1 /* 2 * IPv6 tunneling device 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Ville Nuorvala <vnuorval@tcs.hut.fi> 7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org> 8 * 9 * Based on: 10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c 11 * 12 * RFC 2473 13 * 14 * This program is free software; you can redistribute it and/or 15 * modify it under the terms of the GNU General Public License 16 * as published by the Free Software Foundation; either version 17 * 2 of the License, or (at your option) any later version. 18 * 19 */ 20 21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 22 23 #include <linux/module.h> 24 #include <linux/capability.h> 25 #include <linux/errno.h> 26 #include <linux/types.h> 27 #include <linux/sockios.h> 28 #include <linux/icmp.h> 29 #include <linux/if.h> 30 #include <linux/in.h> 31 #include <linux/ip.h> 32 #include <linux/net.h> 33 #include <linux/in6.h> 34 #include <linux/netdevice.h> 35 #include <linux/if_arp.h> 36 #include <linux/icmpv6.h> 37 #include <linux/init.h> 38 #include <linux/route.h> 39 #include <linux/rtnetlink.h> 40 #include <linux/netfilter_ipv6.h> 41 #include <linux/slab.h> 42 #include <linux/hash.h> 43 #include <linux/etherdevice.h> 44 45 #include <asm/uaccess.h> 46 #include <linux/atomic.h> 47 48 #include <net/icmp.h> 49 #include <net/ip.h> 50 #include <net/ip_tunnels.h> 51 #include <net/ipv6.h> 52 #include <net/ip6_route.h> 53 #include <net/addrconf.h> 54 #include <net/ip6_tunnel.h> 55 #include <net/xfrm.h> 56 #include <net/dsfield.h> 57 #include <net/inet_ecn.h> 58 #include <net/net_namespace.h> 59 #include <net/netns/generic.h> 60 61 MODULE_AUTHOR("Ville Nuorvala"); 62 MODULE_DESCRIPTION("IPv6 tunneling device"); 63 MODULE_LICENSE("GPL"); 64 MODULE_ALIAS_RTNL_LINK("ip6tnl"); 65 MODULE_ALIAS_NETDEV("ip6tnl0"); 66 67 #ifdef IP6_TNL_DEBUG 68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__) 69 #else 70 #define IP6_TNL_TRACE(x...) do {;} while(0) 71 #endif 72 73 #define HASH_SIZE_SHIFT 5 74 #define HASH_SIZE (1 << HASH_SIZE_SHIFT) 75 76 static bool log_ecn_error = true; 77 module_param(log_ecn_error, bool, 0644); 78 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 79 80 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 81 { 82 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 83 84 return hash_32(hash, HASH_SIZE_SHIFT); 85 } 86 87 static int ip6_tnl_dev_init(struct net_device *dev); 88 static void ip6_tnl_dev_setup(struct net_device *dev); 89 static struct rtnl_link_ops ip6_link_ops __read_mostly; 90 91 static int ip6_tnl_net_id __read_mostly; 92 struct ip6_tnl_net { 93 /* the IPv6 tunnel fallback device */ 94 struct net_device *fb_tnl_dev; 95 /* lists for storing tunnels in use */ 96 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; 97 struct ip6_tnl __rcu *tnls_wc[1]; 98 struct ip6_tnl __rcu **tnls[2]; 99 }; 100 101 static struct net_device_stats *ip6_get_stats(struct net_device *dev) 102 { 103 struct pcpu_sw_netstats tmp, sum = { 0 }; 104 int i; 105 106 for_each_possible_cpu(i) { 107 unsigned int start; 108 const struct pcpu_sw_netstats *tstats = 109 per_cpu_ptr(dev->tstats, i); 110 111 do { 112 start = u64_stats_fetch_begin_irq(&tstats->syncp); 113 tmp.rx_packets = tstats->rx_packets; 114 tmp.rx_bytes = tstats->rx_bytes; 115 tmp.tx_packets = tstats->tx_packets; 116 tmp.tx_bytes = tstats->tx_bytes; 117 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start)); 118 119 sum.rx_packets += tmp.rx_packets; 120 sum.rx_bytes += tmp.rx_bytes; 121 sum.tx_packets += tmp.tx_packets; 122 sum.tx_bytes += tmp.tx_bytes; 123 } 124 dev->stats.rx_packets = sum.rx_packets; 125 dev->stats.rx_bytes = sum.rx_bytes; 126 dev->stats.tx_packets = sum.tx_packets; 127 dev->stats.tx_bytes = sum.tx_bytes; 128 return &dev->stats; 129 } 130 131 /* 132 * Locking : hash tables are protected by RCU and RTNL 133 */ 134 135 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 136 { 137 struct dst_entry *dst = t->dst_cache; 138 139 if (dst && dst->obsolete && 140 dst->ops->check(dst, t->dst_cookie) == NULL) { 141 t->dst_cache = NULL; 142 dst_release(dst); 143 return NULL; 144 } 145 146 return dst; 147 } 148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check); 149 150 void ip6_tnl_dst_reset(struct ip6_tnl *t) 151 { 152 dst_release(t->dst_cache); 153 t->dst_cache = NULL; 154 } 155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset); 156 157 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 158 { 159 struct rt6_info *rt = (struct rt6_info *) dst; 160 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 161 dst_release(t->dst_cache); 162 t->dst_cache = dst; 163 } 164 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store); 165 166 /** 167 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 168 * @remote: the address of the tunnel exit-point 169 * @local: the address of the tunnel entry-point 170 * 171 * Return: 172 * tunnel matching given end-points if found, 173 * else fallback tunnel if its device is up, 174 * else %NULL 175 **/ 176 177 #define for_each_ip6_tunnel_rcu(start) \ 178 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 179 180 static struct ip6_tnl * 181 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local) 182 { 183 unsigned int hash = HASH(remote, local); 184 struct ip6_tnl *t; 185 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 186 187 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 188 if (ipv6_addr_equal(local, &t->parms.laddr) && 189 ipv6_addr_equal(remote, &t->parms.raddr) && 190 (t->dev->flags & IFF_UP)) 191 return t; 192 } 193 t = rcu_dereference(ip6n->tnls_wc[0]); 194 if (t && (t->dev->flags & IFF_UP)) 195 return t; 196 197 return NULL; 198 } 199 200 /** 201 * ip6_tnl_bucket - get head of list matching given tunnel parameters 202 * @p: parameters containing tunnel end-points 203 * 204 * Description: 205 * ip6_tnl_bucket() returns the head of the list matching the 206 * &struct in6_addr entries laddr and raddr in @p. 207 * 208 * Return: head of IPv6 tunnel list 209 **/ 210 211 static struct ip6_tnl __rcu ** 212 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 213 { 214 const struct in6_addr *remote = &p->raddr; 215 const struct in6_addr *local = &p->laddr; 216 unsigned int h = 0; 217 int prio = 0; 218 