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 #include <linux/module.h> 22 #include <linux/capability.h> 23 #include <linux/errno.h> 24 #include <linux/types.h> 25 #include <linux/sockios.h> 26 #include <linux/icmp.h> 27 #include <linux/if.h> 28 #include <linux/in.h> 29 #include <linux/ip.h> 30 #include <linux/if_tunnel.h> 31 #include <linux/net.h> 32 #include <linux/in6.h> 33 #include <linux/netdevice.h> 34 #include <linux/if_arp.h> 35 #include <linux/icmpv6.h> 36 #include <linux/init.h> 37 #include <linux/route.h> 38 #include <linux/rtnetlink.h> 39 #include <linux/netfilter_ipv6.h> 40 #include <linux/slab.h> 41 42 #include <asm/uaccess.h> 43 #include <asm/atomic.h> 44 45 #include <net/icmp.h> 46 #include <net/ip.h> 47 #include <net/ipv6.h> 48 #include <net/ip6_route.h> 49 #include <net/addrconf.h> 50 #include <net/ip6_tunnel.h> 51 #include <net/xfrm.h> 52 #include <net/dsfield.h> 53 #include <net/inet_ecn.h> 54 #include <net/net_namespace.h> 55 #include <net/netns/generic.h> 56 57 MODULE_AUTHOR("Ville Nuorvala"); 58 MODULE_DESCRIPTION("IPv6 tunneling device"); 59 MODULE_LICENSE("GPL"); 60 MODULE_ALIAS_NETDEV("ip6tnl0"); 61 62 #ifdef IP6_TNL_DEBUG 63 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__) 64 #else 65 #define IP6_TNL_TRACE(x...) do {;} while(0) 66 #endif 67 68 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 69 #define IPV6_TCLASS_SHIFT 20 70 71 #define HASH_SIZE 32 72 73 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \ 74 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \ 75 (HASH_SIZE - 1)) 76 77 static int ip6_tnl_dev_init(struct net_device *dev); 78 static void ip6_tnl_dev_setup(struct net_device *dev); 79 80 static int ip6_tnl_net_id __read_mostly; 81 struct ip6_tnl_net { 82 /* the IPv6 tunnel fallback device */ 83 struct net_device *fb_tnl_dev; 84 /* lists for storing tunnels in use */ 85 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE]; 86 struct ip6_tnl __rcu *tnls_wc[1]; 87 struct ip6_tnl __rcu **tnls[2]; 88 }; 89 90 /* often modified stats are per cpu, other are shared (netdev->stats) */ 91 struct pcpu_tstats { 92 unsigned long rx_packets; 93 unsigned long rx_bytes; 94 unsigned long tx_packets; 95 unsigned long tx_bytes; 96 }; 97 98 static struct net_device_stats *ip6_get_stats(struct net_device *dev) 99 { 100 struct pcpu_tstats sum = { 0 }; 101 int i; 102 103 for_each_possible_cpu(i) { 104 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i); 105 106 sum.rx_packets += tstats->rx_packets; 107 sum.rx_bytes += tstats->rx_bytes; 108 sum.tx_packets += tstats->tx_packets; 109 sum.tx_bytes += tstats->tx_bytes; 110 } 111 dev->stats.rx_packets = sum.rx_packets; 112 dev->stats.rx_bytes = sum.rx_bytes; 113 dev->stats.tx_packets = sum.tx_packets; 114 dev->stats.tx_bytes = sum.tx_bytes; 115 return &dev->stats; 116 } 117 118 /* 119 * Locking : hash tables are protected by RCU and RTNL 120 */ 121 122 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 123 { 124 struct dst_entry *dst = t->dst_cache; 125 126 if (dst && dst->obsolete && 127 dst->ops->check(dst, t->dst_cookie) == NULL) { 128 t->dst_cache = NULL; 129 dst_release(dst); 130 return NULL; 131 } 132 133 return dst; 134 } 135 136 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t) 137 { 138 dst_release(t->dst_cache); 139 t->dst_cache = NULL; 140 } 141 142 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 143 { 144 struct rt6_info *rt = (struct rt6_info *) dst; 145 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 146 dst_release(t->dst_cache); 147 t->dst_cache = dst; 148 } 149 150 /** 151 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 152 * @remote: the address of the tunnel exit-point 153 * @local: the address of the tunnel entry-point 154 * 155 * Return: 156 * tunnel matching given end-points if found, 157 * else fallback tunnel if its device is up, 158 * else %NULL 159 **/ 160 161 #define for_each_ip6_tunnel_rcu(start) \ 162 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 163 164 static struct ip6_tnl * 165 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local) 166 { 167 unsigned int h0 = HASH(remote); 168 unsigned int h1 = HASH(local); 169 struct ip6_tnl *t; 170 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 171 172 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) { 173 if (ipv6_addr_equal(local, &t->parms.laddr) && 174 ipv6_addr_equal(remote, &t->parms.raddr) && 175 (t->dev->flags & IFF_UP)) 176 return t; 177 } 178 t = rcu_dereference(ip6n->tnls_wc[0]); 179 if (t && (t->dev->flags & IFF_UP)) 180 return t; 181 182 return NULL; 183 } 184 185 /** 186 * ip6_tnl_bucket - get head of list matching given tunnel parameters 187 * @p: parameters containing tunnel end-points 188 * 189 * Description: 190 * ip6_tnl_bucket() returns the head of the list matching the 191 * &struct in6_addr entries laddr and raddr in @p. 192 * 193 * Return: head of IPv6 tunnel list 194 **/ 195 196 static struct ip6_tnl __rcu ** 197 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p) 198 { 199 struct in6_addr *remote = &p->raddr; 200 struct in6_addr *local = &p->laddr; 201 unsigned