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 41 #include <asm/uaccess.h> 42 #include <asm/atomic.h> 43 44 #include <net/icmp.h> 45 #include <net/ip.h> 46 #include <net/ipv6.h> 47 #include <net/ip6_route.h> 48 #include <net/addrconf.h> 49 #include <net/ip6_tunnel.h> 50 #include <net/xfrm.h> 51 #include <net/dsfield.h> 52 #include <net/inet_ecn.h> 53 #include <net/net_namespace.h> 54 #include <net/netns/generic.h> 55 56 MODULE_AUTHOR("Ville Nuorvala"); 57 MODULE_DESCRIPTION("IPv6 tunneling device"); 58 MODULE_LICENSE("GPL"); 59 60 #define IPV6_TLV_TEL_DST_SIZE 8 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 void ip6_fb_tnl_dev_init(struct net_device *dev); 78 static void ip6_tnl_dev_init(struct net_device *dev); 79 static void ip6_tnl_dev_setup(struct net_device *dev); 80 81 static int ip6_tnl_net_id; 82 struct ip6_tnl_net { 83 /* the IPv6 tunnel fallback device */ 84 struct net_device *fb_tnl_dev; 85 /* lists for storing tunnels in use */ 86 struct ip6_tnl *tnls_r_l[HASH_SIZE]; 87 struct ip6_tnl *tnls_wc[1]; 88 struct ip6_tnl **tnls[2]; 89 }; 90 91 /* lock for the tunnel lists */ 92 static DEFINE_RWLOCK(ip6_tnl_lock); 93 94 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 95 { 96 struct dst_entry *dst = t->dst_cache; 97 98 if (dst && dst->obsolete && 99 dst->ops->check(dst, t->dst_cookie) == NULL) { 100 t->dst_cache = NULL; 101 dst_release(dst); 102 return NULL; 103 } 104 105 return dst; 106 } 107 108 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t) 109 { 110 dst_release(t->dst_cache); 111 t->dst_cache = NULL; 112 } 113 114 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 115 { 116 struct rt6_info *rt = (struct rt6_info *) dst; 117 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 118 dst_release(t->dst_cache); 119 t->dst_cache = dst; 120 } 121 122 /** 123 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 124 * @remote: the address of the tunnel exit-point 125 * @local: the address of the tunnel entry-point 126 * 127 * Return: 128 * tunnel matching given end-points if found, 129 * else fallback tunnel if its device is up, 130 * else %NULL 131 **/ 132 133 static struct ip6_tnl * 134 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local) 135 { 136 unsigned h0 = HASH(remote); 137 unsigned h1 = HASH(local); 138 struct ip6_tnl *t; 139 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 140 141 for (t = ip6n->tnls_r_l[h0 ^ h1]; t; t = t->next) { 142 if (ipv6_addr_equal(local, &t->parms.laddr) && 143 ipv6_addr_equal(remote, &t->parms.raddr) && 144 (t->dev->flags & IFF_UP)) 145 return t; 146 } 147 if ((t = ip6n->tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP)) 148 return t; 149 150 return NULL; 151 } 152 153 /** 154 * ip6_tnl_bucket - get head of list matching given tunnel parameters 155 * @p: parameters containing tunnel end-points 156 * 157 * Description: 158 * ip6_tnl_bucket() returns the head of the list matching the 159 * &struct in6_addr entries laddr and raddr in @p. 160 * 161 * Return: head of IPv6 tunnel list 162 **/ 163 164 static struct ip6_tnl ** 165 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p) 166 { 167 struct in6_addr *remote = &p->raddr; 168 struct in6_addr *local = &p->laddr; 169 unsigned h = 0; 170 int prio = 0; 171 172 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 173 prio = 1; 174 h = HASH(remote) ^ HASH(local); 175 } 176 return &ip6n->tnls[prio][h]; 177 } 178 179 /** 180 * ip6_tnl_link - add tunnel to hash table 181 * @t: tunnel to be added 182 **/ 183 184 static void 185 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 186 { 187 struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms); 188 189 t->next = *tp; 190 write_lock_bh(&ip6_tnl_lock); 191 *tp = t; 192 write_unlock_bh(&ip6_tnl_lock); 193 } 194 195 /** 196 * ip6_tnl_unlink - remove tunnel from hash table 197 * @t: tunnel to be removed 198 **/ 199 200 static void 201 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 202 { 203 struct ip6_tnl **tp; 204 205 for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) { 206 if (t == *tp) { 207 write_lock_bh(&ip6_tnl_lock); 208 *tp = t->next; 209 write_unlock_bh(&ip6_tnl_lock); 210 break; 211 } 212 } 213 } 214 215 /** 216 * ip6_tnl_create() - create a new tunnel 217 * @p: tunnel parameters 218 * @pt: pointer to new tunnel 219 * 220 * Description: 221 * Create tunnel matching given parameters. 