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 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
 130 {
 131         struct dst_entry *dst = t->dst_cache;
 132
 133         if (dst && dst->obsolete &&
 134             !dst->ops->check(dst, t->dst_cookie)) {
 135                 t->dst_cache = NULL;
 136                 dst_release(dst);
 137                 return NULL;
 138         }
 139
 140         return dst;
 141 }
 142 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
 143
 144 void ip6_tnl_dst_reset(struct ip6_tnl *t)
 145 {
 146         dst_release(t->dst_cache);
 147         t->dst_cache = NULL;
 148 }
 149 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
 150
 151 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
 152 {
 153         struct rt6_info *rt = (struct rt6_info *) dst;
 154         t->dst_cookie = rt6_get_cookie(rt);
 155         dst_release(t->dst_cache);
 156         t->dst_cache = dst;
 157 }
 158 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
 159
 160 /**
 161  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 162  *   @remote: the address of the tunnel exit-point
 163  *   @local: the address of the tunnel entry-point
 164  *
 165  * Return:
 166  *   tunnel matching given end-points if found,
 167  *   else fallback tunnel if its device is up,
 168  *   else %NULL
 169  **/
 170
 171 #define for_each_ip6_tunnel_rcu(start) \
 172         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 173
 174 static struct ip6_tnl *
 175 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 176 {
 177         unsigned int hash = HASH(remote, local);
 178         struct ip6_tnl *t;
 179         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 180         struct in6_addr any;
 181
 182         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 183                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 184                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 185                     (t->dev->flags & IFF_UP))
 186                         return t;
 187         }
 188
 189         memset(&any, 0, sizeof(any));
 190         hash = HASH(&any, local);
 191         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 192                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 193                     (t->dev->flags & IFF_UP))
 194                         return t;
 195         }
 196
 197         hash = HASH(remote, &any);
 198         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 199                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 200                     (t->dev->flags & IFF_UP))
 201                         return t;
 202         }
 203
 204         t = rcu_dereference(ip6n->tnls_wc[0]);
 205         if (t && (t->dev->flags & IFF_UP))
 206                 return t;
 207
 208         return NULL;
 209 }
 210
 211 /**
 212  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 213  *   @p: parameters containing tunnel end-points
 214  *
 215  * Description:
 216  *   ip6_tnl_bucket() returns the head of the list matching the
 217  *   &struct in6_addr entries laddr and raddr in @p.
 218  *
 219  * Return: head of IPv6 tunnel list
 220  **/
 221
 222 static struct ip6_tnl __rcu **
 223 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 224 {
 225         const struct in6_addr *remote = &p->raddr;
 226         const struct in6_addr *local = &p->laddr;
 227         unsigned int h = 0;
 228         int prio = 0;
 229
 230         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 231                 prio = 1;
 232                 h = HASH(remote, local);
 233         }
 234         return &ip6n->tnls[prio][h];
 235 }
 236
 237 /**
 238  * ip6_tnl_link - add tunnel to hash table
 239  *   @t: tunnel to be added
 240  **/
 241
 242 static void
 243 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 244 {
 245         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 246
 247         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 248         rcu_assign_pointer(*tp, t);
 249 }
 250
 251 /**
 252  * ip6_tnl_unlink - remove tunnel from hash table
 253  *   @t: tunnel to be removed
 254  **/
 255
 256 static void
 257 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 258 {
 259         struct ip6_tnl __rcu **tp;
 260         struct ip6_tnl *iter;
 261
 262         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 263              (iter = rtnl_dereference(*tp)) != NULL;
 264              tp = &iter->next) {
 265                 if (t == iter) {
 266                         rcu_assign_pointer(*tp, t->next);
 267                         break;
 268                 }
 269         }
 270 }
 271
 272 static void ip6_dev_free(struct net_device *dev)
 273 {
 274         free_percpu(dev->tstats);
 275         free_netdev(dev);
 276 }
 277
 278 static int ip6_tnl_create2(struct net_device *dev)
 279 {
 280         struct ip6_tnl *t = netdev_priv(dev);
 281         struct net *net = dev_net(dev);
 282         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 283         int err;
 284
 285         t = netdev_priv(dev);
 286
 287         err = register_netdevice(dev);
 288         if (err < 0)
 289                 goto out;
 290
 291         strcpy(t->parms.name, dev->name);
 292         dev->rtnl_link_ops = &ip6_link_ops;
 293
 294         dev_hold(dev);
 295         ip6_tnl_link(ip6n, t);
 296         return 0;
 297
 298 out:
 299         return err;
 300 }
 301
 302 /**
 303  * ip6_tnl_create - create a new tunnel
 304  *   @p: tunnel parameters
 305  *   @pt: pointer to new tunnel
 306  *
 307  * Description:
 308  *   Create tunnel matching given parameters.
 309  *
 310  * Return:
 311  *   created tunnel or error pointer
 312  **/
 313
 314 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 315 {
 316         struct net_device *dev;
 317         struct ip6_tnl *t;
 318         char name[IFNAMSIZ];
 319         int err = -ENOMEM;
 320
 321         if (p->name[0])
 322                 strlcpy(name, p->name, IFNAMSIZ);
 323         else
 324                 sprintf(name, "ip6tnl%%d");
 325
 326         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 327                            ip6_tnl_dev_setup);
 328         if (!dev)
 329                 goto failed;
 330
 331         dev_net_set(dev, net);
 332
 333         t = netdev_priv(dev);
 334         t->parms = *p;
 335         t->net = dev_net(dev);
 336         err = ip6_tnl_create2(dev);
 337         if (err < 0)
 338                 goto failed_free;
 339
 340         return t;
 341
 342 failed_free:
 343         ip6_dev_free(dev);
 344 failed:
 345         return ERR_PTR(err);
 346 }
 347
 348 /**
 349  * ip6_tnl_locate - find or create tunnel matching given parameters
 350  *   @p: tunnel parameters
 351  *   @create: != 0 if allowed to create new tunnel if no match found
 352  *
 353  * Description:
 354  *   ip6_tnl_locate() first tries to locate an existing tunnel
 355  *   based on @parms. If this is unsuccessful, but @create is set a new
 356  *   tunnel device is created and registered for use.