h = 0; 202 int prio = 0; 203 204 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 205 prio = 1; 206 h = HASH(remote) ^ HASH(local); 207 } 208 return &ip6n->tnls[prio][h]; 209 } 210 211 /** 212 * ip6_tnl_link - add tunnel to hash table 213 * @t: tunnel to be added 214 **/ 215 216 static void 217 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 218 { 219 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms); 220 221 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 222 rcu_assign_pointer(*tp, t); 223 } 224 225 /** 226 * ip6_tnl_unlink - remove tunnel from hash table 227 * @t: tunnel to be removed 228 **/ 229 230 static void 231 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 232 { 233 struct ip6_tnl __rcu **tp; 234 struct ip6_tnl *iter; 235 236 for (tp = ip6_tnl_bucket(ip6n, &t->parms); 237 (iter = rtnl_dereference(*tp)) != NULL; 238 tp = &iter->next) { 239 if (t == iter) { 240 rcu_assign_pointer(*tp, t->next); 241 break; 242 } 243 } 244 } 245 246 static void ip6_dev_free(struct net_device *dev) 247 { 248 free_percpu(dev->tstats); 249 free_netdev(dev); 250 } 251 252 /** 253 * ip6_tnl_create() - create a new tunnel 254 * @p: tunnel parameters 255 * @pt: pointer to new tunnel 256 * 257 * Description: 258 * Create tunnel matching given parameters. 259 * 260 * Return: 261 * created tunnel or NULL 262 **/ 263 264 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p) 265 { 266 struct net_device *dev; 267 struct ip6_tnl *t; 268 char name[IFNAMSIZ]; 269 int err; 270 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 271 272 if (p->name[0]) 273 strlcpy(name, p->name, IFNAMSIZ); 274 else 275 sprintf(name, "ip6tnl%%d"); 276 277 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup); 278 if (dev == NULL) 279 goto failed; 280 281 dev_net_set(dev, net); 282 283 if (strchr(name, '%')) { 284 if (dev_alloc_name(dev, name) < 0) 285 goto failed_free; 286 } 287 288 t = netdev_priv(dev); 289 t->parms = *p; 290 err = ip6_tnl_dev_init(dev); 291 if (err < 0) 292 goto failed_free; 293 294 if ((err = register_netdevice(dev)) < 0) 295 goto failed_free; 296 297 dev_hold(dev); 298 ip6_tnl_link(ip6n, t); 299 return t; 300 301 failed_free: 302 ip6_dev_free(dev); 303 failed: 304 return NULL; 305 } 306 307 /** 308 * ip6_tnl_locate - find or create tunnel matching given parameters 309 * @p: tunnel parameters 310 * @create: != 0 if allowed to create new tunnel if no match found 311 * 312 * Description: 313 * ip6_tnl_locate() first tries to locate an existing tunnel 314 * based on @parms. If this is unsuccessful, but @create is set a new 315 * tunnel device is created and registered for use. 316 * 317 * Return: 318 * matching tunnel or NULL 319 **/ 320 321 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 322 struct ip6_tnl_parm *p, int create) 323 { 324 struct in6_addr *remote = &p->raddr; 325 struct in6_addr *local = &p->laddr; 326 struct ip6_tnl __rcu **tp; 327 struct ip6_tnl *t; 328 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 329 330 for (tp = ip6_tnl_bucket(ip6n, p); 331 (t = rtnl_dereference(*tp)) != NULL; 332 tp = &t->next) { 333 if (ipv6_addr_equal(local, &t->parms.laddr) && 334 ipv6_addr_equal(remote, &t->parms.raddr)) 335 return t; 336 } 337 if (!create) 338 return NULL; 339 return ip6_tnl_create(net, p); 340 } 341 342 /** 343 * ip6_tnl_dev_uninit - tunnel device uninitializer 344 * @dev: the device to be destroyed 345 * 346 * Description: 347 * ip6_tnl_dev_uninit() removes tunnel from its list 348 **/ 349 350 static void 351 ip6_tnl_dev_uninit(struct net_device *dev) 352 { 353 struct ip6_tnl *t = netdev_priv(dev); 354 struct net *net = dev_net(dev); 355 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 356 357 if (dev == ip6n->fb_tnl_dev) 358 rcu_assign_pointer(ip6n->tnls_wc[0], NULL); 359 else 360 ip6_tnl_unlink(ip6n, t); 361 ip6_tnl_dst_reset(t); 362 dev_put(dev); 363 } 364 365 /** 366 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 367 * @skb: received socket buffer 368 * 369 * Return: 370 * 0 if none was found, 371 * else index to encapsulation limit 372 **/ 373 374 static __u16 375 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw) 376 { 377 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw; 378 __u8 nexthdr = ipv6h->nexthdr; 379 __u16 off = sizeof (*ipv6h); 380 381 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 382 __u16 optlen = 0; 383 struct ipv6_opt_hdr *hdr; 384 if (raw + off + sizeof (*hdr) > skb->data && 385 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 386 break; 387 388 hdr = (struct ipv6_opt_hdr *) (raw + off); 389 if (nexthdr == NEXTHDR_FRAGMENT) { 390 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 391 if (frag_hdr->frag_off) 392 break; 393 optlen = 8; 394 } else if (nexthdr == NEXTHDR_AUTH) { 395 optlen = (hdr->hdrlen + 2) << 2; 396 } else { 397 optlen = ipv6_optlen(hdr); 398 } 399 if (nexthdr == NEXTHDR_DEST) { 400 __u16 i = off + 2; 401 while (1) { 402 struct ipv6_tlv_tnl_enc_lim *tel; 403 404 /* No more room for encapsulation limit */ 405 if (i + sizeof (*tel) > off + optlen) 406 break; 407 408 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 409 /* return index of option if found and valid */ 410 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 411 tel->length == 1) 412 return i; 413 /* else jump to next option */ 414 if (tel->type) 415 i += tel->length + 2; 416 else 417 i++; 418 } 419 } 420 nexthdr = hdr->nexthdr; 421 off += optlen; 422 } 423 return 0; 424 } 425 426 /** 427 * ip6_tnl_err - tunnel error handler 428 * 429 * Description: 430 * ip6_tnl_err() should handle errors in the tunnel according 431 * to the specifications in RFC 2473. 