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 <linux/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 } __attribute__((aligned(4*sizeof(unsigned long)))); 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, const struct in6_addr *remote, const 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, const struct ip6_tnl_parm *p) 198 { 199 const struct in6_addr *remote = &p->raddr; 200 const 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 t = netdev_priv(dev); 284 t->parms = *p; 285 err = ip6_tnl_dev_init(dev); 286 if (err < 0) 287 goto failed_free; 288 289 if ((err = register_netdevice(dev)) < 0) 290 goto failed_free; 291 292 strcpy(t->parms.name, dev->name); 293 294 dev_hold(dev); 295 ip6_tnl_link(ip6n, t); 296 return t; 297 298 failed_free: 299 ip6_dev_free(dev); 300 failed: 301 return NULL; 302 } 303 304 /** 305 * ip6_tnl_locate - find or create tunnel matching given parameters 306 * @p: tunnel parameters 307 * @create: != 0 if allowed to create new tunnel if no match found 308 * 309 * Description: 310 * ip6_tnl_locate() first tries to locate an existing tunnel 311 * based on @parms. If this is unsuccessful, but @create is set a new 312 * tunnel device is created and registered for use. 313 * 314 * Return: 315 * matching tunnel or NULL 316 **/ 317 318 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 319 struct ip6_tnl_parm *p, int create) 320 { 321 const struct in6_addr *remote = &p->raddr; 322 const struct in6_addr *local = &p->laddr; 323 struct ip6_tnl __rcu **tp; 324 struct ip6_tnl *t; 325 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 326 327 for (tp = ip6_tnl_bucket(ip6n, p); 328 (t = rtnl_dereference(*tp)) != NULL; 329 tp = &t->next) { 330 if (ipv6_addr_equal(local, &t->parms.laddr) && 331 ipv6_addr_equal(remote, &t->parms.raddr)) 332 return t; 333 } 334 if (!create) 335 return NULL; 336 return ip6_tnl_create(net, p); 337 } 338 339 /** 340 * ip6_tnl_dev_uninit - tunnel device uninitializer 341 * @dev: the device to be destroyed 342 * 343 * Description: 344 * ip6_tnl_dev_uninit() removes tunnel from its list 345 **/ 346 347 static void 348 ip6_tnl_dev_uninit(struct net_device *dev) 349 { 350 struct ip6_tnl *t = netdev_priv(dev); 351 struct net *net = dev_net(dev); 352 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 353 354 if (dev == ip6n->fb_tnl_dev) 355 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 356 else 357 ip6_tnl_unlink(ip6n, t); 358 ip6_tnl_dst_reset(t); 359 dev_put(dev); 360 } 361 362 /** 363 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 364 * @skb: received socket buffer 365 * 366 * Return: 367 * 0 if none was found, 368 * else index to encapsulation limit 369 **/ 370 371 static __u16 372 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw) 373 { 374 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw; 375 __u8 nexthdr = ipv6h->nexthdr; 376 __u16 off = sizeof (*ipv6h); 377 378 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 379 __u16 optlen = 0; 380 struct ipv6_opt_hdr *hdr; 381 if (raw + off + sizeof (*hdr) > skb->data && 382 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 383 break; 384 385 hdr = (struct ipv6_opt_hdr *) (raw + off); 386 if (nexthdr == NEXTHDR_FRAGMENT) { 387 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 388 if (frag_hdr->frag_off) 389 break; 390 optlen = 8; 391 } else if (nexthdr == NEXTHDR_AUTH) { 392 optlen = (hdr->hdrlen + 2) << 2; 393 } else { 394 optlen = ipv6_optlen(hdr); 395 } 396 if (nexthdr == NEXTHDR_DEST) { 397 __u16 i = off + 2; 398 while (1) { 399 struct ipv6_tlv_tnl_enc_lim *tel; 400 401 /* No more room for encapsulation limit */ 402 if (i + sizeof (*tel) > off + optlen) 403 break; 404 405 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 406 /* return index of option if found and valid */ 407 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 408 tel->length == 1) 409 return i; 410 /* else jump to next option */ 411 if (tel->type) 412 i += tel->length + 2; 413 else 414 i++; 415 } 416 } 417 nexthdr = hdr->nexthdr; 418 off += optlen; 419 } 420 return 0; 421 } 422 423 /** 424 * ip6_tnl_err - tunnel error handler 425 * 426 * Description: 427 * ip6_tnl_err() should handle errors in the tunnel according 428 * to the specifications in RFC 2473. 