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 #ifdef IP6_TNL_DEBUG
  68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
  69 #else
  70 #define IP6_TNL_TRACE(x...) do {;} while(0)
  71 #endif
  72
  73 #define HASH_SIZE_SHIFT  5
  74 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  75
  76 static bool log_ecn_error = true;
  77 module_param(log_ecn_error, bool, 0644);
  78 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  79
  80 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  81 {
  82         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  83
  84         return hash_32(hash, HASH_SIZE_SHIFT);
  85 }
  86
  87 static int ip6_tnl_dev_init(struct net_device *dev);
  88 static void ip6_tnl_dev_setup(struct net_device *dev);
  89 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  90
  91 static int ip6_tnl_net_id __read_mostly;
  92 struct ip6_tnl_net {
  93         /* the IPv6 tunnel fallback device */
  94         struct net_device *fb_tnl_dev;
  95         /* lists for storing tunnels in use */
  96         struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
  97         struct ip6_tnl __rcu *tnls_wc[1];
  98         struct ip6_tnl __rcu **tnls[2];
  99 };
 100
 101 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
 102 {
 103         struct pcpu_sw_netstats tmp, sum = { 0 };
 104         int i;
 105
 106         for_each_possible_cpu(i) {
 107                 unsigned int start;
 108                 const struct pcpu_sw_netstats *tstats =
 109                                                    per_cpu_ptr(dev->tstats, i);
 110
 111                 do {
 112                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 113                         tmp.rx_packets = tstats->rx_packets;
 114                         tmp.rx_bytes = tstats->rx_bytes;
 115                         tmp.tx_packets = tstats->tx_packets;
 116                         tmp.tx_bytes =  tstats->tx_bytes;
 117                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 118
 119                 sum.rx_packets += tmp.rx_packets;
 120                 sum.rx_bytes   += tmp.rx_bytes;
 121                 sum.tx_packets += tmp.tx_packets;
 122                 sum.tx_bytes   += tmp.tx_bytes;
 123         }
 124         dev->stats.rx_packets = sum.rx_packets;
 125         dev->stats.rx_bytes   = sum.rx_bytes;
 126         dev->stats.tx_packets = sum.tx_packets;
 127         dev->stats.tx_bytes   = sum.tx_bytes;
 128         return &dev->stats;
 129 }
 130
 131 /*
 132  * Locking : hash tables are protected by RCU and RTNL
 133  */
 134
 135 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
 136 {
 137         struct dst_entry *dst = t->dst_cache;
 138
 139         if (dst && dst->obsolete &&
 140             dst->ops->check(dst, t->dst_cookie) == NULL) {
 141                 t->dst_cache = NULL;
 142                 dst_release(dst);
 143                 return NULL;
 144         }
 145
 146         return dst;
 147 }
 148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
 149
 150 void ip6_tnl_dst_reset(struct ip6_tnl *t)
 151 {
 152         dst_release(t->dst_cache);
 153         t->dst_cache = NULL;
 154 }
 155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
 156
 157 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
 158 {
 159         struct rt6_info *rt = (struct rt6_info *) dst;
 160         t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
 161         dst_release(t->dst_cache);
 162         t->dst_cache = dst;
 163 }
 164 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
 165
 166 /**
 167  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 168  *   @remote: the address of the tunnel exit-point
 169  *   @local: the address of the tunnel entry-point
 170  *
 171  * Return:
 172  *   tunnel matching given end-points if found,
 173  *   else fallback tunnel if its device is up,
 174  *   else %NULL
 175  **/
 176
 177 #define for_each_ip6_tunnel_rcu(start) \
 178         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 179
 180 static struct ip6_tnl *
 181 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 182 {
 183         unsigned int hash = HASH(remote, local);
 184         struct ip6_tnl *t;
 185         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 186
 187         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 188                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 189                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 190                     (t->dev->flags & IFF_UP))
 191                         return t;
 192         }
 193         t = rcu_dereference(ip6n->tnls_wc[0]);
 194         if (t && (t->dev->flags & IFF_UP))
 195                 return t;
 196
 197         return NULL;
 198 }
 199
 200 /**
 201  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 202  *   @p: parameters containing tunnel end-points
 203  *
 204  * Description:
 205  *   ip6_tnl_bucket() returns the head of the list matching the
 206  *   &struct in6_addr entries laddr and raddr in @p.
 207  *
 208  * Return: head of IPv6 tunnel list
 209  **/
 210
 211 static struct ip6_tnl __rcu **
 212 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 213 {
 214         const struct in6_addr *remote = &p->raddr;
 215         const struct in6_addr *local = &p->laddr;
 216         unsigned int h = 0;
 217         int prio = 0;
 218
 219         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 220                 prio = 1;
 221                 h = HASH(remote, local);
 222         }
 223         return &ip6n->tnls[prio][h];
 224 }
 225
 226 /**
 227  * ip6_tnl_link - add tunnel to hash table
 228  *   @t: tunnel to be added
 229  **/
 230
 231 static void
 232 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 233 {
 234         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 235
 236         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 237         rcu_assign_pointer(*tp, t);
 238 }
 239
 240 /**
 241  * ip6_tnl_unlink - remove tunnel from hash table
 242  *   @t: tunnel to be removed
 243  **/
 244
 245 static void
 246 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 247 {
 248         struct ip6_tnl __rcu **tp;
 249         struct ip6_tnl *iter;
 250
 251         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 252              (iter = rtnl_dereference(*tp)) != NULL;
 253              tp = &iter->next) {
 254                 if (t == iter) {
 255                         rcu_assign_pointer(*tp, t->next);
 256                         break;
 257                 }
 258         }
 259 }
 260
 261 static void ip6_dev_free(struct net_device *dev)
 262 {
 263         free_percpu(dev->tstats);
 264         free_netdev(dev);
 265 }
 266
 267 static int ip6_tnl_create2(struct net_device *dev)
 268 {
 269         struct ip6_tnl *t = netdev_priv(dev);
 270         struct net *net = dev_net(dev);
 271         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 272         int err;
 273
 274         t = netdev_priv(dev);
 275         err = ip6_tnl_dev_init(dev);
 276         if (err < 0)
 277                 goto out;
 278
 279         err = register_netdevice(dev);
 280         if (err < 0)
 281                 goto out;
 282
 283         strcpy(t->parms.name, dev->name);
 284         dev->rtnl_link_ops = &ip6_link_ops;
 285
 286         dev_hold(dev);
 287         ip6_tnl_link(ip6n, t);
 288         return 0;
 289
 290 out:
 291         return err;
 292 }
 293
 294 /**
 295  * ip6_tnl_create - create a new tunnel
 296  *   @p: tunnel parameters
 297  *   @pt: pointer to new tunnel
 298  *
 299  * Description:
 300  *   Create tunnel matching given parameters.