222 * 223 * Return: 224 * created tunnel or NULL 225 **/ 226 227 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p) 228 { 229 struct net_device *dev; 230 struct ip6_tnl *t; 231 char name[IFNAMSIZ]; 232 int err; 233 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 234 235 if (p->name[0]) 236 strlcpy(name, p->name, IFNAMSIZ); 237 else 238 sprintf(name, "ip6tnl%%d"); 239 240 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup); 241 if (dev == NULL) 242 goto failed; 243 244 dev_net_set(dev, net); 245 246 if (strchr(name, '%')) { 247 if (dev_alloc_name(dev, name) < 0) 248 goto failed_free; 249 } 250 251 t = netdev_priv(dev); 252 t->parms = *p; 253 ip6_tnl_dev_init(dev); 254 255 if ((err = register_netdevice(dev)) < 0) 256 goto failed_free; 257 258 dev_hold(dev); 259 ip6_tnl_link(ip6n, t); 260 return t; 261 262 failed_free: 263 free_netdev(dev); 264 failed: 265 return NULL; 266 } 267 268 /** 269 * ip6_tnl_locate - find or create tunnel matching given parameters 270 * @p: tunnel parameters 271 * @create: != 0 if allowed to create new tunnel if no match found 272 * 273 * Description: 274 * ip6_tnl_locate() first tries to locate an existing tunnel 275 * based on @parms. If this is unsuccessful, but @create is set a new 276 * tunnel device is created and registered for use. 277 * 278 * Return: 279 * matching tunnel or NULL 280 **/ 281 282 static struct ip6_tnl *ip6_tnl_locate(struct net *net, 283 struct ip6_tnl_parm *p, int create) 284 { 285 struct in6_addr *remote = &p->raddr; 286 struct in6_addr *local = &p->laddr; 287 struct ip6_tnl *t; 288 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 289 290 for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) { 291 if (ipv6_addr_equal(local, &t->parms.laddr) && 292 ipv6_addr_equal(remote, &t->parms.raddr)) 293 return t; 294 } 295 if (!create) 296 return NULL; 297 return ip6_tnl_create(net, p); 298 } 299 300 /** 301 * ip6_tnl_dev_uninit - tunnel device uninitializer 302 * @dev: the device to be destroyed 303 * 304 * Description: 305 * ip6_tnl_dev_uninit() removes tunnel from its list 306 **/ 307 308 static void 309 ip6_tnl_dev_uninit(struct net_device *dev) 310 { 311 struct ip6_tnl *t = netdev_priv(dev); 312 struct net *net = dev_net(dev); 313 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 314 315 if (dev == ip6n->fb_tnl_dev) { 316 write_lock_bh(&ip6_tnl_lock); 317 ip6n->tnls_wc[0] = NULL; 318 write_unlock_bh(&ip6_tnl_lock); 319 } else { 320 ip6_tnl_unlink(ip6n, t); 321 } 322 ip6_tnl_dst_reset(t); 323 dev_put(dev); 324 } 325 326 /** 327 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 328 * @skb: received socket buffer 329 * 330 * Return: 331 * 0 if none was found, 332 * else index to encapsulation limit 333 **/ 334 335 static __u16 336 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw) 337 { 338 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw; 339 __u8 nexthdr = ipv6h->nexthdr; 340 __u16 off = sizeof (*ipv6h); 341 342 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 343 __u16 optlen = 0; 344 struct ipv6_opt_hdr *hdr; 345 if (raw + off + sizeof (*hdr) > skb->data && 346 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 347 break; 348 349 hdr = (struct ipv6_opt_hdr *) (raw + off); 350 if (nexthdr == NEXTHDR_FRAGMENT) { 351 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 352 if (frag_hdr->frag_off) 353 break; 354 optlen = 8; 355 } else if (nexthdr == NEXTHDR_AUTH) { 356 optlen = (hdr->hdrlen + 2) << 2; 357 } else { 358 optlen = ipv6_optlen(hdr); 359 } 360 if (nexthdr == NEXTHDR_DEST) { 361 __u16 i = off + 2; 362 while (1) { 363 struct ipv6_tlv_tnl_enc_lim *tel; 364 365 /* No more room for encapsulation limit */ 366 if (i + sizeof (*tel) > off + optlen) 367 break; 368 369 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 370 /* return index of option if found and valid */ 371 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 372 tel->length == 1) 373 return i; 374 /* else jump to next option */ 375 if (tel->type) 376 i += tel->length + 2; 377 else 378 i++; 379 } 380 } 381 nexthdr = hdr->nexthdr; 382 off += optlen; 383 } 384 return 0; 385 } 386 387 /** 388 * ip6_tnl_err - tunnel error handler 389 * 390 * Description: 391 * ip6_tnl_err() should handle errors in the tunnel according 392 * to the specifications in RFC 2473. 393 **/ 394 395 static int 396 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 397 u8 *type, u8 *code, int *msg, __u32 *info, int offset) 398 { 399 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data; 400 struct ip6_tnl *t; 401 int rel_msg = 0; 402 u8 rel_type = ICMPV6_DEST_UNREACH; 403 u8 rel_code = ICMPV6_ADDR_UNREACH; 404 __u32 rel_info = 0; 405 __u16 len; 406 int err = -ENOENT; 407 408 /* If the packet doesn't contain the original IPv6 header we are 409 in trouble since we might need the source address for further 410 processing of the error. */ 411 412 read_lock(&ip6_tnl_lock); 413 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 414 &ipv6h->saddr)) == NULL) 415 goto out; 416 417 if (t->parms.proto != ipproto && t->parms.proto != 0) 418 goto out; 419 420 err = 0; 421 422 switch (*type) { 423 __u32 teli; 424 struct ipv6_tlv_tnl_enc_lim *tel; 425 __u32 mtu; 426 case ICMPV6_DEST_UNREACH: 427 if (net_ratelimit()) 428 printk(KERN_WARNING 429 "%s: Path to destination invalid " 430 "or inactive!