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 #include <linux/etherdevice.h> 45 46 #include <asm/uaccess.h> 47 #include <linux/atomic.h> 48 49 #include <net/icmp.h> 50 #include <net/ip.h> 51 #include <net/ip_tunnels.h> 52 #include <net/ipv6.h> 53 #include <net/ip6_route.h> 54 #include <net/addrconf.h> 55 #include <net/ip6_tunnel.h> 56 #include <net/xfrm.h> 57 #include <net/dsfield.h> 58 #include <net/inet_ecn.h> 59 #include <net/net_namespace.h> 60 #include <net/netns/generic.h> 61 62 MODULE_AUTHOR("Ville Nuorvala"); 63 MODULE_DESCRIPTION("IPv6 tunneling device"); 64 MODULE_LICENSE("GPL"); 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 IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 74 #define IPV6_TCLASS_SHIFT 20 75 76 #define HASH_SIZE_SHIFT 5 77 #define HASH_SIZE (1 << HASH_SIZE_SHIFT) 78 79 static bool log_ecn_error = true; 80 module_param(log_ecn_error, bool, 0644); 81 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 82 83 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 84 { 85 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 86 87 return hash_32(hash, HASH_SIZE_SHIFT); 88 } 89 90 static int ip6_tnl_dev_init(struct net_device *dev); 91 static void ip6_tnl_dev_setup(struct net_device *dev); 92 static struct rtnl_link_ops ip6_link_ops __read_mostly; 93 94 static int ip6_tnl_net_id __read_mostly; 95 struct ip6_tnl_net { 96 /* the IPv6 tunnel fallback device */ 97 struct net_device *fb_tnl_dev; 98 /* lists for storing tunnels in use */ 99 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; 100 struct ip6_tnl __rcu *tnls_wc[1]; 101 struct ip6_tnl __rcu **tnls[2]; 102 }; 103 104 static struct net_device_stats *ip6_get_stats(struct net_device *dev) 105 { 106 struct pcpu_tstats tmp, sum = { 0 }; 107 int i; 108 109 for_each_possible_cpu(i) { 110 unsigned int start; 111 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i); 112 113 do { 114 start = u64_stats_fetch_begin_bh(&tstats->syncp); 115 tmp.rx_packets = tstats->rx_packets; 116 tmp.rx_bytes = tstats->rx_bytes; 117 tmp.tx_packets = tstats->tx_packets; 118 tmp.tx_bytes = tstats->tx_bytes; 119 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start)); 120 121 sum.rx_packets += tmp.rx_packets; 122 sum.rx_bytes += tmp.rx_bytes; 123 sum.tx_packets += tmp.tx_packets; 124 sum.tx_bytes += tmp.tx_bytes; 125 } 126 dev->stats.rx_packets = sum.rx_packets; 127 dev->stats.rx_bytes = sum.rx_bytes; 128 dev->stats.tx_packets = sum.tx_packets; 129 dev->stats.tx_bytes = sum.tx_bytes; 130 return &dev->stats; 131 } 132 133 /* 134 * Locking : hash tables are protected by RCU and RTNL 135 */ 136 137 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 138 { 139 struct dst_entry *dst = t->dst_cache; 140 141 if (dst && dst->obsolete && 142 dst->ops->check(dst, t->dst_cookie) == NULL) { 143 t->dst_cache = NULL; 144 dst_release(dst); 145 return NULL; 146 } 147 148 return dst; 149 } 150 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check); 151 152 void ip6_tnl_dst_reset(struct ip6_tnl *t) 153 { 154 dst_release(t->dst_cache); 155 t->dst_cache = NULL; 156 } 157 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset); 158 159 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 160 { 161 struct rt6_info *rt = (struct rt6_info *) dst; 162 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 163 dst_release(t->dst_cache); 164 t->dst_cache = dst; 165 } 166 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store); 167 168 /** 169 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 170 * @remote: the address of the tunnel exit-point 171 * @local: the address of the tunnel entry-point 172 * 173 * Return: 174 * tunnel matching given end-points if found, 175 * else fallback tunnel if its device is up, 176 * else %NULL 177 **/ 178 179 #define for_each_ip6_tunnel_rcu(start) \ 180 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 181 182 static struct ip6_tnl * 183 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local) 184 { 185 unsigned int hash = HASH(remote, local); 186 struct ip6_tnl *t; 187 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 188 189 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 190 if (ipv6_addr_equal(local, &t->parms.laddr) && 191 ipv6_addr_equal(remote, &t->parms.raddr) && 192 (t->dev->flags & IFF_UP)) 193 return t; 194 } 195 t = rcu_dereference(ip6n->tnls_wc[0]); 196 if (t && (t->dev->flags & IFF_UP)) 197 return t; 198 199 return NULL; 200 } 201 202 /** 203 * ip6_tnl_bucket - get head of list matching given tunnel parameters 204 * @p: parameters containing tunnel end-points 205 * 206 * Description: 207 * ip6_tnl_bucket() returns the head of the list matching the 208 * &struct in6_addr entries laddr and raddr in @p. 209 * 210 * Return: head of IPv6 tunnel list 211 **/ 212 213 static struct ip6_tnl __rcu ** 214 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 215 { 216 const struct in6_addr *remote = &p->raddr; 217 const struct in6_addr *local = &p->laddr; 218 unsigned int h = 0; 219 int prio = 0; 220 221 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 222 prio = 1; 223 h = HASH(remote, local); 224 } 225 return &ip6n->tnls[prio][h]; 226 } 227 228 /** 229 * ip6_tnl_link - add tunnel to hash table 230 * @t: tunnel to be added 231 **/ 232 233 static void 234 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 235 { 236 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 237 238 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 239 rcu_assign_pointer(*tp, t); 240 } 241 242 /** 243 * ip6_tnl_unlink - remove tunnel from hash table 244 * @t: tunnel to be removed 245 **/ 246 247 static void 248 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 249 { 250 struct ip6_tnl __rcu **tp; 251 struct ip6_tnl *iter; 252 253 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 254 (iter = rtnl_dereference(*tp)) != NULL; 255 tp = &iter->next) { 256 if (t == iter) { 257 rcu_assign_pointer(*tp, t->next); 258 break; 259 } 260 } 261 } 262 263 static void ip6_dev_free(struct net_device *dev) 264 { 265 free_percpu(dev->tstats); 266 free_netdev(dev); 267 } 268 269 static int ip6_tnl_create2(struct net_device *dev) 270 { 271 struct ip6_tnl *t = netdev_priv(dev); 272 struct net *net = dev_net(dev); 273 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 274 int err; 275 276 t = netdev_priv(dev); 277 err = ip6_tnl_dev_init(dev); 278 if (err < 0) 279 goto out; 280 281 err = register_netdevice(dev); 282 if (err < 0) 283 goto out; 284 285 strcpy(t->parms.name, dev->name); 286 dev->rtnl_link_ops = &ip6_link_ops; 287 288 dev_hold(dev); 289 ip6_tnl_link(ip6n, t); 290 return 0; 291 292 out: 293 return err; 294 } 295 296 /** 297 * ip6_tnl_create - create a new tunnel 298 * @p: tunnel parameters 299 * @pt: pointer to new tunnel 300 * 301 * Description: 302 * Create tunnel matching given parameters. 