219 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 220 prio = 1; 221 h = HASH(remote, local); 222 } 223 return &ip6n->tnls[prio][h]; 224 } 225 226 /** 227 * ip6_tnl_link - add tunnel to hash table 228 * @t: tunnel to be added 229 **/ 230 231 static void 232 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 233 { 234 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 235 236 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 237 rcu_assign_pointer(*tp, t); 238 } 239 240 /** 241 * ip6_tnl_unlink - remove tunnel from hash table 242 * @t: tunnel to be removed 243 **/ 244 245 static void 246 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 247 { 248 struct ip6_tnl __rcu **tp; 249 struct ip6_tnl *iter; 250 251 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 252 (iter = rtnl_dereference(*tp)) != NULL; 253 tp = &iter->next) { 254 if (t == iter) { 255 rcu_assign_pointer(*tp, t->next); 256 break; 257 } 258 } 259 } 260 261 static void ip6_dev_free(struct net_device *dev) 262 { 263 free_percpu(dev->tstats); 264 free_netdev(dev); 265 } 266 267 static int ip6_tnl_create2(struct net_device *dev) 268 { 269 struct ip6_tnl *t = netdev_priv(dev); 270 struct net *net = dev_net(dev); 271 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 272 int err; 273 274 t = netdev_priv(dev); 275 err = ip6_tnl_dev_init(dev); 276 if (err < 0) 277 goto out; 278 279 err = register_netdevice(dev); 280 if (err < 0) 281 goto out; 282 283 strcpy(t->parms.name, dev->name); 284 dev->rtnl_link_ops = &ip6_link_ops; 285 286 dev_hold(dev); 287 ip6_tnl_link(ip6n, t); 288 return 0; 289 290 out: 291 return err; 292 } 293 294 /** 295 * ip6_tnl_create - create a new tunnel 296 * @p: tunnel parameters 297 * @pt: pointer to new tunnel 298 * 299 * Description: 300 * Create tunnel matching given parameters. 301 * 302 * Return: 303 * created tunnel or NULL 304 **/ 305 306 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 307 { 308 struct net_device *dev; 309 struct ip6_tnl *t; 310 char name[IFNAMSIZ]; 311 int err; 312 313 if (p->name[0]) 314 strlcpy(name, p->name, IFNAMSIZ); 315 else 316 sprintf(name, "ip6tnl%%d"); 317 318 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup); 319 if (dev == NULL) 320 goto failed; 321 322 dev_net_set(dev, net); 323 324 t = netdev_priv(dev); 325 t->parms = *p; 326 t->net = dev_net(dev); 327 err = ip6_tnl_create2(dev); 328 if (err < 0) 329 goto failed_free; 330 331 return t; 332 333 failed_free: 334 ip6_dev_free(dev); 335 failed: 336 return NULL; 337 } 338 339 /** 340 * ip6_tnl_locate - find or create tunnel matching given parameters 341 * @p: tunnel parameters 342 * @create: != 0 if allowed to create new tunnel if no match found 343 * 344 * Description: 345 * ip6_tnl_locate() first tries to locate an existing tunnel 346 * based on @parms. If this is unsuccessful, but @create is set a new 347 * tunnel device is created and registered for use. 348 * 349 * Return: 350 * matching tunnel or NULL 351 **/ 352 353 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 354 struct __ip6_tnl_parm *p, int create) 355 { 356 const struct in6_addr *remote = &p->raddr; 357 const struct in6_addr *local = &p->laddr; 358 struct ip6_tnl __rcu **tp; 359 struct ip6_tnl *t; 360 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 361 362 for (tp = ip6_tnl_bucket(ip6n, p); 363 (t = rtnl_dereference(*tp)) != NULL; 364 tp = &t->next) { 365 if (ipv6_addr_equal(local, &t->parms.laddr) && 366 ipv6_addr_equal(remote, &t->parms.raddr)) 367 return t; 368 } 369 if (!create) 370 return NULL; 371 return ip6_tnl_create(net, p); 372 } 373 374 /** 375 * ip6_tnl_dev_uninit - tunnel device uninitializer 376 * @dev: the device to be destroyed 377 * 378 * Description: 379 * ip6_tnl_dev_uninit() removes tunnel from its list 380 **/ 381 382 static void 383 ip6_tnl_dev_uninit(struct net_device *dev) 384 { 385 struct ip6_tnl *t = netdev_priv(dev); 386 struct net *net = t->net; 387 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 388 389 if (dev == ip6n->fb_tnl_dev) 390 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 391 else 392 ip6_tnl_unlink(ip6n, t); 393 ip6_tnl_dst_reset(t); 394 dev_put(dev); 395 } 396 397 /** 398 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 399 * @skb: received socket buffer 400 * 401 * Return: 402 * 0 if none was found, 403 * else index to encapsulation limit 404 **/ 405 406 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 407 { 408 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw; 409 __u8 nexthdr = ipv6h->nexthdr; 410 __u16 off = sizeof (*ipv6h); 411 412 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 413 __u16 optlen = 0; 414 struct ipv6_opt_hdr *hdr; 415 if (raw + off + sizeof (*hdr) > skb->data && 416 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 417 break; 418 419 hdr = (struct ipv6_opt_hdr *) (raw + off); 420 if (nexthdr == NEXTHDR_FRAGMENT) { 421 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 422 if (frag_hdr->frag_off) 423 break; 424 optlen = 8; 425 } else if (nexthdr == NEXTHDR_AUTH) { 426 optlen = (hdr->hdrlen + 2) << 2; 427 } else { 428 optlen = ipv6_optlen(hdr); 429 } 430 if (nexthdr == NEXTHDR_DEST) { 431 __u16 i = off + 2; 432 while (1) { 433 struct ipv6_tlv_tnl_enc_lim *tel; 434 435 /* No more room for encapsulation limit */ 436 if (i + sizeof (*tel) > off + optlen) 437 break; 438 439 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[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; 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 /** 459 * ip6_tnl_err - tunnel error handler 460 * 461 * Description: 462 * ip6_tnl_err() should handle errors in the tunnel according 463 * to the specifications in RFC 2473. 