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("%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("%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 } 1125 } else if (t->parms.proto != 0 && !(t->parms.flags & 1126 (IP6_TNL_F_USE_ORIG_TCLASS | 1127 IP6_TNL_F_USE_ORIG_FWMARK))) { 1128 /* enable the cache only if neither the outer protocol nor the 1129 * routing decision depends on the current inner header value 1130 */ 1131 use_cache = true; 1132 } 1133 1134 if (use_cache) 1135 dst = dst_cache_get(&t->dst_cache); 1136 1137 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1138 goto tx_err_link_failure; 1139 1140 if (!dst) { 1141 route_lookup: 1142 /* add dsfield to flowlabel for route lookup */ 1143 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1144 1145 dst = ip6_route_output(net, NULL, fl6); 1146 1147 if (dst->error) 1148 goto tx_err_link_failure; 1149 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1150 if (IS_ERR(dst)) { 1151 err = PTR_ERR(dst); 1152 dst = NULL; 1153 goto tx_err_link_failure; 1154 } 1155 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) && 1156 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1157 &fl6->daddr, 0, &fl6->saddr)) 1158 goto tx_err_link_failure; 1159 ndst = dst; 1160 } 1161 1162 tdev = dst->dev; 1163 1164 if (tdev == dev) { 1165 stats->collisions++; 1166 net_warn_ratelimited("%s: Local routing loop detected!\n", 1167 t->parms.name); 1168 goto tx_err_dst_release; 1169 } 1170 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen; 1171 if (encap_limit >= 0) { 1172 max_headroom += 8; 1173 mtu -= 8; 1174 } 1175 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ? 1176 IPV6_MIN_MTU : IPV4_MIN_MTU); 1177 1178 skb_dst_update_pmtu_no_confirm(skb, mtu); 1179 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) { 1180 *pmtu = mtu; 1181 err = -EMSGSIZE; 1182 goto tx_err_dst_release; 1183 } 1184 1185 if (t->err_count > 0) { 1186 if (time_before(jiffies, 1187 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1188 t->err_count--; 1189 1190 dst_link_failure(skb); 1191 } else { 1192 t->err_count = 0; 1193 } 1194 } 1195 1196 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1197 1198 /* 1199 * Okay, now see if we can stuff it in the buffer as-is. 1200 */ 1201 max_headroom += LL_RESERVED_SPACE(tdev); 1202 1203 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1204 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1205 struct sk_buff *new_skb; 1206 1207 new_skb = skb_realloc_headroom(skb, max_headroom); 1208 if (!new_skb) 1209 goto tx_err_dst_release; 1210 1211 if (skb->sk) 1212 skb_set_owner_w(new_skb, skb->sk); 1213 consume_skb(skb); 1214 skb = new_skb; 1215 } 1216 1217 if (t->parms.collect_md) { 1218 if (t->encap.type != TUNNEL_ENCAP_NONE) 1219 goto tx_err_dst_release; 1220 } else { 1221 if (use_cache && ndst) 1222 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1223 } 1224 skb_dst_set(skb, dst); 1225 1226 if (hop_limit == 0) { 1227 if (skb->protocol == htons(ETH_P_IP)) 1228 hop_limit = ip_hdr(skb)->ttl; 1229 else if (skb->protocol == htons(ETH_P_IPV6)) 1230 hop_limit = ipv6_hdr(skb)->hop_limit; 1231 else 1232 hop_limit = ip6_dst_hoplimit(dst); 1233 } 1234 1235 /* Calculate max headroom for all the headers and adjust 1236 * needed_headroom if necessary. 1237 */ 1238 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1239 + dst->header_len + t->hlen; 1240 if (max_headroom > dev->needed_headroom) 1241 dev->needed_headroom = max_headroom; 1242 1243 err = ip6_tnl_encap(skb, t, &proto, fl6); 1244 if (err) 1245 return err; 1246 1247 if (encap_limit >= 0) { 1248 init_tel_txopt(&opt, encap_limit); 1249 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1250 } 1251 1252 skb_push(skb, sizeof(struct ipv6hdr)); 1253 skb_reset_network_header(skb); 1254 ipv6h = ipv6_hdr(skb); 1255 ip6_flow_hdr(ipv6h, dsfield, 1256 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1257 ipv6h->hop_limit = hop_limit; 1258 ipv6h->nexthdr = proto; 1259 ipv6h->saddr = fl6->saddr; 1260 ipv6h->daddr = fl6->daddr; 1261 ip6tunnel_xmit(NULL, skb, dev); 1262 return 0; 1263 tx_err_link_failure: 1264 stats->tx_carrier_errors++; 1265 dst_link_failure(skb); 1266 tx_err_dst_release: 1267 dst_release(dst); 1268 return err; 1269 } 1270 EXPORT_SYMBOL(ip6_tnl_xmit); 1271 1272 static inline int 1273 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, 1274 u8 protocol) 1275 { 1276 struct ip6_tnl *t = netdev_priv(dev); 1277 struct ipv6hdr *ipv6h; 1278 const struct iphdr *iph; 1279 int encap_limit = -1; 1280 __u16 offset; 1281 struct flowi6 fl6; 1282 __u8 dsfield, orig_dsfield; 1283 __u32 mtu; 1284 u8 tproto; 1285 int err; 1286 1287 tproto = READ_ONCE(t->parms.proto); 1288 if (tproto != protocol && tproto != 0) 1289 return -1; 1290 1291 if (t->parms.collect_md) { 1292 struct ip_tunnel_info *tun_info; 1293 const struct ip_tunnel_key *key; 1294 1295 tun_info = skb_tunnel_info(skb); 1296 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1297 ip_tunnel_info_af(tun_info) != AF_INET6)) 1298 return -1; 1299 key = &tun_info->key; 1300 memset(&fl6, 0, sizeof(fl6)); 1301 fl6.flowi6_proto = protocol; 1302 fl6.saddr = key->u.ipv6.src; 1303 fl6.daddr = key->u.ipv6.dst; 1304 fl6.flowlabel = key->label; 1305 dsfield = key->tos; 1306 switch (protocol) { 1307 case IPPROTO_IPIP: 1308 iph = ip_hdr(skb); 1309 orig_dsfield = ipv4_get_dsfield(iph); 1310 break; 1311 case IPPROTO_IPV6: 1312 ipv6h = ipv6_hdr(skb); 1313 orig_dsfield = ipv6_get_dsfield(ipv6h); 1314 break; 1315 default: 1316 orig_dsfield = dsfield; 1317 break; 1318 } 1319 } else { 1320 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1321 encap_limit = t->parms.encap_limit; 1322 if (protocol == IPPROTO_IPV6) { 1323 offset = ip6_tnl_parse_tlv_enc_lim(skb, 1324 skb_network_header(skb)); 1325 /* ip6_tnl_parse_tlv_enc_lim() might have 1326 * reallocated skb->head 1327 */ 1328 if (offset > 0) { 1329 struct ipv6_tlv_tnl_enc_lim *tel; 1330 1331 tel = (void *)&skb_network_header(skb)[offset]; 1332 if (tel->encap_limit == 0) { 1333 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 1334 ICMPV6_HDR_FIELD, offset + 2); 1335 return -1; 1336 } 1337 encap_limit = tel->encap_limit - 1; 1338 } 1339 } 1340 1341 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1342 fl6.flowi6_proto = protocol; 1343 1344 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1345 fl6.flowi6_mark = skb->mark; 1346 else 1347 fl6.flowi6_mark = t->parms.fwmark; 1348 switch (protocol) { 1349 case IPPROTO_IPIP: 1350 iph = ip_hdr(skb); 1351 orig_dsfield = ipv4_get_dsfield(iph); 1352 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1353 dsfield = orig_dsfield; 1354 else 1355 dsfield = ip6_tclass(t->parms.flowinfo); 1356 break; 1357 case IPPROTO_IPV6: 1358 ipv6h = ipv6_hdr(skb); 1359 orig_dsfield = ipv6_get_dsfield(ipv6h); 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 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1365 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1366 break; 1367 default: 1368 orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo); 1369 break; 1370 } 1371 } 1372 1373 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1374 dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield); 1375 1376 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1377 return -1; 1378 1379 skb_set_inner_ipproto(skb, protocol); 1380 1381 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1382 protocol); 1383 if (err != 0) { 1384 /* XXX: send ICMP error even if DF is not set. */ 1385 if (err == -EMSGSIZE) 1386 switch (protocol) { 1387 case IPPROTO_IPIP: 1388 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1389 ICMP_FRAG_NEEDED, htonl(mtu)); 1390 break; 1391 case IPPROTO_IPV6: 1392 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1393 break; 1394 default: 1395 break; 1396 } 1397 return -1; 1398 } 1399 1400 return 0; 1401 } 1402 1403 static netdev_tx_t 1404 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1405 { 1406 struct ip6_tnl *t = netdev_priv(dev); 1407 struct net_device_stats *stats = &t->dev->stats; 1408 u8 ipproto; 1409 int ret; 1410 1411 if (!pskb_inet_may_pull(skb)) 1412 goto tx_err; 1413 1414 switch (skb->protocol) { 1415 case htons(ETH_P_IP): 1416 ipproto = IPPROTO_IPIP; 1417 break; 1418 case htons(ETH_P_IPV6): 1419 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb))) 1420 goto tx_err; 1421 ipproto = IPPROTO_IPV6; 1422 break; 1423 case htons(ETH_P_MPLS_UC): 1424 ipproto = IPPROTO_MPLS; 1425 break; 1426 default: 1427 goto tx_err; 1428 } 1429 1430 ret = ipxip6_tnl_xmit(skb, dev, ipproto); 1431 if (ret < 0) 1432 goto tx_err; 1433 1434 return NETDEV_TX_OK; 1435 1436 tx_err: 1437 stats->tx_errors++; 1438 stats->tx_dropped++; 1439 kfree_skb(skb); 1440 return NETDEV_TX_OK; 1441 } 1442 1443 static void ip6_tnl_link_config(struct ip6_tnl *t) 1444 { 1445 struct net_device *dev = t->dev; 1446 struct net_device *tdev = NULL; 1447 struct __ip6_tnl_parm *p = &t->parms; 