\n", t->parms.name); 431 rel_msg = 1; 432 break; 433 case ICMPV6_TIME_EXCEED: 434 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 435 if (net_ratelimit()) 436 printk(KERN_WARNING 437 "%s: Too small hop limit or " 438 "routing loop in tunnel!\n", 439 t->parms.name); 440 rel_msg = 1; 441 } 442 break; 443 case ICMPV6_PARAMPROB: 444 teli = 0; 445 if ((*code) == ICMPV6_HDR_FIELD) 446 teli = parse_tlv_tnl_enc_lim(skb, skb->data); 447 448 if (teli && teli == *info - 2) { 449 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 450 if (tel->encap_limit == 0) { 451 if (net_ratelimit()) 452 printk(KERN_WARNING 453 "%s: Too small encapsulation " 454 "limit or routing loop in " 455 "tunnel!\n", t->parms.name); 456 rel_msg = 1; 457 } 458 } else if (net_ratelimit()) { 459 printk(KERN_WARNING 460 "%s: Recipient unable to parse tunneled " 461 "packet!\n ", t->parms.name); 462 } 463 break; 464 case ICMPV6_PKT_TOOBIG: 465 mtu = *info - offset; 466 if (mtu < IPV6_MIN_MTU) 467 mtu = IPV6_MIN_MTU; 468 t->dev->mtu = mtu; 469 470 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 471 rel_type = ICMPV6_PKT_TOOBIG; 472 rel_code = 0; 473 rel_info = mtu; 474 rel_msg = 1; 475 } 476 break; 477 } 478 479 *type = rel_type; 480 *code = rel_code; 481 *info = rel_info; 482 *msg = rel_msg; 483 484 out: 485 read_unlock(&ip6_tnl_lock); 486 return err; 487 } 488 489 static int 490 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 491 u8 type, u8 code, int offset, __be32 info) 492 { 493 int rel_msg = 0; 494 u8 rel_type = type; 495 u8 rel_code = code; 496 __u32 rel_info = ntohl(info); 497 int err; 498 struct sk_buff *skb2; 499 struct iphdr *eiph; 500 struct flowi fl; 501 struct rtable *rt; 502 503 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 504 &rel_msg, &rel_info, offset); 505 if (err < 0) 506 return err; 507 508 if (rel_msg == 0) 509 return 0; 510 511 switch (rel_type) { 512 case ICMPV6_DEST_UNREACH: 513 if (rel_code != ICMPV6_ADDR_UNREACH) 514 return 0; 515 rel_type = ICMP_DEST_UNREACH; 516 rel_code = ICMP_HOST_UNREACH; 517 break; 518 case ICMPV6_PKT_TOOBIG: 519 if (rel_code != 0) 520 return 0; 521 rel_type = ICMP_DEST_UNREACH; 522 rel_code = ICMP_FRAG_NEEDED; 523 break; 524 default: 525 return 0; 526 } 527 528 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 529 return 0; 530 531 skb2 = skb_clone(skb, GFP_ATOMIC); 532 if (!skb2) 533 return 0; 534 535 skb_dst_drop(skb2); 536 537 skb_pull(skb2, offset); 538 skb_reset_network_header(skb2); 539 eiph = ip_hdr(skb2); 540 541 /* Try to guess incoming interface */ 542 memset(&fl, 0, sizeof(fl)); 543 fl.fl4_dst = eiph->saddr; 544 fl.fl4_tos = RT_TOS(eiph->tos); 545 fl.proto = IPPROTO_IPIP; 546 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl)) 547 goto out; 548 549 skb2->dev = rt->u.dst.dev; 550 551 /* route "incoming" packet */ 552 if (rt->rt_flags & RTCF_LOCAL) { 553 ip_rt_put(rt); 554 rt = NULL; 555 fl.fl4_dst = eiph->daddr; 556 fl.fl4_src = eiph->saddr; 557 fl.fl4_tos = eiph->tos; 558 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) || 559 rt->u.dst.dev->type != ARPHRD_TUNNEL) { 560 ip_rt_put(rt); 561 goto out; 562 } 563 skb_dst_set(skb2, (struct dst_entry *)rt); 564 } else { 565 ip_rt_put(rt); 566 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 567 skb2->dev) || 568 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL) 569 goto out; 570 } 571 572 /* change mtu on this route */ 573 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 574 if (rel_info > dst_mtu(skb_dst(skb2))) 575 goto out; 576 577 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info); 578 } 579 580 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 581 582 out: 583 kfree_skb(skb2); 584 return 0; 585 } 586 587 static int 588 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 589 u8 type, u8 code, int offset, __be32 info) 590 { 591 int rel_msg = 0; 592 u8 rel_type = type; 593 u8 rel_code = code; 594 __u32 rel_info = ntohl(info); 595 int err; 596 597 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 598 &rel_msg, &rel_info, offset); 599 if (err < 0) 600 return err; 601 602 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 603 struct rt6_info *rt; 604 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 605 606 if (!skb2) 607 return 0; 608 609 skb_dst_drop(skb2); 610 skb_pull(skb2, offset); 611 skb_reset_network_header(skb2); 612 613 /* Try to guess incoming interface */ 614 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 615 NULL, 0, 0); 616 617 if (rt && rt->rt6i_dev) 618 skb2->dev = rt->rt6i_dev; 619 620 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev); 