 301  *
 302  * Return:
 303  *   created tunnel or NULL
 304  **/
 305
 306 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 307 {
 308         struct net_device *dev;
 309         struct ip6_tnl *t;
 310         char name[IFNAMSIZ];
 311         int err;
 312
 313         if (p->name[0])
 314                 strlcpy(name, p->name, IFNAMSIZ);
 315         else
 316                 sprintf(name, "ip6tnl%%d");
 317
 318         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 319                            ip6_tnl_dev_setup);
 320         if (dev == NULL)
 321                 goto failed;
 322
 323         dev_net_set(dev, net);
 324
 325         t = netdev_priv(dev);
 326         t->parms = *p;
 327         t->net = dev_net(dev);
 328         err = ip6_tnl_create2(dev);
 329         if (err < 0)
 330                 goto failed_free;
 331
 332         return t;
 333
 334 failed_free:
 335         ip6_dev_free(dev);
 336 failed:
 337         return NULL;
 338 }
 339
 340 /**
 341  * ip6_tnl_locate - find or create tunnel matching given parameters
 342  *   @p: tunnel parameters
 343  *   @create: != 0 if allowed to create new tunnel if no match found
 344  *
 345  * Description:
 346  *   ip6_tnl_locate() first tries to locate an existing tunnel
 347  *   based on @parms. If this is unsuccessful, but @create is set a new
 348  *   tunnel device is created and registered for use.
 349  *
 350  * Return:
 351  *   matching tunnel or NULL
 352  **/
 353
 354 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 355                 struct __ip6_tnl_parm *p, int create)
 356 {
 357         const struct in6_addr *remote = &p->raddr;
 358         const struct in6_addr *local = &p->laddr;
 359         struct ip6_tnl __rcu **tp;
 360         struct ip6_tnl *t;
 361         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 362
 363         for (tp = ip6_tnl_bucket(ip6n, p);
 364              (t = rtnl_dereference(*tp)) != NULL;
 365              tp = &t->next) {
 366                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 367                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 368                         if (create)
 369                                 return NULL;
 370
 371                         return t;
 372                 }
 373         }
 374         if (!create)
 375                 return NULL;
 376         return ip6_tnl_create(net, p);
 377 }
 378
 379 /**
 380  * ip6_tnl_dev_uninit - tunnel device uninitializer
 381  *   @dev: the device to be destroyed
 382  *
 383  * Description:
 384  *   ip6_tnl_dev_uninit() removes tunnel from its list
 385  **/
 386
 387 static void
 388 ip6_tnl_dev_uninit(struct net_device *dev)
 389 {
 390         struct ip6_tnl *t = netdev_priv(dev);
 391         struct net *net = t->net;
 392         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 393
 394         if (dev == ip6n->fb_tnl_dev)
 395                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 396         else
 397                 ip6_tnl_unlink(ip6n, t);
 398         ip6_tnl_dst_reset(t);
 399         dev_put(dev);
 400 }
 401
 402 /**
 403  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 404  *   @skb: received socket buffer
 405  *
 406  * Return:
 407  *   0 if none was found,
 408  *   else index to encapsulation limit
 409  **/
 410
 411 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 412 {
 413         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 414         __u8 nexthdr = ipv6h->nexthdr;
 415         __u16 off = sizeof(*ipv6h);
 416
 417         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 418                 __u16 optlen = 0;
 419                 struct ipv6_opt_hdr *hdr;
 420                 if (raw + off + sizeof(*hdr) > skb->data &&
 421                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 422                         break;
 423
 424                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 425                 if (nexthdr == NEXTHDR_FRAGMENT) {
 426                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 427                         if (frag_hdr->frag_off)
 428                                 break;
 429                         optlen = 8;
 430                 } else if (nexthdr == NEXTHDR_AUTH) {
 431                         optlen = (hdr->hdrlen + 2) << 2;
 432                 } else {
 433                         optlen = ipv6_optlen(hdr);
 434                 }
 435                 if (nexthdr == NEXTHDR_DEST) {
 436                         __u16 i = off + 2;
 437                         while (1) {
 438                                 struct ipv6_tlv_tnl_enc_lim *tel;
 439
 440                                 /* No more room for encapsulation limit */
 441                                 if (i + sizeof (*tel) > off + optlen)
 442                                         break;
 443
 444                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 445                                 /* return index of option if found and valid */
 446                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 447                                     tel->length == 1)
 448                                         return i;
 449                                 /* else jump to next option */
 450                                 if (tel->type)
 451                                         i += tel->length + 2;
 452                                 else
 453                                         i++;
 454                         }
 455                 }
 456                 nexthdr = hdr->nexthdr;
 457                 off += optlen;
 458         }
 459         return 0;
 460 }
 461 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 462
 463 /**
 464  * ip6_tnl_err - tunnel error handler
 465  *
 466  * Description:
 467  *   ip6_tnl_err() should handle errors in the tunnel according
 468  *   to the specifications in RFC 2473.
