1 /* 2 * IPv6 tunneling device 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Ville Nuorvala <vnuorval@tcs.hut.fi> 7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org> 8 * 9 * Based on: 10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c 11 * 12 * RFC 2473 13 * 14 * This program is free software; you can redistribute it and/or 15 * modify it under the terms of the GNU General Public License 16 * as published by the Free Software Foundation; either version 17 * 2 of the License, or (at your option) any later version. 18 * 19 */ 20 21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 22 23 #include <linux/module.h> 24 #include <linux/capability.h> 25 #include <linux/errno.h> 26 #include <linux/types.h> 27 #include <linux/sockios.h> 28 #include <linux/icmp.h> 29 #include <linux/if.h> 30 #include <linux/in.h> 31 #include <linux/ip.h> 32 #include <linux/net.h> 33 #include <linux/in6.h> 34 #include <linux/netdevice.h> 35 #include <linux/if_arp.h> 36 #include <linux/icmpv6.h> 37 #include <linux/init.h> 38 #include <linux/route.h> 39 #include <linux/rtnetlink.h> 40 #include <linux/netfilter_ipv6.h> 41 #include <linux/slab.h> 42 #include <linux/hash.h> 43 #include <linux/etherdevice.h> 44 45 #include <asm/uaccess.h> 46 #include <linux/atomic.h> 47 48 #include <net/icmp.h> 49 #include <net/ip.h> 50 #include <net/ip_tunnels.h> 51 #include <net/ipv6.h> 52 #include <net/ip6_route.h> 53 #include <net/addrconf.h> 54 #include <net/ip6_tunnel.h> 55 #include <net/xfrm.h> 56 #include <net/dsfield.h> 57 #include <net/inet_ecn.h> 58 #include <net/net_namespace.h> 59 #include <net/netns/generic.h> 60 61 MODULE_AUTHOR("Ville Nuorvala"); 62 MODULE_DESCRIPTION("IPv6 tunneling device"); 63 MODULE_LICENSE("GPL"); 64 MODULE_ALIAS_RTNL_LINK("ip6tnl"); 65 MODULE_ALIAS_NETDEV("ip6tnl0"); 66 67 #ifdef IP6_TNL_DEBUG 68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__) 69 #else 70 #define IP6_TNL_TRACE(x...) do {;} while(0) 71 #endif 72 73 #define HASH_SIZE_SHIFT 5 74 #define HASH_SIZE (1 << HASH_SIZE_SHIFT) 75 76 static bool log_ecn_error = true; 77 module_param(log_ecn_error, bool, 0644); 78 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 79 80 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 81 { 82 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 83 84 return hash_32(hash, HASH_SIZE_SHIFT); 85 } 86 87 static int ip6_tnl_dev_init(struct net_device *dev); 88 static void ip6_tnl_dev_setup(struct net_device *dev); 89 static struct rtnl_link_ops ip6_link_ops __read_mostly; 90 91 static int ip6_tnl_net_id __read_mostly; 92 struct ip6_tnl_net { 93 /* the IPv6 tunnel fallback device */ 94 struct net_device *fb_tnl_dev; 95 /* lists for storing tunnels in use */ 96 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; 97 struct ip6_tnl __rcu *tnls_wc[1]; 98 struct ip6_tnl __rcu **tnls[2]; 99 }; 100 101 static struct net_device_stats *ip6_get_stats(struct net_device *dev) 102 { 103 struct pcpu_sw_netstats tmp, sum = { 0 }; 104 int i; 105 106 for_each_possible_cpu(i) { 107 unsigned int start; 108 const struct pcpu_sw_netstats *tstats = 109 per_cpu_ptr(dev->tstats, i); 110 111 do { 112 start = u64_stats_fetch_begin_irq(&tstats->syncp); 113 tmp.rx_packets = tstats->rx_packets; 114 tmp.rx_bytes = tstats->rx_bytes; 115 tmp.tx_packets = tstats->tx_packets; 116 tmp.tx_bytes = tstats->tx_bytes; 117 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start)); 118 119 sum.rx_packets += tmp.rx_packets; 120 sum.rx_bytes += tmp.rx_bytes; 121 sum.tx_packets += tmp.tx_packets; 122 sum.tx_bytes += tmp.tx_bytes; 123 } 124 dev->stats.rx_packets = sum.rx_packets; 125 dev->stats.rx_bytes = sum.rx_bytes; 126 dev->stats.tx_packets = sum.tx_packets; 127 dev->stats.tx_bytes = sum.tx_bytes; 128 return &dev->stats; 129 } 130 131 /* 132 * Locking : hash tables are protected by RCU and RTNL 133 */ 134 135 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 136 { 137 struct dst_entry *dst = t->dst_cache; 138 139 if (dst && dst->obsolete && 140 dst->ops->check(dst, t->dst_cookie) == NULL) { 141 t->dst_cache = NULL; 142 dst_release(dst); 143 return NULL; 144 } 145 