303 * 304 * Return: 305 * created tunnel or NULL 306 **/ 307 308 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 309 { 310 struct net_device *dev; 311 struct ip6_tnl *t; 312 char name[IFNAMSIZ]; 313 int err; 314 315 if (p->name[0]) 316 strlcpy(name, p->name, IFNAMSIZ); 317 else 318 sprintf(name, "ip6tnl%%d"); 319 320 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup); 321 if (dev == NULL) 322 goto failed; 323 324 dev_net_set(dev, net); 325 326 t = netdev_priv(dev); 327 t->parms = *p; 328 t->net = dev_net(dev); 329 err = ip6_tnl_create2(dev); 330 if (err < 0) 331 goto failed_free; 332 333 return t; 334 335 failed_free: 336 ip6_dev_free(dev); 337 failed: 338 return NULL; 339 } 340 341 /** 342 * ip6_tnl_locate - find or create tunnel matching given parameters 343 * @p: tunnel parameters 344 * @create: != 0 if allowed to create new tunnel if no match found 345 * 346 * Description: 347 * ip6_tnl_locate() first tries to locate an existing tunnel 348 * based on @parms. If this is unsuccessful, but @create is set a new 349 * tunnel device is created and registered for use. 350 * 351 * Return: 352 * matching tunnel or NULL 353 **/ 354 355 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 356 struct __ip6_tnl_parm *p, int create) 357 { 358 const struct in6_addr *remote = &p->raddr; 359 const struct in6_addr *local = &p->laddr; 360 struct ip6_tnl __rcu **tp; 361 struct ip6_tnl *t; 362 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 363 364 for (tp = ip6_tnl_bucket(ip6n, p); 365 (t = rtnl_dereference(*tp)) != NULL; 366 tp = &t->next) { 367 if (ipv6_addr_equal(local, &t->parms.laddr) && 368 ipv6_addr_equal(remote, &t->parms.raddr)) 369 return t; 370 } 371 if (!create) 372 return NULL; 373 return ip6_tnl_create(net, p); 374 } 375 376 /** 377 * ip6_tnl_dev_uninit - tunnel device uninitializer 378 * @dev: the device to be destroyed 379 * 380 * Description: 381 * ip6_tnl_dev_uninit() removes tunnel from its list 382 **/ 383 384 static void 385 ip6_tnl_dev_uninit(struct net_device *dev) 386 { 387 struct ip6_tnl *t = netdev_priv(dev); 388 struct net *net = t->net; 389 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 390 391 if (dev == ip6n->fb_tnl_dev) 392 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 393 else 394 ip6_tnl_unlink(ip6n, t); 395 ip6_tnl_dst_reset(t); 396 dev_put(dev); 397 } 398 399 /** 400 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 401 * @skb: received socket buffer 402 * 403 * Return: 404 * 0 if none was found, 405 * else index to encapsulation limit 406 **/ 407 408 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw) 409 { 410 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw; 411 __u8 nexthdr = ipv6h->nexthdr; 412 __u16 off = sizeof (*ipv6h); 413 414 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 415 __u16 optlen = 0; 416 struct ipv6_opt_hdr *hdr; 417 if (raw + off + sizeof (*hdr) > skb->data && 418 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 419 break; 420 421 hdr = (struct ipv6_opt_hdr *) (raw + off); 422 if (nexthdr == NEXTHDR_FRAGMENT) { 423 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 424 if (frag_hdr->frag_off) 425 break; 426 optlen = 8; 427 } else if (nexthdr == NEXTHDR_AUTH) { 428 optlen = (hdr->hdrlen + 2) << 2; 429 } else { 430 optlen = ipv6_optlen(hdr); 431 } 432 if (nexthdr == NEXTHDR_DEST) { 433 __u16 i = off + 2; 434 while (1) { 435 struct ipv6_tlv_tnl_enc_lim *tel; 436 437 /* No more room for encapsulation limit */ 438 if (i + sizeof (*tel) > off + optlen) 439 break; 440 441 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 442 /* return index of option if found and valid */ 443 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 444 tel->length == 1) 445 return i; 446 /* else jump to next option */ 447 if (tel->type) 448 i += tel->length + 2; 449 else 450 i++; 451 } 452 } 453 nexthdr = hdr->nexthdr; 454 off += optlen; 455 } 456 return 0; 457 } 458 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim); 459 460 /** 461 * ip6_tnl_err - tunnel error handler 462 * 463 * Description: 464 * ip6_tnl_err() should handle errors in the tunnel according 465 * to the specifications in RFC 2473. 466 **/ 467 468 static int 469 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 470 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 471 { 472 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data; 473 struct ip6_tnl *t; 474 int rel_msg = 0; 475 u8 rel_type = ICMPV6_DEST_UNREACH; 476 u8 rel_code = ICMPV6_ADDR_UNREACH; 477 __u32 rel_info = 0; 478 __u16 len; 479 int err = -ENOENT; 480 481 /* If the packet doesn't contain the original IPv6 header we are 482 in trouble since we might need the source address for further 483 processing of the error. */ 484 485 rcu_read_lock(); 486 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 487 &ipv6h->saddr)) == NULL) 488 goto out; 489 490 if (t->parms.proto != ipproto && t->parms.proto != 0) 491 goto out; 492 493 err = 0; 494 495 switch (*type) { 496 __u32 teli; 497 struct ipv6_tlv_tnl_enc_lim *tel; 498 __u32 mtu; 499 case ICMPV6_DEST_UNREACH: 500 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n", 501 t->parms.name); 502 rel_msg = 1; 503 break; 504 case ICMPV6_TIME_EXCEED: 505 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 506 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 507 t->parms.name); 508 rel_msg = 1; 509 } 510 break; 511 case ICMPV6_PARAMPROB: 512 teli = 0; 513 if ((*code) == ICMPV6_HDR_FIELD) 514 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 