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 #include <net/inet_dscp.h> 57 58 MODULE_AUTHOR("Ville Nuorvala"); 59 MODULE_DESCRIPTION("IPv6 tunneling device"); 60 MODULE_LICENSE("GPL"); 61 MODULE_ALIAS_RTNL_LINK("ip6tnl"); 62 MODULE_ALIAS_NETDEV("ip6tnl0"); 63 64 #define IP6_TUNNEL_HASH_SIZE_SHIFT 5 65 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT) 66 67 static bool log_ecn_error = true; 68 module_param(log_ecn_error, bool, 0644); 69 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 70 71 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 72 { 73 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 74 75 return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT); 76 } 77 78 static int ip6_tnl_dev_init(struct net_device *dev); 79 static void ip6_tnl_dev_setup(struct net_device *dev); 80 static struct rtnl_link_ops ip6_link_ops __read_mostly; 81 82 static unsigned int ip6_tnl_net_id __read_mostly; 83 struct ip6_tnl_net { 84 /* the IPv6 tunnel fallback device */ 85 struct net_device *fb_tnl_dev; 86 /* lists for storing tunnels in use */ 87 struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE]; 88 struct ip6_tnl __rcu *tnls_wc[1]; 89 struct ip6_tnl __rcu **tnls[2]; 90 struct ip6_tnl __rcu *collect_md_tun; 91 }; 92 93 static inline int ip6_tnl_mpls_supported(void) 94 { 95 return IS_ENABLED(CONFIG_MPLS); 96 } 97 98 #define for_each_ip6_tunnel_rcu(start) \ 99 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 100 101 /** 102 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 103 * @net: network namespace 104 * @link: ifindex of underlying interface 105 * @remote: the address of the tunnel exit-point 106 * @local: the address of the tunnel entry-point 107 * 108 * Return: 109 * tunnel matching given end-points if found, 110 * else fallback tunnel if its device is up, 111 * else %NULL 112 **/ 113 114 static struct ip6_tnl * 115 ip6_tnl_lookup(struct net *net, int link, 116 const struct in6_addr *remote, const struct in6_addr *local) 117 { 118 unsigned int hash = HASH(remote, local); 119 struct ip6_tnl *t, *cand = NULL; 120 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 121 struct in6_addr any; 122 123 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 124 if (!ipv6_addr_equal(local, &t->parms.laddr) || 125 !ipv6_addr_equal(remote, &t->parms.raddr) || 126 !(t->dev->flags & IFF_UP)) 127 continue; 128 129 if (link == t->parms.link) 130 return t; 131 else 132 cand = t; 133 } 134 135 memset(&any, 0, sizeof(any)); 136 hash = HASH(&any, local); 137 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 138 if (!ipv6_addr_equal(local, &t->parms.laddr) || 139 !ipv6_addr_any(&t->parms.raddr) || 140 !(t->dev->flags & IFF_UP)) 141 continue; 142 143 if (link == t->parms.link) 144 return t; 145 else if (!cand) 146 cand = t; 147 } 148 149 hash = HASH(remote, &any); 150 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 151 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 152 !ipv6_addr_any(&t->parms.laddr) || 153 !(t->dev->flags & IFF_UP)) 154 continue; 155 156 if (link == t->parms.link) 157 return t; 158 else if (!cand) 159 cand = t; 160 } 161 162 if (cand) 163 return cand; 164 165 t = rcu_dereference(ip6n->collect_md_tun); 166 if (t && t->dev->flags & IFF_UP) 167 return t; 168 169 t = rcu_dereference(ip6n->tnls_wc[0]); 170 if (t && (t->dev->flags & IFF_UP)) 171 return t; 172 173 return NULL; 174 } 175 176 /** 177 * ip6_tnl_bucket - get head of list matching given tunnel parameters 178 * @ip6n: the private data for ip6_vti in the netns 179 * @p: parameters containing tunnel end-points 180 * 181 * Description: 182 * ip6_tnl_bucket() returns the head of the list matching the 183 * &struct in6_addr entries laddr and raddr in @p. 184 * 185 * Return: head of IPv6 tunnel list 186 **/ 187 188 static struct ip6_tnl __rcu ** 189 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 190 { 191 const struct in6_addr *remote = &p->raddr; 192 const struct in6_addr *local = &p->laddr; 193 unsigned int h = 0; 194 int prio = 0; 195 196 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 197 prio = 1; 198 h = HASH(remote, local); 199 } 200 return &ip6n->tnls[prio][h]; 201 } 202 203 /** 204 * ip6_tnl_link - add tunnel to hash table 205 * @ip6n: the private data for ip6_vti in the netns 206 * @t: tunnel to be added 207 **/ 208 209 static void 210 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 211 { 212 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 213 214 if (t->parms.collect_md) 215 rcu_assign_pointer(ip6n->collect_md_tun, t); 216 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 217 rcu_assign_pointer(*tp, t); 218 } 219 220 /** 221 * ip6_tnl_unlink - remove tunnel from hash table 222 * @ip6n: the private data for ip6_vti in the netns 223 * @t: tunnel to be removed 224 **/ 225 226 static void 227 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 228 { 229 struct ip6_tnl __rcu **tp; 230 struct ip6_tnl *iter; 231 232 if (t->parms.collect_md) 233 rcu_assign_pointer(ip6n->collect_md_tun, NULL); 234 235 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 236 (iter = rtnl_dereference(*tp)) != NULL; 237 tp = &iter->next) { 238 if (t == iter) { 239 rcu_assign_pointer(*tp, t->next); 240 break; 241 } 242 } 243 } 244 245 static void ip6_dev_free(struct net_device *dev) 246 { 247 struct ip6_tnl *t = netdev_priv(dev); 248 249 gro_cells_destroy(&t->gro_cells); 250 dst_cache_destroy(&t->dst_cache); 251 } 252 253 static int ip6_tnl_create2(struct net_device *dev) 254 { 255 struct ip6_tnl *t = netdev_priv(dev); 256 struct ip6_tnl_net *ip6n = net_generic(t->net, ip6_tnl_net_id); 257 int err; 258 259 dev->rtnl_link_ops = &ip6_link_ops; 260 err = register_netdevice(dev); 261 if (err < 0) 262 goto out; 263 264 strcpy(t->parms.name, dev->name); 265 266 ip6_tnl_link(ip6n, t); 267 return 0; 268 269 out: 270 return err; 271 } 272 273 /** 274 * ip6_tnl_create - create a new tunnel 275 * @net: network namespace 276 * @p: tunnel parameters 277 * 278 * Description: 279 * Create tunnel matching given parameters. 280 * 281 * Return: 282 * created tunnel or error pointer 283 **/ 284 285 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 286 { 287 struct net_device *dev; 288 struct ip6_tnl *t; 289 char name[IFNAMSIZ]; 290 int err = -E2BIG; 291 292 if (p->name[0]) { 293 if (!dev_valid_name(p->name)) 294 goto failed; 295 strscpy(name, p->name, IFNAMSIZ); 296 } else { 297 sprintf(name, "ip6tnl%%d"); 298 } 299 err = -ENOMEM; 300 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 301 ip6_tnl_dev_setup); 302 if (!dev) 303 goto failed; 304 305 dev_net_set(dev, net); 306 307 t = netdev_priv(dev); 308 t->parms = *p; 309 t->net = dev_net(dev); 310 err = ip6_tnl_create2(dev); 311 if (err < 0) 312 goto failed_free; 313 314 return t; 315 316 failed_free: 317 free_netdev(dev); 318 failed: 319 return ERR_PTR(err); 320 } 321 322 /** 323 * ip6_tnl_locate - find or create tunnel matching given parameters 324 * @net: network namespace 325 * @p: tunnel parameters 326 * @create: != 0 if allowed to create new tunnel if no match found 327 * 328 * Description: 329 * ip6_tnl_locate() first tries to locate an existing tunnel 330 * based on @parms. If this is unsuccessful, but @create is set a new 331 * tunnel device is created and registered for use. 332 * 333 * Return: 334 * matching tunnel or error pointer 335 **/ 336 337 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 338 struct __ip6_tnl_parm *p, int create) 339 { 340 const struct in6_addr *remote = &p->raddr; 341 const struct in6_addr *local = &p->laddr; 342 struct ip6_tnl __rcu **tp; 343 struct ip6_tnl *t; 344 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 345 346 for (tp = ip6_tnl_bucket(ip6n, p); 347 (t = rtnl_dereference(*tp)) != NULL; 348 tp = &t->next) { 349 if (ipv6_addr_equal(local, &t->parms.laddr) && 350 ipv6_addr_equal(remote, &t->parms.raddr) && 351 p->link == t->parms.link) { 352 if (create) 353 return ERR_PTR(-EEXIST); 354 355 return t; 356 } 357 } 358 if (!create) 359 return ERR_PTR(-ENODEV); 360 