net/ipv6/ip6_tunnel.c

   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  *      $Id$
  10  *
  11  *      Based on:
  12  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  13  *
  14  *      RFC 2473
  15  *
  16  *      This program is free software; you can redistribute it and/or
  17  *      modify it under the terms of the GNU General Public License
  18  *      as published by the Free Software Foundation; either version
  19  *      2 of the License, or (at your option) any later version.
  20  *
  21  */
  22
  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/icmp.h>
  29 #include <linux/if.h>
  30 #include <linux/in.h>
  31 #include <linux/ip.h>
  32 #include <linux/if_tunnel.h>
  33 #include <linux/net.h>
  34 #include <linux/in6.h>
  35 #include <linux/netdevice.h>
  36 #include <linux/if_arp.h>
  37 #include <linux/icmpv6.h>
  38 #include <linux/init.h>
  39 #include <linux/route.h>
  40 #include <linux/rtnetlink.h>
  41 #include <linux/netfilter_ipv6.h>
  42
  43 #include <asm/uaccess.h>
  44 #include <asm/atomic.h>
  45
  46 #include <net/icmp.h>
  47 #include <net/ip.h>
  48 #include <net/ipv6.h>
  49 #include <net/ip6_route.h>
  50 #include <net/addrconf.h>
  51 #include <net/ip6_tunnel.h>
  52 #include <net/xfrm.h>
  53 #include <net/dsfield.h>
  54 #include <net/inet_ecn.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", __FUNCTION__)
  64 #else
  65 #define IP6_TNL_TRACE(x...) do {;} while(0)
  66 #endif
  67
  68 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
  69 #define IPV6_TCLASS_SHIFT 20
  70
  71 #define HASH_SIZE  32
  72
  73 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
  74                      (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
  75                     (HASH_SIZE - 1))
  76
  77 static int ip6_fb_tnl_dev_init(struct net_device *dev);
  78 static int ip6_tnl_dev_init(struct net_device *dev);
  79 static void ip6_tnl_dev_setup(struct net_device *dev);
  80
  81 /* the IPv6 tunnel fallback device */
  82 static struct net_device *ip6_fb_tnl_dev;
  83
  84
  85 /* lists for storing tunnels in use */
  86 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
  87 static struct ip6_tnl *tnls_wc[1];
  88 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
  89
  90 /* lock for the tunnel lists */
  91 static DEFINE_RWLOCK(ip6_tnl_lock);
  92
  93 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
  94 {
  95         struct dst_entry *dst = t->dst_cache;
  96
  97         if (dst && dst->obsolete &&
  98             dst->ops->check(dst, t->dst_cookie) == NULL) {
  99                 t->dst_cache = NULL;
 100                 dst_release(dst);
 101                 return NULL;
 102         }
 103
 104         return dst;
 105 }
 106
 107 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
 108 {
 109         dst_release(t->dst_cache);
 110         t->dst_cache = NULL;
 111 }
 112
 113 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
 114 {
 115         struct rt6_info *rt = (struct rt6_info *) dst;
 116         t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
 117         dst_release(t->dst_cache);
 118         t->dst_cache = dst;
 119 }
 120
 121 /**
 122  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 123  *   @remote: the address of the tunnel exit-point
 124  *   @local: the address of the tunnel entry-point
 125  *
 126  * Return:
 127  *   tunnel matching given end-points if found,
 128  *   else fallback tunnel if its device is up,
 129  *   else %NULL
 130  **/
 131
 132 static struct ip6_tnl *
 133 ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
 134 {
 135         unsigned h0 = HASH(remote);
 136         unsigned h1 = HASH(local);
 137         struct ip6_tnl *t;
 138
 139         for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
 140                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 141                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 142                     (t->dev->flags & IFF_UP))
 143                         return t;
 144         }
 145         if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
 146                 return t;
 147
 148         return NULL;
 149 }
 150
 151 /**
 152  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 153  *   @p: parameters containing tunnel end-points
 154  *
 155  * Description:
 156  *   ip6_tnl_bucket() returns the head of the list matching the
 157  *   &struct in6_addr entries laddr and raddr in @p.
 158  *
 159  * Return: head of IPv6 tunnel list
 160  **/
 161
 162 static struct ip6_tnl **
 163 ip6_tnl_bucket(struct ip6_tnl_parm *p)
 164 {
 165         struct in6_addr *remote = &p->raddr;
 166         struct in6_addr *local = &p->laddr;
 167         unsigned h = 0;
 168         int prio = 0;
 169
 170         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 171                 prio = 1;
 172                 h = HASH(remote) ^ HASH(local);
 173         }
 174         return &tnls[prio][h];
 175 }
 176
 177 /**
 178  * ip6_tnl_link - add tunnel to hash table
 179  *   @t: tunnel to be added
 180  **/
 181
 182 static void
 183 ip6_tnl_link(struct ip6_tnl *t)
 184 {
 185         struct ip6_tnl **tp = ip6_tnl_bucket(&t->parms);
 186
 187         t->next = *tp;
 188         write_lock_bh(&ip6_tnl_lock);
 189         *tp = t;
 190         write_unlock_bh(&ip6_tnl_lock);
 191 }
 192
 193 /**
 194  * ip6_tnl_unlink - remove tunnel from hash table
 195  *   @t: tunnel to be removed
 196  **/
 197
 198 static void
 199 ip6_tnl_unlink(struct ip6_tnl *t)
 200 {
 201         struct ip6_tnl **tp;
 202
 203         for (tp = ip6_tnl_bucket(&t->parms); *tp; tp = &(*tp)->next) {
 204                 if (t == *tp) {
 205                         write_lock_bh(&ip6_tnl_lock);
 206                         *tp = t->next;
 207                         write_unlock_bh(&ip6_tnl_lock);
 208                         break;
 209                 }
 210         }
 211 }
 212
 213 /**
 214  * ip6_tnl_create() - create a new tunnel
 215  *   @p: tunnel parameters
 216  *   @pt: pointer to new tunnel
 217  *
 218  * Description:
 219  *   Create tunnel matching given parameters.