429 **/ 430 431 static int 432 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 433 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 434 { 435 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data; 436 struct ip6_tnl *t; 437 int rel_msg = 0; 438 u8 rel_type = ICMPV6_DEST_UNREACH; 439 u8 rel_code = ICMPV6_ADDR_UNREACH; 440 __u32 rel_info = 0; 441 __u16 len; 442 int err = -ENOENT; 443 444 /* If the packet doesn't contain the original IPv6 header we are 445 in trouble since we might need the source address for further 446 processing of the error. */ 447 448 rcu_read_lock(); 449 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 450 &ipv6h->saddr)) == NULL) 451 goto out; 452 453 if (t->parms.proto != ipproto && t->parms.proto != 0) 454 goto out; 455 456 err = 0; 457 458 switch (*type) { 459 __u32 teli; 460 struct ipv6_tlv_tnl_enc_lim *tel; 461 __u32 mtu; 462 case ICMPV6_DEST_UNREACH: 463 if (net_ratelimit()) 464 printk(KERN_WARNING 465 "%s: Path to destination invalid " 466 "or inactive!\n", t->parms.name); 467 rel_msg = 1; 468 break; 469 case ICMPV6_TIME_EXCEED: 470 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 471 if (net_ratelimit()) 472 printk(KERN_WARNING 473 "%s: Too small hop limit or " 474 "routing loop in tunnel!\n", 475 t->parms.name); 476 rel_msg = 1; 477 } 478 break; 479 case ICMPV6_PARAMPROB: 480 teli = 0; 481 if ((*code) == ICMPV6_HDR_FIELD) 482 teli = parse_tlv_tnl_enc_lim(skb, skb->data); 483 484 if (teli && teli == *info - 2) { 485 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 486 if (tel->encap_limit == 0) { 487 if (net_ratelimit()) 488 printk(KERN_WARNING 489 "%s: Too small encapsulation " 490 "limit or routing loop in " 491 "tunnel!\n", t->parms.name); 492 rel_msg = 1; 493 } 494 } else if (net_ratelimit()) { 495 printk(KERN_WARNING 496 "%s: Recipient unable to parse tunneled " 497 "packet!\n ", t->parms.name); 498 } 499 break; 500 case ICMPV6_PKT_TOOBIG: 501 mtu = *info - offset; 502 if (mtu < IPV6_MIN_MTU) 503 mtu = IPV6_MIN_MTU; 504 t->dev->mtu = mtu; 505 506 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 507 rel_type = ICMPV6_PKT_TOOBIG; 508 rel_code = 0; 509 rel_info = mtu; 510 rel_msg = 1; 511 } 512 break; 513 } 514 515 *type = rel_type; 516 *code = rel_code; 517 *info = rel_info; 518 *msg = rel_msg; 519 520 out: 521 rcu_read_unlock(); 522 return err; 523 } 524 525 static int 526 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 527 u8 type, u8 code, int offset, __be32 info) 528 { 529 int rel_msg = 0; 530 u8 rel_type = type; 531 u8 rel_code = code; 532 __u32 rel_info = ntohl(info); 533 int err; 534 struct sk_buff *skb2; 535 const struct iphdr *eiph; 536 struct rtable *rt; 537 struct flowi4 fl4; 538 539 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 540 &rel_msg, &rel_info, offset); 541 if (err < 0) 542 return err; 543 544 if (rel_msg == 0) 545 return 0; 546 547 switch (rel_type) { 548 case ICMPV6_DEST_UNREACH: 549 if (rel_code != ICMPV6_ADDR_UNREACH) 550 return 0; 551 rel_type = ICMP_DEST_UNREACH; 552 rel_code = ICMP_HOST_UNREACH; 553 break; 554 case ICMPV6_PKT_TOOBIG: 555 if (rel_code != 0) 556 return 0; 557 rel_type = ICMP_DEST_UNREACH; 558 rel_code = ICMP_FRAG_NEEDED; 559 break; 560 default: 561 return 0; 562 } 563 564 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 565 return 0; 566 567 skb2 = skb_clone(skb, GFP_ATOMIC); 568 if (!skb2) 569 return 0; 570 571 skb_dst_drop(skb2); 572 573 skb_pull(skb2, offset); 574 skb_reset_network_header(skb2); 575 eiph = ip_hdr(skb2); 576 577 /* Try to guess incoming interface */ 578 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 579 eiph->saddr, 0, 580 0, 0, 581 IPPROTO_IPIP, RT_TOS(eiph->tos), 0); 582 if (IS_ERR(rt)) 583 goto out; 584 585 skb2->dev = rt->dst.dev; 586 587 /* route "incoming" packet */ 588 if (rt->rt_flags & RTCF_LOCAL) { 589 ip_rt_put(rt); 590 rt = NULL; 591 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, 592 eiph->daddr, eiph->saddr, 593 0, 0, 594 IPPROTO_IPIP, 595 RT_TOS(eiph->tos), 0); 596 if (IS_ERR(rt) || 597 rt->dst.dev->type != ARPHRD_TUNNEL) { 598 if (!IS_ERR(rt)) 599 ip_rt_put(rt); 600 goto out; 601 } 602 skb_dst_set(skb2, &rt->dst); 603 } else { 604 ip_rt_put(rt); 605 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 606 skb2->dev) || 607 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 608 goto out; 609 } 610 611 /* change mtu on this route */ 612 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 