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 #define for_each_ip6_tunnel_rcu(start) \ 98 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 99 100 /** 101 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 102 * @net: network namespace 103 * @link: ifindex of underlying interface 104 * @remote: the address of the tunnel exit-point 105 * @local: the address of the tunnel entry-point 106 * 107 * Return: 108 * tunnel matching given end-points if found, 109 * else fallback tunnel if its device is up, 110 * else %NULL 111 **/ 112 113 static struct ip6_tnl * 114 ip6_tnl_lookup(struct net *net, int link, 115 const struct in6_addr *remote, const struct in6_addr *local) 116 { 117 unsigned int hash = HASH(remote, local); 118 struct ip6_tnl *t, *cand = NULL; 119 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 120 struct in6_addr any; 121 122 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 123 if (!ipv6_addr_equal(local, &t->parms.laddr) || 124 !ipv6_addr_equal(remote, &t->parms.raddr) || 125 !(t->dev->flags & IFF_UP)) 126 continue; 127 128 if (link == t->parms.link) 129 return t; 130 else 131 cand = t; 132 } 133 134 memset(&any, 0, sizeof(any)); 135 hash = HASH(&any, local); 136 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 137 if (!ipv6_addr_equal(local, &t->parms.laddr) || 138 !ipv6_addr_any(&t->parms.raddr) || 139 !(t->dev->flags & IFF_UP)) 140 continue; 141 142 if (link == t->parms.link) 143 return t; 144 else if (!cand) 145 cand = t; 146 } 147 148 hash = HASH(remote, &any); 149 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 150 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 151 !ipv6_addr_any(&t->parms.laddr) || 152 !(t->dev->flags & IFF_UP)) 153 continue; 154 155 if (link == t->parms.link) 156 return t; 157 else if (!cand) 158 cand = t; 159 } 160 161 if (cand) 162 return cand; 163 164 t = rcu_dereference(ip6n->collect_md_tun); 165 if (t && t->dev->flags & IFF_UP) 166 return t; 167 168 t = rcu_dereference(ip6n->tnls_wc[0]); 169 if (t && (t->dev->flags & IFF_UP)) 170 return t; 171 172 return NULL; 173 } 174 175 /** 176 * ip6_tnl_bucket - get head of list matching given tunnel parameters 177 * @ip6n: the private data for ip6_vti in the netns 178 * @p: parameters containing tunnel end-points 179 * 180 * Description: 181 * ip6_tnl_bucket() returns the head of the list matching the 182 * &struct in6_addr entries laddr and raddr in @p. 183 * 184 * Return: head of IPv6 tunnel list 185 **/ 186 187 static struct ip6_tnl __rcu ** 188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 189 { 190 const struct in6_addr *remote = &p->raddr; 191 const struct in6_addr *local = &p->laddr; 192 unsigned int h = 0; 193 int prio = 0; 194 195 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 196 prio = 1; 197 h = HASH(remote, local); 198 } 199 return &ip6n->tnls[prio][h]; 200 } 201 202 /** 203 * ip6_tnl_link - add tunnel to hash table 204 * @ip6n: the private data for ip6_vti in the netns 205 * @t: tunnel to be added 206 **/ 207 208 static void 209 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 210 { 211 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 212 213 if (t->parms.collect_md) 214 rcu_assign_pointer(ip6n->collect_md_tun, t); 215 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 216 rcu_assign_pointer(*tp, t); 217 } 218 219 /** 220 * ip6_tnl_unlink - remove tunnel from hash table 221 * @ip6n: the private data for ip6_vti in the netns 222 * @t: tunnel to be removed 223 **/ 224 225 static void 226 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 227 { 228 struct ip6_tnl __rcu **tp; 229 struct ip6_tnl *iter; 230 231 if (t->parms.collect_md) 232 rcu_assign_pointer(ip6n->collect_md_tun, NULL); 233 234 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 235 (iter = rtnl_dereference(*tp)) != NULL; 236 tp = &iter->next) { 237 if (t == iter) { 238 rcu_assign_pointer(*tp, t->next); 239 break; 240 } 241 } 242 } 243 244 static void ip6_dev_free(struct net_device *dev) 245 { 246 struct ip6_tnl *t = netdev_priv(dev); 247 248 gro_cells_destroy(&t->gro_cells); 249 dst_cache_destroy(&t->dst_cache); 250 free_percpu(dev->tstats); 251 } 252 253 static int ip6_tnl_create2(struct net_device *dev) 254 { 255 struct ip6_tnl *t = netdev_priv(dev); 256 struct net *net = dev_net(dev); 257 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 258 int err; 259 260 t = netdev_priv(dev); 261 262 dev->rtnl_link_ops = &ip6_link_ops; 263 err = register_netdevice(dev); 264 if (err < 0) 265 goto out; 266 267 strcpy(t->parms.name, dev->name); 268 269 ip6_tnl_link(ip6n, t); 270 return 0; 271 272 out: 273 return err; 274 } 275 276 /** 277 * ip6_tnl_create - create a new tunnel 278 * @net: network namespace 279 * @p: tunnel parameters 280 * 281 * Description: 282 * Create tunnel matching given parameters. 283 * 284 * Return: 285 * created tunnel or error pointer 286 **/ 287 288 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 289 { 290 struct net_device *dev; 291 struct ip6_tnl *t; 292 char name[IFNAMSIZ]; 293 int err = -E2BIG; 294 295 if (p->name[0]) { 296 if (!dev_valid_name(p->name)) 297 goto failed; 298 strlcpy(name, p->name, IFNAMSIZ); 299 } else { 300 sprintf(name, "ip6tnl%%d"); 301 } 302 err = -ENOMEM; 303 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 304 ip6_tnl_dev_setup); 305 if (!dev) 306 goto failed; 307 308 dev_net_set(dev, net); 309 310 t = netdev_priv(dev); 311 t->parms = *p; 312 t->net = dev_net(dev); 313 err = ip6_tnl_create2(dev); 314 if (err < 0) 315 goto failed_free; 316 317 return t; 318 319 failed_free: 320 free_netdev(dev); 321 failed: 322 return ERR_PTR(err); 323 } 324 325 /** 326 * ip6_tnl_locate - find or create tunnel matching given parameters 327 * @net: network namespace 328 * @p: tunnel parameters 329 * @create: != 0 if allowed to create new tunnel if no match found 330 * 331 * Description: 332 * ip6_tnl_locate() first tries to locate an existing tunnel 333 * based on @parms. If this is unsuccessful, but @create is set a new 334 * tunnel device is created and registered for use. 335 * 336 * Return: 337 * matching tunnel or error pointer 338 **/ 339 340 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 341 struct __ip6_tnl_parm *p, int create) 342 { 343 const struct in6_addr *remote = &p->raddr; 344 const struct in6_addr *local = &p->laddr; 345 struct ip6_tnl __rcu **tp; 346 struct ip6_tnl *t; 347 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 348 349 for (tp = ip6_tnl_bucket(ip6n, p); 350 (t = rtnl_dereference(*tp)) != NULL; 351 tp = &t->next) { 352 if (ipv6_addr_equal(local, &t->parms.laddr) && 353 ipv6_addr_equal(remote, &t->parms.raddr) && 354 p->link == t->parms.link) { 355 if (create) 356 return ERR_PTR(-EEXIST); 357 358 return t; 359 } 360 } 361 if (!create) 362 return ERR_PTR(-ENODEV); 363 return ip6_tnl_create(net, p); 364 } 365 366 /** 367 * ip6_tnl_dev_uninit - tunnel device uninitializer 368 * @dev: the device to be destroyed 369 * 370 * Description: 371 * ip6_tnl_dev_uninit() removes tunnel from its list 372 **/ 373 374 static void 375 ip6_tnl_dev_uninit(struct net_device *dev) 376 { 377 struct ip6_tnl *t = netdev_priv(dev); 378 struct net *net = t->net; 379 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 380 381 if (dev == ip6n->fb_tnl_dev) 382 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 383 else 384 ip6_tnl_unlink(ip6n, t); 385 dst_cache_reset(&t->dst_cache); 386 dev_put_track(dev, &t->dev_tracker); 387 } 388 389 /** 390 * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option 391 * @skb: received socket buffer 392 * @raw: the ICMPv6 error message data 393 * 394 * Return: 395 * 0 if none was found, 396 * else index to encapsulation limit 397 **/ 398 399 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 400 { 401 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw; 402 unsigned int nhoff = raw - skb->data; 403 unsigned int off = nhoff + sizeof(*ipv6h); 404 u8 