432 **/ 433 434 static int 435 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 436 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 437 { 438 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data; 439 struct ip6_tnl *t; 440 int rel_msg = 0; 441 u8 rel_type = ICMPV6_DEST_UNREACH; 442 u8 rel_code = ICMPV6_ADDR_UNREACH; 443 __u32 rel_info = 0; 444 __u16 len; 445 int err = -ENOENT; 446 447 /* If the packet doesn't contain the original IPv6 header we are 448 in trouble since we might need the source address for further 449 processing of the error. */ 450 451 rcu_read_lock(); 452 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 453 &ipv6h->saddr)) == NULL) 454 goto out; 455 456 if (t->parms.proto != ipproto && t->parms.proto != 0) 457 goto out; 458 459 err = 0; 460 461 switch (*type) { 462 __u32 teli; 463 struct ipv6_tlv_tnl_enc_lim *tel; 464 __u32 mtu; 465 case ICMPV6_DEST_UNREACH: 466 if (net_ratelimit()) 467 printk(KERN_WARNING 468 "%s: Path to destination invalid " 469 "or inactive!\n", t->parms.name); 470 rel_msg = 1; 471 break; 472 case ICMPV6_TIME_EXCEED: 473 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 474 if (net_ratelimit()) 475 printk(KERN_WARNING 476 "%s: Too small hop limit or " 477 "routing loop in tunnel!\n", 478 t->parms.name); 479 rel_msg = 1; 480 } 481 break; 482 case ICMPV6_PARAMPROB: 483 teli = 0; 484 if ((*code) == ICMPV6_HDR_FIELD) 485 teli = parse_tlv_tnl_enc_lim(skb, skb->data); 486 487 if (teli && teli == *info - 2) { 488 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 489 if (tel->encap_limit == 0) { 490 if (net_ratelimit()) 491 printk(KERN_WARNING 492 "%s: Too small encapsulation " 493 "limit or routing loop in " 494 "tunnel!\n", t->parms.name); 495 rel_msg = 1; 496 } 497 } else if (net_ratelimit()) { 498 printk(KERN_WARNING 499 "%s: Recipient unable to parse tunneled " 500 "packet!\n ", t->parms.name); 501 } 502 break; 503 case ICMPV6_PKT_TOOBIG: 504 mtu = *info - offset; 505 if (mtu < IPV6_MIN_MTU) 506 mtu = IPV6_MIN_MTU; 507 t->dev->mtu = mtu; 508 509 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 510 rel_type = ICMPV6_PKT_TOOBIG; 511 rel_code = 0; 512 rel_info = mtu; 513 rel_msg = 1; 514 } 515 break; 516 } 517 518 *type = rel_type; 519 *code = rel_code; 520 *info = rel_info; 521 *msg = rel_msg; 522 523 out: 524 rcu_read_unlock(); 525 return err; 526 } 527 528 static int 529 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 530 u8 type, u8 code, int offset, __be32 info) 531 { 532 int rel_msg = 0; 533 u8 rel_type = type; 534 u8 rel_code = code; 535 __u32 rel_info = ntohl(info); 536 int err; 537 struct sk_buff *skb2; 538 struct iphdr *eiph; 539 struct rtable *rt; 540 541 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 542 &rel_msg, &rel_info, offset); 543 if (err < 0) 544 return err; 545 546 if (rel_msg == 0) 547 return 0; 548 549 switch (rel_type) { 550 case ICMPV6_DEST_UNREACH: 551 if (rel_code != ICMPV6_ADDR_UNREACH) 552 return 0; 553 rel_type = ICMP_DEST_UNREACH; 554 rel_code = ICMP_HOST_UNREACH; 555 break; 556 case ICMPV6_PKT_TOOBIG: 557 if (rel_code != 0) 558 return 0; 559 rel_type = ICMP_DEST_UNREACH; 560 rel_code = ICMP_FRAG_NEEDED; 561 break; 562 default: 563 return 0; 564 } 565 566 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 567 return 0; 568 569 skb2 = skb_clone(skb, GFP_ATOMIC); 570 if (!skb2) 571 return 0; 572 573 skb_dst_drop(skb2); 574 575 skb_pull(skb2, offset); 576 skb_reset_network_header(skb2); 577 eiph = ip_hdr(skb2); 578 579 /* Try to guess incoming interface */ 580 rt = ip_route_output_ports(dev_net(skb->dev), NULL, 581 eiph->saddr, 0, 582 0, 0, 583 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 584 if (IS_ERR(rt)) 585 goto out; 586 587 skb2->dev = rt->dst.dev; 588 589 /* route "incoming" packet */ 590 if (rt->rt_flags & RTCF_LOCAL) { 591 ip_rt_put(rt); 592 rt = NULL; 593 rt = ip_route_output_ports(dev_net(skb->dev), NULL, 594 eiph->daddr, eiph->saddr, 595 0, 0, 596 IPPROTO_IPIP, 597 RT_TOS(eiph->tos), 0); 598 if (IS_ERR(rt) || 599 rt->dst.dev->type != ARPHRD_TUNNEL) { 600 if (!IS_ERR(rt)) 601 ip_rt_put(rt); 602 goto out; 603 } 604 skb_dst_set(skb2, &rt->dst); 605 } else { 606 ip_rt_put(rt); 607 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 608 skb2->dev) || 609 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 610 goto out; 611 } 612 613 /* change mtu on this route */ 614 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 615 if (rel_info > dst_mtu(skb_dst(skb2))) 616 goto out; 617 618 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info); 619 } 620 