 357  *
 358  * Return:
 359  *   matching tunnel or error pointer
 360  **/
 361
 362 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 363                 struct __ip6_tnl_parm *p, int create)
 364 {
 365         const struct in6_addr *remote = &p->raddr;
 366         const struct in6_addr *local = &p->laddr;
 367         struct ip6_tnl __rcu **tp;
 368         struct ip6_tnl *t;
 369         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 370
 371         for (tp = ip6_tnl_bucket(ip6n, p);
 372              (t = rtnl_dereference(*tp)) != NULL;
 373              tp = &t->next) {
 374                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 375                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 376                         if (create)
 377                                 return ERR_PTR(-EEXIST);
 378
 379                         return t;
 380                 }
 381         }
 382         if (!create)
 383                 return ERR_PTR(-ENODEV);
 384         return ip6_tnl_create(net, p);
 385 }
 386
 387 /**
 388  * ip6_tnl_dev_uninit - tunnel device uninitializer
 389  *   @dev: the device to be destroyed
 390  *
 391  * Description:
 392  *   ip6_tnl_dev_uninit() removes tunnel from its list
 393  **/
 394
 395 static void
 396 ip6_tnl_dev_uninit(struct net_device *dev)
 397 {
 398         struct ip6_tnl *t = netdev_priv(dev);
 399         struct net *net = t->net;
 400         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 401
 402         if (dev == ip6n->fb_tnl_dev)
 403                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 404         else
 405                 ip6_tnl_unlink(ip6n, t);
 406         ip6_tnl_dst_reset(t);
 407         dev_put(dev);
 408 }
 409
 410 /**
 411  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 412  *   @skb: received socket buffer
 413  *
 414  * Return:
 415  *   0 if none was found,
 416  *   else index to encapsulation limit
 417  **/
 418
 419 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 420 {
 421         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 422         __u8 nexthdr = ipv6h->nexthdr;
 423         __u16 off = sizeof(*ipv6h);
 424
 425         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 426                 __u16 optlen = 0;
 427                 struct ipv6_opt_hdr *hdr;
 428                 if (raw + off + sizeof(*hdr) > skb->data &&
 429                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 430                         break;
 431
 432                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 433                 if (nexthdr == NEXTHDR_FRAGMENT) {
 434                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 435                         if (frag_hdr->frag_off)
 436                                 break;
 437                         optlen = 8;
 438                 } else if (nexthdr == NEXTHDR_AUTH) {
 439                         optlen = (hdr->hdrlen + 2) << 2;
 440                 } else {
 441                         optlen = ipv6_optlen(hdr);
 442                 }
 443                 if (nexthdr == NEXTHDR_DEST) {
 444                         __u16 i = off + 2;
 445                         while (1) {
 446                                 struct ipv6_tlv_tnl_enc_lim *tel;
 447
 448                                 /* No more room for encapsulation limit */
 449                                 if (i + sizeof (*tel) > off + optlen)
 450                                         break;
 451
 452                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 453                                 /* return index of option if found and valid */
 454                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 455                                     tel->length == 1)
 456                                         return i;
 457                                 /* else jump to next option */
 458                                 if (tel->type)
 459                                         i += tel->length + 2;
 460                                 else
 461                                         i++;
 462                         }
 463                 }
 464                 nexthdr = hdr->nexthdr;
 465                 off += optlen;
 466         }
 467         return 0;
 468 }
 469 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 470
 471 /**
 472  * ip6_tnl_err - tunnel error handler
 473  *
 474  * Description:
 475  *   ip6_tnl_err() should handle errors in the tunnel according
 476  *   to the specifications in RFC 2473.
 477  **/
 478
 479 static int
 480 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 481             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 482 {
 483         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 484         struct ip6_tnl *t;
 485         int rel_msg = 0;
 486         u8 rel_type = ICMPV6_DEST_UNREACH;
 487         u8 rel_code = ICMPV6_ADDR_UNREACH;
 488         u8 tproto;
 489         __u32 rel_info = 0;
 490         __u16 len;
 491         int err = -ENOENT;
 492
 493         /* If the packet doesn't contain the original IPv6 header we are
 494            in trouble since we might need the source address for further
 495            processing of the error. */
 496
 497         rcu_read_lock();
 498         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 499         if (!t)
 500                 goto out;
 501
 502         tproto = ACCESS_ONCE(t->parms.proto);
 503         if (tproto != ipproto && tproto != 0)
 504                 goto out;
 505
 506         err = 0;
 507
 508         switch (*type) {
 509                 __u32 teli;
 510                 struct ipv6_tlv_tnl_enc_lim *tel;
 511                 __u32 mtu;
 512         case ICMPV6_DEST_UNREACH:
 513                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 514                                      t->parms.name);
 515                 rel_msg = 1;
 516                 break;
 517         case ICMPV6_TIME_EXCEED:
 518                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 519                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 520                                              t->parms.name);
 521                         rel_msg = 1;
 522                 }
 523                 break;
 524         case ICMPV6_PARAMPROB:
 525                 teli = 0;
 526                 if ((*code) == ICMPV6_HDR_FIELD)
 527                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 528
 529                 if (teli && teli == *info - 2) {
 530                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 531                         if (tel->encap_limit == 0) {
 532                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 533                                                      t->parms.name);
 534                                 rel_msg = 1;
 535                         }
 536                 } else {
 537                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 538                                              t->parms.name);
 539                 }
 540                 break;
 541         case ICMPV6_PKT_TOOBIG:
 542                 mtu = *info - offset;
 543                 if (mtu < IPV6_MIN_MTU)
 544                         mtu = IPV6_MIN_MTU;
 545                 t->dev->mtu = mtu;
 546
 547                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 548                 if (len > mtu) {
 549                         rel_type = ICMPV6_PKT_TOOBIG;
 550                         rel_code = 0;
 551                         rel_info = mtu;
 552                         rel_msg = 1;
 553                 }
 554                 break;
 555         }
 556
 557         *type = rel_type;
 558         *code = rel_code;
 559         *info = rel_info;
 560         *msg = rel_msg;
 561
 562 out:
 563         rcu_read_unlock();
 564         return err;
 565 }
 566
 567 static int
 568 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 569            u8 type, u8 code, int offset, __be32 info)
 570 {
 571         int rel_msg = 0;
 572         u8 rel_type = type;
 573         u8 rel_code = code;
 574         __u32 rel_info = ntohl(info);
 575         int err;
 576         struct sk_buff *skb2;
 577         const struct iphdr *eiph;
 578         struct rtable *rt;
 579         struct flowi4 fl4;
 580
 581         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 582                           &rel_msg, &rel_info, offset);
 583         if (err < 0)
 584                 return err;
 585
 586         if (rel_msg == 0)
 587                 return 0;
 588
 589         switch (rel_type) {
 590         case ICMPV6_DEST_UNREACH:
 591                 if (rel_code != ICMPV6_ADDR_UNREACH)
 592                         return 0;