 220  *
 221  * Return:
 222  *   created tunnel or NULL
 223  **/
 224
 225 static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
 226 {
 227         struct net_device *dev;
 228         struct ip6_tnl *t;
 229         char name[IFNAMSIZ];
 230         int err;
 231
 232         if (p->name[0])
 233                 strlcpy(name, p->name, IFNAMSIZ);
 234         else
 235                 sprintf(name, "ip6tnl%%d");
 236
 237         dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
 238         if (dev == NULL)
 239                 goto failed;
 240
 241         t = netdev_priv(dev);
 242         dev->init = ip6_tnl_dev_init;
 243         t->parms = *p;
 244
 245         if ((err = register_netdevice(dev)) < 0) {
 246                 free_netdev(dev);
 247                 goto failed;
 248         }
 249         dev_hold(dev);
 250         ip6_tnl_link(t);
 251         return t;
 252 failed:
 253         return NULL;
 254 }
 255
 256 /**
 257  * ip6_tnl_locate - find or create tunnel matching given parameters
 258  *   @p: tunnel parameters
 259  *   @create: != 0 if allowed to create new tunnel if no match found
 260  *
 261  * Description:
 262  *   ip6_tnl_locate() first tries to locate an existing tunnel
 263  *   based on @parms. If this is unsuccessful, but @create is set a new
 264  *   tunnel device is created and registered for use.
 265  *
 266  * Return:
 267  *   matching tunnel or NULL
 268  **/
 269
 270 static struct ip6_tnl *ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
 271 {
 272         struct in6_addr *remote = &p->raddr;
 273         struct in6_addr *local = &p->laddr;
 274         struct ip6_tnl *t;
 275
 276         for (t = *ip6_tnl_bucket(p); t; t = t->next) {
 277                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 278                     ipv6_addr_equal(remote, &t->parms.raddr))
 279                         return t;
 280         }
 281         if (!create)
 282                 return NULL;
 283         return ip6_tnl_create(p);
 284 }
 285
 286 /**
 287  * ip6_tnl_dev_uninit - tunnel device uninitializer
 288  *   @dev: the device to be destroyed
 289  *
 290  * Description:
 291  *   ip6_tnl_dev_uninit() removes tunnel from its list
 292  **/
 293
 294 static void
 295 ip6_tnl_dev_uninit(struct net_device *dev)
 296 {
 297         struct ip6_tnl *t = netdev_priv(dev);
 298
 299         if (dev == ip6_fb_tnl_dev) {
 300                 write_lock_bh(&ip6_tnl_lock);
 301                 tnls_wc[0] = NULL;
 302                 write_unlock_bh(&ip6_tnl_lock);
 303         } else {
 304                 ip6_tnl_unlink(t);
 305         }
 306         ip6_tnl_dst_reset(t);
 307         dev_put(dev);
 308 }
 309
 310 /**
 311  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 312  *   @skb: received socket buffer
 313  *
 314  * Return:
 315  *   0 if none was found,
 316  *   else index to encapsulation limit
 317  **/
 318
 319 static __u16
 320 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
 321 {
 322         struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
 323         __u8 nexthdr = ipv6h->nexthdr;
 324         __u16 off = sizeof (*ipv6h);
 325
 326         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 327                 __u16 optlen = 0;
 328                 struct ipv6_opt_hdr *hdr;
 329                 if (raw + off + sizeof (*hdr) > skb->data &&
 330                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 331                         break;
 332
 333                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 334                 if (nexthdr == NEXTHDR_FRAGMENT) {
 335                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 336                         if (frag_hdr->frag_off)
 337                                 break;
 338                         optlen = 8;
 339                 } else if (nexthdr == NEXTHDR_AUTH) {
 340                         optlen = (hdr->hdrlen + 2) << 2;
 341                 } else {
 342                         optlen = ipv6_optlen(hdr);
 343                 }
 344                 if (nexthdr == NEXTHDR_DEST) {
 345                         __u16 i = off + 2;
 346                         while (1) {
 347                                 struct ipv6_tlv_tnl_enc_lim *tel;
 348
 349                                 /* No more room for encapsulation limit */
 350                                 if (i + sizeof (*tel) > off + optlen)
 351                                         break;
 352
 353                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 354                                 /* return index of option if found and valid */
 355                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 356                                     tel->length == 1)
 357                                         return i;
 358                                 /* else jump to next option */
 359                                 if (tel->type)
 360                                         i += tel->length + 2;
 361                                 else
 362                                         i++;
 363                         }
 364                 }
 365                 nexthdr = hdr->nexthdr;
 366                 off += optlen;
 367         }
 368         return 0;
 369 }
 370
 371 /**
 372  * ip6_tnl_err - tunnel error handler
 373  *
 374  * Description:
 375  *   ip6_tnl_err() should handle errors in the tunnel according
 376  *   to the specifications in RFC 2473.