621 622 if (rt) 623 dst_release(&rt->u.dst); 624 625 kfree_skb(skb2); 626 } 627 628 return 0; 629 } 630 631 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t, 632 struct ipv6hdr *ipv6h, 633 struct sk_buff *skb) 634 { 635 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 636 637 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 638 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 639 640 if (INET_ECN_is_ce(dsfield)) 641 IP_ECN_set_ce(ip_hdr(skb)); 642 } 643 644 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t, 645 struct ipv6hdr *ipv6h, 646 struct sk_buff *skb) 647 { 648 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 649 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 650 651 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h))) 652 IP6_ECN_set_ce(ipv6_hdr(skb)); 653 } 654 655 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t) 656 { 657 struct ip6_tnl_parm *p = &t->parms; 658 int ret = 0; 659 struct net *net = dev_net(t->dev); 660 661 if (p->flags & IP6_TNL_F_CAP_RCV) { 662 struct net_device *ldev = NULL; 663 664 if (p->link) 665 ldev = dev_get_by_index(net, p->link); 666 667 if ((ipv6_addr_is_multicast(&p->laddr) || 668 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) && 669 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0))) 670 ret = 1; 671 672 if (ldev) 673 dev_put(ldev); 674 } 675 return ret; 676 } 677 678 /** 679 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 680 * @skb: received socket buffer 681 * @protocol: ethernet protocol ID 682 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 683 * 684 * Return: 0 685 **/ 686 687 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 688 __u8 ipproto, 689 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t, 690 struct ipv6hdr *ipv6h, 691 struct sk_buff *skb)) 692 { 693 struct ip6_tnl *t; 694 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 695 696 read_lock(&ip6_tnl_lock); 697 698 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 699 &ipv6h->daddr)) != NULL) { 700 if (t->parms.proto != ipproto && t->parms.proto != 0) { 701 read_unlock(&ip6_tnl_lock); 702 goto discard; 703 } 704 705 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 706 read_unlock(&ip6_tnl_lock); 707 goto discard; 708 } 709 710 if (!ip6_tnl_rcv_ctl(t)) { 711 t->dev->stats.rx_dropped++; 712 read_unlock(&ip6_tnl_lock); 713 goto discard; 714 } 715 secpath_reset(skb); 716 skb->mac_header = skb->network_header; 717 skb_reset_network_header(skb); 718 skb->protocol = htons(protocol); 719 skb->pkt_type = PACKET_HOST; 720 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 721 skb->dev = t->dev; 722 skb_dst_drop(skb); 723 nf_reset(skb); 724 725 dscp_ecn_decapsulate(t, ipv6h, skb); 726 727 t->dev->stats.rx_packets++; 728 t->dev->stats.rx_bytes += skb->len; 729 netif_rx(skb); 730 read_unlock(&ip6_tnl_lock); 731 return 0; 732 } 733 read_unlock(&ip6_tnl_lock); 734 return 1; 735 736 discard: 737 kfree_skb(skb); 738 return 0; 739 } 740 741 static int ip4ip6_rcv(struct sk_buff *skb) 742 { 743 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 744 ip4ip6_dscp_ecn_decapsulate); 745 } 746 747 static int ip6ip6_rcv(struct sk_buff *skb) 748 { 749 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 750 ip6ip6_dscp_ecn_decapsulate); 751 } 752 753 struct ipv6_tel_txoption { 754 struct ipv6_txoptions ops; 755 __u8 dst_opt[8]; 756 }; 757 758 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 759 { 760 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 761 762 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 763 opt->dst_opt[3] = 1; 764 opt->dst_opt[4] = encap_limit; 765 opt->dst_opt[5] = IPV6_TLV_PADN; 766 opt->dst_opt[6] = 1; 767 768 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 769 opt->ops.opt_nflen = 8; 770 } 771 772 /** 773 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 774 * @t: the outgoing tunnel device 775 * @hdr: IPv6 header from the incoming packet 776 * 777 * Description: 778 * Avoid trivial tunneling loop by checking that tunnel exit-point 779 * doesn't match source of incoming packet. 780 * 781 * Return: 782 * 1 if conflict, 783 * 0 else 784 **/ 785 786 static inline int 787 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr) 788 { 789 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 790 } 791 792 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 793 { 794 struct ip6_tnl_parm *p = &t->parms; 795 int ret = 0; 796 struct net *net = dev_net(t->dev); 797 798 if (p->flags & IP6_TNL_F_CAP_XMIT) { 799 struct net_device *ldev = NULL; 800 801 if (p->link) 802 ldev = dev_get_by_index(net, p->link); 803 804 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 805 printk(KERN_WARNING 806 "%s xmit: Local address not yet configured!