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/if_tunnel.h> 33 #include <linux/net.h> 34 #include <linux/in6.h> 35 #include <linux/netdevice.h> 36 #include <linux/if_arp.h> 37 #include <linux/icmpv6.h> 38 #include <linux/init.h> 39 #include <linux/route.h> 40 #include <linux/rtnetlink.h> 41 #include <linux/netfilter_ipv6.h> 42 #include <linux/slab.h> 43 #include <linux/hash.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/ipv6.h> 51 #include <net/ip6_route.h> 52 #include <net/addrconf.h> 53 #include <net/ip6_tunnel.h> 54 #include <net/xfrm.h> 55 #include <net/dsfield.h> 56 #include <net/inet_ecn.h> 57 #include <net/net_namespace.h> 58 #include <net/netns/generic.h> 59 60 MODULE_AUTHOR("Ville Nuorvala"); 61 MODULE_DESCRIPTION("IPv6 tunneling device"); 62 MODULE_LICENSE("GPL"); 63 MODULE_ALIAS_NETDEV("ip6tnl0"); 64 65 #ifdef IP6_TNL_DEBUG 66 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__) 67 #else 68 #define IP6_TNL_TRACE(x...) do {;} while(0) 69 #endif 70 71 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 72 #define IPV6_TCLASS_SHIFT 20 73 74 #define HASH_SIZE_SHIFT 5 75 #define HASH_SIZE (1 << HASH_SIZE_SHIFT) 76 77 static bool log_ecn_error = true; 78 module_param(log_ecn_error, bool, 0644); 79 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 80 81 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 82 { 83 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 84 85 return hash_32(hash, HASH_SIZE_SHIFT); 86 } 87 88 static int ip6_tnl_dev_init(struct net_device *dev); 89 static void ip6_tnl_dev_setup(struct net_device *dev); 90 static struct rtnl_link_ops ip6_link_ops __read_mostly; 91 92 static int ip6_tnl_net_id __read_mostly; 93 struct ip6_tnl_net { 94 /* the IPv6 tunnel fallback device */ 95 struct net_device *fb_tnl_dev; 96 /* lists for storing tunnels in use */ 97 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; 98 struct ip6_tnl __rcu *tnls_wc[1]; 99 struct ip6_tnl __rcu **tnls[2]; 100 }; 101 102 static struct net_device_stats *ip6_get_stats(struct net_device *dev) 103 { 104 struct pcpu_tstats sum = { 0 }; 105 int i; 106 107 for_each_possible_cpu(i) { 108 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i); 109 110 sum.rx_packets += tstats->rx_packets; 111 sum.rx_bytes += tstats->rx_bytes; 112 sum.tx_packets += tstats->tx_packets; 113 sum.tx_bytes += tstats->tx_bytes; 114 } 115 dev->stats.rx_packets = sum.rx_packets; 116 dev->stats.rx_bytes = sum.rx_bytes; 117 dev->stats.tx_packets = sum.tx_packets; 118 dev->stats.tx_bytes = sum.tx_bytes; 119 return &dev->stats; 120 } 121 122 /* 123 * Locking : hash tables are protected by RCU and RTNL 124 */ 125 126 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 127 { 128 struct dst_entry *dst = t->dst_cache; 129 130 if (dst && dst->obsolete && 131 dst->ops->check(dst, t->dst_cookie) == NULL) { 132 t->dst_cache = NULL; 133 dst_release(dst); 134 return NULL; 135 } 136 137 return dst; 138 } 139 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check); 140 141 void ip6_tnl_dst_reset(struct ip6_tnl *t) 142 { 143 dst_release(t->dst_cache); 144 t->dst_cache = NULL; 145 } 146 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset); 147 148 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 149 { 150 struct rt6_info *rt = (struct rt6_info *) dst; 151 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 152 dst_release(t->dst_cache); 153 t->dst_cache = dst; 154 } 155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store); 156 157 /** 158 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 159 * @remote: the address of the tunnel exit-point 160 * @local: the address of the tunnel entry-point 161 * 162 * Return: 163 * tunnel matching given end-points if found, 164 * else fallback tunnel if its device is up, 165 * else %NULL 166 **/ 167 168 #define for_each_ip6_tunnel_rcu(start) \ 169 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 170 171 static struct ip6_tnl * 172 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local) 173 { 174 unsigned int hash = HASH(remote, local); 175 struct ip6_tnl *t; 176 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 177 178 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 179 if (ipv6_addr_equal(local, &t->parms.laddr) && 180 ipv6_addr_equal(remote, &t->parms.raddr) && 181 (t->dev->flags & IFF_UP)) 182 return t; 183 } 184 t = rcu_dereference(ip6n->tnls_wc[0]); 185 if (t && (t->dev->flags & IFF_UP)) 186 return t; 187 188 return NULL; 189 } 190 191 /** 192 * ip6_tnl_bucket - get head of list matching given tunnel parameters 193 * @p: parameters containing tunnel end-points 194 * 195 * Description: 196 * ip6_tnl_bucket() returns the head of the list matching the 197 * &struct in6_addr entries laddr and raddr in @p. 198 * 199 * Return: head of IPv6 tunnel list 200 **/ 201 202 static struct ip6_tnl __rcu ** 203 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 204 { 205 const struct in6_addr *remote = &p->raddr; 206 const struct in6_addr *local = &p->laddr; 207 unsigned int h = 0; 208 int prio = 0; 209 210 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 211 prio = 1; 212 h = HASH(remote, local); 213 } 214 return &ip6n->tnls[prio][h]; 215 } 216 217 /** 218 * ip6_tnl_link - add tunnel to hash table 219 * @t: tunnel to be added 220 **/ 221 222 static void 223 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 224 { 225 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 226 227 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 228 rcu_assign_pointer(*tp, t); 229 } 230 231 /** 232 * ip6_tnl_unlink - remove tunnel from hash table 233 * @t: tunnel to be removed 234 **/ 235 236 static void 237 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 238 { 239 struct ip6_tnl __rcu **tp; 240 struct ip6_tnl *iter; 241 242 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 243 (iter = rtnl_dereference(*tp)) != NULL; 244 tp = &iter->next) { 245 if (t == iter) { 246 rcu_assign_pointer(*tp, t->next); 247 break; 248 } 249 } 250 } 251 252 static void ip6_dev_free(struct net_device *dev) 253 { 254 free_percpu(dev->tstats); 255 free_netdev(dev); 256 } 257 258 static int ip6_tnl_create2(struct net_device *dev) 259 { 260 struct ip6_tnl *t = netdev_priv(dev); 261 struct net *net = dev_net(dev); 262 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 263 int err; 264 265 t = netdev_priv(dev); 266 err = ip6_tnl_dev_init(dev); 267 if (err < 0) 268 goto out; 269 270 err = register_netdevice(dev); 271 if (err < 0) 272 goto out; 273 274 strcpy(t->parms.name, dev->name); 275 dev->rtnl_link_ops = &ip6_link_ops; 276 277 dev_hold(dev); 278 ip6_tnl_link(ip6n, t); 279 return 0; 280 281 out: 282 return err; 283 } 284 285 /** 286 * ip6_tnl_create - create a new tunnel 287 * @p: tunnel parameters 288 * @pt: pointer to new tunnel 289 * 290 * Description: 291 * Create tunnel matching given parameters. 