146 return dst; 147 } 148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check); 149 150 void ip6_tnl_dst_reset(struct ip6_tnl *t) 151 { 152 dst_release(t->dst_cache); 153 t->dst_cache = NULL; 154 } 155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset); 156 157 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 158 { 159 struct rt6_info *rt = (struct rt6_info *) dst; 160 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 161 dst_release(t->dst_cache); 162 t->dst_cache = dst; 163 } 164 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store); 165 166 /** 167 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 168 * @remote: the address of the tunnel exit-point 169 * @local: the address of the tunnel entry-point 170 * 171 * Return: 172 * tunnel matching given end-points if found, 173 * else fallback tunnel if its device is up, 174 * else %NULL 175 **/ 176 177 #define for_each_ip6_tunnel_rcu(start) \ 178 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 179 180 static struct ip6_tnl * 181 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local) 182 { 183 unsigned int hash = HASH(remote, local); 184 struct ip6_tnl *t; 185 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 186 187 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 188 if (ipv6_addr_equal(local, &t->parms.laddr) && 189 ipv6_addr_equal(remote, &t->parms.raddr) && 190 (t->dev->flags & IFF_UP)) 191 return t; 192 } 193 t = rcu_dereference(ip6n->tnls_wc[0]); 194 if (t && (t->dev->flags & IFF_UP)) 195 return t; 196 197 return NULL; 198 } 199 200 /** 201 * ip6_tnl_bucket - get head of list matching given tunnel parameters 202 * @p: parameters containing tunnel end-points 203 * 204 * Description: 205 * ip6_tnl_bucket() returns the head of the list matching the 206 * &struct in6_addr entries laddr and raddr in @p. 207 * 208 * Return: head of IPv6 tunnel list 209 **/ 210 211 static struct ip6_tnl __rcu ** 212 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p) 213 { 214 const struct in6_addr *remote = &p->raddr; 215 const struct in6_addr *local = &p->laddr; 216 unsigned int h = 0; 217 int prio = 0; 218 219 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 220 prio = 1; 221 h = HASH(remote, local); 222 } 223 return &ip6n->tnls[prio][h]; 224 } 225 226 /** 227 * ip6_tnl_link - add tunnel to hash table 228 * @t: tunnel to be added 229 **/ 230 231 static void 232 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 233 { 234 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 235 236 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 237 rcu_assign_pointer(*tp, t); 238 } 239 240 /** 241 * ip6_tnl_unlink - remove tunnel from hash table 242 * @t: tunnel to be removed 243 **/ 244 245 static void 246 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 247 { 248 struct ip6_tnl __rcu **tp; 249 struct ip6_tnl *iter; 250 251 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 252 (iter = rtnl_dereference(*tp)) != NULL; 253 tp = &iter->next) { 254 if (t == iter) { 255 rcu_assign_pointer(*tp, t->next); 256 break; 257 } 258 } 259 } 260 261 static void ip6_dev_free(struct net_device *dev) 262 { 263 free_percpu(dev->tstats); 264 free_netdev(dev); 265 } 266 267 static int ip6_tnl_create2(struct net_device *dev) 268 { 269 struct ip6_tnl *t = netdev_priv(dev); 270 struct net *net = dev_net(dev); 271 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 272 int err; 273 274 t = netdev_priv(dev); 275 276 err = register_netdevice(dev); 277 if (err < 0) 278 goto out; 279 280 strcpy(t->parms.name, dev->name); 281 dev->rtnl_link_ops = &ip6_link_ops; 282 283 dev_hold(dev); 284 ip6_tnl_link(ip6n, t); 285 return 0; 286 287 out: 288 return err; 289 } 290 291 /** 292 * ip6_tnl_create - create a new tunnel 293 * @p: tunnel parameters 294 * @pt: pointer to new tunnel 295 * 296 * Description: 297 * Create tunnel matching given parameters. 