net/ipv6/ip6_tunnel.c

   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/capability.h>
  25 #include <linux/errno.h>
  26 #include <linux/types.h>
  27 #include <linux/sockios.h>
  28 #include <linux/if.h>
  29 #include <linux/in.h>
  30 #include <linux/ip.h>
  31 #include <linux/if_tunnel.h>
  32 #include <linux/net.h>
  33 #include <linux/in6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/icmpv6.h>
  37 #include <linux/init.h>
  38 #include <linux/route.h>
  39 #include <linux/rtnetlink.h>
  40 #include <linux/netfilter_ipv6.h>
  41
  42 #include <asm/uaccess.h>
  43 #include <asm/atomic.h>
  44
  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
  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 = netdev_priv(dev);
 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 = netdev_priv(dev);
 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 int
 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         int err = -ENOENT;
 405
 406         /* If the packet doesn't contain the original IPv6 header we are
 407            in trouble since we might need the source address for further
 408            processing of the error. */
 409
 410         read_lock(&ip6ip6_lock);
 411         if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
 412                 goto out;
 413
 414         err = 0;
 415
 416         switch (type) {
 417                 __u32 teli;
 418                 struct ipv6_tlv_tnl_enc_lim *tel;
 419                 __u32 mtu;
 420         case ICMPV6_DEST_UNREACH:
 421                 if (net_ratelimit())
 422                         printk(KERN_WARNING
 423                                "%s: Path to destination invalid "
 424                                "or inactive!\n", t->parms.name);
 425                 rel_msg = 1;
 426                 break;
 427         case ICMPV6_TIME_EXCEED:
 428                 if (code == ICMPV6_EXC_HOPLIMIT) {
 429                         if (net_ratelimit())
 430                                 printk(KERN_WARNING
 431                                        "%s: Too small hop limit or "
 432                                        "routing loop in tunnel!\n",
 433                                        t->parms.name);
 434                         rel_msg = 1;
 435                 }
 436                 break;
 437         case ICMPV6_PARAMPROB:
 438                 /* ignore if parameter problem not caused by a tunnel
 439                    encapsulation limit sub-option */
 440                 if (code != ICMPV6_HDR_FIELD) {
 441                         break;
 442                 }
 443                 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
 444
 445                 if (teli && teli == ntohl(info) - 2) {
 446                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 447                         if (tel->encap_limit == 0) {
 448                                 if (net_ratelimit())
 449                                         printk(KERN_WARNING
 450                                                "%s: Too small encapsulation "
 451                                                "limit or routing loop in "
 452                                                "tunnel!\n", t->parms.name);
 453                                 rel_msg = 1;
 454                         }
 455                 }
 456                 break;
 457         case ICMPV6_PKT_TOOBIG:
 458                 mtu = ntohl(info) - offset;
 459                 if (mtu < IPV6_MIN_MTU)
 460                         mtu = IPV6_MIN_MTU;
 461                 t->dev->mtu = mtu;
 462
 463                 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
 464                         rel_type = ICMPV6_PKT_TOOBIG;
 465                         rel_code = 0;
 466                         rel_info = mtu;
 467                         rel_msg = 1;
 468                 }
 469                 break;
 470         }
 471         if (rel_msg &&  pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
 472                 struct rt6_info *rt;
 473                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 474                 if (!skb2)
 475                         goto out;
 476
 477                 dst_release(skb2->dst);
 478                 skb2->dst = NULL;
 479                 skb_pull(skb2, offset);
 480                 skb2->nh.raw = skb2->data;
 481
 482                 /* Try to guess incoming interface */
 483                 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
 484
 485                 if (rt && rt->rt6i_dev)
 486                         skb2->dev = rt->rt6i_dev;
 487
 488                 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 489
 490                 if (rt)
 491                         dst_release(&rt->u.dst);