1448 struct flowi6 *fl6 = &t->fl.u.ip6; 1449 unsigned int mtu; 1450 int t_hlen; 1451 1452 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1453 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1454 1455 /* Set up flowi template */ 1456 fl6->saddr = p->laddr; 1457 fl6->daddr = p->raddr; 1458 fl6->flowi6_oif = p->link; 1459 fl6->flowlabel = 0; 1460 1461 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1462 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1463 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1464 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1465 1466 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1467 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1468 1469 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1470 dev->flags |= IFF_POINTOPOINT; 1471 else 1472 dev->flags &= ~IFF_POINTOPOINT; 1473 1474 t->tun_hlen = 0; 1475 t->hlen = t->encap_hlen + t->tun_hlen; 1476 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1477 1478 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1479 int strict = (ipv6_addr_type(&p->raddr) & 1480 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1481 1482 struct rt6_info *rt = rt6_lookup(t->net, 1483 &p->raddr, &p->laddr, 1484 p->link, NULL, strict); 1485 if (rt) { 1486 tdev = rt->dst.dev; 1487 ip6_rt_put(rt); 1488 } 1489 1490 if (!tdev && p->link) 1491 tdev = __dev_get_by_index(t->net, p->link); 1492 1493 if (tdev) { 1494 dev->hard_header_len = tdev->hard_header_len + t_hlen; 1495 mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU); 1496 1497 dev->mtu = mtu - t_hlen; 1498 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1499 dev->mtu -= 8; 1500 1501 if (dev->mtu < IPV6_MIN_MTU) 1502 dev->mtu = IPV6_MIN_MTU; 1503 } 1504 } 1505 } 1506 1507 /** 1508 * ip6_tnl_change - update the tunnel parameters 1509 * @t: tunnel to be changed 1510 * @p: tunnel configuration parameters 1511 * 1512 * Description: 1513 * ip6_tnl_change() updates the tunnel parameters 1514 **/ 1515 1516 static int 1517 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1518 { 1519 t->parms.laddr = p->laddr; 1520 t->parms.raddr = p->raddr; 1521 t->parms.flags = p->flags; 1522 t->parms.hop_limit = p->hop_limit; 1523 t->parms.encap_limit = p->encap_limit; 1524 t->parms.flowinfo = p->flowinfo; 1525 t->parms.link = p->link; 1526 t->parms.proto = p->proto; 1527 t->parms.fwmark = p->fwmark; 1528 dst_cache_reset(&t->dst_cache); 1529 ip6_tnl_link_config(t); 1530 return 0; 1531 } 1532 1533 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1534 { 1535 struct net *net = t->net; 1536 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1537 int err; 1538 1539 ip6_tnl_unlink(ip6n, t); 1540 synchronize_net(); 1541 err = ip6_tnl_change(t, p); 1542 ip6_tnl_link(ip6n, t); 1543 netdev_state_change(t->dev); 1544 return err; 1545 } 1546 1547 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1548 { 1549 /* for default tnl0 device allow to change only the proto */ 1550 t->parms.proto = p->proto; 1551 netdev_state_change(t->dev); 1552 return 0; 1553 } 1554 1555 static void 1556 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1557 { 1558 p->laddr = u->laddr; 1559 p->raddr = u->raddr; 1560 p->flags = u->flags; 1561 p->hop_limit = u->hop_limit; 1562 p->encap_limit = u->encap_limit; 1563 p->flowinfo = u->flowinfo; 1564 p->link = u->link; 1565 p->proto = u->proto; 1566 memcpy(p->name, u->name, sizeof(u->name)); 1567 } 1568 1569 static void 1570 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1571 { 1572 u->laddr = p->laddr; 1573 u->raddr = p->raddr; 1574 u->flags = p->flags; 1575 u->hop_limit = p->hop_limit; 1576 u->encap_limit = p->encap_limit; 1577 u->flowinfo = p->flowinfo; 1578 u->link = p->link; 1579 u->proto = p->proto; 1580 memcpy(u->name, p->name, sizeof(u->name)); 1581 } 1582 1583 /** 1584 * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace 1585 * @dev: virtual device associated with tunnel 1586 * @ifr: unused 1587 * @data: parameters passed from userspace 1588 * @cmd: command to be performed 1589 * 1590 * Description: 1591 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1592 * from userspace. 1593 * 1594 * The possible commands are the following: 1595 * %SIOCGETTUNNEL: get tunnel parameters for device 1596 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1597 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1598 * %SIOCDELTUNNEL: delete tunnel 1599 * 1600 * The fallback device "ip6tnl0", created during module 1601 * initialization, can be used for creating other tunnel devices. 