613 if (rel_info > dst_mtu(skb_dst(skb2))) 614 goto out; 615 616 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info); 617 } 618 619 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 620 621 out: 622 kfree_skb(skb2); 623 return 0; 624 } 625 626 static int 627 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 628 u8 type, u8 code, int offset, __be32 info) 629 { 630 int rel_msg = 0; 631 u8 rel_type = type; 632 u8 rel_code = code; 633 __u32 rel_info = ntohl(info); 634 int err; 635 636 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 637 &rel_msg, &rel_info, offset); 638 if (err < 0) 639 return err; 640 641 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 642 struct rt6_info *rt; 643 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 644 645 if (!skb2) 646 return 0; 647 648 skb_dst_drop(skb2); 649 skb_pull(skb2, offset); 650 skb_reset_network_header(skb2); 651 652 /* Try to guess incoming interface */ 653 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 654 NULL, 0, 0); 655 656 if (rt && rt->dst.dev) 657 skb2->dev = rt->dst.dev; 658 659 icmpv6_send(skb2, rel_type, rel_code, rel_info); 660 661 if (rt) 662 dst_release(&rt->dst); 663 664 kfree_skb(skb2); 665 } 666 667 return 0; 668 } 669 670 static void ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 671 const struct ipv6hdr *ipv6h, 672 struct sk_buff *skb) 673 { 674 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 675 676 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 677 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 678 679 if (INET_ECN_is_ce(dsfield)) 680 IP_ECN_set_ce(ip_hdr(skb)); 681 } 682 683 static void ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t, 684 const struct ipv6hdr *ipv6h, 685 struct sk_buff *skb) 686 { 687 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 688 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 689 690 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h))) 691 IP6_ECN_set_ce(ipv6_hdr(skb)); 692 } 693 694 /* called with rcu_read_lock() */ 695 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t) 696 { 697 struct ip6_tnl_parm *p = &t->parms; 698 int ret = 0; 699 struct net *net = dev_net(t->dev); 700 701 if (p->flags & IP6_TNL_F_CAP_RCV) { 702 struct net_device *ldev = NULL; 703 704 if (p->link) 705 ldev = dev_get_by_index_rcu(net, p->link); 706 707 if ((ipv6_addr_is_multicast(&p->laddr) || 708 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) && 709 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0))) 710 ret = 1; 711 712 } 713 return ret; 714 } 715 716 /** 717 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 718 * @skb: received socket buffer 719 * @protocol: ethernet protocol ID 720 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 721 * 722 * Return: 0 723 **/ 724 725 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 726 __u8 ipproto, 727 void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t, 728 const struct ipv6hdr *ipv6h, 729 struct sk_buff *skb)) 730 { 731 struct ip6_tnl *t; 732 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 733 734 rcu_read_lock(); 735 736 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 737 &ipv6h->daddr)) != NULL) { 738 struct pcpu_tstats *tstats; 739 740 if (t->parms.proto != ipproto && t->parms.proto != 0) { 741 rcu_read_unlock(); 742 goto discard; 743 } 744 745 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 746 rcu_read_unlock(); 747 goto discard; 748 } 749 750 if (!ip6_tnl_rcv_ctl(t)) { 751 t->dev->stats.rx_dropped++; 752 rcu_read_unlock(); 753 goto discard; 754 } 755 secpath_reset(skb); 756 skb->mac_header = skb->network_header; 757 skb_reset_network_header(skb); 758 skb->protocol = htons(protocol); 759 skb->pkt_type = PACKET_HOST; 760 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 761 762 tstats = this_cpu_ptr(t->dev->tstats); 763 tstats->rx_packets++; 764 tstats->rx_bytes += skb->len; 765 766 __skb_tunnel_rx(skb, t->dev); 767 768 dscp_ecn_decapsulate(t, ipv6h, skb); 769 770 netif_rx(skb); 771 772 rcu_read_unlock(); 773 return 0; 774 } 775 rcu_read_unlock(); 776 return 1; 777 778 discard: 779 kfree_skb(skb); 780 return 0; 781 } 782 783 static int ip4ip6_rcv(struct sk_buff *skb) 784 { 785 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 786 ip4ip6_dscp_ecn_decapsulate); 787 } 788 789 static int