464 **/ 465 466 static int 467 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 468 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 469 { 470 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data; 471 struct ip6_tnl *t; 472 int rel_msg = 0; 473 u8 rel_type = ICMPV6_DEST_UNREACH; 474 u8 rel_code = ICMPV6_ADDR_UNREACH; 475 __u32 rel_info = 0; 476 __u16 len; 477 int err = -ENOENT; 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 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 485 &ipv6h->saddr)) == NULL) 486 goto out; 487 488 if (t->parms.proto != ipproto && t->parms.proto != 0) 489 goto out; 490 491 err = 0; 492 493 switch (*type) { 494 __u32 teli; 495 struct ipv6_tlv_tnl_enc_lim *tel; 496 __u32 mtu; 497 case ICMPV6_DEST_UNREACH: 498 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n", 499 t->parms.name); 500 rel_msg = 1; 501 break; 502 case ICMPV6_TIME_EXCEED: 503 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 504 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 505 t->parms.name); 506 rel_msg = 1; 507 } 508 break; 509 case ICMPV6_PARAMPROB: 510 teli = 0; 511 if ((*code) == ICMPV6_HDR_FIELD) 512 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 513 514 if (teli && teli == *info - 2) { 515 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 516 if (tel->encap_limit == 0) { 517 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 518 t->parms.name); 519 rel_msg = 1; 520 } 521 } else { 522 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 523 t->parms.name); 524 } 525 break; 526 case ICMPV6_PKT_TOOBIG: 527 mtu = *info - offset; 528 if (mtu < IPV6_MIN_MTU) 529 mtu = IPV6_MIN_MTU; 530 t->dev->mtu = mtu; 531 532 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 533 rel_type = ICMPV6_PKT_TOOBIG; 534 rel_code = 0; 535 rel_info = mtu; 536 rel_msg = 1; 537 } 538 break; 539 } 540 541 *type = rel_type; 542 *code = rel_code; 543 *info = rel_info; 544 *msg = rel_msg; 545 546 out: 547 rcu_read_unlock(); 548 return err; 549 } 550 551 static int 552 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 553 u8 type, u8 code, int offset, __be32 info) 554 { 555 int rel_msg = 0; 556 u8 rel_type = type; 557 u8 rel_code = code; 558 __u32 rel_info = ntohl(info); 559 int err; 560 struct sk_buff *skb2; 561 const struct iphdr *eiph; 562 struct rtable *rt; 563 struct flowi4 fl4; 564 565 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 566 &rel_msg, &rel_info, offset); 567 if (err < 0) 568 return err; 569 570 if (rel_msg == 0) 571 return 0; 572 573 switch (rel_type) { 574 case ICMPV6_DEST_UNREACH: 575 if (rel_code != ICMPV6_ADDR_UNREACH) 576 return 0; 577 rel_type = ICMP_DEST_UNREACH; 578 rel_code = ICMP_HOST_UNREACH; 579 break; 580 case ICMPV6_PKT_TOOBIG: 581 if (rel_code != 0) 582 return 0; 583 rel_type = ICMP_DEST_UNREACH; 584 rel_code = ICMP_FRAG_NEEDED; 585 break; 586 case NDISC_REDIRECT: 587 rel_type = ICMP_REDIRECT; 588 rel_code = ICMP_REDIR_HOST; 589 default: 590 return 0; 591 } 592 593 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 594 return 0; 595 596 skb2 = skb_clone(skb, GFP_ATOMIC); 597 if (!skb2) 598 return 0; 599 600 skb_dst_drop(skb2); 601 602 skb_pull(skb2, offset); 603 skb_reset_network_header(skb2); 604 eiph = ip_hdr(skb2); 605 606 /* Try to guess incoming interface */ 607 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 608 eiph->saddr, 0, 609 0, 0, 610 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 611 if (IS_ERR(rt)) 612 goto out; 613 614 skb2->dev = rt->dst.dev; 615 616 /* route "incoming" packet */ 617 if (rt->rt_flags & RTCF_LOCAL) { 618 ip_rt_put(rt); 619 rt = NULL; 620 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 621 eiph->daddr, eiph->saddr, 622 0, 0, 623 IPPROTO_IPIP, 624 RT_TOS(eiph->tos), 0); 625 if (IS_ERR(rt) || 626 rt->dst.dev->type != ARPHRD_TUNNEL) { 627 if (!IS_ERR(rt)) 628 ip_rt_put(rt); 629 goto out; 630 } 631 skb_dst_set(skb2, &rt->dst); 632 } else { 633 ip_rt_put(rt); 634 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 635 skb2->dev) || 636 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 637 goto out; 638 } 639 640 /* change mtu on this route */ 641 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 642 if (rel_info > dst_mtu(skb_dst(skb2))) 643 goto out; 644 645 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info); 646 } 647 if (rel_type == ICMP_REDIRECT) 648 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2); 649 650 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 651 652 out: 653 kfree_skb(skb2); 654 return 0; 655 } 656 657 static int 658 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 659 u8 type, u8 code, int offset, __be32 info) 660 { 661 int rel_msg = 0; 662 u8 rel_type = type; 663 u8 rel_code = code; 664 __u32 rel_info = ntohl(info); 665 int err; 666 667 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 668 &rel_msg, &rel_info, offset); 669 if (err < 0) 670 return err; 671 672 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 673 struct rt6_info *rt; 674 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 675 676 if (!skb2) 677 return 0; 678 679 skb_dst_drop(skb2); 680 skb_pull(skb2, offset); 681 skb_reset_network_header(skb2); 682 683 /* Try to guess incoming interface */ 684 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 685 NULL, 0, 0); 686 687 if (rt && rt->dst.dev) 688 skb2->dev = rt->dst.dev; 689 690 icmpv6_send(skb2, rel_type, rel_code, rel_info); 691 692 ip6_rt_put(rt); 693 694 kfree_skb(skb2); 695 } 696 697 return 0; 698 } 699 700 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 701 const struct ipv6hdr *ipv6h, 702 struct sk_buff *skb) 703 { 704 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 705 706 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 707 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 708 709 return IP6_ECN_decapsulate(ipv6h, skb); 710 } 711 712 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 713 const struct ipv6hdr *ipv6h, 714 struct sk_buff *skb) 715 { 716 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 717 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 718 719 return IP6_ECN_decapsulate(ipv6h, skb); 720 } 721 722 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 723 const struct in6_addr *laddr, 724 const struct in6_addr *raddr) 725 { 726 struct __ip6_tnl_parm *p = &t->parms; 727 int ltype = ipv6_addr_type(laddr); 728 int rtype = ipv6_addr_type(raddr); 729 __u32 flags = 0; 730 731 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 732 flags = IP6_TNL_F_CAP_PER_PACKET; 733 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 734 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 735 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 736 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 737 if (ltype&IPV6_ADDR_UNICAST) 738 flags |= IP6_TNL_F_CAP_XMIT; 739 if (rtype&IPV6_ADDR_UNICAST) 740 flags |= IP6_TNL_F_CAP_RCV; 741 } 742 return flags; 743 } 744 EXPORT_SYMBOL(ip6_tnl_get_cap); 745 746 /* called with rcu_read_lock() */ 747 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 748 const struct in6_addr *laddr, 749 const struct in6_addr *raddr) 750 { 751 struct __ip6_tnl_parm *p = &t->parms; 752 int ret = 0; 753 struct net *net = t->net; 754 755 if ((p->flags & IP6_TNL_F_CAP_RCV) || 756 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 757 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 758 struct net_device *ldev = NULL; 759 760 if (p->link) 761 ldev = dev_get_by_index_rcu(net, p->link); 762 763 if ((ipv6_addr_is_multicast(laddr) || 764 likely(ipv6_chk_addr(net, laddr, ldev, 0))) && 765 likely(!ipv6_chk_addr(net, raddr, NULL, 0))) 766 ret = 1; 767 } 768 return ret; 769 } 770 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 771 772 /** 773 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 774 * @skb: received socket buffer 775 * @protocol: ethernet protocol ID 776 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 777 * 778 * Return: 0 779 **/ 780 781 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 782 __u8 ipproto, 783 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 784 const struct ipv6hdr *ipv6h, 785 struct sk_buff *skb)) 786 { 787 struct ip6_tnl *t; 788 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 789 int err; 790 791 rcu_read_lock(); 792 793 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 794 &ipv6h->daddr)) != NULL) { 795 struct pcpu_sw_netstats *tstats; 796 797 if (t->parms.proto != ipproto && t->parms.proto != 0) { 798 rcu_read_unlock(); 799 goto discard; 800 } 801 802 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 803 rcu_read_unlock(); 804 goto discard; 805 } 806 807 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 808 t->dev->stats.rx_dropped++; 809 rcu_read_unlock(); 810 goto discard; 811 } 812 skb->mac_header = skb->network_header; 813 skb_reset_network_header(skb); 814 skb->protocol = htons(protocol); 815 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 816 817 __skb_tunnel_rx(skb, t->dev, t->net); 818 819 err = dscp_ecn_decapsulate(t, ipv6h, skb); 820 if (unlikely(err)) { 821 if (log_ecn_error) 822 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n", 823 &ipv6h->saddr, 824 ipv6_get_dsfield(ipv6h)); 825 if (err > 1) { 826 ++t->dev->stats.rx_frame_errors; 827 ++t->dev->stats.rx_errors; 828 rcu_read_unlock(); 829 goto discard; 830 } 831 } 832 833 tstats = this_cpu_ptr(t->dev->tstats); 834 u64_stats_update_begin(&tstats->syncp); 835 tstats->rx_packets++; 836 tstats->rx_bytes += skb->len; 837 u64_stats_update_end(&tstats->syncp); 838 839 netif_rx(skb); 840 841 rcu_read_unlock(); 842 return 0; 843 } 844 rcu_read_unlock(); 845 return 1; 846 847 discard: 848 kfree_skb(skb); 849 return 0; 850 } 851 852 static int ip4ip6_rcv(struct sk_buff *skb) 853 { 854 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 855 ip4ip6_dscp_ecn_decapsulate); 856 } 857 858 static int ip6ip6_rcv(struct sk_buff *skb) 859 { 860 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 861 ip6ip6_dscp_ecn_decapsulate); 862 } 863 864 struct ipv6_tel_txoption { 865 struct ipv6_txoptions ops; 866 __u8 dst_opt[8]; 867 }; 868 869 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 870 { 871 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 872 873 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 874 opt->dst_opt[3] = 1; 875 opt->dst_opt[4] = encap_limit; 876 opt->dst_opt[5] = IPV6_TLV_PADN; 877 opt->dst_opt[6] = 1; 878 879 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 880 opt->ops.opt_nflen = 8; 881 } 882 883 /** 884 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 885 * @t: the outgoing tunnel device 886 * @hdr: IPv6 header from the incoming packet 887 * 888 * Description: 889 * Avoid trivial tunneling loop by checking that tunnel exit-point 890 * doesn't match source of incoming packet. 891 * 892 * Return: 893 * 1 if conflict, 894 * 0 else 895 **/ 896 897 static inline bool 898 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 899 { 900 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 901 } 902 903 int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 904 { 905 struct __ip6_tnl_parm *p = &t->parms; 906 int ret = 0; 907 struct net *net = t->net; 908 909 if (p->flags & IP6_TNL_F_CAP_XMIT) { 910 struct net_device *ldev = NULL; 911 912 rcu_read_lock(); 913 if (p->link) 914 ldev = dev_get_by_index_rcu(net, p->link); 915 916 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 917 pr_warn("%s xmit: Local address not yet configured!\n", 918 p->name); 919 else if (!ipv6_addr_is_multicast(&p->raddr) && 920 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 921 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 922 p->name); 923 else 924 ret = 1; 925 rcu_read_unlock(); 926 } 927 return ret; 928 } 929 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 930 931 /** 932 * ip6_tnl_xmit2 - encapsulate packet and send 933 * @skb: the outgoing socket buffer 934 * @dev: the outgoing tunnel device 935 * @dsfield: dscp code for outer header 936 * @fl: flow of tunneled packet 937 * @encap_limit: encapsulation limit 938 * @pmtu: Path MTU is stored if packet is too big 939 * 940 * Description: 941 * Build new header and do some sanity checks on the packet before sending 942 * it. 943 * 944 * Return: 945 * 0 on success 946 * -1 fail 947 * %-EMSGSIZE message too big. return mtu in this case. 948 **/ 949 950 static int ip6_tnl_xmit2(struct sk_buff *skb, 951 struct net_device *dev, 952 __u8 dsfield, 953 struct flowi6 *fl6, 954 int encap_limit, 955 __u32 *pmtu) 956 { 957 struct ip6_tnl *t = netdev_priv(dev); 958 struct net *net = t->net; 959 struct net_device_stats *stats = &t->dev->stats; 960 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 961 struct ipv6_tel_txoption opt; 962 struct dst_entry *dst = NULL, *ndst = NULL; 963 struct net_device *tdev; 964 int mtu; 965 unsigned int max_headroom = sizeof(struct ipv6hdr); 966 u8 proto; 967 int err = -1; 968 969 if (!fl6->flowi6_mark) 970 dst = ip6_tnl_dst_check(t); 971 if (!dst) { 972 ndst = ip6_route_output(net, NULL, fl6); 973 974 if (ndst->error) 975 goto tx_err_link_failure; 976 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0); 977 if (IS_ERR(ndst)) { 978 err = PTR_ERR(ndst); 979 ndst = NULL; 980 goto tx_err_link_failure; 981 } 982 dst = ndst; 983 } 984 985 tdev = dst->dev; 986 987 if (tdev == dev) { 988 stats->collisions++; 989 net_warn_ratelimited("%s: Local routing loop detected!