621 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 622 623 out: 624 kfree_skb(skb2); 625 return 0; 626 } 627 628 static int 629 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 630 u8 type, u8 code, int offset, __be32 info) 631 { 632 int rel_msg = 0; 633 u8 rel_type = type; 634 u8 rel_code = code; 635 __u32 rel_info = ntohl(info); 636 int err; 637 638 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 639 &rel_msg, &rel_info, offset); 640 if (err < 0) 641 return err; 642 643 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 644 struct rt6_info *rt; 645 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 646 647 if (!skb2) 648 return 0; 649 650 skb_dst_drop(skb2); 651 skb_pull(skb2, offset); 652 skb_reset_network_header(skb2); 653 654 /* Try to guess incoming interface */ 655 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 656 NULL, 0, 0); 657 658 if (rt && rt->rt6i_dev) 659 skb2->dev = rt->rt6i_dev; 660 661 icmpv6_send(skb2, rel_type, rel_code, rel_info); 662 663 if (rt) 664 dst_release(&rt->dst); 665 666 kfree_skb(skb2); 667 } 668 669 return 0; 670 } 671 672 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t, 673 struct ipv6hdr *ipv6h, 674 struct sk_buff *skb) 675 { 676 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 677 678 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 679 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 680 681 if (INET_ECN_is_ce(dsfield)) 682 IP_ECN_set_ce(ip_hdr(skb)); 683 } 684 685 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t, 686 struct ipv6hdr *ipv6h, 687 struct sk_buff *skb) 688 { 689 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 690 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 691 692 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h))) 693 IP6_ECN_set_ce(ipv6_hdr(skb)); 694 } 695 696 /* called with rcu_read_lock() */ 697 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t) 698 { 699 struct ip6_tnl_parm *p = &t->parms; 700 int ret = 0; 701 struct net *net = dev_net(t->dev); 702 703 if (p->flags & IP6_TNL_F_CAP_RCV) { 704 struct net_device *ldev = NULL; 705 706 if (p->link) 707 ldev = dev_get_by_index_rcu(net, p->link); 708 709 if ((ipv6_addr_is_multicast(&p->laddr) || 710 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) && 711 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0))) 712 ret = 1; 713 714 } 715 return ret; 716 } 717 718 /** 719 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 720 * @skb: received socket buffer 721 * @protocol: ethernet protocol ID 722 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 723 * 724 * Return: 0 725 **/ 726 727 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 728 __u8 ipproto, 729 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t, 730 struct ipv6hdr *ipv6h, 731 struct sk_buff *skb)) 732 { 733 struct ip6_tnl *t; 734 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 735 736 rcu_read_lock(); 737 738 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 739 &ipv6h->daddr)) != NULL) { 740 struct pcpu_tstats *tstats; 741 742 if (t->parms.proto != ipproto && t->parms.proto != 0) { 743 rcu_read_unlock(); 744 goto discard; 745 } 746 747 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 748 rcu_read_unlock(); 749 goto discard; 750 } 751 752 if (!ip6_tnl_rcv_ctl(t)) { 753 t->dev->stats.rx_dropped++; 754 rcu_read_unlock(); 755 goto discard; 756 } 757 secpath_reset(skb); 758 skb->mac_header = skb->network_header; 759 skb_reset_network_header(skb); 760 skb->protocol = htons(protocol); 761 skb->pkt_type = PACKET_HOST; 762 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 763 764 tstats = this_cpu_ptr(t->dev->tstats); 765 tstats->rx_packets++; 766 tstats->rx_bytes += skb->len; 767 768 __skb_tunnel_rx(skb, t->dev); 769 770 dscp_ecn_decapsulate(t, ipv6h, skb); 771 772 netif_rx(skb); 773 774 rcu_read_unlock(); 775 return 0; 776 } 777 rcu_read_unlock(); 778 return 1; 779 780 discard: 781 kfree_skb(skb); 782 return 0; 783 } 784 785 static int ip4ip6_rcv(struct sk_buff *skb) 786 { 787 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 788 ip4ip6_dscp_ecn_decapsulate); 789 } 790 791 static int ip6ip6_rcv(struct sk_buff *skb) 792 { 793 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 794 ip6ip6_dscp_ecn_decapsulate); 795 } 796 797 struct ipv6_tel_txoption { 798 struct ipv6_txoptions ops; 799 __u8 dst_opt[8]; 800 }; 801 802 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 803 { 804 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 805 806 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 807 opt->dst_opt[3] = 1; 808 opt->dst_opt[4] = encap_limit; 809 opt->dst_opt[5] = IPV6_TLV_PADN; 810 opt->dst_opt[6] = 1; 811 812 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 813 opt->ops.opt_nflen = 8; 814 } 815 816 /** 817 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 818 * @t: the outgoing tunnel device 819 * @hdr: IPv6 header from the incoming packet 820 * 821 * Description: 822 * Avoid trivial tunneling loop by checking that tunnel exit-point 823 * doesn't match source of incoming packet. 