 492
 493                 kfree_skb(skb2);
 494         }
 495 out:
 496         read_unlock(&ip6ip6_lock);
 497         return err;
 498 }
 499
 500 static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
 501                                           struct sk_buff *skb)
 502 {
 503         struct ipv6hdr *inner_iph = skb->nh.ipv6h;
 504
 505         if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
 506                 IP6_ECN_set_ce(inner_iph);
 507 }
 508
 509 /**
 510  * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
 511  *   @skb: received socket buffer
 512  *
 513  * Return: 0
 514  **/
 515
 516 static int
 517 ip6ip6_rcv(struct sk_buff *skb)
 518 {
 519         struct ipv6hdr *ipv6h;
 520         struct ip6_tnl *t;
 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                         read_unlock(&ip6ip6_lock);
 529                         goto discard;
 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         return 1;
 557
 558 discard:
 559         kfree_skb(skb);
 560         return 0;
 561 }
 562
 563 static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
 564 {
 565         struct ipv6_tlv_tnl_enc_lim *tel;
 566         struct ipv6_txoptions *opt;
 567         __u8 *raw;
 568
 569         int opt_len = sizeof(*opt) + 8;
 570
 571         if (!(opt = kmalloc(opt_len, GFP_ATOMIC))) {
 572                 return NULL;
 573         }
 574         memset(opt, 0, opt_len);
 575         opt->tot_len = opt_len;
 576         opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
 577         opt->opt_nflen = 8;
 578
 579         tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1);
 580         tel->type = IPV6_TLV_TNL_ENCAP_LIMIT;
 581         tel->length = 1;
 582         tel->encap_limit = encap_limit;
 583
 584         raw = (__u8 *) opt->dst0opt;
 585         raw[5] = IPV6_TLV_PADN;
 586         raw[6] = 1;
 587
 588         return opt;
 589 }
 590
 591 /**
 592  * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 593  *   @t: the outgoing tunnel device
 594  *   @hdr: IPv6 header from the incoming packet
 595  *
 596  * Description:
 597  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 598  *   doesn't match source of incoming packet.
 599  *
 600  * Return:
 601  *   1 if conflict,
 602  *   0 else
 603  **/
 604
 605 static inline int
 606 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
 607 {
 608         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 609 }
 610
 611 /**
 612  * ip6ip6_tnl_xmit - encapsulate packet and send
 613  *   @skb: the outgoing socket buffer
 614  *   @dev: the outgoing tunnel device
 615  *
 616  * Description:
 617  *   Build new header and do some sanity checks on the packet before sending
 618  *   it.
 619  *
 620  * Return:
 621  *   0
 622  **/
 623
 624 static int
 625 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 626 {
 627         struct ip6_tnl *t = netdev_priv(dev);
 628         struct net_device_stats *stats = &t->stat;
 629         struct ipv6hdr *ipv6h = skb->nh.ipv6h;
 630         struct ipv6_txoptions *opt = NULL;
 631         int encap_limit = -1;
 632         __u16 offset;
 633         struct flowi fl;
 634         struct dst_entry *dst;
 635         struct net_device *tdev;
 636         int mtu;
 637         int max_headroom = sizeof(struct ipv6hdr);
 638         u8 proto;
 639         int err;
 640         int pkt_len;
 641         int dsfield;
 642
 643         if (t->recursion++) {
 644                 stats->collisions++;
 645                 goto tx_err;
 646         }
 647         if (skb->protocol != htons(ETH_P_IPV6) ||
 648             !(t->parms.flags & IP6_TNL_F_CAP_XMIT) ||
 649             ip6ip6_tnl_addr_conflict(t, ipv6h)) {
 650                 goto tx_err;
 651         }
 652         if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
 653                 struct ipv6_tlv_tnl_enc_lim *tel;
 654                 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
 655                 if (tel->encap_limit == 0) {
 656                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 657                                     ICMPV6_HDR_FIELD, offset + 2, skb->dev);
 658                         goto tx_err;
 659                 }
 660                 encap_limit = tel->encap_limit - 1;
 661         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 662                 encap_limit = t->parms.encap_limit;
 663         }
 664         memcpy(&fl, &t->fl, sizeof (fl));
 665         proto = fl.proto;
 666
 667         dsfield = ipv6_get_dsfield(ipv6h);
 668         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
 669                 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_TCLASS_MASK);
