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 return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate, 919 log_ecn_err); 920 } 921 EXPORT_SYMBOL(ip6_tnl_rcv); 922 923 static const struct tnl_ptk_info tpi_v6 = { 924 /* no tunnel info required for ipxip6. */ 925 .proto = htons(ETH_P_IPV6), 926 }; 927 928 static const struct tnl_ptk_info tpi_v4 = { 929 /* no tunnel info required for ipxip6. */ 930 .proto = htons(ETH_P_IP), 931 }; 932 933 static const struct tnl_ptk_info tpi_mpls = { 934 /* no tunnel info required for mplsip6. */ 935 .proto = htons(ETH_P_MPLS_UC), 936 }; 937 938 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto, 939 const struct tnl_ptk_info *tpi, 940 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 941 const struct ipv6hdr *ipv6h, 942 struct sk_buff *skb)) 943 { 944 struct ip6_tnl *t; 945 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 946 struct metadata_dst *tun_dst = NULL; 947 int ret = -1; 948 949 rcu_read_lock(); 950 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr); 951 952 if (t) { 953 u8 tproto = READ_ONCE(t->parms.proto); 954 955 if (tproto != ipproto && tproto != 0) 956 goto drop; 957 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 958 goto drop; 959 ipv6h = ipv6_hdr(skb); 960 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) 961 goto drop; 962 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 963 goto drop; 964 if (t->parms.collect_md) { 965 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0); 966 if (!tun_dst) 967 goto drop; 968 } 969 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 970 log_ecn_error); 971 } 972 973 rcu_read_unlock(); 974 975 return ret; 976 977 drop: 978 rcu_read_unlock(); 979 kfree_skb(skb); 980 return 0; 981 } 982 983 static int ip4ip6_rcv(struct sk_buff *skb) 984 { 985 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4, 986 ip4ip6_dscp_ecn_decapsulate); 987 } 988 989 static int ip6ip6_rcv(struct sk_buff *skb) 990 { 991 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6, 992 ip6ip6_dscp_ecn_decapsulate); 993 } 994 995 static int mplsip6_rcv(struct sk_buff *skb) 996 { 997 return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls, 998 mplsip6_dscp_ecn_decapsulate); 999 } 1000 1001 struct ipv6_tel_txoption { 1002 struct ipv6_txoptions ops; 1003 __u8 dst_opt[8]; 1004 }; 1005 1006 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 1007 { 1008 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 1009 1010 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 1011 opt->dst_opt[3] = 1; 1012 opt->dst_opt[4] = encap_limit; 1013 opt->dst_opt[5] = IPV6_TLV_PADN; 1014 opt->dst_opt[6] = 1; 1015 1016 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt; 1017 opt->ops.opt_nflen = 8; 1018 } 1019 1020 /** 1021 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 1022 * @t: the outgoing tunnel device 1023 * @hdr: IPv6 header from the incoming packet 1024 * 1025 * Description: 1026 * Avoid trivial tunneling loop by checking that tunnel exit-point 1027 * doesn't match source of incoming packet. 1028 * 1029 * Return: 1030 * 1 if conflict, 1031 * 0 else 1032 **/ 1033 1034 static inline bool 1035 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 1036 { 1037 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 1038 } 1039 1040 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, 1041 const struct in6_addr *laddr, 1042 const struct in6_addr *raddr) 1043 { 1044 struct __ip6_tnl_parm *p = &t->parms; 1045 int ret = 0; 1046 struct net *net = t->net; 1047 1048 if (t->parms.collect_md) 1049 return 1; 1050 1051 if ((p->flags & IP6_TNL_F_CAP_XMIT) || 1052 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 1053 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) { 1054 struct net_device *ldev = NULL; 1055 1056 rcu_read_lock(); 1057 if (p->link) 1058 ldev = dev_get_by_index_rcu(net, p->link); 1059 1060 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false, 1061 0, IFA_F_TENTATIVE))) 1062 pr_warn("%s xmit: Local address not yet configured!\n", 1063 p->name); 1064 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) && 1065 !ipv6_addr_is_multicast(raddr) && 1066 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev, 1067 true, 0, IFA_F_TENTATIVE))) 1068 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 1069 p->name); 1070 else 1071 ret = 1; 1072 rcu_read_unlock(); 1073 } 1074 return ret; 1075 } 1076 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 1077 1078 /** 1079 * ip6_tnl_xmit - encapsulate packet and send 1080 * @skb: the outgoing socket buffer 1081 * @dev: the outgoing tunnel device 1082 * @dsfield: dscp code for outer header 1083 * @fl6: flow of tunneled packet 1084 * @encap_limit: encapsulation limit 1085 * @pmtu: Path MTU is stored if packet is too big 1086 * @proto: next header value 1087 * 1088 * Description: 1089 * Build new header and do some sanity checks on the packet before sending 1090 * it. 1091 * 1092 * Return: 1093 * 0 on success 1094 * -1 fail 1095 * %-EMSGSIZE message too big. return mtu in this case. 1096 **/ 1097 1098 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield, 1099 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, 1100 __u8 proto) 1101 { 1102 