 593                 rel_type = ICMP_DEST_UNREACH;
 594                 rel_code = ICMP_HOST_UNREACH;
 595                 break;
 596         case ICMPV6_PKT_TOOBIG:
 597                 if (rel_code != 0)
 598                         return 0;
 599                 rel_type = ICMP_DEST_UNREACH;
 600                 rel_code = ICMP_FRAG_NEEDED;
 601                 break;
 602         case NDISC_REDIRECT:
 603                 rel_type = ICMP_REDIRECT;
 604                 rel_code = ICMP_REDIR_HOST;
 605         default:
 606                 return 0;
 607         }
 608
 609         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 610                 return 0;
 611
 612         skb2 = skb_clone(skb, GFP_ATOMIC);
 613         if (!skb2)
 614                 return 0;
 615
 616         skb_dst_drop(skb2);
 617
 618         skb_pull(skb2, offset);
 619         skb_reset_network_header(skb2);
 620         eiph = ip_hdr(skb2);
 621
 622         /* Try to guess incoming interface */
 623         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 624                                    eiph->saddr, 0,
 625                                    0, 0,
 626                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 627         if (IS_ERR(rt))
 628                 goto out;
 629
 630         skb2->dev = rt->dst.dev;
 631
 632         /* route "incoming" packet */
 633         if (rt->rt_flags & RTCF_LOCAL) {
 634                 ip_rt_put(rt);
 635                 rt = NULL;
 636                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 637                                            eiph->daddr, eiph->saddr,
 638                                            0, 0,
 639                                            IPPROTO_IPIP,
 640                                            RT_TOS(eiph->tos), 0);
 641                 if (IS_ERR(rt) ||
 642                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 643                         if (!IS_ERR(rt))
 644                                 ip_rt_put(rt);
 645                         goto out;
 646                 }
 647                 skb_dst_set(skb2, &rt->dst);
 648         } else {
 649                 ip_rt_put(rt);
 650                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 651                                    skb2->dev) ||
 652                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 653                         goto out;
 654         }
 655
 656         /* change mtu on this route */
 657         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 658                 if (rel_info > dst_mtu(skb_dst(skb2)))
 659                         goto out;
 660
 661                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 662         }
 663         if (rel_type == ICMP_REDIRECT)
 664                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 665
 666         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 667
 668 out:
 669         kfree_skb(skb2);
 670         return 0;
 671 }
 672
 673 static int
 674 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 675            u8 type, u8 code, int offset, __be32 info)
 676 {
 677         int rel_msg = 0;
 678         u8 rel_type = type;
 679         u8 rel_code = code;
 680         __u32 rel_info = ntohl(info);
 681         int err;
 682
 683         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 684                           &rel_msg, &rel_info, offset);
 685         if (err < 0)
 686                 return err;
 687
 688         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 689                 struct rt6_info *rt;
 690                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 691
 692                 if (!skb2)
 693                         return 0;
 694
 695                 skb_dst_drop(skb2);
 696                 skb_pull(skb2, offset);
 697                 skb_reset_network_header(skb2);
 698
 699                 /* Try to guess incoming interface */
 700                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 701                                 NULL, 0, 0);
 702
 703                 if (rt && rt->dst.dev)
 704                         skb2->dev = rt->dst.dev;
 705
 706                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 707
 708                 ip6_rt_put(rt);
 709
 710                 kfree_skb(skb2);
 711         }
 712
 713         return 0;
 714 }
 715
 716 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 717                                        const struct ipv6hdr *ipv6h,
 718                                        struct sk_buff *skb)
 719 {
 720         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 721
 722         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 723                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 724
 725         return IP6_ECN_decapsulate(ipv6h, skb);
 726 }
 727
 728 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 729                                        const struct ipv6hdr *ipv6h,
 730                                        struct sk_buff *skb)
 731 {
 732         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 733                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 734
 735         return IP6_ECN_decapsulate(ipv6h, skb);
 736 }
 737
 738 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 739                              const struct in6_addr *laddr,
 740                              const struct in6_addr *raddr)
 741 {
 742         struct __ip6_tnl_parm *p = &t->parms;
 743         int ltype = ipv6_addr_type(laddr);
 744         int rtype = ipv6_addr_type(raddr);
 745         __u32 flags = 0;
 746
 747         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 748                 flags = IP6_TNL_F_CAP_PER_PACKET;
 749         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 750                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 751                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 752                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 753                 if (ltype&IPV6_ADDR_UNICAST)
 754                         flags |= IP6_TNL_F_CAP_XMIT;
 755                 if (rtype&IPV6_ADDR_UNICAST)
 756                         flags |= IP6_TNL_F_CAP_RCV;
 757         }
 758         return flags;
 759 }
 760 EXPORT_SYMBOL(ip6_tnl_get_cap);
 761
 762 /* called with rcu_read_lock() */
 763 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 764                                   const struct in6_addr *laddr,
 765                                   const struct in6_addr *raddr)
 766 {
 767         struct __ip6_tnl_parm *p = &t->parms;
 768         int ret = 0;
 769         struct net *net = t->net;
 770
 771         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 772             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 773              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 774                 struct net_device *ldev = NULL;
 775
 776                 if (p->link)
 777                         ldev = dev_get_by_index_rcu(net, p->link);
 778
 779                 if ((ipv6_addr_is_multicast(laddr) ||
 780                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 781                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 782                         ret = 1;
 783         }
 784         return ret;
 785 }
 786 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 787
 788 /**
 789  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 790  *   @skb: received socket buffer
 791  *   @protocol: ethernet protocol ID
 792  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 793  *
 794  * Return: 0
 795  **/
 796
 797 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 798                        __u8 ipproto,
 799                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 800                                                    const struct ipv6hdr *ipv6h,
 801                                                    struct sk_buff *skb))
 802 {
 803         struct ip6_tnl *t;
 804         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 805         u8 tproto;
 806         int err;
 807
 808         rcu_read_lock();
 809         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 810         if (t) {
 811                 struct pcpu_sw_netstats *tstats;
 812
 813                 tproto = ACCESS_ONCE(t->parms.proto);
 814                 if (tproto != ipproto && tproto != 0) {
 815                         rcu_read_unlock();
 816                         goto discard;
 817                 }
 818
 819                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 820                         rcu_read_unlock();