 377  **/
 378
 379 static int
 380 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 381             int *type, int *code, int *msg, __u32 *info, int offset)
 382 {
 383         struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
 384         struct ip6_tnl *t;
 385         int rel_msg = 0;
 386         int rel_type = ICMPV6_DEST_UNREACH;
 387         int rel_code = ICMPV6_ADDR_UNREACH;
 388         __u32 rel_info = 0;
 389         __u16 len;
 390         int err = -ENOENT;
 391
 392         /* If the packet doesn't contain the original IPv6 header we are
 393            in trouble since we might need the source address for further
 394            processing of the error. */
 395
 396         read_lock(&ip6_tnl_lock);
 397         if ((t = ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
 398                 goto out;
 399
 400         if (t->parms.proto != ipproto && t->parms.proto != 0)
 401                 goto out;
 402
 403         err = 0;
 404
 405         switch (*type) {
 406                 __u32 teli;
 407                 struct ipv6_tlv_tnl_enc_lim *tel;
 408                 __u32 mtu;
 409         case ICMPV6_DEST_UNREACH:
 410                 if (net_ratelimit())
 411                         printk(KERN_WARNING
 412                                "%s: Path to destination invalid "
 413                                "or inactive!\n", t->parms.name);
 414                 rel_msg = 1;
 415                 break;
 416         case ICMPV6_TIME_EXCEED:
 417                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 418                         if (net_ratelimit())
 419                                 printk(KERN_WARNING
 420                                        "%s: Too small hop limit or "
 421                                        "routing loop in tunnel!\n",
 422                                        t->parms.name);
 423                         rel_msg = 1;
 424                 }
 425                 break;
 426         case ICMPV6_PARAMPROB:
 427                 teli = 0;
 428                 if ((*code) == ICMPV6_HDR_FIELD)
 429                         teli = parse_tlv_tnl_enc_lim(skb, skb->data);
 430
 431                 if (teli && teli == *info - 2) {
 432                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 433                         if (tel->encap_limit == 0) {
 434                                 if (net_ratelimit())
 435                                         printk(KERN_WARNING
 436                                                "%s: Too small encapsulation "
 437                                                "limit or routing loop in "
 438                                                "tunnel!\n", t->parms.name);
 439                                 rel_msg = 1;
 440                         }
 441                 } else if (net_ratelimit()) {
 442                         printk(KERN_WARNING
 443                                "%s: Recipient unable to parse tunneled "
 444                                "packet!\n ", t->parms.name);
 445                 }
 446                 break;
 447         case ICMPV6_PKT_TOOBIG:
 448                 mtu = *info - offset;
 449                 if (mtu < IPV6_MIN_MTU)
 450                         mtu = IPV6_MIN_MTU;
 451                 t->dev->mtu = mtu;
 452
 453                 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
 454                         rel_type = ICMPV6_PKT_TOOBIG;
 455                         rel_code = 0;
 456                         rel_info = mtu;
 457                         rel_msg = 1;
 458                 }
 459                 break;
 460         }
 461
 462         *type = rel_type;
 463         *code = rel_code;
 464         *info = rel_info;
 465         *msg = rel_msg;
 466
 467 out:
 468         read_unlock(&ip6_tnl_lock);
 469         return err;
 470 }
 471
 472 static int
 473 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 474            int type, int code, int offset, __be32 info)
 475 {
 476         int rel_msg = 0;
 477         int rel_type = type;
 478         int rel_code = code;
 479         __u32 rel_info = ntohl(info);
 480         int err;
 481         struct sk_buff *skb2;
 482         struct iphdr *eiph;
 483         struct flowi fl;
 484         struct rtable *rt;
 485
 486         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 487                           &rel_msg, &rel_info, offset);
 488         if (err < 0)
 489                 return err;
 490
 491         if (rel_msg == 0)
 492                 return 0;
 493
 494         switch (rel_type) {
 495         case ICMPV6_DEST_UNREACH:
 496                 if (rel_code != ICMPV6_ADDR_UNREACH)
 497                         return 0;
 498                 rel_type = ICMP_DEST_UNREACH;
 499                 rel_code = ICMP_HOST_UNREACH;
 500                 break;
 501         case ICMPV6_PKT_TOOBIG:
 502                 if (rel_code != 0)
 503                         return 0;
 504                 rel_type = ICMP_DEST_UNREACH;
 505                 rel_code = ICMP_FRAG_NEEDED;
 506                 break;
 507         default:
 508                 return 0;
 509         }
 510
 511         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 512                 return 0;
 513
 514         skb2 = skb_clone(skb, GFP_ATOMIC);
 515         if (!skb2)
 516                 return 0;
 517
 518         dst_release(skb2->dst);
 519         skb2->dst = NULL;
 520         skb_pull(skb2, offset);
 521         skb_reset_network_header(skb2);
 522         eiph = ip_hdr(skb2);
 523
 524         /* Try to guess incoming interface */
 525         memset(&fl, 0, sizeof(fl));
 526         fl.fl4_dst = eiph->saddr;
 527         fl.fl4_tos = RT_TOS(eiph->tos);
 528         fl.proto = IPPROTO_IPIP;
 529         if (ip_route_output_key(&init_net, &rt, &fl))
 530                 goto out;
 531
 532         skb2->dev = rt->u.dst.dev;
 533
 534         /* route "incoming" packet */
 535         if (rt->rt_flags & RTCF_LOCAL) {
 536                 ip_rt_put(rt);
 537                 rt = NULL;
 538                 fl.fl4_dst = eiph->daddr;
 539                 fl.fl4_src = eiph->saddr;
 540                 fl.fl4_tos = eiph->tos;
 541                 if (ip_route_output_key(&init_net, &rt, &fl) ||
 542                     rt->u.dst.dev->type != ARPHRD_TUNNEL) {
 543                         ip_rt_put(rt);
 544                         goto out;
 545                 }
 546                 skb2->dst = (struct dst_entry *)rt;
 547         } else {
 548                 ip_rt_put(rt);
 549                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 550                                    skb2->dev) ||
 551                     skb2->dst->dev->type != ARPHRD_TUNNEL)
 552                         goto out;
 553         }
 554
 555         /* change mtu on this route */
 556         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 557                 if (rel_info > dst_mtu(skb2->dst))
 558                         goto out;
 559
