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  *      Based on:
  10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  11  *
  12  *      RFC 2473
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  */
  20
  21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  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/net.h>
  33 #include <linux/in6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/icmpv6.h>
  37 #include <linux/init.h>
  38 #include <linux/route.h>
  39 #include <linux/rtnetlink.h>
  40 #include <linux/netfilter_ipv6.h>
  41 #include <linux/slab.h>
  42 #include <linux/hash.h>
  43 #include <linux/etherdevice.h>
  44
  45 #include <asm/uaccess.h>
  46 #include <linux/atomic.h>
  47
  48 #include <net/icmp.h>
  49 #include <net/ip.h>
  50 #include <net/ip_tunnels.h>
  51 #include <net/ipv6.h>
  52 #include <net/ip6_route.h>
  53 #include <net/addrconf.h>
  54 #include <net/ip6_tunnel.h>
  55 #include <net/xfrm.h>
  56 #include <net/dsfield.h>
  57 #include <net/inet_ecn.h>
  58 #include <net/net_namespace.h>
  59 #include <net/netns/generic.h>
  60
  61 MODULE_AUTHOR("Ville Nuorvala");
  62 MODULE_DESCRIPTION("IPv6 tunneling device");
  63 MODULE_LICENSE("GPL");
  64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  65 MODULE_ALIAS_NETDEV("ip6tnl0");
  66
  67 #define HASH_SIZE_SHIFT  5
  68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  69
  70 static bool log_ecn_error = true;
  71 module_param(log_ecn_error, bool, 0644);
  72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  73
  74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  75 {
  76         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  77
  78         return hash_32(hash, HASH_SIZE_SHIFT);
  79 }
  80
  81 static int ip6_tnl_dev_init(struct net_device *dev);
  82 static void ip6_tnl_dev_setup(struct net_device *dev);
  83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  84
  85 static int ip6_tnl_net_id __read_mostly;
  86 struct ip6_tnl_net {
  87         /* the IPv6 tunnel fallback device */
  88         struct net_device *fb_tnl_dev;
  89         /* lists for storing tunnels in use */
  90         struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
  91         struct ip6_tnl __rcu *tnls_wc[1];
  92         struct ip6_tnl __rcu **tnls[2];
  93 };
  94
  95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  96 {
  97         struct pcpu_sw_netstats tmp, sum = { 0 };
  98         int i;
  99
 100         for_each_possible_cpu(i) {
 101                 unsigned int start;
 102                 const struct pcpu_sw_netstats *tstats =
 103                                                    per_cpu_ptr(dev->tstats, i);
 104
 105                 do {
 106                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 107                         tmp.rx_packets = tstats->rx_packets;
 108                         tmp.rx_bytes = tstats->rx_bytes;
 109                         tmp.tx_packets = tstats->tx_packets;
 110                         tmp.tx_bytes =  tstats->tx_bytes;
 111                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 112
 113                 sum.rx_packets += tmp.rx_packets;
 114                 sum.rx_bytes   += tmp.rx_bytes;
 115                 sum.tx_packets += tmp.tx_packets;
 116                 sum.tx_bytes   += tmp.tx_bytes;
 117         }
 118         dev->stats.rx_packets = sum.rx_packets;
 119         dev->stats.rx_bytes   = sum.rx_bytes;
 120         dev->stats.tx_packets = sum.tx_packets;
 121         dev->stats.tx_bytes   = sum.tx_bytes;
 122         return &dev->stats;
 123 }
 124
 125 /*
 126  * Locking : hash tables are protected by RCU and RTNL
 127  */
 128
 129 static void ip6_tnl_per_cpu_dst_set(struct ip6_tnl_dst *idst,
 130                                     struct dst_entry *dst)
 131 {
 132         write_seqlock_bh(&idst->lock);
 133         dst_release(rcu_dereference_protected(
 134                             idst->dst,
 135                             lockdep_is_held(&idst->lock.lock)));
 136         if (dst) {
 137                 dst_hold(dst);
 138                 idst->cookie = rt6_get_cookie((struct rt6_info *)dst);
 139         } else {
 140                 idst->cookie = 0;
 141         }
 142         rcu_assign_pointer(idst->dst, dst);
 143         write_sequnlock_bh(&idst->lock);
 144 }
 145
 146 struct dst_entry *ip6_tnl_dst_get(struct ip6_tnl *t)
 147 {
 148         struct ip6_tnl_dst *idst;
 149         struct dst_entry *dst;
 150         unsigned int seq;
 151         u32 cookie;
 152
 153         idst = raw_cpu_ptr(t->dst_cache);
 154
 155         rcu_read_lock();
 156         do {
 157                 seq = read_seqbegin(&idst->lock);
 158                 dst = rcu_dereference(idst->dst);
 159                 cookie = idst->cookie;
 160         } while (read_seqretry(&idst->lock, seq));
 161
 162         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
 163                 dst = NULL;
 164         rcu_read_unlock();
 165
 166         if (dst && dst->obsolete && !dst->ops->check(dst, cookie)) {
 167                 ip6_tnl_per_cpu_dst_set(idst, NULL);
 168                 dst_release(dst);
 169                 dst = NULL;
 170         }
 171         return dst;
 172 }
 173 EXPORT_SYMBOL_GPL(ip6_tnl_dst_get);
 174
 175 void ip6_tnl_dst_reset(struct ip6_tnl *t)
 176 {
 177         int i;
 178
 179         for_each_possible_cpu(i)
 180                 ip6_tnl_per_cpu_dst_set(per_cpu_ptr(t->dst_cache, i), NULL);
 181 }
 182 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
 183
 184 void ip6_tnl_dst_set(struct ip6_tnl *t, struct dst_entry *dst)
 185 {
 186         ip6_tnl_per_cpu_dst_set(raw_cpu_ptr(t->dst_cache), dst);
 187
 188 }
 189 EXPORT_SYMBOL_GPL(ip6_tnl_dst_set);
 190
 191 void ip6_tnl_dst_destroy(struct ip6_tnl *t)
 192 {
 193         if (!t->dst_cache)
 194                 return;
 195
 196         ip6_tnl_dst_reset(t);
 197         free_percpu(t->dst_cache);
 198 }
 199 EXPORT_SYMBOL_GPL(ip6_tnl_dst_destroy);
 200
 201 int ip6_tnl_dst_init(struct ip6_tnl *t)
 202 {
 203         int i;
 204
 205         t->dst_cache = alloc_percpu(struct ip6_tnl_dst);
 206         if (!t->dst_cache)
 207                 return -ENOMEM;
 208
 209         for_each_possible_cpu(i)
 210                 seqlock_init(&per_cpu_ptr(t->dst_cache, i)->lock);
 211
 212         return 0;
 213 }
 214 EXPORT_SYMBOL_GPL(ip6_tnl_dst_init);
 215
 216 /**
 217  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 218  *   @remote: the address of the tunnel exit-point
 219  *   @local: the address of the tunnel entry-point
 220  *
 221  * Return:
 222  *   tunnel matching given end-points if found,
 223  *   else fallback tunnel if its device is up,
 224  *   else %NULL
 225  **/
 226
 227 #define for_each_ip6_tunnel_rcu(start) \
 228         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 229
 230 static struct ip6_tnl *
 231 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 232 {
 233         unsigned int hash = HASH(remote, local);
 234         struct ip6_tnl *t;
 235         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 236         struct in6_addr any;
 237
 238         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 239                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 240                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 241                     (t->dev->flags & IFF_UP))
 242                         return t;
 243         }
 244
 245         memset(&any, 0, sizeof(any));
 246         hash = HASH(&any, local);
 247         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 248                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 249                     (t->dev->flags & IFF_UP))
 250                         return t;
 251         }
 252
 253         hash = HASH(remote, &any);
 254         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 255                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 256                     (t->dev->flags & IFF_UP))
 257                         return t;
 258         }
 259
 260         t = rcu_dereference(ip6n->tnls_wc[0]);
 261         if (t && (t->dev->flags & IFF_UP))
 262                 return t;
 263
 264         return NULL;
 265 }
 266
 267 /**
 268  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 269  *   @p: parameters containing tunnel end-points
 270  *
 271  * Description:
 272  *   ip6_tnl_bucket() returns the head of the list matching the
 273  *   &struct in6_addr entries laddr and raddr in @p.