1602 * 1603 * Return: 1604 * 0 on success, 1605 * %-EFAULT if unable to copy data to or from userspace, 1606 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1607 * %-EINVAL if passed tunnel parameters are invalid, 1608 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1609 * %-ENODEV if attempting to change or delete a nonexisting device 1610 **/ 1611 1612 static int 1613 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr, 1614 void __user *data, int cmd) 1615 { 1616 int err = 0; 1617 struct ip6_tnl_parm p; 1618 struct __ip6_tnl_parm p1; 1619 struct ip6_tnl *t = netdev_priv(dev); 1620 struct net *net = t->net; 1621 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1622 1623 memset(&p1, 0, sizeof(p1)); 1624 1625 switch (cmd) { 1626 case SIOCGETTUNNEL: 1627 if (dev == ip6n->fb_tnl_dev) { 1628 if (copy_from_user(&p, data, sizeof(p))) { 1629 err = -EFAULT; 1630 break; 1631 } 1632 ip6_tnl_parm_from_user(&p1, &p); 1633 t = ip6_tnl_locate(net, &p1, 0); 1634 if (IS_ERR(t)) 1635 t = netdev_priv(dev); 1636 } else { 1637 memset(&p, 0, sizeof(p)); 1638 } 1639 ip6_tnl_parm_to_user(&p, &t->parms); 1640 if (copy_to_user(data, &p, sizeof(p))) 1641 err = -EFAULT; 1642 break; 1643 case SIOCADDTUNNEL: 1644 case SIOCCHGTUNNEL: 1645 err = -EPERM; 1646 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1647 break; 1648 err = -EFAULT; 1649 if (copy_from_user(&p, data, sizeof(p))) 1650 break; 1651 err = -EINVAL; 1652 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1653 p.proto != 0) 1654 break; 1655 ip6_tnl_parm_from_user(&p1, &p); 1656 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1657 if (cmd == SIOCCHGTUNNEL) { 1658 if (!IS_ERR(t)) { 1659 if (t->dev != dev) { 1660 err = -EEXIST; 1661 break; 1662 } 1663 } else 1664 t = netdev_priv(dev); 1665 if (dev == ip6n->fb_tnl_dev) 1666 err = ip6_tnl0_update(t, &p1); 1667 else 1668 err = ip6_tnl_update(t, &p1); 1669 } 1670 if (!IS_ERR(t)) { 1671 err = 0; 1672 ip6_tnl_parm_to_user(&p, &t->parms); 1673 if (copy_to_user(data, &p, sizeof(p))) 1674 err = -EFAULT; 1675 1676 } else { 1677 err = PTR_ERR(t); 1678 } 1679 break; 1680 case SIOCDELTUNNEL: 1681 err = -EPERM; 1682 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1683 break; 1684 1685 if (dev == ip6n->fb_tnl_dev) { 1686 err = -EFAULT; 1687 if (copy_from_user(&p, data, sizeof(p))) 1688 break; 1689 err = -ENOENT; 1690 ip6_tnl_parm_from_user(&p1, &p); 1691 t = ip6_tnl_locate(net, &p1, 0); 1692 if (IS_ERR(t)) 1693 break; 1694 err = -EPERM; 1695 if (t->dev == ip6n->fb_tnl_dev) 1696 break; 1697 dev = t->dev; 1698 } 1699 err = 0; 1700 unregister_netdevice(dev); 1701 break; 1702 default: 1703 err = -EINVAL; 1704 } 1705 return err; 1706 } 1707 1708 /** 1709 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1710 * @dev: virtual device associated with tunnel 1711 * @new_mtu: the new mtu 1712 * 1713 * Return: 1714 * 0 on success, 1715 * %-EINVAL if mtu too small 1716 **/ 1717 1718 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1719 { 1720 struct ip6_tnl *tnl = netdev_priv(dev); 1721 1722 if (tnl->parms.proto == IPPROTO_IPV6) { 1723 if (new_mtu < IPV6_MIN_MTU) 1724 return -EINVAL; 1725 } else { 1726 if (new_mtu < ETH_MIN_MTU) 1727 return -EINVAL; 1728 } 1729 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) { 1730 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len) 1731 return -EINVAL; 1732 } else { 1733 if (new_mtu > IP_MAX_MTU - dev->hard_header_len) 1734 return -EINVAL; 1735 } 1736 dev->mtu = new_mtu; 1737 return 0; 1738 } 1739 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1740 1741 int ip6_tnl_get_iflink(const struct net_device *dev) 1742 { 1743 struct ip6_tnl *t = netdev_priv(dev); 1744 1745 return t->parms.link; 1746 } 1747 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1748 1749 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1750 unsigned int num) 1751 { 1752 if (num >= MAX_IPTUN_ENCAP_OPS) 1753 return -ERANGE; 1754 1755 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1756 &ip6tun_encaps[num], 1757 NULL, ops) ? 0 : -1; 1758 } 1759 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1760 1761 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1762 unsigned int num) 1763 { 1764 int ret; 1765 1766 if (num >= MAX_IPTUN_ENCAP_OPS) 1767 return -ERANGE; 1768 1769 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1770 &ip6tun_encaps[num], 1771 ops, NULL) == ops) ? 