return ip6_tnl_create(net, p); 361 } 362 363 /** 364 * ip6_tnl_dev_uninit - tunnel device uninitializer 365 * @dev: the device to be destroyed 366 * 367 * Description: 368 * ip6_tnl_dev_uninit() removes tunnel from its list 369 **/ 370 371 static void 372 ip6_tnl_dev_uninit(struct net_device *dev) 373 { 374 struct ip6_tnl *t = netdev_priv(dev); 375 struct net *net = t->net; 376 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 377 378 if (dev == ip6n->fb_tnl_dev) 379 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 380 else 381 ip6_tnl_unlink(ip6n, t); 382 dst_cache_reset(&t->dst_cache); 383 netdev_put(dev, &t->dev_tracker); 384 } 385 386 /** 387 * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option 388 * @skb: received socket buffer 389 * @raw: the ICMPv6 error message data 390 * 391 * Return: 392 * 0 if none was found, 393 * else index to encapsulation limit 394 **/ 395 396 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 397 { 398 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw; 399 unsigned int nhoff = raw - skb->data; 400 unsigned int off = nhoff + sizeof(*ipv6h); 401 u8 nexthdr = ipv6h->nexthdr; 402 403 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 404 struct ipv6_opt_hdr *hdr; 405 u16 optlen; 406 407 if (!pskb_may_pull(skb, off + sizeof(*hdr))) 408 break; 409 410 hdr = (struct ipv6_opt_hdr *)(skb->data + off); 411 if (nexthdr == NEXTHDR_FRAGMENT) { 412 optlen = 8; 413 } else if (nexthdr == NEXTHDR_AUTH) { 414 optlen = ipv6_authlen(hdr); 415 } else { 416 optlen = ipv6_optlen(hdr); 417 } 418 419 if (!pskb_may_pull(skb, off + optlen)) 420 break; 421 422 hdr = (struct ipv6_opt_hdr *)(skb->data + off); 423 if (nexthdr == NEXTHDR_FRAGMENT) { 424 struct frag_hdr *frag_hdr = (struct frag_hdr *)hdr; 425 426 if (frag_hdr->frag_off) 427 break; 428 } 429 if (nexthdr == NEXTHDR_DEST) { 430 u16 i = 2; 431 432 while (1) { 433 struct ipv6_tlv_tnl_enc_lim *tel; 434 435 /* No more room for encapsulation limit */ 436 if (i + sizeof(*tel) > optlen) 437 break; 438 439 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i); 440 /* return index of option if found and valid */ 441 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 442 tel->length == 1) 443 return i + off - nhoff; 444 /* else jump to next option */ 445 if (tel->type) 446 i += tel->length + 2; 447 else 448 i++; 449 } 450 } 451 nexthdr = hdr->nexthdr; 452 off += optlen; 453 } 454 return 0; 455 } 456 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 457 458 /* ip6_tnl_err() should handle errors in the tunnel according to the 459 * specifications in RFC 2473. 460 */ 461 static int 462 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 463 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 464 { 465 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data; 466 struct net *net = dev_net(skb->dev); 467 u8 rel_type = ICMPV6_DEST_UNREACH; 468 u8 rel_code = ICMPV6_ADDR_UNREACH; 469 __u32 rel_info = 0; 470 struct ip6_tnl *t; 471 int err = -ENOENT; 472 int rel_msg = 0; 473 u8 tproto; 474 __u16 len; 475 476 /* If the packet doesn't contain the original IPv6 header we are 477 in trouble since we might need the source address for further 478 processing of the error. */ 479 480 rcu_read_lock(); 481 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr); 482 if (!t) 483 goto out; 484 485 tproto = READ_ONCE(t->parms.proto); 486 if (tproto != ipproto && tproto != 0) 487 goto out; 488 489 err = 0; 490 491 switch (*type) { 492 case ICMPV6_DEST_UNREACH: 493 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 494 t->parms.name); 495 rel_msg = 1; 496 break; 497 case ICMPV6_TIME_EXCEED: 498 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 499 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 500 t->parms.name); 501 rel_msg = 1; 502 } 503 break; 504 case ICMPV6_PARAMPROB: { 505 struct ipv6_tlv_tnl_enc_lim *tel; 506 __u32 teli; 507 508 teli = 0; 509 if ((*code) == ICMPV6_HDR_FIELD) 510 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 511 512 if (teli && teli == *info - 2) { 513 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 514 if (tel->encap_limit == 0) { 515 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 516 t->parms.name); 517 rel_msg = 1; 518 } 519 } else { 520 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 521 t->parms.name); 522 } 523 break; 524 } 525 case ICMPV6_PKT_TOOBIG: { 526 __u32 mtu; 527 528 ip6_update_pmtu(skb, net, htonl(*info), 0, 0, 529 sock_net_uid(net, NULL)); 530 mtu = *info - offset; 531 if (mtu < IPV6_MIN_MTU) 532 mtu = IPV6_MIN_MTU; 533 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len); 534 if (len > mtu) { 535 rel_type = ICMPV6_PKT_TOOBIG; 536 rel_code = 0; 537 rel_info = mtu; 538 rel_msg = 1; 539 } 540 break; 541 } 542 case NDISC_REDIRECT: 543 ip6_redirect(skb, net, skb->dev->ifindex, 0, 544 sock_net_uid(net, NULL)); 545 break; 546 } 547 548 *type = rel_type; 549 *code = rel_code; 550 *info = rel_info; 551 *msg = rel_msg; 552 553 out: 554 rcu_read_unlock(); 555 return err; 556 } 557 558 static int 559 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 560 u8 type, u8 code, int offset, __be32 info) 561 { 562 __u32 rel_info = ntohl(info); 563 const struct iphdr *eiph; 564 struct sk_buff *skb2; 565 int err, rel_msg = 0; 566 u8 rel_type = type; 567 u8 rel_code = code; 568 struct rtable *rt; 569 struct flowi4 fl4; 570 571 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 572 &rel_msg, &rel_info, offset); 573 if (err < 0) 574 return err; 575 576 if (rel_msg == 0) 577 return 0; 578 579 switch (rel_type) { 580 case ICMPV6_DEST_UNREACH: 581 if (rel_code != ICMPV6_ADDR_UNREACH) 582 return 0; 583 rel_type = ICMP_DEST_UNREACH; 584 rel_code = ICMP_HOST_UNREACH; 585 break; 586 case ICMPV6_PKT_TOOBIG: 587 if (rel_code != 0) 588 return 0; 589 rel_type = ICMP_DEST_UNREACH; 590 rel_code = ICMP_FRAG_NEEDED; 591 break; 592 default: 593 return 0; 594 } 595 596 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 597 return 0; 598 599 skb2 = skb_clone(skb, GFP_ATOMIC); 600 if (!skb2) 601 return 0; 602 603 skb_dst_drop(skb2); 604 605 skb_pull(skb2, offset); 606 skb_reset_network_header(skb2); 607 eiph = ip_hdr(skb2); 608 609 /* Try to guess incoming interface */ 610 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr, 611 0, 0, 0, IPPROTO_IPIP, 612 eiph->tos & INET_DSCP_MASK, 0); 613 if (IS_ERR(rt)) 614 goto out; 615 616 skb2->dev = rt->dst.dev; 617 ip_rt_put(rt); 618 619 /* route "incoming" packet */ 620 if (rt->rt_flags & RTCF_LOCAL) { 621 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 622 eiph->daddr, eiph->saddr, 0, 0, 623 IPPROTO_IPIP, 624 eiph->tos & INET_DSCP_MASK, 0); 625 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) { 626 if (!IS_ERR(rt)) 627 ip_rt_put(rt); 628 goto out; 629 } 630 skb_dst_set(skb2, &rt->dst); 631 } else { 632 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, 633 ip4h_dscp(eiph), skb2->dev) || 634 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6) 635 goto out; 636 } 637 638 /* change mtu on this route */ 639 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 640 if (rel_info > dst_mtu(skb_dst(skb2))) 641 goto out; 642 643 skb_dst_update_pmtu_no_confirm(skb2, rel_info); 644 } 645 646 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 647 648 out: 649 kfree_skb(skb2); 650 return 0; 651 } 652 653 static int 654 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 655 u8 type, u8 code, int offset, __be32 info) 656 { 657 __u32 rel_info = ntohl(info); 658 int err, rel_msg = 0; 659 u8 rel_type = type; 660 u8 rel_code = code; 661 662 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 663 &rel_msg, &rel_info, offset); 664 if (err < 0) 665 return err; 666 667 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 668 struct rt6_info *rt; 669 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 670 671 if (!skb2) 672 return 0; 673 674 skb_dst_drop(skb2); 675 skb_pull(skb2, offset); 676 skb_reset_network_header(skb2); 677 678 /* Try to guess incoming interface */ 679 