1 /* 2 * IPv6 over IPv6 tunnel device 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Ville Nuorvala <vnuorval@tcs.hut.fi> 7 * 8 * $Id$ 9 * 10 * Based on: 11 * linux/net/ipv6/sit.c 12 * 13 * RFC 2473 14 * 15 * This program is free software; you can redistribute it and/or 16 * modify it under the terms of the GNU General Public License 17 * as published by the Free Software Foundation; either version 18 * 2 of the License, or (at your option) any later version. 19 * 20 */ 21 22 #include <linux/config.h> 23 #include <linux/module.h> 24 #include <linux/errno.h> 25 #include <linux/types.h> 26 #include <linux/sockios.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/ip.h> 45 #include <net/ipv6.h> 46 #include <net/protocol.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 54 MODULE_AUTHOR("Ville Nuorvala"); 55 MODULE_DESCRIPTION("IPv6-in-IPv6 tunnel"); 56 MODULE_LICENSE("GPL"); 57 58 #define IPV6_TLV_TEL_DST_SIZE 8 59 60 #ifdef IP6_TNL_DEBUG 61 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __FUNCTION__) 62 #else 63 #define IP6_TNL_TRACE(x...) do {;} while(0) 64 #endif 65 66 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 67 68 #define HASH_SIZE 32 69 70 #define HASH(addr) (((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \ 71 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \ 72 (HASH_SIZE - 1)) 73 74 static int ip6ip6_fb_tnl_dev_init(struct net_device *dev); 75 static int ip6ip6_tnl_dev_init(struct net_device *dev); 76 static void ip6ip6_tnl_dev_setup(struct net_device *dev); 77 78 /* the IPv6 tunnel fallback device */ 79 static struct net_device *ip6ip6_fb_tnl_dev; 80 81 82 /* lists for storing tunnels in use */ 83 static struct ip6_tnl *tnls_r_l[HASH_SIZE]; 84 static struct ip6_tnl *tnls_wc[1]; 85 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l }; 86 87 /* lock for the tunnel lists */ 88 static DEFINE_RWLOCK(ip6ip6_lock); 89 90 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 91 { 92 struct dst_entry *dst = t->dst_cache; 93 94 if (dst && dst->obsolete && 95 dst->ops->check(dst, t->dst_cookie) == NULL) { 96 t->dst_cache = NULL; 97 dst_release(dst); 98 return NULL; 99 } 100 101 return dst; 102 } 103 104 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t) 105 { 106 dst_release(t->dst_cache); 107 t->dst_cache = NULL; 108 } 109 110 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 111 { 112 struct rt6_info *rt = (struct rt6_info *) dst; 113 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 114 dst_release(t->dst_cache); 115 t->dst_cache = dst; 116 } 117 118 /** 119 * ip6ip6_tnl_lookup - fetch tunnel matching the end-point addresses 120 * @remote: the address of the tunnel exit-point 121 * @local: the address of the tunnel entry-point 122 * 123 * Return: 124 * tunnel matching given end-points if found, 125 * else fallback tunnel if its device is up, 126 * else %NULL 127 **/ 128 129 static struct ip6_tnl * 130 ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local) 131 { 132 unsigned h0 = HASH(remote); 133 unsigned h1 = HASH(local); 134 struct ip6_tnl *t; 135 136 for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) { 137 if (ipv6_addr_equal(local, &t->parms.laddr) && 138 ipv6_addr_equal(remote, &t->parms.raddr) && 139 (t->dev->flags & IFF_UP)) 140 return t; 141 } 142 if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP)) 143 return t; 144 145 return NULL; 146 } 147 148 /** 149 * ip6ip6_bucket - get head of list matching given tunnel parameters 150 * @p: parameters containing tunnel end-points 151 * 152 * Description: 153 * ip6ip6_bucket() returns the head of the list matching the 154 * &struct in6_addr entries laddr and raddr in @p. 155 * 156 * Return: head of IPv6 tunnel list 157 **/ 158 159 static struct ip6_tnl ** 160 ip6ip6_bucket(struct ip6_tnl_parm *p) 161 { 