292 * 293 * Return: 294 * created tunnel or NULL 295 **/ 296 297 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 298 { 299 struct net_device *dev; 300 struct ip6_tnl *t; 301 char name[IFNAMSIZ]; 302 int err; 303 304 if (p->name[0]) 305 strlcpy(name, p->name, IFNAMSIZ); 306 else 307 sprintf(name, "ip6tnl%%d"); 308 309 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup); 310 if (dev == NULL) 311 goto failed; 312 313 dev_net_set(dev, net); 314 315 t = netdev_priv(dev); 316 t->parms = *p; 317 err = ip6_tnl_create2(dev); 318 if (err < 0) 319 goto failed_free; 320 321 return t; 322 323 failed_free: 324 ip6_dev_free(dev); 325 failed: 326 return NULL; 327 } 328 329 /** 330 * ip6_tnl_locate - find or create tunnel matching given parameters 331 * @p: tunnel parameters 332 * @create: != 0 if allowed to create new tunnel if no match found 333 * 334 * Description: 335 * ip6_tnl_locate() first tries to locate an existing tunnel 336 * based on @parms. If this is unsuccessful, but @create is set a new 337 * tunnel device is created and registered for use. 338 * 339 * Return: 340 * matching tunnel or NULL 341 **/ 342 343 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 344 struct __ip6_tnl_parm *p, int create) 345 { 346 const struct in6_addr *remote = &p->raddr; 347 const struct in6_addr *local = &p->laddr; 348 struct ip6_tnl __rcu **tp; 349 struct ip6_tnl *t; 350 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 351 352 for (tp = ip6_tnl_bucket(ip6n, p); 353 (t = rtnl_dereference(*tp)) != NULL; 354 tp = &t->next) { 355 if (ipv6_addr_equal(local, &t->parms.laddr) && 356 ipv6_addr_equal(remote, &t->parms.raddr)) 357 return t; 358 } 359 if (!create) 360 return NULL; 361 return ip6_tnl_create(net, p); 362 } 363 364 /** 365 * ip6_tnl_dev_uninit - tunnel device uninitializer 366 * @dev: the device to be destroyed 367 * 368 * Description: 369 * ip6_tnl_dev_uninit() removes tunnel from its list 370 **/ 371 372 static void 373 ip6_tnl_dev_uninit(struct net_device *dev) 374 { 375 struct ip6_tnl *t = netdev_priv(dev); 376 struct net *net = dev_net(dev); 377 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 378 379 if (dev == ip6n->fb_tnl_dev) 380 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 381 else 382 ip6_tnl_unlink(ip6n, t); 383 ip6_tnl_dst_reset(t); 384 dev_put(dev); 385 } 386 387 /** 388 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 389 * @skb: received socket buffer 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 __u8 nexthdr = ipv6h->nexthdr; 400 __u16 off = sizeof (*ipv6h); 401 402 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 403 __u16 optlen = 0; 404 struct ipv6_opt_hdr *hdr; 405 if (raw + off + sizeof (*hdr) > skb->data && 406 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 407 break; 408 409 hdr = (struct ipv6_opt_hdr *) (raw + off); 410 if (nexthdr == NEXTHDR_FRAGMENT) { 411 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 412 if (frag_hdr->frag_off) 413 break; 414 optlen = 8; 415 } else if (nexthdr == NEXTHDR_AUTH) { 416 optlen = (hdr->hdrlen + 2) << 2; 417 } else { 418 optlen = ipv6_optlen(hdr); 419 } 420 if (nexthdr == NEXTHDR_DEST) { 421 __u16 i = off + 2; 422 while (1) { 423 struct ipv6_tlv_tnl_enc_lim *tel; 424 425 /* No more room for encapsulation limit */ 426 if (i + sizeof (*tel) > off + optlen) 427 break; 428 429 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 430 /* return index of option if found and valid */ 431 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 432 tel->length == 1) 433 return i; 434 /* else jump to next option */ 435 if (tel->type) 436 i += tel->length + 2; 437 else 438 i++; 439 } 440 } 441 nexthdr = hdr->nexthdr; 442 off += optlen; 443 } 444 return 0; 445 } 446 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 447 448 /** 449 * ip6_tnl_err - tunnel error handler 450 * 451 * Description: 452 * ip6_tnl_err() should handle errors in the tunnel according 453 * to the specifications in RFC 2473. 454 **/ 455 456 static int 457 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 458 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 459 { 460 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data; 461 struct ip6_tnl *t; 462 int rel_msg = 0; 463 u8 rel_type = ICMPV6_DEST_UNREACH; 464 u8 rel_code = ICMPV6_ADDR_UNREACH; 465 __u32 rel_info = 0; 466 __u16 len; 467 int err = -ENOENT; 468 469 /* If the packet doesn't contain the original IPv6 header we are 470 in trouble since we might need the source address for further 471 processing of the error. */ 472 473 rcu_read_lock(); 474 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 475 &ipv6h->saddr)) == NULL) 476 goto out; 477 478 if (t->parms.proto != ipproto && t->parms.proto != 0) 479 goto out; 480 481 err = 0; 482 483 switch (*type) { 484 __u32 teli; 485 struct ipv6_tlv_tnl_enc_lim *tel; 486 __u32 mtu; 487 case ICMPV6_DEST_UNREACH: 488 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n", 489 t->parms.name); 490 rel_msg = 1; 491 break; 492 case ICMPV6_TIME_EXCEED: 493 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 494 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 495 t->parms.name); 496 rel_msg = 1; 