struct ip6_tnl *t = netdev_priv(dev); 1103 struct net *net = t->net; 1104 struct net_device_stats *stats = &t->dev->stats; 1105 struct ipv6hdr *ipv6h; 1106 struct ipv6_tel_txoption opt; 1107 struct dst_entry *dst = NULL, *ndst = NULL; 1108 struct net_device *tdev; 1109 int mtu; 1110 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0; 1111 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen; 1112 unsigned int max_headroom = psh_hlen; 1113 bool use_cache = false; 1114 u8 hop_limit; 1115 int err = -1; 1116 1117 if (t->parms.collect_md) { 1118 hop_limit = skb_tunnel_info(skb)->key.ttl; 1119 goto route_lookup; 1120 } else { 1121 hop_limit = t->parms.hop_limit; 1122 } 1123 1124 /* NBMA tunnel */ 1125 if (ipv6_addr_any(&t->parms.raddr)) { 1126 if (skb->protocol == htons(ETH_P_IPV6)) { 1127 struct in6_addr *addr6; 1128 struct neighbour *neigh; 1129 int addr_type; 1130 1131 if (!skb_dst(skb)) 1132 goto tx_err_link_failure; 1133 1134 neigh = dst_neigh_lookup(skb_dst(skb), 1135 &ipv6_hdr(skb)->daddr); 1136 if (!neigh) 1137 goto tx_err_link_failure; 1138 1139 addr6 = (struct in6_addr *)&neigh->primary_key; 1140 addr_type = ipv6_addr_type(addr6); 1141 1142 if (addr_type == IPV6_ADDR_ANY) 1143 addr6 = &ipv6_hdr(skb)->daddr; 1144 1145 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr)); 1146 neigh_release(neigh); 1147 } 1148 } else if (t->parms.proto != 0 && !(t->parms.flags & 1149 (IP6_TNL_F_USE_ORIG_TCLASS | 1150 IP6_TNL_F_USE_ORIG_FWMARK))) { 1151 /* enable the cache only if neither the outer protocol nor the 1152 * routing decision depends on the current inner header value 1153 */ 1154 use_cache = true; 1155 } 1156 1157 if (use_cache) 1158 dst = dst_cache_get(&t->dst_cache); 1159 1160 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1161 goto tx_err_link_failure; 1162 1163 if (!dst) { 1164 route_lookup: 1165 /* add dsfield to flowlabel for route lookup */ 1166 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1167 1168 dst = ip6_route_output(net, NULL, fl6); 1169 1170 if (dst->error) 1171 goto tx_err_link_failure; 1172 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1173 if (IS_ERR(dst)) { 1174 err = PTR_ERR(dst); 1175 dst = NULL; 1176 goto tx_err_link_failure; 1177 } 1178 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) && 1179 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1180 &fl6->daddr, 0, &fl6->saddr)) 1181 goto tx_err_link_failure; 1182 ndst = dst; 1183 } 1184 1185 tdev = dst->dev; 1186 1187 if (tdev == dev) { 1188 stats->collisions++; 1189 net_warn_ratelimited("%s: Local routing loop detected!\n", 1190 t->parms.name); 1191 goto tx_err_dst_release; 1192 } 1193 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen; 1194 if (encap_limit >= 0) { 1195 max_headroom += 8; 1196 mtu -= 8; 1197 } 1198 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ? 1199 IPV6_MIN_MTU : IPV4_MIN_MTU); 1200 1201 skb_dst_update_pmtu_no_confirm(skb, mtu); 1202 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) { 1203 *pmtu = mtu; 1204 err = -EMSGSIZE; 1205 goto tx_err_dst_release; 1206 } 1207 1208 if (t->err_count > 0) { 1209 if (time_before(jiffies, 1210 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1211 t->err_count--; 1212 1213 dst_link_failure(skb); 1214 } else { 1215 t->err_count = 0; 1216 } 1217 } 1218 1219 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1220 1221 /* 1222 * Okay, now see if we can stuff it in the buffer as-is. 1223 */ 1224 max_headroom += LL_RESERVED_SPACE(tdev); 1225 1226 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1227 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1228 struct sk_buff *new_skb; 1229 1230 new_skb = skb_realloc_headroom(skb, max_headroom); 1231 if (!new_skb) 1232 goto tx_err_dst_release; 1233 1234 if (skb->sk) 1235 skb_set_owner_w(new_skb, skb->sk); 1236 consume_skb(skb); 1237 skb = new_skb; 1238 } 1239 1240 if (t->parms.collect_md) { 1241 if (t->encap.type != TUNNEL_ENCAP_NONE) 1242 goto tx_err_dst_release; 1243 } else { 1244 if (use_cache && ndst) 1245 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1246 } 1247 skb_dst_set(skb, dst); 1248 1249 if (hop_limit == 0) { 1250 if (skb->protocol == htons(ETH_P_IP)) 1251 hop_limit = ip_hdr(skb)->ttl; 1252 else if (skb->protocol == htons(ETH_P_IPV6)) 1253 hop_limit = ipv6_hdr(skb)->hop_limit; 1254 else 1255 hop_limit = ip6_dst_hoplimit(dst); 1256 } 1257 1258 /* Calculate max headroom for all the headers and adjust 1259 * needed_headroom if necessary. 