162 struct in6_addr *remote = &p->raddr; 163 struct in6_addr *local = &p->laddr; 164 unsigned h = 0; 165 int prio = 0; 166 167 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 168 prio = 1; 169 h = HASH(remote) ^ HASH(local); 170 } 171 return &tnls[prio][h]; 172 } 173 174 /** 175 * ip6ip6_tnl_link - add tunnel to hash table 176 * @t: tunnel to be added 177 **/ 178 179 static void 180 ip6ip6_tnl_link(struct ip6_tnl *t) 181 { 182 struct ip6_tnl **tp = ip6ip6_bucket(&t->parms); 183 184 t->next = *tp; 185 write_lock_bh(&ip6ip6_lock); 186 *tp = t; 187 write_unlock_bh(&ip6ip6_lock); 188 } 189 190 /** 191 * ip6ip6_tnl_unlink - remove tunnel from hash table 192 * @t: tunnel to be removed 193 **/ 194 195 static void 196 ip6ip6_tnl_unlink(struct ip6_tnl *t) 197 { 198 struct ip6_tnl **tp; 199 200 for (tp = ip6ip6_bucket(&t->parms); *tp; tp = &(*tp)->next) { 201 if (t == *tp) { 202 write_lock_bh(&ip6ip6_lock); 203 *tp = t->next; 204 write_unlock_bh(&ip6ip6_lock); 205 break; 206 } 207 } 208 } 209 210 /** 211 * ip6_tnl_create() - create a new tunnel 212 * @p: tunnel parameters 213 * @pt: pointer to new tunnel 214 * 215 * Description: 216 * Create tunnel matching given parameters. 217 * 218 * Return: 219 * 0 on success 220 **/ 221 222 static int 223 ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt) 224 { 225 struct net_device *dev; 226 struct ip6_tnl *t; 227 char name[IFNAMSIZ]; 228 int err; 229 230 if (p->name[0]) { 231 strlcpy(name, p->name, IFNAMSIZ); 232 } else { 233 int i; 234 for (i = 1; i < IP6_TNL_MAX; i++) { 235 sprintf(name, "ip6tnl%d", i); 236 if (__dev_get_by_name(name) == NULL) 237 break; 238 } 239 if (i == IP6_TNL_MAX) 240 return -ENOBUFS; 241 } 242 dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup); 243 if (dev == NULL) 244 return -ENOMEM; 245 246 t = dev->priv; 247 dev->init = ip6ip6_tnl_dev_init; 248 t->parms = *p; 249 250 if ((err = register_netdevice(dev)) < 0) { 251 free_netdev(dev); 252 return err; 253 } 254 dev_hold(dev); 255 256 ip6ip6_tnl_link(t); 257 *pt = t; 258 return 0; 259 } 260 261 /** 262 * ip6ip6_tnl_locate - find or create tunnel matching given parameters 263 * @p: tunnel parameters 264 * @create: != 0 if allowed to create new tunnel if no match found 265 * 266 * Description: 267 * ip6ip6_tnl_locate() first tries to locate an existing tunnel 268 * based on @parms. If this is unsuccessful, but @create is set a new 269 * tunnel device is created and registered for use. 270 * 271 * Return: 272 * 0 if tunnel located or created, 273 * -EINVAL if parameters incorrect, 274 * -ENODEV if no matching tunnel available 275 **/ 276 277 static int 278 ip6ip6_tnl_locate(struct ip6_tnl_parm *p, struct ip6_tnl **pt, int create) 279 { 280 struct in6_addr *remote = &p->raddr; 281 struct in6_addr *local = &p->laddr; 282 struct ip6_tnl *t; 283 284 if (p->proto != IPPROTO_IPV6) 285 return -EINVAL; 286 287 for (t = *ip6ip6_bucket(p); t; t = t->next) { 288 if (ipv6_addr_equal(local, &t->parms.laddr) && 289 ipv6_addr_equal(remote, &t->parms.raddr)) { 290 *pt = t; 291 return (create ? -EEXIST : 0); 292 } 293 } 294 if (!create) 295 return -ENODEV; 296 297 return ip6_tnl_create(p, pt); 298 } 299 300 /** 301 * ip6ip6_tnl_dev_uninit - tunnel device uninitializer 302 * @dev: the device to be destroyed 303 * 304 * Description: 305 * ip6ip6_tnl_dev_uninit() removes tunnel from its list 306 **/ 307 308 static void 309 ip6ip6_tnl_dev_uninit(struct net_device *dev) 310 { 311 struct ip6_tnl *t = dev->priv; 312 313 if (dev == ip6ip6_fb_tnl_dev) { 314 write_lock_bh(&ip6ip6_lock); 315 tnls_wc[0] = NULL; 316 write_unlock_bh(&ip6ip6_lock); 317 } else { 318 ip6ip6_tnl_unlink(t); 319 } 320 ip6_tnl_dst_reset(t); 321 dev_put(dev); 322 } 323 324 /** 325 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 326 * @skb: received socket buffer 