git.oblomov.eu Git - linux-2.6/blob - 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                 int i;
 236                 for (i = 1; i < IP6_TNL_MAX; i++) {
 237                         sprintf(name, "ip6tnl%d", i);
 238                         if (__dev_get_by_name(&init_net, name) == NULL)
 239                                 break;
 240                 }
 241                 if (i == IP6_TNL_MAX)
 242                         goto failed;
 243         }
 244         dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
 245         if (dev == NULL)
 246                 goto failed;
 247
 248         t = netdev_priv(dev);
 249         dev->init = ip6_tnl_dev_init;
 250         t->parms = *p;
 251
 252         if ((err = register_netdevice(dev)) < 0) {
 253                 free_netdev(dev);
 254                 goto failed;
 255         }
 256         dev_hold(dev);
 257         ip6_tnl_link(t);
 258         return t;
 259 failed:
 260         return NULL;
 261 }
 262
 263 /**
 264  * ip6_tnl_locate - find or create tunnel matching given parameters
 265  *   @p: tunnel parameters
 266  *   @create: != 0 if allowed to create new tunnel if no match found
 267  *
 268  * Description:
 269  *   ip6_tnl_locate() first tries to locate an existing tunnel
 270  *   based on @parms. If this is unsuccessful, but @create is set a new
 271  *   tunnel device is created and registered for use.
 272  *
 273  * Return:
 274  *   matching tunnel or NULL
 275  **/
 276
 277 static struct ip6_tnl *ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
 278 {
 279         struct in6_addr *remote = &p->raddr;
 280         struct in6_addr *local = &p->laddr;
 281         struct ip6_tnl *t;
 282
 283         for (t = *ip6_tnl_bucket(p); t; t = t->next) {
 284                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 285                     ipv6_addr_equal(remote, &t->parms.raddr))
 286                         return t;
 287         }
 288         if (!create)
 289                 return NULL;
 290         return ip6_tnl_create(p);
 291 }
 292
 293 /**
 294  * ip6_tnl_dev_uninit - tunnel device uninitializer
 295  *   @dev: the device to be destroyed
 296  *
 297  * Description:
 298  *   ip6_tnl_dev_uninit() removes tunnel from its list
 299  **/
 300
 301 static void
 302 ip6_tnl_dev_uninit(struct net_device *dev)
 303 {
 304         struct ip6_tnl *t = netdev_priv(dev);
 305
 306         if (dev == ip6_fb_tnl_dev) {
 307                 write_lock_bh(&ip6_tnl_lock);
 308                 tnls_wc[0] = NULL;
 309                 write_unlock_bh(&ip6_tnl_lock);
 310         } else {
 311                 ip6_tnl_unlink(t);
 312         }
 313         ip6_tnl_dst_reset(t);
 314         dev_put(dev);
 315 }
 316
 317 /**
 318  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 319  *   @skb: received socket buffer
 320  *
 321  * Return:
 322  *   0 if none was found,
 323  *   else index to encapsulation limit
 324  **/
 325
 326 static __u16
 327 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
 328 {
 329         struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
 330         __u8 nexthdr = ipv6h->nexthdr;
 331         __u16 off = sizeof (*ipv6h);
 332
 333         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 334                 __u16 optlen = 0;
 335                 struct ipv6_opt_hdr *hdr;
 336                 if (raw + off + sizeof (*hdr) > skb->data &&
 337                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 338                         break;
 339
 340                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 341                 if (nexthdr == NEXTHDR_FRAGMENT) {
 342                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 343                         if (frag_hdr->frag_off)
 344                                 break;
 345                         optlen = 8;
 346                 } else if (nexthdr == NEXTHDR_AUTH) {
 347                         optlen = (hdr->hdrlen + 2) << 2;
 348                 } else {
 349                         optlen = ipv6_optlen(hdr);
 350                 }
 351                 if (nexthdr == NEXTHDR_DEST) {
 352                         __u16 i = off + 2;
 353                         while (1) {
 354                                 struct ipv6_tlv_tnl_enc_lim *tel;
 355
 356                                 /* No more room for encapsulation limit */
 357                                 if (i + sizeof (*tel) > off + optlen)
 358                                         break;
 359
 360                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 361                                 /* return index of option if found and valid */
 362                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 363                                     tel->length == 1)
 364                                         return i;
 365                                 /* else jump to next option */
 366                                 if (tel->type)
 367                                         i += tel->length + 2;
 368                                 else
 369                                         i++;
 370                         }
 371                 }
 372                 nexthdr = hdr->nexthdr;
 373                 off += optlen;
 374         }
 375         return 0;
 376 }
 377
 378 /**
 379  * ip6_tnl_err - tunnel error handler
 380  *
 381  * Description:
 382  *   ip6_tnl_err() should handle errors in the tunnel according
 383  *   to the specifications in RFC 2473.