 821                         goto discard;
 822                 }
 823
 824                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
 825                         t->dev->stats.rx_dropped++;
 826                         rcu_read_unlock();
 827                         goto discard;
 828                 }
 829                 skb->mac_header = skb->network_header;
 830                 skb_reset_network_header(skb);
 831                 skb->protocol = htons(protocol);
 832                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 833
 834                 __skb_tunnel_rx(skb, t->dev, t->net);
 835
 836                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
 837                 if (unlikely(err)) {
 838                         if (log_ecn_error)
 839                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 840                                                      &ipv6h->saddr,
 841                                                      ipv6_get_dsfield(ipv6h));
 842                         if (err > 1) {
 843                                 ++t->dev->stats.rx_frame_errors;
 844                                 ++t->dev->stats.rx_errors;
 845                                 rcu_read_unlock();
 846                                 goto discard;
 847                         }
 848                 }
 849
 850                 tstats = this_cpu_ptr(t->dev->tstats);
 851                 u64_stats_update_begin(&tstats->syncp);
 852                 tstats->rx_packets++;
 853                 tstats->rx_bytes += skb->len;
 854                 u64_stats_update_end(&tstats->syncp);
 855
 856                 netif_rx(skb);
 857
 858                 rcu_read_unlock();
 859                 return 0;
 860         }
 861         rcu_read_unlock();
 862         return 1;
 863
 864 discard:
 865         kfree_skb(skb);
 866         return 0;
 867 }
 868
 869 static int ip4ip6_rcv(struct sk_buff *skb)
 870 {
 871         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 872                            ip4ip6_dscp_ecn_decapsulate);
 873 }
 874
 875 static int ip6ip6_rcv(struct sk_buff *skb)
 876 {
 877         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 878                            ip6ip6_dscp_ecn_decapsulate);
 879 }
 880
 881 struct ipv6_tel_txoption {
 882         struct ipv6_txoptions ops;
 883         __u8 dst_opt[8];
 884 };
 885
 886 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 887 {
 888         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 889
 890         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 891         opt->dst_opt[3] = 1;
 892         opt->dst_opt[4] = encap_limit;
 893         opt->dst_opt[5] = IPV6_TLV_PADN;
 894         opt->dst_opt[6] = 1;
 895
 896         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 897         opt->ops.opt_nflen = 8;
 898 }
 899
 900 /**
 901  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 902  *   @t: the outgoing tunnel device
 903  *   @hdr: IPv6 header from the incoming packet
 904  *
 905  * Description:
 906  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 907  *   doesn't match source of incoming packet.
 908  *
 909  * Return:
 910  *   1 if conflict,
 911  *   0 else
 912  **/
 913
 914 static inline bool
 915 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 916 {
 917         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 918 }
 919
 920 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 921                      const struct in6_addr *laddr,
 922                      const struct in6_addr *raddr)
 923 {
 924         struct __ip6_tnl_parm *p = &t->parms;
 925         int ret = 0;
 926         struct net *net = t->net;
 927
 928         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 929             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 930              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 931                 struct net_device *ldev = NULL;
 932
 933                 rcu_read_lock();
 934                 if (p->link)
 935                         ldev = dev_get_by_index_rcu(net, p->link);
 936
 937                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 938                         pr_warn("%s xmit: Local address not yet configured!\n",
 939                                 p->name);
 940                 else if (!ipv6_addr_is_multicast(raddr) &&
 941                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
 942                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 943                                 p->name);
 944                 else
 945                         ret = 1;
 946                 rcu_read_unlock();
 947         }
 948         return ret;
 949 }
 950 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
 951
 952 /**
 953  * ip6_tnl_xmit2 - encapsulate packet and send
 954  *   @skb: the outgoing socket buffer
 955  *   @dev: the outgoing tunnel device
 956  *   @dsfield: dscp code for outer header
 957  *   @fl: flow of tunneled packet
 958  *   @encap_limit: encapsulation limit
 959  *   @pmtu: Path MTU is stored if packet is too big
 960  *
 961  * Description:
 962  *   Build new header and do some sanity checks on the packet before sending
 963  *   it.
 964  *
 965  * Return:
 966  *   0 on success
 967  *   -1 fail
 968  *   %-EMSGSIZE message too big. return mtu in this case.
 969  **/
 970
 971 static int ip6_tnl_xmit2(struct sk_buff *skb,
 972                          struct net_device *dev,
 973                          __u8 dsfield,
 974                          struct flowi6 *fl6,
 975                          int encap_limit,
 976                          __u32 *pmtu)
 977 {
 978         struct ip6_tnl *t = netdev_priv(dev);
 979         struct net *net = t->net;
 980         struct net_device_stats *stats = &t->dev->stats;
 981         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 982         struct ipv6_tel_txoption opt;
 983         struct dst_entry *dst = NULL, *ndst = NULL;
 984         struct net_device *tdev;
 985         int mtu;
 986         unsigned int max_headroom = sizeof(struct ipv6hdr);
 987         u8 proto;
 988         int err = -1;
 989
 990         /* NBMA tunnel */
 991         if (ipv6_addr_any(&t->parms.raddr)) {
 992                 struct in6_addr *addr6;
 993                 struct neighbour *neigh;
 994                 int addr_type;
 995
 996                 if (!skb_dst(skb))
 997                         goto tx_err_link_failure;
 998
 999                 neigh = dst_neigh_lookup(skb_dst(skb),
1000                                          &ipv6_hdr(skb)->daddr);
1001                 if (!neigh)
1002                         goto tx_err_link_failure;
1003
1004                 addr6 = (struct in6_addr *)&neigh->primary_key;
1005                 addr_type = ipv6_addr_type(addr6);
1006
1007                 if (addr_type == IPV6_ADDR_ANY)
1008                         addr6 = &ipv6_hdr(skb)->daddr;
1009
1010                 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1011                 neigh_release(neigh);
1012         } else if (!fl6->flowi6_mark)
1013                 dst = ip6_tnl_dst_check(t);
1014
1015         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1016                 goto tx_err_link_failure;
1017
1018         if (!dst) {
1019                 ndst = ip6_route_output(net, NULL, fl6);
1020
1021                 if (ndst->error)
1022                         goto tx_err_link_failure;
1023                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
1024                 if (IS_ERR(ndst)) {
1025                         err = PTR_ERR(ndst);
1026                         ndst = NULL;
1027                         goto tx_err_link_failure;
1028                 }
1029                 dst = ndst;
1030         }
1031
1032         tdev = dst->dev;
1033
1034         if (tdev == dev) {
1035                 stats->collisions++;
1036                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1037                                      t->parms.name);
1038                 goto tx_err_dst_release;
1039         }
1040         mtu = dst_mtu(dst) - sizeof(*ipv6h);
1041         if (encap_limit >= 0) {
1042                 max_headroom += 8;
1043                 mtu -= 8;
1044         }
1045         if (mtu < IPV6_MIN_MTU)
1046                 mtu = IPV6_MIN_MTU;
1047         if (skb_dst(skb))
1048                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1049         if (skb->len > mtu) {
1050                 *pmtu = mtu;
1051                 err = -EMSGSIZE;
1052                 goto tx_err_dst_release;
1053         }
1054
1055         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1056
1057         /*
1058          * Okay, now see if we can stuff it in the buffer as-is.