515 516 if (teli && teli == *info - 2) { 517 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 518 if (tel->encap_limit == 0) { 519 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 520 t->parms.name); 521 rel_msg = 1; 522 } 523 } else { 524 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 525 t->parms.name); 526 } 527 break; 528 case ICMPV6_PKT_TOOBIG: 529 mtu = *info - offset; 530 if (mtu < IPV6_MIN_MTU) 531 mtu = IPV6_MIN_MTU; 532 t->dev->mtu = mtu; 533 534 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 535 rel_type = ICMPV6_PKT_TOOBIG; 536 rel_code = 0; 537 rel_info = mtu; 538 rel_msg = 1; 539 } 540 break; 541 } 542 543 *type = rel_type; 544 *code = rel_code; 545 *info = rel_info; 546 *msg = rel_msg; 547 548 out: 549 rcu_read_unlock(); 550 return err; 551 } 552 553 static int 554 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 555 u8 type, u8 code, int offset, __be32 info) 556 { 557 int rel_msg = 0; 558 u8 rel_type = type; 559 u8 rel_code = code; 560 __u32 rel_info = ntohl(info); 561 int err; 562 struct sk_buff *skb2; 563 const struct iphdr *eiph; 564 struct rtable *rt; 565 struct flowi4 fl4; 566 567 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 568 &rel_msg, &rel_info, offset); 569 if (err < 0) 570 return err; 571 572 if (rel_msg == 0) 573 return 0; 574 575 switch (rel_type) { 576 case ICMPV6_DEST_UNREACH: 577 if (rel_code != ICMPV6_ADDR_UNREACH) 578 return 0; 579 rel_type = ICMP_DEST_UNREACH; 580 rel_code = ICMP_HOST_UNREACH; 581 break; 582 case ICMPV6_PKT_TOOBIG: 583 if (rel_code != 0) 584 return 0; 585 rel_type = ICMP_DEST_UNREACH; 586 rel_code = ICMP_FRAG_NEEDED; 587 break; 588 case NDISC_REDIRECT: 589 rel_type = ICMP_REDIRECT; 590 rel_code = ICMP_REDIR_HOST; 591 default: 592 return 0; 593 } 594 595 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 596 return 0; 597 598 skb2 = skb_clone(skb, GFP_ATOMIC); 599 if (!skb2) 600 return 0; 601 602 skb_dst_drop(skb2); 603 604 skb_pull(skb2, offset); 605 skb_reset_network_header(skb2); 606 eiph = ip_hdr(skb2); 607 608 /* Try to guess incoming interface */ 609 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 610 eiph->saddr, 0, 611 0, 0, 612 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 613 if (IS_ERR(rt)) 614 goto out; 615 616 skb2->dev = rt->dst.dev; 617 618 /* route "incoming" packet */ 619 if (rt->rt_flags & RTCF_LOCAL) { 620 ip_rt_put(rt); 621 rt = NULL; 622 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 623 eiph->daddr, eiph->saddr, 624 0, 0, 625 IPPROTO_IPIP, 626 RT_TOS(eiph->tos), 0); 627 if (IS_ERR(rt) || 628 rt->dst.dev->type != ARPHRD_TUNNEL) { 629 if (!IS_ERR(rt)) 630 ip_rt_put(rt); 631 goto out; 632 } 633 skb_dst_set(skb2, &rt->dst); 634 } else { 635 ip_rt_put(rt); 636 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 637 skb2->dev) || 638 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 639 goto out; 640 } 641 642 /* change mtu on this route */ 643 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 644 if (rel_info > dst_mtu(skb_dst(skb2))) 645 goto out; 646 647 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info); 648 } 649 if (rel_type == ICMP_REDIRECT) 650 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2); 651 652 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 653 654 out: 655 kfree_skb(skb2); 656 return 0; 657 } 658 659 static int 660 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 661 u8 type, u8 code, int offset, __be32 info) 662 { 663 int rel_msg = 0; 664 u8 rel_type = type; 665 u8 rel_code = code; 666 __u32 rel_info = ntohl(info); 667 int err; 668 669 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 670 &rel_msg, &rel_info, offset); 671 if (err < 0) 672 return err; 673 674 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 675 struct rt6_info *rt; 676 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 677 678 if (!skb2) 679 return 0; 680 681 skb_dst_drop(skb2); 682 skb_pull(skb2, offset); 683 skb_reset_network_header(skb2); 684 685 /* Try to guess incoming interface */ 686 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 687 NULL, 0, 0); 688 689 if (rt && rt->dst.dev) 690 skb2->dev = rt->dst.dev; 691 692 icmpv6_send(skb2, rel_type, rel_code, rel_info); 693 694 ip6_rt_put(rt); 695 696 kfree_skb(skb2); 697 } 698 699 return 0; 700 } 701 702 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 703 const struct ipv6hdr *ipv6h, 704 struct sk_buff *skb) 705 { 706 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 707 708 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 709 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 710 711 return IP6_ECN_decapsulate(ipv6h, skb); 712 } 713 714 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 715 const struct ipv6hdr *ipv6h, 716 struct sk_buff *skb) 717 { 718 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 719 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 720 721 return IP6_ECN_decapsulate(ipv6h, skb); 722 } 723 724 __u32 ip6_tnl_get_cap(struct ip6_tnl *t, 725 const struct in6_addr *laddr, 726 const struct in6_addr *raddr) 727 { 728 struct __ip6_tnl_parm *p = &t->parms; 729 int ltype = ipv6_addr_type(laddr); 730 int rtype = ipv6_addr_type(raddr); 731 __u32 flags = 0; 732 733 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) { 734 flags = IP6_TNL_F_CAP_PER_PACKET; 735 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 736 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 737 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 738 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 739 if (ltype&IPV6_ADDR_UNICAST) 740 flags |= IP6_TNL_F_CAP_XMIT; 741 if (rtype&IPV6_ADDR_UNICAST) 742 flags |= IP6_TNL_F_CAP_RCV; 743 } 744 return flags; 745 } 746 EXPORT_SYMBOL(ip6_tnl_get_cap); 747 748 /* called with rcu_read_lock() */ 749 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, 750 const struct in6_addr *laddr, 751 const struct in6_addr *raddr) 752 { 753 struct __ip6_tnl_parm *p = &t->parms; 754 int ret = 0; 755 struct net *net = t->net; 756 757 if ((p->flags & IP6_TNL_F_CAP_RCV) || 758 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) && 759 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) { 