298 * 299 * Return: 300 * created tunnel or NULL 301 **/ 302 303 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 304 { 305 struct net_device *dev; 306 struct ip6_tnl *t; 307 char name[IFNAMSIZ]; 308 int err; 309 310 if (p->name[0]) 311 strlcpy(name, p->name, IFNAMSIZ); 312 else 313 sprintf(name, "ip6tnl%%d"); 314 315 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 316 ip6_tnl_dev_setup); 317 if (dev == NULL) 318 goto failed; 319 320 dev_net_set(dev, net); 321 322 t = netdev_priv(dev); 323 t->parms = *p; 324 t->net = dev_net(dev); 325 err = ip6_tnl_create2(dev); 326 if (err < 0) 327 goto failed_free; 328 329 return t; 330 331 failed_free: 332 ip6_dev_free(dev); 333 failed: 334 return NULL; 335 } 336 337 /** 338 * ip6_tnl_locate - find or create tunnel matching given parameters 339 * @p: tunnel parameters 340 * @create: != 0 if allowed to create new tunnel if no match found 341 * 342 * Description: 343 * ip6_tnl_locate() first tries to locate an existing tunnel 344 * based on @parms. If this is unsuccessful, but @create is set a new 345 * tunnel device is created and registered for use. 346 * 347 * Return: 348 * matching tunnel or NULL 349 **/ 350 351 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 352 struct __ip6_tnl_parm *p, int create) 353 { 354 const struct in6_addr *remote = &p->raddr; 355 const struct in6_addr *local = &p->laddr; 356 struct ip6_tnl __rcu **tp; 357 struct ip6_tnl *t; 358 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 359 360 for (tp = ip6_tnl_bucket(ip6n, p); 361 (t = rtnl_dereference(*tp)) != NULL; 362 tp = &t->next) { 363 if (ipv6_addr_equal(local, &t->parms.laddr) && 364 ipv6_addr_equal(remote, &t->parms.raddr)) { 365 if (create) 366 return NULL; 367 368 return t; 369 } 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_sw_netstats *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), 1052 ip6_make_flowlabel(net, skb, fl6->flowlabel, false)); 1053 ipv6h->hop_limit = t->parms.hop_limit; 1054 ipv6h->nexthdr = proto; 1055 ipv6h->saddr = fl6->saddr; 1056 ipv6h->daddr = fl6->daddr; 1057 ip6tunnel_xmit(skb, dev); 1058 if (ndst) 1059 ip6_tnl_dst_store(t, ndst); 1060 return 0; 1061 tx_err_link_failure: 1062 stats->tx_carrier_errors++; 1063 dst_link_failure(skb); 1064 tx_err_dst_release: 1065 dst_release(ndst); 1066 return err; 1067 } 1068 1069 static inline int 1070 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1071 { 1072 struct ip6_tnl *t = netdev_priv(dev); 1073 const struct iphdr *iph = ip_hdr(skb); 1074 int encap_limit = -1; 1075 struct flowi6 fl6; 1076 __u8 dsfield; 1077 __u32 mtu; 1078 int err; 1079 1080 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 1081 !ip6_tnl_xmit_ctl(t)) 1082 return -1; 1083 1084 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1085 encap_limit = t->parms.encap_limit; 1086 1087 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1088 fl6.flowi6_proto = IPPROTO_IPIP; 1089 1090 dsfield = ipv4_get_dsfield(iph); 1091 1092 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1093 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 1094 & IPV6_TCLASS_MASK; 1095 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1096 fl6.flowi6_mark = skb->mark; 1097 1098 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1099 if (err != 0) { 1100 /* XXX: send ICMP error even if DF is not set. */ 1101 if (err == -EMSGSIZE) 1102 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1103 htonl(mtu)); 1104 return -1; 1105 } 1106 1107 return 0; 1108 } 1109 1110 static inline int 1111 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1112 { 1113 struct ip6_tnl *t = netdev_priv(dev); 1114 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1115 int encap_limit = -1; 1116 __u16 offset; 1117 struct flowi6 fl6; 1118 __u8 dsfield; 1119 __u32 mtu; 1120 int err; 1121 1122 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1123 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1124 return -1; 1125 1126 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 1127 if (offset > 0) { 1128 struct ipv6_tlv_tnl_enc_lim *tel; 1129 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1130 if (tel->encap_limit == 0) { 1131 icmpv6_send(skb, ICMPV6_PARAMPROB, 1132 ICMPV6_HDR_FIELD, offset + 2); 1133 return -1; 1134 } 1135 encap_limit = tel->encap_limit - 1; 1136 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1137 