497 } 498 break; 499 case ICMPV6_PARAMPROB: 500 teli = 0; 501 if ((*code) == ICMPV6_HDR_FIELD) 502 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 503 504 if (teli && teli == *info - 2) { 505 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 506 if (tel->encap_limit == 0) { 507 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 508 t->parms.name); 509 rel_msg = 1; 510 } 511 } else { 512 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 513 t->parms.name); 514 } 515 break; 516 case ICMPV6_PKT_TOOBIG: 517 mtu = *info - offset; 518 if (mtu < IPV6_MIN_MTU) 519 mtu = IPV6_MIN_MTU; 520 t->dev->mtu = mtu; 521 522 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 523 rel_type = ICMPV6_PKT_TOOBIG; 524 rel_code = 0; 525 rel_info = mtu; 526 rel_msg = 1; 527 } 528 break; 529 } 530 531 *type = rel_type; 532 *code = rel_code; 533 *info = rel_info; 534 *msg = rel_msg; 535 536 out: 537 rcu_read_unlock(); 538 return err; 539 } 540 541 static int 542 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 543 u8 type, u8 code, int offset, __be32 info) 544 { 545 int rel_msg = 0; 546 u8 rel_type = type; 547 u8 rel_code = code; 548 __u32 rel_info = ntohl(info); 549 int err; 550 struct sk_buff *skb2; 551 const struct iphdr *eiph; 552 struct rtable *rt; 553 struct flowi4 fl4; 554 555 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 556 &rel_msg, &rel_info, offset); 557 if (err < 0) 558 return err; 559 560 if (rel_msg == 0) 561 return 0; 562 563 switch (rel_type) { 564 case ICMPV6_DEST_UNREACH: 565 if (rel_code != ICMPV6_ADDR_UNREACH) 566 return 0; 567 rel_type = ICMP_DEST_UNREACH; 568 rel_code = ICMP_HOST_UNREACH; 569 break; 570 case ICMPV6_PKT_TOOBIG: 571 if (rel_code != 0) 572 return 0; 573 rel_type = ICMP_DEST_UNREACH; 574 rel_code = ICMP_FRAG_NEEDED; 575 break; 576 case NDISC_REDIRECT: 577 rel_type = ICMP_REDIRECT; 578 rel_code = ICMP_REDIR_HOST; 579 default: 580 return 0; 581 } 582 583 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 584 return 0; 585 586 skb2 = skb_clone(skb, GFP_ATOMIC); 587 if (!skb2) 588 return 0; 589 590 skb_dst_drop(skb2); 591 592 skb_pull(skb2, offset); 593 skb_reset_network_header(skb2); 594 eiph = ip_hdr(skb2); 595 596 /* Try to guess incoming interface */ 597 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 598 eiph->saddr, 0, 599 0, 0, 600 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 601 if (IS_ERR(rt)) 602 goto out; 603 604 skb2->dev = rt->dst.dev; 605 606 /* route "incoming" packet */ 607 if (rt->rt_flags & RTCF_LOCAL) { 608 ip_rt_put(rt); 609 rt = NULL; 610 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 611 eiph->daddr, eiph->saddr, 612 0, 0, 613 IPPROTO_IPIP, 614 RT_TOS(eiph->tos), 0); 615 if (IS_ERR(rt) || 616 rt->dst.dev->type != ARPHRD_TUNNEL) { 617 if (!IS_ERR(rt)) 618 ip_rt_put(rt); 619 goto out; 620 } 621 skb_dst_set(skb2, &rt->dst); 622 } else { 623 ip_rt_put(rt); 624 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 625 skb2->dev) || 626 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 627 goto out; 628 } 629 630 /* change mtu on this route */ 631 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 632 if (rel_info > dst_mtu(skb_dst(skb2))) 633 goto out; 634 635 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info); 636 } 637 if (rel_type == ICMP_REDIRECT) 638 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2); 639 640 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 641 642 out: 643 kfree_skb(skb2); 644 return 0; 645 } 646 647 static int 648 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 649 u8 type, u8 code, int offset, __be32 info) 650 { 651 int rel_msg = 0; 652 u8 rel_type = type; 653 u8 rel_code = code; 654 __u32 rel_info = ntohl(info); 655 int err; 656 657 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 658 &rel_msg, &rel_info, offset); 659 if (err < 0) 660 return err; 661 662 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 663 struct rt6_info *rt; 664 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 665 666 if (!skb2) 667 return 0; 668 669 skb_dst_drop(skb2); 670 skb_pull(skb2, offset); 671 skb_reset_network_header(skb2); 672 673 /* Try to guess incoming interface */ 674 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 675 NULL, 0, 0); 676 677 if (rt && rt->dst.dev) 678 skb2->dev = rt->dst.dev; 679 680 icmpv6_send(skb2, rel_type, rel_code, rel_info); 681 682 ip6_rt_put(rt); 683 684 kfree_skb(skb2); 685 } 686 687 return 0; 688 } 689 690 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 691 const struct ipv6hdr *ipv6h, 692 struct sk_buff *skb) 693 { 694 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 695 696 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 697 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 698 699 return IP6_ECN_decapsulate(ipv6h, skb); 700 } 701 702 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 703 const struct ipv6hdr *ipv6h, 704 struct sk_buff *skb) 705 { 706 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 707 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 708 709 return IP6_ECN_decapsulate(ipv6h, skb); 710 } 711 712 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 713 const struct in6_addr *laddr, 714 const struct in6_addr *raddr) 715 { 716 struct __ip6_tnl_parm *p = &t->parms; 717 int ltype = ipv6_addr_type(laddr); 718 int rtype = ipv6_addr_type(raddr); 719 __u32 flags = 0; 720 721 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 722 flags = IP6_TNL_F_CAP_PER_PACKET; 723 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 724 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 725 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 726 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 727 if (ltype&IPV6_ADDR_UNICAST) 728 flags |= IP6_TNL_F_CAP_XMIT; 729 if (rtype&IPV6_ADDR_UNICAST) 730 flags |= IP6_TNL_F_CAP_RCV; 731 } 732 return flags; 733 } 734 EXPORT_SYMBOL(ip6_tnl_get_cap); 735 736 /* called with rcu_read_lock() */ 737 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 738 