 274  *
 275  * Return: head of IPv6 tunnel list
 276  **/
 277
 278 static struct ip6_tnl __rcu **
 279 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 280 {
 281         const struct in6_addr *remote = &p->raddr;
 282         const struct in6_addr *local = &p->laddr;
 283         unsigned int h = 0;
 284         int prio = 0;
 285
 286         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 287                 prio = 1;
 288                 h = HASH(remote, local);
 289         }
 290         return &ip6n->tnls[prio][h];
 291 }
 292
 293 /**
 294  * ip6_tnl_link - add tunnel to hash table
 295  *   @t: tunnel to be added
 296  **/
 297
 298 static void
 299 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 300 {
 301         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 302
 303         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 304         rcu_assign_pointer(*tp, t);
 305 }
 306
 307 /**
 308  * ip6_tnl_unlink - remove tunnel from hash table
 309  *   @t: tunnel to be removed
 310  **/
 311
 312 static void
 313 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 314 {
 315         struct ip6_tnl __rcu **tp;
 316         struct ip6_tnl *iter;
 317
 318         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 319              (iter = rtnl_dereference(*tp)) != NULL;
 320              tp = &iter->next) {
 321                 if (t == iter) {
 322                         rcu_assign_pointer(*tp, t->next);
 323                         break;
 324                 }
 325         }
 326 }
 327
 328 static void ip6_dev_free(struct net_device *dev)
 329 {
 330         struct ip6_tnl *t = netdev_priv(dev);
 331
 332         ip6_tnl_dst_destroy(t);
 333         free_percpu(dev->tstats);
 334         free_netdev(dev);
 335 }
 336
 337 static int ip6_tnl_create2(struct net_device *dev)
 338 {
 339         struct ip6_tnl *t = netdev_priv(dev);
 340         struct net *net = dev_net(dev);
 341         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 342         int err;
 343
 344         t = netdev_priv(dev);
 345
 346         err = register_netdevice(dev);
 347         if (err < 0)
 348                 goto out;
 349
 350         strcpy(t->parms.name, dev->name);
 351         dev->rtnl_link_ops = &ip6_link_ops;
 352
 353         dev_hold(dev);
 354         ip6_tnl_link(ip6n, t);
 355         return 0;
 356
 357 out:
 358         return err;
 359 }
 360
 361 /**
 362  * ip6_tnl_create - create a new tunnel
 363  *   @p: tunnel parameters
 364  *   @pt: pointer to new tunnel
 365  *
 366  * Description:
 367  *   Create tunnel matching given parameters.
 368  *
 369  * Return:
 370  *   created tunnel or error pointer
 371  **/
 372
 373 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 374 {
 375         struct net_device *dev;
 376         struct ip6_tnl *t;
 377         char name[IFNAMSIZ];
 378         int err = -ENOMEM;
 379
 380         if (p->name[0])
 381                 strlcpy(name, p->name, IFNAMSIZ);
 382         else
 383                 sprintf(name, "ip6tnl%%d");
 384
 385         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 386                            ip6_tnl_dev_setup);
 387         if (!dev)
 388                 goto failed;
 389
 390         dev_net_set(dev, net);
 391
 392         t = netdev_priv(dev);
 393         t->parms = *p;
 394         t->net = dev_net(dev);
 395         err = ip6_tnl_create2(dev);
 396         if (err < 0)
 397                 goto failed_free;
 398
 399         return t;
 400
 401 failed_free:
 402         ip6_dev_free(dev);
 403 failed:
 404         return ERR_PTR(err);
 405 }
 406
 407 /**
 408  * ip6_tnl_locate - find or create tunnel matching given parameters
 409  *   @p: tunnel parameters
 410  *   @create: != 0 if allowed to create new tunnel if no match found
 411  *
 412  * Description:
 413  *   ip6_tnl_locate() first tries to locate an existing tunnel
 414  *   based on @parms. If this is unsuccessful, but @create is set a new
 415  *   tunnel device is created and registered for use.
 416  *
 417  * Return:
 418  *   matching tunnel or error pointer
 419  **/
 420
 421 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 422                 struct __ip6_tnl_parm *p, int create)
 423 {
 424         const struct in6_addr *remote = &p->raddr;
 425         const struct in6_addr *local = &p->laddr;
 426         struct ip6_tnl __rcu **tp;
 427         struct ip6_tnl *t;
 428         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 429
 430         for (tp = ip6_tnl_bucket(ip6n, p);
 431              (t = rtnl_dereference(*tp)) != NULL;
 432              tp = &t->next) {
 433                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 434                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 435                         if (create)
 436                                 return ERR_PTR(-EEXIST);
 437
 438                         return t;
 439                 }
 440         }
 441         if (!create)
 442                 return ERR_PTR(-ENODEV);
 443         return ip6_tnl_create(net, p);
 444 }
 445
 446 /**
 447  * ip6_tnl_dev_uninit - tunnel device uninitializer
 448  *   @dev: the device to be destroyed
 449  *
 450  * Description:
 451  *   ip6_tnl_dev_uninit() removes tunnel from its list
 452  **/
 453
 454 static void
 455 ip6_tnl_dev_uninit(struct net_device *dev)
 456 {
 457         struct ip6_tnl *t = netdev_priv(dev);
 458         struct net *net = t->net;
 459         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 460
 461         if (dev == ip6n->fb_tnl_dev)
 462                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 463         else
 464                 ip6_tnl_unlink(ip6n, t);
 465         ip6_tnl_dst_reset(t);
 466         dev_put(dev);
 467 }
 468
 469 /**
 470  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 471  *   @skb: received socket buffer
 472  *
 473  * Return:
 474  *   0 if none was found,
 475  *   else index to encapsulation limit
 476  **/
 477
 478 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 479 {
 480         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 481         __u8 nexthdr = ipv6h->nexthdr;
 482         __u16 off = sizeof(*ipv6h);
 483
 484         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 485                 __u16 optlen = 0;
 486                 struct ipv6_opt_hdr *hdr;
 487                 if (raw + off + sizeof(*hdr) > skb->data &&
 488                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 489                         break;
 490
 491                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 492                 if (nexthdr == NEXTHDR_FRAGMENT) {
 493                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 494                         if (frag_hdr->frag_off)
 495                                 break;
 496                         optlen = 8;
 497                 } else if (nexthdr == NEXTHDR_AUTH) {
 498                         optlen = (hdr->hdrlen + 2) << 2;
 499                 } else {
 500                         optlen = ipv6_optlen(hdr);
 501                 }
 502                 if (nexthdr == NEXTHDR_DEST) {
 503                         __u16 i = off + 2;
 504                         while (1) {
 505                                 struct ipv6_tlv_tnl_enc_lim *tel;
 506
 507                                 /* No more room for encapsulation limit */
 508                                 if (i + sizeof (*tel) > off + optlen)
 509                                         break;
 510
 511                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 512                                 /* return index of option if found and valid */
 513                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 514                                     tel->length == 1)
 515                                         return i;
 516                                 /* else jump to next option */
 517                                 if (tel->type)
 518                                         i += tel->length + 2;
 519                                 else
 520                                         i++;
 521                         }
 522                 }
 523                 nexthdr = hdr->nexthdr;
 524                 off += optlen;
 525         }
 526         return 0;
 527 }
 528 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 529
 530 /**
 531  * ip6_tnl_err - tunnel error handler
 532  *
 533  * Description:
 534  *   ip6_tnl_err() should handle errors in the tunnel according
 535  *   to the specifications in RFC 2473.