encap_limit = t->parms.encap_limit; 1138 1139 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1140 fl6.flowi6_proto = IPPROTO_IPV6; 1141 1142 dsfield = ipv6_get_dsfield(ipv6h); 1143 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1144 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1145 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1146 fl6.flowlabel |= ip6_flowlabel(ipv6h); 1147 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1148 fl6.flowi6_mark = skb->mark; 1149 1150 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1151 if (err != 0) { 1152 if (err == -EMSGSIZE) 1153 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1154 return -1; 1155 } 1156 1157 return 0; 1158 } 1159 1160 static netdev_tx_t 1161 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1162 { 1163 struct ip6_tnl *t = netdev_priv(dev); 1164 struct net_device_stats *stats = &t->dev->stats; 1165 int ret; 1166 1167 switch (skb->protocol) { 1168 case htons(ETH_P_IP): 1169 ret = ip4ip6_tnl_xmit(skb, dev); 1170 break; 1171 case htons(ETH_P_IPV6): 1172 ret = ip6ip6_tnl_xmit(skb, dev); 1173 break; 1174 default: 1175 goto tx_err; 1176 } 1177 1178 if (ret < 0) 1179 goto tx_err; 1180 1181 return NETDEV_TX_OK; 1182 1183 tx_err: 1184 stats->tx_errors++; 1185 stats->tx_dropped++; 1186 kfree_skb(skb); 1187 return NETDEV_TX_OK; 1188 } 1189 1190 static void ip6_tnl_link_config(struct ip6_tnl *t) 1191 { 1192 struct net_device *dev = t->dev; 1193 struct __ip6_tnl_parm *p = &t->parms; 1194 struct flowi6 *fl6 = &t->fl.u.ip6; 1195 1196 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1197 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1198 1199 /* Set up flowi template */ 1200 fl6->saddr = p->laddr; 1201 fl6->daddr = p->raddr; 1202 fl6->flowi6_oif = p->link; 1203 fl6->flowlabel = 0; 1204 1205 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1206 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1207 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1208 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1209 1210 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1211 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1212 1213 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1214 dev->flags |= IFF_POINTOPOINT; 1215 else 1216 dev->flags &= ~IFF_POINTOPOINT; 1217 1218 dev->iflink = p->link; 1219 1220 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1221 int strict = (ipv6_addr_type(&p->raddr) & 1222 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1223 1224 struct rt6_info *rt = rt6_lookup(t->net, 1225 &p->raddr, &p->laddr, 1226 p->link, strict); 1227 1228 if (rt == NULL) 1229 return; 1230 1231 if (rt->dst.dev) { 1232 dev->hard_header_len = rt->dst.dev->hard_header_len + 1233 sizeof(struct ipv6hdr); 1234 1235 dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr); 1236 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1237 dev->mtu -= 8; 1238 1239 if (dev->mtu < IPV6_MIN_MTU) 1240 dev->mtu = IPV6_MIN_MTU; 1241 } 1242 ip6_rt_put(rt); 1243 } 1244 } 1245 1246 /** 1247 * ip6_tnl_change - update the tunnel parameters 1248 * @t: tunnel to be changed 1249 * @p: tunnel configuration parameters 1250 * 1251 * Description: 1252 * ip6_tnl_change() updates the tunnel parameters 1253 **/ 1254 1255 static int 1256 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 1257 { 1258 t->parms.laddr = p->laddr; 1259 t->parms.raddr = p->raddr; 1260 t->parms.flags = p->flags; 1261 t->parms.hop_limit = p->hop_limit; 1262 t->parms.encap_limit = p->encap_limit; 1263 t->parms.flowinfo = p->flowinfo; 1264 t->parms.link = p->link; 1265 t->parms.proto = p->proto; 1266 ip6_tnl_dst_reset(t); 1267 ip6_tnl_link_config(t); 1268 return 0; 1269 } 1270 1271 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 1272 { 1273 struct net *net = t->net; 1274 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1275 int err; 1276 1277 ip6_tnl_unlink(ip6n, t); 1278 synchronize_net(); 1279 err = ip6_tnl_change(t, p); 1280 ip6_tnl_link(ip6n, t); 1281 netdev_state_change(t->dev); 1282 return err; 1283 } 1284 1285 static void 1286 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u) 