const struct in6_addr *laddr, 739 const struct in6_addr *raddr) 740 { 741 struct __ip6_tnl_parm *p = &t->parms; 742 int ret = 0; 743 struct net *net = dev_net(t->dev); 744 745 if ((p->flags & IP6_TNL_F_CAP_RCV) || 746 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 747 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 748 struct net_device *ldev = NULL; 749 750 if (p->link) 751 ldev = dev_get_by_index_rcu(net, p->link); 752 753 if ((ipv6_addr_is_multicast(laddr) || 754 likely(ipv6_chk_addr(net, laddr, ldev, 0))) && 755 likely(!ipv6_chk_addr(net, raddr, NULL, 0))) 756 ret = 1; 757 } 758 return ret; 759 } 760 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 761 762 /** 763 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 764 * @skb: received socket buffer 765 * @protocol: ethernet protocol ID 766 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 767 * 768 * Return: 0 769 **/ 770 771 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 772 __u8 ipproto, 773 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 774 const struct ipv6hdr *ipv6h, 775 struct sk_buff *skb)) 776 { 777 struct ip6_tnl *t; 778 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 779 int err; 780 781 rcu_read_lock(); 782 783 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 784 &ipv6h->daddr)) != NULL) { 785 struct pcpu_tstats *tstats; 786 787 if (t->parms.proto != ipproto && t->parms.proto != 0) { 788 rcu_read_unlock(); 789 goto discard; 790 } 791 792 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 793 rcu_read_unlock(); 794 goto discard; 795 } 796 797 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 798 t->dev->stats.rx_dropped++; 799 rcu_read_unlock(); 800 goto discard; 801 } 802 secpath_reset(skb); 803 skb->mac_header = skb->network_header; 804 skb_reset_network_header(skb); 805 skb->protocol = htons(protocol); 806 skb->pkt_type = PACKET_HOST; 807 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 808 809 __skb_tunnel_rx(skb, t->dev); 810 811 err = dscp_ecn_decapsulate(t, ipv6h, skb); 812 if (unlikely(err)) { 813 if (log_ecn_error) 814 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n", 815 &ipv6h->saddr, 816 ipv6_get_dsfield(ipv6h)); 817 if (err > 1) { 818 ++t->dev->stats.rx_frame_errors; 819 ++t->dev->stats.rx_errors; 820 rcu_read_unlock(); 821 goto discard; 822 } 823 } 824 825 tstats = this_cpu_ptr(t->dev->tstats); 826 tstats->rx_packets++; 827 tstats->rx_bytes += skb->len; 828 829 netif_rx(skb); 830 831 rcu_read_unlock(); 832 return 0; 833 } 834 rcu_read_unlock(); 835 return 1; 836 837 discard: 838 kfree_skb(skb); 839 return 0; 840 } 841 842 static int ip4ip6_rcv(struct sk_buff *skb) 843 { 844 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 845 ip4ip6_dscp_ecn_decapsulate); 846 } 847 848 static int ip6ip6_rcv(struct sk_buff *skb) 849 { 850 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 851 ip6ip6_dscp_ecn_decapsulate); 852 } 853 854 struct ipv6_tel_txoption { 855 struct ipv6_txoptions ops; 856 __u8 dst_opt[8]; 857 }; 858 859 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 860 { 861 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 862 863 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 864 opt->dst_opt[3] = 1; 865 opt->dst_opt[4] = encap_limit; 866 opt->dst_opt[5] = IPV6_TLV_PADN; 867 opt->dst_opt[6] = 1; 868 869 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 870 opt->ops.opt_nflen = 8; 871 } 872 873 /** 874 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 875 * @t: the outgoing tunnel device 876 * @hdr: IPv6 header from the incoming packet 877 * 878 * Description: 879 * Avoid trivial tunneling loop by checking that tunnel exit-point 880 * doesn't match source of incoming packet. 881 * 882 * Return: 883 * 1 if conflict, 884 * 0 else 885 **/ 886 887 static inline bool 888 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 889 { 890 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 891 } 892 893 int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 894 { 895 struct __ip6_tnl_parm *p = &t->parms; 896 int ret = 0; 897 struct net *net = dev_net(t->dev); 898 899 if (p->flags & IP6_TNL_F_CAP_XMIT) { 900 struct net_device *ldev = NULL; 901 902 rcu_read_lock(); 903 if (p->link) 904 ldev = dev_get_by_index_rcu(net, p->link); 905 906 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 907 pr_warn("%s xmit: Local address not yet configured!\n", 908 p->name); 909 else if (!ipv6_addr_is_multicast(&p->raddr) && 910 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 911 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 912 p->name); 913 else 914 ret = 1; 915 rcu_read_unlock(); 916 } 917 return ret; 918 } 919 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 920 921 /** 922 * ip6_tnl_xmit2 - encapsulate packet and send 923 * @skb: the outgoing socket buffer 924 * @dev: the outgoing tunnel device 925 * @dsfield: dscp code for outer header 926 * @fl: flow of tunneled packet 927 * @encap_limit: encapsulation limit 928 * @pmtu: Path MTU is stored if packet is too big 929 * 930 * Description: 931 * Build new header and do some sanity checks on the packet before sending 932 * it. 933 * 934 * Return: 935 * 0 on success 936 * -1 fail 937 * %-EMSGSIZE message too big. return mtu in this case. 938 **/ 939 940 static int ip6_tnl_xmit2(struct sk_buff *skb, 941 struct net_device *dev, 942 __u8 dsfield, 943 struct flowi6 *fl6, 944 int encap_limit, 945 __u32 *pmtu) 946 { 947 struct net *net = dev_net(dev); 948 struct ip6_tnl *t = netdev_priv(dev); 949 struct net_device_stats *stats = &t->dev->stats; 950 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 951 struct ipv6_tel_txoption opt; 952 struct dst_entry *dst = NULL, *ndst = NULL; 953 struct net_device *tdev; 954 int mtu; 955 unsigned int max_headroom = sizeof(struct ipv6hdr); 956 u8 proto; 957 int err = -1; 958 int pkt_len; 959 960 if (!fl6->flowi6_mark) 961 dst = ip6_tnl_dst_check(t); 962 if (!dst) { 963 ndst = ip6_route_output(net, NULL, fl6); 964 965 if (ndst->error) 966 goto tx_err_link_failure; 967 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0); 968 if (IS_ERR(ndst)) { 969 err = PTR_ERR(ndst); 970 ndst = NULL; 971 goto tx_err_link_failure; 972 } 973 dst = ndst; 974 } 975 976 tdev = dst->dev; 977 978 if (tdev == dev) { 979 stats->collisions++; 980 net_warn_ratelimited("%s: Local routing loop detected!