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         struct in6_addr any;
 187
 188         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 189                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 190                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 191                     (t->dev->flags & IFF_UP))
 192                         return t;
 193         }
 194
 195         memset(&any, 0, sizeof(any));
 196         hash = HASH(&any, local);
 197         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 198                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 199                     (t->dev->flags & IFF_UP))
 200                         return t;
 201         }
 202
 203         hash = HASH(remote, &any);
 204         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 205                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 206                     (t->dev->flags & IFF_UP))
 207                         return t;
 208         }
 209
 210         t = rcu_dereference(ip6n->tnls_wc[0]);
 211         if (t && (t->dev->flags & IFF_UP))
 212                 return t;
 213
 214         return NULL;
 215 }
 216
 217 /**
 218  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 219  *   @p: parameters containing tunnel end-points
 220  *
 221  * Description:
 222  *   ip6_tnl_bucket() returns the head of the list matching the
 223  *   &struct in6_addr entries laddr and raddr in @p.
 224  *
 225  * Return: head of IPv6 tunnel list
 226  **/
 227
 228 static struct ip6_tnl __rcu **
 229 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 230 {
 231         const struct in6_addr *remote = &p->raddr;
 232         const struct in6_addr *local = &p->laddr;
 233         unsigned int h = 0;
 234         int prio = 0;
 235
 236         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 237                 prio = 1;
 238                 h = HASH(remote, local);
 239         }
 240         return &ip6n->tnls[prio][h];
 241 }
 242
 243 /**
 244  * ip6_tnl_link - add tunnel to hash table
 245  *   @t: tunnel to be added
 246  **/
 247
 248 static void
 249 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 250 {
 251         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 252
 253         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 254         rcu_assign_pointer(*tp, t);
 255 }
 256
 257 /**
 258  * ip6_tnl_unlink - remove tunnel from hash table
 259  *   @t: tunnel to be removed
 260  **/
 261
 262 static void
 263 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 264 {
 265         struct ip6_tnl __rcu **tp;
 266         struct ip6_tnl *iter;
 267
 268         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 269              (iter = rtnl_dereference(*tp)) != NULL;
 270              tp = &iter->next) {
 271                 if (t == iter) {
 272                         rcu_assign_pointer(*tp, t->next);
 273                         break;
 274                 }
 275         }
 276 }
 277
 278 static void ip6_dev_free(struct net_device *dev)
 279 {
 280         free_percpu(dev->tstats);
 281         free_netdev(dev);
 282 }
 283
 284 static int ip6_tnl_create2(struct net_device *dev)
 285 {
 286         struct ip6_tnl *t = netdev_priv(dev);
 287         struct net *net = dev_net(dev);
 288         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 289         int err;
 290
 291         t = netdev_priv(dev);
 292
 293         err = register_netdevice(dev);
 294         if (err < 0)
 295                 goto out;
 296
 297         strcpy(t->parms.name, dev->name);
 298         dev->rtnl_link_ops = &ip6_link_ops;
 299
 300         dev_hold(dev);
 301         ip6_tnl_link(ip6n, t);
 302         return 0;
 303
 304 out:
 305         return err;
 306 }
 307
 308 /**
 309  * ip6_tnl_create - create a new tunnel
 310  *   @p: tunnel parameters
 311  *   @pt: pointer to new tunnel
 312  *
 313  * Description:
 314  *   Create tunnel matching given parameters.
 315  *
 316  * Return:
 317  *   created tunnel or error pointer
 318  **/
 319
 320 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 321 {
 322         struct net_device *dev;
 323         struct ip6_tnl *t;
 324         char name[IFNAMSIZ];
 325         int err = -ENOMEM;
 326
 327         if (p->name[0])
 328                 strlcpy(name, p->name, IFNAMSIZ);
 329         else
 330                 sprintf(name, "ip6tnl%%d");
 331
 332         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 333                            ip6_tnl_dev_setup);
 334         if (dev == NULL)
 335                 goto failed;
 336
 337         dev_net_set(dev, net);
 338
 339         t = netdev_priv(dev);
 340         t->parms = *p;
 341         t->net = dev_net(dev);
 342         err = ip6_tnl_create2(dev);
 343         if (err < 0)
 344                 goto failed_free;
 345
 346         return t;
 347
 348 failed_free:
 349         ip6_dev_free(dev);
 350 failed:
 351         return ERR_PTR(err);
 352 }
 353
 354 /**
 355  * ip6_tnl_locate - find or create tunnel matching given parameters
 356  *   @p: tunnel parameters
 357  *   @create: != 0 if allowed to create new tunnel if no match found
 358  *
 359  * Description:
 360  *   ip6_tnl_locate() first tries to locate an existing tunnel
 361  *   based on @parms. If this is unsuccessful, but @create is set a new
 362  *   tunnel device is created and registered for use.