0 : -1; 1772 1773 synchronize_net(); 1774 1775 return ret; 1776 } 1777 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1778 1779 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1780 struct ip_tunnel_encap *ipencap) 1781 { 1782 int hlen; 1783 1784 memset(&t->encap, 0, sizeof(t->encap)); 1785 1786 hlen = ip6_encap_hlen(ipencap); 1787 if (hlen < 0) 1788 return hlen; 1789 1790 t->encap.type = ipencap->type; 1791 t->encap.sport = ipencap->sport; 1792 t->encap.dport = ipencap->dport; 1793 t->encap.flags = ipencap->flags; 1794 1795 t->encap_hlen = hlen; 1796 t->hlen = t->encap_hlen + t->tun_hlen; 1797 1798 return 0; 1799 } 1800 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1801 1802 static const struct net_device_ops ip6_tnl_netdev_ops = { 1803 .ndo_init = ip6_tnl_dev_init, 1804 .ndo_uninit = ip6_tnl_dev_uninit, 1805 .ndo_start_xmit = ip6_tnl_start_xmit, 1806 .ndo_siocdevprivate = ip6_tnl_siocdevprivate, 1807 .ndo_change_mtu = ip6_tnl_change_mtu, 1808 .ndo_get_stats64 = dev_get_tstats64, 1809 .ndo_get_iflink = ip6_tnl_get_iflink, 1810 }; 1811 1812 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1813 NETIF_F_FRAGLIST | \ 1814 NETIF_F_HIGHDMA | \ 1815 NETIF_F_GSO_SOFTWARE | \ 1816 NETIF_F_HW_CSUM) 1817 1818 /** 1819 * ip6_tnl_dev_setup - setup virtual tunnel device 1820 * @dev: virtual device associated with tunnel 1821 * 1822 * Description: 1823 * Initialize function pointers and device parameters 1824 **/ 1825 1826 static void ip6_tnl_dev_setup(struct net_device *dev) 1827 { 1828 dev->netdev_ops = &ip6_tnl_netdev_ops; 1829 dev->header_ops = &ip_tunnel_header_ops; 1830 dev->needs_free_netdev = true; 1831 dev->priv_destructor = ip6_dev_free; 1832 1833 dev->type = ARPHRD_TUNNEL6; 1834 dev->flags |= IFF_NOARP; 1835 dev->addr_len = sizeof(struct in6_addr); 1836 dev->features |= NETIF_F_LLTX; 1837 netif_keep_dst(dev); 1838 1839 dev->features |= IPXIPX_FEATURES; 1840 dev->hw_features |= IPXIPX_FEATURES; 1841 1842 /* This perm addr will be used as interface identifier by IPv6 */ 1843 dev->addr_assign_type = NET_ADDR_RANDOM; 1844 eth_random_addr(dev->perm_addr); 1845 } 1846 1847 1848 /** 1849 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1850 * @dev: virtual device associated with tunnel 1851 **/ 1852 1853 static inline int 1854 ip6_tnl_dev_init_gen(struct net_device *dev) 1855 { 1856 struct ip6_tnl *t = netdev_priv(dev); 1857 int ret; 1858 int t_hlen; 1859 1860 t->dev = dev; 1861 t->net = dev_net(dev); 1862 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1863 if (!dev->tstats) 1864 return -ENOMEM; 1865 1866 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1867 if (ret) 1868 goto free_stats; 1869 1870 ret = gro_cells_init(&t->gro_cells, dev); 1871 if (ret) 1872 goto destroy_dst; 1873 1874 t->tun_hlen = 0; 1875 t->hlen = t->encap_hlen + t->tun_hlen; 1876 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1877 1878 dev->type = ARPHRD_TUNNEL6; 1879 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1880 dev->mtu = ETH_DATA_LEN - t_hlen; 1881 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1882 dev->mtu -= 8; 1883 dev->min_mtu = ETH_MIN_MTU; 1884 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len; 1885 1886 dev_hold_track(dev, &t->dev_tracker, GFP_KERNEL); 1887 return 0; 1888 1889 destroy_dst: 1890 dst_cache_destroy(&t->dst_cache); 1891 free_stats: 1892 free_percpu(dev->tstats); 1893 dev->tstats = NULL; 1894 1895 return ret; 1896 } 1897 1898 /** 1899 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1900 * @dev: virtual device associated with tunnel 1901 **/ 1902 1903 static int ip6_tnl_dev_init(struct net_device *dev) 1904 { 1905 struct ip6_tnl *t = netdev_priv(dev); 1906 int err = ip6_tnl_dev_init_gen(dev); 1907 1908 if (err) 1909 return err; 1910 ip6_tnl_link_config(t); 1911 if (t->parms.collect_md) 1912 netif_keep_dst(dev); 1913 return 0; 1914 } 1915 1916 /** 1917 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1918 * @dev: fallback device 1919 * 1920 * Return: 0 1921 **/ 1922 1923 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1924 { 1925 struct ip6_tnl *t = netdev_priv(dev); 1926 struct net *net = dev_net(dev); 1927 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1928 1929 t->parms.proto = IPPROTO_IPV6; 1930 1931 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1932 return 0; 1933 } 1934 1935 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[], 1936 struct netlink_ext_ack *extack) 1937 { 1938 u8 