ip6ip6_rcv(struct sk_buff *skb) 790 { 791 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 792 ip6ip6_dscp_ecn_decapsulate); 793 } 794 795 struct ipv6_tel_txoption { 796 struct ipv6_txoptions ops; 797 __u8 dst_opt[8]; 798 }; 799 800 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 801 { 802 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 803 804 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 805 opt->dst_opt[3] = 1; 806 opt->dst_opt[4] = encap_limit; 807 opt->dst_opt[5] = IPV6_TLV_PADN; 808 opt->dst_opt[6] = 1; 809 810 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 811 opt->ops.opt_nflen = 8; 812 } 813 814 /** 815 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 816 * @t: the outgoing tunnel device 817 * @hdr: IPv6 header from the incoming packet 818 * 819 * Description: 820 * Avoid trivial tunneling loop by checking that tunnel exit-point 821 * doesn't match source of incoming packet. 822 * 823 * Return: 824 * 1 if conflict, 825 * 0 else 826 **/ 827 828 static inline int 829 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 830 { 831 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 832 } 833 834 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 835 { 836 struct ip6_tnl_parm *p = &t->parms; 837 int ret = 0; 838 struct net *net = dev_net(t->dev); 839 840 if (p->flags & IP6_TNL_F_CAP_XMIT) { 841 struct net_device *ldev = NULL; 842 843 rcu_read_lock(); 844 if (p->link) 845 ldev = dev_get_by_index_rcu(net, p->link); 846 847 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 848 printk(KERN_WARNING 849 "%s xmit: Local address not yet configured!\n", 850 p->name); 851 else if (!ipv6_addr_is_multicast(&p->raddr) && 852 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 853 printk(KERN_WARNING 854 "%s xmit: Routing loop! " 855 "Remote address found on this node!\n", 856 p->name); 857 else 858 ret = 1; 859 rcu_read_unlock(); 860 } 861 return ret; 862 } 863 /** 864 * ip6_tnl_xmit2 - encapsulate packet and send 865 * @skb: the outgoing socket buffer 866 * @dev: the outgoing tunnel device 867 * @dsfield: dscp code for outer header 868 * @fl: flow of tunneled packet 869 * @encap_limit: encapsulation limit 870 * @pmtu: Path MTU is stored if packet is too big 871 * 872 * Description: 873 * Build new header and do some sanity checks on the packet before sending 874 * it. 875 * 876 * Return: 877 * 0 on success 878 * -1 fail 879 * %-EMSGSIZE message too big. return mtu in this case. 880 **/ 881 882 static int ip6_tnl_xmit2(struct sk_buff *skb, 883 struct net_device *dev, 884 __u8 dsfield, 885 struct flowi6 *fl6, 886 int encap_limit, 887 __u32 *pmtu) 888 { 889 struct net *net = dev_net(dev); 890 struct ip6_tnl *t = netdev_priv(dev); 891 struct net_device_stats *stats = &t->dev->stats; 892 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 893 struct ipv6_tel_txoption opt; 894 struct dst_entry *dst = NULL, *ndst = NULL; 895 struct net_device *tdev; 896 int mtu; 897 unsigned int max_headroom = sizeof(struct ipv6hdr); 898 u8 proto; 899 int err = -1; 900 int pkt_len; 901 902 if (!fl6->flowi6_mark) 903 dst = ip6_tnl_dst_check(t); 904 if (!dst) { 905 ndst = ip6_route_output(net, NULL, fl6); 906 907 if (ndst->error) 908 goto tx_err_link_failure; 909 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0); 910 if (IS_ERR(ndst)) { 911 err = PTR_ERR(ndst); 912 ndst = NULL; 913 goto tx_err_link_failure; 914 } 915 dst = ndst; 916 } 917 918 tdev = dst->dev; 919 920 if (tdev == dev) { 921 stats->collisions++; 922 if (net_ratelimit()) 923 printk(KERN_WARNING 924 "%s: Local routing loop detected!\n", 925 t->parms.name); 926 goto tx_err_dst_release; 927 } 928 mtu = dst_mtu(dst) - sizeof (*ipv6h); 929 if (encap_limit >= 0) { 930 max_headroom += 8; 931 mtu -= 8; 932 } 933 if (mtu < IPV6_MIN_MTU) 934 mtu = IPV6_MIN_MTU; 935 if (skb_dst(skb)) 936 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu); 937 if (skb->len > mtu) { 938 *pmtu = mtu; 939 err = -EMSGSIZE; 940 goto tx_err_dst_release; 941 } 942 943 /* 944 * Okay, now see if we can stuff it in the buffer as-is. 945 */ 946 max_headroom += LL_RESERVED_SPACE(tdev); 947 948 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 949 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 950 struct sk_buff *new_skb; 951 952 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 953 goto