327 * 328 * Return: 329 * 0 if none was found, 330 * else index to encapsulation limit 331 **/ 332 333 static __u16 334 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw) 335 { 336 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw; 337 __u8 nexthdr = ipv6h->nexthdr; 338 __u16 off = sizeof (*ipv6h); 339 340 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 341 __u16 optlen = 0; 342 struct ipv6_opt_hdr *hdr; 343 if (raw + off + sizeof (*hdr) > skb->data && 344 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 345 break; 346 347 hdr = (struct ipv6_opt_hdr *) (raw + off); 348 if (nexthdr == NEXTHDR_FRAGMENT) { 349 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 350 if (frag_hdr->frag_off) 351 break; 352 optlen = 8; 353 } else if (nexthdr == NEXTHDR_AUTH) { 354 optlen = (hdr->hdrlen + 2) << 2; 355 } else { 356 optlen = ipv6_optlen(hdr); 357 } 358 if (nexthdr == NEXTHDR_DEST) { 359 __u16 i = off + 2; 360 while (1) { 361 struct ipv6_tlv_tnl_enc_lim *tel; 362 363 /* No more room for encapsulation limit */ 364 if (i + sizeof (*tel) > off + optlen) 365 break; 366 367 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 368 /* return index of option if found and valid */ 369 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 370 tel->length == 1) 371 return i; 372 /* else jump to next option */ 373 if (tel->type) 374 i += tel->length + 2; 375 else 376 i++; 377 } 378 } 379 nexthdr = hdr->nexthdr; 380 off += optlen; 381 } 382 return 0; 383 } 384 385 /** 386 * ip6ip6_err - tunnel error handler 387 * 388 * Description: 389 * ip6ip6_err() should handle errors in the tunnel according 390 * to the specifications in RFC 2473. 391 **/ 392 393 static void 394 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 395 int type, int code, int offset, __u32 info) 396 { 397 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data; 398 struct ip6_tnl *t; 399 int rel_msg = 0; 400 int rel_type = ICMPV6_DEST_UNREACH; 401 int rel_code = ICMPV6_ADDR_UNREACH; 402 __u32 rel_info = 0; 403 __u16 len; 404 405 /* If the packet doesn't contain the original IPv6 header we are 406 in trouble since we might need the source address for further 407 processing of the error. */ 408 409 read_lock(&ip6ip6_lock); 410 if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL) 411 goto out; 412 413 switch (type) { 414 __u32 teli; 415 struct ipv6_tlv_tnl_enc_lim *tel; 416 __u32 mtu; 417 case ICMPV6_DEST_UNREACH: 418 if (net_ratelimit()) 419 printk(KERN_WARNING 420 "%s: Path to destination invalid " 421 "or inactive!\n", t->parms.name); 422 rel_msg = 1; 423 break; 424 case ICMPV6_TIME_EXCEED: 425 if (code == ICMPV6_EXC_HOPLIMIT) { 426 if (net_ratelimit()) 427 printk(KERN_WARNING 428 "%s: Too small hop limit or " 429 "routing loop in tunnel!\n", 430 t->parms.name); 431 rel_msg = 1; 432 } 433 break; 434 case ICMPV6_PARAMPROB: 435 /* ignore if parameter problem not caused by a tunnel 436 encapsulation limit sub-option */ 437 if (code != ICMPV6_HDR_FIELD) { 438 break; 439 } 440 teli = parse_tlv_tnl_enc_lim(skb, skb->data); 441 442 if (teli && teli == ntohl(info) - 2) { 443 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 444 if (tel->encap_limit == 0) { 445 if (net_ratelimit()) 446 printk(KERN_WARNING 447 "%s: Too small encapsulation " 448 "limit or routing loop in " 449 "tunnel!