1260 */ 1261 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1262 + dst->header_len + t->hlen; 1263 if (max_headroom > dev->needed_headroom) 1264 dev->needed_headroom = max_headroom; 1265 1266 err = ip6_tnl_encap(skb, t, &proto, fl6); 1267 if (err) 1268 return err; 1269 1270 if (encap_limit >= 0) { 1271 init_tel_txopt(&opt, encap_limit); 1272 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1273 } 1274 1275 skb_set_inner_ipproto(skb, proto); 1276 1277 skb_push(skb, sizeof(struct ipv6hdr)); 1278 skb_reset_network_header(skb); 1279 ipv6h = ipv6_hdr(skb); 1280 ip6_flow_hdr(ipv6h, dsfield, 1281 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1282 ipv6h->hop_limit = hop_limit; 1283 ipv6h->nexthdr = proto; 1284 ipv6h->saddr = fl6->saddr; 1285 ipv6h->daddr = fl6->daddr; 1286 ip6tunnel_xmit(NULL, skb, dev); 1287 return 0; 1288 tx_err_link_failure: 1289 stats->tx_carrier_errors++; 1290 dst_link_failure(skb); 1291 tx_err_dst_release: 1292 dst_release(dst); 1293 return err; 1294 } 1295 EXPORT_SYMBOL(ip6_tnl_xmit); 1296 1297 static inline int 1298 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, 1299 u8 protocol) 1300 { 1301 struct ip6_tnl *t = netdev_priv(dev); 1302 struct ipv6hdr *ipv6h; 1303 const struct iphdr *iph; 1304 int encap_limit = -1; 1305 __u16 offset; 1306 struct flowi6 fl6; 1307 __u8 dsfield, orig_dsfield; 1308 __u32 mtu; 1309 u8 tproto; 1310 int err; 1311 1312 tproto = READ_ONCE(t->parms.proto); 1313 if (tproto != protocol && tproto != 0) 1314 return -1; 1315 1316 if (t->parms.collect_md) { 1317 struct ip_tunnel_info *tun_info; 1318 const struct ip_tunnel_key *key; 1319 1320 tun_info = skb_tunnel_info(skb); 1321 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1322 ip_tunnel_info_af(tun_info) != AF_INET6)) 1323 return -1; 1324 key = &tun_info->key; 1325 memset(&fl6, 0, sizeof(fl6)); 1326 fl6.flowi6_proto = protocol; 1327 fl6.saddr = key->u.ipv6.src; 1328 fl6.daddr = key->u.ipv6.dst; 1329 fl6.flowlabel = key->label; 1330 dsfield = key->tos; 1331 switch (protocol) { 1332 case IPPROTO_IPIP: 1333 iph = ip_hdr(skb); 1334 orig_dsfield = ipv4_get_dsfield(iph); 1335 break; 1336 case IPPROTO_IPV6: 1337 ipv6h = ipv6_hdr(skb); 1338 orig_dsfield = ipv6_get_dsfield(ipv6h); 1339 break; 1340 default: 1341 orig_dsfield = dsfield; 1342 break; 1343 } 1344 } else { 1345 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1346 encap_limit = t->parms.encap_limit; 1347 if (protocol == IPPROTO_IPV6) { 1348 offset = ip6_tnl_parse_tlv_enc_lim(skb, 1349 skb_network_header(skb)); 1350 /* ip6_tnl_parse_tlv_enc_lim() might have 1351 * reallocated skb->head 1352 */ 1353 if (offset > 0) { 1354 struct ipv6_tlv_tnl_enc_lim *tel; 1355 1356 tel = (void *)&skb_network_header(skb)[offset]; 1357 if (tel->encap_limit == 0) { 1358 icmpv6_send(skb, ICMPV6_PARAMPROB, 1359 ICMPV6_HDR_FIELD, offset + 2); 1360 return -1; 1361 } 1362 encap_limit = tel->encap_limit - 1; 1363 } 1364 } 1365 1366 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1367 fl6.flowi6_proto = protocol; 1368 1369 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1370 fl6.flowi6_mark = skb->mark; 1371 else 1372 fl6.flowi6_mark = t->parms.fwmark; 1373 switch (protocol) { 1374 case IPPROTO_IPIP: 1375 iph = ip_hdr(skb); 1376 orig_dsfield = ipv4_get_dsfield(iph); 1377 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1378 dsfield = orig_dsfield; 1379 else 1380 dsfield = ip6_tclass(t->parms.flowinfo); 1381 break; 1382 case IPPROTO_IPV6: 1383 ipv6h = ipv6_hdr(skb); 1384 orig_dsfield = ipv6_get_dsfield(ipv6h); 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 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1390 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1391 break; 1392 default: 1393 orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo); 1394 break; 1395 } 1396 } 1397 1398 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1399 dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield); 1400 1401 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1402 return -1; 1403 1404 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1405 protocol); 1406 if (err != 0) { 1407 /* XXX: send ICMP error even if DF is not set. */ 1408 if (err == -EMSGSIZE) 1409 switch (protocol) { 1410 case IPPROTO_IPIP: 1411 icmp_send(skb, ICMP_DEST_UNREACH, 1412 ICMP_FRAG_NEEDED, htonl(mtu)); 1413 break; 1414 case IPPROTO_IPV6: 1415 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1416 break; 1417 default: 1418 break; 1419 } 1420 return -1; 1421 } 1422 1423 return 0; 1424 } 1425 1426 static netdev_tx_t 1427 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1428 { 1429 struct ip6_tnl *t = netdev_priv(dev); 1430 struct net_device_stats *stats = &t->dev->stats; 1431 u8 ipproto; 1432 int ret; 1433 1434 if (!pskb_inet_may_pull(skb)) 1435 goto tx_err; 1436 1437 switch (skb->protocol) { 1438 case htons(ETH_P_IP): 1439 ipproto = IPPROTO_IPIP; 1440 break; 1441 case htons(ETH_P_IPV6): 1442 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb))) 1443 goto tx_err; 1444 ipproto = IPPROTO_IPV6; 1445 break; 1446 case htons(ETH_P_MPLS_UC): 1447 ipproto = IPPROTO_MPLS; 1448 break; 1449 default: 1450 goto tx_err; 1451 } 1452 1453 ret = ipxip6_tnl_xmit(skb, dev, ipproto); 1454 if (ret < 0) 1455 goto tx_err; 1456 1457 return NETDEV_TX_OK; 1458 1459 tx_err: 1460 stats->tx_errors++; 1461 stats->tx_dropped++; 1462 kfree_skb(skb); 1463 return NETDEV_TX_OK; 1464 } 1465 1466 static void ip6_tnl_link_config(struct ip6_tnl *t) 1467 { 1468 struct net_device *dev = t->dev; 1469 struct net_device *tdev = NULL; 1470 struct __ip6_tnl_parm *p = &t->parms; 1471 struct flowi6 *fl6 = &t->fl.u.ip6; 1472 unsigned int mtu; 1473 int t_hlen; 1474 1475 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1476 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1477 1478 /* Set up flowi template */ 1479 fl6->saddr = p->laddr; 1480 fl6->daddr = p->raddr; 1481 fl6->flowi6_oif = p->link; 1482 fl6->flowlabel = 0; 1483 1484 