 536  **/
 537
 538 static int
 539 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 540             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 541 {
 542         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 543         struct ip6_tnl *t;
 544         int rel_msg = 0;
 545         u8 rel_type = ICMPV6_DEST_UNREACH;
 546         u8 rel_code = ICMPV6_ADDR_UNREACH;
 547         u8 tproto;
 548         __u32 rel_info = 0;
 549         __u16 len;
 550         int err = -ENOENT;
 551
 552         /* If the packet doesn't contain the original IPv6 header we are
 553            in trouble since we might need the source address for further
 554            processing of the error. */
 555
 556         rcu_read_lock();
 557         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 558         if (!t)
 559                 goto out;
 560
 561         tproto = ACCESS_ONCE(t->parms.proto);
 562         if (tproto != ipproto && tproto != 0)
 563                 goto out;
 564
 565         err = 0;
 566
 567         switch (*type) {
 568                 __u32 teli;
 569                 struct ipv6_tlv_tnl_enc_lim *tel;
 570                 __u32 mtu;
 571         case ICMPV6_DEST_UNREACH:
 572                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 573                                     t->parms.name);
 574                 rel_msg = 1;
 575                 break;
 576         case ICMPV6_TIME_EXCEED:
 577                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 578                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 579                                             t->parms.name);
 580                         rel_msg = 1;
 581                 }
 582                 break;
 583         case ICMPV6_PARAMPROB:
 584                 teli = 0;
 585                 if ((*code) == ICMPV6_HDR_FIELD)
 586                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 587
 588                 if (teli && teli == *info - 2) {
 589                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 590                         if (tel->encap_limit == 0) {
 591                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 592                                                     t->parms.name);
 593                                 rel_msg = 1;
 594                         }
 595                 } else {
 596                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 597                                             t->parms.name);
 598                 }
 599                 break;
 600         case ICMPV6_PKT_TOOBIG:
 601                 mtu = *info - offset;
 602                 if (mtu < IPV6_MIN_MTU)
 603                         mtu = IPV6_MIN_MTU;
 604                 t->dev->mtu = mtu;
 605
 606                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 607                 if (len > mtu) {
 608                         rel_type = ICMPV6_PKT_TOOBIG;
 609                         rel_code = 0;
 610                         rel_info = mtu;
 611                         rel_msg = 1;
 612                 }
 613                 break;
 614         }
 615
 616         *type = rel_type;
 617         *code = rel_code;
 618         *info = rel_info;
 619         *msg = rel_msg;
 620
 621 out:
 622         rcu_read_unlock();
 623         return err;
 624 }
 625
 626 static int
 627 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 628            u8 type, u8 code, int offset, __be32 info)
 629 {
 630         int rel_msg = 0;
 631         u8 rel_type = type;
 632         u8 rel_code = code;
 633         __u32 rel_info = ntohl(info);
 634         int err;
 635         struct sk_buff *skb2;
 636         const struct iphdr *eiph;
 637         struct rtable *rt;
 638         struct flowi4 fl4;
 639
 640         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 641                           &rel_msg, &rel_info, offset);
 642         if (err < 0)
 643                 return err;
 644
 645         if (rel_msg == 0)
 646                 return 0;
 647
 648         switch (rel_type) {
 649         case ICMPV6_DEST_UNREACH:
 650                 if (rel_code != ICMPV6_ADDR_UNREACH)
 651                         return 0;
 652                 rel_type = ICMP_DEST_UNREACH;
 653                 rel_code = ICMP_HOST_UNREACH;
 654                 break;
 655         case ICMPV6_PKT_TOOBIG:
 656                 if (rel_code != 0)
 657                         return 0;
 658                 rel_type = ICMP_DEST_UNREACH;
 659                 rel_code = ICMP_FRAG_NEEDED;
 660                 break;
 661         case NDISC_REDIRECT:
 662                 rel_type = ICMP_REDIRECT;
 663                 rel_code = ICMP_REDIR_HOST;
 664         default:
 665                 return 0;
 666         }
 667
 668         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 669                 return 0;
 670
 671         skb2 = skb_clone(skb, GFP_ATOMIC);
 672         if (!skb2)
 673                 return 0;
 674
 675         skb_dst_drop(skb2);
 676
 677         skb_pull(skb2, offset);
 678         skb_reset_network_header(skb2);
 679         eiph = ip_hdr(skb2);
 680
 681         /* Try to guess incoming interface */
 682         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 683                                    eiph->saddr, 0,
 684                                    0, 0,
 685                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 686         if (IS_ERR(rt))
 687                 goto out;
 688
 689         skb2->dev = rt->dst.dev;
 690
 691         /* route "incoming" packet */
 692         if (rt->rt_flags & RTCF_LOCAL) {
 693                 ip_rt_put(rt);
 694                 rt = NULL;
 695                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 696                                            eiph->daddr, eiph->saddr,
 697                                            0, 0,
 698                                            IPPROTO_IPIP,
 699                                            RT_TOS(eiph->tos), 0);
 700                 if (IS_ERR(rt) ||
 701                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 702                         if (!IS_ERR(rt))
 703                                 ip_rt_put(rt);
 704                         goto out;
 705                 }
 706                 skb_dst_set(skb2, &rt->dst);
 707         } else {
 708                 ip_rt_put(rt);
 709                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 710                                    skb2->dev) ||
 711                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 712                         goto out;
 713         }
 714
 715         /* change mtu on this route */
 716         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 717                 if (rel_info > dst_mtu(skb_dst(skb2)))
 718                         goto out;
 719
 720                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 721         }
 722         if (rel_type == ICMP_REDIRECT)
 723                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 724
 725         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 726
 727 out:
 728         kfree_skb(skb2);
 729         return 0;
 730 }
 731
 732 static int
 733 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 734            u8 type, u8 code, int offset, __be32 info)
 735 {
 736         int rel_msg = 0;
 737         u8 rel_type = type;
 738         u8 rel_code = code;
 739         __u32 rel_info = ntohl(info);
 740         int err;
 741
 742         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 743                           &rel_msg, &rel_info, offset);
 744         if (err < 0)
 745                 return err;
 746
 747         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 748                 struct rt6_info *rt;
 749                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 750
 751                 if (!skb2)
 752                         return 0;
 753
 754                 skb_dst_drop(skb2);
 755                 skb_pull(skb2, offset);
 756                 skb_reset_network_header(skb2);
 757
 758                 /* Try to guess incoming interface */
 759                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 760                                 NULL, 0, 0);
 761
 762                 if (rt && rt->dst.dev)
 763                         skb2->dev = rt->dst.dev;
 764
 765                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 766
 767                 ip6_rt_put(rt);
 768
 769                 kfree_skb(skb2);
 770         }
 771
 772         return 0;
 773 }
 774
 775 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 776                                        const struct ipv6hdr *ipv6h,
 777                                        struct sk_buff *skb)
 778 {
 779         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 780
 781         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 782                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 783
 784         return IP6_ECN_decapsulate(ipv6h, skb);
 785 }
 786