 384  **/
 385
 386 static int
 387 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 388             int *type, int *code, int *msg, __u32 *info, int offset)
 389 {
 390         struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
 391         struct ip6_tnl *t;
 392         int rel_msg = 0;
 393         int rel_type = ICMPV6_DEST_UNREACH;
 394         int rel_code = ICMPV6_ADDR_UNREACH;
 395         __u32 rel_info = 0;
 396         __u16 len;
 397         int err = -ENOENT;
 398
 399         /* If the packet doesn't contain the original IPv6 header we are
 400            in trouble since we might need the source address for further
 401            processing of the error. */
 402
 403         read_lock(&ip6_tnl_lock);
 404         if ((t = ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
 405                 goto out;
 406
 407         if (t->parms.proto != ipproto && t->parms.proto != 0)
 408                 goto out;
 409
 410         err = 0;
 411
 412         switch (*type) {
 413                 __u32 teli;
 414                 struct ipv6_tlv_tnl_enc_lim *tel;
 415                 __u32 mtu;
 416         case ICMPV6_DEST_UNREACH:
 417                 if (net_ratelimit())
 418                         printk(KERN_WARNING
 419                                "%s: Path to destination invalid "
 420                                "or inactive!\n", t->parms.name);
 421                 rel_msg = 1;
 422                 break;
 423         case ICMPV6_TIME_EXCEED:
 424                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 425                         if (net_ratelimit())
 426                                 printk(KERN_WARNING
 427                                        "%s: Too small hop limit or "
 428                                        "routing loop in tunnel!\n",
 429                                        t->parms.name);
 430                         rel_msg = 1;
 431                 }
 432                 break;
 433         case ICMPV6_PARAMPROB:
 434                 teli = 0;
 435                 if ((*code) == ICMPV6_HDR_FIELD)
 436                         teli = parse_tlv_tnl_enc_lim(skb, skb->data);
 437
 438                 if (teli && teli == *info - 2) {
 439                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 440                         if (tel->encap_limit == 0) {
 441                                 if (net_ratelimit())
 442                                         printk(KERN_WARNING
 443                                                "%s: Too small encapsulation "
 444                                                "limit or routing loop in "
 445                                                "tunnel!\n", t->parms.name);
 446                                 rel_msg = 1;
 447                         }
 448                 } else if (net_ratelimit()) {
 449                         printk(KERN_WARNING
 450                                "%s: Recipient unable to parse tunneled "
 451                                "packet!\n ", t->parms.name);
 452                 }
 453                 break;
 454         case ICMPV6_PKT_TOOBIG:
 455                 mtu = *info - offset;
 456                 if (mtu < IPV6_MIN_MTU)
 457                         mtu = IPV6_MIN_MTU;
 458                 t->dev->mtu = mtu;
 459
 460                 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
 461                         rel_type = ICMPV6_PKT_TOOBIG;
 462                         rel_code = 0;
 463                         rel_info = mtu;
 464                         rel_msg = 1;
 465                 }
 466                 break;
 467         }
 468
 469         *type = rel_type;
 470         *code = rel_code;
 471         *info = rel_info;
 472         *msg = rel_msg;
 473
 474 out:
 475         read_unlock(&ip6_tnl_lock);
 476         return err;
 477 }
 478
 479 static int
 480 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 481            int type, int code, int offset, __be32 info)
 482 {
 483         int rel_msg = 0;
 484         int rel_type = type;
 485         int rel_code = code;
 486         __u32 rel_info = ntohl(info);
 487         int err;
 488         struct sk_buff *skb2;
 489         struct iphdr *eiph;
 490         struct flowi fl;
 491         struct rtable *rt;
 492
 493         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 494                           &rel_msg, &rel_info, offset);
 495         if (err < 0)
 496                 return err;
 497
 498         if (rel_msg == 0)
 499                 return 0;
 500
 501         switch (rel_type) {
 502         case ICMPV6_DEST_UNREACH:
 503                 if (rel_code != ICMPV6_ADDR_UNREACH)
 504                         return 0;
 505                 rel_type = ICMP_DEST_UNREACH;
 506                 rel_code = ICMP_HOST_UNREACH;
 507                 break;
 508         case ICMPV6_PKT_TOOBIG:
 509                 if (rel_code != 0)
 510                         return 0;
 511                 rel_type = ICMP_DEST_UNREACH;
 512                 rel_code = ICMP_FRAG_NEEDED;
 513                 break;
 514         default:
 515                 return 0;
 516         }
 517
 518         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 519                 return 0;
 520
 521         skb2 = skb_clone(skb, GFP_ATOMIC);
 522         if (!skb2)