\n", 990 t->parms.name); 991 goto tx_err_dst_release; 992 } 993 mtu = dst_mtu(dst) - sizeof (*ipv6h); 994 if (encap_limit >= 0) { 995 max_headroom += 8; 996 mtu -= 8; 997 } 998 if (mtu < IPV6_MIN_MTU) 999 mtu = IPV6_MIN_MTU; 1000 if (skb_dst(skb)) 1001 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 1002 if (skb->len > mtu) { 1003 *pmtu = mtu; 1004 err = -EMSGSIZE; 1005 goto tx_err_dst_release; 1006 } 1007 1008 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1009 1010 /* 1011 * Okay, now see if we can stuff it in the buffer as-is. 1012 */ 1013 max_headroom += LL_RESERVED_SPACE(tdev); 1014 1015 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1016 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1017 struct sk_buff *new_skb; 1018 1019 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 1020 goto tx_err_dst_release; 1021 1022 if (skb->sk) 1023 skb_set_owner_w(new_skb, skb->sk); 1024 consume_skb(skb); 1025 skb = new_skb; 1026 } 1027 if (fl6->flowi6_mark) { 1028 skb_dst_set(skb, dst); 1029 ndst = NULL; 1030 } else { 1031 skb_dst_set_noref(skb, dst); 1032 } 1033 skb->transport_header = skb->network_header; 1034 1035 proto = fl6->flowi6_proto; 1036 if (encap_limit >= 0) { 1037 init_tel_txopt(&opt, encap_limit); 1038 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 1039 } 1040 1041 if (likely(!skb->encapsulation)) { 1042 skb_reset_inner_headers(skb); 1043 skb->encapsulation = 1; 1044 } 1045 1046 skb_push(skb, sizeof(struct ipv6hdr)); 1047 skb_reset_network_header(skb); 1048 ipv6h = ipv6_hdr(skb); 1049 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel); 1050 ipv6h->hop_limit = t->parms.hop_limit; 1051 ipv6h->nexthdr = proto; 1052 ipv6h->saddr = fl6->saddr; 1053 ipv6h->daddr = fl6->daddr; 1054 ip6tunnel_xmit(skb, dev); 1055 if (ndst) 1056 ip6_tnl_dst_store(t, ndst); 1057 return 0; 1058 tx_err_link_failure: 1059 stats->tx_carrier_errors++; 1060 dst_link_failure(skb); 1061 tx_err_dst_release: 1062 dst_release(ndst); 1063 return err; 1064 } 1065 1066 static inline int 1067 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1068 { 1069 struct ip6_tnl *t = netdev_priv(dev); 1070 const struct iphdr *iph = ip_hdr(skb); 1071 int encap_limit = -1; 1072 struct flowi6 fl6; 1073 __u8 dsfield; 1074 __u32 mtu; 1075 int err; 1076 1077 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 1078 !ip6_tnl_xmit_ctl(t)) 1079 return -1; 1080 1081 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1082 encap_limit = t->parms.encap_limit; 1083 1084 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1085 fl6.flowi6_proto = IPPROTO_IPIP; 1086 1087 dsfield = ipv4_get_dsfield(iph); 1088 1089 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1090 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 1091 & IPV6_TCLASS_MASK; 1092 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1093 fl6.flowi6_mark = skb->mark; 1094 1095 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1096 if (err != 0) { 1097 /* XXX: send ICMP error even if DF is not set. */ 1098 if (err == -EMSGSIZE) 1099 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1100 htonl(mtu)); 1101 return -1; 1102 } 1103 1104 return 0; 1105 } 1106 1107 static inline int 1108 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1109 { 1110 struct ip6_tnl *t = netdev_priv(dev); 1111 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1112 int encap_limit = -1; 1113 __u16 offset; 1114 struct flowi6 fl6; 1115 __u8 dsfield; 1116 __u32 mtu; 1117 int err; 1118 1119 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1120 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1121 return -1; 1122 1123 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 1124 if (offset > 0) { 1125 struct ipv6_tlv_tnl_enc_lim *tel; 1126 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1127 if (tel->encap_limit == 0) { 1128 icmpv6_send(skb, ICMPV6_PARAMPROB, 1129 ICMPV6_HDR_FIELD, offset + 2); 1130 return -1; 1131 } 1132 encap_limit = tel->encap_limit - 1; 1133 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1134 encap_limit = t->parms.encap_limit; 1135 1136 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1137 fl6.flowi6_proto = IPPROTO_IPV6; 1138 1139 dsfield = ipv6_get_dsfield(ipv6h); 1140 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1141 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1142 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1143 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1144 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1145 fl6.flowi6_mark = skb->mark; 1146 1147 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1148 if (err != 0) { 1149 if (err == -EMSGSIZE) 1150 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1151 return -1; 1152 } 1153 1154 return 0; 1155 } 1156 1157 static netdev_tx_t 1158 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1159 { 1160 struct ip6_tnl *t = netdev_priv(dev); 1161 struct net_device_stats *stats = &t->dev->stats; 1162 int ret; 1163 1164 switch (skb->protocol) { 1165 case htons(ETH_P_IP): 1166 ret = ip4ip6_tnl_xmit(skb, dev); 1167 break; 1168 case htons(ETH_P_IPV6): 1169 ret = ip6ip6_tnl_xmit(skb, dev); 1170 break; 1171 default: 1172 goto tx_err; 1173 } 1174 1175 if (ret < 0) 1176 goto tx_err; 1177 1178 return NETDEV_TX_OK; 1179 1180 tx_err: 1181 stats->tx_errors++; 1182 stats->tx_dropped++; 1183 kfree_skb(skb); 1184 return NETDEV_TX_OK; 1185 } 1186 1187 static void ip6_tnl_link_config(struct ip6_tnl *t) 1188 { 1189 struct net_device *dev = t->dev; 1190 struct __ip6_tnl_parm *p = &t->parms; 1191 struct flowi6 *fl6 = &t->fl.u.ip6; 1192 1193 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1194 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1195 1196 /* Set up flowi template */ 1197 fl6->saddr = p->laddr; 1198 fl6->daddr = p->raddr; 1199 fl6->flowi6_oif = p->link; 1200 fl6->flowlabel = 0; 1201 1202 