1059          */
1060         max_headroom += LL_RESERVED_SPACE(tdev);
1061
1062         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1063             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1064                 struct sk_buff *new_skb;
1065
1066                 new_skb = skb_realloc_headroom(skb, max_headroom);
1067                 if (!new_skb)
1068                         goto tx_err_dst_release;
1069
1070                 if (skb->sk)
1071                         skb_set_owner_w(new_skb, skb->sk);
1072                 consume_skb(skb);
1073                 skb = new_skb;
1074         }
1075         if (fl6->flowi6_mark) {
1076                 skb_dst_set(skb, dst);
1077                 ndst = NULL;
1078         } else {
1079                 skb_dst_set_noref(skb, dst);
1080         }
1081         skb->transport_header = skb->network_header;
1082
1083         proto = fl6->flowi6_proto;
1084         if (encap_limit >= 0) {
1085                 init_tel_txopt(&opt, encap_limit);
1086                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1087         }
1088
1089         if (likely(!skb->encapsulation)) {
1090                 skb_reset_inner_headers(skb);
1091                 skb->encapsulation = 1;
1092         }
1093
1094         skb_push(skb, sizeof(struct ipv6hdr));
1095         skb_reset_network_header(skb);
1096         ipv6h = ipv6_hdr(skb);
1097         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1098                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1099         ipv6h->hop_limit = t->parms.hop_limit;
1100         ipv6h->nexthdr = proto;
1101         ipv6h->saddr = fl6->saddr;
1102         ipv6h->daddr = fl6->daddr;
1103         ip6tunnel_xmit(NULL, skb, dev);
1104         if (ndst)
1105                 ip6_tnl_dst_store(t, ndst);
1106         return 0;
1107 tx_err_link_failure:
1108         stats->tx_carrier_errors++;
1109         dst_link_failure(skb);
1110 tx_err_dst_release:
1111         dst_release(ndst);
1112         return err;
1113 }
1114
1115 static inline int
1116 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1117 {
1118         struct ip6_tnl *t = netdev_priv(dev);
1119         const struct iphdr  *iph = ip_hdr(skb);
1120         int encap_limit = -1;
1121         struct flowi6 fl6;
1122         __u8 dsfield;
1123         __u32 mtu;
1124         u8 tproto;
1125         int err;
1126
1127         tproto = ACCESS_ONCE(t->parms.proto);
1128         if (tproto != IPPROTO_IPIP && tproto != 0)
1129                 return -1;
1130
1131         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1132                 encap_limit = t->parms.encap_limit;
1133
1134         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1135         fl6.flowi6_proto = IPPROTO_IPIP;
1136
1137         dsfield = ipv4_get_dsfield(iph);
1138
1139         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1140                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1141                                           & IPV6_TCLASS_MASK;
1142         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1143                 fl6.flowi6_mark = skb->mark;
1144
1145         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1146         if (err != 0) {
1147                 /* XXX: send ICMP error even if DF is not set. */
1148                 if (err == -EMSGSIZE)
1149                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1150                                   htonl(mtu));
1151                 return -1;
1152         }
1153
1154         return 0;
1155 }
1156
1157 static inline int
1158 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1159 {
1160         struct ip6_tnl *t = netdev_priv(dev);
1161         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1162         int encap_limit = -1;
1163         __u16 offset;
1164         struct flowi6 fl6;
1165         __u8 dsfield;
1166         __u32 mtu;
1167         u8 tproto;
1168         int err;
1169
1170         tproto = ACCESS_ONCE(t->parms.proto);
1171         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1172             ip6_tnl_addr_conflict(t, ipv6h))
1173                 return -1;
1174
1175         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1176         if (offset > 0) {
1177                 struct ipv6_tlv_tnl_enc_lim *tel;
1178                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1179                 if (tel->encap_limit == 0) {
1180                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1181                                     ICMPV6_HDR_FIELD, offset + 2);
1182                         return -1;
1183                 }
1184                 encap_limit = tel->encap_limit - 1;
1185         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1186                 encap_limit = t->parms.encap_limit;
1187
1188         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1189         fl6.flowi6_proto = IPPROTO_IPV6;
1190
1191         dsfield = ipv6_get_dsfield(ipv6h);
1192         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1193                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1194         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1195                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1196         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1197                 fl6.flowi6_mark = skb->mark;
1198
1199         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1200         if (err != 0) {
1201                 if (err == -EMSGSIZE)
1202                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1203                 return -1;
1204         }
1205
1206         return 0;
1207 }
1208
1209 static netdev_tx_t
1210 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1211 {
1212         struct ip6_tnl *t = netdev_priv(dev);
1213         struct net_device_stats *stats = &t->dev->stats;
1214         int ret;
1215
1216         switch (skb->protocol) {
1217         case htons(ETH_P_IP):
1218                 ret = ip4ip6_tnl_xmit(skb, dev);
1219                 break;
1220         case htons(ETH_P_IPV6):
1221                 ret = ip6ip6_tnl_xmit(skb, dev);
1222                 break;
1223         default:
1224                 goto tx_err;
1225         }
1226
1227         if (ret < 0)
1228                 goto tx_err;
1229
1230         return NETDEV_TX_OK;
1231
1232 tx_err:
1233         stats->tx_errors++;
1234         stats->tx_dropped++;
1235         kfree_skb(skb);
1236         return NETDEV_TX_OK;
1237 }
1238
1239 static void ip6_tnl_link_config(struct ip6_tnl *t)
1240 {
1241         struct net_device *dev = t->dev;
1242         struct __ip6_tnl_parm *p = &t->parms;
1243         struct flowi6 *fl6 = &t->fl.u.ip6;
1244
1245         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1246         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1247
1248         /* Set up flowi template */
1249         fl6->saddr = p->laddr;
1250         fl6->daddr = p->raddr;
1251         fl6->flowi6_oif = p->link;
1252         fl6->flowlabel = 0;
1253
1254         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1255                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1256         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1257                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1258
1259         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1260         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1261
1262         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1263                 dev->flags |= IFF_POINTOPOINT;
1264         else
1265                 dev->flags &= ~IFF_POINTOPOINT;
1266
1267         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1268                 int strict = (ipv6_addr_type(&p->raddr) &
1269                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1270
1271                 struct rt6_info *rt = rt6_lookup(t->net,