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1485 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1486 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1487 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1488 1489 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1490 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1491 1492 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1493 dev->flags |= IFF_POINTOPOINT; 1494 else 1495 dev->flags &= ~IFF_POINTOPOINT; 1496 1497 t->tun_hlen = 0; 1498 t->hlen = t->encap_hlen + t->tun_hlen; 1499 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1500 1501 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1502 int strict = (ipv6_addr_type(&p->raddr) & 1503 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1504 1505 struct rt6_info *rt = rt6_lookup(t->net, 1506 &p->raddr, &p->laddr, 1507 p->link, NULL, strict); 1508 if (rt) { 1509 tdev = rt->dst.dev; 1510 ip6_rt_put(rt); 1511 } 1512 1513 if (!tdev && p->link) 1514 tdev = __dev_get_by_index(t->net, p->link); 1515 1516 if (tdev) { 1517 dev->hard_header_len = tdev->hard_header_len + t_hlen; 1518 mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU); 1519 1520 dev->mtu = mtu - t_hlen; 1521 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1522 dev->mtu -= 8; 1523 1524 if (dev->mtu < IPV6_MIN_MTU) 1525 dev->mtu = IPV6_MIN_MTU; 1526 } 1527 } 1528 } 1529 1530 /** 1531 * ip6_tnl_change - update the tunnel parameters 1532 * @t: tunnel to be changed 1533 * @p: tunnel configuration parameters 1534 * 1535 * Description: 1536 * ip6_tnl_change() updates the tunnel parameters 1537 **/ 1538 1539 static int 1540 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1541 { 1542 t->parms.laddr = p->laddr; 1543 t->parms.raddr = p->raddr; 1544 t->parms.flags = p->flags; 1545 t->parms.hop_limit = p->hop_limit; 1546 t->parms.encap_limit = p->encap_limit; 1547 t->parms.flowinfo = p->flowinfo; 1548 t->parms.link = p->link; 1549 t->parms.proto = p->proto; 1550 t->parms.fwmark = p->fwmark; 1551 dst_cache_reset(&t->dst_cache); 1552 ip6_tnl_link_config(t); 1553 return 0; 1554 } 1555 1556 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1557 { 1558 struct net *net = t->net; 1559 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1560 int err; 1561 1562 ip6_tnl_unlink(ip6n, t); 1563 synchronize_net(); 1564 err = ip6_tnl_change(t, p); 1565 ip6_tnl_link(ip6n, t); 1566 netdev_state_change(t->dev); 1567 return err; 1568 } 1569 1570 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1571 { 1572 /* for default tnl0 device allow to change only the proto */ 1573 t->parms.proto = p->proto; 1574 netdev_state_change(t->dev); 1575 return 0; 1576 } 1577 1578 static void 1579 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1580 { 1581 p->laddr = u->laddr; 1582 p->raddr = u->raddr; 1583 p->flags = u->flags; 1584 p->hop_limit = u->hop_limit; 1585 p->encap_limit = u->encap_limit; 1586 p->flowinfo = u->flowinfo; 1587 p->link = u->link; 1588 p->proto = u->proto; 1589 memcpy(p->name, u->name, sizeof(u->name)); 1590 } 1591 1592 static void 1593 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1594 { 1595 u->laddr = p->laddr; 1596 u->raddr = p->raddr; 1597 u->flags = p->flags; 1598 u->hop_limit = p->hop_limit; 1599 u->encap_limit = p->encap_limit; 1600 u->flowinfo = p->flowinfo; 1601 u->link = p->link; 1602 u->proto = p->proto; 1603 memcpy(u->name, p->name, sizeof(u->name)); 1604 } 1605 1606 /** 1607 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1608 * @dev: virtual device associated with tunnel 1609 * @ifr: parameters passed from userspace 1610 * @cmd: command to be performed 1611 * 1612 * Description: 1613 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1614 * from userspace. 1615 * 1616 * The possible commands are the following: 1617 * %SIOCGETTUNNEL: get tunnel parameters for device 1618 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1619 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1620 * %SIOCDELTUNNEL: delete tunnel 1621 * 1622 * The fallback device "ip6tnl0", created during module 1623 * initialization, can be used for creating other tunnel devices. 1624 * 1625 * Return: 1626 * 0 on success, 1627 * %-EFAULT if unable to copy data to or from userspace, 1628 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1629 * %-EINVAL if passed tunnel parameters are invalid, 1630 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1631 * %-ENODEV if attempting to change or delete a nonexisting device 1632 **/ 1633 1634 static int 1635 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1636 { 1637 int err = 0; 1638 struct ip6_tnl_parm p; 1639 struct __ip6_tnl_parm p1; 1640 struct ip6_tnl *t = netdev_priv(dev); 1641 struct net *net = t->net; 1642 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1643 1644 memset(&p1, 0, sizeof(p1)); 1645 1646 switch (cmd) { 1647 case SIOCGETTUNNEL: 1648 if (dev == ip6n->fb_tnl_dev) { 1649 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1650 err = -EFAULT; 1651 break; 1652 } 1653 ip6_tnl_parm_from_user(&p1, &p); 1654 t = ip6_tnl_locate(net, &p1, 0); 1655 if (IS_ERR(t)) 1656 t = netdev_priv(dev); 1657 } else { 1658 memset(&p, 0, sizeof(p)); 1659 } 