 469  **/
 470
 471 static int
 472 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 473             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 474 {
 475         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 476         struct ip6_tnl *t;
 477         int rel_msg = 0;
 478         u8 rel_type = ICMPV6_DEST_UNREACH;
 479         u8 rel_code = ICMPV6_ADDR_UNREACH;
 480         __u32 rel_info = 0;
 481         __u16 len;
 482         int err = -ENOENT;
 483
 484         /* If the packet doesn't contain the original IPv6 header we are
 485            in trouble since we might need the source address for further
 486            processing of the error. */
 487
 488         rcu_read_lock();
 489         if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
 490                                         &ipv6h->saddr)) == NULL)
 491                 goto out;
 492
 493         if (t->parms.proto != ipproto && t->parms.proto != 0)
 494                 goto out;
 495
 496         err = 0;
 497
 498         switch (*type) {
 499                 __u32 teli;
 500                 struct ipv6_tlv_tnl_enc_lim *tel;
 501                 __u32 mtu;
 502         case ICMPV6_DEST_UNREACH:
 503                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 504                                      t->parms.name);
 505                 rel_msg = 1;
 506                 break;
 507         case ICMPV6_TIME_EXCEED:
 508                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 509                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 510                                              t->parms.name);
 511                         rel_msg = 1;
 512                 }
 513                 break;
 514         case ICMPV6_PARAMPROB:
 515                 teli = 0;
 516                 if ((*code) == ICMPV6_HDR_FIELD)
 517                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 518
 519                 if (teli && teli == *info - 2) {
 520                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 521                         if (tel->encap_limit == 0) {
 522                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 523                                                      t->parms.name);
 524                                 rel_msg = 1;
 525                         }
 526                 } else {
 527                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 528                                              t->parms.name);
 529                 }
 530                 break;
 531         case ICMPV6_PKT_TOOBIG:
 532                 mtu = *info - offset;
 533                 if (mtu < IPV6_MIN_MTU)
 534                         mtu = IPV6_MIN_MTU;
 535                 t->dev->mtu = mtu;
 536
 537                 if ((len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
 538                         rel_type = ICMPV6_PKT_TOOBIG;
 539                         rel_code = 0;
 540                         rel_info = mtu;
 541                         rel_msg = 1;
 542                 }
 543                 break;
 544         }
 545
 546         *type = rel_type;
 547         *code = rel_code;
 548         *info = rel_info;
 549         *msg = rel_msg;
 550
 551 out:
 552         rcu_read_unlock();
 553         return err;
 554 }
 555
 556 static int
 557 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 558            u8 type, u8 code, int offset, __be32 info)
 559 {
 560         int rel_msg = 0;
 561         u8 rel_type = type;
 562         u8 rel_code = code;
 563         __u32 rel_info = ntohl(info);
 564         int err;
 565         struct sk_buff *skb2;
 566         const struct iphdr *eiph;
 567         struct rtable *rt;
 568         struct flowi4 fl4;
 569
 570         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 571                           &rel_msg, &rel_info, offset);
 572         if (err < 0)
 573                 return err;
 574
 575         if (rel_msg == 0)
 576                 return 0;
 577
 578         switch (rel_type) {
 579         case ICMPV6_DEST_UNREACH:
 580                 if (rel_code != ICMPV6_ADDR_UNREACH)
 581                         return 0;
 582                 rel_type = ICMP_DEST_UNREACH;
 583                 rel_code = ICMP_HOST_UNREACH;
 584                 break;
 585         case ICMPV6_PKT_TOOBIG:
 586                 if (rel_code != 0)
 587                         return 0;
 588                 rel_type = ICMP_DEST_UNREACH;
 589                 rel_code = ICMP_FRAG_NEEDED;
 590                 break;
 591         case NDISC_REDIRECT:
 592                 rel_type = ICMP_REDIRECT;
 593                 rel_code = ICMP_REDIR_HOST;
 594         default:
 595                 return 0;
 596         }
 597
 598         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 599                 return 0;
 600
 601         skb2 = skb_clone(skb, GFP_ATOMIC);
 602         if (!skb2)
 603                 return 0;
 604
 605         skb_dst_drop(skb2);
 606
 607         skb_pull(skb2, offset);
 608         skb_reset_network_header(skb2);
 609         eiph = ip_hdr(skb2);
 610
 611         /* Try to guess incoming interface */
 612         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 613                                    eiph->saddr, 0,
 614                                    0, 0,
 615                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 616         if (IS_ERR(rt))
 617                 goto out;
 618
 619         skb2->dev = rt->dst.dev;
 620
 621         /* route "incoming" packet */
 622         if (rt->rt_flags & RTCF_LOCAL) {
 623                 ip_rt_put(rt);
 624                 rt = NULL;
 625                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 626                                            eiph->daddr, eiph->saddr,
 627                                            0, 0,
 628                                            IPPROTO_IPIP,
 629                                            RT_TOS(eiph->tos), 0);
 630                 if (IS_ERR(rt) ||
 631                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 632                         if (!IS_ERR(rt))
 633                                 ip_rt_put(rt);
 634                         goto out;
 635                 }
 636                 skb_dst_set(skb2, &rt->dst);
 637         } else {
 638                 ip_rt_put(rt);
 639                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 640                                    skb2->dev) ||
 641                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 642                         goto out;
 643         }
 644
 645         /* change mtu on this route */
 646         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 647                 if (rel_info > dst_mtu(skb_dst(skb2)))
 648                         goto out;
 649
 650                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 651         }
 652         if (rel_type == ICMP_REDIRECT)
 653                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 654
 655         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 656
 657 out:
 658         kfree_skb(skb2);
 659         return 0;
 660 }
 661
 662 static int
 663 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 664            u8 type, u8 code, int offset, __be32 info)
 665 {
 666         int rel_msg = 0;
 667         u8 rel_type = type;
 668         u8 rel_code = code;
 669         __u32 rel_info = ntohl(info);
 670         int err;
 671
 672         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 673                           &rel_msg, &rel_info, offset);
 674         if (err < 0)
 675                 return err;
 676
 677         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 678                 struct rt6_info *rt;
 679                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 680
 681                 if (!skb2)
 682                         return 0;
 683
 684                 skb_dst_drop(skb2);
 685                 skb_pull(skb2, offset);
 686                 skb_reset_network_header(skb2);
 687
 688                 /* Try to guess incoming interface */
 689                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 690                                 NULL, 0, 0);
 691
 692                 if (rt && rt->dst.dev)
 693                         skb2->dev = rt->dst.dev;
 694
 695                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 696
 697                 ip6_rt_put(rt);
 698
 699                 kfree_skb(skb2);
 700         }
 701
 702         return 0;
 703 }
 704
 705 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 706                                        const struct ipv6hdr *ipv6h,
 707                                        struct sk_buff *skb)
 708 {
 709         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 710
 711         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 712                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 713