1272                                                  &p->raddr, &p->laddr,
1273                                                  p->link, strict);
1274
1275                 if (!rt)
1276                         return;
1277
1278                 if (rt->dst.dev) {
1279                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1280                                 sizeof(struct ipv6hdr);
1281
1282                         dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1283                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1284                                 dev->mtu -= 8;
1285
1286                         if (dev->mtu < IPV6_MIN_MTU)
1287                                 dev->mtu = IPV6_MIN_MTU;
1288                 }
1289                 ip6_rt_put(rt);
1290         }
1291 }
1292
1293 /**
1294  * ip6_tnl_change - update the tunnel parameters
1295  *   @t: tunnel to be changed
1296  *   @p: tunnel configuration parameters
1297  *
1298  * Description:
1299  *   ip6_tnl_change() updates the tunnel parameters
1300  **/
1301
1302 static int
1303 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1304 {
1305         t->parms.laddr = p->laddr;
1306         t->parms.raddr = p->raddr;
1307         t->parms.flags = p->flags;
1308         t->parms.hop_limit = p->hop_limit;
1309         t->parms.encap_limit = p->encap_limit;
1310         t->parms.flowinfo = p->flowinfo;
1311         t->parms.link = p->link;
1312         t->parms.proto = p->proto;
1313         ip6_tnl_dst_reset(t);
1314         ip6_tnl_link_config(t);
1315         return 0;
1316 }
1317
1318 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1319 {
1320         struct net *net = t->net;
1321         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1322         int err;
1323
1324         ip6_tnl_unlink(ip6n, t);
1325         synchronize_net();
1326         err = ip6_tnl_change(t, p);
1327         ip6_tnl_link(ip6n, t);
1328         netdev_state_change(t->dev);
1329         return err;
1330 }
1331
1332 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1333 {
1334         /* for default tnl0 device allow to change only the proto */
1335         t->parms.proto = p->proto;
1336         netdev_state_change(t->dev);
1337         return 0;
1338 }
1339
1340 static void
1341 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1342 {
1343         p->laddr = u->laddr;
1344         p->raddr = u->raddr;
1345         p->flags = u->flags;
1346         p->hop_limit = u->hop_limit;
1347         p->encap_limit = u->encap_limit;
1348         p->flowinfo = u->flowinfo;
1349         p->link = u->link;
1350         p->proto = u->proto;
1351         memcpy(p->name, u->name, sizeof(u->name));
1352 }
1353
1354 static void
1355 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1356 {
1357         u->laddr = p->laddr;
1358         u->raddr = p->raddr;
1359         u->flags = p->flags;
1360         u->hop_limit = p->hop_limit;
1361         u->encap_limit = p->encap_limit;
1362         u->flowinfo = p->flowinfo;
1363         u->link = p->link;
1364         u->proto = p->proto;
1365         memcpy(u->name, p->name, sizeof(u->name));
1366 }
1367
1368 /**
1369  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1370  *   @dev: virtual device associated with tunnel
1371  *   @ifr: parameters passed from userspace
1372  *   @cmd: command to be performed
1373  *
1374  * Description:
1375  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1376  *   from userspace.
1377  *
1378  *   The possible commands are the following:
1379  *     %SIOCGETTUNNEL: get tunnel parameters for device
1380  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1381  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1382  *     %SIOCDELTUNNEL: delete tunnel
1383  *
1384  *   The fallback device "ip6tnl0", created during module
1385  *   initialization, can be used for creating other tunnel devices.
1386  *
1387  * Return:
1388  *   0 on success,
1389  *   %-EFAULT if unable to copy data to or from userspace,
1390  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1391  *   %-EINVAL if passed tunnel parameters are invalid,
1392  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1393  *   %-ENODEV if attempting to change or delete a nonexisting device
1394  **/
1395
1396 static int
1397 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1398 {
1399         int err = 0;
1400         struct ip6_tnl_parm p;
1401         struct __ip6_tnl_parm p1;
1402         struct ip6_tnl *t = netdev_priv(dev);
1403         struct net *net = t->net;
1404         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1405
1406         switch (cmd) {
1407         case SIOCGETTUNNEL:
1408                 if (dev == ip6n->fb_tnl_dev) {
1409                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1410                                 err = -EFAULT;
1411                                 break;
1412                         }
1413                         ip6_tnl_parm_from_user(&p1, &p);
1414                         t = ip6_tnl_locate(net, &p1, 0);
1415                         if (IS_ERR(t))
1416                                 t = netdev_priv(dev);
1417                 } else {
1418                         memset(&p, 0, sizeof(p));
1419                 }
1420                 ip6_tnl_parm_to_user(&p, &t->parms);
1421                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1422                         err = -EFAULT;
1423                 }
1424                 break;
1425         case SIOCADDTUNNEL:
1426         case SIOCCHGTUNNEL:
1427                 err = -EPERM;
1428                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1429                         break;
1430                 err = -EFAULT;
1431                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1432                         break;
1433                 err = -EINVAL;
1434                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1435                     p.proto != 0)
1436                         break;
1437                 ip6_tnl_parm_from_user(&p1, &p);
1438                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1439                 if (cmd == SIOCCHGTUNNEL) {
1440                         if (!IS_ERR(t)) {
1441                                 if (t->dev != dev) {
1442                                         err = -EEXIST;
1443                                         break;
1444                                 }
1445                         } else
1446                                 t = netdev_priv(dev);
1447                         if (dev == ip6n->fb_tnl_dev)
1448                                 err = ip6_tnl0_update(t, &p1);
1449                         else
1450                                 err = ip6_tnl_update(t, &p1);
1451                 }
1452                 if (!IS_ERR(t)) {
1453                         err = 0;
1454                         ip6_tnl_parm_to_user(&p, &t->parms);
1455                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1456                                 err = -EFAULT;
1457
1458                 } else {
1459                         err = PTR_ERR(t);
1460                 }
1461                 break;
1462         case SIOCDELTUNNEL:
1463                 err = -EPERM;
1464                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1465                         break;
1466
1467                 if (dev == ip6n->fb_tnl_dev) {
1468                         err = -EFAULT;
1469                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1470                                 break;
1471                         err = -ENOENT;
1472                         ip6_tnl_parm_from_user(&p1, &p);
1473                         t = ip6_tnl_locate(net, &p1, 0);
1474                         if (IS_ERR(t))
1475                                 break;
1476                         err = -EPERM;
1477                         if (t->dev == ip6n->fb_tnl_dev)