1660 ip6_tnl_parm_to_user(&p, &t->parms); 1661 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) { 1662 err = -EFAULT; 1663 } 1664 break; 1665 case SIOCADDTUNNEL: 1666 case SIOCCHGTUNNEL: 1667 err = -EPERM; 1668 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1669 break; 1670 err = -EFAULT; 1671 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1672 break; 1673 err = -EINVAL; 1674 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1675 p.proto != 0) 1676 break; 1677 ip6_tnl_parm_from_user(&p1, &p); 1678 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1679 if (cmd == SIOCCHGTUNNEL) { 1680 if (!IS_ERR(t)) { 1681 if (t->dev != dev) { 1682 err = -EEXIST; 1683 break; 1684 } 1685 } else 1686 t = netdev_priv(dev); 1687 if (dev == ip6n->fb_tnl_dev) 1688 err = ip6_tnl0_update(t, &p1); 1689 else 1690 err = ip6_tnl_update(t, &p1); 1691 } 1692 if (!IS_ERR(t)) { 1693 err = 0; 1694 ip6_tnl_parm_to_user(&p, &t->parms); 1695 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1696 err = -EFAULT; 1697 1698 } else { 1699 err = PTR_ERR(t); 1700 } 1701 break; 1702 case SIOCDELTUNNEL: 1703 err = -EPERM; 1704 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1705 break; 1706 1707 if (dev == ip6n->fb_tnl_dev) { 1708 err = -EFAULT; 1709 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1710 break; 1711 err = -ENOENT; 1712 ip6_tnl_parm_from_user(&p1, &p); 1713 t = ip6_tnl_locate(net, &p1, 0); 1714 if (IS_ERR(t)) 1715 break; 1716 err = -EPERM; 1717 if (t->dev == ip6n->fb_tnl_dev) 1718 break; 1719 dev = t->dev; 1720 } 1721 err = 0; 1722 unregister_netdevice(dev); 1723 break; 1724 default: 1725 err = -EINVAL; 1726 } 1727 return err; 1728 } 1729 1730 /** 1731 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1732 * @dev: virtual device associated with tunnel 1733 * @new_mtu: the new mtu 1734 * 1735 * Return: 1736 * 0 on success, 1737 * %-EINVAL if mtu too small 1738 **/ 1739 1740 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1741 { 1742 struct ip6_tnl *tnl = netdev_priv(dev); 1743 1744 if (tnl->parms.proto == IPPROTO_IPV6) { 1745 if (new_mtu < IPV6_MIN_MTU) 1746 return -EINVAL; 1747 } else { 1748 if (new_mtu < ETH_MIN_MTU) 1749 return -EINVAL; 1750 } 1751 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) { 1752 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len) 1753 return -EINVAL; 1754 } else { 1755 if (new_mtu > IP_MAX_MTU - dev->hard_header_len) 1756 return -EINVAL; 1757 } 1758 dev->mtu = new_mtu; 1759 return 0; 1760 } 1761 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1762 1763 int ip6_tnl_get_iflink(const struct net_device *dev) 1764 { 1765 struct ip6_tnl *t = netdev_priv(dev); 1766 1767 return t->parms.link; 1768 } 1769 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1770 1771 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1772 unsigned int num) 1773 { 1774 if (num >= MAX_IPTUN_ENCAP_OPS) 1775 return -ERANGE; 1776 1777 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1778 &ip6tun_encaps[num], 1779 NULL, ops) ? 0 : -1; 1780 } 1781 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1782 1783 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1784 unsigned int num) 1785 { 1786 int ret; 1787 1788 if (num >= MAX_IPTUN_ENCAP_OPS) 1789 return -ERANGE; 1790 1791 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1792 &ip6tun_encaps[num], 1793 ops, NULL) == ops) ? 0 : -1; 1794 1795 synchronize_net(); 1796 1797 return ret; 1798 } 1799 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1800 1801 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1802 struct ip_tunnel_encap *ipencap) 1803 { 1804 int hlen; 1805 1806 memset(&t->encap, 0, sizeof(t->encap)); 1807 1808 hlen = ip6_encap_hlen(ipencap); 1809 if (hlen < 0) 1810 return hlen; 1811 1812 t->encap.type = ipencap->type; 1813 t->encap.sport = ipencap->sport; 1814 t->encap.dport = ipencap->dport; 1815 t->encap.flags = ipencap->flags; 1816 1817 t->encap_hlen = hlen; 1818 t->hlen = t->encap_hlen + t->tun_hlen; 1819 1820 return 0; 1821 } 1822 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1823 1824 static const struct net_device_ops ip6_tnl_netdev_ops = { 1825 .ndo_init = ip6_tnl_dev_init, 1826 .ndo_uninit = ip6_tnl_dev_uninit, 1827 .ndo_start_xmit = ip6_tnl_start_xmit, 1828 .ndo_do_ioctl = ip6_tnl_ioctl, 1829 .ndo_change_mtu = ip6_tnl_change_mtu, 1830 .ndo_get_stats = ip6_get_stats, 1831 .ndo_get_iflink = ip6_tnl_get_iflink, 1832 }; 1833 1834 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1835 NETIF_F_FRAGLIST | \ 1836 NETIF_F_HIGHDMA | \ 1837 NETIF_F_GSO_SOFTWARE | \ 1838 NETIF_F_HW_CSUM) 1839 1840 /** 1841 * ip6_tnl_dev_setup - setup virtual tunnel device 1842 * @dev: virtual device associated with tunnel 1843 * 1844 * Description: 1845 * Initialize function pointers and device parameters 1846 **/ 1847 1848 static void ip6_tnl_dev_setup(struct net_device *dev) 1849 { 1850 dev->netdev_ops = &ip6_tnl_netdev_ops; 1851 dev->header_ops = &ip_tunnel_header_ops; 1852 dev->needs_free_netdev = true; 1853 dev->priv_destructor = ip6_dev_free; 1854 1855 dev->type = ARPHRD_TUNNEL6; 1856 dev->flags |= IFF_NOARP; 1857 dev->addr_len = sizeof(struct in6_addr); 1858 dev->features |= NETIF_F_LLTX; 1859 netif_keep_dst(dev); 