tx_err_dst_release; 954 955 if (skb->sk) 956 skb_set_owner_w(new_skb, skb->sk); 957 kfree_skb(skb); 958 skb = new_skb; 959 } 960 skb_dst_drop(skb); 961 if (fl6->flowi6_mark) { 962 skb_dst_set(skb, dst); 963 ndst = NULL; 964 } else { 965 skb_dst_set_noref(skb, dst); 966 } 967 skb->transport_header = skb->network_header; 968 969 proto = fl6->flowi6_proto; 970 if (encap_limit >= 0) { 971 init_tel_txopt(&opt, encap_limit); 972 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 973 } 974 skb_push(skb, sizeof(struct ipv6hdr)); 975 skb_reset_network_header(skb); 976 ipv6h = ipv6_hdr(skb); 977 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000); 978 dsfield = INET_ECN_encapsulate(0, dsfield); 979 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield); 980 ipv6h->hop_limit = t->parms.hop_limit; 981 ipv6h->nexthdr = proto; 982 ipv6h->saddr = fl6->saddr; 983 ipv6h->daddr = fl6->daddr; 984 nf_reset(skb); 985 pkt_len = skb->len; 986 err = ip6_local_out(skb); 987 988 if (net_xmit_eval(err) == 0) { 989 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats); 990 991 tstats->tx_bytes += pkt_len; 992 tstats->tx_packets++; 993 } else { 994 stats->tx_errors++; 995 stats->tx_aborted_errors++; 996 } 997 if (ndst) 998 ip6_tnl_dst_store(t, ndst); 999 return 0; 1000 tx_err_link_failure: 1001 stats->tx_carrier_errors++; 1002 dst_link_failure(skb); 1003 tx_err_dst_release: 1004 dst_release(ndst); 1005 return err; 1006 } 1007 1008 static inline int 1009 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1010 { 1011 struct ip6_tnl *t = netdev_priv(dev); 1012 const struct iphdr *iph = ip_hdr(skb); 1013 int encap_limit = -1; 1014 struct flowi6 fl6; 1015 __u8 dsfield; 1016 __u32 mtu; 1017 int err; 1018 1019 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 1020 !ip6_tnl_xmit_ctl(t)) 1021 return -1; 1022 1023 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1024 encap_limit = t->parms.encap_limit; 1025 1026 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1027 fl6.flowi6_proto = IPPROTO_IPIP; 1028 1029 dsfield = ipv4_get_dsfield(iph); 1030 1031 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1032 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 1033 & IPV6_TCLASS_MASK; 1034 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1035 fl6.flowi6_mark = skb->mark; 1036 1037 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1038 if (err != 0) { 1039 /* XXX: send ICMP error even if DF is not set. */ 1040 if (err == -EMSGSIZE) 1041 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 1042 htonl(mtu)); 1043 return -1; 1044 } 1045 1046 return 0; 1047 } 1048 1049 static inline int 1050 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1051 { 1052 struct ip6_tnl *t = netdev_priv(dev); 1053 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 1054 int encap_limit = -1; 1055 __u16 offset; 1056 struct flowi6 fl6; 1057 __u8 dsfield; 1058 __u32 mtu; 1059 int err; 1060 1061 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1062 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1063 return -1; 1064 1065 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb)); 1066 if (offset > 0) { 1067 struct ipv6_tlv_tnl_enc_lim *tel; 1068 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1069 if (tel->encap_limit == 0) { 1070 icmpv6_send(skb, ICMPV6_PARAMPROB, 1071 ICMPV6_HDR_FIELD, offset + 2); 1072 return -1; 1073 } 1074 encap_limit = tel->encap_limit - 1; 1075 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1076 encap_limit = t->parms.encap_limit; 1077 1078 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6)); 1079 fl6.flowi6_proto = IPPROTO_IPV6; 1080 1081 dsfield = ipv6_get_dsfield(ipv6h); 1082 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 1083 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1084 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 1085 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); 1086 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 1087 fl6.flowi6_mark = skb->mark; 1088 1089 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu); 1090 if (err != 0) { 1091 if (err == -EMSGSIZE) 1092 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1093 return -1; 1094 } 1095 1096 return 0; 1097 } 1098 1099 static netdev_tx_t 1100 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1101 { 1102 struct ip6_tnl *t = netdev_priv(dev); 