 787 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 788                                        const struct ipv6hdr *ipv6h,
 789                                        struct sk_buff *skb)
 790 {
 791         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 792                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 793
 794         return IP6_ECN_decapsulate(ipv6h, skb);
 795 }
 796
 797 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 798                              const struct in6_addr *laddr,
 799                              const struct in6_addr *raddr)
 800 {
 801         struct __ip6_tnl_parm *p = &t->parms;
 802         int ltype = ipv6_addr_type(laddr);
 803         int rtype = ipv6_addr_type(raddr);
 804         __u32 flags = 0;
 805
 806         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 807                 flags = IP6_TNL_F_CAP_PER_PACKET;
 808         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 809                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 810                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 811                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 812                 if (ltype&IPV6_ADDR_UNICAST)
 813                         flags |= IP6_TNL_F_CAP_XMIT;
 814                 if (rtype&IPV6_ADDR_UNICAST)
 815                         flags |= IP6_TNL_F_CAP_RCV;
 816         }
 817         return flags;
 818 }
 819 EXPORT_SYMBOL(ip6_tnl_get_cap);
 820
 821 /* called with rcu_read_lock() */
 822 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 823                                   const struct in6_addr *laddr,
 824                                   const struct in6_addr *raddr)
 825 {
 826         struct __ip6_tnl_parm *p = &t->parms;
 827         int ret = 0;
 828         struct net *net = t->net;
 829
 830         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 831             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 832              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 833                 struct net_device *ldev = NULL;
 834
 835                 if (p->link)
 836                         ldev = dev_get_by_index_rcu(net, p->link);
 837
 838                 if ((ipv6_addr_is_multicast(laddr) ||
 839                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 840                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 841                         ret = 1;
 842         }
 843         return ret;
 844 }
 845 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 846
 847 /**
 848  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 849  *   @skb: received socket buffer
 850  *   @protocol: ethernet protocol ID
 851  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 852  *
 853  * Return: 0
 854  **/
 855
 856 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 857                        __u8 ipproto,
 858                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 859                                                    const struct ipv6hdr *ipv6h,
 860                                                    struct sk_buff *skb))
 861 {
 862         struct ip6_tnl *t;
 863         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 864         u8 tproto;
 865         int err;
 866
 867         rcu_read_lock();
 868         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 869         if (t) {
 870                 struct pcpu_sw_netstats *tstats;
 871
 872                 tproto = ACCESS_ONCE(t->parms.proto);
 873                 if (tproto != ipproto && tproto != 0) {
 874                         rcu_read_unlock();
 875                         goto discard;
 876                 }
 877
 878                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 879                         rcu_read_unlock();
 880                         goto discard;
 881                 }
 882
 883                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
 884                         t->dev->stats.rx_dropped++;
 885                         rcu_read_unlock();
 886                         goto discard;
 887                 }
 888                 skb->mac_header = skb->network_header;
 889                 skb_reset_network_header(skb);
 890                 skb->protocol = htons(protocol);
 891                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 892
 893                 __skb_tunnel_rx(skb, t->dev, t->net);
 894
 895                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
 896                 if (unlikely(err)) {
 897                         if (log_ecn_error)
 898                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 899                                                      &ipv6h->saddr,
 900                                                      ipv6_get_dsfield(ipv6h));
 901                         if (err > 1) {
 902                                 ++t->dev->stats.rx_frame_errors;
 903                                 ++t->dev->stats.rx_errors;
 904                                 rcu_read_unlock();
 905                                 goto discard;
 906                         }
 907                 }
 908
 909                 tstats = this_cpu_ptr(t->dev->tstats);
 910                 u64_stats_update_begin(&tstats->syncp);
 911                 tstats->rx_packets++;
 912                 tstats->rx_bytes += skb->len;
 913                 u64_stats_update_end(&tstats->syncp);
 914
 915                 netif_rx(skb);
 916
 917                 rcu_read_unlock();
 918                 return 0;
 919         }
 920         rcu_read_unlock();
 921         return 1;
 922
 923 discard:
 924         kfree_skb(skb);
 925         return 0;
 926 }
 927
 928 static int ip4ip6_rcv(struct sk_buff *skb)
 929 {
 930         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 931                            ip4ip6_dscp_ecn_decapsulate);
 932 }
 933
 934 static int ip6ip6_rcv(struct sk_buff *skb)
 935 {
 936         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 937                            ip6ip6_dscp_ecn_decapsulate);
 938 }
 939
 940 struct ipv6_tel_txoption {
 941         struct ipv6_txoptions ops;
 942         __u8 dst_opt[8];
 943 };
 944
 945 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 946 {
 947         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 948
 949         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 950         opt->dst_opt[3] = 1;
 951         opt->dst_opt[4] = encap_limit;
 952         opt->dst_opt[5] = IPV6_TLV_PADN;
 953         opt->dst_opt[6] = 1;
 954
 955         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 956         opt->ops.opt_nflen = 8;
 957 }
 958
 959 /**
 960  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 961  *   @t: the outgoing tunnel device
 962  *   @hdr: IPv6 header from the incoming packet
 963  *
 964  * Description:
 965  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 966  *   doesn't match source of incoming packet.
 967  *
 968  * Return:
 969  *   1 if conflict,
 970  *   0 else
 971  **/
 972
 973 static inline bool
 974 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 975 {
 976         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 977 }
 978
 979 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 980                      const struct in6_addr *laddr,
 981                      const struct in6_addr *raddr)
 982 {
 983         struct __ip6_tnl_parm *p = &t->parms;
 984         int ret = 0;
 985         struct net *net = t->net;
 986
 987         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 988             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 989              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 990                 struct net_device *ldev = NULL;
 991
 992                 rcu_read_lock();
 993                 if (p->link)
 994                         ldev = dev_get_by_index_rcu(net, p->link);
 995
 996                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 997                         pr_warn("%s xmit: Local address not yet configured!\n",
 998                                 p->name);
 999                 else if (!ipv6_addr_is_multicast(raddr) &&
1000                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1001                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1002                                 p->name);
1003                 else
1004                         ret = 1;
1005                 rcu_read_unlock();
1006         }
1007         return ret;
1008 }
1009 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1010
1011 /**
1012  * ip6_tnl_xmit2 - encapsulate packet and send
1013  *   @skb: the outgoing socket buffer
1014  *   @dev: the outgoing tunnel device
1015  *   @dsfield: dscp code for outer header
1016  *   @fl: flow of tunneled packet
1017  *   @encap_limit: encapsulation limit
1018  *   @pmtu: Path MTU is stored if packet is too big
1019  *
1020  * Description:
1021  *   Build new header and do some sanity checks on the packet before sending
1022  *   it.
1023  *
1024  * Return:
1025  *   0 on success
1026  *   -1 fail
1027  *   %-EMSGSIZE message too big. return mtu in this case.