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 680 NULL, 0, skb2, 0); 681 682 if (rt && rt->dst.dev) 683 skb2->dev = rt->dst.dev; 684 685 icmpv6_send(skb2, rel_type, rel_code, rel_info); 686 687 ip6_rt_put(rt); 688 689 kfree_skb(skb2); 690 } 691 692 return 0; 693 } 694 695 static int 696 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 697 u8 type, u8 code, int offset, __be32 info) 698 { 699 __u32 rel_info = ntohl(info); 700 int err, rel_msg = 0; 701 u8 rel_type = type; 702 u8 rel_code = code; 703 704 err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code, 705 &rel_msg, &rel_info, offset); 706 return err; 707 } 708 709 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 710 const struct ipv6hdr *ipv6h, 711 struct sk_buff *skb) 712 { 713 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 714 715 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 716 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 717 718 return IP6_ECN_decapsulate(ipv6h, skb); 719 } 720 721 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 722 const struct ipv6hdr *ipv6h, 723 struct sk_buff *skb) 724 { 725 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 726 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 727 728 return IP6_ECN_decapsulate(ipv6h, skb); 729 } 730 731 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 732 const struct ipv6hdr *ipv6h, 733 struct sk_buff *skb) 734 { 735 /* ECN is not supported in AF_MPLS */ 736 return 0; 737 } 738 739 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 740 const struct in6_addr *laddr, 741 const struct in6_addr *raddr) 742 { 743 struct __ip6_tnl_parm *p = &t->parms; 744 int ltype = ipv6_addr_type(laddr); 745 int rtype = ipv6_addr_type(raddr); 746 __u32 flags = 0; 747 748 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 749 flags = IP6_TNL_F_CAP_PER_PACKET; 750 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 751 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 752 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 753 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 754 if (ltype&IPV6_ADDR_UNICAST) 755 flags |= IP6_TNL_F_CAP_XMIT; 756 if (rtype&IPV6_ADDR_UNICAST) 757 flags |= IP6_TNL_F_CAP_RCV; 758 } 759 return flags; 760 } 761 EXPORT_SYMBOL(ip6_tnl_get_cap); 762 763 /* called with rcu_read_lock() */ 764 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 765 const struct in6_addr *laddr, 766 const struct in6_addr *raddr) 767 { 768 struct __ip6_tnl_parm *p = &t->parms; 769 int ret = 0; 770 struct net *net = t->net; 771 772 if ((p->flags & IP6_TNL_F_CAP_RCV) || 773 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 774 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 775 struct net_device *ldev = NULL; 776 777 if (p->link) 778 ldev = dev_get_by_index_rcu(net, p->link); 779 780 if ((ipv6_addr_is_multicast(laddr) || 781 likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false, 782 0, IFA_F_TENTATIVE))) && 783 ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) || 784 likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true, 785 0, IFA_F_TENTATIVE)))) 786 ret = 1; 787 } 788 return ret; 789 } 790 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 791 792 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb, 793 const struct tnl_ptk_info *tpi, 794 struct metadata_dst *tun_dst, 795 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 796 const struct ipv6hdr *ipv6h, 797 struct sk_buff *skb), 798 bool log_ecn_err) 799 { 800 const struct ipv6hdr *ipv6h; 801 int nh, err; 802 803 if (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.i_flags) != 804 test_bit(IP_TUNNEL_CSUM_BIT, tpi->flags)) { 805 DEV_STATS_INC(tunnel->dev, rx_crc_errors); 806 DEV_STATS_INC(tunnel->dev, rx_errors); 807 goto drop; 808 } 809 810 if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.i_flags)) { 811 if (!test_bit(IP_TUNNEL_SEQ_BIT, tpi->flags) || 812 (tunnel->i_seqno && 813 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) { 814 DEV_STATS_INC(tunnel->dev, rx_fifo_errors); 815 DEV_STATS_INC(tunnel->dev, rx_errors); 816 goto drop; 817 } 818 tunnel->i_seqno = ntohl(tpi->seq) + 1; 819 } 820 821 skb->protocol = tpi->proto; 822 823 /* Warning: All skb pointers will be invalidated! */ 824 if (tunnel->dev->type == ARPHRD_ETHER) { 825 if (!pskb_may_pull(skb, ETH_HLEN)) { 826 DEV_STATS_INC(tunnel->dev, rx_length_errors); 827 DEV_STATS_INC(tunnel->dev, rx_errors); 828 goto drop; 829 } 830 831 skb->protocol = eth_type_trans(skb, tunnel->dev); 832 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); 833 } else { 834 skb->dev = tunnel->dev; 835 skb_reset_mac_header(skb); 836 } 837 838 /* Save offset of outer header relative to skb->head, 839 * because we are going to reset the network header to the inner header 840 * and might change skb->head. 841 */ 842 nh = skb_network_header(skb) - skb->head; 843 844 skb_reset_network_header(skb); 845 846 if (!pskb_inet_may_pull(skb)) { 847 DEV_STATS_INC(tunnel->dev, rx_length_errors); 848 DEV_STATS_INC(tunnel->dev, rx_errors); 849 goto drop; 850 } 851 852 /* Get the outer header. */ 853 ipv6h = (struct ipv6hdr *)(skb->head + nh); 854 855 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 856 857 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 858 859 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb); 860 if (unlikely(err)) { 861 if (log_ecn_err) 862 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n", 863 &ipv6h->saddr, 864 ipv6_get_dsfield(ipv6h)); 865 if (err > 1) { 866 DEV_STATS_INC(tunnel->dev, rx_frame_errors); 867 DEV_STATS_INC(tunnel->dev, rx_errors); 868 goto drop; 869 } 870 } 871 872 dev_sw_netstats_rx_add(tunnel->dev, skb->len); 873 874 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev))); 875 876 if (tun_dst) 877 skb_dst_set(skb, (struct dst_entry *)tun_dst); 878 879 gro_cells_receive(&tunnel->gro_cells, skb); 880 return 0; 881 882 drop: 883 if (tun_dst) 884 dst_release((struct dst_entry *)tun_dst); 885 kfree_skb(skb); 886 return 0; 887 } 888 889 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb, 890 const struct tnl_ptk_info *tpi, 891 struct metadata_dst *tun_dst, 892 bool log_ecn_err) 893 { 894 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 895 const struct ipv6hdr *ipv6h, 896 struct sk_buff *skb); 897 898 dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate; 899 if (tpi->proto == htons(ETH_P_IP)) 900 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate; 901 902 return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 903 log_ecn_err); 904 } 905 EXPORT_SYMBOL(ip6_tnl_rcv); 906 907 static const struct tnl_ptk_info tpi_v6 = { 908 /* no tunnel info required for ipxip6. */ 909 .proto = htons(ETH_P_IPV6), 910 }; 911 912 static const struct tnl_ptk_info tpi_v4 = { 913 /* no tunnel info required for ipxip6. */ 914 .proto = htons(ETH_P_IP), 915 }; 916 917 static const struct tnl_ptk_info tpi_mpls = { 918 /* no tunnel info required for mplsip6. */ 919 .proto = htons(ETH_P_MPLS_UC), 920 }; 921 922 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto, 923 const struct tnl_ptk_info *tpi, 924 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 925 const struct ipv6hdr *ipv6h, 926 struct sk_buff *skb)) 927 { 928 struct ip6_tnl *t; 929 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 930 struct metadata_dst *tun_dst = NULL; 931 int ret = -1; 932 933 rcu_read_lock(); 934 t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr); 935 936 if (t) { 937 u8 tproto = READ_ONCE(t->parms.proto); 938 939 if (tproto != ipproto && tproto != 0) 940 goto drop; 941 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 942 goto drop; 943 ipv6h = ipv6_hdr(skb); 944 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) 945 goto drop; 946 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 947 goto drop; 948 if (t->parms.collect_md) { 949 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 950 951 tun_dst = ipv6_tun_rx_dst(skb, flags, 0, 0); 952 if (!tun_dst) 953 goto drop; 954 } 955 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate, 