1103 struct net_device_stats *stats = &t->dev->stats; 1104 int ret; 1105 1106 switch (skb->protocol) { 1107 case htons(ETH_P_IP): 1108 ret = ip4ip6_tnl_xmit(skb, dev); 1109 break; 1110 case htons(ETH_P_IPV6): 1111 ret = ip6ip6_tnl_xmit(skb, dev); 1112 break; 1113 default: 1114 goto tx_err; 1115 } 1116 1117 if (ret < 0) 1118 goto tx_err; 1119 1120 return NETDEV_TX_OK; 1121 1122 tx_err: 1123 stats->tx_errors++; 1124 stats->tx_dropped++; 1125 kfree_skb(skb); 1126 return NETDEV_TX_OK; 1127 } 1128 1129 static void ip6_tnl_set_cap(struct ip6_tnl *t) 1130 { 1131 struct ip6_tnl_parm *p = &t->parms; 1132 int ltype = ipv6_addr_type(&p->laddr); 1133 int rtype = ipv6_addr_type(&p->raddr); 1134 1135 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV); 1136 1137 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1138 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1139 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 1140 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 1141 if (ltype&IPV6_ADDR_UNICAST) 1142 p->flags |= IP6_TNL_F_CAP_XMIT; 1143 if (rtype&IPV6_ADDR_UNICAST) 1144 p->flags |= IP6_TNL_F_CAP_RCV; 1145 } 1146 } 1147 1148 static void ip6_tnl_link_config(struct ip6_tnl *t) 1149 { 1150 struct net_device *dev = t->dev; 1151 struct ip6_tnl_parm *p = &t->parms; 1152 struct flowi6 *fl6 = &t->fl.u.ip6; 1153 1154 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1155 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1156 1157 /* Set up flowi template */ 1158 fl6->saddr = p->laddr; 1159 fl6->daddr = p->raddr; 1160 fl6->flowi6_oif = p->link; 1161 fl6->flowlabel = 0; 1162 1163 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1164 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1165 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1166 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1167 1168 ip6_tnl_set_cap(t); 1169 1170 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1171 dev->flags |= IFF_POINTOPOINT; 1172 else 1173 dev->flags &= ~IFF_POINTOPOINT; 1174 1175 dev->iflink = p->link; 1176 1177 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1178 int strict = (ipv6_addr_type(&p->raddr) & 1179 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1180 1181 struct rt6_info *rt = rt6_lookup(dev_net(dev), 1182 &p->raddr, &p->laddr, 1183 p->link, strict); 1184 1185 if (rt == NULL) 1186 return; 1187 1188 if (rt->dst.dev) { 1189 dev->hard_header_len = rt->dst.dev->hard_header_len + 1190 sizeof (struct ipv6hdr); 1191 1192 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr); 1193 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1194 dev->mtu-=8; 1195 1196 if (dev->mtu < IPV6_MIN_MTU) 1197 dev->mtu = IPV6_MIN_MTU; 1198 } 1199 dst_release(&rt->dst); 1200 } 1201 } 1202 1203 /** 1204 * ip6_tnl_change - update the tunnel parameters 1205 * @t: tunnel to be changed 1206 * @p: tunnel configuration parameters 1207 * 1208 * Description: 1209 * ip6_tnl_change() updates the tunnel parameters 1210 **/ 1211 1212 static int 1213 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p) 1214 { 1215 t->parms.laddr = p->laddr; 1216 t->parms.raddr = p->raddr; 1217 t->parms.flags = p->flags; 1218 t->parms.hop_limit = p->hop_limit; 1219 t->parms.encap_limit = p->encap_limit; 1220 t->parms.flowinfo = p->flowinfo; 1221 t->parms.link = p->link; 1222 t->parms.proto = p->proto; 1223 ip6_tnl_dst_reset(t); 1224 ip6_tnl_link_config(t); 1225 return 0; 1226 } 1227 1228 /** 1229 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1230 * @dev: virtual device associated with tunnel 1231 * @ifr: parameters passed from userspace 1232 * @cmd: command to be performed 1233 * 1234 * Description: 1235 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1236 * from userspace. 1237 * 1238 * The possible commands are the following: 1239 * %SIOCGETTUNNEL: get tunnel parameters for device 1240 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1241 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1242 * %SIOCDELTUNNEL: delete tunnel 1243 * 1244 * The fallback device "ip6tnl0", created during module 1245 * initialization, can be used for creating other tunnel devices. 