\n", 807 p->name); 808 else if (!ipv6_addr_is_multicast(&p->raddr) && 809 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 810 printk(KERN_WARNING 811 "%s xmit: Routing loop! " 812 "Remote address found on this node!\n", 813 p->name); 814 else 815 ret = 1; 816 if (ldev) 817 dev_put(ldev); 818 } 819 return ret; 820 } 821 /** 822 * ip6_tnl_xmit2 - encapsulate packet and send 823 * @skb: the outgoing socket buffer 824 * @dev: the outgoing tunnel device 825 * @dsfield: dscp code for outer header 826 * @fl: flow of tunneled packet 827 * @encap_limit: encapsulation limit 828 * @pmtu: Path MTU is stored if packet is too big 829 * 830 * Description: 831 * Build new header and do some sanity checks on the packet before sending 832 * it. 833 * 834 * Return: 835 * 0 on success 836 * -1 fail 837 * %-EMSGSIZE message too big. return mtu in this case. 838 **/ 839 840 static int ip6_tnl_xmit2(struct sk_buff *skb, 841 struct net_device *dev, 842 __u8 dsfield, 843 struct flowi *fl, 844 int encap_limit, 845 __u32 *pmtu) 846 { 847 struct net *net = dev_net(dev); 848 struct ip6_tnl *t = netdev_priv(dev); 849 struct net_device_stats *stats = &t->dev->stats; 850 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 851 struct ipv6_tel_txoption opt; 852 struct dst_entry *dst; 853 struct net_device *tdev; 854 int mtu; 855 unsigned int max_headroom = sizeof(struct ipv6hdr); 856 u8 proto; 857 int err = -1; 858 int pkt_len; 859 860 if ((dst = ip6_tnl_dst_check(t)) != NULL) 861 dst_hold(dst); 862 else { 863 dst = ip6_route_output(net, NULL, fl); 864 865 if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0) 866 goto tx_err_link_failure; 867 } 868 869 tdev = dst->dev; 870 871 if (tdev == dev) { 872 stats->collisions++; 873 if (net_ratelimit()) 874 printk(KERN_WARNING 875 "%s: Local routing loop detected!\n", 876 t->parms.name); 877 goto tx_err_dst_release; 878 } 879 mtu = dst_mtu(dst) - sizeof (*ipv6h); 880 if (encap_limit >= 0) { 881 max_headroom += 8; 882 mtu -= 8; 883 } 884 if (mtu < IPV6_MIN_MTU) 885 mtu = IPV6_MIN_MTU; 886 if (skb_dst(skb)) 887 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu); 888 if (skb->len > mtu) { 889 *pmtu = mtu; 890 err = -EMSGSIZE; 891 goto tx_err_dst_release; 892 } 893 894 /* 895 * Okay, now see if we can stuff it in the buffer as-is. 896 */ 897 max_headroom += LL_RESERVED_SPACE(tdev); 898 899 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 900 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 901 struct sk_buff *new_skb; 902 903 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 904 goto tx_err_dst_release; 905 906 if (skb->sk) 907 skb_set_owner_w(new_skb, skb->sk); 908 kfree_skb(skb); 909 skb = new_skb; 910 } 911 skb_dst_drop(skb); 912 skb_dst_set(skb, dst_clone(dst)); 913 914 skb->transport_header = skb->network_header; 915 916 proto = fl->proto; 917 if (encap_limit >= 0) { 918 init_tel_txopt(&opt, encap_limit); 919 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 920 } 921 skb_push(skb, sizeof(struct ipv6hdr)); 922 skb_reset_network_header(skb); 923 ipv6h = ipv6_hdr(skb); 924 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000); 925 dsfield = INET_ECN_encapsulate(0, dsfield); 926 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield); 927 ipv6h->hop_limit = t->parms.hop_limit; 928 ipv6h->nexthdr = proto; 929 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src); 930 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst); 931 nf_reset(skb); 932 pkt_len = skb->len; 933 err = ip6_local_out(skb); 934 935 if (net_xmit_eval(err) == 0) { 936 stats->tx_bytes += pkt_len; 937 stats->tx_packets++; 938 } else { 939 stats->tx_errors++; 940 stats->tx_aborted_errors++; 941 } 942 ip6_tnl_dst_store(t, dst); 943 return 0; 944 tx_err_link_failure: 945 stats->tx_carrier_errors++; 946 dst_link_failure(skb); 947 tx_err_dst_release: 948 dst_release(dst); 949 return err; 950 } 951 952 static inline int 953 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 954 { 955 struct ip6_tnl *t = netdev_priv(dev); 956 struct iphdr *iph = ip_hdr(skb); 957 int encap_limit = -1; 958 struct flowi fl; 959 __u8 dsfield; 960 __u32 mtu; 961 int err; 962 963 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 964 !ip6_tnl_xmit_ctl(t)) 965 return -1; 966 967 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 968 encap_limit = t->parms.encap_limit; 969 970 memcpy(&fl, &t->fl, sizeof (fl)); 971 fl.proto = IPPROTO_IPIP; 972 973 dsfield = ipv4_get_dsfield(iph); 974 975 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 976 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 977 & IPV6_TCLASS_MASK; 978 979 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu); 980 if (err != 0) { 981 /* XXX: send ICMP error even if DF is not set. */ 982 if (err == -EMSGSIZE) 