 523                 return 0;
 524
 525         dst_release(skb2->dst);
 526         skb2->dst = NULL;
 527         skb_pull(skb2, offset);
 528         skb_reset_network_header(skb2);
 529         eiph = ip_hdr(skb2);
 530
 531         /* Try to guess incoming interface */
 532         memset(&fl, 0, sizeof(fl));
 533         fl.fl4_dst = eiph->saddr;
 534         fl.fl4_tos = RT_TOS(eiph->tos);
 535         fl.proto = IPPROTO_IPIP;
 536         if (ip_route_output_key(&rt, &fl))
 537                 goto out;
 538
 539         skb2->dev = rt->u.dst.dev;
 540
 541         /* route "incoming" packet */
 542         if (rt->rt_flags & RTCF_LOCAL) {
 543                 ip_rt_put(rt);
 544                 rt = NULL;
 545                 fl.fl4_dst = eiph->daddr;
 546                 fl.fl4_src = eiph->saddr;
 547                 fl.fl4_tos = eiph->tos;
 548                 if (ip_route_output_key(&rt, &fl) ||
 549                     rt->u.dst.dev->type != ARPHRD_TUNNEL) {
 550                         ip_rt_put(rt);
 551                         goto out;
 552                 }
 553         } else {
 554                 ip_rt_put(rt);
 555                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 556                                    skb2->dev) ||
 557                     skb2->dst->dev->type != ARPHRD_TUNNEL)
 558                         goto out;
 559         }
 560
 561         /* change mtu on this route */
 562         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 563                 if (rel_info > dst_mtu(skb2->dst))
 564                         goto out;
 565
 566                 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
 567         }
 568
 569         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 570
 571 out:
 572         kfree_skb(skb2);
 573         return 0;
 574 }
 575
 576 static int
 577 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 578            int type, int code, int offset, __be32 info)
 579 {
 580         int rel_msg = 0;
 581         int rel_type = type;
 582         int rel_code = code;
 583         __u32 rel_info = ntohl(info);
 584         int err;
 585
 586         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 587                           &rel_msg, &rel_info, offset);
 588         if (err < 0)
 589                 return err;
 590
 591         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 592                 struct rt6_info *rt;
 593                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 594
 595                 if (!skb2)
 596                         return 0;
 597
 598                 dst_release(skb2->dst);
 599                 skb2->dst = NULL;
 600                 skb_pull(skb2, offset);
 601                 skb_reset_network_header(skb2);
 602
 603                 /* Try to guess incoming interface */
 604                 rt = rt6_lookup(&ipv6_hdr(skb2)->saddr, NULL, 0, 0);
 605
 606                 if (rt && rt->rt6i_dev)
 607                         skb2->dev = rt->rt6i_dev;
 608
 609                 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 610
 611                 if (rt)
 612                         dst_release(&rt->u.dst);
 613
 614                 kfree_skb(skb2);
 615         }
 616
 617         return 0;
 618 }
 619
 620 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
 621                                         struct ipv6hdr *ipv6h,
 622                                         struct sk_buff *skb)
 623 {
 624         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 625
 626         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 627                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 628
 629         if (INET_ECN_is_ce(dsfield))
 630                 IP_ECN_set_ce(ip_hdr(skb));
 631 }
 632
 633 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
 634                                         struct ipv6hdr *ipv6h,
 635                                         struct sk_buff *skb)
 636 {
 637         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 638                 ipv6_copy_dscp(ipv6h, ipv6_hdr(skb));
 639
 640         if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
 641                 IP6_ECN_set_ce(ipv6_hdr(skb));
 642 }
 643
 644 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
 645 {
 646         struct ip6_tnl_parm *p = &t->parms;
 647         int ret = 0;
 648
 649         if (p->flags & IP6_TNL_F_CAP_RCV) {
 650                 struct net_device *ldev = NULL;
 651
 652                 if (p->link)
 653                         ldev = dev_get_by_index(&init_net, p->link);
 654
 655                 if ((ipv6_addr_is_multicast(&p->laddr) ||
 656                      likely(ipv6_chk_addr(&p->laddr, ldev, 0))) &&
 657                     likely(!ipv6_chk_addr(&p->raddr, NULL, 0)))
 658                         ret = 1;
 659
 660                 if (ldev)
 661                         dev_put(ldev);
 662         }
 663         return ret;
 664 }
 665
 666 /**
 667  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 668  *   @skb: received socket buffer
 669  *   @protocol: ethernet protocol ID
 670  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 671  *
 672  * Return: 0
 673  **/
 674
 675 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 676                        __u8 ipproto,