next, nexthdr = ipv6h->nexthdr; 405 406 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 407 struct ipv6_opt_hdr *hdr; 408 u16 optlen; 409 410 if (!pskb_may_pull(skb, off + sizeof(*hdr))) 411 break; 412 413 hdr = (struct ipv6_opt_hdr *)(skb->data + off); 414 if (nexthdr == NEXTHDR_FRAGMENT) { 415 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 416 if (frag_hdr->frag_off) 417 break; 418 optlen = 8; 419 } else if (nexthdr == NEXTHDR_AUTH) { 420 optlen = ipv6_authlen(hdr); 421 } else { 422 optlen = ipv6_optlen(hdr); 423 } 424 /* cache hdr->nexthdr, since pskb_may_pull() might 425 * invalidate hdr 426 */ 427 next = hdr->nexthdr; 428 if (nexthdr == NEXTHDR_DEST) { 429 u16 i = 2; 430 431 /* Remember : hdr is no longer valid at this point. */ 432 if (!pskb_may_pull(skb, off + optlen)) 433 break; 434 435 while (1) { 436 struct ipv6_tlv_tnl_enc_lim *tel; 437 438 /* No more room for encapsulation limit */ 439 if (i + sizeof(*tel) > optlen) 440 break; 441 442 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i); 443 /* return index of option if found and valid */ 444 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 445 tel->length == 1) 446 return i + off - nhoff; 447 /* else jump to next option */ 448 if (tel->type) 449 i += tel->length + 2; 450 else 451 i++; 452 } 453 } 454 nexthdr = next; 455 off += optlen; 456 } 457 return 0; 458 } 459 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 460 461 /* ip6_tnl_err() should handle errors in the tunnel according to the 462 * specifications in RFC 2473. 463 */ 464 static int 465 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 466 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 467 { 468 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data; 469 struct net *net = dev_net(skb->dev); 470 u8 rel_type = ICMPV6_DEST_UNREACH; 471 u8 rel_code = ICMPV6_ADDR_UNREACH; 472 __u32 rel_info = 0; 473 struct ip6_tnl *t; 474 int err = -ENOENT; 475 int rel_msg = 0; 476 u8 tproto; 477 __u16 len; 478 479 /* If the packet doesn't contain the original IPv6 header we are 480 in trouble since we might need the source address for further 481 processing of the error. */ 482 483 rcu_read_lock(); 484 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr); 485 if (!t) 486 goto out; 487 488 tproto = READ_ONCE(t->parms.proto); 489 if (tproto != ipproto && tproto != 0) 490 goto out; 491 492 err = 0; 493 494 switch (*type) { 495 case ICMPV6_DEST_UNREACH: 496 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 497 t->parms.name); 498 rel_msg = 1; 499 break; 500 case ICMPV6_TIME_EXCEED: 501 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 502 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 503 t->parms.name); 504 rel_msg = 1; 505 } 506 break; 507 case ICMPV6_PARAMPROB: { 508 struct ipv6_tlv_tnl_enc_lim *tel; 509 __u32 teli; 510 511 teli = 0; 512 if ((*code) == ICMPV6_HDR_FIELD) 513 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 514 515 if (teli && teli == *info - 2) { 516 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 517 if (tel->encap_limit == 0) { 518 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 519 t->parms.name); 520 rel_msg = 1; 521 } 522 } else { 523 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 524 t->parms.name); 525 } 526 break; 527 } 528 case ICMPV6_PKT_TOOBIG: { 529 __u32 mtu; 530 531 ip6_update_pmtu(skb, net, htonl(*info), 0, 0, 532 sock_net_uid(net, NULL)); 533 mtu = *info - offset; 534 if (mtu < IPV6_MIN_MTU) 535 mtu = IPV6_MIN_MTU; 536 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len); 537 if (len > mtu) { 538 rel_type = ICMPV6_PKT_TOOBIG; 539 rel_code = 0; 540 rel_info = mtu; 541 rel_msg = 1; 542 } 543 break; 544 } 545 case NDISC_REDIRECT: 546 ip6_redirect(skb, net, skb->dev->ifindex, 0, 547 sock_net_uid(net, NULL)); 548 break; 549 } 550 551 *type = rel_type; 552 *code = rel_code; 553 *info = rel_info; 554 *msg = rel_msg; 555 556 out: 557 rcu_read_unlock(); 558 return err; 559 } 560 561 static int 562 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 563 u8 type, u8 code, int offset, __be32 info) 564 { 565 __u32 rel_info = ntohl(info); 566 const struct iphdr *eiph; 567 struct sk_buff *skb2; 568 int err, rel_msg = 0; 569 u8 rel_type = type; 570 u8 rel_code = code; 571 struct rtable *rt; 572 struct flowi4 fl4; 573 574 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 575 &rel_msg, &rel_info, offset); 576 if (err < 0) 577 return err; 578 579 if (rel_msg == 0) 580 return 0; 581 582 switch (rel_type) { 583 case ICMPV6_DEST_UNREACH: 584 if (rel_code != ICMPV6_ADDR_UNREACH) 585 return 0; 586 rel_type = ICMP_DEST_UNREACH; 587 rel_code = ICMP_HOST_UNREACH; 588 break; 589 case ICMPV6_PKT_TOOBIG: 590 if (rel_code != 0) 591 return 0; 592 rel_type = ICMP_DEST_UNREACH; 593 rel_code = ICMP_FRAG_NEEDED; 594 break; 595 default: 596 return 0; 597 } 598 599 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 600 return 0; 601 602 skb2 = skb_clone(skb, GFP_ATOMIC); 603 if (!skb2) 604 return 0; 605 606 skb_dst_drop(skb2); 607 608 skb_pull(skb2, offset); 609 skb_reset_network_header(skb2); 610 eiph = ip_hdr(skb2); 611 612 /* Try to guess incoming interface */ 613 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr, 614 0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 615 if (IS_ERR(rt)) 616 goto out; 617 618 skb2->dev = rt->dst.dev; 619 ip_rt_put(rt); 620 621 /* route "incoming" packet */ 622 if (rt->rt_flags & RTCF_LOCAL) { 623 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 624 eiph->daddr, eiph->saddr, 0, 0, 625 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 626 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) { 627 if (!IS_ERR(rt)) 628 ip_rt_put(rt); 629 goto out; 630 } 631 skb_dst_set(skb2, &rt->dst); 632 } else { 633 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 634 skb2->dev) || 635 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6) 636 goto out; 637 } 638 639 /* change mtu on this route */ 640 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 641 if (rel_info > dst_mtu(skb_dst(skb2))) 642 goto out; 643 644 skb_dst_update_pmtu_no_confirm(skb2, rel_info); 645 } 646 647 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 648 649 out: 650 kfree_skb(skb2); 651 return 0; 652 } 653 654 static int 655 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 656 u8 type, u8 code, int offset, __be32 info) 657 { 658 __u32 rel_info = ntohl(info); 659 int err, rel_msg = 0; 660 u8 rel_type = type; 661 u8 rel_code = code; 662 663 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 664 &rel_msg, &rel_info, offset); 665 if (err < 0) 666 return err; 667 668 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 669 struct rt6_info *rt; 670 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 671 672 if (!skb2) 673 return 0; 674 675 skb_dst_drop(skb2); 676 skb_pull(skb2, offset); 677 skb_reset_network_header(skb2); 678 679 /* Try to guess incoming interface */ 680 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 681 NULL, 0, skb2, 0); 682 683 if (rt && rt->dst.dev) 684 skb2->dev = rt->dst.dev; 685 686 icmpv6_send(skb2, rel_type, rel_code, rel_info); 687 688 ip6_rt_put(rt); 689 690 kfree_skb(skb2); 691 } 692 693 return 0; 694 } 695 696 static int 697 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 698 u8 type, u8 code, int offset, __be32 info) 699 { 700 __u32 rel_info = ntohl(info); 701 int err, rel_msg = 0; 702 u8 rel_type = type; 703 u8 rel_code = code; 704 705 err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code, 706 &rel_msg, &rel_info, offset); 707 return err; 708 } 709 710 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 711 const struct ipv6hdr *ipv6h, 712 struct sk_buff *skb) 713 { 714 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 715 716 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 717 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 718 719 return IP6_ECN_decapsulate(ipv6h, skb); 720 } 721 722 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 723 const struct ipv6hdr *ipv6h, 724 struct sk_buff *skb) 725 { 726 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 727 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 728 729 return IP6_ECN_decapsulate(ipv6h, skb); 730 } 731 732 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 733 const struct ipv6hdr *ipv6h, 734 struct sk_buff *skb) 735 { 736 /* ECN is not supported in AF_MPLS */ 737 return 0; 738 } 739 740 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 741 const struct in6_addr *laddr, 742 const struct in6_addr *raddr) 743 { 744 struct __ip6_tnl_parm *p = &t->parms; 745 int ltype = ipv6_addr_type(laddr); 746 int rtype = ipv6_addr_type(raddr); 747 __u32 flags = 0; 748 749 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 750 flags = IP6_TNL_F_CAP_PER_PACKET; 751 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 752 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 753 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 754 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 755 if (ltype&IPV6_ADDR_UNICAST) 756 flags |= IP6_TNL_F_CAP_XMIT; 757 if (rtype&IPV6_ADDR_UNICAST) 758 flags |= IP6_TNL_F_CAP_RCV; 759 } 760 return flags; 761 } 762 EXPORT_SYMBOL(ip6_tnl_get_cap); 763 764 /* called with rcu_read_lock() */ 765 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 766 const struct in6_addr *laddr, 767 const struct in6_addr *raddr) 768 { 769 struct __ip6_tnl_parm *p = &t->parms; 770 int ret = 0; 771 struct net *net = t->net; 772 773 if ((p->flags & IP6_TNL_F_CAP_RCV) || 774 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 775 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 776 struct net_device *ldev = NULL; 777 778 if (p->link) 779 ldev = dev_get_by_index_rcu(net, p->link); 780 781 if ((ipv6_addr_is_multicast(laddr) || 782 likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false, 783 0, IFA_F_TENTATIVE))) && 784 ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) || 785 likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true, 786 0, IFA_F_TENTATIVE)))) 787 ret = 1; 788 } 789 return ret; 790 } 791 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 792 793 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb, 794 const struct tnl_ptk_info *tpi, 795 struct metadata_dst *tun_dst, 796 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 797 const struct ipv6hdr *ipv6h, 798 struct sk_buff *skb), 799 bool log_ecn_err) 800 { 801 struct pcpu_sw_netstats *tstats; 802 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 803 int err; 804 805 if ((!(tpi->flags & TUNNEL_CSUM) && 806 (tunnel->parms.i_flags & TUNNEL_CSUM)) || 807 ((tpi->flags & TUNNEL_CSUM) && 808 !(tunnel->parms.i_flags & TUNNEL_CSUM))) { 809 tunnel->dev->stats.rx_crc_errors++; 810 tunnel->dev->stats.rx_errors++; 811 goto drop; 812 } 813 814 if (tunnel->parms.i_flags & TUNNEL_SEQ) { 815 if (!(tpi->flags & TUNNEL_SEQ) || 816 (tunnel->i_seqno && 817 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) { 818 tunnel->dev->stats.rx_fifo_errors++; 819 tunnel->dev->stats.rx_errors++; 820 goto drop; 821 } 822 tunnel->i_seqno = ntohl(tpi->seq) + 1; 823 } 824 825 skb->protocol = tpi->proto; 826 827 /* Warning: All skb pointers will be invalidated! */ 828 if (tunnel->dev->type == ARPHRD_ETHER) { 829 if (!pskb_may_pull(skb, ETH_HLEN)) { 830 tunnel->dev->stats.rx_length_errors++; 831 tunnel->dev->stats.rx_errors++; 832 goto drop; 833 } 834 835 ipv6h = ipv6_hdr(skb); 836 skb->protocol = eth_type_trans(skb, tunnel->dev); 837 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 838 } else { 839 skb->dev = tunnel->dev; 840 skb_reset_mac_header(skb); 841 } 842 843 skb_reset_network_header(skb); 844 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 845 846 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 847 848 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb); 849 if (unlikely(err)) { 850 if (log_ecn_err) 851 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n", 852 &ipv6h->saddr, 853 ipv6_get_dsfield(ipv6h)); 854 if (err > 1) { 855 ++tunnel->dev->stats.rx_frame_errors; 856 ++tunnel->dev->stats.rx_errors; 857 goto drop; 858 } 859 } 860 861 tstats = this_cpu_ptr(tunnel->dev->tstats); 862 u64_stats_update_begin(&tstats->syncp); 863 tstats->rx_packets++; 864 tstats->rx_bytes += skb->len; 865 u64_stats_update_end(&tstats->syncp); 866 867 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev))); 868 869 if (tun_dst) 870 skb_dst_set(skb, (struct dst_entry *)tun_dst); 871 872 gro_cells_receive(&tunnel->gro_cells, skb); 873 return 0; 874 875 drop: 876 if (tun_dst) 877 dst_release((struct dst_entry *)tun_dst); 878 kfree_skb(skb); 879 return 0; 880 } 881 882 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb, 883 const struct tnl_ptk_info *tpi, 884 struct metadata_dst *tun_dst, 885 bool log_ecn_err) 886 { 887 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 888 const struct ipv6hdr *ipv6h, 889 struct sk_buff *skb); 890 891 dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate; 892 if (tpi->proto == htons(ETH_P_IP)) 893 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate; 894 895 return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 896 log_ecn_err); 897 } 898 EXPORT_SYMBOL(ip6_tnl_rcv); 899 900 static const struct tnl_ptk_info tpi_v6 = { 901 /* no tunnel info required for ipxip6. */ 902 .proto = htons(ETH_P_IPV6), 903 }; 904 905 static const struct tnl_ptk_info tpi_v4 = { 906 /* no tunnel info required for ipxip6. */ 907 .proto = htons(ETH_P_IP), 908 }; 909 910 static const struct tnl_ptk_info tpi_mpls = { 911 /* no tunnel info required for mplsip6. */ 912 .proto = htons(ETH_P_MPLS_UC), 913 }; 914 915 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto, 916 const struct tnl_ptk_info *tpi, 917 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 918 const struct ipv6hdr *ipv6h, 919 struct sk_buff *skb)) 920 { 921 struct ip6_tnl *t; 922 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 923 struct metadata_dst *tun_dst = NULL; 924 int ret = -1; 925 926 rcu_read_lock(); 927 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr); 928 929 if (t) { 930 u8 tproto = READ_ONCE(t->parms.proto); 931 932 if (tproto != ipproto && tproto != 0) 933 goto drop; 934 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 935 goto drop; 936 ipv6h = ipv6_hdr(skb); 937 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) 938 goto drop; 939 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 940 goto drop; 941 if (t->parms.collect_md) { 942 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0); 943 if (!tun_dst) 944 goto drop; 945 } 946 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 947 log_ecn_error); 948 } 949 950 rcu_read_unlock(); 951 952 return ret; 953 954 drop: 955 rcu_read_unlock(); 956 kfree_skb(skb); 957 return 0; 958 } 959 960 static int ip4ip6_rcv(struct sk_buff *skb) 961 { 962 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4, 963 ip4ip6_dscp_ecn_decapsulate); 964 } 965 966 static int ip6ip6_rcv(struct sk_buff *skb) 967 { 968 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6, 969 ip6ip6_dscp_ecn_decapsulate); 970 } 971 972 static int mplsip6_rcv(struct sk_buff *skb) 973 { 974 return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls, 975 mplsip6_dscp_ecn_decapsulate); 976 } 977 978 struct ipv6_tel_txoption { 979 struct ipv6_txoptions ops; 980 __u8 dst_opt[8]; 981 }; 982 983 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 984 { 985 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 986 987 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 988 opt->dst_opt[3] = 1; 989 opt->dst_opt[4] = encap_limit; 990 opt->dst_opt[5] = IPV6_TLV_PADN; 991 opt->dst_opt[6] = 1; 992 993 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt; 994 opt->ops.opt_nflen = 8; 995 } 996 997 /** 998 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 999 * @t: the outgoing tunnel device 1000 * @hdr: IPv6 header from the incoming packet 1001 * 1002 * Description: 1003 * Avoid trivial tunneling loop by checking that tunnel exit-point 1004 * doesn't match source of incoming packet. 