 670         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
 671                 fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 672
 673         if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
 674                 goto tx_err;
 675
 676         if ((dst = ip6_tnl_dst_check(t)) != NULL)
 677                 dst_hold(dst);
 678         else {
 679                 dst = ip6_route_output(NULL, &fl);
 680
 681                 if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
 682                         goto tx_err_link_failure;
 683         }
 684
 685         tdev = dst->dev;
 686
 687         if (tdev == dev) {
 688                 stats->collisions++;
 689                 if (net_ratelimit())
 690                         printk(KERN_WARNING
 691                                "%s: Local routing loop detected!\n",
 692                                t->parms.name);
 693                 goto tx_err_dst_release;
 694         }
 695         mtu = dst_mtu(dst) - sizeof (*ipv6h);
 696         if (opt) {
 697                 max_headroom += 8;
 698                 mtu -= 8;
 699         }
 700         if (mtu < IPV6_MIN_MTU)
 701                 mtu = IPV6_MIN_MTU;
 702         if (skb->dst && mtu < dst_mtu(skb->dst)) {
 703                 struct rt6_info *rt = (struct rt6_info *) skb->dst;
 704                 rt->rt6i_flags |= RTF_MODIFIED;
 705                 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
 706         }
 707         if (skb->len > mtu) {
 708                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 709                 goto tx_err_dst_release;
 710         }
 711
 712         /*
 713          * Okay, now see if we can stuff it in the buffer as-is.
 714          */
 715         max_headroom += LL_RESERVED_SPACE(tdev);
 716
 717         if (skb_headroom(skb) < max_headroom ||
 718             skb_cloned(skb) || skb_shared(skb)) {
 719                 struct sk_buff *new_skb;
 720
 721                 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
 722                         goto tx_err_dst_release;
 723
 724                 if (skb->sk)
 725                         skb_set_owner_w(new_skb, skb->sk);
 726                 kfree_skb(skb);
 727                 skb = new_skb;
 728         }
 729         dst_release(skb->dst);
 730         skb->dst = dst_clone(dst);
 731
 732         skb->h.raw = skb->nh.raw;
 733
 734         if (opt)
 735                 ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
 736
 737         skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
 738         ipv6h = skb->nh.ipv6h;
 739         *(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
 740         dsfield = INET_ECN_encapsulate(0, dsfield);
 741         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 742         ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
 743         ipv6h->hop_limit = t->parms.hop_limit;
 744         ipv6h->nexthdr = proto;
 745         ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
 746         ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
 747         nf_reset(skb);
 748         pkt_len = skb->len;
 749         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
 750                       skb->dst->dev, dst_output);
 751
 752         if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) {
 753                 stats->tx_bytes += pkt_len;
 754                 stats->tx_packets++;
 755         } else {
 756                 stats->tx_errors++;
 757                 stats->tx_aborted_errors++;
 758         }
 759         ip6_tnl_dst_store(t, dst);
 760
 761         kfree(opt);
 762
 763         t->recursion--;
 764         return 0;
 765 tx_err_link_failure:
 766         stats->tx_carrier_errors++;
 767         dst_link_failure(skb);
 768 tx_err_dst_release:
 769         dst_release(dst);
 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         ip6_tnl_dst_reset(t);
 888         ip6ip6_tnl_link_config(t);
 889         return 0;
 890 }
 891
 892 /**
 893  * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
 894  *   @dev: virtual device associated with tunnel
 895  *   @ifr: parameters passed from userspace
 896  *   @cmd: command to be performed
 897  *
 898  * Description:
 899  *   ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
 900  *   from userspace.
 901  *
 902  *   The possible commands are the following:
 903  *     %SIOCGETTUNNEL: get tunnel parameters for device
 904  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
 905  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
 906  *     %SIOCDELTUNNEL: delete tunnel
 907  *
 908  *   The fallback device "ip6tnl0", created during module
 909  *   initialization, can be used for creating other tunnel devices.