824 * 825 * Return: 826 * 1 if conflict, 827 * 0 else 828 **/ 829 830 static inline int 831 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr) 832 { 833 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 834 } 835 836 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 837 { 838 struct ip6_tnl_parm *p = &t->parms; 839 int ret = 0; 840 struct net *net = dev_net(t->dev); 841 842 if (p->flags & IP6_TNL_F_CAP_XMIT) { 843 struct net_device *ldev = NULL; 844 845 rcu_read_lock(); 846 if (p->link) 847 ldev = dev_get_by_index_rcu(net, p->link); 848 849 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 850 printk(KERN_WARNING 851 "%s xmit: Local address not yet configured!\n", 852 p->name); 853 else if (!ipv6_addr_is_multicast(&p->raddr) && 854 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 855 printk(KERN_WARNING 856 "%s xmit: Routing loop! " 857 "Remote address found on this node!\n", 858 p->name); 859 else 860 ret = 1; 861 rcu_read_unlock(); 862 } 863 return ret; 864 } 865 /** 866 * ip6_tnl_xmit2 - encapsulate packet and send 867 * @skb: the outgoing socket buffer 868 * @dev: the outgoing tunnel device 869 * @dsfield: dscp code for outer header 870 * @fl: flow of tunneled packet 871 * @encap_limit: encapsulation limit 872 * @pmtu: Path MTU is stored if packet is too big 873 * 874 * Description: 875 * Build new header and do some sanity checks on the packet before sending 876 * it. 877 * 878 * Return: 879 * 0 on success 880 * -1 fail 881 * %-EMSGSIZE message too big. return mtu in this case. 882 **/ 883 884 static int ip6_tnl_xmit2(struct sk_buff *skb, 885 struct net_device *dev, 886 __u8 dsfield, 887 struct flowi6 *fl6, 888 int encap_limit, 889 __u32 *pmtu) 890 { 891 struct net *net = dev_net(dev); 892 struct ip6_tnl *t = netdev_priv(dev); 893 struct net_device_stats *stats = &t->dev->stats; 894 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 895 struct ipv6_tel_txoption opt; 896 struct dst_entry *dst; 897 struct net_device *tdev; 898 int mtu; 899 unsigned int max_headroom = sizeof(struct ipv6hdr); 900 u8 proto; 901 int err = -1; 902 int pkt_len; 903 904 if ((dst = ip6_tnl_dst_check(t)) != NULL) 905 dst_hold(dst); 906 else { 907 dst = ip6_route_output(net, NULL, fl6); 908 909 if (dst->error) 910 goto tx_err_link_failure; 911 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0); 912 if (IS_ERR(dst)) { 913 err = PTR_ERR(dst); 914 dst = NULL; 915 goto tx_err_link_failure; 916 } 917 } 918 919 tdev = dst->dev; 920 921 if (tdev == dev) { 922 stats->collisions++; 923 if (net_ratelimit()) 924 printk(KERN_WARNING 925 "%s: Local routing loop detected!\n", 926 t->parms.name); 927 goto tx_err_dst_release; 928 } 929 mtu = dst_mtu(dst) - sizeof (*ipv6h); 930 if (encap_limit >= 0) { 931 max_headroom += 8; 932 mtu -= 8; 933 } 934 if (mtu < IPV6_MIN_MTU) 935 mtu = IPV6_MIN_MTU; 936 if (skb_dst(skb)) 937 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu); 938 if (skb->len > mtu) { 939 *pmtu = mtu; 940 err = -EMSGSIZE; 941 goto tx_err_dst_release; 942 } 943 944 /* 945 * Okay, now see if we can stuff it in the buffer as-is. 946 */ 947 max_headroom += LL_RESERVED_SPACE(tdev); 948 949 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 950 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 951 struct sk_buff *new_skb; 952 953 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 954 goto tx_err_dst_release; 955 956 if (skb->sk) 957 skb_set_owner_w(new_skb, skb->sk); 958 kfree_skb(skb); 959 skb = new_skb; 960 } 961 skb_dst_drop(skb); 962 skb_dst_set(skb, dst_clone(dst)); 963 964 skb->transport_header = skb->network_header; 965 966 proto = fl6->flowi6_proto; 967 if (encap_limit >= 0) { 968 init_tel_txopt(&opt, encap_limit); 969 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 970 } 971 skb_push(skb, sizeof(struct ipv6hdr)); 972 skb_reset_network_header(skb); 973 ipv6h = ipv6_hdr(skb); 974 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000); 975 dsfield = INET_ECN_encapsulate(0, dsfield); 976 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield); 977 ipv6h->hop_limit = t->parms.hop_limit; 978 ipv6h->nexthdr = proto; 979 ipv6_addr_copy(&ipv6h->saddr, &fl6->saddr); 980 ipv6_addr_copy(&ipv6h->daddr, &fl6->daddr); 981 nf_reset(skb); 982 pkt_len = skb->len; 983 err = ip6_local_out(skb); 984 985 if (net_xmit_eval(err) == 0) { 986 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats); 987 988 tstats->tx_bytes += pkt_len; 989 tstats->tx_packets++; 990 } else { 991 stats->tx_errors++; 992 stats->tx_aborted_errors++; 993 } 994 ip6_tnl_dst_store(t, dst); 995 return 0; 996 tx_err_link_failure: 997 stats->tx_carrier_errors++; 998 dst_link_failure(skb); 999 tx_err_dst_release: 1000 dst_release(dst); 1001 return err; 1002 } 1003 1004 static inline int 1005 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1006 { 1007 struct ip6_tnl *t = netdev_priv(dev); 1008 struct iphdr *iph = ip_hdr(skb); 1009 int encap_limit = -1; 1010 struct flowi6 fl6; 1011 __u8 dsfield; 1012 __u32 mtu; 1013 int err; 1014 1015 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 1016 !ip6_tnl_xmit_ctl(t)) 1017 return -1; 1018 1019 