760 struct net_device *ldev = NULL; 761 762 if (p->link) 763 ldev = dev_get_by_index_rcu(net, p->link); 764 765 if ((ipv6_addr_is_multicast(laddr) || 766 likely(ipv6_chk_addr(net, laddr, ldev, 0))) && 767 likely(!ipv6_chk_addr(net, raddr, NULL, 0))) 768 ret = 1; 769 } 770 return ret; 771 } 772 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl); 773 774 /** 775 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 776 * @skb: received socket buffer 777 * @protocol: ethernet protocol ID 778 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 779 * 780 * Return: 0 781 **/ 782 783 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 784 __u8 ipproto, 785 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 786 const struct ipv6hdr *ipv6h, 787 struct sk_buff *skb)) 788 { 789 struct ip6_tnl *t; 790 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 791 int err; 792 793 rcu_read_lock(); 794 795 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 796 &ipv6h->daddr)) != NULL) { 797 struct pcpu_tstats *tstats; 798 799 if (t->parms.proto != ipproto && t->parms.proto != 0) { 800 rcu_read_unlock(); 801 goto discard; 802 } 803 804 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 805 rcu_read_unlock(); 806 goto discard; 807 } 808 809 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 810 t->dev->stats.rx_dropped++; 811 rcu_read_unlock(); 812 goto discard; 813 } 814 skb->mac_header = skb->network_header; 815 skb_reset_network_header(skb); 816 skb->protocol = htons(protocol); 817 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 818 819 __skb_tunnel_rx(skb, t->dev, t->net); 820 821 err = dscp_ecn_decapsulate(t, ipv6h, skb); 822 if (unlikely(err)) { 823 if (log_ecn_error) 824 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n", 825 &ipv6h->saddr, 826 ipv6_get_dsfield(ipv6h)); 827 if (err > 1) { 828 ++t->dev->stats.rx_frame_errors; 829 ++t->dev->stats.rx_errors; 830 rcu_read_unlock(); 831 goto discard; 832 } 833 } 834 835 tstats = this_cpu_ptr(t->dev->tstats); 836 u64_stats_update_begin(&tstats->syncp); 837 tstats->rx_packets++; 838 tstats->rx_bytes += skb->len; 839 u64_stats_update_end(&tstats->syncp); 840 841 netif_rx(skb); 842 843 rcu_read_unlock(); 844 return 0; 845 } 846 rcu_read_unlock(); 847 return 1; 848 849 discard: 850 kfree_skb(skb); 851 return 0; 852 } 853 854 static int ip4ip6_rcv(struct sk_buff *skb) 855 { 856 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 857 ip4ip6_dscp_ecn_decapsulate); 858 } 859 860 static int ip6ip6_rcv(struct sk_buff *skb) 861 { 862 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 863 ip6ip6_dscp_ecn_decapsulate); 864 } 865 866 struct ipv6_tel_txoption { 867 struct ipv6_txoptions ops; 868 __u8 dst_opt[8]; 869 }; 870 871 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 872 { 873 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 874 875 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 876 opt->dst_opt[3] = 1; 877 opt->dst_opt[4] = encap_limit; 878 opt->dst_opt[5] = IPV6_TLV_PADN; 879 opt->dst_opt[6] = 1; 880 881 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 882 opt->ops.opt_nflen = 8; 883 } 884 885 /** 886 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 887 * @t: the outgoing tunnel device 888 * @hdr: IPv6 header from the incoming packet 889 * 890 * Description: 891 * Avoid trivial tunneling loop by checking that tunnel exit-point 892 * doesn't match source of incoming packet. 893 * 894 * Return: 895 * 1 if conflict, 896 * 0 else 897 **/ 898 899 static inline bool 900 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 901 { 902 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 903 } 904 905 int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 906 { 907 struct __ip6_tnl_parm *p = &t->parms; 908 int ret = 0; 909 struct net *net = t->net; 910 911 if (p->flags & IP6_TNL_F_CAP_XMIT) { 912 struct net_device *ldev = NULL; 913 914 rcu_read_lock(); 915 if (p->link) 916 ldev = dev_get_by_index_rcu(net, p->link); 917 918 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 919 pr_warn("%s xmit: Local address not yet configured!\n", 920 p->name); 921 else if (!ipv6_addr_is_multicast(&p->raddr) && 922 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 923 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n", 924 p->name); 925 else 926 ret = 1; 927 rcu_read_unlock(); 928 } 929 return ret; 930 } 931 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl); 932 933 /** 934 * ip6_tnl_xmit2 - encapsulate packet and send 935 * @skb: the outgoing socket buffer 936 * @dev: the outgoing tunnel device 937 * @dsfield: dscp code for outer header 938 * @fl: flow of tunneled packet 939 * @encap_limit: encapsulation limit 940 * @pmtu: Path MTU is stored if packet is too big 941 * 942 * Description: 943 * Build new header and do some sanity checks on the packet before sending 944 * it. 945 * 946 * Return: 947 * 0 on success 948 * -1 fail 949 * %-EMSGSIZE message too big. return mtu in this case. 950 **/ 951 952 static int ip6_tnl_xmit2(struct sk_buff *skb, 953 struct net_device *dev, 954 __u8 dsfield, 955 struct flowi6 *fl6, 956 int encap_limit, 957 __u32 *pmtu) 958 { 959 struct ip6_tnl *t = netdev_priv(dev); 960 struct net *net = t->net; 961 struct net_device_stats *stats = &t->dev->stats; 962 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 963 struct ipv6_tel_txoption opt; 964 struct dst_entry *dst = NULL, *ndst = NULL; 965 struct net_device *tdev; 966 int mtu; 967 unsigned int max_headroom = sizeof(struct ipv6hdr); 968 u8 proto; 969 int err = -1; 970 971 if (!fl6->flowi6_mark) 972 dst = ip6_tnl_dst_check(t); 973 if (!dst) { 974 ndst = ip6_route_output(net, NULL, fl6); 975 976 if (ndst->error) 977 goto tx_err_link_failure; 978 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0); 979 if (IS_ERR(ndst)) { 980 err = PTR_ERR(ndst); 981 ndst = NULL; 982 goto tx_err_link_failure; 983 } 984 dst = ndst; 985 } 986 987 tdev = dst->dev; 988 989 if (tdev == dev) { 990 stats->collisions++; 991 net_warn_ratelimited("%s: Local routing loop detected!