 363  *
 364  * Return:
 365  *   matching tunnel or error pointer
 366  **/
 367
 368 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 369                 struct __ip6_tnl_parm *p, int create)
 370 {
 371         const struct in6_addr *remote = &p->raddr;
 372         const struct in6_addr *local = &p->laddr;
 373         struct ip6_tnl __rcu **tp;
 374         struct ip6_tnl *t;
 375         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 376
 377         for (tp = ip6_tnl_bucket(ip6n, p);
 378              (t = rtnl_dereference(*tp)) != NULL;
 379              tp = &t->next) {
 380                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 381                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 382                         if (create)
 383                                 return ERR_PTR(-EEXIST);
 384
 385                         return t;
 386                 }
 387         }
 388         if (!create)
 389                 return ERR_PTR(-ENODEV);
 390         return ip6_tnl_create(net, p);
 391 }
 392
 393 /**
 394  * ip6_tnl_dev_uninit - tunnel device uninitializer
 395  *   @dev: the device to be destroyed
 396  *
 397  * Description:
 398  *   ip6_tnl_dev_uninit() removes tunnel from its list
 399  **/
 400
 401 static void
 402 ip6_tnl_dev_uninit(struct net_device *dev)
 403 {
 404         struct ip6_tnl *t = netdev_priv(dev);
 405         struct net *net = t->net;
 406         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 407
 408         if (dev == ip6n->fb_tnl_dev)
 409                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 410         else
 411                 ip6_tnl_unlink(ip6n, t);
 412         ip6_tnl_dst_reset(t);
 413         dev_put(dev);
 414 }
 415
 416 /**
 417  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 418  *   @skb: received socket buffer
 419  *
 420  * Return:
 421  *   0 if none was found,
 422  *   else index to encapsulation limit
 423  **/
 424
 425 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 426 {
 427         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 428         __u8 nexthdr = ipv6h->nexthdr;
 429         __u16 off = sizeof(*ipv6h);
 430
 431         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 432                 __u16 optlen = 0;
 433                 struct ipv6_opt_hdr *hdr;
 434                 if (raw + off + sizeof(*hdr) > skb->data &&
 435                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 436                         break;
 437
 438                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 439                 if (nexthdr == NEXTHDR_FRAGMENT) {
 440                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 441                         if (frag_hdr->frag_off)
 442                                 break;
 443                         optlen = 8;
 444                 } else if (nexthdr == NEXTHDR_AUTH) {
 445                         optlen = (hdr->hdrlen + 2) << 2;
 446                 } else {
 447                         optlen = ipv6_optlen(hdr);
 448                 }
 449                 if (nexthdr == NEXTHDR_DEST) {
 450                         __u16 i = off + 2;
 451                         while (1) {
 452                                 struct ipv6_tlv_tnl_enc_lim *tel;
 453
 454                                 /* No more room for encapsulation limit */
 455                                 if (i + sizeof (*tel) > off + optlen)
 456                                         break;
 457
 458                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 459                                 /* return index of option if found and valid */
 460                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 461                                     tel->length == 1)
 462                                         return i;
 463                                 /* else jump to next option */
 464                                 if (tel->type)
 465                                         i += tel->length + 2;
 466                                 else
 467                                         i++;
 468                         }
 469                 }
 470                 nexthdr = hdr->nexthdr;
 471                 off += optlen;
 472         }
 473         return 0;
 474 }
 475 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 476
 477 /**
 478  * ip6_tnl_err - tunnel error handler
 479  *
 480  * Description:
 481  *   ip6_tnl_err() should handle errors in the tunnel according
 482  *   to the specifications in RFC 2473.