1028  **/
1029
1030 static int ip6_tnl_xmit2(struct sk_buff *skb,
1031                          struct net_device *dev,
1032                          __u8 dsfield,
1033                          struct flowi6 *fl6,
1034                          int encap_limit,
1035                          __u32 *pmtu)
1036 {
1037         struct ip6_tnl *t = netdev_priv(dev);
1038         struct net *net = t->net;
1039         struct net_device_stats *stats = &t->dev->stats;
1040         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1041         struct ipv6_tel_txoption opt;
1042         struct dst_entry *dst = NULL, *ndst = NULL;
1043         struct net_device *tdev;
1044         int mtu;
1045         unsigned int max_headroom = sizeof(struct ipv6hdr);
1046         u8 proto;
1047         int err = -1;
1048
1049         /* NBMA tunnel */
1050         if (ipv6_addr_any(&t->parms.raddr)) {
1051                 struct in6_addr *addr6;
1052                 struct neighbour *neigh;
1053                 int addr_type;
1054
1055                 if (!skb_dst(skb))
1056                         goto tx_err_link_failure;
1057
1058                 neigh = dst_neigh_lookup(skb_dst(skb),
1059                                          &ipv6_hdr(skb)->daddr);
1060                 if (!neigh)
1061                         goto tx_err_link_failure;
1062
1063                 addr6 = (struct in6_addr *)&neigh->primary_key;
1064                 addr_type = ipv6_addr_type(addr6);
1065
1066                 if (addr_type == IPV6_ADDR_ANY)
1067                         addr6 = &ipv6_hdr(skb)->daddr;
1068
1069                 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1070                 neigh_release(neigh);
1071         } else if (!fl6->flowi6_mark)
1072                 dst = ip6_tnl_dst_get(t);
1073
1074         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1075                 goto tx_err_link_failure;
1076
1077         if (!dst) {
1078                 dst = ip6_route_output(net, NULL, fl6);
1079
1080                 if (dst->error)
1081                         goto tx_err_link_failure;
1082                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1083                 if (IS_ERR(dst)) {
1084                         err = PTR_ERR(dst);
1085                         dst = NULL;
1086                         goto tx_err_link_failure;
1087                 }
1088                 ndst = dst;
1089         }
1090
1091         tdev = dst->dev;
1092
1093         if (tdev == dev) {
1094                 stats->collisions++;
1095                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1096                                      t->parms.name);
1097                 goto tx_err_dst_release;
1098         }
1099         mtu = dst_mtu(dst) - sizeof(*ipv6h);
1100         if (encap_limit >= 0) {
1101                 max_headroom += 8;
1102                 mtu -= 8;
1103         }
1104         if (mtu < IPV6_MIN_MTU)
1105                 mtu = IPV6_MIN_MTU;
1106         if (skb_dst(skb))
1107                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1108         if (skb->len > mtu) {
1109                 *pmtu = mtu;
1110                 err = -EMSGSIZE;
1111                 goto tx_err_dst_release;
1112         }
1113
1114         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1115
1116         /*
1117          * Okay, now see if we can stuff it in the buffer as-is.
1118          */
1119         max_headroom += LL_RESERVED_SPACE(tdev);
1120
1121         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1122             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1123                 struct sk_buff *new_skb;
1124
1125                 new_skb = skb_realloc_headroom(skb, max_headroom);
1126                 if (!new_skb)
1127                         goto tx_err_dst_release;
1128
1129                 if (skb->sk)
1130                         skb_set_owner_w(new_skb, skb->sk);
1131                 consume_skb(skb);
1132                 skb = new_skb;
1133         }
1134
1135         if (!fl6->flowi6_mark && ndst)
1136                 ip6_tnl_dst_set(t, ndst);
1137         skb_dst_set(skb, dst);
1138
1139         skb->transport_header = skb->network_header;
1140
1141         proto = fl6->flowi6_proto;
1142         if (encap_limit >= 0) {
1143                 init_tel_txopt(&opt, encap_limit);
1144                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1145         }
1146
1147         if (likely(!skb->encapsulation)) {
1148                 skb_reset_inner_headers(skb);
1149                 skb->encapsulation = 1;
1150         }
1151
1152         skb_push(skb, sizeof(struct ipv6hdr));
1153         skb_reset_network_header(skb);
1154         ipv6h = ipv6_hdr(skb);
1155         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1156                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1157         ipv6h->hop_limit = t->parms.hop_limit;
1158         ipv6h->nexthdr = proto;
1159         ipv6h->saddr = fl6->saddr;
1160         ipv6h->daddr = fl6->daddr;
1161         ip6tunnel_xmit(NULL, skb, dev);
1162         return 0;
1163 tx_err_link_failure:
1164         stats->tx_carrier_errors++;
1165         dst_link_failure(skb);
1166 tx_err_dst_release:
1167         dst_release(dst);
1168         return err;
1169 }
1170
1171 static inline int
1172 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1173 {
1174         struct ip6_tnl *t = netdev_priv(dev);
1175         const struct iphdr  *iph = ip_hdr(skb);
1176         int encap_limit = -1;
1177         struct flowi6 fl6;
1178         __u8 dsfield;
1179         __u32 mtu;
1180         u8 tproto;
1181         int err;
1182
1183         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1184
1185         tproto = ACCESS_ONCE(t->parms.proto);
1186         if (tproto != IPPROTO_IPIP && tproto != 0)
1187                 return -1;
1188
1189         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1190                 encap_limit = t->parms.encap_limit;
1191
1192         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1193         fl6.flowi6_proto = IPPROTO_IPIP;
1194
1195         dsfield = ipv4_get_dsfield(iph);
1196
1197         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1198                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1199                                           & IPV6_TCLASS_MASK;
1200         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1201                 fl6.flowi6_mark = skb->mark;
1202
1203         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1204         if (err != 0) {
1205                 /* XXX: send ICMP error even if DF is not set. */
1206                 if (err == -EMSGSIZE)
1207                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1208                                   htonl(mtu));
1209                 return -1;
1210         }
1211
1212         return 0;
1213 }
1214
1215 static inline int
1216 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1217 {
1218         struct ip6_tnl *t = netdev_priv(dev);
1219         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1220         int encap_limit = -1;
1221         __u16 offset;
1222         struct flowi6 fl6;
1223         __u8 dsfield;
1224         __u32 mtu;
1225         u8 tproto;
1226         int err;
1227
1228         tproto = ACCESS_ONCE(t->parms.proto);
1229         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1230             ip6_tnl_addr_conflict(t, ipv6h))
1231                 return -1;
1232
1233         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1234         if (offset > 0) {