1005 * 1006 * Return: 1007 * 1 if conflict, 1008 * 0 else 1009 **/ 1010 1011 static inline bool 1012 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 1013 { 1014 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 1015 } 1016 1017 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, 1018 const struct in6_addr *laddr, 1019 const struct in6_addr *raddr) 1020 { 1021 struct __ip6_tnl_parm *p = &t->parms; 1022 int ret = 0; 1023 struct net *net = t->net; 1024 1025 if (t->parms.collect_md) 1026 return 1; 1027 1028 if ((p->flags & IP6_TNL_F_CAP_XMIT) || 1029 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 1030 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) { 1031 struct net_device *ldev = NULL; 1032 1033 rcu_read_lock(); 1034 if (p->link) 1035 ldev = dev_get_by_index_rcu(net, p->link); 1036 1037 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false, 1038 0, IFA_F_TENTATIVE))) 1039 pr_warn_ratelimited("%s xmit: Local address not yet configured!\n", 1040 p->name); 1041 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) && 1042 !ipv6_addr_is_multicast(raddr) && 1043 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev, 1044 true, 0, IFA_F_TENTATIVE))) 1045 pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n", 1046 p->name); 1047 else 1048 ret = 1; 1049 rcu_read_unlock(); 1050 } 1051 return ret; 1052 } 1053 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 1054 1055 /** 1056 * ip6_tnl_xmit - encapsulate packet and send 1057 * @skb: the outgoing socket buffer 1058 * @dev: the outgoing tunnel device 1059 * @dsfield: dscp code for outer header 1060 * @fl6: flow of tunneled packet 1061 * @encap_limit: encapsulation limit 1062 * @pmtu: Path MTU is stored if packet is too big 1063 * @proto: next header value 1064 * 1065 * Description: 1066 * Build new header and do some sanity checks on the packet before sending 1067 * it. 1068 * 1069 * Return: 1070 * 0 on success 1071 * -1 fail 1072 * %-EMSGSIZE message too big. return mtu in this case. 1073 **/ 1074 1075 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield, 1076 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, 1077 __u8 proto) 1078 { 1079 struct ip6_tnl *t = netdev_priv(dev); 1080 struct net *net = t->net; 1081 struct net_device_stats *stats = &t->dev->stats; 1082 struct ipv6hdr *ipv6h; 1083 struct ipv6_tel_txoption opt; 1084 struct dst_entry *dst = NULL, *ndst = NULL; 1085 struct net_device *tdev; 1086 int mtu; 1087 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0; 1088 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen; 1089 unsigned int max_headroom = psh_hlen; 1090 bool use_cache = false; 1091 u8 hop_limit; 1092 int err = -1; 1093 1094 if (t->parms.collect_md) { 1095 hop_limit = skb_tunnel_info(skb)->key.ttl; 1096 goto route_lookup; 1097 } else { 1098 hop_limit = t->parms.hop_limit; 1099 } 1100 1101 /* NBMA tunnel */ 1102 if (ipv6_addr_any(&t->parms.raddr)) { 1103 if (skb->protocol == htons(ETH_P_IPV6)) { 1104 struct in6_addr *addr6; 1105 struct neighbour *neigh; 1106 int addr_type; 1107 1108 if (!skb_dst(skb)) 1109 goto tx_err_link_failure; 1110 1111 neigh = dst_neigh_lookup(skb_dst(skb), 1112 &ipv6_hdr(skb)->daddr); 1113 if (!neigh) 1114 goto tx_err_link_failure; 1115 1116 addr6 = (struct in6_addr *)&neigh->primary_key; 1117 addr_type = ipv6_addr_type(addr6); 1118 1119 if (addr_type == IPV6_ADDR_ANY) 1120 addr6 = &ipv6_hdr(skb)->daddr; 1121 1122 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr)); 1123 neigh_release(neigh); 1124 } else if (skb->protocol == htons(ETH_P_IP)) { 1125 const struct rtable *rt = skb_rtable(skb); 1126 1127 if (!rt) 1128 goto tx_err_link_failure; 1129 1130 if (rt->rt_gw_family == AF_INET6) 1131 memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr)); 1132 } 1133 } else if (t->parms.proto != 0 && !(t->parms.flags & 1134 (IP6_TNL_F_USE_ORIG_TCLASS | 1135 IP6_TNL_F_USE_ORIG_FWMARK))) { 1136 /* enable the cache only if neither the outer protocol nor the 1137 * routing decision depends on the current inner header value 1138 */ 1139 use_cache = true; 1140 } 1141 1142 if (use_cache) 1143 dst = dst_cache_get(&t->dst_cache); 1144 1145 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1146 goto tx_err_link_failure; 1147 1148 if (!dst) { 1149 route_lookup: 1150 /* add dsfield to flowlabel for route lookup */ 1151 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1152 1153 dst = ip6_route_output(net, NULL, fl6); 1154 1155 if (dst->error) 1156 goto tx_err_link_failure; 1157 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1158 if (IS_ERR(dst)) { 1159 err = PTR_ERR(dst); 1160 dst = NULL; 1161 goto tx_err_link_failure; 1162 } 1163 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) && 1164 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1165 &fl6->daddr, 0, &fl6->saddr)) 1166 goto tx_err_link_failure; 1167 ndst = dst; 1168 } 1169 1170 tdev = dst->dev; 1171 1172 if (tdev == dev) { 1173 stats->collisions++; 1174 net_warn_ratelimited("%s: Local routing loop detected!\n", 1175 t->parms.name); 1176 goto tx_err_dst_release; 1177 } 1178 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen; 1179 if (encap_limit >= 0) { 1180 max_headroom += 8; 1181 mtu -= 8; 1182 } 1183 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ? 1184 IPV6_MIN_MTU : IPV4_MIN_MTU); 1185 1186 skb_dst_update_pmtu_no_confirm(skb, mtu); 1187 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) { 1188 *pmtu = mtu; 1189 err = -EMSGSIZE; 1190 goto tx_err_dst_release; 1191 } 1192 1193 if (t->err_count > 0) { 1194 if (time_before(jiffies, 1195 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1196 t->err_count--; 1197 1198 dst_link_failure(skb); 1199 } else { 1200 t->err_count = 0; 1201 } 1202 } 1203 1204 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1205 1206 /* 1207 * Okay, now see if we can stuff it in the buffer as-is. 1208 */ 1209 max_headroom += LL_RESERVED_SPACE(tdev); 1210 1211 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1212 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1213 struct sk_buff *new_skb; 1214 1215 new_skb = skb_realloc_headroom(skb, max_headroom); 1216 if (!new_skb) 1217 goto tx_err_dst_release; 1218 1219 if (skb->sk) 1220 skb_set_owner_w(new_skb, skb->sk); 1221 consume_skb(skb); 1222 skb = new_skb; 1223 } 1224 1225 if (t->parms.collect_md) { 1226 if (t->encap.type != TUNNEL_ENCAP_NONE) 1227 goto tx_err_dst_release; 1228 } else { 1229 if (use_cache && ndst) 1230 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1231 } 1232 skb_dst_set(skb, dst); 1233 1234 if (hop_limit == 0) { 1235 if (skb->protocol == htons(ETH_P_IP)) 1236 hop_limit = ip_hdr(skb)->ttl; 1237 else if (skb->protocol == htons(ETH_P_IPV6)) 1238 hop_limit = ipv6_hdr(skb)->hop_limit; 1239 else 1240 hop_limit = ip6_dst_hoplimit(dst); 1241 } 1242 1243 /* Calculate max headroom for all the headers and adjust 1244 * needed_headroom if necessary. 