956 log_ecn_error); 957 } 958 959 rcu_read_unlock(); 960 961 return ret; 962 963 drop: 964 rcu_read_unlock(); 965 kfree_skb(skb); 966 return 0; 967 } 968 969 static int ip4ip6_rcv(struct sk_buff *skb) 970 { 971 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4, 972 ip4ip6_dscp_ecn_decapsulate); 973 } 974 975 static int ip6ip6_rcv(struct sk_buff *skb) 976 { 977 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6, 978 ip6ip6_dscp_ecn_decapsulate); 979 } 980 981 static int mplsip6_rcv(struct sk_buff *skb) 982 { 983 return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls, 984 mplsip6_dscp_ecn_decapsulate); 985 } 986 987 struct ipv6_tel_txoption { 988 struct ipv6_txoptions ops; 989 __u8 dst_opt[8]; 990 }; 991 992 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 993 { 994 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 995 996 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 997 opt->dst_opt[3] = 1; 998 opt->dst_opt[4] = encap_limit; 999 opt->dst_opt[5] = IPV6_TLV_PADN; 1000 opt->dst_opt[6] = 1; 1001 1002 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt; 1003 opt->ops.opt_nflen = 8; 1004 } 1005 1006 /** 1007 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 1008 * @t: the outgoing tunnel device 1009 * @hdr: IPv6 header from the incoming packet 1010 * 1011 * Description: 1012 * Avoid trivial tunneling loop by checking that tunnel exit-point 1013 * doesn't match source of incoming packet. 1014 * 1015 * Return: 1016 * 1 if conflict, 1017 * 0 else 1018 **/ 1019 1020 static inline bool 1021 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 1022 { 1023 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 1024 } 1025 1026 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, 1027 const struct in6_addr *laddr, 1028 const struct in6_addr *raddr) 1029 { 1030 struct __ip6_tnl_parm *p = &t->parms; 1031 int ret = 0; 1032 struct net *net = t->net; 1033 1034 if (t->parms.collect_md) 1035 return 1; 1036 1037 if ((p->flags & IP6_TNL_F_CAP_XMIT) || 1038 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 1039 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) { 1040 struct net_device *ldev = NULL; 1041 1042 rcu_read_lock(); 1043 if (p->link) 1044 ldev = dev_get_by_index_rcu(net, p->link); 1045 1046 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false, 1047 0, IFA_F_TENTATIVE))) 1048 pr_warn_ratelimited("%s xmit: Local address not yet configured!\n", 1049 p->name); 1050 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) && 1051 !ipv6_addr_is_multicast(raddr) && 1052 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev, 1053 true, 0, IFA_F_TENTATIVE))) 1054 pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n", 1055 p->name); 1056 else 1057 ret = 1; 1058 rcu_read_unlock(); 1059 } 1060 return ret; 1061 } 1062 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 1063 1064 /** 1065 * ip6_tnl_xmit - encapsulate packet and send 1066 * @skb: the outgoing socket buffer 1067 * @dev: the outgoing tunnel device 1068 * @dsfield: dscp code for outer header 1069 * @fl6: flow of tunneled packet 1070 * @encap_limit: encapsulation limit 1071 * @pmtu: Path MTU is stored if packet is too big 1072 * @proto: next header value 1073 * 1074 * Description: 1075 * Build new header and do some sanity checks on the packet before sending 1076 * it. 1077 * 1078 * Return: 1079 * 0 on success 1080 * -1 fail 1081 * %-EMSGSIZE message too big. return mtu in this case. 1082 **/ 1083 1084 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield, 1085 struct flowi6 *fl6, int encap_limit, __u32 *pmtu, 1086 __u8 proto) 1087 { 1088 struct ip6_tnl *t = netdev_priv(dev); 1089 struct net *net = t->net; 1090 struct ipv6hdr *ipv6h; 1091 struct ipv6_tel_txoption opt; 1092 struct dst_entry *dst = NULL, *ndst = NULL; 1093 struct net_device *tdev; 1094 int mtu; 1095 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0; 1096 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen; 1097 unsigned int max_headroom = psh_hlen; 1098 __be16 payload_protocol; 1099 bool use_cache = false; 1100 u8 hop_limit; 1101 int err = -1; 1102 1103 payload_protocol = skb_protocol(skb, true); 1104 1105 if (t->parms.collect_md) { 1106 hop_limit = skb_tunnel_info(skb)->key.ttl; 1107 goto route_lookup; 1108 } else { 1109 hop_limit = t->parms.hop_limit; 1110 } 1111 1112 /* NBMA tunnel */ 1113 if (ipv6_addr_any(&t->parms.raddr)) { 1114 if (payload_protocol == htons(ETH_P_IPV6)) { 1115 struct in6_addr *addr6; 1116 struct neighbour *neigh; 1117 int addr_type; 1118 1119 if (!skb_dst(skb)) 1120 goto tx_err_link_failure; 1121 1122 neigh = dst_neigh_lookup(skb_dst(skb), 1123 &ipv6_hdr(skb)->daddr); 1124 if (!neigh) 1125 goto tx_err_link_failure; 1126 1127 addr6 = (struct in6_addr *)&neigh->primary_key; 1128 addr_type = ipv6_addr_type(addr6); 1129 1130 if (addr_type == IPV6_ADDR_ANY) 1131 addr6 = &ipv6_hdr(skb)->daddr; 1132 1133 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr)); 1134 neigh_release(neigh); 1135 } else if (payload_protocol == htons(ETH_P_IP)) { 1136 const struct rtable *rt = skb_rtable(skb); 1137 1138 if (!rt) 1139 goto tx_err_link_failure; 1140 1141 if (rt->rt_gw_family == AF_INET6) 1142 memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr)); 1143 } 1144 } else if (t->parms.proto != 0 && !(t->parms.flags & 1145 (IP6_TNL_F_USE_ORIG_TCLASS | 1146 IP6_TNL_F_USE_ORIG_FWMARK))) { 1147 /* enable the cache only if neither the outer protocol nor the 1148 * routing decision depends on the current inner header value 1149 */ 1150 use_cache = true; 1151 } 1152 1153 if (use_cache) 1154 dst = dst_cache_get(&t->dst_cache); 1155 1156 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr)) 1157 goto tx_err_link_failure; 1158 1159 if (!dst) { 1160 route_lookup: 1161 /* add dsfield to flowlabel for route lookup */ 1162 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel); 1163 1164 dst = ip6_route_output(net, NULL, fl6); 1165 1166 if (dst->error) 1167 goto tx_err_link_failure; 1168 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 1169 if (IS_ERR(dst)) { 1170 err = PTR_ERR(dst); 1171 dst = NULL; 1172 goto tx_err_link_failure; 1173 } 1174 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) && 1175 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev, 1176 &fl6->daddr, 0, &fl6->saddr)) 1177 goto tx_err_link_failure; 1178 ndst = dst; 1179 } 1180 1181 tdev = dst->dev; 1182 1183 if (tdev == dev) { 1184 DEV_STATS_INC(dev, collisions); 1185 net_warn_ratelimited("%s: Local routing loop detected!\n", 1186 t->parms.name); 1187 goto tx_err_dst_release; 1188 } 1189 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen; 1190 if (encap_limit >= 0) { 1191 max_headroom += 8; 1192 mtu -= 8; 1193 } 1194 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ? 1195 IPV6_MIN_MTU : IPV4_MIN_MTU); 1196 1197 skb_dst_update_pmtu_no_confirm(skb, mtu); 1198 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) { 1199 *pmtu = mtu; 1200 err = -EMSGSIZE; 1201 goto tx_err_dst_release; 1202 } 1203 1204 if (t->err_count > 0) { 1205 if (time_before(jiffies, 1206 t->err_time + IP6TUNNEL_ERR_TIMEO)) { 1207 t->err_count--; 1208 1209 dst_link_failure(skb); 1210 } else { 1211 t->err_count = 0; 1212 } 1213 } 1214 1215 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1216 1217 /* 1218 * Okay, now see if we can stuff it in the buffer as-is. 1219 */ 1220 max_headroom += LL_RESERVED_SPACE(tdev); 1221 1222 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1223 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1224 struct sk_buff *new_skb; 1225 1226 new_skb = skb_realloc_headroom(skb, max_headroom); 1227 if (!new_skb) 1228 goto tx_err_dst_release; 1229 1230 if (skb->sk) 1231 skb_set_owner_w(new_skb, skb->sk); 1232 consume_skb(skb); 1233 skb = new_skb; 1234 } 1235 1236 if (t->parms.collect_md) { 1237 if (t->encap.type != TUNNEL_ENCAP_NONE) 1238 goto tx_err_dst_release; 1239 } else { 1240 if (use_cache && ndst) 1241 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr); 1242 } 1243 skb_dst_set(skb, dst); 1244 1245 if (hop_limit == 0) { 1246 if (payload_protocol == htons(ETH_P_IP)) 1247 hop_limit = ip_hdr(skb)->ttl; 1248 else if (payload_protocol == htons(ETH_P_IPV6)) 1249 hop_limit = ipv6_hdr(skb)->hop_limit; 1250 else 1251 hop_limit = ip6_dst_hoplimit(dst); 1252 } 1253 1254 /* Calculate max headroom for all the headers and adjust 1255 * needed_headroom if necessary. 