\n", t->parms.name); 450 rel_msg = 1; 451 } 452 } 453 break; 454 case ICMPV6_PKT_TOOBIG: 455 mtu = ntohl(info) - offset; 456 if (mtu < IPV6_MIN_MTU) 457 mtu = IPV6_MIN_MTU; 458 t->dev->mtu = mtu; 459 460 if ((len = sizeof (*ipv6h) + ipv6h->payload_len) > mtu) { 461 rel_type = ICMPV6_PKT_TOOBIG; 462 rel_code = 0; 463 rel_info = mtu; 464 rel_msg = 1; 465 } 466 break; 467 } 468 if (rel_msg && pskb_may_pull(skb, offset + sizeof (*ipv6h))) { 469 struct rt6_info *rt; 470 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 471 if (!skb2) 472 goto out; 473 474 dst_release(skb2->dst); 475 skb2->dst = NULL; 476 skb_pull(skb2, offset); 477 skb2->nh.raw = skb2->data; 478 479 /* Try to guess incoming interface */ 480 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0); 481 482 if (rt && rt->rt6i_dev) 483 skb2->dev = rt->rt6i_dev; 484 485 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev); 486 487 if (rt) 488 dst_release(&rt->u.dst); 489 490 kfree_skb(skb2); 491 } 492 out: 493 read_unlock(&ip6ip6_lock); 494 } 495 496 static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph, 497 struct sk_buff *skb) 498 { 499 struct ipv6hdr *inner_iph = skb->nh.ipv6h; 500 501 if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph))) 502 IP6_ECN_set_ce(inner_iph); 503 } 504 505 /** 506 * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally 507 * @skb: received socket buffer 508 * 509 * Return: 0 510 **/ 511 512 static int 513 ip6ip6_rcv(struct sk_buff **pskb, unsigned int *nhoffp) 514 { 515 struct sk_buff *skb = *pskb; 516 struct ipv6hdr *ipv6h; 517 struct ip6_tnl *t; 518 519 if (!pskb_may_pull(skb, sizeof (*ipv6h))) 520 goto discard; 521 522 ipv6h = skb->nh.ipv6h; 523 524 read_lock(&ip6ip6_lock); 525 526 if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) { 527 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 528 kfree_skb(skb); 529 return 0; 530 } 531 532 if (!(t->parms.flags & IP6_TNL_F_CAP_RCV)) { 533 t->stat.rx_dropped++; 534 read_unlock(&ip6ip6_lock); 535 goto discard; 536 } 537 secpath_reset(skb); 538 skb->mac.raw = skb->nh.raw; 539 skb->nh.raw = skb->data; 540 skb->protocol = htons(ETH_P_IPV6); 541 skb->pkt_type = PACKET_HOST; 542 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 543 skb->dev = t->dev; 544 dst_release(skb->dst); 545 skb->dst = NULL; 546 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 547 ipv6_copy_dscp(ipv6h, skb->nh.ipv6h); 548 ip6ip6_ecn_decapsulate(ipv6h, skb); 549 t->stat.rx_packets++; 550 t->stat.rx_bytes += skb->len; 551 netif_rx(skb); 552 read_unlock(&ip6ip6_lock); 553 return 0; 554 } 555 read_unlock(&ip6ip6_lock); 556 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev); 557 discard: 558 return 1; 559 } 560 561 static inline struct ipv6_txoptions *create_tel(__u8 encap_limit) 562 { 563 struct ipv6_tlv_tnl_enc_lim *tel; 564 struct ipv6_txoptions *opt; 565 __u8 *raw; 566 567 int opt_len = sizeof(*opt) + 8; 568 569 if (!(opt = kmalloc(opt_len, GFP_ATOMIC))) { 570 return NULL; 571 } 572 memset(opt, 0, opt_len); 573 opt->tot_len = opt_len; 574 opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1); 575 opt->opt_nflen = 8; 576 577 tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1); 578 tel->type = IPV6_TLV_TNL_ENCAP_LIMIT; 579 tel->length = 1; 580 tel->encap_limit = encap_limit; 581 582 raw = (__u8 *) opt->dst0opt; 583 raw[5] = IPV6_TLV_PADN; 584 raw[6] = 1; 585 586 return opt; 587 } 588 589 /** 590 * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 591 * @t: the outgoing tunnel device 592 * @hdr: IPv6 header from the incoming packet 593 * 594 * Description: 595 * Avoid trivial tunneling loop by checking that tunnel exit-point 596 * doesn't match source of incoming packet. 597 * 598 * Return: 599 * 1 if conflict, 600 * 0 else 601 **/ 602 603 static inline int 604 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr) 605 { 606 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 607 } 608 609 /** 610 * ip6ip6_tnl_xmit - encapsulate packet and send 611 * @skb: the outgoing socket buffer 612 * @dev: the outgoing tunnel device 613 * 614 * Description: 615 * Build new header and do some sanity checks on the packet before sending 616 * it. 617 * 618 * Return: 619 * 0 620 **/ 621 622 static int 623 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 624 { 625 struct ip6_tnl *t = (struct ip6_tnl *) dev->priv; 626 struct net_device_stats *stats = &t->stat; 627 struct ipv6hdr *ipv6h = skb->nh.ipv6h; 628 struct ipv6_txoptions *opt = NULL; 629 int encap_limit = -1; 630 __u16 offset; 631 struct flowi fl; 632 struct dst_entry *dst; 633 struct net_device *tdev; 634 int mtu; 635 int max_headroom = sizeof(struct ipv6hdr); 636 u8 proto; 637 int err; 638 int pkt_len; 639 int dsfield; 640 641 if (t->recursion++) { 642 stats->collisions++; 643 goto tx_err; 644 } 645 if (skb->protocol != htons(ETH_P_IPV6) || 646 !