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1020 encap_limit = t->parms.encap_limit; 1021 1022 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1023 fl6.flowi6_proto = IPPROTO_IPIP; 1024 1025 dsfield = ipv4_get_dsfield(iph); 1026 1027 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 1028 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 1029 & IPV6_TCLASS_MASK; 1030 1031 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1032 if (err != 0) { 1033 /* XXX: send ICMP error even if DF is not set. */ 1034 if (err == -EMSGSIZE) 1035 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1036 htonl(mtu)); 1037 return -1; 1038 } 1039 1040 return 0; 1041 } 1042 1043 static inline int 1044 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1045 { 1046 struct ip6_tnl *t = netdev_priv(dev); 1047 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1048 int encap_limit = -1; 1049 __u16 offset; 1050 struct flowi6 fl6; 1051 __u8 dsfield; 1052 __u32 mtu; 1053 int err; 1054 1055 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1056 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1057 return -1; 1058 1059 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb)); 1060 if (offset > 0) { 1061 struct ipv6_tlv_tnl_enc_lim *tel; 1062 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1063 if (tel->encap_limit == 0) { 1064 icmpv6_send(skb, ICMPV6_PARAMPROB, 1065 ICMPV6_HDR_FIELD, offset + 2); 1066 return -1; 1067 } 1068 encap_limit = tel->encap_limit - 1; 1069 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1070 encap_limit = t->parms.encap_limit; 1071 1072 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1073 fl6.flowi6_proto = IPPROTO_IPV6; 1074 1075 dsfield = ipv6_get_dsfield(ipv6h); 1076 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 1077 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1078 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1079 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); 1080 1081 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1082 if (err != 0) { 1083 if (err == -EMSGSIZE) 1084 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1085 return -1; 1086 } 1087 1088 return 0; 1089 } 1090 1091 static netdev_tx_t 1092 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1093 { 1094 struct ip6_tnl *t = netdev_priv(dev); 1095 struct net_device_stats *stats = &t->dev->stats; 1096 int ret; 1097 1098 switch (skb->protocol) { 1099 case htons(ETH_P_IP): 1100 ret = ip4ip6_tnl_xmit(skb, dev); 1101 break; 1102 case htons(ETH_P_IPV6): 1103 ret = ip6ip6_tnl_xmit(skb, dev); 1104 break; 1105 default: 1106 goto tx_err; 1107 } 1108 1109 if (ret < 0) 1110 goto tx_err; 1111 1112 return NETDEV_TX_OK; 1113 1114 tx_err: 1115 stats->tx_errors++; 1116 stats->tx_dropped++; 1117 kfree_skb(skb); 1118 return NETDEV_TX_OK; 1119 } 1120 1121 static void ip6_tnl_set_cap(struct ip6_tnl *t) 1122 { 1123 struct ip6_tnl_parm *p = &t->parms; 1124 int ltype = ipv6_addr_type(&p->laddr); 1125 int rtype = ipv6_addr_type(&p->raddr); 1126 1127 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV); 1128 1129 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1130 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1131 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 1132 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 1133 if (ltype&IPV6_ADDR_UNICAST) 1134 p->flags |= IP6_TNL_F_CAP_XMIT; 1135 if (rtype&IPV6_ADDR_UNICAST) 1136 p->flags |= IP6_TNL_F_CAP_RCV; 1137 } 1138 } 1139 1140 static void ip6_tnl_link_config(struct ip6_tnl *t) 1141 { 1142 struct net_device *dev = t->dev; 1143 struct ip6_tnl_parm *p = &t->parms; 1144 struct flowi6 *fl6 = &t->fl.u.ip6; 1145 1146 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1147 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1148 1149 /* Set up flowi template */ 1150 ipv6_addr_copy(&fl6->saddr, &p->laddr); 1151 ipv6_addr_copy(&fl6->daddr, &p->raddr); 1152 fl6->flowi6_oif = p->link; 1153 fl6->flowlabel = 0; 1154 1155 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1156 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1157 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1158 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1159 1160 ip6_tnl_set_cap(t); 1161 1162 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1163 dev->flags |= IFF_POINTOPOINT; 1164 else 1165 dev->flags &= ~IFF_POINTOPOINT; 1166 1167 dev->iflink = p->link; 1168 1169 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1170 int strict = (ipv6_addr_type(&p->raddr) & 1171 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1172 1173 struct rt6_info *rt = rt6_lookup(dev_net(dev), 1174 &p->raddr, &p->laddr, 1175 p->link, strict); 1176 1177 if (rt == NULL) 1178 return; 1179 1180 if (rt->rt6i_dev) { 1181 dev->hard_header_len = rt->rt6i_dev->hard_header_len + 1182 sizeof (struct ipv6hdr); 1183 1184 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr); 1185 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1186 dev->mtu-=8; 1187 1188 if (dev->mtu < IPV6_MIN_MTU) 1189 dev->mtu = IPV6_MIN_MTU; 1190 } 1191 dst_release(&rt->dst); 1192 } 1193 } 