1256 */ 1257 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr) 1258 + dst->header_len + t->hlen; 1259 if (max_headroom > READ_ONCE(dev->needed_headroom)) 1260 WRITE_ONCE(dev->needed_headroom, max_headroom); 1261 1262 err = ip6_tnl_encap(skb, t, &proto, fl6); 1263 if (err) 1264 return err; 1265 1266 if (encap_limit >= 0) { 1267 init_tel_txopt(&opt, encap_limit); 1268 ipv6_push_frag_opts(skb, &opt.ops, &proto); 1269 } 1270 1271 skb_push(skb, sizeof(struct ipv6hdr)); 1272 skb_reset_network_header(skb); 1273 ipv6h = ipv6_hdr(skb); 1274 ip6_flow_hdr(ipv6h, dsfield, 1275 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6)); 1276 ipv6h->hop_limit = hop_limit; 1277 ipv6h->nexthdr = proto; 1278 ipv6h->saddr = fl6->saddr; 1279 ipv6h->daddr = fl6->daddr; 1280 ip6tunnel_xmit(NULL, skb, dev); 1281 return 0; 1282 tx_err_link_failure: 1283 DEV_STATS_INC(dev, tx_carrier_errors); 1284 dst_link_failure(skb); 1285 tx_err_dst_release: 1286 dst_release(dst); 1287 return err; 1288 } 1289 EXPORT_SYMBOL(ip6_tnl_xmit); 1290 1291 static inline int 1292 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, 1293 u8 protocol) 1294 { 1295 struct ip6_tnl *t = netdev_priv(dev); 1296 struct ipv6hdr *ipv6h; 1297 const struct iphdr *iph; 1298 int encap_limit = -1; 1299 __u16 offset; 1300 struct flowi6 fl6; 1301 __u8 dsfield, orig_dsfield; 1302 __u32 mtu; 1303 u8 tproto; 1304 int err; 1305 1306 tproto = READ_ONCE(t->parms.proto); 1307 if (tproto != protocol && tproto != 0) 1308 return -1; 1309 1310 if (t->parms.collect_md) { 1311 struct ip_tunnel_info *tun_info; 1312 const struct ip_tunnel_key *key; 1313 1314 tun_info = skb_tunnel_info(skb); 1315 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) || 1316 ip_tunnel_info_af(tun_info) != AF_INET6)) 1317 return -1; 1318 key = &tun_info->key; 1319 memset(&fl6, 0, sizeof(fl6)); 1320 fl6.flowi6_proto = protocol; 1321 fl6.saddr = key->u.ipv6.src; 1322 fl6.daddr = key->u.ipv6.dst; 1323 fl6.flowlabel = key->label; 1324 dsfield = key->tos; 1325 switch (protocol) { 1326 case IPPROTO_IPIP: 1327 iph = ip_hdr(skb); 1328 orig_dsfield = ipv4_get_dsfield(iph); 1329 break; 1330 case IPPROTO_IPV6: 1331 ipv6h = ipv6_hdr(skb); 1332 orig_dsfield = ipv6_get_dsfield(ipv6h); 1333 break; 1334 default: 1335 orig_dsfield = dsfield; 1336 break; 1337 } 1338 } else { 1339 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1340 encap_limit = t->parms.encap_limit; 1341 if (protocol == IPPROTO_IPV6) { 1342 offset = ip6_tnl_parse_tlv_enc_lim(skb, 1343 skb_network_header(skb)); 1344 /* ip6_tnl_parse_tlv_enc_lim() might have 1345 * reallocated skb->head 1346 */ 1347 if (offset > 0) { 1348 struct ipv6_tlv_tnl_enc_lim *tel; 1349 1350 tel = (void *)&skb_network_header(skb)[offset]; 1351 if (tel->encap_limit == 0) { 1352 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 1353 ICMPV6_HDR_FIELD, offset + 2); 1354 return -1; 1355 } 1356 encap_limit = tel->encap_limit - 1; 1357 } 1358 } 1359 1360 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1361 fl6.flowi6_proto = protocol; 1362 1363 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1364 fl6.flowi6_mark = skb->mark; 1365 else 1366 fl6.flowi6_mark = t->parms.fwmark; 1367 switch (protocol) { 1368 case IPPROTO_IPIP: 1369 iph = ip_hdr(skb); 1370 orig_dsfield = ipv4_get_dsfield(iph); 1371 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1372 dsfield = orig_dsfield; 1373 else 1374 dsfield = ip6_tclass(t->parms.flowinfo); 1375 break; 1376 case IPPROTO_IPV6: 1377 ipv6h = ipv6_hdr(skb); 1378 orig_dsfield = ipv6_get_dsfield(ipv6h); 1379 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1380 dsfield = orig_dsfield; 1381 else 1382 dsfield = ip6_tclass(t->parms.flowinfo); 1383 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1384 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1385 break; 1386 default: 1387 orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo); 1388 break; 1389 } 1390 } 1391 1392 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1393 dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield); 1394 1395 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6)) 1396 return -1; 1397 1398 skb_set_inner_ipproto(skb, protocol); 1399 1400 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1401 protocol); 1402 if (err != 0) { 1403 /* XXX: send ICMP error even if DF is not set. */ 1404 if (err == -EMSGSIZE) 1405 switch (protocol) { 1406 case IPPROTO_IPIP: 1407 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1408 ICMP_FRAG_NEEDED, htonl(mtu)); 1409 break; 1410 case IPPROTO_IPV6: 1411 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1412 break; 1413 default: 1414 break; 1415 } 1416 return -1; 1417 } 1418 1419 return 0; 1420 } 1421 1422 static netdev_tx_t 1423 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev) 1424 { 1425 struct ip6_tnl *t = netdev_priv(dev); 1426 u8 ipproto; 1427 int ret; 1428 1429 if (!pskb_inet_may_pull(skb)) 1430 goto tx_err; 1431 1432 switch (skb->protocol) { 1433 case htons(ETH_P_IP): 1434 ipproto = IPPROTO_IPIP; 1435 break; 1436 case htons(ETH_P_IPV6): 1437 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb))) 1438 goto tx_err; 1439 ipproto = IPPROTO_IPV6; 1440 break; 1441 case htons(ETH_P_MPLS_UC): 1442 ipproto = IPPROTO_MPLS; 1443 break; 1444 default: 1445 goto tx_err; 1446 } 1447 1448 ret = ipxip6_tnl_xmit(skb, dev, ipproto); 1449 if (ret < 0) 1450 goto tx_err; 1451 1452 return NETDEV_TX_OK; 1453 1454 tx_err: 1455 DEV_STATS_INC(dev, tx_errors); 1456 DEV_STATS_INC(dev, tx_dropped); 1457 kfree_skb(skb); 1458 return NETDEV_TX_OK; 1459 } 1460 1461 static void ip6_tnl_link_config(struct ip6_tnl *t) 1462 { 1463 struct net_device *dev = t->dev; 1464 struct net_device *tdev = NULL; 1465 struct __ip6_tnl_parm *p = &t->parms; 1466 struct flowi6 *fl6 = &t->fl.u.ip6; 1467 int t_hlen; 1468 int mtu; 1469 1470 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1471 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1472 1473 /* Set up flowi template */ 1474 fl6->saddr = p->laddr; 1475 fl6->daddr = p->raddr; 1476 fl6->flowi6_oif = p->link; 1477 fl6->flowlabel = 0; 1478 1479 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1480 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1481 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1482 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1483 1484 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1485 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1486 1487 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1488 dev->flags |= IFF_POINTOPOINT; 1489 else 1490 dev->flags &= ~IFF_POINTOPOINT; 1491 1492 t->tun_hlen = 0; 1493 t->hlen = t->encap_hlen + t->tun_hlen; 1494 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1495 1496 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1497 int strict = (ipv6_addr_type(&p->raddr) & 1498 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1499 1500 struct rt6_info *rt = rt6_lookup(t->net, 1501 &p->raddr, &p->laddr, 1502 p->link, NULL, strict); 1503 if (rt) { 1504 tdev = rt->dst.dev; 1505 ip6_rt_put(rt); 1506 } 1507 1508 if (!tdev && p->link) 1509 tdev = __dev_get_by_index(t->net, p->link); 1510 1511 if (tdev) { 1512 dev->needed_headroom = tdev->hard_header_len + 1513 tdev->needed_headroom + t_hlen; 1514 mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU); 1515 1516 mtu = mtu - t_hlen; 1517 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1518 mtu -= 8; 1519 1520 if (mtu < IPV6_MIN_MTU) 1521 mtu = IPV6_MIN_MTU; 1522 WRITE_ONCE(dev->mtu, mtu); 1523 } 1524 } 1525 } 1526 1527 /** 1528 * ip6_tnl_change - update the tunnel parameters 1529 * @t: tunnel to be changed 1530 * @p: tunnel