(t->parms.flags & IP6_TNL_F_CAP_XMIT) || 647 ip6ip6_tnl_addr_conflict(t, ipv6h)) { 648 goto tx_err; 649 } 650 if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) { 651 struct ipv6_tlv_tnl_enc_lim *tel; 652 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset]; 653 if (tel->encap_limit == 0) { 654 icmpv6_send(skb, ICMPV6_PARAMPROB, 655 ICMPV6_HDR_FIELD, offset + 2, skb->dev); 656 goto tx_err; 657 } 658 encap_limit = tel->encap_limit - 1; 659 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 660 encap_limit = t->parms.encap_limit; 661 } 662 memcpy(&fl, &t->fl, sizeof (fl)); 663 proto = fl.proto; 664 665 dsfield = ipv6_get_dsfield(ipv6h); 666 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 667 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_TCLASS_MASK); 668 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)) 669 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_FLOWLABEL_MASK); 670 671 if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL) 672 goto tx_err; 673 674 if ((dst = ip6_tnl_dst_check(t)) != NULL) 675 dst_hold(dst); 676 else { 677 dst = ip6_route_output(NULL, &fl); 678 679 if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0) 680 goto tx_err_link_failure; 681 } 682 683 tdev = dst->dev; 684 685 if (tdev == dev) { 686 stats->collisions++; 687 if (net_ratelimit()) 688 printk(KERN_WARNING 689 "%s: Local routing loop detected!\n", 690 t->parms.name); 691 goto tx_err_dst_release; 692 } 693 mtu = dst_mtu(dst) - sizeof (*ipv6h); 694 if (opt) { 695 max_headroom += 8; 696 mtu -= 8; 697 } 698 if (mtu < IPV6_MIN_MTU) 699 mtu = IPV6_MIN_MTU; 700 if (skb->dst && mtu < dst_mtu(skb->dst)) { 701 struct rt6_info *rt = (struct rt6_info *) skb->dst; 702 rt->rt6i_flags |= RTF_MODIFIED; 703 rt->u.dst.metrics[RTAX_MTU-1] = mtu; 704 } 705 if (skb->len > mtu) { 706 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev); 707 goto tx_err_dst_release; 708 } 709 710 /* 711 * Okay, now see if we can stuff it in the buffer as-is. 712 */ 713 max_headroom += LL_RESERVED_SPACE(tdev); 714 715 if (skb_headroom(skb) < max_headroom || 716 skb_cloned(skb) || skb_shared(skb)) { 717 struct sk_buff *new_skb; 718 719 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 720 goto tx_err_dst_release; 721 722 if (skb->sk) 723 skb_set_owner_w(new_skb, skb->sk); 724 kfree_skb(skb); 725 skb = new_skb; 726 } 727 dst_release(skb->dst); 728 skb->dst = dst_clone(dst); 729 730 skb->h.raw = skb->nh.raw; 731 732 if (opt) 733 ipv6_push_nfrag_opts(skb, opt, &proto, NULL); 734 735 skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr)); 736 ipv6h = skb->nh.ipv6h; 737 *(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000); 738 dsfield = INET_ECN_encapsulate(0, dsfield); 739 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield); 740 ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr)); 741 ipv6h->hop_limit = t->parms.hop_limit; 742 ipv6h->nexthdr = proto; 743 ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src); 744 ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst); 745 nf_reset(skb); 746 pkt_len = skb->len; 747 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, 748 skb->dst->dev, dst_output); 749 750 if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) { 