 910  *
 911  * Return:
 912  *   0 on success,
 913  *   %-EFAULT if unable to copy data to or from userspace,
 914  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
 915  *   %-EINVAL if passed tunnel parameters are invalid,
 916  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
 917  *   %-ENODEV if attempting to change or delete a nonexisting device
 918  **/
 919
 920 static int
 921 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 922 {
 923         int err = 0;
 924         int create;
 925         struct ip6_tnl_parm p;
 926         struct ip6_tnl *t = NULL;
 927
 928         switch (cmd) {
 929         case SIOCGETTUNNEL:
 930                 if (dev == ip6ip6_fb_tnl_dev) {
 931                         if (copy_from_user(&p,
 932                                            ifr->ifr_ifru.ifru_data,
 933                                            sizeof (p))) {
 934                                 err = -EFAULT;
 935                                 break;
 936                         }
 937                         if ((err = ip6ip6_tnl_locate(&p, &t, 0)) == -ENODEV)
 938                                 t = netdev_priv(dev);
 939                         else if (err)
 940                                 break;
 941                 } else
 942                         t = netdev_priv(dev);
 943
 944                 memcpy(&p, &t->parms, sizeof (p));
 945                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
 946                         err = -EFAULT;
 947                 }
 948                 break;
 949         case SIOCADDTUNNEL:
 950         case SIOCCHGTUNNEL:
 951                 err = -EPERM;
 952                 create = (cmd == SIOCADDTUNNEL);
 953                 if (!capable(CAP_NET_ADMIN))
 954                         break;
 955                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
 956                         err = -EFAULT;
 957                         break;
 958                 }
 959                 if (!create && dev != ip6ip6_fb_tnl_dev) {
 960                         t = netdev_priv(dev);
 961                 }
 962                 if (!t && (err = ip6ip6_tnl_locate(&p, &t, create))) {
 963                         break;
 964                 }
 965                 if (cmd == SIOCCHGTUNNEL) {
 966                         if (t->dev != dev) {
 967                                 err = -EEXIST;
 968                                 break;
 969                         }
 970                         ip6ip6_tnl_unlink(t);
 971                         err = ip6ip6_tnl_change(t, &p);
 972                         ip6ip6_tnl_link(t);
 973                         netdev_state_change(dev);
 974                 }
 975                 if (copy_to_user(ifr->ifr_ifru.ifru_data,
 976                                  &t->parms, sizeof (p))) {
 977                         err = -EFAULT;
 978                 } else {
 979                         err = 0;
 980                 }
 981                 break;
 982         case SIOCDELTUNNEL:
 983                 err = -EPERM;
 984                 if (!capable(CAP_NET_ADMIN))
 985                         break;
 986
 987                 if (dev == ip6ip6_fb_tnl_dev) {
 988                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
 989                                            sizeof (p))) {
 990                                 err = -EFAULT;
 991                                 break;
 992                         }
 993                         err = ip6ip6_tnl_locate(&p, &t, 0);
 994                         if (err)
 995                                 break;
 996                         if (t == netdev_priv(ip6ip6_fb_tnl_dev)) {
 997                                 err = -EPERM;
 998                                 break;
 999                         }
1000                 } else {
1001                         t = netdev_priv(dev);
1002                 }
1003                 err = unregister_netdevice(t->dev);
1004                 break;
1005         default:
1006                 err = -EINVAL;
1007         }
1008         return err;
1009 }
1010
1011 /**
1012  * ip6ip6_tnl_get_stats - return the stats for tunnel device
1013  *   @dev: virtual device associated with tunnel
1014  *
1015  * Return: stats for device
1016  **/
1017
1018 static struct net_device_stats *