1235                 struct ipv6_tlv_tnl_enc_lim *tel;
1236                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1237                 if (tel->encap_limit == 0) {
1238                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1239                                     ICMPV6_HDR_FIELD, offset + 2);
1240                         return -1;
1241                 }
1242                 encap_limit = tel->encap_limit - 1;
1243         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1244                 encap_limit = t->parms.encap_limit;
1245
1246         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1247         fl6.flowi6_proto = IPPROTO_IPV6;
1248
1249         dsfield = ipv6_get_dsfield(ipv6h);
1250         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1251                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1252         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1253                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1254         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1255                 fl6.flowi6_mark = skb->mark;
1256
1257         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1258         if (err != 0) {
1259                 if (err == -EMSGSIZE)
1260                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1261                 return -1;
1262         }
1263
1264         return 0;
1265 }
1266
1267 static netdev_tx_t
1268 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1269 {
1270         struct ip6_tnl *t = netdev_priv(dev);
1271         struct net_device_stats *stats = &t->dev->stats;
1272         int ret;
1273
1274         switch (skb->protocol) {
1275         case htons(ETH_P_IP):
1276                 ret = ip4ip6_tnl_xmit(skb, dev);
1277                 break;
1278         case htons(ETH_P_IPV6):
1279                 ret = ip6ip6_tnl_xmit(skb, dev);
1280                 break;
1281         default:
1282                 goto tx_err;
1283         }
1284
1285         if (ret < 0)
1286                 goto tx_err;
1287
1288         return NETDEV_TX_OK;
1289
1290 tx_err:
1291         stats->tx_errors++;
1292         stats->tx_dropped++;
1293         kfree_skb(skb);
1294         return NETDEV_TX_OK;
1295 }
1296
1297 static void ip6_tnl_link_config(struct ip6_tnl *t)
1298 {
1299         struct net_device *dev = t->dev;
1300         struct __ip6_tnl_parm *p = &t->parms;
1301         struct flowi6 *fl6 = &t->fl.u.ip6;
1302
1303         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1304         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1305
1306         /* Set up flowi template */
1307         fl6->saddr = p->laddr;
1308         fl6->daddr = p->raddr;
1309         fl6->flowi6_oif = p->link;
1310         fl6->flowlabel = 0;
1311
1312         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1313                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1314         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1315                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1316
1317         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1318         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1319
1320         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1321                 dev->flags |= IFF_POINTOPOINT;
1322         else
1323                 dev->flags &= ~IFF_POINTOPOINT;
1324
1325         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1326                 int strict = (ipv6_addr_type(&p->raddr) &
1327                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1328
1329                 struct rt6_info *rt = rt6_lookup(t->net,
1330                                                  &p->raddr, &p->laddr,
1331                                                  p->link, strict);
1332
1333                 if (!rt)
1334                         return;
1335
1336                 if (rt->dst.dev) {
1337                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1338                                 sizeof(struct ipv6hdr);
1339
1340                         dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1341                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1342                                 dev->mtu -= 8;
1343
1344                         if (dev->mtu < IPV6_MIN_MTU)
1345                                 dev->mtu = IPV6_MIN_MTU;
1346                 }
1347                 ip6_rt_put(rt);
1348         }
1349 }
1350
1351 /**
1352  * ip6_tnl_change - update the tunnel parameters
1353  *   @t: tunnel to be changed
1354  *   @p: tunnel configuration parameters
1355  *
1356  * Description:
1357  *   ip6_tnl_change() updates the tunnel parameters
1358  **/
1359
1360 static int
1361 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1362 {
1363         t->parms.laddr = p->laddr;
1364         t->parms.raddr = p->raddr;
1365         t->parms.flags = p->flags;
1366         t->parms.hop_limit = p->hop_limit;
1367         t->parms.encap_limit = p->encap_limit;
1368         t->parms.flowinfo = p->flowinfo;
1369         t->parms.link = p->link;
1370         t->parms.proto = p->proto;
1371         ip6_tnl_dst_reset(t);
1372         ip6_tnl_link_config(t);
1373         return 0;
1374 }
1375
1376 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1377 {
1378         struct net *net = t->net;
1379         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1380         int err;
1381
1382         ip6_tnl_unlink(ip6n, t);
1383         synchronize_net();
1384         err = ip6_tnl_change(t, p);
1385         ip6_tnl_link(ip6n, t);
1386         netdev_state_change(t->dev);
1387         return err;
1388 }
1389
1390 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1391 {
1392         /* for default tnl0 device allow to change only the proto */
1393         t->parms.proto = p->proto;
1394         netdev_state_change(t->dev);
1395         return 0;
1396 }
1397
1398 static void
1399 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1400 {
1401         p->laddr = u->laddr;
1402         p->raddr = u->raddr;
1403         p->flags = u->flags;
1404         p->hop_limit = u->hop_limit;
1405         p->encap_limit = u->encap_limit;
1406         p->flowinfo = u->flowinfo;
1407         p->link = u->link;
1408         p->proto = u->proto;
1409         memcpy(p->name, u->name, sizeof(u->name));
1410 }
1411
1412 static void
1413 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1414 {
1415         u->laddr = p->laddr;
1416         u->raddr = p->raddr;
1417         u->flags = p->flags;
1418         u->hop_limit = p->hop_limit;
1419         u->encap_limit = p->encap_limit;
1420         u->flowinfo = p->flowinfo;
1421         u->link = p->link;
1422         u->proto = p->proto;
1423         memcpy(u->name, p->name, sizeof(u->name));
1424 }
1425
1426 /**
1427  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1428  *   @dev: virtual device associated with tunnel
1429  *   @ifr: parameters passed from userspace
1430  *   @cmd: command to be performed
1431  *
1432  * Description:
1433  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1434  *   from userspace.
1435  *
1436  *   The possible commands are the following:
1437  *     %SIOCGETTUNNEL: get tunnel parameters for device
1438  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1439  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1440  *     %SIOCDELTUNNEL: delete tunnel
1441  *
1442  *   The fallback device "ip6tnl0", created during module
1443  *   initialization, can be used for creating other tunnel devices.