\n", 981 t->parms.name); 982 goto tx_err_dst_release; 983 } 984 mtu = dst_mtu(dst) - sizeof (*ipv6h); 985 if (encap_limit >= 0) { 986 max_headroom += 8; 987 mtu -= 8; 988 } 989 if (mtu < IPV6_MIN_MTU) 990 mtu = IPV6_MIN_MTU; 991 if (skb_dst(skb)) 992 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 993 if (skb->len > mtu) { 994 *pmtu = mtu; 995 err = -EMSGSIZE; 996 goto tx_err_dst_release; 997 } 998 999 /* 1000 * Okay, now see if we can stuff it in the buffer as-is. 1001 */ 1002 max_headroom += LL_RESERVED_SPACE(tdev); 1003 1004 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1005 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1006 struct sk_buff *new_skb; 1007 1008 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 1009 goto tx_err_dst_release; 1010 1011 if (skb->sk) 1012 skb_set_owner_w(new_skb, skb->sk); 1013 consume_skb(skb); 1014 skb = new_skb; 1015 } 1016 skb_dst_drop(skb); 1017 if (fl6->flowi6_mark) { 1018 skb_dst_set(skb, dst); 1019 ndst = NULL; 1020 } else { 1021 skb_dst_set_noref(skb, dst); 1022 } 1023 skb->transport_header = skb->network_header; 1024 1025 proto = fl6->flowi6_proto; 1026 if (encap_limit >= 0) { 1027 init_tel_txopt(&opt, encap_limit); 1028 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 1029 } 1030 skb_push(skb, sizeof(struct ipv6hdr)); 1031 skb_reset_network_header(skb); 1032 ipv6h = ipv6_hdr(skb); 1033 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000); 1034 dsfield = INET_ECN_encapsulate(0, dsfield); 1035 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield); 1036 ipv6h->hop_limit = t->parms.hop_limit; 1037 ipv6h->nexthdr = proto; 1038 ipv6h->saddr = fl6->saddr; 1039 ipv6h->daddr = fl6->daddr; 1040 nf_reset(skb); 1041 pkt_len = skb->len; 1042 err = ip6_local_out(skb); 1043 1044 if (net_xmit_eval(err) == 0) { 1045 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats); 1046 1047 tstats->tx_bytes += pkt_len; 1048 tstats->tx_packets++; 1049 } else { 1050 stats->tx_errors++; 1051 stats->tx_aborted_errors++; 1052 } 1053 if (ndst) 1054 ip6_tnl_dst_store(t, ndst); 1055 return 0; 1056 tx_err_link_failure: 1057 stats->tx_carrier_errors++; 1058 dst_link_failure(skb); 1059 tx_err_dst_release: 1060 dst_release(ndst); 1061 return err; 1062 } 1063 1064 static inline int 1065 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1066 { 1067 struct ip6_tnl *t = netdev_priv(dev); 1068 const struct iphdr *iph = ip_hdr(skb); 1069 int encap_limit = -1; 1070 struct flowi6 fl6; 1071 __u8 dsfield; 1072 __u32 mtu; 1073 int err; 1074 1075 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 1076 !ip6_tnl_xmit_ctl(t)) 1077 return -1; 1078 1079 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1080 encap_limit = t->parms.encap_limit; 1081 1082 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1083 fl6.flowi6_proto = IPPROTO_IPIP; 1084 1085 dsfield = ipv4_get_dsfield(iph); 1086 1087 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1088 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 1089 & IPV6_TCLASS_MASK; 1090 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1091 fl6.flowi6_mark = skb->mark; 1092 1093 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1094 if (err != 0) { 1095 /* XXX: send ICMP error even if DF is not set. */ 1096 if (err == -EMSGSIZE) 1097 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1098 htonl(mtu)); 1099 return -1; 1100 } 1101 1102 return 0; 1103 } 1104 1105 static inline int 1106 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1107 { 1108 struct ip6_tnl *t = netdev_priv(dev); 1109 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1110 int encap_limit = -1; 1111 __u16 offset; 1112 struct flowi6 fl6; 1113 __u8 dsfield; 1114 __u32 mtu; 1115 int err; 1116 1117 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1118 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1119 return -1; 1120 1121 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 1122 if (offset > 0) { 1123 struct ipv6_tlv_tnl_enc_lim *tel; 1124 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1125 if (tel->encap_limit == 0) { 1126 icmpv6_send(skb, ICMPV6_PARAMPROB, 1127 ICMPV6_HDR_FIELD, offset + 2); 1128 return -1; 1129 } 1130 encap_limit = tel->encap_limit - 1; 1131 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1132 encap_limit = t->parms.encap_limit; 1133 1134 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1135 fl6.flowi6_proto = IPPROTO_IPV6; 1136 1137 dsfield = ipv6_get_dsfield(ipv6h); 1138 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1139 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1140 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1141 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); 1142 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1143 fl6.flowi6_mark = skb->mark; 1144 1145 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1146 if (err != 0) { 1147 if (err == -EMSGSIZE) 1148 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1149 return -1; 1150 } 1151 1152 return 0; 1153 } 1154 1155 static netdev_tx_t 1156 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1157 { 1158 struct ip6_tnl *t = netdev_priv(dev); 1159 struct net_device_stats *stats = &t->dev->stats; 1160 int ret; 1161 1162 switch (skb->protocol) { 1163 case htons(ETH_P_IP): 1164 ret = ip4ip6_tnl_xmit(skb, dev); 1165 break; 1166 case htons(ETH_P_IPV6): 1167 ret = ip6ip6_tnl_xmit(skb, dev); 1168 break; 1169 default: 1170 goto tx_err; 1171 } 1172 1173 if (ret < 0) 1174 goto tx_err; 1175 1176 return NETDEV_TX_OK; 1177 1178 tx_err: 1179 stats->tx_errors++; 1180 stats->tx_dropped++; 1181 kfree_skb(skb); 1182 return NETDEV_TX_OK; 1183 } 1184 1185 static void ip6_tnl_link_config(struct