **/ 1855 1856 static void ip6_tnl_dev_setup(struct net_device *dev) 1857 { 1858 dev->netdev_ops = &ip6_tnl_netdev_ops; 1859 dev->header_ops = &ip_tunnel_header_ops; 1860 dev->needs_free_netdev = true; 1861 dev->priv_destructor = ip6_dev_free; 1862 1863 dev->type = ARPHRD_TUNNEL6; 1864 dev->flags |= IFF_NOARP; 1865 dev->addr_len = sizeof(struct in6_addr); 1866 dev->features |= NETIF_F_LLTX; 1867 netif_keep_dst(dev); 1868 1869 dev->features |= IPXIPX_FEATURES; 1870 dev->hw_features |= IPXIPX_FEATURES; 1871 1872 /* This perm addr will be used as interface identifier by IPv6 */ 1873 dev->addr_assign_type = NET_ADDR_RANDOM; 1874 eth_random_addr(dev->perm_addr); 1875 } 1876 1877 1878 /** 1879 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1880 * @dev: virtual device associated with tunnel 1881 **/ 1882 1883 static inline int 1884 ip6_tnl_dev_init_gen(struct net_device *dev) 1885 { 1886 struct ip6_tnl *t = netdev_priv(dev); 1887 int ret; 1888 int t_hlen; 1889 1890 t->dev = dev; 1891 t->net = dev_net(dev); 1892 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1893 if (!dev->tstats) 1894 return -ENOMEM; 1895 1896 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1897 if (ret) 1898 goto free_stats; 1899 1900 ret = gro_cells_init(&t->gro_cells, dev); 1901 if (ret) 1902 goto destroy_dst; 1903 1904 t->tun_hlen = 0; 1905 t->hlen = t->encap_hlen + t->tun_hlen; 1906 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1907 1908 dev->type = ARPHRD_TUNNEL6; 1909 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1910 dev->mtu = ETH_DATA_LEN - t_hlen; 1911 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1912 dev->mtu -= 8; 1913 dev->min_mtu = ETH_MIN_MTU; 1914 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len; 1915 1916 return 0; 1917 1918 destroy_dst: 1919 dst_cache_destroy(&t->dst_cache); 1920 free_stats: 1921 free_percpu(dev->tstats); 1922 dev->tstats = NULL; 1923 1924 return ret; 1925 } 1926 1927 /** 1928 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1929 * @dev: virtual device associated with tunnel 1930 **/ 1931 1932 static int ip6_tnl_dev_init(struct net_device *dev) 1933 { 1934 struct ip6_tnl *t = netdev_priv(dev); 1935 int err = ip6_tnl_dev_init_gen(dev); 1936 1937 if (err) 1938 return err; 1939 ip6_tnl_link_config(t); 1940 if (t->parms.collect_md) 1941 netif_keep_dst(dev); 1942 return 0; 1943 } 1944 1945 /** 1946 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1947 * @dev: fallback device 1948 * 1949 * Return: 0 1950 **/ 1951 1952 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1953 { 1954 struct ip6_tnl *t = netdev_priv(dev); 1955 struct net *net = dev_net(dev); 1956 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1957 1958 t->parms.proto = IPPROTO_IPV6; 1959 dev_hold(dev); 1960 1961 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1962 return 0; 1963 } 1964 1965 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[], 1966 struct netlink_ext_ack *extack) 1967 { 1968 u8 proto; 1969 1970 if (!data || !data[IFLA_IPTUN_PROTO]) 1971 return 0; 1972 1973 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1974 if (proto != IPPROTO_IPV6 && 1975 proto != IPPROTO_IPIP && 1976 proto != 0) 1977 return -EINVAL; 1978 1979 return 0; 1980 } 1981 1982 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1983 struct __ip6_tnl_parm *parms) 1984 { 1985 memset(parms, 0, sizeof(*parms)); 1986 1987 if (!data) 1988 return; 1989 1990 if (data[IFLA_IPTUN_LINK]) 1991 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1992 1993 if (data[IFLA_IPTUN_LOCAL]) 1994 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1995 1996 if (data[IFLA_IPTUN_REMOTE]) 1997 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1998 1999 if (data[IFLA_IPTUN_TTL]) 2000 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 2001 2002 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 2003 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 2004 2005 if (data[IFLA_IPTUN_FLOWINFO]) 2006 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 