1444  *
1445  * Return:
1446  *   0 on success,
1447  *   %-EFAULT if unable to copy data to or from userspace,
1448  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1449  *   %-EINVAL if passed tunnel parameters are invalid,
1450  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1451  *   %-ENODEV if attempting to change or delete a nonexisting device
1452  **/
1453
1454 static int
1455 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1456 {
1457         int err = 0;
1458         struct ip6_tnl_parm p;
1459         struct __ip6_tnl_parm p1;
1460         struct ip6_tnl *t = netdev_priv(dev);
1461         struct net *net = t->net;
1462         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1463
1464         switch (cmd) {
1465         case SIOCGETTUNNEL:
1466                 if (dev == ip6n->fb_tnl_dev) {
1467                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1468                                 err = -EFAULT;
1469                                 break;
1470                         }
1471                         ip6_tnl_parm_from_user(&p1, &p);
1472                         t = ip6_tnl_locate(net, &p1, 0);
1473                         if (IS_ERR(t))
1474                                 t = netdev_priv(dev);
1475                 } else {
1476                         memset(&p, 0, sizeof(p));
1477                 }
1478                 ip6_tnl_parm_to_user(&p, &t->parms);
1479                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1480                         err = -EFAULT;
1481                 }
1482                 break;
1483         case SIOCADDTUNNEL:
1484         case SIOCCHGTUNNEL:
1485                 err = -EPERM;
1486                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1487                         break;
1488                 err = -EFAULT;
1489                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1490                         break;
1491                 err = -EINVAL;
1492                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1493                     p.proto != 0)
1494                         break;
1495                 ip6_tnl_parm_from_user(&p1, &p);
1496                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1497                 if (cmd == SIOCCHGTUNNEL) {
1498                         if (!IS_ERR(t)) {
1499                                 if (t->dev != dev) {
1500                                         err = -EEXIST;
1501                                         break;
1502                                 }
1503                         } else
1504                                 t = netdev_priv(dev);
1505                         if (dev == ip6n->fb_tnl_dev)
1506                                 err = ip6_tnl0_update(t, &p1);
1507                         else
1508                                 err = ip6_tnl_update(t, &p1);
1509                 }
1510                 if (!IS_ERR(t)) {
1511                         err = 0;
1512                         ip6_tnl_parm_to_user(&p, &t->parms);
1513                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1514                                 err = -EFAULT;
1515
1516                 } else {
1517                         err = PTR_ERR(t);
1518                 }
1519                 break;
1520         case SIOCDELTUNNEL:
1521                 err = -EPERM;
1522                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1523                         break;
1524
1525                 if (dev == ip6n->fb_tnl_dev) {
1526                         err = -EFAULT;
1527                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1528                                 break;
1529                         err = -ENOENT;
1530                         ip6_tnl_parm_from_user(&p1, &p);
1531                         t = ip6_tnl_locate(net, &p1, 0);
1532                         if (IS_ERR(t))
1533                                 break;
1534                         err = -EPERM;
1535                         if (t->dev == ip6n->fb_tnl_dev)
1536                                 break;
1537                         dev = t->dev;
1538                 }
1539                 err = 0;
1540                 unregister_netdevice(dev);
1541                 break;
1542         default:
1543                 err = -EINVAL;
1544         }
1545         return err;
1546 }
1547
1548 /**
1549  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1550  *   @dev: virtual device associated with tunnel
1551  *   @new_mtu: the new mtu
1552  *
1553  * Return:
1554  *   0 on success,
1555  *   %-EINVAL if mtu too small
1556  **/
1557
1558 static int
1559 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1560 {
1561         struct ip6_tnl *tnl = netdev_priv(dev);
1562
1563         if (tnl->parms.proto == IPPROTO_IPIP) {
1564                 if (new_mtu < 68)
1565                         return -EINVAL;
1566         } else {
1567                 if (new_mtu < IPV6_MIN_MTU)
1568                         return -EINVAL;
1569         }
1570         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1571                 return -EINVAL;
1572         dev->mtu = new_mtu;
1573         return 0;
1574 }
1575
1576 int ip6_tnl_get_iflink(const struct net_device *dev)
1577 {
1578         struct ip6_tnl *t = netdev_priv(dev);
1579
1580         return t->parms.link;
1581 }
1582 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1583
1584 static const struct net_device_ops ip6_tnl_netdev_ops = {
1585         .ndo_init       = ip6_tnl_dev_init,
1586         .ndo_uninit     = ip6_tnl_dev_uninit,
1587         .ndo_start_xmit = ip6_tnl_xmit,
1588         .ndo_do_ioctl   = ip6_tnl_ioctl,
1589         .ndo_change_mtu = ip6_tnl_change_mtu,
1590         .ndo_get_stats  = ip6_get_stats,
1591         .ndo_get_iflink = ip6_tnl_get_iflink,
1592 };
1593
1594
1595 /**
1596  * ip6_tnl_dev_setup - setup virtual tunnel device
1597  *   @dev: virtual device associated with tunnel
1598  *
1599  * Description:
1600  *   Initialize function pointers and device parameters
1601  **/
1602
1603 static void ip6_tnl_dev_setup(struct net_device *dev)
1604 {
1605         struct ip6_tnl *t;
1606
1607         dev->netdev_ops = &ip6_tnl_netdev_ops;
1608         dev->destructor = ip6_dev_free;
1609
1610         dev->type = ARPHRD_TUNNEL6;
1611         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1612         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1613         t = netdev_priv(dev);
1614         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1615                 dev->mtu -= 8;
1616         dev->flags |= IFF_NOARP;
1617         dev->addr_len = sizeof(struct in6_addr);
1618         netif_keep_dst(dev);
1619         /* This perm addr will be used as interface identifier by IPv6 */
1620         dev->addr_assign_type = NET_ADDR_RANDOM;
1621         eth_random_addr(dev->perm_addr);
1622 }
1623
1624
1625 /**
1626  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1627  *   @dev: virtual device associated with tunnel
1628  **/
1629
1630 static inline int
1631 ip6_tnl_dev_init_gen(struct net_device *dev)
1632 {
1633         struct ip6_tnl *t = netdev_priv(dev);
1634         int ret;
1635
1636         t->dev = dev;
1637         t->net = dev_net(dev);
1638         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1639         if (!dev->tstats)
1640                 return -ENOMEM;
1641
1642         ret = ip6_tnl_dst_init(t);
1643         if (ret) {
1644                 free_percpu(dev->tstats);
1645                 dev->tstats = NULL;
1646                 return ret;
1647         }
1648
1649         return 0;
1650 }
1651
1652 /**
1653  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1654  *   @dev: virtual device associated with tunnel
1655  **/
1656
1657 static int ip6_tnl_dev_init(struct net_device *dev)
1658 {
1659         struct ip6_tnl *t = netdev_priv(dev);
1660         int err = ip6_tnl_dev_init_gen(dev);
1661
1662         if (err)
1663                 return err;
1664         ip6_tnl_link_config(t);
1665         return 0;
1666 }
1667
1668 /**
1669  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1670  *   @dev: fallback device
1671  *
1672  * Return: 0
1673  **/
1674
1675 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1676 {
1677         struct ip6_tnl *t = netdev_priv(dev);
1678         struct net *net = dev_net(dev);
1679         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1680
1681         t->parms.proto = IPPROTO_IPV6;
1682         dev_hold(dev);
1683
1684         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1685         return 0;
1686 }
1687
1688 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1689 {
1690         u8 proto;
1691
1692         if (!data || !data[IFLA_IPTUN_PROTO])
1693                 return 0;
1694
1695         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1696         if (proto != IPPROTO_IPV6 &&
1697             proto != IPPROTO_IPIP &&
1698             proto != 0)
1699                 return -EINVAL;
1700
1701         return 0;
1702 }
1703
1704 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1705                                   struct __ip6_tnl_parm *parms)
1706 {
1707         memset(parms, 0, sizeof(*parms));
1708
1709         if (!data)
1710                 return;
1711
1712         if (data[IFLA_IPTUN_LINK])
1713                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1714
1715         if (data[IFLA_IPTUN_LOCAL])
1716                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1717
1718         if (data[IFLA_IPTUN_REMOTE])
1719                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1720
1721         if (data[IFLA_IPTUN_TTL])
1722                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1723
1724         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1725                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1726
1727         if (data[IFLA_IPTUN_FLOWINFO])
1728                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1729
1730         if (data[IFLA_IPTUN_FLAGS])