\n", 992 t->parms.name); 993 goto tx_err_dst_release; 994 } 995 mtu = dst_mtu(dst) - sizeof (*ipv6h); 996 if (encap_limit >= 0) { 997 max_headroom += 8; 998 mtu -= 8; 999 } 1000 if (mtu < IPV6_MIN_MTU) 1001 mtu = IPV6_MIN_MTU; 1002 if (skb_dst(skb)) 1003 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 1004 if (skb->len > mtu) { 1005 *pmtu = mtu; 1006 err = -EMSGSIZE; 1007 goto tx_err_dst_release; 1008 } 1009 1010 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 1011 1012 /* 1013 * Okay, now see if we can stuff it in the buffer as-is. 1014 */ 1015 max_headroom += LL_RESERVED_SPACE(tdev); 1016 1017 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1018 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1019 struct sk_buff *new_skb; 1020 1021 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 1022 goto tx_err_dst_release; 1023 1024 if (skb->sk) 1025 skb_set_owner_w(new_skb, skb->sk); 1026 consume_skb(skb); 1027 skb = new_skb; 1028 } 1029 if (fl6->flowi6_mark) { 1030 skb_dst_set(skb, dst); 1031 ndst = NULL; 1032 } else { 1033 skb_dst_set_noref(skb, dst); 1034 } 1035 skb->transport_header = skb->network_header; 1036 1037 proto = fl6->flowi6_proto; 1038 if (encap_limit >= 0) { 1039 init_tel_txopt(&opt, encap_limit); 1040 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 1041 } 1042 1043 if (likely(!skb->encapsulation)) { 1044 skb_reset_inner_headers(skb); 1045 skb->encapsulation = 1; 1046 } 1047 1048 skb_push(skb, sizeof(struct ipv6hdr)); 1049 skb_reset_network_header(skb); 1050 ipv6h = ipv6_hdr(skb); 1051 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel); 1052 ipv6h->hop_limit = t->parms.hop_limit; 1053 ipv6h->nexthdr = proto; 1054 ipv6h->saddr = fl6->saddr; 1055 ipv6h->daddr = fl6->daddr; 1056 ip6tunnel_xmit(skb, dev); 1057 if (ndst) 1058 ip6_tnl_dst_store(t, ndst); 1059 return 0; 1060 tx_err_link_failure: 1061 stats->tx_carrier_errors++; 1062 dst_link_failure(skb); 1063 tx_err_dst_release: 1064 dst_release(ndst); 1065 return err; 1066 } 1067 1068 static inline int 1069 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1070 { 1071 struct ip6_tnl *t = netdev_priv(dev); 1072 const struct iphdr *iph = ip_hdr(skb); 1073 int encap_limit = -1; 1074 struct flowi6 fl6; 1075 __u8 dsfield; 1076 __u32 mtu; 1077 int err; 1078 1079 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 1080 !ip6_tnl_xmit_ctl(t)) 1081 return -1; 1082 1083 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1084 encap_limit = t->parms.encap_limit; 1085 1086 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1087 fl6.flowi6_proto = IPPROTO_IPIP; 1088 1089 dsfield = ipv4_get_dsfield(iph); 1090 1091 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1092 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 1093 & IPV6_TCLASS_MASK; 1094 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1095 fl6.flowi6_mark = skb->mark; 1096 1097 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1098 if (err != 0) { 1099 /* XXX: send ICMP error even if DF is not set. */ 1100 if (err == -EMSGSIZE) 1101 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1102 htonl(mtu)); 1103 return -1; 1104 } 1105 1106 return 0; 1107 } 1108 1109 static inline int 1110 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1111 { 1112 struct ip6_tnl *t = netdev_priv(dev); 1113 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1114 int encap_limit = -1; 1115 __u16 offset; 1116 struct flowi6 fl6; 1117 __u8 dsfield; 1118 __u32 mtu; 1119 int err; 1120 1121 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1122 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1123 return -1; 1124 1125 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 1126 if (offset > 0) { 1127 struct ipv6_tlv_tnl_enc_lim *tel; 1128 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1129 if (tel->encap_limit == 0) { 1130 icmpv6_send(skb, ICMPV6_PARAMPROB, 1131 ICMPV6_HDR_FIELD, offset + 2); 1132 return -1; 1133 } 1134 encap_limit = tel->encap_limit - 1; 1135 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1136 encap_limit = t->parms.encap_limit; 1137 1138 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1139 fl6.flowi6_proto = IPPROTO_IPV6; 1140 1141 dsfield = ipv6_get_dsfield(ipv6h); 1142 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1143 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1144 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1145 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); 1146 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1147 fl6.flowi6_mark = skb->mark; 1148 1149 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1150 if (err != 0) { 1151 if (err == -EMSGSIZE) 1152 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1153 return -1; 1154 } 1155 1156 return 0; 1157 } 1158 1159 static netdev_tx_t 1160 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1161 { 1162 struct ip6_tnl *t = netdev_priv(dev); 1163 struct net_device_stats *stats = &t->dev->stats; 1164 int ret; 1165 1166 switch (skb->protocol) { 1167 case htons(ETH_P_IP): 1168 ret = ip4ip6_tnl_xmit(skb, dev); 1169 break; 1170 case htons(ETH_P_IPV6): 1171 ret = ip6ip6_tnl_xmit(skb, dev); 1172 break; 1173 default: 1174 goto tx_err; 1175 } 1176 1177 if (ret < 0) 1178 goto tx_err; 1179 1180 return NETDEV_TX_OK; 1181 1182 tx_err: 1183 stats->tx_errors++; 1184 stats->tx_dropped++; 1185 kfree_skb(skb); 1186 return NETDEV_TX_OK; 1187 } 1188 1189 static void ip6_tnl_link_config(struct ip6_tnl *t) 1190 { 1191 struct net_device *dev = t->dev; 1192 struct __ip6_tnl_parm *p = &t->parms; 1193 struct flowi6 *fl6 = &t->fl.u.ip6; 1194 1195 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1196 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1197 1198 /* Set up flowi template */ 1199 fl6->saddr = p->laddr; 1200 fl6->daddr = p->raddr; 1201 fl6->flowi6_oif = p->link; 1202 fl6->flowlabel = 0; 1203 1204 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1205 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1206 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1207 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1208 1209 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1210 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1211 1212 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1213 