1287 { 1288 p->laddr = u->laddr; 1289 p->raddr = u->raddr; 1290 p->flags = u->flags; 1291 p->hop_limit = u->hop_limit; 1292 p->encap_limit = u->encap_limit; 1293 p->flowinfo = u->flowinfo; 1294 p->link = u->link; 1295 p->proto = u->proto; 1296 memcpy(p->name, u->name, sizeof(u->name)); 1297 } 1298 1299 static void 1300 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p) 1301 { 1302 u->laddr = p->laddr; 1303 u->raddr = p->raddr; 1304 u->flags = p->flags; 1305 u->hop_limit = p->hop_limit; 1306 u->encap_limit = p->encap_limit; 1307 u->flowinfo = p->flowinfo; 1308 u->link = p->link; 1309 u->proto = p->proto; 1310 memcpy(u->name, p->name, sizeof(u->name)); 1311 } 1312 1313 /** 1314 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1315 * @dev: virtual device associated with tunnel 1316 * @ifr: parameters passed from userspace 1317 * @cmd: command to be performed 1318 * 1319 * Description: 1320 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1321 * from userspace. 1322 * 1323 * The possible commands are the following: 1324 * %SIOCGETTUNNEL: get tunnel parameters for device 1325 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1326 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1327 * %SIOCDELTUNNEL: delete tunnel 1328 * 1329 * The fallback device "ip6tnl0", created during module 1330 * initialization, can be used for creating other tunnel devices. 1331 * 1332 * Return: 1333 * 0 on success, 1334 * %-EFAULT if unable to copy data to or from userspace, 1335 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1336 * %-EINVAL if passed tunnel parameters are invalid, 1337 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1338 * %-ENODEV if attempting to change or delete a nonexisting device 1339 **/ 1340 1341 static int 1342 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1343 { 1344 int err = 0; 1345 struct ip6_tnl_parm p; 1346 struct __ip6_tnl_parm p1; 1347 struct ip6_tnl *t = netdev_priv(dev); 1348 struct net *net = t->net; 1349 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1350 1351 switch (cmd) { 1352 case SIOCGETTUNNEL: 1353 if (dev == ip6n->fb_tnl_dev) { 1354 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1355 err = -EFAULT; 1356 break; 1357 } 1358 ip6_tnl_parm_from_user(&p1, &p); 1359 t = ip6_tnl_locate(net, &p1, 0); 1360 if (t == NULL) 1361 t = netdev_priv(dev); 1362 } else { 1363 memset(&p, 0, sizeof(p)); 1364 } 1365 ip6_tnl_parm_to_user(&p, &t->parms); 1366 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) { 1367 err = -EFAULT; 1368 } 1369 break; 1370 case SIOCADDTUNNEL: 1371 case SIOCCHGTUNNEL: 1372 err = -EPERM; 1373 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1374 break; 1375 err = -EFAULT; 1376 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1377 break; 1378 err = -EINVAL; 1379 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1380 p.proto != 0) 1381 break; 1382 ip6_tnl_parm_from_user(&p1, &p); 1383 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL); 1384 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1385 if (t != NULL) { 1386 if (t->dev != dev) { 1387 err = -EEXIST; 1388 break; 1389 } 1390 } else 1391 t = netdev_priv(dev); 1392 1393 err = ip6_tnl_update(t, &p1); 1394 } 1395 if (t) { 1396 err = 0; 1397 ip6_tnl_parm_to_user(&p, &t->parms); 1398 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1399 err = -EFAULT; 1400 1401 } else 1402 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1403 break; 1404 case SIOCDELTUNNEL: 1405 err = -EPERM; 1406 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1407 break; 1408 1409 if (dev == ip6n->fb_tnl_dev) { 1410 err = -EFAULT; 1411 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1412 break; 1413 err = -ENOENT; 1414 ip6_tnl_parm_from_user(&p1, &p); 1415 t = ip6_tnl_locate(net, &p1, 0); 1416 if (t == NULL) 1417 break; 1418 err = -EPERM; 1419 if (t->dev == ip6n->fb_tnl_dev) 1420 break; 1421 dev = t->dev; 1422 } 1423 err = 0; 1424 unregister_netdevice(dev); 1425 break; 1426 default: 1427 err = -EINVAL; 1428 } 1429 return err; 1430 } 1431 1432 /** 1433 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1434 * @dev: virtual device associated with tunnel 1435 * @new_mtu: the new mtu 1436 * 1437 * Return: 1438 * 0 on success, 1439 * %-EINVAL if mtu too small 1440 **/ 1441 1442 static int 1443 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1444 { 1445 struct ip6_tnl *tnl = netdev_priv(dev); 1446 1447 if (tnl->parms.proto == IPPROTO_IPIP) { 1448 if (new_mtu < 68) 1449 return -EINVAL; 1450 } else { 1451 if (new_mtu < IPV6_MIN_MTU) 1452 return -EINVAL; 1453 } 1454 if (new_mtu > 0xFFF8 - dev->hard_header_len) 1455 return -EINVAL; 1456 dev->mtu = new_mtu; 1457 return 0; 1458 } 1459 1460 1461 static const struct net_device_ops ip6_tnl_netdev_ops = { 1462 .ndo_init = ip6_tnl_dev_init, 1463 .ndo_uninit = ip6_tnl_dev_uninit, 1464 .ndo_start_xmit = ip6_tnl_xmit, 1465 .ndo_do_ioctl = ip6_tnl_ioctl, 1466 .ndo_change_mtu = ip6_tnl_change_mtu, 1467 .ndo_get_stats = ip6_get_stats, 1468 }; 1469 1470 1471 /** 1472 * ip6_tnl_dev_setup - setup virtual tunnel device 1473 * @dev: virtual device associated with tunnel 1474 * 1475 * Description: 1476 * Initialize function pointers and device parameters 1477 **/ 1478 1479 static void ip6_tnl_dev_setup(struct net_device *dev) 1480 { 1481 struct ip6_tnl *t; 1482 1483 dev->netdev_ops = &ip6_tnl_netdev_ops; 1484 dev->destructor = ip6_dev_free; 1485 1486 dev->type = ARPHRD_TUNNEL6; 1487 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr); 1488 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr); 1489 t = netdev_priv(dev); 1490 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1491 dev->mtu -= 8; 1492 dev->flags |= IFF_NOARP; 1493 dev->addr_len = sizeof(struct in6_addr); 1494 netif_keep_dst(dev); 1495 /* This perm addr will be used as interface identifier by IPv6 */ 1496 dev->addr_assign_type = NET_ADDR_RANDOM; 1497 eth_random_addr(dev->perm_addr); 1498 } 1499 1500 1501 /** 1502 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1503 * @dev: virtual device associated with tunnel 1504 **/ 1505 1506 static inline int 1507 ip6_tnl_dev_init_gen(struct net_device *dev) 1508 { 1509 struct ip6_tnl *t = netdev_priv(dev); 1510 1511 t->dev = dev; 1512 t->net = dev_net(dev); 1513 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1514 if (!dev->tstats) 1515 return -ENOMEM; 1516 return 0; 1517 } 1518 1519 /** 1520 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1521 * @dev: virtual device associated with tunnel 1522 **/ 1523 1524 static int ip6_tnl_dev_init(struct net_device *dev) 1525 { 1526 struct ip6_tnl *t = netdev_priv(dev); 1527 int err = ip6_tnl_dev_init_gen(dev); 1528 1529 if (err) 1530 return err; 1531 ip6_tnl_link_config(t); 1532 return 0; 1533 } 1534 1535 /** 1536 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1537 * @dev: fallback device 1538 * 1539 * Return: 0 1540 **/ 1541 1542 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1543 { 1544 struct ip6_tnl *t = netdev_priv(dev); 1545 struct net *net = dev_net(dev); 1546 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1547 1548 t->parms.proto = IPPROTO_IPV6; 1549 dev_hold(dev); 1550 1551 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1552 return 0; 1553 } 1554 1555 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[]) 1556 { 1557 u8 proto; 1558 1559 if (!data || !data[IFLA_IPTUN_PROTO]) 1560 return 0; 1561 1562 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1563 if (proto != IPPROTO_IPV6 && 1564 proto != IPPROTO_IPIP && 1565 proto != 0) 1566 return -EINVAL; 1567 1568 return 0; 1569 } 1570 1571 static void ip6_tnl_netlink_parms(struct nlattr *data[], 1572 struct __ip6_tnl_parm *parms) 1573 { 1574 memset(parms, 0, sizeof(*parms)); 1575 1576 if (!data) 1577 return; 1578 1579 if (data[IFLA_IPTUN_LINK]) 1580 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1581 1582 if (data[IFLA_IPTUN_LOCAL]) 1583 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL], 1584 sizeof(struct in6_addr)); 1585 1586 if (data[IFLA_IPTUN_REMOTE]) 1587 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE], 1588 sizeof(struct in6_addr)); 1589 1590 if (data[IFLA_IPTUN_TTL]) 1591 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]); 