ip6_tnl *t) 1186 { 1187 struct net_device *dev = t->dev; 1188 struct __ip6_tnl_parm *p = &t->parms; 1189 struct flowi6 *fl6 = &t->fl.u.ip6; 1190 1191 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1192 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1193 1194 /* Set up flowi template */ 1195 fl6->saddr = p->laddr; 1196 fl6->daddr = p->raddr; 1197 fl6->flowi6_oif = p->link; 1198 fl6->flowlabel = 0; 1199 1200 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1201 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1202 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1203 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1204 1205 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1206 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1207 1208 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1209 dev->flags |= IFF_POINTOPOINT; 1210 else 1211 dev->flags &= ~IFF_POINTOPOINT; 1212 1213 dev->iflink = p->link; 1214 1215 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1216 int strict = (ipv6_addr_type(&p->raddr) & 1217 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1218 1219 struct rt6_info *rt = rt6_lookup(dev_net(dev), 1220 &p->raddr, &p->laddr, 1221 p->link, strict); 1222 1223 if (rt == NULL) 1224 return; 1225 1226 if (rt->dst.dev) { 1227 dev->hard_header_len = rt->dst.dev->hard_header_len + 1228 sizeof (struct ipv6hdr); 1229 1230 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr); 1231 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1232 dev->mtu-=8; 1233 1234 if (dev->mtu < IPV6_MIN_MTU) 1235 dev->mtu = IPV6_MIN_MTU; 1236 } 1237 ip6_rt_put(rt); 1238 } 1239 } 1240 1241 /** 1242 * ip6_tnl_change - update the tunnel parameters 1243 * @t: tunnel to be changed 1244 * @p: tunnel configuration parameters 1245 * 1246 * Description: 1247 * ip6_tnl_change() updates the tunnel parameters 1248 **/ 1249 1250 static int 1251 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1252 { 1253 t->parms.laddr = p->laddr; 1254 t->parms.raddr = p->raddr; 1255 t->parms.flags = p->flags; 1256 t->parms.hop_limit = p->hop_limit; 1257 t->parms.encap_limit = p->encap_limit; 1258 t->parms.flowinfo = p->flowinfo; 1259 t->parms.link = p->link; 1260 t->parms.proto = p->proto; 1261 ip6_tnl_dst_reset(t); 1262 ip6_tnl_link_config(t); 1263 return 0; 1264 } 1265 1266 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1267 { 1268 struct net *net = dev_net(t->dev); 1269 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1270 int err; 1271 1272 ip6_tnl_unlink(ip6n, t); 1273 synchronize_net(); 1274 err = ip6_tnl_change(t, p); 1275 ip6_tnl_link(ip6n, t); 1276 netdev_state_change(t->dev); 1277 return err; 1278 } 1279 1280 static void 1281 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1282 { 1283 p->laddr = u->laddr; 1284 p->raddr = u->raddr; 1285 p->flags = u->flags; 1286 p->hop_limit = u->hop_limit; 1287 p->encap_limit = u->encap_limit; 1288 p->flowinfo = u->flowinfo; 1289 p->link = u->link; 1290 p->proto = u->proto; 1291 memcpy(p->name, u->name, sizeof(u->name)); 1292 } 1293 1294 static void 1295 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1296 { 1297 u->laddr = p->laddr; 1298 u->raddr = p->raddr; 1299 u->flags = p->flags; 1300 u->hop_limit = p->hop_limit; 1301 u->encap_limit = p->encap_limit; 1302 u->flowinfo = p->flowinfo; 1303 u->link = p->link; 1304 u->proto = p->proto; 1305 memcpy(u->name, p->name, sizeof(u->name)); 1306 } 1307 1308 /** 1309 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1310 * @dev: virtual device associated with tunnel 1311 * @ifr: parameters passed from userspace 1312 * @cmd: command to be performed 1313 * 1314 * Description: 1315 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1316 * from userspace. 1317 * 1318 * The possible commands are the following: 1319 * %SIOCGETTUNNEL: get tunnel parameters for device 1320 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1321 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1322 * %SIOCDELTUNNEL: delete tunnel 1323 * 1324 * The fallback device "ip6tnl0", created during module 1325 * initialization, can be used for creating other tunnel devices. 1326 * 1327 * Return: 1328 * 0 on success, 1329 * %-EFAULT if unable to copy data to or from userspace, 1330 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1331 * %-EINVAL if passed tunnel parameters are invalid, 1332 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1333 * %-ENODEV if attempting to change or delete a nonexisting device 1334 **/ 1335 1336 static int 1337 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1338 { 1339 int err = 0; 1340 struct ip6_tnl_parm p; 1341 struct __ip6_tnl_parm p1; 1342 struct ip6_tnl *t = NULL; 1343 struct net *net = dev_net(dev); 1344 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1345 1346 switch (cmd) { 1347 case SIOCGETTUNNEL: 1348 if (dev == ip6n->fb_tnl_dev) { 1349 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1350 err = -EFAULT; 1351 break; 1352 } 1353 ip6_tnl_parm_from_user(&p1, &p); 1354 t = ip6_tnl_locate(net, &p1, 0); 1355 } else { 1356 memset(&p, 0, sizeof(p)); 1357 } 1358 if (t == NULL) 1359 t = netdev_priv(dev); 1360 ip6_tnl_parm_to_user(&p, &t->parms); 1361 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1362 err = -EFAULT; 1363 } 1364 break; 1365 case SIOCADDTUNNEL: 1366 case SIOCCHGTUNNEL: 1367 err = -EPERM; 1368 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1369 break; 1370 err = -EFAULT; 1371 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1372 break; 1373 err = -EINVAL; 1374 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1375 p.proto != 0) 1376 break; 1377 ip6_tnl_parm_from_user(&p1, &p); 1378 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1379 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1380 if (t != NULL) { 1381 if (t->dev != dev) { 1382 err = -EEXIST; 1383 break; 1384 } 1385 } else 1386 t = netdev_priv(dev); 1387 1388 err = ip6_tnl_update(t, &p1); 1389 } 1390 if (t) { 1391 err = 0; 1392 ip6_tnl_parm_to_user(&p, &t->parms); 