 677                        void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
 678                                                     struct ipv6hdr *ipv6h,
 679                                                     struct sk_buff *skb))
 680 {
 681         struct ip6_tnl *t;
 682         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 683
 684         read_lock(&ip6_tnl_lock);
 685
 686         if ((t = ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
 687                 if (t->parms.proto != ipproto && t->parms.proto != 0) {
 688                         read_unlock(&ip6_tnl_lock);
 689                         goto discard;
 690                 }
 691
 692                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 693                         read_unlock(&ip6_tnl_lock);
 694                         goto discard;
 695                 }
 696
 697                 if (!ip6_tnl_rcv_ctl(t)) {
 698                         t->stat.rx_dropped++;
 699                         read_unlock(&ip6_tnl_lock);
 700                         goto discard;
 701                 }
 702                 secpath_reset(skb);
 703                 skb->mac_header = skb->network_header;
 704                 skb_reset_network_header(skb);
 705                 skb->protocol = htons(protocol);
 706                 skb->pkt_type = PACKET_HOST;
 707                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 708                 skb->dev = t->dev;
 709                 dst_release(skb->dst);
 710                 skb->dst = NULL;
 711                 nf_reset(skb);
 712
 713                 dscp_ecn_decapsulate(t, ipv6h, skb);
 714
 715                 t->stat.rx_packets++;
 716                 t->stat.rx_bytes += skb->len;
 717                 netif_rx(skb);
 718                 read_unlock(&ip6_tnl_lock);
 719                 return 0;
 720         }
 721         read_unlock(&ip6_tnl_lock);
 722         return 1;
 723
 724 discard:
 725         kfree_skb(skb);
 726         return 0;
 727 }
 728
 729 static int ip4ip6_rcv(struct sk_buff *skb)
 730 {
 731         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 732                            ip4ip6_dscp_ecn_decapsulate);
 733 }
 734
 735 static int ip6ip6_rcv(struct sk_buff *skb)
 736 {
 737         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 738                            ip6ip6_dscp_ecn_decapsulate);
 739 }
 740
 741 struct ipv6_tel_txoption {
 742         struct ipv6_txoptions ops;
 743         __u8 dst_opt[8];
 744 };
 745
 746 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 747 {
 748         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 749
 750         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 751         opt->dst_opt[3] = 1;
 752         opt->dst_opt[4] = encap_limit;
 753         opt->dst_opt[5] = IPV6_TLV_PADN;
 754         opt->dst_opt[6] = 1;
 755
 756         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 757         opt->ops.opt_nflen = 8;
 758 }
 759
 760 /**
 761  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 762  *   @t: the outgoing tunnel device
 763  *   @hdr: IPv6 header from the incoming packet
 764  *
 765  * Description:
 766  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 767  *   doesn't match source of incoming packet.
 768  *
 769  * Return:
 770  *   1 if conflict,
 771  *   0 else
 772  **/
 773
 774 static inline int
 775 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
 776 {
 777         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 778 }
 779
 780 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
 781 {
 782         struct ip6_tnl_parm *p = &t->parms;
 783         int ret = 0;
 784
 785         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 786                 struct net_device *ldev = NULL;
 787
 788                 if (p->link)
 789                         ldev = dev_get_by_index(&init_net, p->link);
 790
 791                 if (unlikely(!ipv6_chk_addr(&p->laddr, ldev, 0)))
 792                         printk(KERN_WARNING
 793                                "%s xmit: Local address not yet configured!\n",
 794                                p->name);
 795                 else if (!ipv6_addr_is_multicast(&p->raddr) &&
 796                          unlikely(ipv6_chk_addr(&p->raddr, NULL, 0)))
 797                         printk(KERN_WARNING
 798                                "%s xmit: Routing loop! "
 799                                "Remote address found on this node!\n",
 800                                p->name);
 801                 else
 802                         ret = 1;
 803                 if (ldev)
 804                         dev_put(ldev);
 805         }
 806         return ret;
 807 }
 808 /**
 809  * ip6_tnl_xmit2 - encapsulate packet and send
 810  *   @skb: the outgoing socket buffer
 811  *   @dev: the outgoing tunnel device
 812  *   @dsfield: dscp code for outer header
 813  *   @fl: flow of tunneled packet
 814  *   @encap_limit: encapsulation limit
 815  *   @pmtu: Path MTU is stored if packet is too big
 816  *
 817  * Description:
 818  *   Build new header and do some sanity checks on the packet before sending
 819  *   it.
 820  *
 821  * Return:
 822  *   0 on success
 823  *   -1 fail
 824  *   %-EMSGSIZE message too big. return mtu in this case.