1592 1593 if (data[IFLA_IPTUN_ENCAP_LIMIT]) 1594 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]); 1595 1596 if (data[IFLA_IPTUN_FLOWINFO]) 1597 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]); 1598 1599 if (data[IFLA_IPTUN_FLAGS]) 1600 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]); 1601 1602 if (data[IFLA_IPTUN_PROTO]) 1603 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1604 } 1605 1606 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev, 1607 struct nlattr *tb[], struct nlattr *data[]) 1608 { 1609 struct net *net = dev_net(dev); 1610 struct ip6_tnl *nt; 1611 1612 nt = netdev_priv(dev); 1613 ip6_tnl_netlink_parms(data, &nt->parms); 1614 1615 if (ip6_tnl_locate(net, &nt->parms, 0)) 1616 return -EEXIST; 1617 1618 return ip6_tnl_create2(dev); 1619 } 1620 1621 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[], 1622 struct nlattr *data[]) 1623 { 1624 struct ip6_tnl *t = netdev_priv(dev); 1625 struct __ip6_tnl_parm p; 1626 struct net *net = t->net; 1627 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1628 1629 if (dev == ip6n->fb_tnl_dev) 1630 return -EINVAL; 1631 1632 ip6_tnl_netlink_parms(data, &p); 1633 1634 t = ip6_tnl_locate(net, &p, 0); 1635 1636 if (t) { 1637 if (t->dev != dev) 1638 return -EEXIST; 1639 } else 1640 t = netdev_priv(dev); 1641 1642 return ip6_tnl_update(t, &p); 1643 } 1644 1645 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head) 1646 { 1647 struct net *net = dev_net(dev); 1648 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1649 1650 if (dev != ip6n->fb_tnl_dev) 1651 unregister_netdevice_queue(dev, head); 1652 } 1653 1654 static size_t ip6_tnl_get_size(const struct net_device *dev) 1655 { 1656 return 1657 /* IFLA_IPTUN_LINK */ 1658 nla_total_size(4) + 1659 /* IFLA_IPTUN_LOCAL */ 1660 nla_total_size(sizeof(struct in6_addr)) + 1661 /* IFLA_IPTUN_REMOTE */ 1662 nla_total_size(sizeof(struct in6_addr)) + 1663 /* IFLA_IPTUN_TTL */ 1664 nla_total_size(1) + 1665 /* IFLA_IPTUN_ENCAP_LIMIT */ 1666 nla_total_size(1) + 1667 /* IFLA_IPTUN_FLOWINFO */ 1668 nla_total_size(4) + 1669 /* IFLA_IPTUN_FLAGS */ 1670 nla_total_size(4) + 1671 /* IFLA_IPTUN_PROTO */ 1672 nla_total_size(1) + 1673 0; 1674 } 1675 1676 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev) 1677 { 1678 struct ip6_tnl *tunnel = netdev_priv(dev); 1679 struct __ip6_tnl_parm *parm = &tunnel->parms; 1680 1681 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1682 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr), 1683 &parm->laddr) || 1684 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr), 1685 &parm->raddr) || 1686 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) || 1687 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) || 1688 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) || 1689 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) || 1690 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto)) 1691 goto nla_put_failure; 1692 return 0; 1693 1694 nla_put_failure: 1695 return -EMSGSIZE; 1696 } 1697 1698 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = { 1699 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1700 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) }, 1701 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) }, 1702 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1703 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 }, 1704 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 }, 1705 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 }, 1706 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1707 }; 1708 1709 static struct rtnl_link_ops ip6_link_ops __read_mostly = { 1710 .kind = "ip6tnl", 1711 .maxtype = IFLA_IPTUN_MAX, 1712 .policy = ip6_tnl_policy, 1713 .priv_size = sizeof(struct ip6_tnl), 1714 .setup = ip6_tnl_dev_setup, 1715 .validate = ip6_tnl_validate, 1716 .newlink = ip6_tnl_newlink, 1717 .changelink = ip6_tnl_changelink, 1718 .dellink = ip6_tnl_dellink, 1719 .get_size = ip6_tnl_get_size, 1720 .fill_info = ip6_tnl_fill_info, 1721 }; 1722 1723 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 1724 .handler = ip4ip6_rcv, 