1393 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1394 err = -EFAULT; 1395 1396 } else 1397 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1398 break; 1399 case SIOCDELTUNNEL: 1400 err = -EPERM; 1401 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1402 break; 1403 1404 if (dev == ip6n->fb_tnl_dev) { 1405 err = -EFAULT; 1406 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1407 break; 1408 err = -ENOENT; 1409 ip6_tnl_parm_from_user(&p1, &p); 1410 t = ip6_tnl_locate(net, &p1, 0); 1411 if (t == NULL) 1412 break; 1413 err = -EPERM; 1414 if (t->dev == ip6n->fb_tnl_dev) 1415 break; 1416 dev = t->dev; 1417 } 1418 err = 0; 1419 unregister_netdevice(dev); 1420 break; 1421 default: 1422 err = -EINVAL; 1423 } 1424 return err; 1425 } 1426 1427 /** 1428 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1429 * @dev: virtual device associated with tunnel 1430 * @new_mtu: the new mtu 1431 * 1432 * Return: 1433 * 0 on success, 1434 * %-EINVAL if mtu too small 1435 **/ 1436 1437 static int 1438 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1439 { 1440 if (new_mtu < IPV6_MIN_MTU) { 1441 return -EINVAL; 1442 } 1443 dev->mtu = new_mtu; 1444 return 0; 1445 } 1446 1447 1448 static const struct net_device_ops ip6_tnl_netdev_ops = { 1449 .ndo_uninit = ip6_tnl_dev_uninit, 1450 .ndo_start_xmit = ip6_tnl_xmit, 1451 .ndo_do_ioctl = ip6_tnl_ioctl, 1452 .ndo_change_mtu = ip6_tnl_change_mtu, 1453 .ndo_get_stats = ip6_get_stats, 1454 }; 1455 1456 1457 /** 1458 * ip6_tnl_dev_setup - setup virtual tunnel device 1459 * @dev: virtual device associated with tunnel 1460 * 1461 * Description: 1462 * Initialize function pointers and device parameters 1463 **/ 1464 1465 static void ip6_tnl_dev_setup(struct net_device *dev) 1466 { 1467 struct ip6_tnl *t; 1468 1469 dev->netdev_ops = &ip6_tnl_netdev_ops; 1470 dev->destructor = ip6_dev_free; 1471 1472 dev->type = ARPHRD_TUNNEL6; 1473 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1474 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1475 t = netdev_priv(dev); 1476 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1477 dev->mtu-=8; 1478 dev->flags |= IFF_NOARP; 1479 dev->addr_len = sizeof(struct in6_addr); 1480 dev->features |= NETIF_F_NETNS_LOCAL; 1481 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1482 } 1483 1484 1485 /** 1486 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1487 * @dev: virtual device associated with tunnel 1488 **/ 1489 1490 static inline int 1491 ip6_tnl_dev_init_gen(struct net_device *dev) 1492 { 1493 struct ip6_tnl *t = netdev_priv(dev); 1494 1495 t->dev = dev; 1496 dev->tstats = alloc_percpu(struct pcpu_tstats); 1497 if (!dev->tstats) 1498 return -ENOMEM; 1499 return 0; 1500 } 1501 1502 /** 1503 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1504 * @dev: virtual device associated with tunnel 1505 **/ 1506 1507 static int ip6_tnl_dev_init(struct net_device *dev) 1508 { 1509 struct ip6_tnl *t = netdev_priv(dev); 1510 int err = ip6_tnl_dev_init_gen(dev); 1511 1512 if (err) 1513 return err; 1514 ip6_tnl_link_config(t); 1515 return 0; 1516 } 1517 1518 /** 1519 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1520 * @dev: fallback device 1521 * 1522 * Return: 0 1523 **/ 1524 1525 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1526 { 1527 struct ip6_tnl *t = netdev_priv(dev); 1528 struct net *net = dev_net(dev); 1529 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1530 int err = ip6_tnl_dev_init_gen(dev); 1531 1532 if (err) 1533 return err; 1534 1535 t->parms.proto = IPPROTO_IPV6; 1536 dev_hold(dev); 1537 1538 ip6_tnl_link_config(t); 1539 1540 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1541 return 0; 1542 } 1543 1544 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[]) 1545 { 1546 u8 proto; 1547 1548 if (!data) 1549 return 0; 1550 1551 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1552 if (proto != IPPROTO_IPV6 && 1553 proto != IPPROTO_IPIP && 1554 proto != 0) 1555 return -EINVAL; 1556 1557 return 0; 1558 } 1559 1560 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1561 struct __ip6_tnl_parm *parms) 1562 { 1563 memset(parms, 0, sizeof(*parms)); 1564 1565 if (!data) 1566 return; 1567 1568 if (data[IFLA_IPTUN_LINK]) 1569 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1570 1571 if (data[IFLA_IPTUN_LOCAL]) 1572 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL], 1573 sizeof(struct in6_addr)); 1574 1575 if (data[IFLA_IPTUN_REMOTE]) 1576 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE], 1577 sizeof(struct in6_addr)); 1578 1579 if (data[IFLA_IPTUN_TTL]) 1580 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1581 1582 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1583 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1584 1585 if (data[IFLA_IPTUN_FLOWINFO]) 1586 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1587 1588 if (data[IFLA_IPTUN_FLAGS]) 1589 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1590 1591 if (data[IFLA_IPTUN_PROTO]) 1592 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1593 } 1594 1595 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 1596 struct nlattr *tb[], struct nlattr *data[]) 1597 { 1598 struct net *net = dev_net(dev); 1599 struct ip6_tnl *nt; 1600 1601 nt = netdev_priv(dev); 1602 ip6_tnl_netlink_parms(data, &nt->parms); 1603 1604 if (ip6_tnl_locate(net, &nt->parms, 0)) 1605 return -EEXIST; 1606 1607 return ip6_tnl_create2(dev); 1608 } 1609 1610 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 1611 struct nlattr *data[]) 1612 { 1613 struct ip6_tnl *t; 1614 struct __ip6_tnl_parm p; 1615 struct net *net = dev_net(dev); 1616 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1617 1618 if (dev == ip6n->fb_tnl_dev) 1619 return -EINVAL; 1620 1621 ip6_tnl_netlink_parms(data, &p); 1622 1623 t = ip6_tnl_locate(net, &p, 0); 1624 1625 if (t) { 1626 if (t->dev != dev) 1627 return -EEXIST; 1628 } else 1629 t = netdev_priv(dev); 1630 1631 return ip6_tnl_update(t, &p); 1632 } 1633 1634 static