 483  **/
 484
 485 static int
 486 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 487             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 488 {
 489         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 490         struct ip6_tnl *t;
 491         int rel_msg = 0;
 492         u8 rel_type = ICMPV6_DEST_UNREACH;
 493         u8 rel_code = ICMPV6_ADDR_UNREACH;
 494         u8 tproto;
 495         __u32 rel_info = 0;
 496         __u16 len;
 497         int err = -ENOENT;
 498
 499         /* If the packet doesn't contain the original IPv6 header we are
 500            in trouble since we might need the source address for further
 501            processing of the error. */
 502
 503         rcu_read_lock();
 504         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 505         if (t == NULL)
 506                 goto out;
 507
 508         tproto = ACCESS_ONCE(t->parms.proto);
 509         if (tproto != ipproto && tproto != 0)
 510                 goto out;
 511
 512         err = 0;
 513
 514         switch (*type) {
 515                 __u32 teli;
 516                 struct ipv6_tlv_tnl_enc_lim *tel;
 517                 __u32 mtu;
 518         case ICMPV6_DEST_UNREACH:
 519                 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
 520                                      t->parms.name);
 521                 rel_msg = 1;
 522                 break;
 523         case ICMPV6_TIME_EXCEED:
 524                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 525                         net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 526                                              t->parms.name);
 527                         rel_msg = 1;
 528                 }
 529                 break;
 530         case ICMPV6_PARAMPROB:
 531                 teli = 0;
 532                 if ((*code) == ICMPV6_HDR_FIELD)
 533                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 534
 535                 if (teli && teli == *info - 2) {
 536                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 537                         if (tel->encap_limit == 0) {
 538                                 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 539                                                      t->parms.name);
 540                                 rel_msg = 1;
 541                         }
 542                 } else {
 543                         net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 544                                              t->parms.name);
 545                 }
 546                 break;
 547         case ICMPV6_PKT_TOOBIG:
 548                 mtu = *info - offset;
 549                 if (mtu < IPV6_MIN_MTU)
 550                         mtu = IPV6_MIN_MTU;
 551                 t->dev->mtu = mtu;
 552
 553                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 554                 if (len > mtu) {
 555                         rel_type = ICMPV6_PKT_TOOBIG;
 556                         rel_code = 0;
 557                         rel_info = mtu;
 558                         rel_msg = 1;
 559                 }
 560                 break;
 561         }
 562
 563         *type = rel_type;
 564         *code = rel_code;
 565         *info = rel_info;
 566         *msg = rel_msg;
 567
 568 out:
 569         rcu_read_unlock();
 570         return err;
 571 }
 572
 573 static int
 574 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 575            u8 type, u8 code, int offset, __be32 info)
 576 {
 577         int rel_msg = 0;
 578         u8 rel_type = type;
 579         u8 rel_code = code;
 580         __u32 rel_info = ntohl(info);
 581         int err;
 582         struct sk_buff *skb2;
 583         const struct iphdr *eiph;
 584         struct rtable *rt;
 585         struct flowi4 fl4;
 586
 587         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 588                           &rel_msg, &rel_info, offset);
 589         if (err < 0)
 590                 return err;
 591
 592         if (rel_msg == 0)
 593                 return 0;
 594
 595         switch (rel_type) {
 596         case ICMPV6_DEST_UNREACH:
 597                 if (rel_code != ICMPV6_ADDR_UNREACH)
 598                         return 0;
 599                 rel_type = ICMP_DEST_UNREACH;
 600                 rel_code = ICMP_HOST_UNREACH;
 601                 break;
 602         case ICMPV6_PKT_TOOBIG:
 603                 if (rel_code != 0)
 604                         return 0;
 605                 rel_type = ICMP_DEST_UNREACH;
 606                 rel_code = ICMP_FRAG_NEEDED;
 607                 break;
 608         case NDISC_REDIRECT:
 609                 rel_type = ICMP_REDIRECT;
 610                 rel_code = ICMP_REDIR_HOST;
 611         default:
 612                 return 0;
 613         }
 614
 615         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 616                 return 0;
 617
 618         skb2 = skb_clone(skb, GFP_ATOMIC);
 619         if (!skb2)
 620                 return 0;
 621
 622         skb_dst_drop(skb2);
 623
 624         skb_pull(skb2, offset);
 625         skb_reset_network_header(skb2);
 626         eiph = ip_hdr(skb2);
 627
 628         /* Try to guess incoming interface */
 629         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 630                                    eiph->saddr, 0,