983 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 984 htonl(mtu)); 985 return -1; 986 } 987 988 return 0; 989 } 990 991 static inline int 992 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 993 { 994 struct ip6_tnl *t = netdev_priv(dev); 995 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 996 int encap_limit = -1; 997 __u16 offset; 998 struct flowi fl; 999 __u8 dsfield; 1000 __u32 mtu; 1001 int err; 1002 1003 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1004 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1005 return -1; 1006 1007 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb)); 1008 if (offset > 0) { 1009 struct ipv6_tlv_tnl_enc_lim *tel; 1010 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1011 if (tel->encap_limit == 0) { 1012 icmpv6_send(skb, ICMPV6_PARAMPROB, 1013 ICMPV6_HDR_FIELD, offset + 2, skb->dev); 1014 return -1; 1015 } 1016 encap_limit = tel->encap_limit - 1; 1017 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1018 encap_limit = t->parms.encap_limit; 1019 1020 memcpy(&fl, &t->fl, sizeof (fl)); 1021 fl.proto = IPPROTO_IPV6; 1022 1023 dsfield = ipv6_get_dsfield(ipv6h); 1024 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 1025 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1026 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1027 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); 1028 1029 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu); 1030 if (err != 0) { 1031 if (err == -EMSGSIZE) 1032 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev); 1033 return -1; 1034 } 1035 1036 return 0; 1037 } 1038 1039 static netdev_tx_t 1040 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1041 { 1042 struct ip6_tnl *t = netdev_priv(dev); 1043 struct net_device_stats *stats = &t->dev->stats; 1044 int ret; 1045 1046 if (t->recursion++) { 1047 stats->collisions++; 1048 goto tx_err; 1049 } 1050 1051 switch (skb->protocol) { 1052 case htons(ETH_P_IP): 1053 ret = ip4ip6_tnl_xmit(skb, dev); 1054 break; 1055 case htons(ETH_P_IPV6): 1056 ret = ip6ip6_tnl_xmit(skb, dev); 1057 break; 1058 default: 1059 goto tx_err; 1060 } 1061 1062 if (ret < 0) 1063 goto tx_err; 1064 1065 t->recursion--; 1066 return NETDEV_TX_OK; 1067 1068 tx_err: 1069 stats->tx_errors++; 1070 stats->tx_dropped++; 1071 kfree_skb(skb); 1072 t->recursion--; 1073 return NETDEV_TX_OK; 1074 } 1075 1076 static void ip6_tnl_set_cap(struct ip6_tnl *t) 1077 { 1078 struct ip6_tnl_parm *p = &t->parms; 1079 int ltype = ipv6_addr_type(&p->laddr); 1080 int rtype = ipv6_addr_type(&p->raddr); 1081 1082 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV); 1083 1084 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1085 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1086 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 1087 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 1088 if (ltype&IPV6_ADDR_UNICAST) 1089 p->flags |= IP6_TNL_F_CAP_XMIT; 1090 if (rtype&IPV6_ADDR_UNICAST) 1091 p->flags |= IP6_TNL_F_CAP_RCV; 1092 } 1093 } 1094 1095 static void ip6_tnl_link_config(struct ip6_tnl *t) 1096 { 1097 struct net_device *dev = t->dev; 1098 struct ip6_tnl_parm *p = &t->parms; 1099 struct flowi *fl = &t->fl; 1100 1101 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1102 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1103 1104 /* Set up flowi template */ 1105 ipv6_addr_copy(&fl->fl6_src, &p->laddr); 1106 ipv6_addr_copy(&fl->fl6_dst, &p->raddr); 1107 fl->oif = p->link; 1108 fl->fl6_flowlabel = 0; 1109 1110 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1111 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1112 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1113 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1114 1115 ip6_tnl_set_cap(t); 1116 1117 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1118 dev->flags |= IFF_POINTOPOINT; 1119 else 1120 dev->flags &= ~IFF_POINTOPOINT; 1121 1122 dev->iflink = p->link; 1123 1124 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1125 int strict = (ipv6_addr_type(&p->raddr) & 1126 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1127 1128 struct rt6_info *rt = rt6_lookup(dev_net(dev), 1129 &p->raddr, &p->laddr, 1130 p->link, strict); 1131 1132 if (rt == NULL) 1133 return; 1134 1135 if (rt->rt6i_dev) { 1136 dev->hard_header_len = rt->rt6i_dev->hard_header_len + 1137 sizeof (struct ipv6hdr); 1138 1139 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr); 1140 1141 if (dev->mtu < IPV6_MIN_MTU) 1142 dev->mtu = IPV6_MIN_MTU; 1143 } 1144 dst_release(&rt->u.dst); 1145 } 1146 } 1147 1148 /** 1149 * ip6_tnl_change - update the tunnel parameters 1150 * @t: tunnel to be changed 1151 * @p: tunnel configuration parameters 1152 * 1153 * Description: 1154 * ip6_tnl_change() updates the tunnel parameters 1155 **/ 1156 1157 static int 1158 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p) 1159 { 1160 ipv6_addr_copy(&t->parms.laddr, &p->laddr); 1161 ipv6_addr_copy(&t->parms.raddr, &p->raddr); 1162 t->parms.flags = p->flags; 1163 t->parms.hop_limit = p->hop_limit; 1164 t->parms.encap_limit = p->encap_limit; 1165 t->parms.flowinfo = p->flowinfo; 1166 t->parms.link = p->link; 1167 t->parms.proto = p->proto; 1168 ip6_tnl_dst_reset(t); 1169 ip6_tnl_link_config(t); 1170 return 0; 1171 } 1172 1173 /** 1174 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1175 * @dev: virtual device associated with tunnel 1176 * @ifr: parameters passed from userspace 1177 * @cmd: command to be performed 1178 * 1179 * Description: 1180 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1181 * from userspace. 1182 * 1183 * The possible commands are the following: 1184 * %SIOCGETTUNNEL: get tunnel parameters for device 1185 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1186 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1187 * %SIOCDELTUNNEL: delete tunnel 1188 * 1189 * The fallback device "ip6tnl0", created during module 1190 * initialization, can be used for creating other tunnel devices. 1191 * 1192 * Return: 1193 * 0 on success, 1194 * %-EFAULT if unable to copy data to or from userspace, 1195 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1196 * %-EINVAL if passed tunnel parameters are invalid, 1197 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1198 * %-ENODEV if attempting to change or delete a nonexisting device 1199 **/ 1200 1201 static int 1202 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1203 { 1204 int err = 0; 1205 struct ip6_tnl_parm p; 1206 struct ip6_tnl *t = NULL; 1207 struct net *net = dev_net(dev); 1208 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1209 1210 switch (cmd) { 1211 case SIOCGETTUNNEL: 1212 if (dev == ip6n->fb_tnl_dev) { 1213 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1214 err = -EFAULT; 1215 break; 1216 } 1217 t = ip6_tnl_locate(net, &p, 0); 1218 } 1219 if (t == NULL) 1220 t = netdev_priv(dev); 1221 memcpy(&p, &t->parms, sizeof (p)); 1222 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1223 err = -EFAULT; 1224 } 1225 break; 1226 case SIOCADDTUNNEL: 1227 case SIOCCHGTUNNEL: 1228 err = -EPERM; 1229 if (!capable(CAP_NET_ADMIN)) 1230 break; 1231 err = -EFAULT; 1232 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1233 break; 1234 err = -EINVAL; 1235 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1236 p.proto != 0) 1237 break; 1238 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL); 1239 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1240 if (t != NULL) { 1241 if (t->dev != dev) { 1242 err = -EEXIST; 1243 break; 1244 } 1245 } else 1246 t = netdev_priv(dev); 1247 1248 ip6_tnl_unlink(ip6n, t); 1249 err = ip6_tnl_change(t, &p); 1250 ip6_tnl_link(ip6n, t); 1251 netdev_state_change(dev); 1252 } 1253 if (t) { 1254 err = 0; 1255 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p))) 1256 err = -EFAULT; 1257 1258 } else 1259 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1260 break; 1261 case SIOCDELTUNNEL: 1262 err = -EPERM; 1263 if (!capable(CAP_NET_ADMIN)) 1264 break; 1265 1266 if (dev == ip6n->fb_tnl_dev) { 1267 err = -EFAULT; 1268 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1269 break; 1270 err = -ENOENT; 1271 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL) 1272 break; 1273 err = -EPERM; 1274 if (t->dev == ip6n->fb_tnl_dev) 1275 break; 1276 dev = t->dev; 1277 } 1278 err = 0; 1279 unregister_netdevice(dev); 1280 break; 1281 default: 1282 err = -EINVAL; 1283 } 1284 return err; 1285 } 1286 1287 /** 1288 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1289 * @dev: virtual device associated with tunnel 1290 * @new_mtu: the new mtu 1291 * 1292 * Return: 1293 * 0 on success, 1294 * %-EINVAL if mtu too small 1295 **/ 1296 1297 static int 1298 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1299 { 1300 if (new_mtu < IPV6_MIN_MTU) { 1301 return -EINVAL; 1302 } 1303 dev->mtu = new_mtu; 1304 return 0; 1305 } 1306 1307 1308 static const struct net_device_ops ip6_tnl_netdev_ops = { 1309 .ndo_uninit = ip6_tnl_dev_uninit, 1310 .ndo_start_xmit = ip6_tnl_xmit, 1311 .ndo_do_ioctl = ip6_tnl_ioctl, 1312 .ndo_change_mtu = ip6_tnl_change_mtu, 1313 }; 1314 1315 /** 1316 * ip6_tnl_dev_setup - setup virtual tunnel device 1317 * @dev: virtual device associated with tunnel 1318 * 1319 * Description: 1320 * Initialize function pointers and device parameters 1321 **/ 1322 1323 static void ip6_tnl_dev_setup(struct net_device *dev) 1324 { 1325 dev->netdev_ops = &ip6_tnl_netdev_ops; 1326 dev->destructor = free_netdev; 1327 1328 dev->type = ARPHRD_TUNNEL6; 