size_t ip6_tnl_get_size(const struct net_device *dev) 1635 { 1636 return 1637 /* IFLA_IPTUN_LINK */ 1638 nla_total_size(4) + 1639 /* IFLA_IPTUN_LOCAL */ 1640 nla_total_size(sizeof(struct in6_addr)) + 1641 /* IFLA_IPTUN_REMOTE */ 1642 nla_total_size(sizeof(struct in6_addr)) + 1643 /* IFLA_IPTUN_TTL */ 1644 nla_total_size(1) + 1645 /* IFLA_IPTUN_ENCAP_LIMIT */ 1646 nla_total_size(1) + 1647 /* IFLA_IPTUN_FLOWINFO */ 1648 nla_total_size(4) + 1649 /* IFLA_IPTUN_FLAGS */ 1650 nla_total_size(4) + 1651 /* IFLA_IPTUN_PROTO */ 1652 nla_total_size(1) + 1653 0; 1654 } 1655 1656 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 1657 { 1658 struct ip6_tnl *tunnel = netdev_priv(dev); 1659 struct __ip6_tnl_parm *parm = &tunnel->parms; 1660 1661 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1662 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr), 1663 &parm->raddr) || 1664 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr), 1665 &parm->laddr) || 1666 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 1667 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 1668 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 1669 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 1670 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto)) 1671 goto nla_put_failure; 1672 return 0; 1673 1674 nla_put_failure: 1675 return -EMSGSIZE; 1676 } 1677 1678 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 1679 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1680 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 1681 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 1682 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1683 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 1684 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 1685 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 1686 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1687 }; 1688 1689 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 1690 .kind = "ip6tnl", 1691 .maxtype = IFLA_IPTUN_MAX, 1692 .policy = ip6_tnl_policy, 1693 .priv_size = sizeof(struct ip6_tnl), 1694 .setup = ip6_tnl_dev_setup, 1695 .validate = ip6_tnl_validate, 1696 .newlink = ip6_tnl_newlink, 1697 .changelink = ip6_tnl_changelink, 1698 .get_size = ip6_tnl_get_size, 1699 .fill_info = ip6_tnl_fill_info, 1700 }; 1701 1702 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 1703 .handler = ip4ip6_rcv, 1704 .err_handler = ip4ip6_err, 1705 .priority = 1, 1706 }; 1707 1708 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 1709 .handler = ip6ip6_rcv, 1710 .err_handler = ip6ip6_err, 1711 .priority = 1, 1712 }; 1713 1714 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n) 1715 { 1716 int h; 1717 struct ip6_tnl *t; 1718 LIST_HEAD(list); 1719 1720 for (h = 0; h < HASH_SIZE; h++) { 1721 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1722 while (t != NULL) { 1723 unregister_netdevice_queue(t->dev, &list); 1724 t = rtnl_dereference(t->next); 1725 } 1726 } 1727 1728 t = rtnl_dereference(ip6n->tnls_wc[0]); 1729 unregister_netdevice_queue(t->dev, &list); 1730 unregister_netdevice_many(&list); 1731 } 1732 1733 static int __net_init ip6_tnl_init_net(struct net *net) 1734 { 1735 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1736 struct ip6_tnl *t = NULL; 1737 int err; 1738 1739 ip6n->tnls[0] = ip6n->tnls_wc; 1740 ip6n->tnls[1] = ip6n->tnls_r_l; 1741 1742 err = -ENOMEM; 1743 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1744 ip6_tnl_dev_setup); 1745 1746 if (!ip6n->fb_tnl_dev) 1747 goto err_alloc_dev; 1748 dev_net_set(ip6n->fb_tnl_dev, net); 1749 1750 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1751 if (err < 0) 1752 goto err_register; 1753 1754 err = register_netdev(ip6n->fb_tnl_dev); 1755 if (err < 0) 1756 goto err_register; 1757 1758 t = netdev_priv(ip6n->fb_tnl_dev); 1759 1760 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1761 return 0; 1762 1763 err_register: 1764 ip6_dev_free(ip6n->fb_tnl_dev); 1765 err_alloc_dev: 1766 return err; 1767 } 1768 1769 static void __net_exit ip6_tnl_exit_net(struct net *net) 1770 { 1771 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1772 1773 rtnl_lock(); 1774 ip6_tnl_destroy_tunnels(ip6n); 1775 rtnl_unlock(); 1776 } 1777 1778 static struct pernet_operations ip6_tnl_net_ops = { 1779 .init = ip6_tnl_init_net, 1780 .exit = ip6_tnl_exit_net, 1781 .id = &ip6_tnl_net_id, 1782 .size = sizeof(struct ip6_tnl_net), 1783 }; 1784 1785 /** 1786 * ip6_tunnel_init - register protocol and reserve needed resources 1787 * 1788 * Return: 0 on success 1789 **/ 1790 1791 static int __init ip6_tunnel_init(void) 1792 { 1793 int err; 1794 1795 err = register_pernet_device(&ip6_tnl_net_ops); 1796 if (err < 0) 1797 goto out_pernet; 1798 1799 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 1800 if (err < 0) { 1801 pr_err("%s: can't register ip4ip6\n", __func__); 1802 goto out_ip4ip6; 1803 } 1804 1805 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 1806 if (err < 0) { 1807 pr_err("%s: can't register ip6ip6\n", __func__); 1808 goto out_ip6ip6; 1809 } 1810 err = rtnl_link_register(&ip6_link_ops); 1811 if (err < 0) 1812 goto rtnl_link_failed; 1813 1814 return 0; 1815 1816 rtnl_link_failed: 1817 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 1818 out_ip6ip6: 1819 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1820 out_ip4ip6: 1821 unregister_pernet_device(&ip6_tnl_net_ops); 1822 out_pernet: 1823 return err; 1824 } 1825 1826 /** 1827 * ip6_tunnel_cleanup - free resources and unregister protocol 1828 **/ 1829 1830 static void __exit ip6_tunnel_cleanup(void) 1831 { 1832 rtnl_link_unregister(&ip6_link_ops); 1833 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1834 pr_info("%s: can't deregister ip4ip6\n", __func__); 1835 1836 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1837 pr_info("%s: can't deregister ip6ip6\n", __func__); 1838 1839 unregister_pernet_device(&ip6_tnl_net_ops); 1840 } 1841 1842 module_init(ip6_tunnel_init); 1843 module_exit(ip6_tunnel_cleanup); 1844