1725 .err_handler = ip4ip6_err, 1726 .priority = 1, 1727 }; 1728 1729 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 1730 .handler = ip6ip6_rcv, 1731 .err_handler = ip6ip6_err, 1732 .priority = 1, 1733 }; 1734 1735 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net) 1736 { 1737 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1738 struct net_device *dev, *aux; 1739 int h; 1740 struct ip6_tnl *t; 1741 LIST_HEAD(list); 1742 1743 for_each_netdev_safe(net, dev, aux) 1744 if (dev->rtnl_link_ops == &ip6_link_ops) 1745 unregister_netdevice_queue(dev, &list); 1746 1747 for (h = 0; h < HASH_SIZE; h++) { 1748 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1749 while (t != NULL) { 1750 /* If dev is in the same netns, it has already 1751 * been added to the list by the previous loop. 1752 */ 1753 if (!net_eq(dev_net(t->dev), net)) 1754 unregister_netdevice_queue(t->dev, &list); 1755 t = rtnl_dereference(t->next); 1756 } 1757 } 1758 1759 unregister_netdevice_many(&list); 1760 } 1761 1762 static int __net_init ip6_tnl_init_net(struct net *net) 1763 { 1764 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1765 struct ip6_tnl *t = NULL; 1766 int err; 1767 1768 ip6n->tnls[0] = ip6n->tnls_wc; 1769 ip6n->tnls[1] = ip6n->tnls_r_l; 1770 1771 err = -ENOMEM; 1772 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1773 NET_NAME_UNKNOWN, ip6_tnl_dev_setup); 1774 1775 if (!ip6n->fb_tnl_dev) 1776 goto err_alloc_dev; 1777 dev_net_set(ip6n->fb_tnl_dev, net); 1778 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops; 1779 /* FB netdevice is special: we have one, and only one per netns. 1780 * Allowing to move it to another netns is clearly unsafe. 1781 */ 1782 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL; 1783 1784 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1785 if (err < 0) 1786 goto err_register; 1787 1788 err = register_netdev(ip6n->fb_tnl_dev); 1789 if (err < 0) 1790 goto err_register; 1791 1792 t = netdev_priv(ip6n->fb_tnl_dev); 1793 1794 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1795 return 0; 1796 1797 err_register: 1798 ip6_dev_free(ip6n->fb_tnl_dev); 1799 err_alloc_dev: 1800 return err; 1801 } 1802 1803 static void __net_exit ip6_tnl_exit_net(struct net *net) 1804 { 1805 rtnl_lock(); 1806 ip6_tnl_destroy_tunnels(net); 1807 rtnl_unlock(); 1808 } 1809 1810 static struct pernet_operations ip6_tnl_net_ops = { 1811 .init = ip6_tnl_init_net, 1812 .exit = ip6_tnl_exit_net, 1813 .id = &ip6_tnl_net_id, 1814 .size = sizeof(struct ip6_tnl_net), 1815 }; 1816 1817 /** 1818 * ip6_tunnel_init - register protocol and reserve needed resources 1819 * 1820 * Return: 0 on success 1821 **/ 1822 1823 static int __init ip6_tunnel_init(void) 1824 { 1825 int err; 1826 1827 err = register_pernet_device(&ip6_tnl_net_ops); 1828 if (err < 0) 1829 goto out_pernet; 1830 1831 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 1832 if (err < 0) { 1833 pr_err("%s: can't register ip4ip6\n", __func__); 1834 goto out_ip4ip6; 1835 } 1836 1837 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 1838 if (err < 0) { 1839 pr_err("%s: can't register ip6ip6\n", __func__); 1840 goto out_ip6ip6; 1841 } 1842 err = rtnl_link_register(&ip6_link_ops); 1843 if (err < 0) 1844 goto rtnl_link_failed; 1845 1846 return 0; 1847 1848 rtnl_link_failed: 1849 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 1850 out_ip6ip6: 1851 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1852 out_ip4ip6: 1853 unregister_pernet_device(&ip6_tnl_net_ops); 1854 out_pernet: 1855 return err; 1856 } 1857 1858 /** 1859 * ip6_tunnel_cleanup - free resources and unregister protocol 1860 **/ 1861 1862 static void __exit ip6_tunnel_cleanup(void) 1863 { 1864 rtnl_link_unregister(&ip6_link_ops); 1865 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1866 pr_info("%s: can't deregister ip4ip6\n", __func__); 1867 1868 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1869 pr_info("%s: can't deregister ip6ip6\n", __func__); 1870 1871 unregister_pernet_device(&ip6_tnl_net_ops); 1872 } 1873 1874 module_init(ip6_tunnel_init); 1875 module_exit(ip6_tunnel_cleanup); 1876