1245 */ 1246 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1247 + dst->header_len + t->hlen; 1248 if (max_headroom > dev->needed_headroom) 1249 dev->needed_headroom = max_headroom; 1250 1251 err = ip6_tnl_encap(skb, t, &proto, fl6); 1252 if (err) 1253 return err; 1254 1255 if (encap_limit >= 0) { 1256 init_tel_txopt(&opt, encap_limit); 1257 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1258 } 1259 1260 skb_push(skb, sizeof(struct ipv6hdr)); 1261 skb_reset_network_header(skb); 1262 ipv6h = ipv6_hdr(skb); 1263 ip6_flow_hdr(ipv6h, dsfield, 1264 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1265 ipv6h->hop_limit = hop_limit; 1266 ipv6h->nexthdr = proto; 1267 ipv6h->saddr = fl6->saddr; 1268 ipv6h->daddr = fl6->daddr; 1269 ip6tunnel_xmit(NULL, skb, dev); 1270 return 0; 1271 tx_err_link_failure: 1272 stats->tx_carrier_errors++; 1273 dst_link_failure(skb); 1274 tx_err_dst_release: 1275 dst_release(dst); 1276 return err; 1277 } 1278 EXPORT_SYMBOL(ip6_tnl_xmit); 1279 1280 static inline int 1281 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, 1282 u8 protocol) 1283 { 1284 struct ip6_tnl *t = netdev_priv(dev); 1285 struct ipv6hdr *ipv6h; 1286 const struct iphdr *iph; 1287 int encap_limit = -1; 1288 __u16 offset; 1289 struct flowi6 fl6; 1290 __u8 dsfield, orig_dsfield; 1291 __u32 mtu; 1292 u8 tproto; 1293 int err; 1294 1295 tproto = READ_ONCE(t->parms.proto); 1296 if (tproto != protocol && tproto != 0) 1297 return -1; 1298 1299 if (t->parms.collect_md) { 1300 struct ip_tunnel_info *tun_info; 1301 const struct ip_tunnel_key *key; 1302 1303 tun_info = skb_tunnel_info(skb); 1304 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1305 ip_tunnel_info_af(tun_info) != AF_INET6)) 1306 return -1; 1307 key = &tun_info->key; 1308 memset(&fl6, 0, sizeof(fl6)); 1309 fl6.flowi6_proto = protocol; 1310 fl6.saddr = key->u.ipv6.src; 1311 fl6.daddr = key->u.ipv6.dst; 1312 fl6.flowlabel = key->label; 1313 dsfield = key->tos; 1314 switch (protocol) { 1315 case IPPROTO_IPIP: 1316 iph = ip_hdr(skb); 1317 orig_dsfield = ipv4_get_dsfield(iph); 1318 break; 1319 case IPPROTO_IPV6: 1320 ipv6h = ipv6_hdr(skb); 1321 orig_dsfield = ipv6_get_dsfield(ipv6h); 1322 break; 1323 default: 1324 orig_dsfield = dsfield; 1325 break; 1326 } 1327 } else { 1328 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1329 encap_limit = t->parms.encap_limit; 1330 if (protocol == IPPROTO_IPV6) { 1331 offset = ip6_tnl_parse_tlv_enc_lim(skb, 1332 skb_network_header(skb)); 1333 /* ip6_tnl_parse_tlv_enc_lim() might have 1334 * reallocated skb->head 1335 */ 1336 if (offset > 0) { 1337 struct ipv6_tlv_tnl_enc_lim *tel; 1338 1339 tel = (void *)&skb_network_header(skb)[offset]; 1340 if (tel->encap_limit == 0) { 1341 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 1342 ICMPV6_HDR_FIELD, offset + 2); 1343 return -1; 1344 } 1345 encap_limit = tel->encap_limit - 1; 1346 } 1347 } 1348 1349 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1350 fl6.flowi6_proto = protocol; 1351 1352 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1353 fl6.flowi6_mark = skb->mark; 1354 else 1355 fl6.flowi6_mark = t->parms.fwmark; 1356 switch (protocol) { 1357 case IPPROTO_IPIP: 1358 iph = ip_hdr(skb); 1359 orig_dsfield = ipv4_get_dsfield(iph); 1360 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1361 dsfield = orig_dsfield; 1362 else 1363 dsfield = ip6_tclass(t->parms.flowinfo); 1364 break; 1365 case IPPROTO_IPV6: 1366 ipv6h = ipv6_hdr(skb); 1367 orig_dsfield = ipv6_get_dsfield(ipv6h); 1368 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1369 dsfield = orig_dsfield; 1370 else 1371 dsfield = ip6_tclass(t->parms.flowinfo); 1372 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1373 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1374 break; 1375 default: 1376 orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo); 1377 break; 1378 } 1379 } 1380 1381 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1382 dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield); 1383 1384 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1385 return -1; 1386 1387 skb_set_inner_ipproto(skb, protocol); 1388 1389 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1390 protocol); 1391 if (err != 0) { 1392 /* XXX: send ICMP error even if DF is not set. */ 1393 if (err == -EMSGSIZE) 1394 switch (protocol) { 1395 case IPPROTO_IPIP: 1396 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1397 ICMP_FRAG_NEEDED, htonl(mtu)); 1398 break; 1399 case IPPROTO_IPV6: 1400 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1401 break; 1402 default: 1403 break; 1404 } 1405 return -1; 1406 } 1407 1408 return 0; 1409 } 1410 1411 static netdev_tx_t 1412 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1413 { 1414 struct ip6_tnl *t = netdev_priv(dev); 1415 struct net_device_stats *stats = &t->dev->stats; 1416 u8 ipproto; 1417 int ret; 1418 1419 if (!pskb_inet_may_pull(skb)) 1420 goto tx_err; 1421 1422 switch (skb->protocol) { 1423 case htons(ETH_P_IP): 1424 ipproto = IPPROTO_IPIP; 1425 break; 1426 case htons(ETH_P_IPV6): 1427 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb))) 1428 goto tx_err; 1429 ipproto = IPPROTO_IPV6; 1430 break; 1431 case htons(ETH_P_MPLS_UC): 1432 ipproto = IPPROTO_MPLS; 1433 break; 1434 default: 1435 goto tx_err; 1436 } 1437 1438 ret = ipxip6_tnl_xmit(skb, dev, ipproto); 1439 if (ret < 0) 1440 goto tx_err; 1441 1442 return NETDEV_TX_OK; 1443 1444 tx_err: 1445 stats->tx_errors++; 1446 stats->tx_dropped++; 1447 kfree_skb(skb); 1448 return NETDEV_TX_OK; 1449 } 1450 1451 static void ip6_tnl_link_config(struct ip6_tnl *t) 1452 { 1453 struct net_device *dev = t->dev; 1454 struct net_device *tdev = NULL; 1455 struct __ip6_tnl_parm *p = &t->parms; 1456 struct flowi6 *fl6 = &t->fl.u.ip6; 1457 unsigned int mtu; 1458 int t_hlen; 1459 1460 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1461 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1462 1463 /* Set up flowi template */ 1464 fl6->saddr = p->laddr; 1465 fl6->daddr = p->raddr; 1466 fl6->flowi6_oif = p->link; 1467 fl6->flowlabel = 0; 1468 1469 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1470 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1471 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1472 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1473 1474 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1475 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1476 1477 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1478 dev->flags |= IFF_POINTOPOINT; 1479 else 1480 dev->flags &= ~IFF_POINTOPOINT; 1481 1482 t->tun_hlen = 0; 1483 t->hlen = t->encap_hlen + t->tun_hlen; 1484 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1485 1486 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1487 int strict = (ipv6_addr_type(&p->raddr) & 1488 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1489 1490 struct rt6_info *rt = rt6_lookup(t->net, 1491 &p->raddr, &p->laddr, 1492 p->link, NULL, strict); 1493 if (rt) { 1494 tdev = rt->dst.dev; 1495 ip6_rt_put(rt); 1496 } 1497 1498 if (!tdev && p->link) 1499 tdev = __dev_get_by_index(t->net, p->link); 1500 1501 if (tdev) { 1502 dev->hard_header_len = tdev->hard_header_len + t_hlen; 1503 mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU); 1504 1505 dev->mtu = mtu - t_hlen; 1506 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1507 dev->mtu -= 8; 1508 1509 if (dev->mtu < IPV6_MIN_MTU) 1510 dev->mtu = IPV6_MIN_MTU; 1511 } 1512 } 1513 } 1514 1515 /** 1516 * ip6_tnl_change - update the tunnel parameters 1517 * @t: tunnel to be changed 1518 * @p: tunnel configuration parameters 1519 * 1520 * Description: 1521 * ip6_tnl_change() updates the tunnel parameters 1522 **/ 1523 1524 static int 1525 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1526 { 1527 t->parms.laddr = p->laddr; 1528 t->parms.raddr = p->raddr; 1529 t->parms.flags = p->flags; 1530 t->parms.hop_limit = p->hop_limit; 1531 t->parms.encap_limit = p->encap_limit; 1532 t->parms.flowinfo = p->flowinfo; 1533 t->parms.link = p->link; 1534 t->parms.proto = p->proto; 1535 t->parms.fwmark = p->fwmark; 1536 dst_cache_reset(&t->dst_cache); 1537 ip6_tnl_link_config(t); 1538 return 0; 1539 } 1540 1541 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1542 { 1543 struct net *net = t->net; 1544 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1545 int err; 1546 1547 ip6_tnl_unlink(ip6n, t); 1548 synchronize_net(); 1549 err = ip6_tnl_change(t, p); 1550 ip6_tnl_link(ip6n, t); 1551 netdev_state_change(t->dev); 1552 return err; 1553 } 1554 1555 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1556 { 1557 /* for default tnl0 device allow to change only the proto */ 1558 t->parms.proto = p->proto; 1559 netdev_state_change(t->dev); 1560 return 0; 1561 } 1562 1563 static void 1564 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1565 { 1566 p->laddr = u->laddr; 1567 p->raddr = u->raddr; 1568 p->flags = u->flags; 1569 p->hop_limit = u->hop_limit; 1570 p->encap_limit = u->encap_limit; 1571 p->flowinfo = u->flowinfo; 1572 p->link = u->link; 1573 p->proto = u->proto; 1574 memcpy(p->name, u->name, sizeof(u->name)); 1575 } 1576 1577 static void 1578 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1579 { 1580 u->laddr = p->laddr; 1581 u->raddr = p->raddr; 1582 u->flags = p->flags; 1583 u->hop_limit = p->hop_limit; 1584 u->encap_limit = p->encap_limit; 1585 u->flowinfo = p->flowinfo; 1586 u->link = p->link; 1587 u->proto = p->proto; 1588 memcpy(u->name, p->name, sizeof(u->name)); 1589 } 1590 1591 /** 1592 * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace 1593 * @dev: virtual device associated with tunnel 1594 * @ifr: unused 1595 * @data: parameters passed from userspace 1596 * @cmd: command to be performed 1597 * 1598 * Description: 1599 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1600 * from userspace. 