proto; 1939 1940 if (!data || !data[IFLA_IPTUN_PROTO]) 1941 return 0; 1942 1943 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1944 if (proto != IPPROTO_IPV6 && 1945 proto != IPPROTO_IPIP && 1946 proto != 0) 1947 return -EINVAL; 1948 1949 return 0; 1950 } 1951 1952 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1953 struct __ip6_tnl_parm *parms) 1954 { 1955 memset(parms, 0, sizeof(*parms)); 1956 1957 if (!data) 1958 return; 1959 1960 if (data[IFLA_IPTUN_LINK]) 1961 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1962 1963 if (data[IFLA_IPTUN_LOCAL]) 1964 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1965 1966 if (data[IFLA_IPTUN_REMOTE]) 1967 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1968 1969 if (data[IFLA_IPTUN_TTL]) 1970 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1971 1972 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1973 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1974 1975 if (data[IFLA_IPTUN_FLOWINFO]) 1976 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1977 1978 if (data[IFLA_IPTUN_FLAGS]) 1979 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1980 1981 if (data[IFLA_IPTUN_PROTO]) 1982 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1983 1984 if (data[IFLA_IPTUN_COLLECT_METADATA]) 1985 parms->collect_md = true; 1986 1987 if (data[IFLA_IPTUN_FWMARK]) 1988 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 1989 } 1990 1991 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[], 1992 struct ip_tunnel_encap *ipencap) 1993 { 1994 bool ret = false; 1995 1996 memset(ipencap, 0, sizeof(*ipencap)); 1997 1998 if (!data) 1999 return ret; 2000 2001 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 2002 ret = true; 2003 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 2004 } 2005 2006 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 2007 ret = true; 2008 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 2009 } 2010 2011 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 2012 ret = true; 2013 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 2014 } 2015 2016 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 2017 ret = true; 2018 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 2019 } 2020 2021 return ret; 2022 } 2023 2024 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 2025 struct nlattr *tb[], struct nlattr *data[], 2026 struct netlink_ext_ack *extack) 2027 { 2028 struct net *net = dev_net(dev); 2029 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2030 struct ip_tunnel_encap ipencap; 2031 struct ip6_tnl *nt, *t; 2032 int err; 2033 2034 nt = netdev_priv(dev); 2035 2036 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2037 err = ip6_tnl_encap_setup(nt, &ipencap); 2038 if (err < 0) 2039 return err; 2040 } 2041 2042 ip6_tnl_netlink_parms(data, &nt->parms); 2043 2044 if (nt->parms.collect_md) { 2045 if (rtnl_dereference(ip6n->collect_md_tun)) 2046 return -EEXIST; 2047 } else { 2048 t = ip6_tnl_locate(net, &nt->parms, 0); 2049 if (!IS_ERR(t)) 2050 return -EEXIST; 2051 } 2052 2053 err = ip6_tnl_create2(dev); 2054 if (!err && tb[IFLA_MTU]) 2055 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2056 2057 return err; 2058 } 2059 2060 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2061 struct nlattr *data[], 2062 struct netlink_ext_ack *extack) 2063 { 2064 struct ip6_tnl *t = netdev_priv(dev); 2065 struct __ip6_tnl_parm p; 2066 struct net *net = t->net; 2067 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2068 struct ip_tunnel_encap ipencap; 2069 2070 if (dev == ip6n->fb_tnl_dev) 2071 return -EINVAL; 2072 2073 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) { 2074 int err = ip6_tnl_encap_setup(t, &ipencap); 2075 2076 if (err < 0) 2077 return err; 2078 } 2079 ip6_tnl_netlink_parms(data, &p); 2080 if (p.collect_md) 2081 return -EINVAL; 2082 2083 t = ip6_tnl_locate(net, &p, 0); 2084 if (!IS_ERR(t)) { 2085 if (t->dev != dev) 2086 return -EEXIST; 2087 } else 2088 t = netdev_priv(dev); 2089 2090 return ip6_tnl_update(t, &p); 2091 } 2092 2093 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2094 { 2095 struct net *net = dev_net(dev); 2096 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2097 2098 if (dev != ip6n->fb_tnl_dev) 2099 unregister_netdevice_queue(dev, head); 2100 } 2101 2102 static size_t ip6_tnl_get_size(const struct net_device *dev) 2103 { 2104 return 2105 /* IFLA_IPTUN_LINK */ 2106 nla_total_size(4) + 2107 /* IFLA_IPTUN_LOCAL */ 2108 