dev->flags |= IFF_POINTOPOINT; 1214 else 1215 dev->flags &= ~IFF_POINTOPOINT; 1216 1217 dev->iflink = p->link; 1218 1219 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1220 int strict = (ipv6_addr_type(&p->raddr) & 1221 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1222 1223 struct rt6_info *rt = rt6_lookup(t->net, 1224 &p->raddr, &p->laddr, 1225 p->link, strict); 1226 1227 if (rt == NULL) 1228 return; 1229 1230 if (rt->dst.dev) { 1231 dev->hard_header_len = rt->dst.dev->hard_header_len + 1232 sizeof (struct ipv6hdr); 1233 1234 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr); 1235 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1236 dev->mtu-=8; 1237 1238 if (dev->mtu < IPV6_MIN_MTU) 1239 dev->mtu = IPV6_MIN_MTU; 1240 } 1241 ip6_rt_put(rt); 1242 } 1243 } 1244 1245 /** 1246 * ip6_tnl_change - update the tunnel parameters 1247 * @t: tunnel to be changed 1248 * @p: tunnel configuration parameters 1249 * 1250 * Description: 1251 * ip6_tnl_change() updates the tunnel parameters 1252 **/ 1253 1254 static int 1255 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1256 { 1257 t->parms.laddr = p->laddr; 1258 t->parms.raddr = p->raddr; 1259 t->parms.flags = p->flags; 1260 t->parms.hop_limit = p->hop_limit; 1261 t->parms.encap_limit = p->encap_limit; 1262 t->parms.flowinfo = p->flowinfo; 1263 t->parms.link = p->link; 1264 t->parms.proto = p->proto; 1265 ip6_tnl_dst_reset(t); 1266 ip6_tnl_link_config(t); 1267 return 0; 1268 } 1269 1270 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1271 { 1272 struct net *net = t->net; 1273 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1274 int err; 1275 1276 ip6_tnl_unlink(ip6n, t); 1277 synchronize_net(); 1278 err = ip6_tnl_change(t, p); 1279 ip6_tnl_link(ip6n, t); 1280 netdev_state_change(t->dev); 1281 return err; 1282 } 1283 1284 static void 1285 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1286 { 1287 p->laddr = u->laddr; 1288 p->raddr = u->raddr; 1289 p->flags = u->flags; 1290 p->hop_limit = u->hop_limit; 1291 p->encap_limit = u->encap_limit; 1292 p->flowinfo = u->flowinfo; 1293 p->link = u->link; 1294 p->proto = u->proto; 1295 memcpy(p->name, u->name, sizeof(u->name)); 1296 } 1297 1298 static void 1299 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1300 { 1301 u->laddr = p->laddr; 1302 u->raddr = p->raddr; 1303 u->flags = p->flags; 1304 u->hop_limit = p->hop_limit; 1305 u->encap_limit = p->encap_limit; 1306 u->flowinfo = p->flowinfo; 1307 u->link = p->link; 1308 u->proto = p->proto; 1309 memcpy(u->name, p->name, sizeof(u->name)); 1310 } 1311 1312 /** 1313 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1314 * @dev: virtual device associated with tunnel 1315 * @ifr: parameters passed from userspace 1316 * @cmd: command to be performed 1317 * 1318 * Description: 1319 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1320 * from userspace. 1321 * 1322 * The possible commands are the following: 1323 * %SIOCGETTUNNEL: get tunnel parameters for device 1324 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1325 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1326 * %SIOCDELTUNNEL: delete tunnel 1327 * 1328 * The fallback device "ip6tnl0", created during module 1329 * initialization, can be used for creating other tunnel devices. 1330 * 1331 * Return: 1332 * 0 on success, 1333 * %-EFAULT if unable to copy data to or from userspace, 1334 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1335 * %-EINVAL if passed tunnel parameters are invalid, 1336 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1337 * %-ENODEV if attempting to change or delete a nonexisting device 1338 **/ 1339 1340 static int 1341 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1342 { 1343 int err = 0; 1344 struct ip6_tnl_parm p; 1345 struct __ip6_tnl_parm p1; 1346 struct ip6_tnl *t = NULL; 1347 struct net *net = dev_net(dev); 1348 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1349 1350 switch (cmd) { 1351 case SIOCGETTUNNEL: 1352 if (dev == ip6n->fb_tnl_dev) { 1353 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1354 err = -EFAULT; 1355 break; 1356 } 1357 ip6_tnl_parm_from_user(&p1, &p); 1358 t = ip6_tnl_locate(net, &p1, 0); 1359 } else { 1360 memset(&p, 0, sizeof(p)); 1361 } 1362 if (t == NULL) 1363 t = netdev_priv(dev); 1364 ip6_tnl_parm_to_user(&p, &t->parms); 1365 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1366 err = -EFAULT; 1367 } 1368 break; 1369 case SIOCADDTUNNEL: 1370 case SIOCCHGTUNNEL: 1371 err = -EPERM; 1372 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1373 break; 1374 err = -EFAULT; 1375 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1376 break; 1377 err = -EINVAL; 1378 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1379 p.proto != 0) 1380 break; 1381 ip6_tnl_parm_from_user(&p1, &p); 1382 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1383 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1384 if (t != NULL) { 1385 if (t->dev != dev) { 1386 err = -EEXIST; 1387 break; 1388 } 1389 } else 1390 t = netdev_priv(dev); 1391 1392 err = ip6_tnl_update(t, &p1); 1393 } 1394 if (t) { 1395 err = 0; 1396 ip6_tnl_parm_to_user(&p, &t->parms); 1397 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1398 err = -EFAULT; 1399 1400 } else 1401 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1402 break; 1403 case SIOCDELTUNNEL: 1404 err = -EPERM; 1405 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1406 break; 1407 1408 if (dev == ip6n->fb_tnl_dev) { 1409 err = -EFAULT; 1410 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1411 break; 1412 err = -ENOENT; 1413 ip6_tnl_parm_from_user(&p1, &p); 1414 t = ip6_tnl_locate(net, &p1, 0); 1415 if (t == NULL) 1416 break; 1417 err = -EPERM; 1418 if (t->dev == ip6n->fb_tnl_dev) 1419 break; 1420 dev = t->dev; 1421 } 1422 err = 0; 1423 unregister_netdevice(dev); 1424 break; 1425 default: 1426 err = -EINVAL; 1427 } 1428 return err; 1429 } 1430 1431 /** 1432 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1433 * @dev: virtual device associated with tunnel 1434 * @new_mtu: the new mtu 1435 * 1436 * Return: 1437 * 0 on success, 1438 * %-EINVAL if mtu too small 1439 **/ 1440 1441 static