1194 1195 /** 1196 * ip6_tnl_change - update the tunnel parameters 1197 * @t: tunnel to be changed 1198 * @p: tunnel configuration parameters 1199 * 1200 * Description: 1201 * ip6_tnl_change() updates the tunnel parameters 1202 **/ 1203 1204 static int 1205 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p) 1206 { 1207 ipv6_addr_copy(&t->parms.laddr, &p->laddr); 1208 ipv6_addr_copy(&t->parms.raddr, &p->raddr); 1209 t->parms.flags = p->flags; 1210 t->parms.hop_limit = p->hop_limit; 1211 t->parms.encap_limit = p->encap_limit; 1212 t->parms.flowinfo = p->flowinfo; 1213 t->parms.link = p->link; 1214 t->parms.proto = p->proto; 1215 ip6_tnl_dst_reset(t); 1216 ip6_tnl_link_config(t); 1217 return 0; 1218 } 1219 1220 /** 1221 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1222 * @dev: virtual device associated with tunnel 1223 * @ifr: parameters passed from userspace 1224 * @cmd: command to be performed 1225 * 1226 * Description: 1227 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1228 * from userspace. 1229 * 1230 * The possible commands are the following: 1231 * %SIOCGETTUNNEL: get tunnel parameters for device 1232 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1233 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1234 * %SIOCDELTUNNEL: delete tunnel 1235 * 1236 * The fallback device "ip6tnl0", created during module 1237 * initialization, can be used for creating other tunnel devices. 1238 * 1239 * Return: 1240 * 0 on success, 1241 * %-EFAULT if unable to copy data to or from userspace, 1242 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1243 * %-EINVAL if passed tunnel parameters are invalid, 1244 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1245 * %-ENODEV if attempting to change or delete a nonexisting device 1246 **/ 1247 1248 static int 1249 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1250 { 1251 int err = 0; 1252 struct ip6_tnl_parm p; 1253 struct ip6_tnl *t = NULL; 1254 struct net *net = dev_net(dev); 1255 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1256 1257 switch (cmd) { 1258 case SIOCGETTUNNEL: 1259 if (dev == ip6n->fb_tnl_dev) { 1260 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1261 err = -EFAULT; 1262 break; 1263 } 1264 t = ip6_tnl_locate(net, &p, 0); 1265 } 1266 if (t == NULL) 1267 t = netdev_priv(dev); 1268 memcpy(&p, &t->parms, sizeof (p)); 1269 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1270 err = -EFAULT; 1271 } 1272 break; 1273 case SIOCADDTUNNEL: 1274 case SIOCCHGTUNNEL: 1275 err = -EPERM; 1276 if (!capable(CAP_NET_ADMIN)) 1277 break; 1278 err = -EFAULT; 1279 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1280 break; 1281 err = -EINVAL; 1282 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1283 p.proto != 0) 1284 break; 1285 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL); 1286 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1287 if (t != NULL) { 1288 if (t->dev != dev) { 1289 err = -EEXIST; 1290 break; 1291 } 1292 } else 1293 t = netdev_priv(dev); 1294 1295 ip6_tnl_unlink(ip6n, t); 1296 synchronize_net(); 1297 err = ip6_tnl_change(t, &p); 1298 ip6_tnl_link(ip6n, t); 1299 netdev_state_change(dev); 1300 } 1301 if (t) { 1302 err = 0; 1303 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p))) 1304 err = -EFAULT; 1305 1306 } else 1307 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1308 break; 1309 case SIOCDELTUNNEL: 1310 err = -EPERM; 1311 if (!capable(CAP_NET_ADMIN)) 1312 break; 1313 1314 if (dev == ip6n->fb_tnl_dev) { 1315 err = -EFAULT; 1316 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1317 break; 1318 err = -ENOENT; 1319 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL) 1320 break; 1321 err = -EPERM; 1322 if (t->dev == ip6n->fb_tnl_dev) 1323 break; 1324 dev = t->dev; 1325 } 1326 err = 0; 1327 unregister_netdevice(dev); 1328 break; 1329 default: 1330 err = -EINVAL; 1331 } 1332 return err; 1333 } 1334 1335 /** 1336 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1337 * @dev: virtual device associated with tunnel 1338 * @new_mtu: the new mtu 1339 * 1340 * Return: 1341 * 0 on success, 1342 * %-EINVAL if mtu too small 1343 **/ 1344 1345 static int 1346 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1347 { 1348 if (new_mtu < IPV6_MIN_MTU) { 1349 return -EINVAL; 1350 } 1351 dev->mtu = new_mtu; 1352 return 0; 1353 } 1354 1355 1356 static const struct net_device_ops ip6_tnl_netdev_ops = { 1357 .ndo_uninit = ip6_tnl_dev_uninit, 1358 .ndo_start_xmit = ip6_tnl_xmit, 1359 .ndo_do_ioctl = ip6_tnl_ioctl, 1360 .ndo_change_mtu = ip6_tnl_change_mtu, 1361 .ndo_get_stats = ip6_get_stats, 1362 }; 1363 1364 1365 /** 1366 * ip6_tnl_dev_setup - setup virtual tunnel device 1367 * @dev: virtual device associated with tunnel 1368 * 1369 * Description: 1370 * Initialize function pointers and device parameters 1371 **/ 1372 1373 static void ip6_tnl_dev_setup(struct net_device *dev) 1374 { 1375 struct ip6_tnl *t; 1376 1377 dev->netdev_ops = &ip6_tnl_netdev_ops; 1378 dev->destructor = ip6_dev_free; 1379 1380 dev->type = ARPHRD_TUNNEL6; 1381 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1382 