nla_total_size(sizeof(struct in6_addr)) + 2109 /* IFLA_IPTUN_REMOTE */ 2110 nla_total_size(sizeof(struct in6_addr)) + 2111 /* IFLA_IPTUN_TTL */ 2112 nla_total_size(1) + 2113 /* IFLA_IPTUN_ENCAP_LIMIT */ 2114 nla_total_size(1) + 2115 /* IFLA_IPTUN_FLOWINFO */ 2116 nla_total_size(4) + 2117 /* IFLA_IPTUN_FLAGS */ 2118 nla_total_size(4) + 2119 /* IFLA_IPTUN_PROTO */ 2120 nla_total_size(1) + 2121 /* IFLA_IPTUN_ENCAP_TYPE */ 2122 nla_total_size(2) + 2123 /* IFLA_IPTUN_ENCAP_FLAGS */ 2124 nla_total_size(2) + 2125 /* IFLA_IPTUN_ENCAP_SPORT */ 2126 nla_total_size(2) + 2127 /* IFLA_IPTUN_ENCAP_DPORT */ 2128 nla_total_size(2) + 2129 /* IFLA_IPTUN_COLLECT_METADATA */ 2130 nla_total_size(0) + 2131 /* IFLA_IPTUN_FWMARK */ 2132 nla_total_size(4) + 2133 0; 2134 } 2135 2136 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2137 { 2138 struct ip6_tnl *tunnel = netdev_priv(dev); 2139 struct __ip6_tnl_parm *parm = &tunnel->parms; 2140 2141 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2142 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2143 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2144 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2145 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2146 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2147 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2148 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2149 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2150 goto nla_put_failure; 2151 2152 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2153 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2154 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2155 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2156 goto nla_put_failure; 2157 2158 if (parm->collect_md) 2159 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2160 goto nla_put_failure; 2161 2162 return 0; 2163 2164 nla_put_failure: 2165 return -EMSGSIZE; 2166 } 2167 2168 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2169 { 2170 struct ip6_tnl *tunnel = netdev_priv(dev); 2171 2172 return tunnel->net; 2173 } 2174 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2175 2176 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2177 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2178 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2179 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2180 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2181 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2182 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2183 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2184 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2185 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2186 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2187 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2188 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2189 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2190 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2191 }; 2192 2193 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2194 .kind = "ip6tnl", 2195 .maxtype = IFLA_IPTUN_MAX, 2196 .policy = ip6_tnl_policy, 2197 .priv_size = sizeof(struct ip6_tnl), 2198 .setup = ip6_tnl_dev_setup, 2199 .validate = ip6_tnl_validate, 2200 .newlink = ip6_tnl_newlink, 2201 .changelink = ip6_tnl_changelink, 2202 .dellink = ip6_tnl_dellink, 2203 .get_size = ip6_tnl_get_size, 2204 .fill_info = ip6_tnl_fill_info, 2205 .get_link_net = ip6_tnl_get_link_net, 2206 }; 2207 2208 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2209 .handler = ip4ip6_rcv, 2210 .err_handler = ip4ip6_err, 2211 .priority = 1, 2212 }; 2213 2214 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2215 .handler = ip6ip6_rcv, 2216 .err_handler = ip6ip6_err, 2217 .priority = 1, 2218 }; 2219 2220 static struct xfrm6_tunnel mplsip6_handler __read_mostly = { 2221 .handler = mplsip6_rcv, 2222 .err_handler = mplsip6_err, 2223 .priority = 1, 2224 }; 2225 2226 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list) 2227 { 2228 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2229 struct net_device *dev, *aux; 2230 int h; 2231 struct ip6_tnl *t; 2232 2233 for_each_netdev_safe(net, dev, aux) 2234 if (dev->rtnl_link_ops == &ip6_link_ops) 2235 unregister_netdevice_queue(dev, list); 2236 2237 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2238 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2239 