1246 * 1247 * Return: 1248 * 0 on success, 1249 * %-EFAULT if unable to copy data to or from userspace, 1250 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1251 * %-EINVAL if passed tunnel parameters are invalid, 1252 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1253 * %-ENODEV if attempting to change or delete a nonexisting device 1254 **/ 1255 1256 static int 1257 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1258 { 1259 int err = 0; 1260 struct ip6_tnl_parm p; 1261 struct ip6_tnl *t = NULL; 1262 struct net *net = dev_net(dev); 1263 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1264 1265 switch (cmd) { 1266 case SIOCGETTUNNEL: 1267 if (dev == ip6n->fb_tnl_dev) { 1268 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1269 err = -EFAULT; 1270 break; 1271 } 1272 t = ip6_tnl_locate(net, &p, 0); 1273 } 1274 if (t == NULL) 1275 t = netdev_priv(dev); 1276 memcpy(&p, &t->parms, sizeof (p)); 1277 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1278 err = -EFAULT; 1279 } 1280 break; 1281 case SIOCADDTUNNEL: 1282 case SIOCCHGTUNNEL: 1283 err = -EPERM; 1284 if (!capable(CAP_NET_ADMIN)) 1285 break; 1286 err = -EFAULT; 1287 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1288 break; 1289 err = -EINVAL; 1290 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1291 p.proto != 0) 1292 break; 1293 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL); 1294 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1295 if (t != NULL) { 1296 if (t->dev != dev) { 1297 err = -EEXIST; 1298 break; 1299 } 1300 } else 1301 t = netdev_priv(dev); 1302 1303 ip6_tnl_unlink(ip6n, t); 1304 synchronize_net(); 1305 err = ip6_tnl_change(t, &p); 1306 ip6_tnl_link(ip6n, t); 1307 netdev_state_change(dev); 1308 } 1309 if (t) { 1310 err = 0; 1311 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p))) 1312 err = -EFAULT; 1313 1314 } else 1315 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1316 break; 1317 case SIOCDELTUNNEL: 1318 err = -EPERM; 1319 if (!capable(CAP_NET_ADMIN)) 1320 break; 1321 1322 if (dev == ip6n->fb_tnl_dev) { 1323 err = -EFAULT; 1324 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1325 break; 1326 err = -ENOENT; 1327 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL) 1328 break; 1329 err = -EPERM; 1330 if (t->dev == ip6n->fb_tnl_dev) 1331 break; 1332 dev = t->dev; 1333 } 1334 err = 0; 1335 unregister_netdevice(dev); 1336 break; 1337 default: 1338 err = -EINVAL; 1339 } 1340 return err; 1341 } 1342 1343 /** 1344 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1345 * @dev: virtual device associated with tunnel 1346 * @new_mtu: the new mtu 1347 * 1348 * Return: 1349 * 0 on success, 1350 * %-EINVAL if mtu too small 1351 **/ 1352 1353 static int 1354 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1355 { 1356 if (new_mtu < IPV6_MIN_MTU) { 1357 return -EINVAL; 1358 } 1359 dev->mtu = new_mtu; 1360 return 0; 1361 } 1362 1363 1364 static const struct net_device_ops ip6_tnl_netdev_ops = { 1365 .ndo_uninit = ip6_tnl_dev_uninit, 1366 .ndo_start_xmit = ip6_tnl_xmit, 1367 .ndo_do_ioctl = ip6_tnl_ioctl, 1368 .ndo_change_mtu = ip6_tnl_change_mtu, 1369 .ndo_get_stats = ip6_get_stats, 1370 }; 1371 1372 1373 /** 1374 * ip6_tnl_dev_setup - setup virtual tunnel device 1375 * @dev: virtual device associated with tunnel 1376 * 1377 * Description: 1378 * Initialize function pointers and device parameters 1379 **/ 1380 1381 static void ip6_tnl_dev_setup(struct net_device *dev) 1382 { 1383 struct ip6_tnl *t; 1384 1385 dev->netdev_ops = &ip6_tnl_netdev_ops; 1386 dev->destructor = ip6_dev_free; 1387 1388 dev->type = ARPHRD_TUNNEL6; 1389 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1390 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1391 t = netdev_priv(dev); 1392 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1393 dev->mtu-=8; 1394 dev->flags |= IFF_NOARP; 1395 dev->addr_len = sizeof(struct in6_addr); 1396 dev->features |= NETIF_F_NETNS_LOCAL; 1397 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1398 } 1399 1400 1401 /** 1402 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1403 * @dev: virtual device associated with tunnel 1404 **/ 1405 1406 static inline int 1407 ip6_tnl_dev_init_gen(struct net_device *dev) 1408 { 1409 struct ip6_tnl *t = netdev_priv(dev); 1410 1411 t->dev = dev; 1412 dev->tstats = alloc_percpu(struct pcpu_tstats); 1413 if (!dev->tstats) 1414 return -ENOMEM; 1415 return 0; 1416 } 1417 1418 /** 1419 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1420 * @dev: virtual device associated