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