 560                 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
 561         }
 562
 563         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 564
 565 out:
 566         kfree_skb(skb2);
 567         return 0;
 568 }
 569
 570 static int
 571 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 572            int type, int code, int offset, __be32 info)
 573 {
 574         int rel_msg = 0;
 575         int rel_type = type;
 576         int rel_code = code;
 577         __u32 rel_info = ntohl(info);
 578         int err;
 579
 580         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 581                           &rel_msg, &rel_info, offset);
 582         if (err < 0)
 583                 return err;
 584
 585         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 586                 struct rt6_info *rt;
 587                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 588
 589                 if (!skb2)
 590                         return 0;
 591
 592                 dst_release(skb2->dst);
 593                 skb2->dst = NULL;
 594                 skb_pull(skb2, offset);
 595                 skb_reset_network_header(skb2);
 596
 597                 /* Try to guess incoming interface */
 598                 rt = rt6_lookup(&ipv6_hdr(skb2)->saddr, NULL, 0, 0);
 599
 600                 if (rt && rt->rt6i_dev)
 601                         skb2->dev = rt->rt6i_dev;
 602
 603                 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 604
 605                 if (rt)
 606                         dst_release(&rt->u.dst);
 607
 608                 kfree_skb(skb2);
 609         }
 610
 611         return 0;
 612 }
 613
 614 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
 615                                         struct ipv6hdr *ipv6h,
 616                                         struct sk_buff *skb)
 617 {
 618         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 619
 620         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 621                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 622
 623         if (INET_ECN_is_ce(dsfield))
 624                 IP_ECN_set_ce(ip_hdr(skb));
 625 }
 626
 627 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
 628                                         struct ipv6hdr *ipv6h,
 629                                         struct sk_buff *skb)
 630 {
 631         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 632                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 633
 634         if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
 635                 IP6_ECN_set_ce(ipv6_hdr(skb));
 636 }
 637
 638 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
 639 {
 640         struct ip6_tnl_parm *p = &t->parms;
 641         int ret = 0;
 642
 643         if (p->flags & IP6_TNL_F_CAP_RCV) {
 644                 struct net_device *ldev = NULL;
 645
 646                 if (p->link)
 647                         ldev = dev_get_by_index(&init_net, p->link);
 648
 649                 if ((ipv6_addr_is_multicast(&p->laddr) ||
 650                      likely(ipv6_chk_addr(&init_net, &p->laddr, ldev, 0))) &&
 651                     likely(!ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
 652                         ret = 1;
 653
 654                 if (ldev)
 655                         dev_put(ldev);
 656         }
 657         return ret;
 658 }
 659
 660 /**
 661  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 662  *   @skb: received socket buffer
 663  *   @protocol: ethernet protocol ID
 664  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 665  *
 666  * Return: 0
 667  **/
 668
 669 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 670                        __u8 ipproto,
 671                        void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
 672                                                     struct ipv6hdr *ipv6h,
 673                                                     struct sk_buff *skb))
 674 {
 675         struct ip6_tnl *t;
 676         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 677
 678         read_lock(&ip6_tnl_lock);
 679
 680         if ((t = ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
 681                 if (t->parms.proto != ipproto && t->parms.proto != 0) {
 682                         read_unlock(&ip6_tnl_lock);
 683                         goto discard;
 684                 }
 685
 686                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 687                         read_unlock(&ip6_tnl_lock);
 688                         goto discard;
 689                 }
 690
 691                 if (!ip6_tnl_rcv_ctl(t)) {
 692                         t->stat.rx_dropped++;
 693                         read_unlock(&ip6_tnl_lock);
 694                         goto discard;
 695                 }
 696                 secpath_reset(skb);
 697                 skb->mac_header = skb->network_header;
 698                 skb_reset_network_header(skb);
 699                 skb->protocol = htons(protocol);
 700                 skb->pkt_type = PACKET_HOST;
 701                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 702                 skb->dev = t->dev;
 703                 dst_release(skb->dst);
 704                 skb->dst = NULL;
 705                 nf_reset(skb);
 706
 707                 dscp_ecn_decapsulate(t, ipv6h, skb);
 708
 709                 t->stat.rx_packets++;
 710                 t->stat.rx_bytes += skb->len;
 711                 netif_rx(skb);
 712                 read_unlock(&ip6_tnl_lock);
 713                 return 0;
 714         }
 715         read_unlock(&ip6_tnl_lock);
 716         return 1;
 717
 718 discard:
 719         kfree_skb(skb);
 720         return 0;
 721 }
 722
 723 static int ip4ip6_rcv(struct sk_buff *skb)
 724 {
 725         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 726                            ip4ip6_dscp_ecn_decapsulate);
 727 }
 728
 729 static int ip6ip6_rcv(struct sk_buff *skb)
 730 {
 731         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 732                            ip6ip6_dscp_ecn_decapsulate);
 733 }
 734
 735 struct ipv6_tel_txoption {
 736         struct ipv6_txoptions ops;
 737         __u8 dst_opt[8];
 738 };
 739
 740 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 741 {
 742         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 743
 744         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 745         opt->dst_opt[3] = 1;
 746         opt->dst_opt[4] = encap_limit;
 747         opt->dst_opt[5] = IPV6_TLV_PADN;
 748         opt->dst_opt[6] = 1;
 749
 750         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 751         opt->ops.opt_nflen = 8;
 752 }
 753
 754 /**
 755  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 756  *   @t: the outgoing tunnel device
 757  *   @hdr: IPv6 header from the incoming packet
 758  *
 759  * Description:
 760  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 761  *   doesn't match source of incoming packet.