1860 1861 dev->features |= IPXIPX_FEATURES; 1862 dev->hw_features |= IPXIPX_FEATURES; 1863 1864 /* This perm addr will be used as interface identifier by IPv6 */ 1865 dev->addr_assign_type = NET_ADDR_RANDOM; 1866 eth_random_addr(dev->perm_addr); 1867 } 1868 1869 1870 /** 1871 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1872 * @dev: virtual device associated with tunnel 1873 **/ 1874 1875 static inline int 1876 ip6_tnl_dev_init_gen(struct net_device *dev) 1877 { 1878 struct ip6_tnl *t = netdev_priv(dev); 1879 int ret; 1880 int t_hlen; 1881 1882 t->dev = dev; 1883 t->net = dev_net(dev); 1884 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1885 if (!dev->tstats) 1886 return -ENOMEM; 1887 1888 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1889 if (ret) 1890 goto free_stats; 1891 1892 ret = gro_cells_init(&t->gro_cells, dev); 1893 if (ret) 1894 goto destroy_dst; 1895 1896 t->tun_hlen = 0; 1897 t->hlen = t->encap_hlen + t->tun_hlen; 1898 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1899 1900 dev->type = ARPHRD_TUNNEL6; 1901 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1902 dev->mtu = ETH_DATA_LEN - t_hlen; 1903 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1904 dev->mtu -= 8; 1905 dev->min_mtu = ETH_MIN_MTU; 1906 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len; 1907 1908 return 0; 1909 1910 destroy_dst: 1911 dst_cache_destroy(&t->dst_cache); 1912 free_stats: 1913 free_percpu(dev->tstats); 1914 dev->tstats = NULL; 1915 1916 return ret; 1917 } 1918 1919 /** 1920 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1921 * @dev: virtual device associated with tunnel 1922 **/ 1923 1924 static int ip6_tnl_dev_init(struct net_device *dev) 1925 { 1926 struct ip6_tnl *t = netdev_priv(dev); 1927 int err = ip6_tnl_dev_init_gen(dev); 1928 1929 if (err) 1930 return err; 1931 ip6_tnl_link_config(t); 1932 if (t->parms.collect_md) 1933 netif_keep_dst(dev); 1934 return 0; 1935 } 1936 1937 /** 1938 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1939 * @dev: fallback device 1940 * 1941 * Return: 0 1942 **/ 1943 1944 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1945 { 1946 struct ip6_tnl *t = netdev_priv(dev); 1947 struct net *net = dev_net(dev); 1948 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1949 1950 t->parms.proto = IPPROTO_IPV6; 1951 dev_hold(dev); 1952 1953 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1954 return 0; 1955 } 1956 1957 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[], 1958 struct netlink_ext_ack *extack) 1959 { 1960 u8 proto; 1961 1962 if (!data || !data[IFLA_IPTUN_PROTO]) 1963 return 0; 1964 1965 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1966 if (proto != IPPROTO_IPV6 && 1967 proto != IPPROTO_IPIP && 1968 proto != 0) 1969 return -EINVAL; 1970 1971 return 0; 1972 } 1973 1974 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1975 struct __ip6_tnl_parm *parms) 1976 { 1977 memset(parms, 0, sizeof(*parms)); 1978 1979 if (!data) 1980 return; 1981 1982 if (data[IFLA_IPTUN_LINK]) 1983 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1984 1985 if (data[IFLA_IPTUN_LOCAL]) 1986 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1987 1988 if (data[IFLA_IPTUN_REMOTE]) 1989 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1990 1991 if (data[IFLA_IPTUN_TTL]) 1992 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1993 1994 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1995 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1996 1997 if (data[IFLA_IPTUN_FLOWINFO]) 1998 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1999 2000 if (data[IFLA_IPTUN_FLAGS]) 2001 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 2002 2003 if (data[IFLA_IPTUN_PROTO]) 2004 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 2005 2006 if (data[IFLA_IPTUN_COLLECT_METADATA]) 2007 parms->collect_md = true; 2008 2009 if (data[IFLA_IPTUN_FWMARK]) 2010 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 2011 } 2012 2013 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[], 2014 struct ip_tunnel_encap *ipencap) 2015 { 2016 bool ret = false; 2017 2018 memset(ipencap, 0, sizeof(*ipencap)); 2019 2020 if (!data) 2021 return ret; 2022 2023 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 2024 ret = true; 2025 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 2026 } 2027 2028 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 2029 ret = true; 2030 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 2031 } 2032 2033 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 2034 ret = true; 2035 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 2036 } 2037 2038 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 2039 ret = true; 2040 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 2041 } 2042 2043 return ret; 2044 } 2045 2046 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 2047 struct nlattr *tb[], struct nlattr *data[], 2048 struct netlink_ext_ack *extack) 2049 { 2050 struct net *net = dev_net(dev); 2051 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2052 struct ip_tunnel_encap ipencap; 2053 struct ip6_tnl *nt, *t; 2054 int err; 2055 2056 nt = netdev_priv(dev); 2057 2058 