751 stats->tx_bytes += pkt_len; 752 stats->tx_packets++; 753 } else { 754 stats->tx_errors++; 755 stats->tx_aborted_errors++; 756 } 757 ip6_tnl_dst_store(t, dst); 758 759 if (opt) 760 kfree(opt); 761 762 t->recursion--; 763 return 0; 764 tx_err_link_failure: 765 stats->tx_carrier_errors++; 766 dst_link_failure(skb); 767 tx_err_dst_release: 768 dst_release(dst); 769 if (opt) 770 kfree(opt); 771 tx_err: 772 stats->tx_errors++; 773 stats->tx_dropped++; 774 kfree_skb(skb); 775 t->recursion--; 776 return 0; 777 } 778 779 static void ip6_tnl_set_cap(struct ip6_tnl *t) 780 { 781 struct ip6_tnl_parm *p = &t->parms; 782 struct in6_addr *laddr = &p->laddr; 783 struct in6_addr *raddr = &p->raddr; 784 int ltype = ipv6_addr_type(laddr); 785 int rtype = ipv6_addr_type(raddr); 786 787 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV); 788 789 if (ltype != IPV6_ADDR_ANY && rtype != IPV6_ADDR_ANY && 790 ((ltype|rtype) & 791 (IPV6_ADDR_UNICAST| 792 IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL| 793 IPV6_ADDR_MAPPED|IPV6_ADDR_RESERVED)) == IPV6_ADDR_UNICAST) { 794 struct net_device *ldev = NULL; 795 int l_ok = 1; 796 int r_ok = 1; 797 798 if (p->link) 799 ldev = dev_get_by_index(p->link); 800 801 if (ltype&IPV6_ADDR_UNICAST && !ipv6_chk_addr(laddr, ldev, 0)) 802 l_ok = 0; 803 804 if (rtype&IPV6_ADDR_UNICAST && ipv6_chk_addr(raddr, NULL, 0)) 805 r_ok = 0; 806 807 if (l_ok && r_ok) { 808 if (ltype&IPV6_ADDR_UNICAST) 809 p->flags |= IP6_TNL_F_CAP_XMIT; 810 if (rtype&IPV6_ADDR_UNICAST) 811 p->flags |= IP6_TNL_F_CAP_RCV; 812 } 813 if (ldev) 814 dev_put(ldev); 815 } 816 } 817 818 static void ip6ip6_tnl_link_config(struct ip6_tnl *t) 819 { 820 struct net_device *dev = t->dev; 821 struct ip6_tnl_parm *p = &t->parms; 822 struct flowi *fl = &t->fl; 823 824 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 825 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 826 827 /* Set up flowi template */ 828 ipv6_addr_copy(&fl->fl6_src, &p->laddr); 829 ipv6_addr_copy(&fl->fl6_dst, &p->raddr); 830 fl->oif = p->link; 831 fl->fl6_flowlabel = 0; 832 833 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 834 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 835 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 836 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 837 838 ip6_tnl_set_cap(t); 839 840 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 841 dev->flags |= IFF_POINTOPOINT; 842 else 843 dev->flags &= ~IFF_POINTOPOINT; 844 845 dev->iflink = p->link; 846 847 if (p->flags & IP6_TNL_F_CAP_XMIT) { 848 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr, 849 p->link, 0); 850 851 if (rt == NULL) 852 return; 853 854 if (rt->rt6i_dev) { 855 dev->hard_header_len = rt->rt6i_dev->hard_header_len + 856 sizeof (struct ipv6hdr); 857 858 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr); 859 860 if (dev->mtu < IPV6_MIN_MTU) 861 dev->mtu = IPV6_MIN_MTU; 862 } 863 dst_release(&rt->u.dst); 864 } 865 } 866 867 /** 868 * ip6ip6_tnl_change - update the tunnel parameters 869 * @t: tunnel to be changed 870 * @p: tunnel configuration parameters 871 * @active: != 0 if tunnel is ready for use 872 * 873 * Description: 874 * ip6ip6_tnl_change() updates the tunnel parameters 875 **/ 876 877 static int 878 ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p) 879 { 880 ipv6_addr_copy(&t->parms.laddr, &p->laddr); 881 ipv6_addr_copy(&t->parms.raddr, &p->raddr); 882 t->parms.flags = p->flags; 883 t->parms.hop_limit = p->hop_limit; 884 t->parms.encap_limit = p->encap_limit; 885 t->parms.flowinfo = p->flowinfo; 886 t->parms.link = p->link; 887 ip6ip6_tnl_link_config(t); 888 return 0; 889 } 890 891 /** 892 * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace 893 * @dev: virtual device associated with tunnel 894 * @ifr: parameters passed from userspace 895 * @cmd: command to be performed 896 * 897 * Description: 898 * ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels 899 * from userspace. 