int 1442 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1443 { 1444 struct ip6_tnl *tnl = netdev_priv(dev); 1445 1446 if (tnl->parms.proto == IPPROTO_IPIP) { 1447 if (new_mtu < 68) 1448 return -EINVAL; 1449 } else { 1450 if (new_mtu < IPV6_MIN_MTU) 1451 return -EINVAL; 1452 } 1453 if (new_mtu > 0xFFF8 - dev->hard_header_len) 1454 return -EINVAL; 1455 dev->mtu = new_mtu; 1456 return 0; 1457 } 1458 1459 1460 static const struct net_device_ops ip6_tnl_netdev_ops = { 1461 .ndo_uninit = ip6_tnl_dev_uninit, 1462 .ndo_start_xmit = ip6_tnl_xmit, 1463 .ndo_do_ioctl = ip6_tnl_ioctl, 1464 .ndo_change_mtu = ip6_tnl_change_mtu, 1465 .ndo_get_stats = ip6_get_stats, 1466 }; 1467 1468 1469 /** 1470 * ip6_tnl_dev_setup - setup virtual tunnel device 1471 * @dev: virtual device associated with tunnel 1472 * 1473 * Description: 1474 * Initialize function pointers and device parameters 1475 **/ 1476 1477 static void ip6_tnl_dev_setup(struct net_device *dev) 1478 { 1479 struct ip6_tnl *t; 1480 1481 dev->netdev_ops = &ip6_tnl_netdev_ops; 1482 dev->destructor = ip6_dev_free; 1483 1484 dev->type = ARPHRD_TUNNEL6; 1485 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1486 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1487 t = netdev_priv(dev); 1488 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1489 dev->mtu-=8; 1490 dev->flags |= IFF_NOARP; 1491 dev->addr_len = sizeof(struct in6_addr); 1492 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1493 /* This perm addr will be used as interface identifier by IPv6 */ 1494 dev->addr_assign_type = NET_ADDR_RANDOM; 1495 eth_random_addr(dev->perm_addr); 1496 } 1497 1498 1499 /** 1500 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1501 * @dev: virtual device associated with tunnel 1502 **/ 1503 1504 static inline int 1505 ip6_tnl_dev_init_gen(struct net_device *dev) 1506 { 1507 struct ip6_tnl *t = netdev_priv(dev); 1508 int i; 1509 1510 t->dev = dev; 1511 t->net = dev_net(dev); 1512 dev->tstats = alloc_percpu(struct pcpu_tstats); 1513 if (!dev->tstats) 1514 return -ENOMEM; 1515 1516 for_each_possible_cpu(i) { 1517 struct pcpu_tstats *ip6_tnl_stats; 1518 ip6_tnl_stats = per_cpu_ptr(dev->tstats, i); 1519 u64_stats_init(&ip6_tnl_stats->syncp); 1520 } 1521 return 0; 1522 } 1523 1524 /** 1525 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1526 * @dev: virtual device associated with tunnel 1527 **/ 1528 1529 static int ip6_tnl_dev_init(struct net_device *dev) 1530 { 1531 struct ip6_tnl *t = netdev_priv(dev); 1532 int err = ip6_tnl_dev_init_gen(dev); 1533 1534 if (err) 1535 return err; 1536 ip6_tnl_link_config(t); 1537 return 0; 1538 } 1539 1540 /** 1541 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1542 * @dev: fallback device 1543 * 1544 * Return: 0 1545 **/ 1546 1547 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1548 { 1549 struct ip6_tnl *t = netdev_priv(dev); 1550 struct net *net = dev_net(dev); 1551 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1552 int err = ip6_tnl_dev_init_gen(dev); 1553 1554 if (err) 1555 return err; 1556 1557 t->parms.proto = IPPROTO_IPV6; 1558 dev_hold(dev); 1559 1560 ip6_tnl_link_config(t); 1561 1562 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1563 return 0; 1564 } 1565 1566 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[]) 1567 { 1568 u8 proto; 1569 1570 if (!data) 1571 return 0; 1572 1573 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1574 if (proto != IPPROTO_IPV6 && 1575 proto != IPPROTO_IPIP && 1576 proto != 0) 1577 return -EINVAL; 1578 1579 return 0; 1580 } 1581 1582 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1583 struct __ip6_tnl_parm *parms) 1584 { 1585 memset(parms, 0, sizeof(*parms)); 1586 1587 if (!data) 1588 return; 1589 1590 if (data[IFLA_IPTUN_LINK]) 1591 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1592 1593 if (data[IFLA_IPTUN_LOCAL]) 1594 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL], 1595 sizeof(struct in6_addr)); 1596 1597 if (data[IFLA_IPTUN_REMOTE]) 1598 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE], 1599 sizeof(struct in6_addr)); 1600 1601 if (data[IFLA_IPTUN_TTL]) 1602 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1603 1604 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1605 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1606 1607 if (data[IFLA_IPTUN_FLOWINFO]) 1608 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1609 1610 if (data[IFLA_IPTUN_FLAGS]) 1611 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1612 1613 if (data[IFLA_IPTUN_PROTO]) 1614 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1615 } 1616 1617 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 1618 struct nlattr *tb[], struct nlattr *data[]) 1619 { 1620 struct net *net = dev_net(dev); 1621 struct ip6_tnl *nt; 1622 1623 nt = netdev_priv(dev); 1624 ip6_tnl_netlink_parms(data, &nt->parms); 1625 1626 if (ip6_tnl_locate(net, &nt->parms, 0)) 1627 return -EEXIST; 1628 1629 return ip6_tnl_create2(dev); 1630 } 1631 1632 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 1633 struct nlattr *data[]) 1634 { 1635 struct ip6_tnl *t = netdev_priv(dev); 1636 struct __ip6_tnl_parm p; 1637 struct net *net = t->net; 1638 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1639 1640 if (dev == ip6n->fb_tnl_dev) 1641 return -EINVAL; 1642 1643 ip6_tnl_netlink_parms(data, &p); 1644 1645 t = ip6_tnl_locate(net, &p, 0); 1646 1647 if (t) { 1648 if (t->dev != dev) 1649 return -EEXIST; 1650 } else 1651 t = netdev_priv(dev); 1652 1653 return ip6_tnl_update(t, &p); 1654 } 1655 1656 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 1657 { 1658 struct net *net = dev_net(dev); 1659 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1660 1661 if (dev != ip6n->fb_tnl_dev) 1662 unregister_netdevice_queue(dev, head); 1663 } 1664 1665 static size_t ip6_tnl_get_size(const struct net_device *dev) 1666 { 1667 return 1668 /* IFLA_IPTUN_LINK */ 1669 nla_total_size(4) + 1670 /* IFLA_IPTUN_LOCAL */ 1671 nla_total_size(sizeof(struct in6_addr)) + 1672 /* IFLA_IPTUN_REMOTE */ 1673 nla_total_size(sizeof(struct in6_addr)) + 1674 /* IFLA_IPTUN_TTL */ 1675 nla_total_size(1) + 1676 /* IFLA_IPTUN_ENCAP_LIMIT */ 1677 