1478                                 break;
1479                         dev = t->dev;
1480                 }
1481                 err = 0;
1482                 unregister_netdevice(dev);
1483                 break;
1484         default:
1485                 err = -EINVAL;
1486         }
1487         return err;
1488 }
1489
1490 /**
1491  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1492  *   @dev: virtual device associated with tunnel
1493  *   @new_mtu: the new mtu
1494  *
1495  * Return:
1496  *   0 on success,
1497  *   %-EINVAL if mtu too small
1498  **/
1499
1500 static int
1501 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1502 {
1503         struct ip6_tnl *tnl = netdev_priv(dev);
1504
1505         if (tnl->parms.proto == IPPROTO_IPIP) {
1506                 if (new_mtu < 68)
1507                         return -EINVAL;
1508         } else {
1509                 if (new_mtu < IPV6_MIN_MTU)
1510                         return -EINVAL;
1511         }
1512         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1513                 return -EINVAL;
1514         dev->mtu = new_mtu;
1515         return 0;
1516 }
1517
1518 int ip6_tnl_get_iflink(const struct net_device *dev)
1519 {
1520         struct ip6_tnl *t = netdev_priv(dev);
1521
1522         return t->parms.link;
1523 }
1524 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1525
1526 static const struct net_device_ops ip6_tnl_netdev_ops = {
1527         .ndo_init       = ip6_tnl_dev_init,
1528         .ndo_uninit     = ip6_tnl_dev_uninit,
1529         .ndo_start_xmit = ip6_tnl_xmit,
1530         .ndo_do_ioctl   = ip6_tnl_ioctl,
1531         .ndo_change_mtu = ip6_tnl_change_mtu,
1532         .ndo_get_stats  = ip6_get_stats,
1533         .ndo_get_iflink = ip6_tnl_get_iflink,
1534 };
1535
1536
1537 /**
1538  * ip6_tnl_dev_setup - setup virtual tunnel device
1539  *   @dev: virtual device associated with tunnel
1540  *
1541  * Description:
1542  *   Initialize function pointers and device parameters
1543  **/
1544
1545 static void ip6_tnl_dev_setup(struct net_device *dev)
1546 {
1547         struct ip6_tnl *t;
1548
1549         dev->netdev_ops = &ip6_tnl_netdev_ops;
1550         dev->destructor = ip6_dev_free;
1551
1552         dev->type = ARPHRD_TUNNEL6;
1553         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1554         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1555         t = netdev_priv(dev);
1556         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1557                 dev->mtu -= 8;
1558         dev->flags |= IFF_NOARP;
1559         dev->addr_len = sizeof(struct in6_addr);
1560         netif_keep_dst(dev);
1561         /* This perm addr will be used as interface identifier by IPv6 */
1562         dev->addr_assign_type = NET_ADDR_RANDOM;
1563         eth_random_addr(dev->perm_addr);
1564 }
1565
1566
1567 /**
1568  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1569  *   @dev: virtual device associated with tunnel
1570  **/
1571
1572 static inline int
1573 ip6_tnl_dev_init_gen(struct net_device *dev)
1574 {
1575         struct ip6_tnl *t = netdev_priv(dev);
1576
1577         t->dev = dev;
1578         t->net = dev_net(dev);
1579         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1580         if (!dev->tstats)
1581                 return -ENOMEM;
1582         return 0;
1583 }
1584
1585 /**
1586  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1587  *   @dev: virtual device associated with tunnel
1588  **/
1589
1590 static int ip6_tnl_dev_init(struct net_device *dev)
1591 {
1592         struct ip6_tnl *t = netdev_priv(dev);
1593         int err = ip6_tnl_dev_init_gen(dev);
1594
1595         if (err)
1596                 return err;
1597         ip6_tnl_link_config(t);
1598         return 0;
1599 }
1600
1601 /**
1602  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1603  *   @dev: fallback device
1604  *
1605  * Return: 0
1606  **/
1607
1608 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1609 {
1610         struct ip6_tnl *t = netdev_priv(dev);
1611         struct net *net = dev_net(dev);
1612         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1613
1614         t->parms.proto = IPPROTO_IPV6;
1615         dev_hold(dev);
1616
1617         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1618         return 0;
1619 }
1620
1621 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1622 {
1623         u8 proto;
1624
1625         if (!data || !data[IFLA_IPTUN_PROTO])
1626                 return 0;
1627
1628         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1629         if (proto != IPPROTO_IPV6 &&
1630             proto != IPPROTO_IPIP &&
1631             proto != 0)
1632                 return -EINVAL;
1633
1634         return 0;
1635 }
1636
1637 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1638                                   struct __ip6_tnl_parm *parms)
1639 {
1640         memset(parms, 0, sizeof(*parms));
1641
1642         if (!data)
1643                 return;
1644
1645         if (data[IFLA_IPTUN_LINK])
1646                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1647
1648         if (data[IFLA_IPTUN_LOCAL])
1649                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1650
1651         if (data[IFLA_IPTUN_REMOTE])
1652                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1653
1654         if (data[IFLA_IPTUN_TTL])
1655                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1656
1657         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1658                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1659
1660         if (data[IFLA_IPTUN_FLOWINFO])
1661                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1662
1663         if (data[IFLA_IPTUN_FLAGS])
1664                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1665
1666         if (data[IFLA_IPTUN_PROTO])
1667                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1668 }
1669
1670 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1671                            struct nlattr *tb[], struct nlattr *data[])
1672 {
1673         struct net *net = dev_net(dev);
1674         struct ip6_tnl *nt, *t;
1675
1676         nt = netdev_priv(dev);
1677         ip6_tnl_netlink_parms(data, &nt->parms);
1678
1679         t = ip6_tnl_locate(net, &nt->parms, 0);
1680         if (!IS_ERR(t))
1681                 return -EEXIST;
1682
1683         return ip6_tnl_create2(dev);
1684 }
1685
1686 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1687                               struct nlattr *data[])
1688 {
1689         struct ip6_tnl *t = netdev_priv(dev);
1690         struct __ip6_tnl_parm p;
1691         struct net *net = t->net;
1692         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1693
1694         if (dev == ip6n->fb_tnl_dev)
1695                 return -EINVAL;
1696
1697         ip6_tnl_netlink_parms(data, &p);
1698
1699         t = ip6_tnl_locate(net, &p, 0);
1700         if (!IS_ERR(t)) {
1701                 if (t->dev != dev)
1702                         return -EEXIST;
1703         } else
1704                 t = netdev_priv(dev);
1705
1706         return ip6_tnl_update(t, &p);
1707 }
1708
1709 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1710 {
1711         struct net *net = dev_net(dev);
1712         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1713
1714         if (dev != ip6n->fb_tnl_dev)
1715                 unregister_netdevice_queue(dev, head);
1716 }
1717
1718 static size_t ip6_tnl_get_size(const struct net_device *dev)
1719 {
1720         return
1721                 /* IFLA_IPTUN_LINK */
1722                 nla_total_size(4) +
1723                 /* IFLA_IPTUN_LOCAL */
1724                 nla_total_size(sizeof(struct in6_addr)) +
1725                 /* IFLA_IPTUN_REMOTE */
1726                 nla_total_size(sizeof(struct in6_addr)) +
1727                 /* IFLA_IPTUN_TTL */