 714         return IP6_ECN_decapsulate(ipv6h, skb);
 715 }
 716
 717 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 718                                        const struct ipv6hdr *ipv6h,
 719                                        struct sk_buff *skb)
 720 {
 721         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 722                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 723
 724         return IP6_ECN_decapsulate(ipv6h, skb);
 725 }
 726
 727 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 728                              const struct in6_addr *laddr,
 729                              const struct in6_addr *raddr)
 730 {
 731         struct __ip6_tnl_parm *p = &t->parms;
 732         int ltype = ipv6_addr_type(laddr);
 733         int rtype = ipv6_addr_type(raddr);
 734         __u32 flags = 0;
 735
 736         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 737                 flags = IP6_TNL_F_CAP_PER_PACKET;
 738         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 739                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 740                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 741                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 742                 if (ltype&IPV6_ADDR_UNICAST)
 743                         flags |= IP6_TNL_F_CAP_XMIT;
 744                 if (rtype&IPV6_ADDR_UNICAST)
 745                         flags |= IP6_TNL_F_CAP_RCV;
 746         }
 747         return flags;
 748 }
 749 EXPORT_SYMBOL(ip6_tnl_get_cap);
 750
 751 /* called with rcu_read_lock() */
 752 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 753                                   const struct in6_addr *laddr,
 754                                   const struct in6_addr *raddr)
 755 {
 756         struct __ip6_tnl_parm *p = &t->parms;
 757         int ret = 0;
 758         struct net *net = t->net;
 759
 760         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 761             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 762              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 763                 struct net_device *ldev = NULL;
 764
 765                 if (p->link)
 766                         ldev = dev_get_by_index_rcu(net, p->link);
 767
 768                 if ((ipv6_addr_is_multicast(laddr) ||
 769                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 770                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 771                         ret = 1;
 772         }
 773         return ret;
 774 }
 775 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 776
 777 /**
 778  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 779  *   @skb: received socket buffer
 780  *   @protocol: ethernet protocol ID
 781  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 782  *
 783  * Return: 0
 784  **/
 785
 786 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 787                        __u8 ipproto,
 788                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 789                                                    const struct ipv6hdr *ipv6h,
 790                                                    struct sk_buff *skb))
 791 {
 792         struct ip6_tnl *t;
 793         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 794         int err;
 795
 796         rcu_read_lock();
 797
 798         if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
 799                                         &ipv6h->daddr)) != NULL) {
 800                 struct pcpu_sw_netstats *tstats;
 801
 802                 if (t->parms.proto != ipproto && t->parms.proto != 0) {
 803                         rcu_read_unlock();
 804                         goto discard;
 805                 }
 806
 807                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 808                         rcu_read_unlock();
 809                         goto discard;
 810                 }
 811
 812                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
 813                         t->dev->stats.rx_dropped++;
 814                         rcu_read_unlock();
 815                         goto discard;
 816                 }
 817                 skb->mac_header = skb->network_header;
 818                 skb_reset_network_header(skb);
 819                 skb->protocol = htons(protocol);
 820                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 821
 822                 __skb_tunnel_rx(skb, t->dev, t->net);
 823
 824                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
 825                 if (unlikely(err)) {
 826                         if (log_ecn_error)
 827                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 828                                                      &ipv6h->saddr,
 829                                                      ipv6_get_dsfield(ipv6h));
 830                         if (err > 1) {
 831                                 ++t->dev->stats.rx_frame_errors;
 832                                 ++t->dev->stats.rx_errors;
 833                                 rcu_read_unlock();
 834                                 goto discard;
 835                         }
 836                 }
 837
 838                 tstats = this_cpu_ptr(t->dev->tstats);
 839                 u64_stats_update_begin(&tstats->syncp);
 840                 tstats->rx_packets++;
 841                 tstats->rx_bytes += skb->len;
 842                 u64_stats_update_end(&tstats->syncp);
 843
 844                 netif_rx(skb);
 845
 846                 rcu_read_unlock();
 847                 return 0;
 848         }
 849         rcu_read_unlock();
 850         return 1;
 851
 852 discard:
 853         kfree_skb(skb);
 854         return 0;
 855 }
 856
 857 static int ip4ip6_rcv(struct sk_buff *skb)
 858 {
 859         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 860                            ip4ip6_dscp_ecn_decapsulate);
 861 }
 862
 863 static int ip6ip6_rcv(struct sk_buff *skb)
 864 {
 865         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 866                            ip6ip6_dscp_ecn_decapsulate);
 867 }
 868
 869 struct ipv6_tel_txoption {
 870         struct ipv6_txoptions ops;
 871         __u8 dst_opt[8];
 872 };
 873
 874 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 875 {
 876         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 877
 878         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 879         opt->dst_opt[3] = 1;
 880         opt->dst_opt[4] = encap_limit;
 881         opt->dst_opt[5] = IPV6_TLV_PADN;
 882         opt->dst_opt[6] = 1;
 883
 884         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 885         opt->ops.opt_nflen = 8;
 886 }
 887
 888 /**
 889  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 890  *   @t: the outgoing tunnel device
 891  *   @hdr: IPv6 header from the incoming packet
 892  *
 893  * Description:
 894  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 895  *   doesn't match source of incoming packet.
 896  *
 897  * Return:
 898  *   1 if conflict,
 899  *   0 else
 900  **/
 901
 902 static inline bool
 903 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 904 {
 905         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 906 }
 907
 908 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
 909 {
 910         struct __ip6_tnl_parm *p = &t->parms;
 911         int ret = 0;
 912         struct net *net = t->net;
 913
 914         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 915                 struct net_device *ldev = NULL;
 916
 917                 rcu_read_lock();
 918                 if (p->link)
 919                         ldev = dev_get_by_index_rcu(net, p->link);
 920
 921                 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
 922                         pr_warn("%s xmit: Local address not yet configured!\n",
 923                                 p->name);
 924                 else if (!ipv6_addr_is_multicast(&p->raddr) &&
 925                          unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
 926                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 927                                 p->name);
 928                 else
 929                         ret = 1;
 930                 rcu_read_unlock();
 931         }
 932         return ret;
 933 }
 934 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
 935
 936 /**
 937  * ip6_tnl_xmit2 - encapsulate packet and send
 938  *   @skb: the outgoing socket buffer
 939  *   @dev: the outgoing tunnel device
 940  *   @dsfield: dscp code for outer header
 941  *   @fl: flow of tunneled packet
 942  *   @encap_limit: encapsulation limit
 943  *   @pmtu: Path MTU is stored if packet is too big
 944  *
 945  * Description:
 946  *   Build new header and do some sanity checks on the packet before sending
 947  *   it.
 948  *
 949  * Return:
 950  *   0 on success
 951  *   -1 fail
 952  *   %-EMSGSIZE message too big. return mtu in this case.