1019 ip6ip6_tnl_get_stats(struct net_device *dev)
1020 {
1021         return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1022 }
1023
1024 /**
1025  * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1026  *   @dev: virtual device associated with tunnel
1027  *   @new_mtu: the new mtu
1028  *
1029  * Return:
1030  *   0 on success,
1031  *   %-EINVAL if mtu too small
1032  **/
1033
1034 static int
1035 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1036 {
1037         if (new_mtu < IPV6_MIN_MTU) {
1038                 return -EINVAL;
1039         }
1040         dev->mtu = new_mtu;
1041         return 0;
1042 }
1043
1044 /**
1045  * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1046  *   @dev: virtual device associated with tunnel
1047  *
1048  * Description:
1049  *   Initialize function pointers and device parameters
1050  **/
1051
1052 static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1053 {
1054         SET_MODULE_OWNER(dev);
1055         dev->uninit = ip6ip6_tnl_dev_uninit;
1056         dev->destructor = free_netdev;
1057         dev->hard_start_xmit = ip6ip6_tnl_xmit;
1058         dev->get_stats = ip6ip6_tnl_get_stats;
1059         dev->do_ioctl = ip6ip6_tnl_ioctl;
1060         dev->change_mtu = ip6ip6_tnl_change_mtu;
1061
1062         dev->type = ARPHRD_TUNNEL6;
1063         dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1064         dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1065         dev->flags |= IFF_NOARP;
1066         dev->addr_len = sizeof(struct in6_addr);
1067 }
1068
1069
1070 /**
1071  * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1072  *   @dev: virtual device associated with tunnel
1073  **/
1074
1075 static inline void
1076 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1077 {
1078         struct ip6_tnl *t = netdev_priv(dev);
1079         t->fl.proto = IPPROTO_IPV6;
1080         t->dev = dev;
1081         strcpy(t->parms.name, dev->name);
1082 }
1083
1084 /**
1085  * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1086  *   @dev: virtual device associated with tunnel
1087  **/
1088
1089 static int
1090 ip6ip6_tnl_dev_init(struct net_device *dev)
1091 {
1092         struct ip6_tnl *t = netdev_priv(dev);
1093         ip6ip6_tnl_dev_init_gen(dev);
1094         ip6ip6_tnl_link_config(t);
1095         return 0;
1096 }
1097
1098 /**
1099  * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1100  *   @dev: fallback device
1101  *
1102  * Return: 0
1103  **/
1104
1105 static int
1106 ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1107 {
1108         struct ip6_tnl *t = netdev_priv(dev);
1109         ip6ip6_tnl_dev_init_gen(dev);
1110         dev_hold(dev);
1111         tnls_wc[0] = t;
1112         return 0;
1113 }
1114
1115 static struct xfrm6_tunnel ip6ip6_handler = {
1116         .handler        = ip6ip6_rcv,
1117         .err_handler    = ip6ip6_err,
1118         .priority       =       1,
1119 };
1120
1121 /**
1122  * ip6_tunnel_init - register protocol and reserve needed resources
1123  *
1124  * Return: 0 on success
1125  **/
1126
1127 static int __init ip6_tunnel_init(void)
1128 {
1129         int  err;
1130
1131         if (xfrm6_tunnel_register(&ip6ip6_handler)) {
1132                 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1133                 return -EAGAIN;
1134         }
1135         ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1136                                          ip6ip6_tnl_dev_setup);
1137
1138         if (!ip6ip6_fb_tnl_dev) {
1139                 err = -ENOMEM;
1140                 goto fail;
1141         }
1142         ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1143
1144         if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1145                 free_netdev(ip6ip6_fb_tnl_dev);
1146                 goto fail;
1147         }
1148         return 0;
1149 fail:
1150         xfrm6_tunnel_deregister(&ip6ip6_handler);
1151         return err;
1152 }
1153
1154 /**
1155  * ip6_tunnel_cleanup - free resources and unregister protocol
1156  **/
1157
1158 static void __exit ip6_tunnel_cleanup(void)
1159 {
1160         if (xfrm6_tunnel_deregister(&ip6ip6_handler))
1161                 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1162
1163         unregister_netdev(ip6ip6_fb_tnl_dev);
1164 }
1165
1166 module_init(ip6_tunnel_init);
1167 module_exit(ip6_tunnel_cleanup);