 762  *
 763  * Return:
 764  *   1 if conflict,
 765  *   0 else
 766  **/
 767
 768 static inline int
 769 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
 770 {
 771         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 772 }
 773
 774 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
 775 {
 776         struct ip6_tnl_parm *p = &t->parms;
 777         int ret = 0;
 778
 779         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 780                 struct net_device *ldev = NULL;
 781
 782                 if (p->link)
 783                         ldev = dev_get_by_index(&init_net, p->link);
 784
 785                 if (unlikely(!ipv6_chk_addr(&init_net, &p->laddr, ldev, 0)))
 786                         printk(KERN_WARNING
 787                                "%s xmit: Local address not yet configured!\n",
 788                                p->name);
 789                 else if (!ipv6_addr_is_multicast(&p->raddr) &&
 790                          unlikely(ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
 791                         printk(KERN_WARNING
 792                                "%s xmit: Routing loop! "
 793                                "Remote address found on this node!\n",
 794                                p->name);
 795                 else
 796                         ret = 1;
 797                 if (ldev)
 798                         dev_put(ldev);
 799         }
 800         return ret;
 801 }
 802 /**
 803  * ip6_tnl_xmit2 - encapsulate packet and send
 804  *   @skb: the outgoing socket buffer
 805  *   @dev: the outgoing tunnel device
 806  *   @dsfield: dscp code for outer header
 807  *   @fl: flow of tunneled packet
 808  *   @encap_limit: encapsulation limit
 809  *   @pmtu: Path MTU is stored if packet is too big
 810  *
 811  * Description:
 812  *   Build new header and do some sanity checks on the packet before sending
 813  *   it.
 814  *
 815  * Return:
 816  *   0 on success
 817  *   -1 fail
 818  *   %-EMSGSIZE message too big. return mtu in this case.
 819  **/
 820
 821 static int ip6_tnl_xmit2(struct sk_buff *skb,
 822                          struct net_device *dev,
 823                          __u8 dsfield,
 824                          struct flowi *fl,
 825                          int encap_limit,
 826                          __u32 *pmtu)
 827 {
 828         struct ip6_tnl *t = netdev_priv(dev);
 829         struct net_device_stats *stats = &t->stat;
 830         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 831         struct ipv6_tel_txoption opt;
 832         struct dst_entry *dst;
 833         struct net_device *tdev;
 834         int mtu;
 835         unsigned int max_headroom = sizeof(struct ipv6hdr);
 836         u8 proto;
 837         int err = -1;
 838         int pkt_len;
 839
 840         if ((dst = ip6_tnl_dst_check(t)) != NULL)
 841                 dst_hold(dst);
 842         else {
 843                 dst = ip6_route_output(NULL, fl);
 844
 845                 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
 846                         goto tx_err_link_failure;
 847         }
 848
 849         tdev = dst->dev;
 850
 851         if (tdev == dev) {
 852                 stats->collisions++;
 853                 if (net_ratelimit())
 854                         printk(KERN_WARNING
 855                                "%s: Local routing loop detected!\n",
 856                                t->parms.name);
 857                 goto tx_err_dst_release;
 858         }
 859         mtu = dst_mtu(dst) - sizeof (*ipv6h);
 860         if (encap_limit >= 0) {
 861                 max_headroom += 8;
 862                 mtu -= 8;
 863         }
 864         if (mtu < IPV6_MIN_MTU)
 865                 mtu = IPV6_MIN_MTU;
 866         if (skb->dst)
 867                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 868         if (skb->len > mtu) {
 869                 *pmtu = mtu;
 870                 err = -EMSGSIZE;
 871                 goto tx_err_dst_release;
 872         }
 873
 874         /*
 875          * Okay, now see if we can stuff it in the buffer as-is.