 953  **/
 954
 955 static int ip6_tnl_xmit2(struct sk_buff *skb,
 956                          struct net_device *dev,
 957                          __u8 dsfield,
 958                          struct flowi6 *fl6,
 959                          int encap_limit,
 960                          __u32 *pmtu)
 961 {
 962         struct ip6_tnl *t = netdev_priv(dev);
 963         struct net *net = t->net;
 964         struct net_device_stats *stats = &t->dev->stats;
 965         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 966         struct ipv6_tel_txoption opt;
 967         struct dst_entry *dst = NULL, *ndst = NULL;
 968         struct net_device *tdev;
 969         int mtu;
 970         unsigned int max_headroom = sizeof(struct ipv6hdr);
 971         u8 proto;
 972         int err = -1;
 973
 974         if (!fl6->flowi6_mark)
 975                 dst = ip6_tnl_dst_check(t);
 976         if (!dst) {
 977                 ndst = ip6_route_output(net, NULL, fl6);
 978
 979                 if (ndst->error)
 980                         goto tx_err_link_failure;
 981                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
 982                 if (IS_ERR(ndst)) {
 983                         err = PTR_ERR(ndst);
 984                         ndst = NULL;
 985                         goto tx_err_link_failure;
 986                 }
 987                 dst = ndst;
 988         }
 989
 990         tdev = dst->dev;
 991
 992         if (tdev == dev) {
 993                 stats->collisions++;
 994                 net_warn_ratelimited("%s: Local routing loop detected!\n",
 995                                      t->parms.name);
 996                 goto tx_err_dst_release;
 997         }
 998         mtu = dst_mtu(dst) - sizeof(*ipv6h);
 999         if (encap_limit >= 0) {
1000                 max_headroom += 8;
1001                 mtu -= 8;
1002         }
1003         if (mtu < IPV6_MIN_MTU)
1004                 mtu = IPV6_MIN_MTU;
1005         if (skb_dst(skb))
1006                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1007         if (skb->len > mtu) {
1008                 *pmtu = mtu;
1009                 err = -EMSGSIZE;
1010                 goto tx_err_dst_release;
1011         }
1012
1013         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1014
1015         /*
1016          * Okay, now see if we can stuff it in the buffer as-is.
1017          */
1018         max_headroom += LL_RESERVED_SPACE(tdev);
1019
1020         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1021             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1022                 struct sk_buff *new_skb;
1023
1024                 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1025                         goto tx_err_dst_release;
1026
1027                 if (skb->sk)
1028                         skb_set_owner_w(new_skb, skb->sk);
1029                 consume_skb(skb);
1030                 skb = new_skb;
1031         }
1032         if (fl6->flowi6_mark) {
1033                 skb_dst_set(skb, dst);
1034                 ndst = NULL;
1035         } else {
1036                 skb_dst_set_noref(skb, dst);
1037         }
1038         skb->transport_header = skb->network_header;
1039
1040         proto = fl6->flowi6_proto;
1041         if (encap_limit >= 0) {
1042                 init_tel_txopt(&opt, encap_limit);
1043                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1044         }
1045
1046         if (likely(!skb->encapsulation)) {
1047                 skb_reset_inner_headers(skb);
1048                 skb->encapsulation = 1;
1049         }
1050
1051         skb_push(skb, sizeof(struct ipv6hdr));
1052         skb_reset_network_header(skb);
1053         ipv6h = ipv6_hdr(skb);
1054         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1055                      ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
1056         ipv6h->hop_limit = t->parms.hop_limit;
1057         ipv6h->nexthdr = proto;
1058         ipv6h->saddr = fl6->saddr;
1059         ipv6h->daddr = fl6->daddr;
1060         ip6tunnel_xmit(skb, dev);
1061         if (ndst)
1062                 ip6_tnl_dst_store(t, ndst);
1063         return 0;
1064 tx_err_link_failure:
1065         stats->tx_carrier_errors++;
1066         dst_link_failure(skb);
1067 tx_err_dst_release:
1068         dst_release(ndst);
1069         return err;
1070 }
1071
1072 static inline int
1073 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1074 {
1075         struct ip6_tnl *t = netdev_priv(dev);
1076         const struct iphdr  *iph = ip_hdr(skb);
1077         int encap_limit = -1;
1078         struct flowi6 fl6;
1079         __u8 dsfield;
1080         __u32 mtu;
1081         int err;
1082
1083         if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1084             !ip6_tnl_xmit_ctl(t))
1085                 return -1;
1086
1087         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1088                 encap_limit = t->parms.encap_limit;
1089
1090         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1091         fl6.flowi6_proto = IPPROTO_IPIP;
1092
1093         dsfield = ipv4_get_dsfield(iph);
1094
1095         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1096                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1097                                           & IPV6_TCLASS_MASK;
1098         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1099                 fl6.flowi6_mark = skb->mark;
1100
1101         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1102         if (err != 0) {
1103                 /* XXX: send ICMP error even if DF is not set. */
1104                 if (err == -EMSGSIZE)
1105                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1106                                   htonl(mtu));
1107                 return -1;
1108         }
1109
1110         return 0;
1111 }
1112
1113 static inline int
1114 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1115 {
1116         struct ip6_tnl *t = netdev_priv(dev);
1117         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1118         int encap_limit = -1;
1119         __u16 offset;
1120         struct flowi6 fl6;
1121         __u8 dsfield;
1122         __u32 mtu;
1123         int err;
1124
1125         if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1126             !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1127                 return -1;
1128
1129         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1130         if (offset > 0) {
1131                 struct ipv6_tlv_tnl_enc_lim *tel;
1132                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1133                 if (tel->encap_limit == 0) {
1134                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1135                                     ICMPV6_HDR_FIELD, offset + 2);
1136                         return -1;
1137                 }
1138                 encap_limit = tel->encap_limit - 1;
1139         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1140                 encap_limit = t->parms.encap_limit;
1141
1142         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1143         fl6.flowi6_proto = IPPROTO_IPV6;
1144
1145         dsfield = ipv6_get_dsfield(ipv6h);
1146         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1147                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1148         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1149                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1150         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1151                 fl6.flowi6_mark = skb->mark;
1152
1153         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1154         if (err != 0) {
1155                 if (err == -EMSGSIZE)
1156                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1157                 return -1;
1158         }
1159
1160         return 0;
1161 }
1162
1163 static netdev_tx_t
1164 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1165 {
1166         struct ip6_tnl *t = netdev_priv(dev);
1167         struct net_device_stats *stats = &t->dev->stats;
1168         int ret;
1169
1170         switch (skb->protocol) {
1171         case htons(ETH_P_IP):
1172                 ret = ip4ip6_tnl_xmit(skb, dev);
1173                 break;