configuration parameters 1531 * 1532 * Description: 1533 * ip6_tnl_change() updates the tunnel parameters 1534 **/ 1535 1536 static void 1537 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1538 { 1539 t->parms.laddr = p->laddr; 1540 t->parms.raddr = p->raddr; 1541 t->parms.flags = p->flags; 1542 t->parms.hop_limit = p->hop_limit; 1543 t->parms.encap_limit = p->encap_limit; 1544 t->parms.flowinfo = p->flowinfo; 1545 t->parms.link = p->link; 1546 t->parms.proto = p->proto; 1547 t->parms.fwmark = p->fwmark; 1548 dst_cache_reset(&t->dst_cache); 1549 ip6_tnl_link_config(t); 1550 } 1551 1552 static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1553 { 1554 struct net *net = t->net; 1555 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1556 1557 ip6_tnl_unlink(ip6n, t); 1558 synchronize_net(); 1559 ip6_tnl_change(t, p); 1560 ip6_tnl_link(ip6n, t); 1561 netdev_state_change(t->dev); 1562 } 1563 1564 static void ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1565 { 1566 /* for default tnl0 device allow to change only the proto */ 1567 t->parms.proto = p->proto; 1568 netdev_state_change(t->dev); 1569 } 1570 1571 static void 1572 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1573 { 1574 p->laddr = u->laddr; 1575 p->raddr = u->raddr; 1576 p->flags = u->flags; 1577 p->hop_limit = u->hop_limit; 1578 p->encap_limit = u->encap_limit; 1579 p->flowinfo = u->flowinfo; 1580 p->link = u->link; 1581 p->proto = u->proto; 1582 memcpy(p->name, u->name, sizeof(u->name)); 1583 } 1584 1585 static void 1586 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1587 { 1588 u->laddr = p->laddr; 1589 u->raddr = p->raddr; 1590 u->flags = p->flags; 1591 u->hop_limit = p->hop_limit; 1592 u->encap_limit = p->encap_limit; 1593 u->flowinfo = p->flowinfo; 1594 u->link = p->link; 1595 u->proto = p->proto; 1596 memcpy(u->name, p->name, sizeof(u->name)); 1597 } 1598 1599 /** 1600 * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace 1601 * @dev: virtual device associated with tunnel 1602 * @ifr: unused 1603 * @data: parameters passed from userspace 1604 * @cmd: command to be performed 1605 * 1606 * Description: 1607 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1608 * from userspace. 1609 * 1610 * The possible commands are the following: 1611 * %SIOCGETTUNNEL: get tunnel parameters for device 1612 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1613 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1614 * %SIOCDELTUNNEL: delete tunnel 1615 * 1616 * The fallback device "ip6tnl0", created during module 1617 * initialization, can be used for creating other tunnel devices. 1618 * 1619 * Return: 1620 * 0 on success, 1621 * %-EFAULT if unable to copy data to or from userspace, 1622 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1623 * %-EINVAL if passed tunnel parameters are invalid, 1624 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1625 * %-ENODEV if attempting to change or delete a nonexisting device 1626 **/ 1627 1628 static int 1629 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr, 1630 void __user *data, int cmd) 1631 { 1632 int err = 0; 1633 struct ip6_tnl_parm p; 1634 struct __ip6_tnl_parm p1; 1635 struct ip6_tnl *t = netdev_priv(dev); 1636 struct net *net = t->net; 1637 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1638 1639 memset(&p1, 0, sizeof(p1)); 1640 1641 switch (cmd) { 1642 case SIOCGETTUNNEL: 1643 if (dev == ip6n->fb_tnl_dev) { 1644 if (copy_from_user(&p, data, sizeof(p))) { 1645 err = -EFAULT; 1646 break; 1647 } 1648 ip6_tnl_parm_from_user(&p1, &p); 1649 t = ip6_tnl_locate(net, &p1, 0); 1650 if (IS_ERR(t)) 1651 t = netdev_priv(dev); 1652 } else { 1653 memset(&p, 0, sizeof(p)); 1654 } 1655 ip6_tnl_parm_to_user(&p, &t->parms); 1656 if (copy_to_user(data, &p, sizeof(p))) 1657 err = -EFAULT; 1658 break; 1659 case SIOCADDTUNNEL: 1660 case SIOCCHGTUNNEL: 1661 err = -EPERM; 1662 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1663 break; 1664 err = -EFAULT; 1665 if (copy_from_user(&p, data, sizeof(p))) 1666 break; 1667 err = -EINVAL; 1668 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1669 p.proto != 0) 1670 break; 1671 ip6_tnl_parm_from_user(&p1, &p); 1672 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1673 if (cmd == SIOCCHGTUNNEL) { 1674 if (!IS_ERR(t)) { 1675 if (t->dev != dev) { 1676 err = -EEXIST; 1677 break; 1678 } 1679 } else 1680 t = netdev_priv(dev); 1681 if (dev == ip6n->fb_tnl_dev) 1682 ip6_tnl0_update(t, &p1); 1683 else 1684 ip6_tnl_update(t, &p1); 1685 } 1686 if (!IS_ERR(t)) { 1687 err = 0; 1688 ip6_tnl_parm_to_user(&p, &t->parms); 1689 if (copy_to_user(data, &p, sizeof(p))) 1690 err = -EFAULT; 1691 1692 } else { 1693 err = PTR_ERR(t); 1694 } 1695 break; 1696 case SIOCDELTUNNEL: 1697 err = -EPERM; 1698 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1699 break; 1700 1701 if (dev == ip6n->fb_tnl_dev) { 1702 err = -EFAULT; 1703 if (copy_from_user(&p, data, sizeof(p))) 1704 break; 1705 err = -ENOENT; 1706 ip6_tnl_parm_from_user(&p1, &p); 1707 t = ip6_tnl_locate(net, &p1, 0); 1708 if (IS_ERR(t)) 1709 break; 1710 err = -EPERM; 1711 if (t->dev == ip6n->fb_tnl_dev) 1712 break; 1713 dev = t->dev; 1714 } 1715 err = 0; 1716 unregister_netdevice(dev); 1717 break; 1718 default: 1719 err = -EINVAL; 1720 } 1721 return err; 1722 } 1723 1724 /** 1725 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1726 * @dev: virtual device associated with tunnel 1727 * @new_mtu: the new mtu 1728 * 1729 * Return: 1730 * 0 on success, 1731 * %-EINVAL if mtu too small 1732 **/ 1733 1734 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1735 { 1736 struct ip6_tnl *tnl = netdev_priv(dev); 1737 int t_hlen; 1738 1739 t_hlen = tnl->hlen + sizeof(struct ipv6hdr); 1740 if (tnl->parms.proto == IPPROTO_IPV6) { 1741 if (new_mtu < IPV6_MIN_MTU) 1742 return -EINVAL; 1743 } else { 1744 if (new_mtu < ETH_MIN_MTU) 1745 return -EINVAL; 1746 } 1747 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) { 1748 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len - t_hlen) 1749 return -EINVAL; 1750 } else { 1751 if (new_mtu > IP_MAX_MTU - dev->hard_header_len - t_hlen) 1752 return -EINVAL; 1753 } 1754 WRITE_ONCE(dev->mtu, new_mtu); 1755 return 0; 1756 } 1757 EXPORT_SYMBOL(ip6_tnl_change_mtu); 1758 1759 int ip6_tnl_get_iflink(const struct net_device *dev) 1760 { 1761 struct ip6_tnl *t = netdev_priv(dev); 1762 1763 return READ_ONCE(t->parms.link); 1764 } 1765 EXPORT_SYMBOL(ip6_tnl_get_iflink); 1766 1767 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops, 1768 unsigned int num) 1769 { 1770 if (num >= MAX_IPTUN_ENCAP_OPS) 1771 return -ERANGE; 1772 1773 return !cmpxchg((const struct ip6_tnl_encap_ops **) 1774 &ip6tun_encaps[num], 1775 NULL, ops) ? 0 : -1; 1776 } 1777 EXPORT_SYMBOL(ip6_tnl_encap_add_ops); 1778 1779 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops, 1780 unsigned int num) 1781 { 1782 int ret; 1783 1784 if (num >= MAX_IPTUN_ENCAP_OPS) 1785 return -ERANGE; 1786 1787 ret = (cmpxchg((const struct ip6_tnl_encap_ops **) 1788 &ip6tun_encaps[num], 1789 ops, NULL) == ops) ? 