900 * 901 * The possible commands are the following: 902 * %SIOCGETTUNNEL: get tunnel parameters for device 903 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 904 * %SIOCCHGTUNNEL: change tunnel parameters to those given 905 * %SIOCDELTUNNEL: delete tunnel 906 * 907 * The fallback device "ip6tnl0", created during module 908 * initialization, can be used for creating other tunnel devices. 909 * 910 * Return: 911 * 0 on success, 912 * %-EFAULT if unable to copy data to or from userspace, 913 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 914 * %-EINVAL if passed tunnel parameters are invalid, 915 * %-EEXIST if changing a tunnel's parameters would cause a conflict 916 * %-ENODEV if attempting to change or delete a nonexisting device 917 **/ 918 919 static int 920 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 921 { 922 int err = 0; 923 int create; 924 struct ip6_tnl_parm p; 925 struct ip6_tnl *t = NULL; 926 927 switch (cmd) { 928 case SIOCGETTUNNEL: 929 if (dev == ip6ip6_fb_tnl_dev) { 930 if (copy_from_user(&p, 931 ifr->ifr_ifru.ifru_data, 932 sizeof (p))) { 933 err = -EFAULT; 934 break; 935 } 936 if ((err = ip6ip6_tnl_locate(&p, &t, 0)) == -ENODEV) 937 t = (struct ip6_tnl *) dev->priv; 938 else if (err) 939 break; 940 } else 941 t = (struct ip6_tnl *) dev->priv; 942 943 memcpy(&p, &t->parms, sizeof (p)); 944 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 945 err = -EFAULT; 946 } 947 break; 948 case SIOCADDTUNNEL: 949 case SIOCCHGTUNNEL: 950 err = -EPERM; 951 create = (cmd == SIOCADDTUNNEL); 952 if (!capable(CAP_NET_ADMIN)) 953 break; 954 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 955 err = -EFAULT; 956 break; 957 } 958 if (!create && dev != ip6ip6_fb_tnl_dev) { 959 t = (struct ip6_tnl *) dev->priv; 960 } 961 if (!t && (err = ip6ip6_tnl_locate(&p, &t, create))) { 962 break; 963 } 964 if (cmd == SIOCCHGTUNNEL) { 965 if (t->dev != dev) { 966 err = -EEXIST; 967 break; 968 } 969 ip6ip6_tnl_unlink(t); 970 err = ip6ip6_tnl_change(t, &p); 971 ip6ip6_tnl_link(t); 972 netdev_state_change(dev); 973 } 974 if (copy_to_user(ifr->ifr_ifru.ifru_data, 975 &t->parms, sizeof (p))) { 976 err = -EFAULT; 977 } else { 978 err = 0; 979 } 980 break; 981 case SIOCDELTUNNEL: 982 err = -EPERM; 983 if (!capable(CAP_NET_ADMIN)) 984 break; 985 986 if (dev == ip6ip6_fb_tnl_dev) { 987 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, 988 sizeof (p))) { 989 err = -EFAULT; 990 break; 991 } 992 err = ip6ip6_tnl_locate(&p, &t, 0); 993 if (err) 994 break; 995 if (t == ip6ip6_fb_tnl_dev->priv) { 996 err = -EPERM; 997 break; 998 } 999 } else { 1000 t = (struct ip6_tnl *) dev->priv; 1001 } 1002 err = unregister_netdevice(t->dev); 1003 break; 1004 default: 1005 err = -EINVAL; 1006 } 1007 return err; 1008 } 1009 1010 /** 1011 * ip6ip6_tnl_get_stats - return the stats for tunnel device 1012 * @dev: virtual device associated with tunnel 1013 * 1014 * Return: stats for device 1015 **/ 1016 1017 static struct net_device_stats * 1018 ip6ip6_tnl_get_stats(struct net_device *dev) 1019 { 1020 return &(((struct ip6_tnl *) dev->priv)->stat); 1021 } 1022 1023 /** 1024 * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device 1025 * @dev: virtual device associated with tunnel 1026 * @new_mtu: the new mtu 1027 * 1028 * Return: 1029 * 0 on success, 1030 * %-EINVAL if mtu too small 1031 **/ 1032 1033 static int 1034 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1035 { 1036 if (new_mtu < IPV6_MIN_MTU) { 1037 return -EINVAL; 1038 } 1039 dev->mtu = new_mtu; 1040 return 0; 1041 } 1042 1043 /** 1044 * ip6ip6_tnl_dev_setup - setup virtual tunnel device 1045 * @dev: virtual device