1174         case htons(ETH_P_IPV6):
1175                 ret = ip6ip6_tnl_xmit(skb, dev);
1176                 break;
1177         default:
1178                 goto tx_err;
1179         }
1180
1181         if (ret < 0)
1182                 goto tx_err;
1183
1184         return NETDEV_TX_OK;
1185
1186 tx_err:
1187         stats->tx_errors++;
1188         stats->tx_dropped++;
1189         kfree_skb(skb);
1190         return NETDEV_TX_OK;
1191 }
1192
1193 static void ip6_tnl_link_config(struct ip6_tnl *t)
1194 {
1195         struct net_device *dev = t->dev;
1196         struct __ip6_tnl_parm *p = &t->parms;
1197         struct flowi6 *fl6 = &t->fl.u.ip6;
1198
1199         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1200         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1201
1202         /* Set up flowi template */
1203         fl6->saddr = p->laddr;
1204         fl6->daddr = p->raddr;
1205         fl6->flowi6_oif = p->link;
1206         fl6->flowlabel = 0;
1207
1208         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1209                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1210         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1211                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1212
1213         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1214         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1215
1216         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1217                 dev->flags |= IFF_POINTOPOINT;
1218         else
1219                 dev->flags &= ~IFF_POINTOPOINT;
1220
1221         dev->iflink = p->link;
1222
1223         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1224                 int strict = (ipv6_addr_type(&p->raddr) &
1225                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1226
1227                 struct rt6_info *rt = rt6_lookup(t->net,
1228                                                  &p->raddr, &p->laddr,
1229                                                  p->link, strict);
1230
1231                 if (rt == NULL)
1232                         return;
1233
1234                 if (rt->dst.dev) {
1235                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1236                                 sizeof(struct ipv6hdr);
1237
1238                         dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1239                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1240                                 dev->mtu -= 8;
1241
1242                         if (dev->mtu < IPV6_MIN_MTU)
1243                                 dev->mtu = IPV6_MIN_MTU;
1244                 }
1245                 ip6_rt_put(rt);
1246         }
1247 }
1248
1249 /**
1250  * ip6_tnl_change - update the tunnel parameters
1251  *   @t: tunnel to be changed
1252  *   @p: tunnel configuration parameters
1253  *
1254  * Description:
1255  *   ip6_tnl_change() updates the tunnel parameters
1256  **/
1257
1258 static int
1259 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1260 {
1261         t->parms.laddr = p->laddr;
1262         t->parms.raddr = p->raddr;
1263         t->parms.flags = p->flags;
1264         t->parms.hop_limit = p->hop_limit;
1265         t->parms.encap_limit = p->encap_limit;
1266         t->parms.flowinfo = p->flowinfo;
1267         t->parms.link = p->link;
1268         t->parms.proto = p->proto;
1269         ip6_tnl_dst_reset(t);
1270         ip6_tnl_link_config(t);
1271         return 0;
1272 }
1273
1274 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1275 {
1276         struct net *net = t->net;
1277         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1278         int err;
1279
1280         ip6_tnl_unlink(ip6n, t);
1281         synchronize_net();
1282         err = ip6_tnl_change(t, p);
1283         ip6_tnl_link(ip6n, t);
1284         netdev_state_change(t->dev);
1285         return err;
1286 }
1287
1288 static void
1289 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1290 {
1291         p->laddr = u->laddr;
1292         p->raddr = u->raddr;
1293         p->flags = u->flags;
1294         p->hop_limit = u->hop_limit;
1295         p->encap_limit = u->encap_limit;
1296         p->flowinfo = u->flowinfo;
1297         p->link = u->link;
1298         p->proto = u->proto;
1299         memcpy(p->name, u->name, sizeof(u->name));
1300 }
1301
1302 static void
1303 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1304 {
1305         u->laddr = p->laddr;
1306         u->raddr = p->raddr;
1307         u->flags = p->flags;
1308         u->hop_limit = p->hop_limit;
1309         u->encap_limit = p->encap_limit;
1310         u->flowinfo = p->flowinfo;
1311         u->link = p->link;
1312         u->proto = p->proto;
1313         memcpy(u->name, p->name, sizeof(u->name));
1314 }
1315
1316 /**
1317  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1318  *   @dev: virtual device associated with tunnel
1319  *   @ifr: parameters passed from userspace
1320  *   @cmd: command to be performed
1321  *
1322  * Description:
1323  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1324  *   from userspace.
1325  *
1326  *   The possible commands are the following:
1327  *     %SIOCGETTUNNEL: get tunnel parameters for device
1328  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1329  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1330  *     %SIOCDELTUNNEL: delete tunnel
1331  *
1332  *   The fallback device "ip6tnl0", created during module
1333  *   initialization, can be used for creating other tunnel devices.
1334  *
1335  * Return:
1336  *   0 on success,
1337  *   %-EFAULT if unable to copy data to or from userspace,
1338  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1339  *   %-EINVAL if passed tunnel parameters are invalid,
1340  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1341  *   %-ENODEV if attempting to change or delete a nonexisting device
1342  **/
1343
1344 static int
1345 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1346 {
1347         int err = 0;
1348         struct ip6_tnl_parm p;
1349         struct __ip6_tnl_parm p1;
1350         struct ip6_tnl *t = netdev_priv(dev);
1351         struct net *net = t->net;
1352         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1353
1354         switch (cmd) {
1355         case SIOCGETTUNNEL:
1356                 if (dev == ip6n->fb_tnl_dev) {
1357                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1358                                 err = -EFAULT;
1359                                 break;
1360                         }
1361                         ip6_tnl_parm_from_user(&p1, &p);
1362                         t = ip6_tnl_locate(net, &p1, 0);
1363                         if (t == NULL)
1364                                 t = netdev_priv(dev);
1365                 } else {
1366                         memset(&p, 0, sizeof(p));
1367                 }
1368                 ip6_tnl_parm_to_user(&p, &t->parms);
1369                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1370                         err = -EFAULT;
1371                 }
1372                 break;
1373         case SIOCADDTUNNEL:
1374         case SIOCCHGTUNNEL:
1375                 err = -EPERM;
1376                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1377                         break;
1378                 err = -EFAULT;
1379                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1380                         break;
1381                 err = -EINVAL;
1382                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1383                     p.proto != 0)
1384                         break;
1385                 ip6_tnl_parm_from_user(&p1, &p);
1386                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1387                 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1388                         if (t != NULL) {
1389                                 if (t->dev != dev) {
1390                                         err = -EEXIST;
1391                                         break;
1392                                 }
1393                         } else
1394                                 t = netdev_priv(dev);
1395
1396                         err = ip6_tnl_update(t, &p1);
1397                 }
1398                 if (t) {