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2059 err = ip6_tnl_encap_setup(nt, &ipencap); 2060 if (err < 0) 2061 return err; 2062 } 2063 2064 ip6_tnl_netlink_parms(data, &nt->parms); 2065 2066 if (nt->parms.collect_md) { 2067 if (rtnl_dereference(ip6n->collect_md_tun)) 2068 return -EEXIST; 2069 } else { 2070 t = ip6_tnl_locate(net, &nt->parms, 0); 2071 if (!IS_ERR(t)) 2072 return -EEXIST; 2073 } 2074 2075 err = ip6_tnl_create2(dev); 2076 if (!err && tb[IFLA_MTU]) 2077 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2078 2079 return err; 2080 } 2081 2082 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2083 struct nlattr *data[], 2084 struct netlink_ext_ack *extack) 2085 { 2086 struct ip6_tnl *t = netdev_priv(dev); 2087 struct __ip6_tnl_parm p; 2088 struct net *net = t->net; 2089 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2090 struct ip_tunnel_encap ipencap; 2091 2092 if (dev == ip6n->fb_tnl_dev) 2093 return -EINVAL; 2094 2095 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2096 int err = ip6_tnl_encap_setup(t, &ipencap); 2097 2098 if (err < 0) 2099 return err; 2100 } 2101 ip6_tnl_netlink_parms(data, &p); 2102 if (p.collect_md) 2103 return -EINVAL; 2104 2105 t = ip6_tnl_locate(net, &p, 0); 2106 if (!IS_ERR(t)) { 2107 if (t->dev != dev) 2108 return -EEXIST; 2109 } else 2110 t = netdev_priv(dev); 2111 2112 return ip6_tnl_update(t, &p); 2113 } 2114 2115 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2116 { 2117 struct net *net = dev_net(dev); 2118 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2119 2120 if (dev != ip6n->fb_tnl_dev) 2121 unregister_netdevice_queue(dev, head); 2122 } 2123 2124 static size_t ip6_tnl_get_size(const struct net_device *dev) 2125 { 2126 return 2127 /* IFLA_IPTUN_LINK */ 2128 nla_total_size(4) + 2129 /* IFLA_IPTUN_LOCAL */ 2130 nla_total_size(sizeof(struct in6_addr)) + 2131 /* IFLA_IPTUN_REMOTE */ 2132 nla_total_size(sizeof(struct in6_addr)) + 2133 /* IFLA_IPTUN_TTL */ 2134 nla_total_size(1) + 2135 /* IFLA_IPTUN_ENCAP_LIMIT */ 2136 nla_total_size(1) + 2137 /* IFLA_IPTUN_FLOWINFO */ 2138 nla_total_size(4) + 2139 /* IFLA_IPTUN_FLAGS */ 2140 nla_total_size(4) + 2141 /* IFLA_IPTUN_PROTO */ 2142 nla_total_size(1) + 2143 /* IFLA_IPTUN_ENCAP_TYPE */ 2144 nla_total_size(2) + 2145 /* IFLA_IPTUN_ENCAP_FLAGS */ 2146 nla_total_size(2) + 2147 /* IFLA_IPTUN_ENCAP_SPORT */ 2148 nla_total_size(2) + 2149 /* IFLA_IPTUN_ENCAP_DPORT */ 2150 nla_total_size(2) + 2151 /* IFLA_IPTUN_COLLECT_METADATA */ 2152 nla_total_size(0) + 2153 /* IFLA_IPTUN_FWMARK */ 2154 nla_total_size(4) + 2155 0; 2156 } 2157 2158 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2159 { 2160 struct ip6_tnl *tunnel = netdev_priv(dev); 2161 struct __ip6_tnl_parm *parm = &tunnel->parms; 2162 2163 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2164 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2165 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2166 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2167 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2168 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2169 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2170 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2171 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2172 goto nla_put_failure; 2173 2174 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2175 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2176 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2177 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2178 goto nla_put_failure; 2179 2180 if (parm->collect_md) 2181 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2182 goto nla_put_failure; 2183 2184 return 0; 2185 2186 nla_put_failure: 2187 return -EMSGSIZE; 2188 } 2189 2190 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2191 { 2192 struct ip6_tnl *tunnel = netdev_priv(dev); 2193 2194 return tunnel->net; 2195 } 2196 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2197 2198 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2199 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2200 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2201 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2202 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2203 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2204 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2205 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2206 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2207 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2208 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2209 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2210 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2211 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2212 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2213 }; 2214 2215 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2216 .kind = "ip6tnl", 2217 .maxtype = IFLA_IPTUN_MAX, 2218 .policy = ip6_tnl_policy, 2219 .priv_size = sizeof(struct ip6_tnl), 2220 .setup = ip6_tnl_dev_setup, 2221 .validate = ip6_tnl_validate, 2222 .newlink = ip6_tnl_newlink, 2223 .changelink = ip6_tnl_changelink, 2224 .dellink = ip6_tnl_dellink, 