associated with tunnel 1046 * 1047 * Description: 1048 * Initialize function pointers and device parameters 1049 **/ 1050 1051 static void ip6ip6_tnl_dev_setup(struct net_device *dev) 1052 { 1053 SET_MODULE_OWNER(dev); 1054 dev->uninit = ip6ip6_tnl_dev_uninit; 1055 dev->destructor = free_netdev; 1056 dev->hard_start_xmit = ip6ip6_tnl_xmit; 1057 dev->get_stats = ip6ip6_tnl_get_stats; 1058 dev->do_ioctl = ip6ip6_tnl_ioctl; 1059 dev->change_mtu = ip6ip6_tnl_change_mtu; 1060 1061 dev->type = ARPHRD_TUNNEL6; 1062 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1063 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1064 dev->flags |= IFF_NOARP; 1065 dev->addr_len = sizeof(struct in6_addr); 1066 } 1067 1068 1069 /** 1070 * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1071 * @dev: virtual device associated with tunnel 1072 **/ 1073 1074 static inline void 1075 ip6ip6_tnl_dev_init_gen(struct net_device *dev) 1076 { 1077 struct ip6_tnl *t = (struct ip6_tnl *) dev->priv; 1078 t->fl.proto = IPPROTO_IPV6; 1079 t->dev = dev; 1080 strcpy(t->parms.name, dev->name); 1081 } 1082 1083 /** 1084 * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1085 * @dev: virtual device associated with tunnel 1086 **/ 1087 1088 static int 1089 ip6ip6_tnl_dev_init(struct net_device *dev) 1090 { 1091 struct ip6_tnl *t = (struct ip6_tnl *) dev->priv; 1092 ip6ip6_tnl_dev_init_gen(dev); 1093 ip6ip6_tnl_link_config(t); 1094 return 0; 1095 } 1096 1097 /** 1098 * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1099 * @dev: fallback device 1100 * 1101 * Return: 0 1102 **/ 1103 1104 static int 1105 ip6ip6_fb_tnl_dev_init(struct net_device *dev) 1106 { 1107 struct ip6_tnl *t = dev->priv; 1108 ip6ip6_tnl_dev_init_gen(dev); 1109 dev_hold(dev); 1110 tnls_wc[0] = t; 1111 return 0; 1112 } 1113 1114 #ifdef CONFIG_INET6_TUNNEL 1115 static struct xfrm6_tunnel ip6ip6_handler = { 1116 .handler = ip6ip6_rcv, 1117 .err_handler = ip6ip6_err, 1118 }; 1119 1120 static inline int ip6ip6_register(void) 1121 { 1122 return xfrm6_tunnel_register(&ip6ip6_handler); 1123 } 1124 1125 static inline int ip6ip6_unregister(void) 1126 { 1127 return xfrm6_tunnel_deregister(&ip6ip6_handler); 1128 } 1129 #else 1130 static struct inet6_protocol xfrm6_tunnel_protocol = { 1131 .handler = ip6ip6_rcv, 1132 .err_handler = ip6ip6_err, 1133 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1134 }; 1135 1136 static inline int ip6ip6_register(void) 1137 { 1138 return inet6_add_protocol(&xfrm6_tunnel_protocol, IPPROTO_IPV6); 1139 } 1140 1141 static inline int ip6ip6_unregister(void) 1142 { 1143 return inet6_del_protocol(&xfrm6_tunnel_protocol, IPPROTO_IPV6); 1144 } 1145 #endif 1146 1147 /** 1148 * ip6_tunnel_init - register protocol and reserve needed resources 1149 * 1150 * Return: 0 on success 1151 **/ 1152 1153 static int __init ip6_tunnel_init(void) 1154 { 1155 int err; 1156 1157 if (ip6ip6_register() < 0) { 1158 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n"); 1159 return -EAGAIN; 1160 } 1161 ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1162 ip6ip6_tnl_dev_setup); 1163 1164 if (!ip6ip6_fb_tnl_dev) { 1165 err = -ENOMEM; 1166 goto fail; 1167 } 1168 ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init; 1169 1170 if ((err = register_netdev(ip6ip6_fb_tnl_dev))) { 1171 free_netdev(ip6ip6_fb_tnl_dev); 1172 goto fail; 1173 } 1174 return 0; 1175 fail: 1176 ip6ip6_unregister(); 1177 return err; 1178 } 1179 1180 /** 1181 * ip6_tunnel_cleanup - free resources and unregister protocol 1182 **/ 1183 1184 static void __exit ip6_tunnel_cleanup(void) 1185 { 1186 if (ip6ip6_unregister() < 0) 1187 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n"); 1188 1189 unregister_netdev(ip6ip6_fb_tnl_dev); 1190 } 1191 1192 module_init(ip6_tunnel_init); 1193 module_exit(ip6_tunnel_cleanup); 1194