1399                         err = 0;
1400                         ip6_tnl_parm_to_user(&p, &t->parms);
1401                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1402                                 err = -EFAULT;
1403
1404                 } else
1405                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1406                 break;
1407         case SIOCDELTUNNEL:
1408                 err = -EPERM;
1409                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1410                         break;
1411
1412                 if (dev == ip6n->fb_tnl_dev) {
1413                         err = -EFAULT;
1414                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1415                                 break;
1416                         err = -ENOENT;
1417                         ip6_tnl_parm_from_user(&p1, &p);
1418                         t = ip6_tnl_locate(net, &p1, 0);
1419                         if (t == NULL)
1420                                 break;
1421                         err = -EPERM;
1422                         if (t->dev == ip6n->fb_tnl_dev)
1423                                 break;
1424                         dev = t->dev;
1425                 }
1426                 err = 0;
1427                 unregister_netdevice(dev);
1428                 break;
1429         default:
1430                 err = -EINVAL;
1431         }
1432         return err;
1433 }
1434
1435 /**
1436  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1437  *   @dev: virtual device associated with tunnel
1438  *   @new_mtu: the new mtu
1439  *
1440  * Return:
1441  *   0 on success,
1442  *   %-EINVAL if mtu too small
1443  **/
1444
1445 static int
1446 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1447 {
1448         struct ip6_tnl *tnl = netdev_priv(dev);
1449
1450         if (tnl->parms.proto == IPPROTO_IPIP) {
1451                 if (new_mtu < 68)
1452                         return -EINVAL;
1453         } else {
1454                 if (new_mtu < IPV6_MIN_MTU)
1455                         return -EINVAL;
1456         }
1457         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1458                 return -EINVAL;
1459         dev->mtu = new_mtu;
1460         return 0;
1461 }
1462
1463
1464 static const struct net_device_ops ip6_tnl_netdev_ops = {
1465         .ndo_uninit     = ip6_tnl_dev_uninit,
1466         .ndo_start_xmit = ip6_tnl_xmit,
1467         .ndo_do_ioctl   = ip6_tnl_ioctl,
1468         .ndo_change_mtu = ip6_tnl_change_mtu,
1469         .ndo_get_stats  = ip6_get_stats,
1470 };
1471
1472
1473 /**
1474  * ip6_tnl_dev_setup - setup virtual tunnel device
1475  *   @dev: virtual device associated with tunnel
1476  *
1477  * Description:
1478  *   Initialize function pointers and device parameters
1479  **/
1480
1481 static void ip6_tnl_dev_setup(struct net_device *dev)
1482 {
1483         struct ip6_tnl *t;
1484
1485         dev->netdev_ops = &ip6_tnl_netdev_ops;
1486         dev->destructor = ip6_dev_free;
1487
1488         dev->type = ARPHRD_TUNNEL6;
1489         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1490         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1491         t = netdev_priv(dev);
1492         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1493                 dev->mtu -= 8;
1494         dev->flags |= IFF_NOARP;
1495         dev->addr_len = sizeof(struct in6_addr);
1496         netif_keep_dst(dev);
1497         /* This perm addr will be used as interface identifier by IPv6 */
1498         dev->addr_assign_type = NET_ADDR_RANDOM;
1499         eth_random_addr(dev->perm_addr);
1500 }
1501
1502
1503 /**
1504  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1505  *   @dev: virtual device associated with tunnel
1506  **/
1507
1508 static inline int
1509 ip6_tnl_dev_init_gen(struct net_device *dev)
1510 {
1511         struct ip6_tnl *t = netdev_priv(dev);
1512
1513         t->dev = dev;
1514         t->net = dev_net(dev);
1515         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1516         if (!dev->tstats)
1517                 return -ENOMEM;
1518         return 0;
1519 }
1520
1521 /**
1522  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1523  *   @dev: virtual device associated with tunnel
1524  **/
1525
1526 static int ip6_tnl_dev_init(struct net_device *dev)
1527 {
1528         struct ip6_tnl *t = netdev_priv(dev);
1529         int err = ip6_tnl_dev_init_gen(dev);
1530
1531         if (err)
1532                 return err;
1533         ip6_tnl_link_config(t);
1534         return 0;
1535 }
1536
1537 /**
1538  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1539  *   @dev: fallback device
1540  *
1541  * Return: 0
1542  **/
1543
1544 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1545 {
1546         struct ip6_tnl *t = netdev_priv(dev);
1547         struct net *net = dev_net(dev);
1548         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1549         int err = ip6_tnl_dev_init_gen(dev);
1550
1551         if (err)
1552                 return err;
1553
1554         t->parms.proto = IPPROTO_IPV6;
1555         dev_hold(dev);
1556
1557         ip6_tnl_link_config(t);
1558
1559         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1560         return 0;
1561 }
1562
1563 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1564 {
1565         u8 proto;
1566
1567         if (!data || !data[IFLA_IPTUN_PROTO])
1568                 return 0;
1569
1570         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1571         if (proto != IPPROTO_IPV6 &&
1572             proto != IPPROTO_IPIP &&
1573             proto != 0)
1574                 return -EINVAL;
1575
1576         return 0;
1577 }
1578
1579 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1580                                   struct __ip6_tnl_parm *parms)
1581 {
1582         memset(parms, 0, sizeof(*parms));
1583
1584         if (!data)
1585                 return;
1586
1587         if (data[IFLA_IPTUN_LINK])
1588                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1589
1590         if (data[IFLA_IPTUN_LOCAL])
1591                 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1592                            sizeof(struct in6_addr));
1593
1594         if (data[IFLA_IPTUN_REMOTE])
1595                 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1596                            sizeof(struct in6_addr));
1597
1598         if (data[IFLA_IPTUN_TTL])
1599                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1600
1601         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1602                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1603
1604         if (data[IFLA_IPTUN_FLOWINFO])
1605                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1606
1607         if (data[IFLA_IPTUN_FLAGS])
1608                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1609
1610         if (data[IFLA_IPTUN_PROTO])
1611                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1612 }
1613
1614 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1615                            struct nlattr *tb[], struct nlattr *data[])
1616 {
1617         struct net *net = dev_net(dev);
1618         struct ip6_tnl *nt;
1619
1620         nt = netdev_priv(dev);
1621         ip6_tnl_netlink_parms(data, &nt->parms);
1622
1623         if (ip6_tnl_locate(net, &nt->parms, 0))
1624                 return -EEXIST;
1625
1626         return ip6_tnl_create2(dev);
1627 }
1628
1629 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1630                               struct nlattr *data[])
1631 {
1632         struct ip6_tnl *t = netdev_priv(dev);
1633         struct __ip6_tnl_parm p;
1634         struct net *net = t->net;
1635         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1636
1637         if (dev == ip6n->fb_tnl_dev)
1638                 return -EINVAL;
1639
1640         ip6_tnl_netlink_parms(data, &p);
1641
1642         t = ip6_tnl_locate(net, &p, 0);
1643
1644         if (t) {
1645                 if (t->dev != dev)
1646                         return -EEXIST;
1647         } else
1648                 t = netdev_priv(dev);
1649
1650         return ip6_tnl_update(t, &p);
1651 }
1652
1653 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1654 {
1655         struct net *net = dev_net(dev);