2225 .get_size = ip6_tnl_get_size, 2226 .fill_info = ip6_tnl_fill_info, 2227 .get_link_net = ip6_tnl_get_link_net, 2228 }; 2229 2230 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2231 .handler = ip4ip6_rcv, 2232 .err_handler = ip4ip6_err, 2233 .priority = 1, 2234 }; 2235 2236 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2237 .handler = ip6ip6_rcv, 2238 .err_handler = ip6ip6_err, 2239 .priority = 1, 2240 }; 2241 2242 static struct xfrm6_tunnel mplsip6_handler __read_mostly = { 2243 .handler = mplsip6_rcv, 2244 .err_handler = mplsip6_err, 2245 .priority = 1, 2246 }; 2247 2248 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list) 2249 { 2250 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2251 struct net_device *dev, *aux; 2252 int h; 2253 struct ip6_tnl *t; 2254 2255 for_each_netdev_safe(net, dev, aux) 2256 if (dev->rtnl_link_ops == &ip6_link_ops) 2257 unregister_netdevice_queue(dev, list); 2258 2259 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2260 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2261 while (t) { 2262 /* If dev is in the same netns, it has already 2263 * been added to the list by the previous loop. 2264 */ 2265 if (!net_eq(dev_net(t->dev), net)) 2266 unregister_netdevice_queue(t->dev, list); 2267 t = rtnl_dereference(t->next); 2268 } 2269 } 2270 } 2271 2272 static int __net_init ip6_tnl_init_net(struct net *net) 2273 { 2274 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2275 struct ip6_tnl *t = NULL; 2276 int err; 2277 2278 ip6n->tnls[0] = ip6n->tnls_wc; 2279 ip6n->tnls[1] = ip6n->tnls_r_l; 2280 2281 if (!net_has_fallback_tunnels(net)) 2282 return 0; 2283 err = -ENOMEM; 2284 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2285 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2286 2287 if (!ip6n->fb_tnl_dev) 2288 goto err_alloc_dev; 2289 dev_net_set(ip6n->fb_tnl_dev, net); 2290 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2291 /* FB netdevice is special: we have one, and only one per netns. 2292 * Allowing to move it to another netns is clearly unsafe. 2293 */ 2294 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 2295 2296 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2297 if (err < 0) 2298 goto err_register; 2299 2300 err = register_netdev(ip6n->fb_tnl_dev); 2301 if (err < 0) 2302 goto err_register; 2303 2304 t = netdev_priv(ip6n->fb_tnl_dev); 2305 2306 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2307 return 0; 2308 2309 err_register: 2310 free_netdev(ip6n->fb_tnl_dev); 2311 err_alloc_dev: 2312 return err; 2313 } 2314 2315 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list) 2316 { 2317 struct net *net; 2318 LIST_HEAD(list); 2319 2320 rtnl_lock(); 2321 list_for_each_entry(net, net_list, exit_list) 2322 ip6_tnl_destroy_tunnels(net, &list); 2323 unregister_netdevice_many(&list); 2324 rtnl_unlock(); 2325 } 2326 2327 static struct pernet_operations ip6_tnl_net_ops = { 2328 .init = ip6_tnl_init_net, 2329 .exit_batch = ip6_tnl_exit_batch_net, 2330 .id = &ip6_tnl_net_id, 2331 .size = sizeof(struct ip6_tnl_net), 2332 }; 2333 2334 /** 2335 * ip6_tunnel_init - register protocol and reserve needed resources 2336 * 2337 * Return: 0 on success 2338 **/ 2339 2340 static int __init ip6_tunnel_init(void) 2341 { 2342 int err; 2343 2344 if (!ipv6_mod_enabled()) 2345 return -EOPNOTSUPP; 2346 2347 err = register_pernet_device(&ip6_tnl_net_ops); 2348 if (err < 0) 2349 goto out_pernet; 2350 2351 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2352 if (err < 0) { 2353 pr_err("%s: can't register ip4ip6\n", __func__); 2354 goto out_ip4ip6; 2355 } 2356 2357 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2358 if (err < 0) { 2359 pr_err("%s: can't register ip6ip6\n", __func__); 2360 goto out_ip6ip6; 2361 } 2362 2363 if (ip6_tnl_mpls_supported()) { 2364 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS); 2365 if (err < 0) { 2366 pr_err("%s: can't register mplsip6\n", __func__); 2367 goto out_mplsip6; 2368 } 2369 } 2370 2371 err = rtnl_link_register(&ip6_link_ops); 2372 if (err < 0) 2373 goto rtnl_link_failed; 2374 2375 return 0; 2376 2377 rtnl_link_failed: 2378 if (ip6_tnl_mpls_supported()) 2379 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS); 2380 out_mplsip6: 2381 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2382 out_ip6ip6: 2383 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2384 out_ip4ip6: 2385 unregister_pernet_device(&ip6_tnl_net_ops); 2386 out_pernet: 2387 return err; 2388 } 2389 2390 /** 2391 * ip6_tunnel_cleanup - free resources and unregister protocol 2392 **/ 2393 2394 static void __exit ip6_tunnel_cleanup(void) 2395 { 2396 rtnl_link_unregister(&ip6_link_ops); 2397 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2398 pr_info("%s: can't deregister ip4ip6\n", __func__); 2399 2400 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2401 pr_info("%s: can't deregister ip6ip6\n", __func__); 2402 2403 if (ip6_tnl_mpls_supported() && 2404 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS)) 2405 pr_info("%s: can't deregister mplsip6\n", __func__); 2406 unregister_pernet_device(&ip6_tnl_net_ops); 2407 } 2408 2409 module_init(ip6_tunnel_init); 2410 module_exit(ip6_tunnel_cleanup); 2411