 876          */
 877         max_headroom += LL_RESERVED_SPACE(tdev);
 878
 879         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 880             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 881                 struct sk_buff *new_skb;
 882
 883                 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
 884                         goto tx_err_dst_release;
 885
 886                 if (skb->sk)
 887                         skb_set_owner_w(new_skb, skb->sk);
 888                 kfree_skb(skb);
 889                 skb = new_skb;
 890         }
 891         dst_release(skb->dst);
 892         skb->dst = dst_clone(dst);
 893
 894         skb->transport_header = skb->network_header;
 895
 896         proto = fl->proto;
 897         if (encap_limit >= 0) {
 898                 init_tel_txopt(&opt, encap_limit);
 899                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 900         }
 901         skb_push(skb, sizeof(struct ipv6hdr));
 902         skb_reset_network_header(skb);
 903         ipv6h = ipv6_hdr(skb);
 904         *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
 905         dsfield = INET_ECN_encapsulate(0, dsfield);
 906         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 907         ipv6h->hop_limit = t->parms.hop_limit;
 908         ipv6h->nexthdr = proto;
 909         ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
 910         ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
 911         nf_reset(skb);
 912         pkt_len = skb->len;
 913         err = ip6_local_out(skb);
 914
 915         if (net_xmit_eval(err) == 0) {
 916                 stats->tx_bytes += pkt_len;
 917                 stats->tx_packets++;
 918         } else {
 919                 stats->tx_errors++;
 920                 stats->tx_aborted_errors++;
 921         }
 922         ip6_tnl_dst_store(t, dst);
 923         return 0;
 924 tx_err_link_failure:
 925         stats->tx_carrier_errors++;
 926         dst_link_failure(skb);
 927 tx_err_dst_release:
 928         dst_release(dst);
 929         return err;
 930 }
 931
 932 static inline int
 933 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 934 {
 935         struct ip6_tnl *t = netdev_priv(dev);
 936         struct iphdr  *iph = ip_hdr(skb);
 937         int encap_limit = -1;
 938         struct flowi fl;
 939         __u8 dsfield;
 940         __u32 mtu;
 941         int err;
 942
 943         if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
 944             !ip6_tnl_xmit_ctl(t))
 945                 return -1;
 946
 947         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 948                 encap_limit = t->parms.encap_limit;
 949
 950         memcpy(&fl, &t->fl, sizeof (fl));
 951         fl.proto = IPPROTO_IPIP;
 952
 953         dsfield = ipv4_get_dsfield(iph);
 954
 955         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
 956                 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 957                                           & IPV6_TCLASS_MASK;
 958
 959         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
 960         if (err != 0) {
 961                 /* XXX: send ICMP error even if DF is not set. */
 962                 if (err == -EMSGSIZE)
 963                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 964                                   htonl(mtu));
 965                 return -1;
 966         }
 967
 968         return 0;
 969 }
 970
 971 static inline int
 972 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 973 {
 974         struct ip6_tnl *t = netdev_priv(dev);
 975         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 976         int encap_limit = -1;
 977         __u16 offset;
 978         struct flowi fl;
 979         __u8 dsfield;
 980         __u32 mtu;
 981         int err;
 982
 983         if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
 984             !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
 985                 return -1;
 986
 987         offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
 988         if (offset > 0) {
 989                 struct ipv6_tlv_tnl_enc_lim *tel;
 990                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 991                 if (tel->encap_limit == 0) {
 992                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 993                                     ICMPV6_HDR_FIELD, offset + 2, skb->dev);
 994                         return -1;
 995                 }
 996                 encap_limit = tel->encap_limit - 1;
 997         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 998                 encap_limit = t->parms.encap_limit;
 999
1000         memcpy(&fl, &t->fl, sizeof (fl));
1001         fl.proto = IPPROTO_IPV6;
1002
1003         dsfield = ipv6_get_dsfield(ipv6h);
1004         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1005                 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1006         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1007                 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1008
1009         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1010         if (err != 0) {
1011                 if (err == -EMSGSIZE)
1012                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1013                 return -1;
1014         }
1015
1016         return 0;
1017 }
1018
1019 static int
1020 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1021 {
1022         struct ip6_tnl *t = netdev_priv(dev);
1023         struct net_device_stats *stats = &t->stat;
1024         int ret;
1025
1026         if (t->recursion++) {
1027                 t->stat.collisions++;
1028                 goto tx_err;
1029         }
1030
1031         switch (skb->protocol) {
1032         case __constant_htons(ETH_P_IP):
1033                 ret = ip4ip6_tnl_xmit(skb, dev);
1034                 break;
1035         case __constant_htons(ETH_P_IPV6):
1036                 ret = ip6ip6_tnl_xmit(skb, dev);
1037                 break;
1038         default:
1039                 goto tx_err;
1040         }
1041
1042         if (ret < 0)
1043                 goto tx_err;
1044
1045         t->recursion--;
1046         return 0;
1047
1048 tx_err:
1049         stats->tx_errors++;
1050         stats->tx_dropped++;
1051         kfree_skb(skb);
1052         t->recursion--;
1053         return 0;
1054 }
1055
1056 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1057 {
1058         struct ip6_tnl_parm *p = &t->parms;
1059         int ltype = ipv6_addr_type(&p->laddr);
1060         int rtype = ipv6_addr_type(&p->raddr);
1061
1062         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1063