1656         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1657
1658         if (dev != ip6n->fb_tnl_dev)
1659                 unregister_netdevice_queue(dev, head);
1660 }
1661
1662 static size_t ip6_tnl_get_size(const struct net_device *dev)
1663 {
1664         return
1665                 /* IFLA_IPTUN_LINK */
1666                 nla_total_size(4) +
1667                 /* IFLA_IPTUN_LOCAL */
1668                 nla_total_size(sizeof(struct in6_addr)) +
1669                 /* IFLA_IPTUN_REMOTE */
1670                 nla_total_size(sizeof(struct in6_addr)) +
1671                 /* IFLA_IPTUN_TTL */
1672                 nla_total_size(1) +
1673                 /* IFLA_IPTUN_ENCAP_LIMIT */
1674                 nla_total_size(1) +
1675                 /* IFLA_IPTUN_FLOWINFO */
1676                 nla_total_size(4) +
1677                 /* IFLA_IPTUN_FLAGS */
1678                 nla_total_size(4) +
1679                 /* IFLA_IPTUN_PROTO */
1680                 nla_total_size(1) +
1681                 0;
1682 }
1683
1684 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1685 {
1686         struct ip6_tnl *tunnel = netdev_priv(dev);
1687         struct __ip6_tnl_parm *parm = &tunnel->parms;
1688
1689         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1690             nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1691                     &parm->laddr) ||
1692             nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1693                     &parm->raddr) ||
1694             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1695             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1696             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1697             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1698             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1699                 goto nla_put_failure;
1700         return 0;
1701
1702 nla_put_failure:
1703         return -EMSGSIZE;
1704 }
1705
1706 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1707         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1708         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1709         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1710         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1711         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1712         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1713         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1714         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1715 };
1716
1717 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1718         .kind           = "ip6tnl",
1719         .maxtype        = IFLA_IPTUN_MAX,
1720         .policy         = ip6_tnl_policy,
1721         .priv_size      = sizeof(struct ip6_tnl),
1722         .setup          = ip6_tnl_dev_setup,
1723         .validate       = ip6_tnl_validate,
1724         .newlink        = ip6_tnl_newlink,
1725         .changelink     = ip6_tnl_changelink,
1726         .dellink        = ip6_tnl_dellink,
1727         .get_size       = ip6_tnl_get_size,
1728         .fill_info      = ip6_tnl_fill_info,
1729 };
1730
1731 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1732         .handler        = ip4ip6_rcv,
1733         .err_handler    = ip4ip6_err,
1734         .priority       =       1,
1735 };
1736
1737 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1738         .handler        = ip6ip6_rcv,
1739         .err_handler    = ip6ip6_err,
1740         .priority       =       1,
1741 };
1742
1743 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1744 {
1745         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1746         struct net_device *dev, *aux;
1747         int h;
1748         struct ip6_tnl *t;
1749         LIST_HEAD(list);
1750
1751         for_each_netdev_safe(net, dev, aux)
1752                 if (dev->rtnl_link_ops == &ip6_link_ops)
1753                         unregister_netdevice_queue(dev, &list);
1754
1755         for (h = 0; h < HASH_SIZE; h++) {
1756                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1757                 while (t != NULL) {
1758                         /* If dev is in the same netns, it has already
1759                          * been added to the list by the previous loop.
1760                          */
1761                         if (!net_eq(dev_net(t->dev), net))
1762                                 unregister_netdevice_queue(t->dev, &list);
1763                         t = rtnl_dereference(t->next);
1764                 }
1765         }
1766
1767         unregister_netdevice_many(&list);
1768 }
1769
1770 static int __net_init ip6_tnl_init_net(struct net *net)
1771 {
1772         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1773         struct ip6_tnl *t = NULL;
1774         int err;
1775
1776         ip6n->tnls[0] = ip6n->tnls_wc;
1777         ip6n->tnls[1] = ip6n->tnls_r_l;
1778
1779         err = -ENOMEM;
1780         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1781                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1782
1783         if (!ip6n->fb_tnl_dev)
1784                 goto err_alloc_dev;
1785         dev_net_set(ip6n->fb_tnl_dev, net);
1786         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1787         /* FB netdevice is special: we have one, and only one per netns.
1788          * Allowing to move it to another netns is clearly unsafe.
1789          */
1790         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1791
1792         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1793         if (err < 0)
1794                 goto err_register;
1795
1796         err = register_netdev(ip6n->fb_tnl_dev);
1797         if (err < 0)
1798                 goto err_register;
1799
1800         t = netdev_priv(ip6n->fb_tnl_dev);
1801
1802         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1803         return 0;
1804
1805 err_register:
1806         ip6_dev_free(ip6n->fb_tnl_dev);
1807 err_alloc_dev:
1808         return err;
1809 }
1810
1811 static void __net_exit ip6_tnl_exit_net(struct net *net)
1812 {
1813         rtnl_lock();
1814         ip6_tnl_destroy_tunnels(net);
1815         rtnl_unlock();
1816 }
1817
1818 static struct pernet_operations ip6_tnl_net_ops = {
1819         .init = ip6_tnl_init_net,
1820         .exit = ip6_tnl_exit_net,
1821         .id   = &ip6_tnl_net_id,
1822         .size = sizeof(struct ip6_tnl_net),
1823 };
1824
1825 /**
1826  * ip6_tunnel_init - register protocol and reserve needed resources
1827  *
1828  * Return: 0 on success
1829  **/
1830
1831 static int __init ip6_tunnel_init(void)
1832 {
1833         int  err;
1834
1835         err = register_pernet_device(&ip6_tnl_net_ops);
1836         if (err < 0)
1837                 goto out_pernet;
1838
1839         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1840         if (err < 0) {
1841                 pr_err("%s: can't register ip4ip6\n", __func__);
1842                 goto out_ip4ip6;
1843         }
1844
1845         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1846         if (err < 0) {
1847                 pr_err("%s: can't register ip6ip6\n", __func__);
1848                 goto out_ip6ip6;
1849         }
1850         err = rtnl_link_register(&ip6_link_ops);
1851         if (err < 0)
1852                 goto rtnl_link_failed;
1853
1854         return 0;
1855
1856 rtnl_link_failed:
1857         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1858 out_ip6ip6:
1859         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1860 out_ip4ip6:
1861         unregister_pernet_device(&ip6_tnl_net_ops);
1862 out_pernet:
1863         return err;
1864 }
1865
1866 /**
1867  * ip6_tunnel_cleanup - free resources and unregister protocol
1868  **/
1869
1870 static void __exit ip6_tunnel_cleanup(void)
1871 {
1872         rtnl_link_unregister(&ip6_link_ops);
1873         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1874                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1875
1876         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1877                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1878
1879         unregister_pernet_device(&ip6_tnl_net_ops);
1880 }
1881
1882 module_init(ip6_tunnel_init);
1883 module_exit(ip6_tunnel_cleanup);