 825  **/
 826
 827 static int ip6_tnl_xmit2(struct sk_buff *skb,
 828                          struct net_device *dev,
 829                          __u8 dsfield,
 830                          struct flowi *fl,
 831                          int encap_limit,
 832                          __u32 *pmtu)
 833 {
 834         struct ip6_tnl *t = netdev_priv(dev);
 835         struct net_device_stats *stats = &t->stat;
 836         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 837         struct ipv6_tel_txoption opt;
 838         struct dst_entry *dst;
 839         struct net_device *tdev;
 840         int mtu;
 841         int max_headroom = sizeof(struct ipv6hdr);
 842         u8 proto;
 843         int err = -1;
 844         int pkt_len;
 845
 846         if ((dst = ip6_tnl_dst_check(t)) != NULL)
 847                 dst_hold(dst);
 848         else {
 849                 dst = ip6_route_output(NULL, fl);
 850
 851                 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
 852                         goto tx_err_link_failure;
 853         }
 854
 855         tdev = dst->dev;
 856
 857         if (tdev == dev) {
 858                 stats->collisions++;
 859                 if (net_ratelimit())
 860                         printk(KERN_WARNING
 861                                "%s: Local routing loop detected!\n",
 862                                t->parms.name);
 863                 goto tx_err_dst_release;
 864         }
 865         mtu = dst_mtu(dst) - sizeof (*ipv6h);
 866         if (encap_limit >= 0) {
 867                 max_headroom += 8;
 868                 mtu -= 8;
 869         }
 870         if (mtu < IPV6_MIN_MTU)
 871                 mtu = IPV6_MIN_MTU;
 872         if (skb->dst)
 873                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 874         if (skb->len > mtu) {
 875                 *pmtu = mtu;
 876                 err = -EMSGSIZE;
 877                 goto tx_err_dst_release;
 878         }
 879
 880         /*
 881          * Okay, now see if we can stuff it in the buffer as-is.
 882          */
 883         max_headroom += LL_RESERVED_SPACE(tdev);
 884
 885         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 886             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 887                 struct sk_buff *new_skb;
 888
 889                 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
 890                         goto tx_err_dst_release;
 891
 892                 if (skb->sk)
 893                         skb_set_owner_w(new_skb, skb->sk);
 894                 kfree_skb(skb);
 895                 skb = new_skb;
 896         }
 897         dst_release(skb->dst);
 898         skb->dst = dst_clone(dst);
 899
 900         skb->transport_header = skb->network_header;
 901
 902         proto = fl->proto;
 903         if (encap_limit >= 0) {
 904                 init_tel_txopt(&opt, encap_limit);
 905                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
 906         }
 907         skb_push(skb, sizeof(struct ipv6hdr));
 908         skb_reset_network_header(skb);
 909         ipv6h = ipv6_hdr(skb);
 910         *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
 911         dsfield = INET_ECN_encapsulate(0, dsfield);
 912         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 913         ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
 914         ipv6h->hop_limit = t->parms.hop_limit;
 915         ipv6h->nexthdr = proto;
 916         ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
 917         ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
 918         nf_reset(skb);
 919         pkt_len = skb->len;
 920         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
 921                       skb->dst->dev, dst_output);
 922
 923         if (net_xmit_eval(err) == 0) {
 924                 stats->tx_bytes += pkt_len;
 925                 stats->tx_packets++;
 926         } else {
 927                 stats->tx_errors++;
 928                 stats->tx_aborted_errors++;
 929         }
 930         ip6_tnl_dst_store(t, dst);
 931         return 0;
 932 tx_err_link_failure:
 933         stats->tx_carrier_errors++;
 934         dst_link_failure(skb);
 935 tx_err_dst_release:
 936         dst_release(dst);
 937         return err;
 938 }
 939
 940 static inline int
 941 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 942 {
 943         struct ip6_tnl *t = netdev_priv(dev);
 944         struct iphdr  *iph = ip_hdr(skb);
 945         int encap_limit = -1;
 946         struct flowi fl;
 947         __u8 dsfield;
 948         __u32 mtu;
 949         int err;
 950
 951         if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
 952             !ip6_tnl_xmit_ctl(t))
 953                 return -1;
 954
 955         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
 956                 encap_limit = t->parms.encap_limit;
 957
 958         memcpy(&fl, &t->fl, sizeof (fl));
 959         fl.proto = IPPROTO_IPIP;
 960
 961         dsfield = ipv4_get_dsfield(iph);
 962
 963         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
 964                 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 965                                           & IPV6_TCLASS_MASK;
 966
 967         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
 968         if (err != 0) {
 969                 /* XXX: send ICMP error even if DF is not set. */
 970                 if (err == -EMSGSIZE)
 971                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
 972                                   htonl(mtu));
 973                 return -1;
 974         }
 975
 976         return 0;
 977 }
 978
 979 static inline int
 980 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 981 {
 982         struct ip6_tnl *t = netdev_priv(dev);
 983         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 984         int encap_limit = -1;
 985         __u16 offset;
 986         struct flowi fl;
 987         __u8 dsfield;
 988         __u32 mtu;
 989         int err;
 990
 991         if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
 992             !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
 993                 return -1;
 994
 995         offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
 996         if (offset > 0) {
 997                 struct ipv6_tlv_tnl_enc_lim *tel;