1731                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1732
1733         if (data[IFLA_IPTUN_PROTO])
1734                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1735 }
1736
1737 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1738                            struct nlattr *tb[], struct nlattr *data[])
1739 {
1740         struct net *net = dev_net(dev);
1741         struct ip6_tnl *nt, *t;
1742
1743         nt = netdev_priv(dev);
1744         ip6_tnl_netlink_parms(data, &nt->parms);
1745
1746         t = ip6_tnl_locate(net, &nt->parms, 0);
1747         if (!IS_ERR(t))
1748                 return -EEXIST;
1749
1750         return ip6_tnl_create2(dev);
1751 }
1752
1753 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1754                               struct nlattr *data[])
1755 {
1756         struct ip6_tnl *t = netdev_priv(dev);
1757         struct __ip6_tnl_parm p;
1758         struct net *net = t->net;
1759         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1760
1761         if (dev == ip6n->fb_tnl_dev)
1762                 return -EINVAL;
1763
1764         ip6_tnl_netlink_parms(data, &p);
1765
1766         t = ip6_tnl_locate(net, &p, 0);
1767         if (!IS_ERR(t)) {
1768                 if (t->dev != dev)
1769                         return -EEXIST;
1770         } else
1771                 t = netdev_priv(dev);
1772
1773         return ip6_tnl_update(t, &p);
1774 }
1775
1776 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1777 {
1778         struct net *net = dev_net(dev);
1779         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1780
1781         if (dev != ip6n->fb_tnl_dev)
1782                 unregister_netdevice_queue(dev, head);
1783 }
1784
1785 static size_t ip6_tnl_get_size(const struct net_device *dev)
1786 {
1787         return
1788                 /* IFLA_IPTUN_LINK */
1789                 nla_total_size(4) +
1790                 /* IFLA_IPTUN_LOCAL */
1791                 nla_total_size(sizeof(struct in6_addr)) +
1792                 /* IFLA_IPTUN_REMOTE */
1793                 nla_total_size(sizeof(struct in6_addr)) +
1794                 /* IFLA_IPTUN_TTL */
1795                 nla_total_size(1) +
1796                 /* IFLA_IPTUN_ENCAP_LIMIT */
1797                 nla_total_size(1) +
1798                 /* IFLA_IPTUN_FLOWINFO */
1799                 nla_total_size(4) +
1800                 /* IFLA_IPTUN_FLAGS */
1801                 nla_total_size(4) +
1802                 /* IFLA_IPTUN_PROTO */
1803                 nla_total_size(1) +
1804                 0;
1805 }
1806
1807 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1808 {
1809         struct ip6_tnl *tunnel = netdev_priv(dev);
1810         struct __ip6_tnl_parm *parm = &tunnel->parms;
1811
1812         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1813             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1814             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1815             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1816             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1817             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1818             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1819             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1820                 goto nla_put_failure;
1821         return 0;
1822
1823 nla_put_failure:
1824         return -EMSGSIZE;
1825 }
1826
1827 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1828 {
1829         struct ip6_tnl *tunnel = netdev_priv(dev);
1830
1831         return tunnel->net;
1832 }
1833 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1834
1835 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1836         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1837         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1838         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1839         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1840         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1841         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1842         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1843         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1844 };
1845
1846 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1847         .kind           = "ip6tnl",
1848         .maxtype        = IFLA_IPTUN_MAX,
1849         .policy         = ip6_tnl_policy,
1850         .priv_size      = sizeof(struct ip6_tnl),
1851         .setup          = ip6_tnl_dev_setup,
1852         .validate       = ip6_tnl_validate,
1853         .newlink        = ip6_tnl_newlink,
1854         .changelink     = ip6_tnl_changelink,
1855         .dellink        = ip6_tnl_dellink,
1856         .get_size       = ip6_tnl_get_size,
1857         .fill_info      = ip6_tnl_fill_info,
1858         .get_link_net   = ip6_tnl_get_link_net,
1859 };
1860
1861 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1862         .handler        = ip4ip6_rcv,
1863         .err_handler    = ip4ip6_err,
1864         .priority       =       1,
1865 };
1866
1867 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1868         .handler        = ip6ip6_rcv,
1869         .err_handler    = ip6ip6_err,
1870         .priority       =       1,
1871 };
1872
1873 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1874 {
1875         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1876         struct net_device *dev, *aux;
1877         int h;
1878         struct ip6_tnl *t;
1879         LIST_HEAD(list);
1880
1881         for_each_netdev_safe(net, dev, aux)
1882                 if (dev->rtnl_link_ops == &ip6_link_ops)
1883                         unregister_netdevice_queue(dev, &list);
1884
1885         for (h = 0; h < HASH_SIZE; h++) {
1886                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1887                 while (t) {
1888                         /* If dev is in the same netns, it has already
1889                          * been added to the list by the previous loop.
1890                          */
1891                         if (!net_eq(dev_net(t->dev), net))
1892                                 unregister_netdevice_queue(t->dev, &list);
1893                         t = rtnl_dereference(t->next);
1894                 }
1895         }
1896
1897         unregister_netdevice_many(&list);
1898 }
1899
1900 static int __net_init ip6_tnl_init_net(struct net *net)
1901 {
1902         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1903         struct ip6_tnl *t = NULL;
1904         int err;
1905
1906         ip6n->tnls[0] = ip6n->tnls_wc;
1907         ip6n->tnls[1] = ip6n->tnls_r_l;
1908
1909         err = -ENOMEM;
1910         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1911                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1912
1913         if (!ip6n->fb_tnl_dev)
1914                 goto err_alloc_dev;
1915         dev_net_set(ip6n->fb_tnl_dev, net);
1916         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1917         /* FB netdevice is special: we have one, and only one per netns.
1918          * Allowing to move it to another netns is clearly unsafe.
1919          */
1920         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1921
1922         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1923         if (err < 0)
1924                 goto err_register;
1925
1926         err = register_netdev(ip6n->fb_tnl_dev);
1927         if (err < 0)
1928                 goto err_register;
1929
1930         t = netdev_priv(ip6n->fb_tnl_dev);
1931
1932         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1933         return 0;
1934
1935 err_register:
1936         ip6_dev_free(ip6n->fb_tnl_dev);
1937 err_alloc_dev:
1938         return err;
1939 }
1940
1941 static void __net_exit ip6_tnl_exit_net(struct net *net)
1942 {
1943         rtnl_lock();
1944         ip6_tnl_destroy_tunnels(net);
1945         rtnl_unlock();
1946 }
1947
1948 static struct pernet_operations ip6_tnl_net_ops = {
1949         .init = ip6_tnl_init_net,
1950         .exit = ip6_tnl_exit_net,
1951         .id   = &ip6_tnl_net_id,
1952         .size = sizeof(struct ip6_tnl_net),
1953 };
1954
1955 /**
1956  * ip6_tunnel_init - register protocol and reserve needed resources
1957  *
1958  * Return: 0 on success
1959  **/
1960
1961 static int __init ip6_tunnel_init(void)
1962 {
1963         int  err;
1964
1965         err = register_pernet_device(&ip6_tnl_net_ops);
1966         if (err < 0)
1967                 goto out_pernet;
1968
1969         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1970         if (err < 0) {
1971                 pr_err("%s: can't register ip4ip6\n", __func__);
1972                 goto out_ip4ip6;
1973         }
1974
1975         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1976         if (err < 0) {
1977                 pr_err("%s: can't register ip6ip6\n", __func__);
1978                 goto out_ip6ip6;
1979         }
1980         err = rtnl_link_register(&ip6_link_ops);
1981         if (err < 0)
1982                 goto rtnl_link_failed;
1983
1984         return 0;
1985
1986 rtnl_link_failed:
1987         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1988 out_ip6ip6:
1989         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1990 out_ip4ip6:
1991         unregister_pernet_device(&ip6_tnl_net_ops);
1992 out_pernet:
1993         return err;
1994 }
1995
1996 /**
1997  * ip6_tunnel_cleanup - free resources and unregister protocol
1998  **/
1999
2000 static void __exit ip6_tunnel_cleanup(void)
2001 {
2002         rtnl_link_unregister(&ip6_link_ops);
2003         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2004                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2005
2006         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2007                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2008
2009         unregister_pernet_device(&ip6_tnl_net_ops);
2010 }
2011
2012 module_init(ip6_tunnel_init);
2013 module_exit(ip6_tunnel_cleanup);