1601 * 1602 * The possible commands are the following: 1603 * %SIOCGETTUNNEL: get tunnel parameters for device 1604 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1605 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1606 * %SIOCDELTUNNEL: delete tunnel 1607 * 1608 * The fallback device "ip6tnl0", created during module 1609 * initialization, can be used for creating other tunnel devices. 1610 * 1611 * Return: 1612 * 0 on success, 1613 * %-EFAULT if unable to copy data to or from userspace, 1614 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1615 * %-EINVAL if passed tunnel parameters are invalid, 1616 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1617 * %-ENODEV if attempting to change or delete a nonexisting device 1618 **/ 1619 1620 static int 1621 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr, 1622 void __user *data, int cmd) 1623 { 1624 int err = 0; 1625 struct ip6_tnl_parm p; 1626 struct __ip6_tnl_parm p1; 1627 struct ip6_tnl *t = netdev_priv(dev); 1628 struct net *net = t->net; 1629 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1630 1631 memset(&p1, 0, sizeof(p1)); 1632 1633 switch (cmd) { 1634 case SIOCGETTUNNEL: 1635 if (dev == ip6n->fb_tnl_dev) { 1636 if (copy_from_user(&p, data, sizeof(p))) { 1637 err = -EFAULT; 1638 break; 1639 } 1640 ip6_tnl_parm_from_user(&p1, &p); 1641 t = ip6_tnl_locate(net, &p1, 0); 1642 if (IS_ERR(t)) 1643 t = netdev_priv(dev); 1644 } else { 1645 memset(&p, 0, sizeof(p)); 1646 } 1647 ip6_tnl_parm_to_user(&p, &t->parms); 1648 if (copy_to_user(data, &p, sizeof(p))) 1649 err = -EFAULT; 1650 break; 1651 case SIOCADDTUNNEL: 1652 case SIOCCHGTUNNEL: 1653 err = -EPERM; 1654 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1655 break; 1656 err = -EFAULT; 1657 if (copy_from_user(&p, data, sizeof(p))) 1658 break; 1659 err = -EINVAL; 1660 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1661 p.proto != 0) 1662 break; 1663 ip6_tnl_parm_from_user(&p1, &p); 1664 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1665 if (cmd == SIOCCHGTUNNEL) { 1666 if (!IS_ERR(t)) { 1667 if (t->dev != dev) { 1668 err = -EEXIST; 1669 break; 1670 } 1671 } else 1672 t = netdev_priv(dev); 1673 if (dev == ip6n->fb_tnl_dev) 1674 err = ip6_tnl0_update(t, &p1); 1675 else 1676 err = ip6_tnl_update(t, &p1); 1677 } 1678 if (!IS_ERR(t)) { 1679 err = 0; 1680 ip6_tnl_parm_to_user(&p, &t->parms); 1681 if (copy_to_user(data, &p, sizeof(p))) 1682 err = -EFAULT; 1683 1684 } else { 1685 err = PTR_ERR(t); 1686 } 1687 break; 1688 case SIOCDELTUNNEL: 1689 err = -EPERM; 1690 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1691 break; 1692 1693 if (dev == ip6n->fb_tnl_dev) { 1694 err = -EFAULT; 1695 if (copy_from_user(&p, data, sizeof(p))) 1696 break; 1697 err = -ENOENT; 1698 ip6_tnl_parm_from_user(&p1, &p); 1699 t = ip6_tnl_locate(net, &p1, 0); 1700 if (IS_ERR(t)) 1701 break; 1702 err = -EPERM; 1703 if (t->dev == ip6n->fb_tnl_dev) 1704 break; 1705 dev = t->dev; 1706 } 1707 err = 0; 1708 unregister_netdevice(dev); 1709 break; 1710 default: 1711 err = -EINVAL; 1712 } 1713 return err; 1714 } 1715 1716 /** 1717 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1718 * @dev: virtual device associated with tunnel 1719 * @new_mtu: the new mtu 1720 * 1721 * Return: 1722 * 0 on success, 1723 * %-EINVAL if mtu too small 1724 **/ 1725 1726 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1727 { 1728 struct ip6_tnl *tnl = netdev_priv(dev); 1729 1730 if (tnl->parms.proto == IPPROTO_IPV6) { 1731 if (new_mtu < IPV6_MIN_MTU) 1732 return -EINVAL; 1733 } else { 1734 if (new_mtu < ETH_MIN_MTU) 1735 return -EINVAL; 1736 } 1737 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) { 1738 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len) 1739 return -EINVAL; 1740 } else { 1741 if (new_mtu > IP_MAX_MTU - dev->hard_header_len) 1742 return -EINVAL; 1743 } 1744 dev->mtu = new_mtu; 1745 return 0; 1746 } 1747 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1748 1749 int ip6_tnl_get_iflink(const struct net_device *dev) 1750 { 1751 struct ip6_tnl *t = netdev_priv(dev); 1752 1753 return t->parms.link; 1754 } 1755 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1756 1757 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1758 unsigned int num) 1759 { 1760 if (num >= MAX_IPTUN_ENCAP_OPS) 1761 return -ERANGE; 1762 1763 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1764 &ip6tun_encaps[num], 1765 NULL, ops) ? 0 : -1; 1766 } 1767 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1768 1769 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1770 unsigned int num) 1771 { 1772 int ret; 1773 1774 if (num >= MAX_IPTUN_ENCAP_OPS) 1775 return -ERANGE; 1776 1777 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1778 &ip6tun_encaps[num], 1779 ops, NULL) == ops) ? 0 : -1; 1780 1781 synchronize_net(); 1782 1783 return ret; 1784 } 1785 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1786 1787 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1788 struct ip_tunnel_encap *ipencap) 1789 { 1790 int hlen; 1791 1792 memset(&t->encap, 0, sizeof(t->encap)); 1793 1794 hlen = ip6_encap_hlen(ipencap); 1795 if (hlen < 0) 1796 return hlen; 1797 1798 t->encap.type = ipencap->type; 1799 t->encap.sport = ipencap->sport; 1800 t->encap.dport = ipencap->dport; 1801 t->encap.flags = ipencap->flags; 1802 1803 t->encap_hlen = hlen; 1804 t->hlen = t->encap_hlen + t->tun_hlen; 1805 1806 return 0; 1807 } 1808 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1809 1810 static const struct net_device_ops ip6_tnl_netdev_ops = { 1811 .ndo_init = ip6_tnl_dev_init, 1812 .ndo_uninit = ip6_tnl_dev_uninit, 1813 .ndo_start_xmit = ip6_tnl_start_xmit, 1814 .ndo_siocdevprivate = ip6_tnl_siocdevprivate, 1815 .ndo_change_mtu = ip6_tnl_change_mtu, 1816 .ndo_get_stats64 = dev_get_tstats64, 1817 .ndo_get_iflink = ip6_tnl_get_iflink, 1818 }; 1819 1820 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1821 NETIF_F_FRAGLIST | \ 1822 NETIF_F_HIGHDMA | \ 1823 NETIF_F_GSO_SOFTWARE | \ 1824 NETIF_F_HW_CSUM) 1825 1826 /** 1827 * ip6_tnl_dev_setup - setup virtual tunnel device 1828 * @dev: virtual device associated with tunnel 1829 * 1830 * Description: 1831 * Initialize function pointers and device parameters 1832 **/ 1833 1834 static void ip6_tnl_dev_setup(struct net_device *dev) 1835 { 1836 dev->netdev_ops = &ip6_tnl_netdev_ops; 1837 dev->header_ops = &ip_tunnel_header_ops; 1838 dev->needs_free_netdev = true; 1839 dev->priv_destructor = ip6_dev_free; 1840 1841 dev->type = ARPHRD_TUNNEL6; 1842 dev->flags |= IFF_NOARP; 1843 dev->addr_len = sizeof(struct in6_addr); 1844 dev->features |= NETIF_F_LLTX; 1845 netif_keep_dst(dev); 1846 1847 dev->features |= IPXIPX_FEATURES; 1848 dev->hw_features |= IPXIPX_FEATURES; 1849 1850 /* This perm addr will be