 998                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
 999                 if (tel->encap_limit == 0) {
1000                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1001                                     ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1002                         return -1;
1003                 }
1004                 encap_limit = tel->encap_limit - 1;
1005         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1006                 encap_limit = t->parms.encap_limit;
1007
1008         memcpy(&fl, &t->fl, sizeof (fl));
1009         fl.proto = IPPROTO_IPV6;
1010
1011         dsfield = ipv6_get_dsfield(ipv6h);
1012         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1013                 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1014         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1015                 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1016
1017         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1018         if (err != 0) {
1019                 if (err == -EMSGSIZE)
1020                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1021                 return -1;
1022         }
1023
1024         return 0;
1025 }
1026
1027 static int
1028 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1029 {
1030         struct ip6_tnl *t = netdev_priv(dev);
1031         struct net_device_stats *stats = &t->stat;
1032         int ret;
1033
1034         if (t->recursion++) {
1035                 t->stat.collisions++;
1036                 goto tx_err;
1037         }
1038
1039         switch (skb->protocol) {
1040         case __constant_htons(ETH_P_IP):
1041                 ret = ip4ip6_tnl_xmit(skb, dev);
1042                 break;
1043         case __constant_htons(ETH_P_IPV6):
1044                 ret = ip6ip6_tnl_xmit(skb, dev);
1045                 break;
1046         default:
1047                 goto tx_err;
1048         }
1049
1050         if (ret < 0)
1051                 goto tx_err;
1052
1053         t->recursion--;
1054         return 0;
1055
1056 tx_err:
1057         stats->tx_errors++;
1058         stats->tx_dropped++;
1059         kfree_skb(skb);
1060         t->recursion--;
1061         return 0;
1062 }
1063
1064 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1065 {
1066         struct ip6_tnl_parm *p = &t->parms;
1067         int ltype = ipv6_addr_type(&p->laddr);
1068         int rtype = ipv6_addr_type(&p->raddr);
1069
1070         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1071
1072         if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1073             rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1074             !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1075             (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1076                 if (ltype&IPV6_ADDR_UNICAST)
1077                         p->flags |= IP6_TNL_F_CAP_XMIT;
1078                 if (rtype&IPV6_ADDR_UNICAST)
1079                         p->flags |= IP6_TNL_F_CAP_RCV;
1080         }
1081 }
1082
1083 static void ip6_tnl_link_config(struct ip6_tnl *t)
1084 {
1085         struct net_device *dev = t->dev;
1086         struct ip6_tnl_parm *p = &t->parms;
1087         struct flowi *fl = &t->fl;
1088
1089         memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1090         memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1091
1092         /* Set up flowi template */
1093         ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1094         ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1095         fl->oif = p->link;
1096         fl->fl6_flowlabel = 0;
1097
1098         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1099                 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1100         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1101                 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1102
1103         ip6_tnl_set_cap(t);
1104
1105         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1106                 dev->flags |= IFF_POINTOPOINT;
1107         else
1108                 dev->flags &= ~IFF_POINTOPOINT;
1109
1110         dev->iflink = p->link;
1111
1112         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1113                 int strict = (ipv6_addr_type(&p->raddr) &
1114                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1115
1116                 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
1117                                                  p->link, strict);
1118
1119                 if (rt == NULL)
1120                         return;
1121
1122                 if (rt->rt6i_dev) {
1123                         dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1124                                 sizeof (struct ipv6hdr);
1125
1126                         dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1127
1128                         if (dev->mtu < IPV6_MIN_MTU)
1129                                 dev->mtu = IPV6_MIN_MTU;
1130                 }
1131                 dst_release(&rt->u.dst);
1132         }
1133 }
1134
1135 /**
1136  * ip6_tnl_change - update the tunnel parameters
1137  *   @t: tunnel to be changed
1138  *   @p: tunnel configuration parameters
1139  *   @active: != 0 if tunnel is ready for use
1140  *
1141  * Description:
1142  *   ip6_tnl_change() updates the tunnel parameters
1143  **/
1144
1145 static int
1146 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1147 {
1148         ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1149         ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1150         t->parms.flags = p->flags;
1151         t->parms.hop_limit = p->hop_limit;
1152         t->parms.encap_limit = p->encap_limit;
1153         t->parms.flowinfo = p->flowinfo;
1154         t->parms.link = p->link;
1155         t->parms.proto = p->proto;
1156         ip6_tnl_dst_reset(t);
1157         ip6_tnl_link_config(t);
1158         return 0;
1159 }
1160
1161 /**
1162  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1163  *   @dev: virtual device associated with tunnel
1164  *   @ifr: parameters passed from userspace
1165  *   @cmd: command to be performed
1166  *
1167  * Description:
1168  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1169  *   from userspace.