 631                                    0, 0,
 632                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 633         if (IS_ERR(rt))
 634                 goto out;
 635
 636         skb2->dev = rt->dst.dev;
 637
 638         /* route "incoming" packet */
 639         if (rt->rt_flags & RTCF_LOCAL) {
 640                 ip_rt_put(rt);
 641                 rt = NULL;
 642                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 643                                            eiph->daddr, eiph->saddr,
 644                                            0, 0,
 645                                            IPPROTO_IPIP,
 646                                            RT_TOS(eiph->tos), 0);
 647                 if (IS_ERR(rt) ||
 648                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 649                         if (!IS_ERR(rt))
 650                                 ip_rt_put(rt);
 651                         goto out;
 652                 }
 653                 skb_dst_set(skb2, &rt->dst);
 654         } else {
 655                 ip_rt_put(rt);
 656                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 657                                    skb2->dev) ||
 658                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 659                         goto out;
 660         }
 661
 662         /* change mtu on this route */
 663         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 664                 if (rel_info > dst_mtu(skb_dst(skb2)))
 665                         goto out;
 666
 667                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 668         }
 669         if (rel_type == ICMP_REDIRECT)
 670                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 671
 672         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 673
 674 out:
 675         kfree_skb(skb2);
 676         return 0;
 677 }
 678
 679 static int
 680 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 681            u8 type, u8 code, int offset, __be32 info)
 682 {
 683         int rel_msg = 0;
 684         u8 rel_type = type;
 685         u8 rel_code = code;
 686         __u32 rel_info = ntohl(info);
 687         int err;
 688
 689         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 690                           &rel_msg, &rel_info, offset);
 691         if (err < 0)
 692                 return err;
 693
 694         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 695                 struct rt6_info *rt;
 696                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 697
 698                 if (!skb2)
 699                         return 0;
 700
 701                 skb_dst_drop(skb2);
 702                 skb_pull(skb2, offset);
 703                 skb_reset_network_header(skb2);
 704
 705                 /* Try to guess incoming interface */
 706                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 707                                 NULL, 0, 0);
 708
 709                 if (rt && rt->dst.dev)
 710                         skb2->dev = rt->dst.dev;
 711
 712                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 713
 714                 ip6_rt_put(rt);
 715
 716                 kfree_skb(skb2);
 717         }
 718
 719         return 0;
 720 }
 721
 722 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 723                                        const struct ipv6hdr *ipv6h,
 724                                        struct sk_buff *skb)
 725 {
 726         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 727
 728         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 729                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 730
 731         return IP6_ECN_decapsulate(ipv6h, skb);
 732 }
 733
 734 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 735                                        const struct ipv6hdr *ipv6h,
 736                                        struct sk_buff *skb)
 737 {
 738         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 739                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 740
 741         return IP6_ECN_decapsulate(ipv6h, skb);
 742 }
 743
 744 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 745                              const struct in6_addr *laddr,
 746                              const struct in6_addr *raddr)
 747 {
 748         struct __ip6_tnl_parm *p = &t->parms;
 749         int ltype = ipv6_addr_type(laddr);
 750         int rtype = ipv6_addr_type(raddr);
 751         __u32 flags = 0;
 752
 753         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 754                 flags = IP6_TNL_F_CAP_PER_PACKET;
 755         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 756                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 757                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 758                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 759                 if (ltype&IPV6_ADDR_UNICAST)
 760                         flags |= IP6_TNL_F_CAP_XMIT;
 761                 if (rtype&IPV6_ADDR_UNICAST)
 762                         flags |= IP6_TNL_F_CAP_RCV;
 763         }
 764         return flags;
 765 }
 766 EXPORT_SYMBOL(ip6_tnl_get_cap);
 767
 768 /* called with rcu_read_lock() */
 769 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 770                                   const struct in6_addr *laddr,