2007 2008 if (data[IFLA_IPTUN_FLAGS]) 2009 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 2010 2011 if (data[IFLA_IPTUN_PROTO]) 2012 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 2013 2014 if (data[IFLA_IPTUN_COLLECT_METADATA]) 2015 parms->collect_md = true; 2016 2017 if (data[IFLA_IPTUN_FWMARK]) 2018 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 2019 } 2020 2021 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[], 2022 struct ip_tunnel_encap *ipencap) 2023 { 2024 bool ret = false; 2025 2026 memset(ipencap, 0, sizeof(*ipencap)); 2027 2028 if (!data) 2029 return ret; 2030 2031 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 2032 ret = true; 2033 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 2034 } 2035 2036 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 2037 ret = true; 2038 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 2039 } 2040 2041 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 2042 ret = true; 2043 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 2044 } 2045 2046 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 2047 ret = true; 2048 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 2049 } 2050 2051 return ret; 2052 } 2053 2054 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 2055 struct nlattr *tb[], struct nlattr *data[], 2056 struct netlink_ext_ack *extack) 2057 { 2058 struct net *net = dev_net(dev); 2059 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2060 struct ip_tunnel_encap ipencap; 2061 struct ip6_tnl *nt, *t; 2062 int err; 2063 2064 nt = netdev_priv(dev); 2065 2066 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2067 err = ip6_tnl_encap_setup(nt, &ipencap); 2068 if (err < 0) 2069 return err; 2070 } 2071 2072 ip6_tnl_netlink_parms(data, &nt->parms); 2073 2074 if (nt->parms.collect_md) { 2075 if (rtnl_dereference(ip6n->collect_md_tun)) 2076 return -EEXIST; 2077 } else { 2078 t = ip6_tnl_locate(net, &nt->parms, 0); 2079 if (!IS_ERR(t)) 2080 return -EEXIST; 2081 } 2082 2083 err = ip6_tnl_create2(dev); 2084 if (!err && tb[IFLA_MTU]) 2085 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2086 2087 return err; 2088 } 2089 2090 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2091 struct nlattr *data[], 2092 struct netlink_ext_ack *extack) 2093 { 2094 struct ip6_tnl *t = netdev_priv(dev); 2095 struct __ip6_tnl_parm p; 2096 struct net *net = t->net; 2097 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2098 struct ip_tunnel_encap ipencap; 2099 2100 if (dev == ip6n->fb_tnl_dev) 2101 return -EINVAL; 2102 2103 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2104 int err = ip6_tnl_encap_setup(t, &ipencap); 2105 2106 if (err < 0) 2107 return err; 2108 } 2109 ip6_tnl_netlink_parms(data, &p); 2110 if (p.collect_md) 2111 return -EINVAL; 2112 2113 t = ip6_tnl_locate(net, &p, 0); 2114 if (!IS_ERR(t)) { 2115 if (t->dev != dev) 2116 return -EEXIST; 2117 } else 2118 t = netdev_priv(dev); 2119 2120 return ip6_tnl_update(t, &p); 2121 } 2122 2123 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2124 { 2125 struct net *net = dev_net(dev); 2126 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2127 2128 if (dev != ip6n->fb_tnl_dev) 2129 unregister_netdevice_queue(dev, head); 2130 } 2131 2132 static size_t ip6_tnl_get_size(const struct net_device *dev) 2133 { 2134 return 2135 /* IFLA_IPTUN_LINK */ 2136 nla_total_size(4) + 2137 /* IFLA_IPTUN_LOCAL */ 2138 nla_total_size(sizeof(struct in6_addr)) + 2139 /* IFLA_IPTUN_REMOTE */ 2140 nla_total_size(sizeof(struct in6_addr)) + 2141 /* IFLA_IPTUN_TTL */ 2142 nla_total_size(1) + 2143 /* IFLA_IPTUN_ENCAP_LIMIT */ 2144 nla_total_size(1) + 2145 /* IFLA_IPTUN_FLOWINFO */ 2146 nla_total_size(4) + 2147 /* IFLA_IPTUN_FLAGS */ 2148 nla_total_size(4) + 2149 /* IFLA_IPTUN_PROTO */ 2150 nla_total_size(1) + 2151 /* IFLA_IPTUN_ENCAP_TYPE */ 2152 nla_total_size(2) + 2153 /* IFLA_IPTUN_ENCAP_FLAGS */ 2154 nla_total_size(2) + 2155 /* IFLA_IPTUN_ENCAP_SPORT */ 2156 nla_total_size(2) + 2157 /* IFLA_IPTUN_ENCAP_DPORT */ 2158 nla_total_size(2) + 2159 /* IFLA_IPTUN_COLLECT_METADATA */ 2160 nla_total_size(0) + 2161 /* IFLA_IPTUN_FWMARK */ 2162 nla_total_size(4) + 2163 0; 2164 } 2165 2166 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2167 { 2168 struct ip6_tnl *tunnel = netdev_priv(dev); 2169 struct __ip6_tnl_parm *parm = &tunnel->parms; 2170 2171 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2172 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2173 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2174 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2175 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2176 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2177 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2178 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2179 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2180 goto nla_put_failure; 2181 2182 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2183 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2184 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2185 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2186 goto nla_put_failure; 2187 2188 if (parm->collect_md) 2189 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2190 goto nla_put_failure; 2191 2192 return 0; 2193 2194 nla_put_failure: 2195 return -EMSGSIZE; 2196 } 2197 2198 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2199 { 2200 struct ip6_tnl *tunnel = netdev_priv(dev); 2201 2202 return tunnel->net; 2203 } 2204 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2205 2206 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2207 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2208 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2209 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2210 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2211 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2212 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2213 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2214 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2215 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2216 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2217 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2218 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2219 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2220 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2221 }; 2222 2223 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2224 .kind = "ip6tnl", 2225 .maxtype = IFLA_IPTUN_MAX, 2226 .policy = ip6_tnl_policy, 2227 .priv_size = sizeof(struct ip6_tnl), 2228 .setup = ip6_tnl_dev_setup, 2229 .validate = ip6_tnl_validate, 2230 .newlink = ip6_tnl_newlink, 2231 .changelink = ip6_tnl_changelink, 2232 .dellink = ip6_tnl_dellink, 2233 .get_size = ip6_tnl_get_size, 2234 .fill_info = ip6_tnl_fill_info, 2235 .get_link_net = ip6_tnl_get_link_net, 2236 }; 2237 2238 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2239 .handler = ip4ip6_rcv, 2240 .err_handler = ip4ip6_err, 2241 .priority = 1, 2242 }; 2243 2244 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2245 .handler = ip6ip6_rcv, 2246 .err_handler = ip6ip6_err, 2247 .priority = 1, 2248 }; 2249 2250 static struct xfrm6_tunnel mplsip6_handler __read_mostly = { 2251 .handler = mplsip6_rcv, 2252 .err_handler = mplsip6_err, 2253 .priority = 1, 2254 }; 2255 2256 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list) 2257 { 2258 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2259 struct net_device *dev, *aux; 2260 int h; 2261 struct ip6_tnl *t; 2262 2263 for_each_netdev_safe(net, dev, aux) 2264 if (dev->rtnl_link_ops == &ip6_link_ops) 2265 unregister_netdevice_queue(dev, list); 2266 2267 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2268 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2269 while (t) { 2270 /* If dev is in the same netns, it has already 2271 * been added to the list by the previous loop. 2272 */ 2273 if (!net_eq(dev_net(t->dev), net)) 2274 unregister_netdevice_queue(t->dev, list); 2275 t = rtnl_dereference(t->next); 2276 } 2277 } 2278 } 2279 2280 static int __net_init ip6_tnl_init_net(struct net *net) 2281 { 2282 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2283 struct ip6_tnl *t = NULL; 2284 int err; 2285 2286 ip6n->tnls[0] = ip6n->tnls_wc; 2287 ip6n->tnls[1] = ip6n->tnls_r_l; 2288 2289 if (!net_has_fallback_tunnels(net)) 2290 return 0; 2291 err = -ENOMEM; 2292 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2293 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2294 2295 if (!ip6n->fb_tnl_dev) 2296 goto err_alloc_dev; 2297 dev_net_set(ip6n->fb_tnl_dev, net); 2298 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2299 /* FB netdevice is special: we have one, and only one per netns. 2300 * Allowing to move it to another netns is clearly unsafe. 2301 */ 2302 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 2303 2304 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2305 if (err < 0) 2306 goto err_register; 2307 2308 err = register_netdev(ip6n->fb_tnl_dev); 2309 if (err < 0) 2310 goto err_register; 2311 2312 t = netdev_priv(ip6n->fb_tnl_dev); 2313 2314 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2315 return 0; 2316 2317 err_register: 2318 free_netdev(ip6n->fb_tnl_dev); 2319 err_alloc_dev: 2320 return err; 2321 } 2322 2323 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list) 2324 { 2325 struct net *net; 2326 LIST_HEAD(list); 2327 2328 rtnl_lock(); 2329 list_for_each_entry(net, net_list, exit_list) 2330 ip6_tnl_destroy_tunnels(net, &list); 2331 unregister_netdevice_many(&list); 2332 rtnl_unlock(); 2333 } 2334 2335 static struct pernet_operations ip6_tnl_net_ops = { 2336 .init = ip6_tnl_init_net, 2337 .exit_batch = ip6_tnl_exit_batch_net, 2338 .id = &ip6_tnl_net_id, 2339 .size = sizeof(struct ip6_tnl_net), 2340 }; 2341 2342 /** 2343 * ip6_tunnel_init - register protocol and reserve needed resources 2344 * 2345 * Return: 0 on success 2346 **/ 2347 2348 static int __init ip6_tunnel_init(void) 2349 { 2350 int err; 2351 2352 if (!ipv6_mod_enabled()) 2353 return -EOPNOTSUPP; 2354 2355 err = register_pernet_device(&ip6_tnl_net_ops); 2356 if (err < 0) 2357 goto out_pernet; 2358 2359 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2360 if (err < 0) { 2361 pr_err("%s: can't register ip4ip6\n", __func__); 2362 goto out_ip4ip6; 2363 } 2364 2365 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2366 if (err < 0) { 2367 pr_err("%s: can't register ip6ip6\n", __func__); 2368 goto out_ip6ip6; 2369 } 2370 2371 if (ip6_tnl_mpls_supported()) { 2372 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS); 2373 if (err < 0) { 2374 pr_err("%s: can't register mplsip6\n", __func__); 2375 goto out_mplsip6; 2376 } 2377 } 2378 2379 err = rtnl_link_register(&ip6_link_ops); 2380 if (err < 0) 2381 goto rtnl_link_failed; 2382 2383 return 0; 2384 2385 rtnl_link_failed: 2386 if (ip6_tnl_mpls_supported()) 2387 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS); 2388 out_mplsip6: 2389 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2390 out_ip6ip6: 2391 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2392 out_ip4ip6: 2393 unregister_pernet_device(&ip6_tnl_net_ops); 2394 out_pernet: 2395 return err; 2396 } 2397 2398 /** 2399 * ip6_tunnel_cleanup - free resources and unregister protocol 2400 **/ 2401 2402 static void __exit ip6_tunnel_cleanup(void) 2403 { 2404 rtnl_link_unregister(&ip6_link_ops); 2405 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2406 pr_info("%s: can't deregister ip4ip6\n", __func__); 2407 2408 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2409 pr_info("%s: can't deregister ip6ip6\n", __func__); 2410 2411 if (ip6_tnl_mpls_supported() && 2412 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS)) 2413 pr_info("%s: can't deregister mplsip6\n", __func__); 2414 unregister_pernet_device(&ip6_tnl_net_ops); 2415 } 2416 2417 module_init(ip6_tunnel_init); 2418 module_exit(ip6_tunnel_cleanup); 2419