used as interface identifier by IPv6 */ 1851 dev->addr_assign_type = NET_ADDR_RANDOM; 1852 eth_random_addr(dev->perm_addr); 1853 } 1854 1855 1856 /** 1857 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1858 * @dev: virtual device associated with tunnel 1859 **/ 1860 1861 static inline int 1862 ip6_tnl_dev_init_gen(struct net_device *dev) 1863 { 1864 struct ip6_tnl *t = netdev_priv(dev); 1865 int ret; 1866 int t_hlen; 1867 1868 t->dev = dev; 1869 t->net = dev_net(dev); 1870 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1871 if (!dev->tstats) 1872 return -ENOMEM; 1873 1874 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1875 if (ret) 1876 goto free_stats; 1877 1878 ret = gro_cells_init(&t->gro_cells, dev); 1879 if (ret) 1880 goto destroy_dst; 1881 1882 t->tun_hlen = 0; 1883 t->hlen = t->encap_hlen + t->tun_hlen; 1884 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1885 1886 dev->type = ARPHRD_TUNNEL6; 1887 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1888 dev->mtu = ETH_DATA_LEN - t_hlen; 1889 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1890 dev->mtu -= 8; 1891 dev->min_mtu = ETH_MIN_MTU; 1892 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len; 1893 1894 dev_hold_track(dev, &t->dev_tracker, GFP_KERNEL); 1895 return 0; 1896 1897 destroy_dst: 1898 dst_cache_destroy(&t->dst_cache); 1899 free_stats: 1900 free_percpu(dev->tstats); 1901 dev->tstats = NULL; 1902 1903 return ret; 1904 } 1905 1906 /** 1907 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1908 * @dev: virtual device associated with tunnel 1909 **/ 1910 1911 static int ip6_tnl_dev_init(struct net_device *dev) 1912 { 1913 struct ip6_tnl *t = netdev_priv(dev); 1914 int err = ip6_tnl_dev_init_gen(dev); 1915 1916 if (err) 1917 return err; 1918 ip6_tnl_link_config(t); 1919 if (t->parms.collect_md) 1920 netif_keep_dst(dev); 1921 return 0; 1922 } 1923 1924 /** 1925 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1926 * @dev: fallback device 1927 * 1928 * Return: 0 1929 **/ 1930 1931 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1932 { 1933 struct ip6_tnl *t = netdev_priv(dev); 1934 struct net *net = dev_net(dev); 1935 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1936 1937 t->parms.proto = IPPROTO_IPV6; 1938 1939 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1940 return 0; 1941 } 1942 1943 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[], 1944 struct netlink_ext_ack *extack) 1945 { 1946 u8 proto; 1947 1948 if (!data || !data[IFLA_IPTUN_PROTO]) 1949 return 0; 1950 1951 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1952 if (proto != IPPROTO_IPV6 && 1953 proto != IPPROTO_IPIP && 1954 proto != 0) 1955 return -EINVAL; 1956 1957 return 0; 1958 } 1959 1960 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1961 struct __ip6_tnl_parm *parms) 1962 { 1963 memset(parms, 0, sizeof(*parms)); 1964 1965 if (!data) 1966 return; 1967 1968 if (data[IFLA_IPTUN_LINK]) 1969 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1970 1971 if (data[IFLA_IPTUN_LOCAL]) 1972 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1973 1974 if (data[IFLA_IPTUN_REMOTE]) 1975 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1976 1977 if (data[IFLA_IPTUN_TTL]) 1978 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1979 1980 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1981 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1982 1983 if (data[IFLA_IPTUN_FLOWINFO]) 1984 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1985 1986 if (data[IFLA_IPTUN_FLAGS]) 1987 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1988 1989 if (data[IFLA_IPTUN_PROTO]) 1990 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1991 1992 if (data[IFLA_IPTUN_COLLECT_METADATA]) 1993 parms->collect_md = true; 1994 1995 if (data[IFLA_IPTUN_FWMARK]) 1996 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 1997 } 1998 1999 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[], 2000 struct ip_tunnel_encap *ipencap) 2001 { 2002 bool ret = false; 2003 2004 memset(ipencap, 0, sizeof(*ipencap)); 2005 2006 if (!data) 2007 return ret; 2008 2009 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 2010 ret = true; 2011 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 2012 } 2013 2014 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 2015 ret = true; 2016 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 2017 } 2018 2019 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 2020 ret = true; 2021 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 2022 } 2023 2024 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 2025 ret = true; 2026 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 2027 } 2028 2029 return ret; 2030 } 2031 2032 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 2033 struct nlattr *tb[], struct nlattr *data[], 2034 struct netlink_ext_ack *extack) 2035 { 2036 struct net *net = dev_net(dev); 2037 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2038 struct ip_tunnel_encap ipencap; 2039 struct ip6_tnl *nt, *t; 2040 int err; 2041 2042 nt = netdev_priv(dev); 2043 2044 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2045 err = ip6_tnl_encap_setup(nt, &ipencap); 2046 if (err < 0) 2047 return err; 2048 } 2049 2050 ip6_tnl_netlink_parms(data, &nt->parms); 2051 2052 if (nt->parms.collect_md) { 2053 if (rtnl_dereference(ip6n->collect_md_tun)) 2054 return -EEXIST; 2055 } else { 2056 t = ip6_tnl_locate(net, &nt->parms, 0); 2057 if (!IS_ERR(t)) 2058 return -EEXIST; 2059 } 2060 2061 err = ip6_tnl_create2(dev); 2062 if (!err && tb[IFLA_MTU]) 2063 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2064 2065 return err; 2066 } 2067 2068 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2069 struct nlattr *data[], 2070 struct netlink_ext_ack *extack) 2071 { 2072 struct ip6_tnl *t = netdev_priv(dev); 2073 struct __ip6_tnl_parm p; 2074 struct net *net = t->net; 2075 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2076 struct ip_tunnel_encap ipencap; 2077 2078 if (dev == ip6n->fb_tnl_dev) 2079 return -EINVAL; 2080 2081 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2082 int err = ip6_tnl_encap_setup(t, &ipencap); 2083 2084 if (err < 0) 2085 return err; 2086 } 2087 ip6_tnl_netlink_parms(data, &p); 2088 if (p.collect_md) 2089 return -EINVAL; 2090 2091 t = ip6_tnl_locate(net, &p, 0); 2092 if (!IS_ERR(t)) { 2093 if (t->dev != dev) 2094 return -EEXIST; 2095 } else 2096 t = netdev_priv(dev); 2097 2098 return ip6_tnl_update(t, &p); 2099 } 2100 2101 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2102 { 2103 struct net *net = dev_net(dev); 2104 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2105 2106 if (dev != ip6n->fb_tnl_dev) 2107 unregister_netdevice_queue(dev, head); 2108 } 2109 2110 static size_t ip6_tnl_get_size(const struct net_device *dev) 2111 { 2112 return 2113 /* IFLA_IPTUN_LINK */ 2114 nla_total_size(4) + 2115 /* IFLA_IPTUN_LOCAL */ 2116 nla_total_size(sizeof(struct in6_addr)) + 2117 /* IFLA_IPTUN_REMOTE */ 2118 nla_total_size(sizeof(struct in6_addr)) + 2119 /* IFLA_IPTUN_TTL */ 2120 nla_total_size(1) + 2121 /* IFLA_IPTUN_ENCAP_LIMIT */ 2122 nla_total_size(1) + 2123 /* IFLA_IPTUN_FLOWINFO */ 2124 nla_total_size(4) + 2125 /* IFLA_IPTUN_FLAGS */ 2126 nla_total_size(4) + 2127 /* IFLA_IPTUN_PROTO */ 2128 nla_total_size(1) + 2129 /* IFLA_IPTUN_ENCAP_TYPE */ 2130 nla_total_size(2) + 2131 /* IFLA_IPTUN_ENCAP_FLAGS */ 2132 nla_total_size(2) + 2133 /* IFLA_IPTUN_ENCAP_SPORT */ 2134 nla_total_size(2) + 2135 /* IFLA_IPTUN_ENCAP_DPORT */ 2136 nla_total_size(2) + 2137 /* IFLA_IPTUN_COLLECT_METADATA */ 2138 nla_total_size(0) + 2139 /* IFLA_IPTUN_FWMARK */ 2140 nla_total_size(4) + 2141 0; 2142 } 2143 2144 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2145 { 2146 struct ip6_tnl *tunnel = netdev_priv(dev); 2147 struct __ip6_tnl_parm *parm = &tunnel->parms; 2148 2149 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2150 