 771                                   const struct in6_addr *raddr)
 772 {
 773         struct __ip6_tnl_parm *p = &t->parms;
 774         int ret = 0;
 775         struct net *net = t->net;
 776
 777         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 778             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 779              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 780                 struct net_device *ldev = NULL;
 781
 782                 if (p->link)
 783                         ldev = dev_get_by_index_rcu(net, p->link);
 784
 785                 if ((ipv6_addr_is_multicast(laddr) ||
 786                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 787                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 788                         ret = 1;
 789         }
 790         return ret;
 791 }
 792 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 793
 794 /**
 795  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 796  *   @skb: received socket buffer
 797  *   @protocol: ethernet protocol ID
 798  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 799  *
 800  * Return: 0
 801  **/
 802
 803 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 804                        __u8 ipproto,
 805                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 806                                                    const struct ipv6hdr *ipv6h,
 807                                                    struct sk_buff *skb))
 808 {
 809         struct ip6_tnl *t;
 810         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 811         u8 tproto;
 812         int err;
 813
 814         rcu_read_lock();
 815         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 816         if (t != NULL) {
 817                 struct pcpu_sw_netstats *tstats;
 818
 819                 tproto = ACCESS_ONCE(t->parms.proto);
 820                 if (tproto != ipproto && tproto != 0) {
 821                         rcu_read_unlock();
 822                         goto discard;
 823                 }
 824
 825                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 826                         rcu_read_unlock();
 827                         goto discard;
 828                 }
 829
 830                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
 831                         t->dev->stats.rx_dropped++;
 832                         rcu_read_unlock();
 833                         goto discard;
 834                 }
 835                 skb->mac_header = skb->network_header;
 836                 skb_reset_network_header(skb);
 837                 skb->protocol = htons(protocol);
 838                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 839
 840                 __skb_tunnel_rx(skb, t->dev, t->net);
 841
 842                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
 843                 if (unlikely(err)) {
 844                         if (log_ecn_error)
 845                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 846                                                      &ipv6h->saddr,
 847                                                      ipv6_get_dsfield(ipv6h));
 848                         if (err > 1) {
 849                                 ++t->dev->stats.rx_frame_errors;
 850                                 ++t->dev->stats.rx_errors;
 851                                 rcu_read_unlock();
 852                                 goto discard;
 853                         }
 854                 }
 855
 856                 tstats = this_cpu_ptr(t->dev->tstats);
 857                 u64_stats_update_begin(&tstats->syncp);
 858                 tstats->rx_packets++;
 859                 tstats->rx_bytes += skb->len;
 860                 u64_stats_update_end(&tstats->syncp);
 861
 862                 netif_rx(skb);
 863
 864                 rcu_read_unlock();
 865                 return 0;
 866         }
 867         rcu_read_unlock();
 868         return 1;
 869
 870 discard:
 871         kfree_skb(skb);
 872         return 0;
 873 }
 874
 875 static int ip4ip6_rcv(struct sk_buff *skb)
 876 {
 877         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 878                            ip4ip6_dscp_ecn_decapsulate);
 879 }
 880
 881 static int ip6ip6_rcv(struct sk_buff *skb)
 882 {
 883         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 884                            ip6ip6_dscp_ecn_decapsulate);
 885 }
 886
 887 struct ipv6_tel_txoption {
 888         struct ipv6_txoptions ops;
 889         __u8 dst_opt[8];
 890 };
 891
 892 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 893 {
 894         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 895
 896         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 897         opt->dst_opt[3] = 1;
 898         opt->dst_opt[4] = encap_limit;
 899         opt->dst_opt[5] = IPV6_TLV_PADN;
 900         opt->dst_opt[6] = 1;
 901
 902         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 903         opt->ops.opt_nflen = 8;
 904 }
 905
 906 /**
 907  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 908  *   @t: the outgoing tunnel device
 909  *   @hdr: IPv6 header from the incoming packet
 910  *
 911  * Description:
 912  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 913  *   doesn't match source of incoming packet.