1728                 nla_total_size(1) +
1729                 /* IFLA_IPTUN_ENCAP_LIMIT */
1730                 nla_total_size(1) +
1731                 /* IFLA_IPTUN_FLOWINFO */
1732                 nla_total_size(4) +
1733                 /* IFLA_IPTUN_FLAGS */
1734                 nla_total_size(4) +
1735                 /* IFLA_IPTUN_PROTO */
1736                 nla_total_size(1) +
1737                 0;
1738 }
1739
1740 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1741 {
1742         struct ip6_tnl *tunnel = netdev_priv(dev);
1743         struct __ip6_tnl_parm *parm = &tunnel->parms;
1744
1745         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1746             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1747             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1748             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1749             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1750             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1751             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1752             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1753                 goto nla_put_failure;
1754         return 0;
1755
1756 nla_put_failure:
1757         return -EMSGSIZE;
1758 }
1759
1760 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1761 {
1762         struct ip6_tnl *tunnel = netdev_priv(dev);
1763
1764         return tunnel->net;
1765 }
1766 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1767
1768 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1769         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1770         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1771         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1772         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1773         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1774         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1775         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1776         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1777 };
1778
1779 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1780         .kind           = "ip6tnl",
1781         .maxtype        = IFLA_IPTUN_MAX,
1782         .policy         = ip6_tnl_policy,
1783         .priv_size      = sizeof(struct ip6_tnl),
1784         .setup          = ip6_tnl_dev_setup,
1785         .validate       = ip6_tnl_validate,
1786         .newlink        = ip6_tnl_newlink,
1787         .changelink     = ip6_tnl_changelink,
1788         .dellink        = ip6_tnl_dellink,
1789         .get_size       = ip6_tnl_get_size,
1790         .fill_info      = ip6_tnl_fill_info,
1791         .get_link_net   = ip6_tnl_get_link_net,
1792 };
1793
1794 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1795         .handler        = ip4ip6_rcv,
1796         .err_handler    = ip4ip6_err,
1797         .priority       =       1,
1798 };
1799
1800 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1801         .handler        = ip6ip6_rcv,
1802         .err_handler    = ip6ip6_err,
1803         .priority       =       1,
1804 };
1805
1806 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1807 {
1808         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1809         struct net_device *dev, *aux;
1810         int h;
1811         struct ip6_tnl *t;
1812         LIST_HEAD(list);
1813
1814         for_each_netdev_safe(net, dev, aux)
1815                 if (dev->rtnl_link_ops == &ip6_link_ops)
1816                         unregister_netdevice_queue(dev, &list);
1817
1818         for (h = 0; h < HASH_SIZE; h++) {
1819                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1820                 while (t) {
1821                         /* If dev is in the same netns, it has already
1822                          * been added to the list by the previous loop.
1823                          */
1824                         if (!net_eq(dev_net(t->dev), net))
1825                                 unregister_netdevice_queue(t->dev, &list);
1826                         t = rtnl_dereference(t->next);
1827                 }
1828         }
1829
1830         unregister_netdevice_many(&list);
1831 }
1832
1833 static int __net_init ip6_tnl_init_net(struct net *net)
1834 {
1835         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1836         struct ip6_tnl *t = NULL;
1837         int err;
1838
1839         ip6n->tnls[0] = ip6n->tnls_wc;
1840         ip6n->tnls[1] = ip6n->tnls_r_l;
1841
1842         err = -ENOMEM;
1843         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1844                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1845
1846         if (!ip6n->fb_tnl_dev)
1847                 goto err_alloc_dev;
1848         dev_net_set(ip6n->fb_tnl_dev, net);
1849         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1850         /* FB netdevice is special: we have one, and only one per netns.
1851          * Allowing to move it to another netns is clearly unsafe.
1852          */
1853         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1854
1855         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1856         if (err < 0)
1857                 goto err_register;
1858
1859         err = register_netdev(ip6n->fb_tnl_dev);
1860         if (err < 0)
1861                 goto err_register;
1862
1863         t = netdev_priv(ip6n->fb_tnl_dev);
1864
1865         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1866         return 0;
1867
1868 err_register:
1869         ip6_dev_free(ip6n->fb_tnl_dev);
1870 err_alloc_dev:
1871         return err;
1872 }
1873
1874 static void __net_exit ip6_tnl_exit_net(struct net *net)
1875 {
1876         rtnl_lock();
1877         ip6_tnl_destroy_tunnels(net);
1878         rtnl_unlock();
1879 }
1880
1881 static struct pernet_operations ip6_tnl_net_ops = {
1882         .init = ip6_tnl_init_net,
1883         .exit = ip6_tnl_exit_net,
1884         .id   = &ip6_tnl_net_id,
1885         .size = sizeof(struct ip6_tnl_net),
1886 };
1887
1888 /**
1889  * ip6_tunnel_init - register protocol and reserve needed resources
1890  *
1891  * Return: 0 on success
1892  **/
1893
1894 static int __init ip6_tunnel_init(void)
1895 {
1896         int  err;
1897
1898         err = register_pernet_device(&ip6_tnl_net_ops);
1899         if (err < 0)
1900                 goto out_pernet;
1901
1902         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1903         if (err < 0) {
1904                 pr_err("%s: can't register ip4ip6\n", __func__);
1905                 goto out_ip4ip6;
1906         }
1907
1908         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1909         if (err < 0) {
1910                 pr_err("%s: can't register ip6ip6\n", __func__);
1911                 goto out_ip6ip6;
1912         }
1913         err = rtnl_link_register(&ip6_link_ops);
1914         if (err < 0)
1915                 goto rtnl_link_failed;
1916
1917         return 0;
1918
1919 rtnl_link_failed:
1920         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1921 out_ip6ip6:
1922         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1923 out_ip4ip6:
1924         unregister_pernet_device(&ip6_tnl_net_ops);
1925 out_pernet:
1926         return err;
1927 }
1928
1929 /**
1930  * ip6_tunnel_cleanup - free resources and unregister protocol
1931  **/
1932
1933 static void __exit ip6_tunnel_cleanup(void)
1934 {
1935         rtnl_link_unregister(&ip6_link_ops);
1936         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1937                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1938
1939         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1940                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1941
1942         unregister_pernet_device(&ip6_tnl_net_ops);
1943 }
1944
1945 module_init(ip6_tunnel_init);
1946 module_exit(ip6_tunnel_cleanup);