1064         if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1065             rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1066             !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1067             (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1068                 if (ltype&IPV6_ADDR_UNICAST)
1069                         p->flags |= IP6_TNL_F_CAP_XMIT;
1070                 if (rtype&IPV6_ADDR_UNICAST)
1071                         p->flags |= IP6_TNL_F_CAP_RCV;
1072         }
1073 }
1074
1075 static void ip6_tnl_link_config(struct ip6_tnl *t)
1076 {
1077         struct net_device *dev = t->dev;
1078         struct ip6_tnl_parm *p = &t->parms;
1079         struct flowi *fl = &t->fl;
1080
1081         memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1082         memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1083
1084         /* Set up flowi template */
1085         ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1086         ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1087         fl->oif = p->link;
1088         fl->fl6_flowlabel = 0;
1089
1090         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1091                 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1092         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1093                 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1094
1095         ip6_tnl_set_cap(t);
1096
1097         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1098                 dev->flags |= IFF_POINTOPOINT;
1099         else
1100                 dev->flags &= ~IFF_POINTOPOINT;
1101
1102         dev->iflink = p->link;
1103
1104         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1105                 int strict = (ipv6_addr_type(&p->raddr) &
1106                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1107
1108                 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
1109                                                  p->link, strict);
1110
1111                 if (rt == NULL)
1112                         return;
1113
1114                 if (rt->rt6i_dev) {
1115                         dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1116                                 sizeof (struct ipv6hdr);
1117
1118                         dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1119
1120                         if (dev->mtu < IPV6_MIN_MTU)
1121                                 dev->mtu = IPV6_MIN_MTU;
1122                 }
1123                 dst_release(&rt->u.dst);
1124         }
1125 }
1126
1127 /**
1128  * ip6_tnl_change - update the tunnel parameters
1129  *   @t: tunnel to be changed
1130  *   @p: tunnel configuration parameters
1131  *   @active: != 0 if tunnel is ready for use
1132  *
1133  * Description:
1134  *   ip6_tnl_change() updates the tunnel parameters
1135  **/
1136
1137 static int
1138 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1139 {
1140         ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1141         ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1142         t->parms.flags = p->flags;
1143         t->parms.hop_limit = p->hop_limit;
1144         t->parms.encap_limit = p->encap_limit;
1145         t->parms.flowinfo = p->flowinfo;
1146         t->parms.link = p->link;
1147         t->parms.proto = p->proto;
1148         ip6_tnl_dst_reset(t);
1149         ip6_tnl_link_config(t);
1150         return 0;
1151 }
1152
1153 /**
1154  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1155  *   @dev: virtual device associated with tunnel
1156  *   @ifr: parameters passed from userspace
1157  *   @cmd: command to be performed
1158  *
1159  * Description:
1160  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1161  *   from userspace.
1162  *
1163  *   The possible commands are the following:
1164  *     %SIOCGETTUNNEL: get tunnel parameters for device
1165  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1166  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1167  *     %SIOCDELTUNNEL: delete tunnel
1168  *
1169  *   The fallback device "ip6tnl0", created during module
1170  *   initialization, can be used for creating other tunnel devices.
1171  *
1172  * Return:
1173  *   0 on success,
1174  *   %-EFAULT if unable to copy data to or from userspace,
1175  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1176  *   %-EINVAL if passed tunnel parameters are invalid,
1177  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1178  *   %-ENODEV if attempting to change or delete a nonexisting device
1179  **/
1180
1181 static int
1182 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1183 {
1184         int err = 0;
1185         struct ip6_tnl_parm p;
1186         struct ip6_tnl *t = NULL;
1187
1188         switch (cmd) {
1189         case SIOCGETTUNNEL:
1190                 if (dev == ip6_fb_tnl_dev) {
1191                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1192                                 err = -EFAULT;
1193                                 break;
1194                         }
1195                         t = ip6_tnl_locate(&p, 0);
1196                 }
1197                 if (t == NULL)
1198                         t = netdev_priv(dev);
1199                 memcpy(&p, &t->parms, sizeof (p));
1200                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1201                         err = -EFAULT;
1202                 }
1203                 break;
1204         case SIOCADDTUNNEL:
1205         case SIOCCHGTUNNEL:
1206                 err = -EPERM;
1207                 if (!capable(CAP_NET_ADMIN))
1208                         break;
1209                 err = -EFAULT;
1210                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1211                         break;
1212                 err = -EINVAL;
1213                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1214                     p.proto != 0)
1215                         break;
1216                 t = ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
1217                 if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1218                         if (t != NULL) {
1219                                 if (t->dev != dev) {
1220                                         err = -EEXIST;
1221                                         break;
1222                                 }
1223                         } else
1224                                 t = netdev_priv(dev);
1225
1226                         ip6_tnl_unlink(t);
1227                         err = ip6_tnl_change(t, &p);
1228                         ip6_tnl_link(t);
1229                         netdev_state_change(dev);
1230                 }
1231                 if (t) {
1232                         err = 0;