1170  *
1171  *   The possible commands are the following:
1172  *     %SIOCGETTUNNEL: get tunnel parameters for device
1173  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1174  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1175  *     %SIOCDELTUNNEL: delete tunnel
1176  *
1177  *   The fallback device "ip6tnl0", created during module
1178  *   initialization, can be used for creating other tunnel devices.
1179  *
1180  * Return:
1181  *   0 on success,
1182  *   %-EFAULT if unable to copy data to or from userspace,
1183  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1184  *   %-EINVAL if passed tunnel parameters are invalid,
1185  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1186  *   %-ENODEV if attempting to change or delete a nonexisting device
1187  **/
1188
1189 static int
1190 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1191 {
1192         int err = 0;
1193         struct ip6_tnl_parm p;
1194         struct ip6_tnl *t = NULL;
1195
1196         switch (cmd) {
1197         case SIOCGETTUNNEL:
1198                 if (dev == ip6_fb_tnl_dev) {
1199                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1200                                 err = -EFAULT;
1201                                 break;
1202                         }
1203                         t = ip6_tnl_locate(&p, 0);
1204                 }
1205                 if (t == NULL)
1206                         t = netdev_priv(dev);
1207                 memcpy(&p, &t->parms, sizeof (p));
1208                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1209                         err = -EFAULT;
1210                 }
1211                 break;
1212         case SIOCADDTUNNEL:
1213         case SIOCCHGTUNNEL:
1214                 err = -EPERM;
1215                 if (!capable(CAP_NET_ADMIN))
1216                         break;
1217                 err = -EFAULT;
1218                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1219                         break;
1220                 err = -EINVAL;
1221                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1222                     p.proto != 0)
1223                         break;
1224                 t = ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
1225                 if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1226                         if (t != NULL) {
1227                                 if (t->dev != dev) {
1228                                         err = -EEXIST;
1229                                         break;
1230                                 }
1231                         } else
1232                                 t = netdev_priv(dev);
1233
1234                         ip6_tnl_unlink(t);
1235                         err = ip6_tnl_change(t, &p);
1236                         ip6_tnl_link(t);
1237                         netdev_state_change(dev);
1238                 }
1239                 if (t) {
1240                         err = 0;
1241                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1242                                 err = -EFAULT;
1243
1244                 } else
1245                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1246                 break;
1247         case SIOCDELTUNNEL:
1248                 err = -EPERM;
1249                 if (!capable(CAP_NET_ADMIN))
1250                         break;
1251
1252                 if (dev == ip6_fb_tnl_dev) {
1253                         err = -EFAULT;
1254                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1255                                 break;
1256                         err = -ENOENT;
1257                         if ((t = ip6_tnl_locate(&p, 0)) == NULL)
1258                                 break;
1259                         err = -EPERM;
1260                         if (t->dev == ip6_fb_tnl_dev)
1261                                 break;
1262                         dev = t->dev;
1263                 }
1264                 err = 0;
1265                 unregister_netdevice(dev);
1266                 break;
1267         default:
1268                 err = -EINVAL;
1269         }
1270         return err;
1271 }
1272
1273 /**
1274  * ip6_tnl_get_stats - return the stats for tunnel device
1275  *   @dev: virtual device associated with tunnel
1276  *
1277  * Return: stats for device
1278  **/
1279
1280 static struct net_device_stats *
1281 ip6_tnl_get_stats(struct net_device *dev)
1282 {
1283         return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1284 }
1285
1286 /**
1287  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1288  *   @dev: virtual device associated with tunnel
1289  *   @new_mtu: the new mtu
1290  *
1291  * Return:
1292  *   0 on success,
1293  *   %-EINVAL if mtu too small
1294  **/
1295
1296 static int
1297 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1298 {