0 : -1; 1790 1791 synchronize_net(); 1792 1793 return ret; 1794 } 1795 EXPORT_SYMBOL(ip6_tnl_encap_del_ops); 1796 1797 int ip6_tnl_encap_setup(struct ip6_tnl *t, 1798 struct ip_tunnel_encap *ipencap) 1799 { 1800 int hlen; 1801 1802 memset(&t->encap, 0, sizeof(t->encap)); 1803 1804 hlen = ip6_encap_hlen(ipencap); 1805 if (hlen < 0) 1806 return hlen; 1807 1808 t->encap.type = ipencap->type; 1809 t->encap.sport = ipencap->sport; 1810 t->encap.dport = ipencap->dport; 1811 t->encap.flags = ipencap->flags; 1812 1813 t->encap_hlen = hlen; 1814 t->hlen = t->encap_hlen + t->tun_hlen; 1815 1816 return 0; 1817 } 1818 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup); 1819 1820 static const struct net_device_ops ip6_tnl_netdev_ops = { 1821 .ndo_init = ip6_tnl_dev_init, 1822 .ndo_uninit = ip6_tnl_dev_uninit, 1823 .ndo_start_xmit = ip6_tnl_start_xmit, 1824 .ndo_siocdevprivate = ip6_tnl_siocdevprivate, 1825 .ndo_change_mtu = ip6_tnl_change_mtu, 1826 .ndo_get_stats64 = dev_get_tstats64, 1827 .ndo_get_iflink = ip6_tnl_get_iflink, 1828 }; 1829 1830 #define IPXIPX_FEATURES (NETIF_F_SG | \ 1831 NETIF_F_FRAGLIST | \ 1832 NETIF_F_HIGHDMA | \ 1833 NETIF_F_GSO_SOFTWARE | \ 1834 NETIF_F_HW_CSUM) 1835 1836 /** 1837 * ip6_tnl_dev_setup - setup virtual tunnel device 1838 * @dev: virtual device associated with tunnel 1839 * 1840 * Description: 1841 * Initialize function pointers and device parameters 1842 **/ 1843 1844 static void ip6_tnl_dev_setup(struct net_device *dev) 1845 { 1846 dev->netdev_ops = &ip6_tnl_netdev_ops; 1847 dev->header_ops = &ip_tunnel_header_ops; 1848 dev->needs_free_netdev = true; 1849 dev->priv_destructor = ip6_dev_free; 1850 1851 dev->type = ARPHRD_TUNNEL6; 1852 dev->flags |= IFF_NOARP; 1853 dev->addr_len = sizeof(struct in6_addr); 1854 dev->lltx = true; 1855 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 1856 netif_keep_dst(dev); 1857 1858 dev->features |= IPXIPX_FEATURES; 1859 dev->hw_features |= IPXIPX_FEATURES; 1860 1861 /* This perm addr will be used as interface identifier by IPv6 */ 1862 dev->addr_assign_type = NET_ADDR_RANDOM; 1863 eth_random_addr(dev->perm_addr); 1864 } 1865 1866 1867 /** 1868 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1869 * @dev: virtual device associated with tunnel 1870 **/ 1871 1872 static inline int 1873 ip6_tnl_dev_init_gen(struct net_device *dev) 1874 { 1875 struct ip6_tnl *t = netdev_priv(dev); 1876 int ret; 1877 int t_hlen; 1878 1879 t->dev = dev; 1880 1881 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL); 1882 if (ret) 1883 return ret; 1884 1885 ret = gro_cells_init(&t->gro_cells, dev); 1886 if (ret) 1887 goto destroy_dst; 1888 1889 t->tun_hlen = 0; 1890 t->hlen = t->encap_hlen + t->tun_hlen; 1891 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1892 1893 dev->type = ARPHRD_TUNNEL6; 1894 dev->mtu = ETH_DATA_LEN - t_hlen; 1895 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1896 dev->mtu -= 8; 1897 dev->min_mtu = ETH_MIN_MTU; 1898 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len - t_hlen; 1899 1900 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL); 1901 netdev_lockdep_set_classes(dev); 1902 return 0; 1903 1904 destroy_dst: 1905 dst_cache_destroy(&t->dst_cache); 1906 1907 return ret; 1908 } 1909 1910 /** 1911 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1912 * @dev: virtual device associated with tunnel 1913 **/ 1914 1915 static int ip6_tnl_dev_init(struct net_device *dev) 1916 { 1917 struct ip6_tnl *t = netdev_priv(dev); 1918 int err = ip6_tnl_dev_init_gen(dev); 1919 1920 if (err) 1921 return err; 1922 ip6_tnl_link_config(t); 1923 if (t->parms.collect_md) 1924 netif_keep_dst(dev); 1925 return 0; 1926 } 1927 1928 /** 1929 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1930 * @dev: fallback device 1931 * 1932 * Return: 0 1933 **/ 1934 1935 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1936 { 1937 struct ip6_tnl *t = netdev_priv(dev); 1938 struct net *net = dev_net(dev); 1939 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1940 1941 t->net = net; 1942 t->parms.proto = IPPROTO_IPV6; 1943 1944 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1945 return 0; 1946 } 1947 1948 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[], 1949 struct netlink_ext_ack *extack) 1950 { 1951 u8 proto; 1952 1953 if (!data || !data[IFLA_IPTUN_PROTO]) 1954 return 0; 1955 1956 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1957 if (proto != IPPROTO_IPV6 && 1958 proto != IPPROTO_IPIP && 1959 proto != 0) 1960 return -EINVAL; 1961 1962 return 0; 1963 } 1964 1965 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1966 struct __ip6_tnl_parm *parms) 1967 { 1968 memset(parms, 0, sizeof(*parms)); 1969 1970 if (!data) 1971 return; 1972 1973 if (data[IFLA_IPTUN_LINK]) 1974 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1975 1976 if (data[IFLA_IPTUN_LOCAL]) 1977 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]); 1978 1979 if (data[IFLA_IPTUN_REMOTE]) 1980 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]); 1981 1982 if (data[IFLA_IPTUN_TTL]) 1983 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1984 1985 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1986 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1987 1988 if (data[IFLA_IPTUN_FLOWINFO]) 1989 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1990 1991 if (data[IFLA_IPTUN_FLAGS]) 1992 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1993 1994 if (data[IFLA_IPTUN_PROTO]) 1995 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1996 1997 if (data[IFLA_IPTUN_COLLECT_METADATA]) 1998 parms->collect_md = true; 1999 2000 if (data[IFLA_IPTUN_FWMARK]) 2001 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 2002 } 2003 2004 static int ip6_tnl_newlink(struct net_device *dev, 2005 struct rtnl_newlink_params *params, 2006 struct netlink_ext_ack *extack) 2007 { 2008 struct nlattr **data = params->data; 2009 struct nlattr **tb = params->tb; 2010 struct ip_tunnel_encap ipencap; 2011 struct ip6_tnl_net *ip6n; 2012 struct ip6_tnl *nt, *t; 2013 struct net *net; 2014 int err; 2015 2016 net = params->link_net ? : dev_net(dev); 2017 ip6n = net_generic(net, ip6_tnl_net_id); 2018 nt = netdev_priv(dev); 2019 nt->net = net; 2020 2021 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) { 2022 err = ip6_tnl_encap_setup(nt, &ipencap); 2023 if (err < 0) 2024 return err; 2025 } 2026 2027 ip6_tnl_netlink_parms(data, &nt->parms); 2028 2029 if (nt->parms.collect_md) { 2030 if (rtnl_dereference(ip6n->collect_md_tun)) 2031 return -EEXIST; 2032 } else { 2033 t = ip6_tnl_locate(net, &nt->parms, 0); 2034 if (!IS_ERR(t)) 2035 return -EEXIST; 2036 } 2037 2038 err = ip6_tnl_create2(dev); 2039 if (!err && tb[IFLA_MTU]) 2040 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2041 2042 return err; 2043 } 2044 2045 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 2046 struct nlattr *data[], 2047 struct netlink_ext_ack *extack) 2048 { 2049 struct ip6_tnl *t = netdev_priv(dev); 2050 struct __ip6_tnl_parm p; 2051 struct net *net = t->net; 2052 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2053 struct ip_tunnel_encap ipencap; 2054 2055 if (dev == ip6n->fb_tnl_dev) 2056 return -EINVAL; 2057 2058 if (ip_tunnel_netlink_encap_parms(data, &ipencap)) { 2059 int err = ip6_tnl_encap_setup(t, &ipencap); 2060 2061 if (err < 0) 2062 return err; 2063 } 2064 ip6_tnl_netlink_parms(data, &p); 2065 if (p.collect_md) 2066 return -EINVAL; 2067 2068 t = ip6_tnl_locate(net, &p, 0); 2069 if (!IS_ERR(t)) { 2070 if (t->dev != dev) 2071 return -EEXIST; 2072 } else 2073 t = netdev_priv(dev); 2074 2075 ip6_tnl_update(t, &p); 2076 return 0; 2077 } 2078 2079 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 2080 { 2081 struct net *net = dev_net(dev); 2082 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2083 2084 if (dev != ip6n->fb_tnl_dev) 2085 unregister_netdevice_queue(dev, head); 2086 } 2087 2088 static size_t ip6_tnl_get_size(const struct net_device *dev) 2089 { 2090 return 2091 /* IFLA_IPTUN_LINK */ 2092 nla_total_size(4) + 2093 /* IFLA_IPTUN_LOCAL */ 2094 nla_total_size(sizeof(struct in6_addr)) + 2095 /* IFLA_IPTUN_REMOTE */ 2096 nla_total_size(sizeof(struct in6_addr)) + 2097 /* IFLA_IPTUN_TTL */ 2098 nla_total_size(1) + 2099 /* IFLA_IPTUN_ENCAP_LIMIT */ 2100 nla_total_size(1) + 2101 /* IFLA_IPTUN_FLOWINFO */ 2102 nla_total_size(4) + 2103 /* IFLA_IPTUN_FLAGS */ 2104 nla_total_size(4) + 2105 /* IFLA_IPTUN_PROTO */ 2106 nla_total_size(1) + 2107 /* IFLA_IPTUN_ENCAP_TYPE */ 2108 nla_total_size(2) + 2109 /* IFLA_IPTUN_ENCAP_FLAGS */ 2110 nla_total_size(2) + 2111 /* IFLA_IPTUN_ENCAP_SPORT */ 2112 nla_total_size(2) + 2113 /* IFLA_IPTUN_ENCAP_DPORT */ 2114 nla_total_size(2) + 2115 /* IFLA_IPTUN_COLLECT_METADATA */ 2116 nla_total_size(0) + 2117 /* IFLA_IPTUN_FWMARK */ 2118 nla_total_size(4) + 2119 0; 2120 } 2121 2122 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 2123 { 2124 struct ip6_tnl *tunnel = netdev_priv(dev); 2125 struct __ip6_tnl_parm *parm = &tunnel->parms; 2126 2127 