1233                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1234                                 err = -EFAULT;
1235
1236                 } else
1237                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1238                 break;
1239         case SIOCDELTUNNEL:
1240                 err = -EPERM;
1241                 if (!capable(CAP_NET_ADMIN))
1242                         break;
1243
1244                 if (dev == ip6_fb_tnl_dev) {
1245                         err = -EFAULT;
1246                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1247                                 break;
1248                         err = -ENOENT;
1249                         if ((t = ip6_tnl_locate(&p, 0)) == NULL)
1250                                 break;
1251                         err = -EPERM;
1252                         if (t->dev == ip6_fb_tnl_dev)
1253                                 break;
1254                         dev = t->dev;
1255                 }
1256                 err = 0;
1257                 unregister_netdevice(dev);
1258                 break;
1259         default:
1260                 err = -EINVAL;
1261         }
1262         return err;
1263 }
1264
1265 /**
1266  * ip6_tnl_get_stats - return the stats for tunnel device
1267  *   @dev: virtual device associated with tunnel
1268  *
1269  * Return: stats for device
1270  **/
1271
1272 static struct net_device_stats *
1273 ip6_tnl_get_stats(struct net_device *dev)
1274 {
1275         return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1276 }
1277
1278 /**
1279  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1280  *   @dev: virtual device associated with tunnel
1281  *   @new_mtu: the new mtu
1282  *
1283  * Return:
1284  *   0 on success,
1285  *   %-EINVAL if mtu too small
1286  **/
1287
1288 static int
1289 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1290 {
1291         if (new_mtu < IPV6_MIN_MTU) {
1292                 return -EINVAL;
1293         }
1294         dev->mtu = new_mtu;
1295         return 0;
1296 }
1297
1298 /**
1299  * ip6_tnl_dev_setup - setup virtual tunnel device
1300  *   @dev: virtual device associated with tunnel
1301  *
1302  * Description:
1303  *   Initialize function pointers and device parameters
1304  **/
1305
1306 static void ip6_tnl_dev_setup(struct net_device *dev)
1307 {
1308         dev->uninit = ip6_tnl_dev_uninit;
1309         dev->destructor = free_netdev;
1310         dev->hard_start_xmit = ip6_tnl_xmit;
1311         dev->get_stats = ip6_tnl_get_stats;
1312         dev->do_ioctl = ip6_tnl_ioctl;
1313         dev->change_mtu = ip6_tnl_change_mtu;
1314
1315         dev->type = ARPHRD_TUNNEL6;
1316         dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1317         dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1318         dev->flags |= IFF_NOARP;
1319         dev->addr_len = sizeof(struct in6_addr);
1320 }
1321
1322
1323 /**
1324  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1325  *   @dev: virtual device associated with tunnel
1326  **/
1327
1328 static inline void
1329 ip6_tnl_dev_init_gen(struct net_device *dev)
1330 {
1331         struct ip6_tnl *t = netdev_priv(dev);
1332         t->dev = dev;
1333         strcpy(t->parms.name, dev->name);
1334 }
1335
1336 /**
1337  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1338  *   @dev: virtual device associated with tunnel
1339  **/
1340
1341 static int
1342 ip6_tnl_dev_init(struct net_device *dev)
1343 {
1344         struct ip6_tnl *t = netdev_priv(dev);
1345         ip6_tnl_dev_init_gen(dev);
1346         ip6_tnl_link_config(t);
1347         return 0;
1348 }
1349
1350 /**
1351  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1352  *   @dev: fallback device
1353  *
1354  * Return: 0
1355  **/
1356
1357 static int
1358 ip6_fb_tnl_dev_init(struct net_device *dev)
1359 {
1360         struct ip6_tnl *t = netdev_priv(dev);
1361         ip6_tnl_dev_init_gen(dev);
1362         t->parms.proto = IPPROTO_IPV6;
1363         dev_hold(dev);
1364         tnls_wc[0] = t;
1365         return 0;
1366 }
1367
1368 static struct xfrm6_tunnel ip4ip6_handler = {
1369         .handler        = ip4ip6_rcv,
1370         .err_handler    = ip4ip6_err,
1371         .priority       =       1,
1372 };
1373
1374 static struct xfrm6_tunnel ip6ip6_handler = {
1375         .handler        = ip6ip6_rcv,
1376         .err_handler    = ip6ip6_err,
1377         .priority       =       1,
1378 };
1379
1380 /**
1381  * ip6_tunnel_init - register protocol and reserve needed resources
1382  *
1383  * Return: 0 on success
1384  **/
1385
1386 static int __init ip6_tunnel_init(void)
1387 {
1388         int  err;
1389
1390         if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1391                 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1392                 err = -EAGAIN;
1393                 goto out;
1394         }
1395
1396         if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1397                 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1398                 err = -EAGAIN;
1399                 goto unreg_ip4ip6;
1400         }
1401         ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1402                                       ip6_tnl_dev_setup);
1403
1404         if (!ip6_fb_tnl_dev) {
1405                 err = -ENOMEM;
1406                 goto fail;
1407         }
1408         ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1409
1410         if ((err = register_netdev(ip6_fb_tnl_dev))) {
1411                 free_netdev(ip6_fb_tnl_dev);
1412                 goto fail;
1413         }
1414         return 0;
1415 fail:
1416         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1417 unreg_ip4ip6:
1418         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1419 out:
1420         return err;
1421 }
1422
1423 static void __exit ip6_tnl_destroy_tunnels(void)
1424 {
1425         int h;
1426         struct ip6_tnl *t;
1427
1428         for (h = 0; h < HASH_SIZE; h++) {
1429                 while ((t = tnls_r_l[h]) != NULL)
1430                         unregister_netdevice(t->dev);
1431         }
1432
1433         t = tnls_wc[0];
1434         unregister_netdevice(t->dev);
1435 }
1436
1437 /**
1438  * ip6_tunnel_cleanup - free resources and unregister protocol
1439  **/
1440
1441 static void __exit ip6_tunnel_cleanup(void)
1442 {
1443         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1444                 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1445
1446         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1447                 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1448
1449         rtnl_lock();
1450         ip6_tnl_destroy_tunnels();
1451         rtnl_unlock();
1452 }
1453
1454 module_init(ip6_tunnel_init);
1455 module_exit(ip6_tunnel_cleanup);