1299         if (new_mtu < IPV6_MIN_MTU) {
1300                 return -EINVAL;
1301         }
1302         dev->mtu = new_mtu;
1303         return 0;
1304 }
1305
1306 /**
1307  * ip6_tnl_dev_setup - setup virtual tunnel device
1308  *   @dev: virtual device associated with tunnel
1309  *
1310  * Description:
1311  *   Initialize function pointers and device parameters
1312  **/
1313
1314 static void ip6_tnl_dev_setup(struct net_device *dev)
1315 {
1316         dev->uninit = ip6_tnl_dev_uninit;
1317         dev->destructor = free_netdev;
1318         dev->hard_start_xmit = ip6_tnl_xmit;
1319         dev->get_stats = ip6_tnl_get_stats;
1320         dev->do_ioctl = ip6_tnl_ioctl;
1321         dev->change_mtu = ip6_tnl_change_mtu;
1322
1323         dev->type = ARPHRD_TUNNEL6;
1324         dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1325         dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1326         dev->flags |= IFF_NOARP;
1327         dev->addr_len = sizeof(struct in6_addr);
1328 }
1329
1330
1331 /**
1332  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1333  *   @dev: virtual device associated with tunnel
1334  **/
1335
1336 static inline void
1337 ip6_tnl_dev_init_gen(struct net_device *dev)
1338 {
1339         struct ip6_tnl *t = netdev_priv(dev);
1340         t->dev = dev;
1341         strcpy(t->parms.name, dev->name);
1342 }
1343
1344 /**
1345  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1346  *   @dev: virtual device associated with tunnel
1347  **/
1348
1349 static int
1350 ip6_tnl_dev_init(struct net_device *dev)
1351 {
1352         struct ip6_tnl *t = netdev_priv(dev);
1353         ip6_tnl_dev_init_gen(dev);
1354         ip6_tnl_link_config(t);
1355         return 0;
1356 }
1357
1358 /**
1359  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1360  *   @dev: fallback device
1361  *
1362  * Return: 0
1363  **/
1364
1365 static int
1366 ip6_fb_tnl_dev_init(struct net_device *dev)
1367 {
1368         struct ip6_tnl *t = netdev_priv(dev);
1369         ip6_tnl_dev_init_gen(dev);
1370         t->parms.proto = IPPROTO_IPV6;
1371         dev_hold(dev);
1372         tnls_wc[0] = t;
1373         return 0;
1374 }
1375
1376 static struct xfrm6_tunnel ip4ip6_handler = {
1377         .handler        = ip4ip6_rcv,
1378         .err_handler    = ip4ip6_err,
1379         .priority       =       1,
1380 };
1381
1382 static struct xfrm6_tunnel ip6ip6_handler = {
1383         .handler        = ip6ip6_rcv,
1384         .err_handler    = ip6ip6_err,
1385         .priority       =       1,
1386 };
1387
1388 /**
1389  * ip6_tunnel_init - register protocol and reserve needed resources
1390  *
1391  * Return: 0 on success
1392  **/
1393
1394 static int __init ip6_tunnel_init(void)
1395 {
1396         int  err;
1397
1398         if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1399                 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1400                 err = -EAGAIN;
1401                 goto out;
1402         }
1403
1404         if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1405                 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1406                 err = -EAGAIN;
1407                 goto unreg_ip4ip6;
1408         }
1409         ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1410                                       ip6_tnl_dev_setup);
1411
1412         if (!ip6_fb_tnl_dev) {
1413                 err = -ENOMEM;
1414                 goto fail;
1415         }
1416         ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1417
1418         if ((err = register_netdev(ip6_fb_tnl_dev))) {
1419                 free_netdev(ip6_fb_tnl_dev);
1420                 goto fail;
1421         }
1422         return 0;
1423 fail:
1424         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1425 unreg_ip4ip6:
1426         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1427 out:
1428         return err;
1429 }
1430
1431 static void __exit ip6_tnl_destroy_tunnels(void)
1432 {
1433         int h;
1434         struct ip6_tnl *t;
1435
1436         for (h = 0; h < HASH_SIZE; h++) {
1437                 while ((t = tnls_r_l[h]) != NULL)
1438                         unregister_netdevice(t->dev);
1439         }
1440
1441         t = tnls_wc[0];
1442         unregister_netdevice(t->dev);
1443 }
1444
1445 /**
1446  * ip6_tunnel_cleanup - free resources and unregister protocol
1447  **/
1448
1449 static void __exit ip6_tunnel_cleanup(void)
1450 {
1451         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1452                 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1453
1454         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1455                 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1456
1457         rtnl_lock();
1458         ip6_tnl_destroy_tunnels();
1459         rtnl_unlock();
1460 }
1461
1462 module_init(ip6_tunnel_init);
1463 module_exit(ip6_tunnel_cleanup);