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1203 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1204 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1205 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1206 1207 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1208 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1209 1210 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1211 dev->flags |= IFF_POINTOPOINT; 1212 else 1213 dev->flags &= ~IFF_POINTOPOINT; 1214 1215 dev->iflink = p->link; 1216 1217 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1218 int strict = (ipv6_addr_type(&p->raddr) & 1219 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1220 1221 struct rt6_info *rt = rt6_lookup(t->net, 1222 &p->raddr, &p->laddr, 1223 p->link, strict); 1224 1225 if (rt == NULL) 1226 return; 1227 1228 if (rt->dst.dev) { 1229 dev->hard_header_len = rt->dst.dev->hard_header_len + 1230 sizeof (struct ipv6hdr); 1231 1232 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr); 1233 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1234 dev->mtu-=8; 1235 1236 if (dev->mtu < IPV6_MIN_MTU) 1237 dev->mtu = IPV6_MIN_MTU; 1238 } 1239 ip6_rt_put(rt); 1240 } 1241 } 1242 1243 /** 1244 * ip6_tnl_change - update the tunnel parameters 1245 * @t: tunnel to be changed 1246 * @p: tunnel configuration parameters 1247 * 1248 * Description: 1249 * ip6_tnl_change() updates the tunnel parameters 1250 **/ 1251 1252 static int 1253 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1254 { 1255 t->parms.laddr = p->laddr; 1256 t->parms.raddr = p->raddr; 1257 t->parms.flags = p->flags; 1258 t->parms.hop_limit = p->hop_limit; 1259 t->parms.encap_limit = p->encap_limit; 1260 t->parms.flowinfo = p->flowinfo; 1261 t->parms.link = p->link; 1262 t->parms.proto = p->proto; 1263 ip6_tnl_dst_reset(t); 1264 ip6_tnl_link_config(t); 1265 return 0; 1266 } 1267 1268 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1269 { 1270 struct net *net = t->net; 1271 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1272 int err; 1273 1274 ip6_tnl_unlink(ip6n, t); 1275 synchronize_net(); 1276 err = ip6_tnl_change(t, p); 1277 ip6_tnl_link(ip6n, t); 1278 netdev_state_change(t->dev); 1279 return err; 1280 } 1281 1282 static void 1283 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1284 { 1285 p->laddr = u->laddr; 1286 p->raddr = u->raddr; 1287 p->flags = u->flags; 1288 p->hop_limit = u->hop_limit; 1289 p->encap_limit = u->encap_limit; 1290 p->flowinfo = u->flowinfo; 1291 p->link = u->link; 1292 p->proto = u->proto; 1293 memcpy(p->name, u->name, sizeof(u->name)); 1294 } 1295 1296 static void 1297 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1298 { 1299 u->laddr = p->laddr; 1300 u->raddr = p->raddr; 1301 u->flags = p->flags; 1302 u->hop_limit = p->hop_limit; 1303 u->encap_limit = p->encap_limit; 1304 u->flowinfo = p->flowinfo; 1305 u->link = p->link; 1306 u->proto = p->proto; 1307 memcpy(u->name, p->name, sizeof(u->name)); 1308 } 1309 1310 /** 1311 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1312 * @dev: virtual device associated with tunnel 1313 * @ifr: parameters passed from userspace 1314 * @cmd: command to be performed 1315 * 1316 * Description: 1317 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1318 * from userspace. 1319 * 1320 * The possible commands are the following: 1321 * %SIOCGETTUNNEL: get tunnel parameters for device 1322 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1323 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1324 * %SIOCDELTUNNEL: delete tunnel 1325 * 1326 * The fallback device "ip6tnl0", created during module 1327 * initialization, can be used for creating other tunnel devices. 1328 * 1329 * Return: 1330 * 0 on success, 1331 * %-EFAULT if unable to copy data to or from userspace, 1332 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1333 * %-EINVAL if passed tunnel parameters are invalid, 1334 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1335 * %-ENODEV if attempting to change or delete a nonexisting device 1336 **/ 1337 1338 static int 1339 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1340 { 1341 int err = 0; 1342 struct ip6_tnl_parm p; 1343 struct __ip6_tnl_parm p1; 1344 struct ip6_tnl *t = netdev_priv(dev); 1345 struct net *net = t->net; 1346 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1347 1348 switch (cmd) { 1349 case SIOCGETTUNNEL: 1350 if (dev == ip6n->fb_tnl_dev) { 1351 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1352 err = -EFAULT; 1353 break; 1354 } 1355 ip6_tnl_parm_from_user(&p1, &p); 1356 t = ip6_tnl_locate(net, &p1, 0); 1357 if (t == NULL) 1358 t = netdev_priv(dev); 1359 } else { 1360 memset(&p, 0, sizeof(p)); 1361 } 1362 ip6_tnl_parm_to_user(&p, &t->parms); 1363 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1364 err = -EFAULT; 1365 } 1366 break; 1367 case SIOCADDTUNNEL: 1368 case SIOCCHGTUNNEL: 1369 err = -EPERM; 1370 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1371 break; 1372 err = -EFAULT; 1373 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1374 break; 1375 err = -EINVAL; 1376 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1377 p.proto != 0) 1378 break; 1379 ip6_tnl_parm_from_user(&p1, &p); 1380 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1381 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1382 if (t != NULL) { 1383 if (t->dev != dev) { 1384 err = -EEXIST; 1385 break; 1386 } 1387 } else 1388 t = netdev_priv(dev); 1389 1390 err = ip6_tnl_update(t, &p1); 1391 } 1392 if (t) { 1393 err = 0; 1394 ip6_tnl_parm_to_user(&p, &t->parms); 1395 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1396 err = -EFAULT; 1397 1398 } else 1399 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1400 break; 1401 case SIOCDELTUNNEL: 1402 err = -EPERM; 1403 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1404 break; 1405 1406 if (dev == ip6n->fb_tnl_dev) { 1407 err = -EFAULT; 1408 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1409 break; 1410 err = -ENOENT; 1411 ip6_tnl_parm_from_user(&p1, &p); 1412 t = ip6_tnl_locate(net, &p1, 0); 1413 if (t == NULL) 1414 break; 1415 err = -EPERM; 1416 if (t->dev == ip6n->fb_tnl_dev) 1417 break; 1418 dev = t->dev; 1419 } 1420 err = 0; 1421 unregister_netdevice(dev); 1422 break; 