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1383 t = netdev_priv(dev); 1384 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1385 dev->mtu-=8; 1386 dev->flags |= IFF_NOARP; 1387 dev->addr_len = sizeof(struct in6_addr); 1388 dev->features |= NETIF_F_NETNS_LOCAL; 1389 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1390 } 1391 1392 1393 /** 1394 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1395 * @dev: virtual device associated with tunnel 1396 **/ 1397 1398 static inline int 1399 ip6_tnl_dev_init_gen(struct net_device *dev) 1400 { 1401 struct ip6_tnl *t = netdev_priv(dev); 1402 1403 t->dev = dev; 1404 strcpy(t->parms.name, dev->name); 1405 dev->tstats = alloc_percpu(struct pcpu_tstats); 1406 if (!dev->tstats) 1407 return -ENOMEM; 1408 return 0; 1409 } 1410 1411 /** 1412 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1413 * @dev: virtual device associated with tunnel 1414 **/ 1415 1416 static int ip6_tnl_dev_init(struct net_device *dev) 1417 { 1418 struct ip6_tnl *t = netdev_priv(dev); 1419 int err = ip6_tnl_dev_init_gen(dev); 1420 1421 if (err) 1422 return err; 1423 ip6_tnl_link_config(t); 1424 return 0; 1425 } 1426 1427 /** 1428 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1429 * @dev: fallback device 1430 * 1431 * Return: 0 1432 **/ 1433 1434 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1435 { 1436 struct ip6_tnl *t = netdev_priv(dev); 1437 struct net *net = dev_net(dev); 1438 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1439 int err = ip6_tnl_dev_init_gen(dev); 1440 1441 if (err) 1442 return err; 1443 1444 t->parms.proto = IPPROTO_IPV6; 1445 dev_hold(dev); 1446 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1447 return 0; 1448 } 1449 1450 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 1451 .handler = ip4ip6_rcv, 1452 .err_handler = ip4ip6_err, 1453 .priority = 1, 1454 }; 1455 1456 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 1457 .handler = ip6ip6_rcv, 1458 .err_handler = ip6ip6_err, 1459 .priority = 1, 1460 }; 1461 1462 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n) 1463 { 1464 int h; 1465 struct ip6_tnl *t; 1466 LIST_HEAD(list); 1467 1468 for (h = 0; h < HASH_SIZE; h++) { 1469 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1470 while (t != NULL) { 1471 unregister_netdevice_queue(t->dev, &list); 1472 t = rtnl_dereference(t->next); 1473 } 1474 } 1475 1476 t = rtnl_dereference(ip6n->tnls_wc[0]); 1477 unregister_netdevice_queue(t->dev, &list); 1478 unregister_netdevice_many(&list); 1479 } 1480 1481 static int __net_init ip6_tnl_init_net(struct net *net) 1482 { 1483 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1484 int err; 1485 1486 ip6n->tnls[0] = ip6n->tnls_wc; 1487 ip6n->tnls[1] = ip6n->tnls_r_l; 1488 1489 err = -ENOMEM; 1490 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1491 ip6_tnl_dev_setup); 1492 1493 if (!ip6n->fb_tnl_dev) 1494 goto err_alloc_dev; 1495 dev_net_set(ip6n->fb_tnl_dev, net); 1496 1497 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1498 if (err < 0) 1499 goto err_register; 1500 1501 err = register_netdev(ip6n->fb_tnl_dev); 1502 if (err < 0) 1503 goto err_register; 1504 return 0; 1505 1506 err_register: 1507 ip6_dev_free(ip6n->fb_tnl_dev); 1508 err_alloc_dev: 1509 return err; 1510 } 1511 1512 static void __net_exit ip6_tnl_exit_net(struct net *net) 1513 { 1514 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1515 1516 rtnl_lock(); 1517 ip6_tnl_destroy_tunnels(ip6n); 1518 rtnl_unlock(); 1519 } 1520 1521 static struct pernet_operations ip6_tnl_net_ops = { 1522 .init = ip6_tnl_init_net, 1523 .exit = ip6_tnl_exit_net, 1524 .id = &ip6_tnl_net_id, 1525 .size = sizeof(struct ip6_tnl_net), 1526 }; 1527 1528 /** 1529 * ip6_tunnel_init - register protocol and reserve needed resources 1530 * 1531 * Return: 0 on success 1532 **/ 1533 1534 static int __init ip6_tunnel_init(void) 1535 { 1536 int err; 1537 1538 err = register_pernet_device(&ip6_tnl_net_ops); 1539 if (err < 0) 1540 goto out_pernet; 1541 1542 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 1543 if (err < 0) { 1544 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n"); 1545 goto out_ip4ip6; 1546 } 1547 1548 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 1549 if (err < 0) { 1550 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n"); 1551 goto out_ip6ip6; 1552 } 1553 1554 return 0; 1555 1556 out_ip6ip6: 1557 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1558 out_ip4ip6: 1559 unregister_pernet_device(&ip6_tnl_net_ops); 1560 out_pernet: 1561 return err; 1562 } 1563 1564 /** 1565 * ip6_tunnel_cleanup - free resources and unregister protocol 1566 **/ 1567 1568 static void __exit ip6_tunnel_cleanup(void) 1569 { 1570 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1571 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n"); 1572 1573 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1574 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n"); 1575 1576 unregister_pernet_device(&ip6_tnl_net_ops); 1577 } 1578 1579 module_init(ip6_tunnel_init); 1580 module_exit(ip6_tunnel_cleanup); 1581