1329 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1330 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1331 dev->flags |= IFF_NOARP; 1332 dev->addr_len = sizeof(struct in6_addr); 1333 dev->features |= NETIF_F_NETNS_LOCAL; 1334 } 1335 1336 1337 /** 1338 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1339 * @dev: virtual device associated with tunnel 1340 **/ 1341 1342 static inline void 1343 ip6_tnl_dev_init_gen(struct net_device *dev) 1344 { 1345 struct ip6_tnl *t = netdev_priv(dev); 1346 t->dev = dev; 1347 strcpy(t->parms.name, dev->name); 1348 } 1349 1350 /** 1351 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1352 * @dev: virtual device associated with tunnel 1353 **/ 1354 1355 static void ip6_tnl_dev_init(struct net_device *dev) 1356 { 1357 struct ip6_tnl *t = netdev_priv(dev); 1358 ip6_tnl_dev_init_gen(dev); 1359 ip6_tnl_link_config(t); 1360 } 1361 1362 /** 1363 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1364 * @dev: fallback device 1365 * 1366 * Return: 0 1367 **/ 1368 1369 static void ip6_fb_tnl_dev_init(struct net_device *dev) 1370 { 1371 struct ip6_tnl *t = netdev_priv(dev); 1372 struct net *net = dev_net(dev); 1373 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1374 1375 ip6_tnl_dev_init_gen(dev); 1376 t->parms.proto = IPPROTO_IPV6; 1377 dev_hold(dev); 1378 ip6n->tnls_wc[0] = t; 1379 } 1380 1381 static struct xfrm6_tunnel ip4ip6_handler = { 1382 .handler = ip4ip6_rcv, 1383 .err_handler = ip4ip6_err, 1384 .priority = 1, 1385 }; 1386 1387 static struct xfrm6_tunnel ip6ip6_handler = { 1388 .handler = ip6ip6_rcv, 1389 .err_handler = ip6ip6_err, 1390 .priority = 1, 1391 }; 1392 1393 static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n) 1394 { 1395 int h; 1396 struct ip6_tnl *t; 1397 1398 for (h = 0; h < HASH_SIZE; h++) { 1399 while ((t = ip6n->tnls_r_l[h]) != NULL) 1400 unregister_netdevice(t->dev); 1401 } 1402 1403 t = ip6n->tnls_wc[0]; 1404 unregister_netdevice(t->dev); 1405 } 1406 1407 static int ip6_tnl_init_net(struct net *net) 1408 { 1409 int err; 1410 struct ip6_tnl_net *ip6n; 1411 1412 err = -ENOMEM; 1413 ip6n = kzalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL); 1414 if (ip6n == NULL) 1415 goto err_alloc; 1416 1417 err = net_assign_generic(net, ip6_tnl_net_id, ip6n); 1418 if (err < 0) 1419 goto err_assign; 1420 1421 ip6n->tnls[0] = ip6n->tnls_wc; 1422 ip6n->tnls[1] = ip6n->tnls_r_l; 1423 1424 err = -ENOMEM; 1425 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1426 ip6_tnl_dev_setup); 1427 1428 if (!ip6n->fb_tnl_dev) 1429 goto err_alloc_dev; 1430 dev_net_set(ip6n->fb_tnl_dev, net); 1431 1432 ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1433 1434 err = register_netdev(ip6n->fb_tnl_dev); 1435 if (err < 0) 1436 goto err_register; 1437 return 0; 1438 1439 err_register: 1440 free_netdev(ip6n->fb_tnl_dev); 1441 err_alloc_dev: 1442 /* nothing */ 1443 err_assign: 1444 kfree(ip6n); 1445 err_alloc: 1446 return err; 1447 } 1448 1449 static void ip6_tnl_exit_net(struct net *net) 1450 { 1451 struct ip6_tnl_net *ip6n; 1452 1453 ip6n = net_generic(net, ip6_tnl_net_id); 1454 rtnl_lock(); 1455 ip6_tnl_destroy_tunnels(ip6n); 1456 rtnl_unlock(); 1457 kfree(ip6n); 1458 } 1459 1460 static struct pernet_operations ip6_tnl_net_ops = { 1461 .init = ip6_tnl_init_net, 1462 .exit = ip6_tnl_exit_net, 1463 }; 1464 1465 /** 1466 * ip6_tunnel_init - register protocol and reserve needed resources 1467 * 1468 * Return: 0 on success 1469 **/ 1470 1471 static int __init ip6_tunnel_init(void) 1472 { 1473 int err; 1474 1475 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) { 1476 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n"); 1477 err = -EAGAIN; 1478 goto out; 1479 } 1480 1481 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) { 1482 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n"); 1483 err = -EAGAIN; 1484 goto unreg_ip4ip6; 1485 } 1486 1487 err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops); 1488 if (err < 0) 1489 goto err_pernet; 1490 return 0; 1491 err_pernet: 1492 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 1493 unreg_ip4ip6: 1494 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1495 out: 1496 return err; 1497 } 1498 1499 /** 1500 * ip6_tunnel_cleanup - free resources and unregister protocol 1501 **/ 1502 1503 static void __exit ip6_tunnel_cleanup(void) 1504 { 1505 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1506 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n"); 1507 1508 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1509 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n"); 1510 1511 unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops); 1512 } 1513 1514 module_init(ip6_tunnel_init); 1515 module_exit(ip6_tunnel_cleanup); 1516