with tunnel 1421 **/ 1422 1423 static int ip6_tnl_dev_init(struct net_device *dev) 1424 { 1425 struct ip6_tnl *t = netdev_priv(dev); 1426 int err = ip6_tnl_dev_init_gen(dev); 1427 1428 if (err) 1429 return err; 1430 ip6_tnl_link_config(t); 1431 return 0; 1432 } 1433 1434 /** 1435 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1436 * @dev: fallback device 1437 * 1438 * Return: 0 1439 **/ 1440 1441 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev) 1442 { 1443 struct ip6_tnl *t = netdev_priv(dev); 1444 struct net *net = dev_net(dev); 1445 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1446 int err = ip6_tnl_dev_init_gen(dev); 1447 1448 if (err) 1449 return err; 1450 1451 t->parms.proto = IPPROTO_IPV6; 1452 dev_hold(dev); 1453 rcu_assign_pointer(ip6n->tnls_wc[0], t); 1454 return 0; 1455 } 1456 1457 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = { 1458 .handler = ip4ip6_rcv, 1459 .err_handler = ip4ip6_err, 1460 .priority = 1, 1461 }; 1462 1463 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = { 1464 .handler = ip6ip6_rcv, 1465 .err_handler = ip6ip6_err, 1466 .priority = 1, 1467 }; 1468 1469 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n) 1470 { 1471 int h; 1472 struct ip6_tnl *t; 1473 LIST_HEAD(list); 1474 1475 for (h = 0; h < HASH_SIZE; h++) { 1476 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1477 while (t != NULL) { 1478 unregister_netdevice_queue(t->dev, &list); 1479 t = rtnl_dereference(t->next); 1480 } 1481 } 1482 1483 t = rtnl_dereference(ip6n->tnls_wc[0]); 1484 unregister_netdevice_queue(t->dev, &list); 1485 unregister_netdevice_many(&list); 1486 } 1487 1488 static int __net_init ip6_tnl_init_net(struct net *net) 1489 { 1490 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1491 struct ip6_tnl *t = NULL; 1492 int err; 1493 1494 ip6n->tnls[0] = ip6n->tnls_wc; 1495 ip6n->tnls[1] = ip6n->tnls_r_l; 1496 1497 err = -ENOMEM; 1498 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1499 ip6_tnl_dev_setup); 1500 1501 if (!ip6n->fb_tnl_dev) 1502 goto err_alloc_dev; 1503 dev_net_set(ip6n->fb_tnl_dev, net); 1504 1505 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1506 if (err < 0) 1507 goto err_register; 1508 1509 err = register_netdev(ip6n->fb_tnl_dev); 1510 if (err < 0) 1511 goto err_register; 1512 1513 t = netdev_priv(ip6n->fb_tnl_dev); 1514 1515 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1516 return 0; 1517 1518 err_register: 1519 ip6_dev_free(ip6n->fb_tnl_dev); 1520 err_alloc_dev: 1521 return err; 1522 } 1523 1524 static void __net_exit ip6_tnl_exit_net(struct net *net) 1525 { 1526 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1527 1528 rtnl_lock(); 1529 ip6_tnl_destroy_tunnels(ip6n); 1530 rtnl_unlock(); 1531 } 1532 1533 static struct pernet_operations ip6_tnl_net_ops = { 1534 .init = ip6_tnl_init_net, 1535 .exit = ip6_tnl_exit_net, 1536 .id = &ip6_tnl_net_id, 1537 .size = sizeof(struct ip6_tnl_net), 1538 }; 1539 1540 /** 1541 * ip6_tunnel_init - register protocol and reserve needed resources 1542 * 1543 * Return: 0 on success 1544 **/ 1545 1546 static int __init ip6_tunnel_init(void) 1547 { 1548 int err; 1549 1550 err = register_pernet_device(&ip6_tnl_net_ops); 1551 if (err < 0) 1552 goto out_pernet; 1553 1554 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET); 1555 if (err < 0) { 1556 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n"); 1557 goto out_ip4ip6; 1558 } 1559 1560 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6); 1561 if (err < 0) { 1562 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n"); 1563 goto out_ip6ip6; 1564 } 1565 1566 return 0; 1567 1568 out_ip6ip6: 1569 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1570 out_ip4ip6: 1571 unregister_pernet_device(&ip6_tnl_net_ops); 1572 out_pernet: 1573 return err; 1574 } 1575 1576 /** 1577 * ip6_tunnel_cleanup - free resources and unregister protocol 1578 **/ 1579 1580 static void __exit ip6_tunnel_cleanup(void) 1581 { 1582 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1583 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n"); 1584 1585 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1586 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n"); 1587 1588 unregister_pernet_device(&ip6_tnl_net_ops); 1589 } 1590 1591 module_init(ip6_tunnel_init); 1592 module_exit(ip6_tunnel_cleanup); 1593