1423 default: 1424 err = -EINVAL; 1425 } 1426 return err; 1427 } 1428 1429 /** 1430 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1431 * @dev: virtual device associated with tunnel 1432 * @new_mtu: the new mtu 1433 * 1434 * Return: 1435 * 0 on success, 1436 * %-EINVAL if mtu too small 1437 **/ 1438 1439 static int 1440 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1441 { 1442 struct ip6_tnl *tnl = netdev_priv(dev); 1443 1444 if (tnl->parms.proto == IPPROTO_IPIP) { 1445 if (new_mtu < 68) 1446 return -EINVAL; 1447 } else { 1448 if (new_mtu < IPV6_MIN_MTU) 1449 return -EINVAL; 1450 } 1451 if (new_mtu > 0xFFF8 - dev->hard_header_len) 1452 return -EINVAL; 1453 dev->mtu = new_mtu; 1454 return 0; 1455 } 1456 1457 1458 static const struct net_device_ops ip6_tnl_netdev_ops = { 1459 .ndo_uninit = ip6_tnl_dev_uninit, 1460 .ndo_start_xmit = ip6_tnl_xmit, 1461 .ndo_do_ioctl = ip6_tnl_ioctl, 1462 .ndo_change_mtu = ip6_tnl_change_mtu, 1463 .ndo_get_stats = ip6_get_stats, 1464 }; 1465 1466 1467 /** 1468 * ip6_tnl_dev_setup - setup virtual tunnel device 1469 * @dev: virtual device associated with tunnel 1470 * 1471 * Description: 1472 * Initialize function pointers and device parameters 1473 **/ 1474 1475 static void ip6_tnl_dev_setup(struct net_device *dev) 1476 { 1477 struct ip6_tnl *t; 1478 1479 dev->netdev_ops = &ip6_tnl_netdev_ops; 1480 dev->destructor = ip6_dev_free; 1481 1482 dev->type = ARPHRD_TUNNEL6; 1483 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1484 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1485 t = netdev_priv(dev); 1486 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1487 dev->mtu-=8; 1488 dev->flags |= IFF_NOARP; 1489 dev->addr_len = sizeof(struct in6_addr); 1490 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1491 /* This perm addr will be used as interface identifier by IPv6 */ 1492 dev->addr_assign_type = NET_ADDR_RANDOM; 1493 eth_random_addr(dev->perm_addr); 1494 } 1495 1496 1497 /** 1498 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1499 * @dev: virtual device associated with tunnel 1500 **/ 1501 1502 static inline int 1503 ip6_tnl_dev_init_gen(struct net_device *dev) 1504 { 1505 struct ip6_tnl *t = netdev_priv(dev); 1506 1507 t->dev = dev; 1508 t->net = dev_net(dev); 1509 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1510 if (!dev->tstats) 1511 return -ENOMEM; 1512 return 0; 1513 } 1514 1515 /** 1516 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1517 * @dev: virtual device associated with tunnel 1518 **/ 1519 1520 static int ip6_tnl_dev_init(struct net_device *dev) 1521 { 1522 struct ip6_tnl *t = netdev_priv(dev); 1523 int err = ip6_tnl_dev_init_gen(dev); 1524 1525 if (err) 1526 return err; 1527 ip6_tnl_link_config(t); 1528 return 0; 1529 } 1530 1531 /** 1532 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1533 * @dev: fallback device 1534 * 1535 * Return: 0 1536 **/ 1537 1538 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1539 { 1540 struct ip6_tnl *t = netdev_priv(dev); 1541 struct net *net = dev_net(dev); 1542 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1543 int err = ip6_tnl_dev_init_gen(dev); 1544 1545 if (err) 1546 return err; 1547 1548 t->parms.proto = IPPROTO_IPV6; 1549 dev_hold(dev); 1550 1551 ip6_tnl_link_config(t); 1552 1553 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1554 return 0; 1555 } 1556 1557 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[]) 1558 { 1559 u8 proto; 1560 1561 if (!data || !data[IFLA_IPTUN_PROTO]) 1562 return 0; 1563 1564 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1565 if (proto != IPPROTO_IPV6 && 1566 proto != IPPROTO_IPIP && 1567 proto != 0) 1568 return -EINVAL; 1569 1570 return 0; 1571 } 1572 1573 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1574 struct __ip6_tnl_parm *parms) 1575 { 1576 memset(parms, 0, sizeof(*parms)); 1577 1578 if (!data) 1579 return; 1580 1581 if (data[IFLA_IPTUN_LINK]) 1582 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1583 1584 if (data[IFLA_IPTUN_LOCAL]) 1585 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL], 1586 sizeof(struct in6_addr)); 1587 1588 if (data[IFLA_IPTUN_REMOTE]) 1589 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE], 1590 sizeof(struct in6_addr)); 1591 1592 if (data[IFLA_IPTUN_TTL]) 1593 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1594 1595 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1596 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1597 1598 if (data[IFLA_IPTUN_FLOWINFO]) 1599 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1600 1601 if (data[IFLA_IPTUN_FLAGS]) 1602 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1603 1604 if (data[IFLA_IPTUN_PROTO]) 1605 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1606 } 1607 1608 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 1609 struct nlattr *tb[], struct nlattr *data[]) 1610 { 1611 struct net *net = dev_net(dev); 1612 struct ip6_tnl *nt; 1613 1614 nt = netdev_priv(dev); 1615 ip6_tnl_netlink_parms(data, &nt->parms); 1616 1617 if (ip6_tnl_locate(net, &nt->parms, 0)) 1618 return -EEXIST; 1619 1620 return ip6_tnl_create2(dev); 1621 } 1622 1623 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 1624 struct nlattr *data[]) 1625 { 1626 struct ip6_tnl *t = netdev_priv(dev); 1627 struct __ip6_tnl_parm p; 1628 struct net *net = t->net; 1629 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1630 1631 if (dev == ip6n->fb_tnl_dev) 1632 return -EINVAL; 1633 1634 ip6_tnl_netlink_parms(data, &p); 1635 1636 t = ip6_tnl_locate(net, &p, 0); 1637 1638 if (t) { 1639 if (t->dev != dev) 1640 return -EEXIST; 1641 } else 1642 t = netdev_priv(dev); 1643 1644 return ip6_tnl_update(t, &p); 1645 } 1646 1647 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 1648 { 1649 struct net *net = dev_net(dev); 1650 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1651 1652 if (dev != ip6n->fb_tnl_dev) 1653 unregister_netdevice_queue(dev, head); 1654 } 1655 1656 static size_t ip6_tnl_get_size(const struct net_device *dev) 1657 { 1658 return 1659 /* IFLA_IPTUN_LINK */ 1660 nla_total_size(4) + 1661 /* IFLA_IPTUN_LOCAL */ 1662 nla_total_size(sizeof(struct in6_addr)) + 1663 /* IFLA_IPTUN_REMOTE */ 1664 nla_total_size(sizeof(struct in6_addr)) + 1665 /* IFLA_IPTUN_TTL */ 1666 nla_total_size(1) + 1667 /* IFLA_IPTUN_ENCAP_LIMIT */ 1668 nla_total_size(1) + 1669 /* IFLA_IPTUN_FLOWINFO */ 1670 nla_total_size(4) + 1671 /* IFLA_IPTUN_FLAGS */ 1672 nla_total_size(4) + 1673 /* IFLA_IPTUN_PROTO */ 1674 nla_total_size(1) + 1675 0; 1676 } 1677 1678 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 1679 { 1680 struct ip6_tnl *tunnel = netdev_priv(dev); 1681 struct __ip6_tnl_parm *parm = &tunnel->parms; 1682 1683 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1684 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr), 1685 &parm->laddr) || 1686 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr), 1687 &parm->raddr) || 1688 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 1689 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 1690 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 1691 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 1692 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto)) 1693 goto nla_put_failure; 1694 return 0; 1695 1696 nla_put_failure: 1697 return -EMSGSIZE; 1698 } 1699 1700 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 1701 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1702 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 1703 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 1704 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1705 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 1706 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 1707 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 1708 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1709 }; 1710 1711 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 1712 .kind = "ip6tnl", 1713 .maxtype = IFLA_IPTUN_MAX, 1714 .policy = ip6_tnl_policy, 1715 .priv_size = sizeof(struct ip6_tnl), 1716 .setup = ip6_tnl_dev_setup, 1717 .validate = ip6_tnl_validate, 1718 .newlink = ip6_tnl_newlink, 1719 .changelink = ip6_tnl_changelink, 1720 .dellink = ip6_tnl_dellink, 1721 .get_size = ip6_tnl_get_size, 1722 .fill_info = ip6_tnl_fill_info, 1723 }; 1724 1725 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 1726 .handler = ip4ip6_rcv, 1727 .err_handler = ip4ip6_err, 1728 .priority = 1, 1729 }; 1730 1731 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 1732 .handler = ip6ip6_rcv, 1733 .err_handler = ip6ip6_err, 1734 .priority = 1, 1735 }; 1736 1737 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net) 1738 { 1739 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1740 struct net_device *dev, *aux; 1741 int h; 1742 struct ip6_tnl *t; 1743 LIST_HEAD(list); 1744 1745 for_each_netdev_safe(net, dev, aux) 1746 if (dev->rtnl_link_ops == &ip6_link_ops) 1747 unregister_netdevice_queue(dev, &list); 1748 1749 for (h = 0; h < HASH_SIZE; h++) { 1750 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1751 while (t != NULL) { 1752 /* If dev is in the same netns, it has already 1753 * been added to the list by the previous loop. 1754 */ 1755 if (!net_eq(dev_net(t->dev), net)) 1756 unregister_netdevice_queue(t->dev, &list); 1757 t = rtnl_dereference(t->next); 1758 } 1759 } 1760 1761 unregister_netdevice_many(&list); 1762 } 1763 1764 static int __net_init ip6_tnl_init_net(struct net *net) 1765 { 1766 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1767 struct ip6_tnl *t = NULL; 1768 int err; 1769 1770 ip6n->tnls[0] = ip6n->tnls_wc; 1771 ip6n->tnls[1] = ip6n->tnls_r_l; 1772 1773 err = -ENOMEM; 1774 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1775 ip6_tnl_dev_setup); 1776 1777 if (!ip6n->fb_tnl_dev) 1778 goto err_alloc_dev; 1779 dev_net_set(ip6n->fb_tnl_dev, net); 1780 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 1781 /* FB netdevice is special: we have one, and only one per netns. 1782 * Allowing to move it to another netns is clearly unsafe. 1783 */ 1784 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 1785 1786 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1787 if (err < 0) 1788 goto err_register; 1789 1790 err = register_netdev(ip6n->fb_tnl_dev); 1791 if (err < 0) 1792 goto err_register; 1793 1794 t = netdev_priv(ip6n->fb_tnl_dev); 1795 1796 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1797 return 0; 1798 1799 err_register: 1800 ip6_dev_free(ip6n->fb_tnl_dev); 1801 err_alloc_dev: 1802 return err; 1803 } 1804 1805 static void __net_exit ip6_tnl_exit_net(struct net *net) 1806 { 1807 rtnl_lock(); 1808 ip6_tnl_destroy_tunnels(net); 1809 rtnl_unlock(); 1810 } 1811 1812 static struct pernet_operations ip6_tnl_net_ops = { 1813 .init = ip6_tnl_init_net, 1814 .exit = ip6_tnl_exit_net, 1815 .id = &ip6_tnl_net_id, 1816 .size = sizeof(struct ip6_tnl_net), 1817 }; 1818 1819 /** 1820 * ip6_tunnel_init - register protocol and reserve needed resources 1821 * 1822 * Return: 0 on success 1823 **/ 1824 1825 static int __init ip6_tunnel_init(void) 1826 { 1827 int err; 1828 1829 err = register_pernet_device(&ip6_tnl_net_ops); 1830 if (err < 0) 1831 goto out_pernet; 1832 1833 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 1834 if (err < 0) { 1835 pr_err("%s: can't register ip4ip6\n", __func__); 1836 goto out_ip4ip6; 1837 } 1838 1839 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 1840 if (err < 0) { 1841 pr_err("%s: can't register ip6ip6\n", __func__); 1842 goto out_ip6ip6; 1843 } 1844 err = rtnl_link_register(&ip6_link_ops); 1845 if (err < 0) 1846 goto rtnl_link_failed; 1847 1848 return 0; 1849 1850 rtnl_link_failed: 1851 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 1852 out_ip6ip6: 1853 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1854 out_ip4ip6: 1855 unregister_pernet_device(&ip6_tnl_net_ops); 1856 out_pernet: 1857 return err; 1858 } 1859 1860 /** 1861 * ip6_tunnel_cleanup - free resources and unregister protocol 1862 **/ 1863 1864 static void __exit ip6_tunnel_cleanup(void) 1865 { 1866 rtnl_link_unregister(&ip6_link_ops); 1867 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1868 pr_info("%s: can't deregister ip4ip6\n", __func__); 1869 1870 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1871 pr_info("%s: can't deregister ip6ip6\n", __func__); 1872 1873 unregister_pernet_device(&ip6_tnl_net_ops); 1874 } 1875 1876 module_init(ip6_tunnel_init); 1877 module_exit(ip6_tunnel_cleanup); 1878