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 2128 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) || 2129 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) || 2130 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 2131 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 2132 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 2133 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 2134 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) || 2135 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark)) 2136 goto nla_put_failure; 2137 2138 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) || 2139 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) || 2140 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) || 2141 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags)) 2142 goto nla_put_failure; 2143 2144 if (parm->collect_md) 2145 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA)) 2146 goto nla_put_failure; 2147 2148 return 0; 2149 2150 nla_put_failure: 2151 return -EMSGSIZE; 2152 } 2153 2154 struct net *ip6_tnl_get_link_net(const struct net_device *dev) 2155 { 2156 struct ip6_tnl *tunnel = netdev_priv(dev); 2157 2158 return READ_ONCE(tunnel->net); 2159 } 2160 EXPORT_SYMBOL(ip6_tnl_get_link_net); 2161 2162 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 2163 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 2164 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 2165 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 2166 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 2167 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 2168 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 2169 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 2170 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 2171 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 2172 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 2173 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 2174 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 2175 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG }, 2176 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 2177 }; 2178 2179 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 2180 .kind = "ip6tnl", 2181 .maxtype = IFLA_IPTUN_MAX, 2182 .policy = ip6_tnl_policy, 2183 .priv_size = sizeof(struct ip6_tnl), 2184 .setup = ip6_tnl_dev_setup, 2185 .validate = ip6_tnl_validate, 2186 .newlink = ip6_tnl_newlink, 2187 .changelink = ip6_tnl_changelink, 2188 .dellink = ip6_tnl_dellink, 2189 .get_size = ip6_tnl_get_size, 2190 .fill_info = ip6_tnl_fill_info, 2191 .get_link_net = ip6_tnl_get_link_net, 2192 }; 2193 2194 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 2195 .handler = ip4ip6_rcv, 2196 .err_handler = ip4ip6_err, 2197 .priority = 1, 2198 }; 2199 2200 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 2201 .handler = ip6ip6_rcv, 2202 .err_handler = ip6ip6_err, 2203 .priority = 1, 2204 }; 2205 2206 static struct xfrm6_tunnel mplsip6_handler __read_mostly = { 2207 .handler = mplsip6_rcv, 2208 .err_handler = mplsip6_err, 2209 .priority = 1, 2210 }; 2211 2212 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list) 2213 { 2214 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2215 struct net_device *dev, *aux; 2216 int h; 2217 struct ip6_tnl *t; 2218 2219 for_each_netdev_safe(net, dev, aux) 2220 if (dev->rtnl_link_ops == &ip6_link_ops) 2221 unregister_netdevice_queue(dev, list); 2222 2223 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) { 2224 t = rtnl_dereference(ip6n->tnls_r_l[h]); 2225 while (t) { 2226 /* If dev is in the same netns, it has already 2227 * been added to the list by the previous loop. 2228 */ 2229 if (!net_eq(dev_net(t->dev), net)) 2230 unregister_netdevice_queue(t->dev, list); 2231 t = rtnl_dereference(t->next); 2232 } 2233 } 2234 2235 t = rtnl_dereference(ip6n->tnls_wc[0]); 2236 while (t) { 2237 /* If dev is in the same netns, it has already 2238 * been added to the list by the previous loop. 2239 */ 2240 if (!net_eq(dev_net(t->dev), net)) 2241 unregister_netdevice_queue(t->dev, list); 2242 t = rtnl_dereference(t->next); 2243 } 2244 } 2245 2246 static int __net_init ip6_tnl_init_net(struct net *net) 2247 { 2248 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 2249 struct ip6_tnl *t = NULL; 2250 int err; 2251 2252 ip6n->tnls[0] = ip6n->tnls_wc; 2253 ip6n->tnls[1] = ip6n->tnls_r_l; 2254 2255 if (!net_has_fallback_tunnels(net)) 2256 return 0; 2257 err = -ENOMEM; 2258 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 2259 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 2260 2261 if (!ip6n->fb_tnl_dev) 2262 goto err_alloc_dev; 2263 dev_net_set(ip6n->fb_tnl_dev, net); 2264 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 2265 /* FB netdevice is special: we have one, and only one per netns. 2266 * Allowing to move it to another netns is clearly unsafe. 2267 */ 2268 ip6n->fb_tnl_dev->netns_immutable = true; 2269 2270 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 2271 if (err < 0) 2272 goto err_register; 2273 2274 err = register_netdev(ip6n->fb_tnl_dev); 2275 if (err < 0) 2276 goto err_register; 2277 2278 t = netdev_priv(ip6n->fb_tnl_dev); 2279 2280 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 2281 return 0; 2282 2283 err_register: 2284 free_netdev(ip6n->fb_tnl_dev); 2285 err_alloc_dev: 2286 return err; 2287 } 2288 2289 static void __net_exit ip6_tnl_exit_batch_rtnl(struct list_head *net_list, 2290 struct list_head *dev_to_kill) 2291 { 2292 struct net *net; 2293 2294 ASSERT_RTNL(); 2295 list_for_each_entry(net, net_list, exit_list) 2296 ip6_tnl_destroy_tunnels(net, dev_to_kill); 2297 } 2298 2299 static struct pernet_operations ip6_tnl_net_ops = { 2300 .init = ip6_tnl_init_net, 2301 .exit_batch_rtnl = ip6_tnl_exit_batch_rtnl, 2302 .id = &ip6_tnl_net_id, 2303 .size = sizeof(struct ip6_tnl_net), 2304 }; 2305 2306 /** 2307 * ip6_tunnel_init - register protocol and reserve needed resources 2308 * 2309 * Return: 0 on success 2310 **/ 2311 2312 static int __init ip6_tunnel_init(void) 2313 { 2314 int err; 2315 2316 if (!ipv6_mod_enabled()) 2317 return -EOPNOTSUPP; 2318 2319 err = register_pernet_device(&ip6_tnl_net_ops); 2320 if (err < 0) 2321 goto out_pernet; 2322 2323 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 2324 if (err < 0) { 2325 pr_err("%s: can't register ip4ip6\n", __func__); 2326 goto out_ip4ip6; 2327 } 2328 2329 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 2330 if (err < 0) { 2331 pr_err("%s: can't register ip6ip6\n", __func__); 2332 goto out_ip6ip6; 2333 } 2334 2335 if (ip6_tnl_mpls_supported()) { 2336 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS); 2337 if (err < 0) { 2338 pr_err("%s: can't register mplsip6\n", __func__); 2339 goto out_mplsip6; 2340 } 2341 } 2342 2343 err = rtnl_link_register(&ip6_link_ops); 2344 if (err < 0) 2345 goto rtnl_link_failed; 2346 2347 return 0; 2348 2349 rtnl_link_failed: 2350 if (ip6_tnl_mpls_supported()) 2351 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS); 2352 out_mplsip6: 2353 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 2354 out_ip6ip6: 2355 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 2356 out_ip4ip6: 2357 unregister_pernet_device(&ip6_tnl_net_ops); 2358 out_pernet: 2359 return err; 2360 } 2361 2362 /** 2363 * ip6_tunnel_cleanup - free resources and unregister protocol 2364 **/ 2365 2366 static void __exit ip6_tunnel_cleanup(void) 2367 { 2368 rtnl_link_unregister(&ip6_link_ops); 2369 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 2370 pr_info("%s: can't deregister ip4ip6\n", __func__); 2371 2372 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 2373 pr_info("%s: can't deregister ip6ip6\n", __func__); 2374 2375 if (ip6_tnl_mpls_supported() && 2376 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS)) 2377 pr_info("%s: can't deregister mplsip6\n", __func__); 2378 unregister_pernet_device(&ip6_tnl_net_ops); 2379 } 2380 2381 module_init(ip6_tunnel_init); 2382 module_exit(ip6_tunnel_cleanup); 2383