while (t) { 2240 /* If dev is in the same netns, it has already 2241 * been added to the list by the previous loop. 2242 */ 2243 if (!net_eq(dev_net(t->dev), net)) 2244 unregister_netdevice_queue(t->dev, list); 2245 t = rtnl_dereference(t->next); 2246 } 2247 } 2248 2249 t = rtnl_dereference(ip6n->tnls_wc[0]); 2250 while (t) { 2251 /* If dev is in the same netns, it has already 2252 * been added to the list by the previous loop. 2253 */ 2254 if (!net_eq(dev_net(t->dev), net)) 2255 unregister_netdevice_queue(t->dev, list); 2256 t = rtnl_dereference(t->next); 2257 } 2258 } 2259 2260 static int __net_init ip6_tnl_init_net(struct net *net) 2261 { 2262 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2263 struct ip6_tnl *t = NULL; 2264 int err; 2265 2266 ip6n->tnls[0] = ip6n->tnls_wc; 2267 ip6n->tnls[1] = ip6n->tnls_r_l; 2268 2269 if (!net_has_fallback_tunnels(net)) 2270 return 0; 2271 err = -ENOMEM; 2272 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2273 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2274 2275 if (!ip6n->fb_tnl_dev) 2276 goto err_alloc_dev; 2277 dev_net_set(ip6n->fb_tnl_dev, net); 2278 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2279 /* FB netdevice is special: we have one, and only one per netns. 2280 * Allowing to move it to another netns is clearly unsafe. 2281 */ 2282 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 2283 2284 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2285 if (err < 0) 2286 goto err_register; 2287 2288 err = register_netdev(ip6n->fb_tnl_dev); 2289 if (err < 0) 2290 goto err_register; 2291 2292 t = netdev_priv(ip6n->fb_tnl_dev); 2293 2294 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2295 return 0; 2296 2297 err_register: 2298 free_netdev(ip6n->fb_tnl_dev); 2299 err_alloc_dev: 2300 return err; 2301 } 2302 2303 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list) 2304 { 2305 struct net *net; 2306 LIST_HEAD(list); 2307 2308 rtnl_lock(); 2309 list_for_each_entry(net, net_list, exit_list) 2310 ip6_tnl_destroy_tunnels(net, &list); 2311 unregister_netdevice_many(&list); 2312 rtnl_unlock(); 2313 } 2314 2315 static struct pernet_operations ip6_tnl_net_ops = { 2316 .init = ip6_tnl_init_net, 2317 .exit_batch = ip6_tnl_exit_batch_net, 2318 .id = &ip6_tnl_net_id, 2319 .size = sizeof(struct ip6_tnl_net), 2320 }; 2321 2322 /** 2323 * ip6_tunnel_init - register protocol and reserve needed resources 2324 * 2325 * Return: 0 on success 2326 **/ 2327 2328 static int __init ip6_tunnel_init(void) 2329 { 2330 int err; 2331 2332 if (!ipv6_mod_enabled()) 2333 return -EOPNOTSUPP; 2334 2335 err = register_pernet_device(&ip6_tnl_net_ops); 2336 if (err < 0) 2337 goto out_pernet; 2338 2339 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2340 if (err < 0) { 2341 pr_err("%s: can't register ip4ip6\n", __func__); 2342 goto out_ip4ip6; 2343 } 2344 2345 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2346 if (err < 0) { 2347 pr_err("%s: can't register ip6ip6\n", __func__); 2348 goto out_ip6ip6; 2349 } 2350 2351 if (ip6_tnl_mpls_supported()) { 2352 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS); 2353 if (err < 0) { 2354 pr_err("%s: can't register mplsip6\n", __func__); 2355 goto out_mplsip6; 2356 } 2357 } 2358 2359 err = rtnl_link_register(&ip6_link_ops); 2360 if (err < 0) 2361 goto rtnl_link_failed; 2362 2363 return 0; 2364 2365 rtnl_link_failed: 2366 if (ip6_tnl_mpls_supported()) 2367 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS); 2368 out_mplsip6: 2369 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2370 out_ip6ip6: 2371 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2372 out_ip4ip6: 2373 unregister_pernet_device(&ip6_tnl_net_ops); 2374 out_pernet: 2375 return err; 2376 } 2377 2378 /** 2379 * ip6_tunnel_cleanup - free resources and unregister protocol 2380 **/ 2381 2382 static void __exit ip6_tunnel_cleanup(void) 2383 { 2384 rtnl_link_unregister(&ip6_link_ops); 2385 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2386 pr_info("%s: can't deregister ip4ip6\n", __func__); 2387 2388 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2389 pr_info("%s: can't deregister ip6ip6\n", __func__); 2390 2391 if (ip6_tnl_mpls_supported() && 2392 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS)) 2393 pr_info("%s: can't deregister mplsip6\n", __func__); 2394 unregister_pernet_device(&ip6_tnl_net_ops); 2395 } 2396 2397 module_init(ip6_tunnel_init); 2398 module_exit(ip6_tunnel_cleanup); 2399