nla_total_size(1) + 1678 /* IFLA_IPTUN_FLOWINFO */ 1679 nla_total_size(4) + 1680 /* IFLA_IPTUN_FLAGS */ 1681 nla_total_size(4) + 1682 /* IFLA_IPTUN_PROTO */ 1683 nla_total_size(1) + 1684 0; 1685 } 1686 1687 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 1688 { 1689 struct ip6_tnl *tunnel = netdev_priv(dev); 1690 struct __ip6_tnl_parm *parm = &tunnel->parms; 1691 1692 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1693 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr), 1694 &parm->laddr) || 1695 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr), 1696 &parm->raddr) || 1697 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 1698 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 1699 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 1700 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 1701 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto)) 1702 goto nla_put_failure; 1703 return 0; 1704 1705 nla_put_failure: 1706 return -EMSGSIZE; 1707 } 1708 1709 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 1710 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1711 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 1712 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 1713 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1714 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 1715 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 1716 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 1717 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1718 }; 1719 1720 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 1721 .kind = "ip6tnl", 1722 .maxtype = IFLA_IPTUN_MAX, 1723 .policy = ip6_tnl_policy, 1724 .priv_size = sizeof(struct ip6_tnl), 1725 .setup = ip6_tnl_dev_setup, 1726 .validate = ip6_tnl_validate, 1727 .newlink = ip6_tnl_newlink, 1728 .changelink = ip6_tnl_changelink, 1729 .dellink = ip6_tnl_dellink, 1730 .get_size = ip6_tnl_get_size, 1731 .fill_info = ip6_tnl_fill_info, 1732 }; 1733 1734 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 1735 .handler = ip4ip6_rcv, 1736 .err_handler = ip4ip6_err, 1737 .priority = 1, 1738 }; 1739 1740 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 1741 .handler = ip6ip6_rcv, 1742 .err_handler = ip6ip6_err, 1743 .priority = 1, 1744 }; 1745 1746 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net) 1747 { 1748 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1749 struct net_device *dev, *aux; 1750 int h; 1751 struct ip6_tnl *t; 1752 LIST_HEAD(list); 1753 1754 for_each_netdev_safe(net, dev, aux) 1755 if (dev->rtnl_link_ops == &ip6_link_ops) 1756 unregister_netdevice_queue(dev, &list); 1757 1758 for (h = 0; h < HASH_SIZE; h++) { 1759 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1760 while (t != NULL) { 1761 /* If dev is in the same netns, it has already 1762 * been added to the list by the previous loop. 1763 */ 1764 if (!net_eq(dev_net(t->dev), net)) 1765 unregister_netdevice_queue(t->dev, &list); 1766 t = rtnl_dereference(t->next); 1767 } 1768 } 1769 1770 unregister_netdevice_many(&list); 1771 } 1772 1773 static int __net_init ip6_tnl_init_net(struct net *net) 1774 { 1775 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1776 struct ip6_tnl *t = NULL; 1777 int err; 1778 1779 ip6n->tnls[0] = ip6n->tnls_wc; 1780 ip6n->tnls[1] = ip6n->tnls_r_l; 1781 1782 err = -ENOMEM; 1783 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1784 ip6_tnl_dev_setup); 1785 1786 if (!ip6n->fb_tnl_dev) 1787 goto err_alloc_dev; 1788 dev_net_set(ip6n->fb_tnl_dev, net); 1789 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 1790 /* FB netdevice is special: we have one, and only one per netns. 1791 * Allowing to move it to another netns is clearly unsafe. 1792 */ 1793 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 1794 1795 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1796 if (err < 0) 1797 goto err_register; 1798 1799 err = register_netdev(ip6n->fb_tnl_dev); 1800 if (err < 0) 1801 goto err_register; 1802 1803 t = netdev_priv(ip6n->fb_tnl_dev); 1804 1805 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1806 return 0; 1807 1808 err_register: 1809 ip6_dev_free(ip6n->fb_tnl_dev); 1810 err_alloc_dev: 1811 return err; 1812 } 1813 1814 static void __net_exit ip6_tnl_exit_net(struct net *net) 1815 { 1816 rtnl_lock(); 1817 ip6_tnl_destroy_tunnels(net); 1818 rtnl_unlock(); 1819 } 1820 1821 static struct pernet_operations ip6_tnl_net_ops = { 1822 .init = ip6_tnl_init_net, 1823 .exit = ip6_tnl_exit_net, 1824 .id = &ip6_tnl_net_id, 1825 .size = sizeof(struct ip6_tnl_net), 1826 }; 1827 1828 /** 1829 * ip6_tunnel_init - register protocol and reserve needed resources 1830 * 1831 * Return: 0 on success 1832 **/ 1833 1834 static int __init ip6_tunnel_init(void) 1835 { 1836 int err; 1837 1838 err = register_pernet_device(&ip6_tnl_net_ops); 1839 if (err < 0) 1840 goto out_pernet; 1841 1842 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 1843 if (err < 0) { 1844 pr_err("%s: can't register ip4ip6\n", __func__); 1845 goto out_ip4ip6; 1846 } 1847 1848 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 1849 if (err < 0) { 1850 pr_err("%s: can't register ip6ip6\n", __func__); 1851 goto out_ip6ip6; 1852 } 1853 err = rtnl_link_register(&ip6_link_ops); 1854 if (err < 0) 1855 goto rtnl_link_failed; 1856 1857 return 0; 1858 1859 rtnl_link_failed: 1860 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 1861 out_ip6ip6: 1862 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1863 out_ip4ip6: 1864 unregister_pernet_device(&ip6_tnl_net_ops); 1865 out_pernet: 1866 return err; 1867 } 1868 1869 /** 1870 * ip6_tunnel_cleanup - free resources and unregister protocol 1871 **/ 1872 1873 static void __exit ip6_tunnel_cleanup(void) 1874 { 1875 rtnl_link_unregister(&ip6_link_ops); 1876 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1877 pr_info("%s: can't deregister ip4ip6\n", __func__); 1878 1879 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1880 pr_info("%s: can't deregister ip6ip6\n", __func__); 1881 1882 unregister_pernet_device(&ip6_tnl_net_ops); 1883 } 1884 1885 module_init(ip6_tunnel_init); 1886 module_exit(ip6_tunnel_cleanup); 1887