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2151 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2152 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2153 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2154 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2155 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2156 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2157 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2158 goto nla_put_failure; 2159 2160 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2161 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2162 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2163 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2164 goto nla_put_failure; 2165 2166 if (parm->collect_md) 2167 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2168 goto nla_put_failure; 2169 2170 return 0; 2171 2172 nla_put_failure: 2173 return -EMSGSIZE; 2174 } 2175 2176 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2177 { 2178 struct ip6_tnl *tunnel = netdev_priv(dev); 2179 2180 return tunnel->net; 2181 } 2182 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2183 2184 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2185 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2186 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2187 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2188 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2189 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2190 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2191 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2192 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2193 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2194 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2195 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2196 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2197 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2198 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2199 }; 2200 2201 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2202 .kind = "ip6tnl", 2203 .maxtype = IFLA_IPTUN_MAX, 2204 .policy = ip6_tnl_policy, 2205 .priv_size = sizeof(struct ip6_tnl), 2206 .setup = ip6_tnl_dev_setup, 2207 .validate = ip6_tnl_validate, 2208 .newlink = ip6_tnl_newlink, 2209 .changelink = ip6_tnl_changelink, 2210 .dellink = ip6_tnl_dellink, 2211 .get_size = ip6_tnl_get_size, 2212 .fill_info = ip6_tnl_fill_info, 2213 .get_link_net = ip6_tnl_get_link_net, 2214 }; 2215 2216 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2217 .handler = ip4ip6_rcv, 2218 .err_handler = ip4ip6_err, 2219 .priority = 1, 2220 }; 2221 2222 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2223 .handler = ip6ip6_rcv, 2224 .err_handler = ip6ip6_err, 2225 .priority = 1, 2226 }; 2227 2228 static struct xfrm6_tunnel mplsip6_handler __read_mostly = { 2229 .handler = mplsip6_rcv, 2230 .err_handler = mplsip6_err, 2231 .priority = 1, 2232 }; 2233 2234 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list) 2235 { 2236 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2237 struct net_device *dev, *aux; 2238 int h; 2239 struct ip6_tnl *t; 2240 2241 for_each_netdev_safe(net, dev, aux) 2242 if (dev->rtnl_link_ops == &ip6_link_ops) 2243 unregister_netdevice_queue(dev, list); 2244 2245 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2246 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2247 while (t) { 2248 /* If dev is in the same netns, it has already 2249 * been added to the list by the previous loop. 2250 */ 2251 if (!net_eq(dev_net(t->dev), net)) 2252 unregister_netdevice_queue(t->dev, list); 2253 t = rtnl_dereference(t->next); 2254 } 2255 } 2256 2257 t = rtnl_dereference(ip6n->tnls_wc[0]); 2258 while (t) { 2259 /* If dev is in the same netns, it has already 2260 * been added to the list by the previous loop. 2261 */ 2262 if (!net_eq(dev_net(t->dev), net)) 2263 unregister_netdevice_queue(t->dev, list); 2264 t = rtnl_dereference(t->next); 2265 } 2266 } 2267 2268 static int __net_init ip6_tnl_init_net(struct net *net) 2269 { 2270 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2271 struct ip6_tnl *t = NULL; 2272 int err; 2273 2274 ip6n->tnls[0] = ip6n->tnls_wc; 2275 ip6n->tnls[1] = ip6n->tnls_r_l; 2276 2277 if (!net_has_fallback_tunnels(net)) 2278 return 0; 2279 err = -ENOMEM; 2280 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2281 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2282 2283 if (!ip6n->fb_tnl_dev) 2284 goto err_alloc_dev; 2285 dev_net_set(ip6n->fb_tnl_dev, net); 2286 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2287 /* FB netdevice is special: we have one, and only one per netns. 2288 * Allowing to move it to another netns is clearly unsafe. 2289 */ 2290 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 2291 2292 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2293 if (err < 0) 2294 goto err_register; 2295 2296 err = register_netdev(ip6n->fb_tnl_dev); 2297 if (err < 0) 2298 goto err_register; 2299 2300 t = netdev_priv(ip6n->fb_tnl_dev); 2301 2302 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2303 return 0; 2304 2305 err_register: 2306 free_netdev(ip6n->fb_tnl_dev); 2307 err_alloc_dev: 2308 return err; 2309 } 2310 2311 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list) 2312 { 2313 struct net *net; 2314 LIST_HEAD(list); 2315 2316 rtnl_lock(); 2317 list_for_each_entry(net, net_list, exit_list) 2318 ip6_tnl_destroy_tunnels(net, &list); 2319 unregister_netdevice_many(&list); 2320 rtnl_unlock(); 2321 } 2322 2323 static struct pernet_operations ip6_tnl_net_ops = { 2324 .init = ip6_tnl_init_net, 2325 .exit_batch = ip6_tnl_exit_batch_net, 2326 .id = &ip6_tnl_net_id, 2327 .size = sizeof(struct ip6_tnl_net), 2328 }; 2329 2330 /** 2331 * ip6_tunnel_init - register protocol and reserve needed resources 2332 * 2333 * Return: 0 on success 2334 **/ 2335 2336 static int __init ip6_tunnel_init(void) 2337 { 2338 int err; 2339 2340 if (!ipv6_mod_enabled()) 2341 return -EOPNOTSUPP; 2342 2343 err = register_pernet_device(&ip6_tnl_net_ops); 2344 if (err < 0) 2345 goto out_pernet; 2346 2347 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2348 if (err < 0) { 2349 pr_err("%s: can't register ip4ip6\n", __func__); 2350 goto out_ip4ip6; 2351 } 2352 2353 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2354 if (err < 0) { 2355 pr_err("%s: can't register ip6ip6\n", __func__); 2356 goto out_ip6ip6; 2357 } 2358 2359 if (ip6_tnl_mpls_supported()) { 2360 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS); 2361 if (err < 0) { 2362 pr_err("%s: can't register mplsip6\n", __func__); 2363 goto out_mplsip6; 2364 } 2365 } 2366 2367 err = rtnl_link_register(&ip6_link_ops); 2368 if (err < 0) 2369 goto rtnl_link_failed; 2370 2371 return 0; 2372 2373 rtnl_link_failed: 2374 if (ip6_tnl_mpls_supported()) 2375 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS); 2376 out_mplsip6: 2377 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2378 out_ip6ip6: 2379 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2380 out_ip4ip6: 2381 unregister_pernet_device(&ip6_tnl_net_ops); 2382 out_pernet: 2383 return err; 2384 } 2385 2386 /** 2387 * ip6_tunnel_cleanup - free resources and unregister protocol 2388 **/ 2389 2390 static void __exit ip6_tunnel_cleanup(void) 2391 { 2392 rtnl_link_unregister(&ip6_link_ops); 2393 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2394 pr_info("%s: can't deregister ip4ip6\n", __func__); 2395 2396 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2397 pr_info("%s: can't deregister ip6ip6\n", __func__); 2398 2399 if (ip6_tnl_mpls_supported() && 2400 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS)) 2401 pr_info("%s: can't deregister mplsip6\n", __func__); 2402 unregister_pernet_device(&ip6_tnl_net_ops); 2403 } 2404 2405 module_init(ip6_tunnel_init); 2406 module_exit(ip6_tunnel_cleanup); 2407