 914  *
 915  * Return:
 916  *   1 if conflict,
 917  *   0 else
 918  **/
 919
 920 static inline bool
 921 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 922 {
 923         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 924 }
 925
 926 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 927                      const struct in6_addr *laddr,
 928                      const struct in6_addr *raddr)
 929 {
 930         struct __ip6_tnl_parm *p = &t->parms;
 931         int ret = 0;
 932         struct net *net = t->net;
 933
 934         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 935             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 936              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 937                 struct net_device *ldev = NULL;
 938
 939                 rcu_read_lock();
 940                 if (p->link)
 941                         ldev = dev_get_by_index_rcu(net, p->link);
 942
 943                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 944                         pr_warn("%s xmit: Local address not yet configured!\n",
 945                                 p->name);
 946                 else if (!ipv6_addr_is_multicast(raddr) &&
 947                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
 948                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 949                                 p->name);
 950                 else
 951                         ret = 1;
 952                 rcu_read_unlock();
 953         }
 954         return ret;
 955 }
 956 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
 957
 958 /**
 959  * ip6_tnl_xmit2 - encapsulate packet and send
 960  *   @skb: the outgoing socket buffer
 961  *   @dev: the outgoing tunnel device
 962  *   @dsfield: dscp code for outer header
 963  *   @fl: flow of tunneled packet
 964  *   @encap_limit: encapsulation limit
 965  *   @pmtu: Path MTU is stored if packet is too big
 966  *
 967  * Description:
 968  *   Build new header and do some sanity checks on the packet before sending
 969  *   it.
 970  *
 971  * Return:
 972  *   0 on success
 973  *   -1 fail
 974  *   %-EMSGSIZE message too big. return mtu in this case.
 975  **/
 976
 977 static int ip6_tnl_xmit2(struct sk_buff *skb,
 978                          struct net_device *dev,
 979                          __u8 dsfield,
 980                          struct flowi6 *fl6,
 981                          int encap_limit,
 982                          __u32 *pmtu)
 983 {
 984         struct ip6_tnl *t = netdev_priv(dev);
 985         struct net *net = t->net;
 986         struct net_device_stats *stats = &t->dev->stats;
 987         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 988         struct ipv6_tel_txoption opt;
 989         struct dst_entry *dst = NULL, *ndst = NULL;
 990         struct net_device *tdev;
 991         int mtu;
 992         unsigned int max_headroom = sizeof(struct ipv6hdr);
 993         u8 proto;
 994         int err = -1;
 995
 996         /* NBMA tunnel */
 997         if (ipv6_addr_any(&t->parms.raddr)) {
 998                 struct in6_addr *addr6;
 999                 struct neighbour *neigh;
1000                 int addr_type;
1001
1002                 if (!skb_dst(skb))
1003                         goto tx_err_link_failure;
1004
1005                 neigh = dst_neigh_lookup(skb_dst(skb),
1006                                          &ipv6_hdr(skb)->daddr);
1007                 if (!neigh)
1008                         goto tx_err_link_failure;
1009
1010                 addr6 = (struct in6_addr *)&neigh->primary_key;
1011                 addr_type = ipv6_addr_type(addr6);
1012
1013                 if (addr_type == IPV6_ADDR_ANY)
1014                         addr6 = &ipv6_hdr(skb)->daddr;
1015
1016                 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1017                 neigh_release(neigh);
1018         } else if (!fl6->flowi6_mark)
1019                 dst = ip6_tnl_dst_check(t);
1020
1021         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1022                 goto tx_err_link_failure;
1023
1024         if (!dst) {
1025                 ndst = ip6_route_output(net, NULL, fl6);
1026
1027                 if (ndst->error)
1028                         goto tx_err_link_failure;
1029                 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
1030                 if (IS_ERR(ndst)) {
1031                         err = PTR_ERR(ndst);
1032                         ndst = NULL;
1033                         goto tx_err_link_failure;
1034                 }
1035                 dst = ndst;
1036         }
1037
1038         tdev = dst->dev;
1039
1040         if (tdev == dev) {
1041                 stats->collisions++;
1042                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1043                                      t->parms.name);
1044                 goto tx_err_dst_release;
1045         }
1046         mtu = dst_mtu(dst) - sizeof(*ipv6h);
1047         if (encap_limit >= 0) {
1048                 max_headroom += 8;
1049                 mtu -= 8;
1050         }
1051         if (mtu < IPV6_MIN_MTU)
1052                 mtu = IPV6_MIN_MTU;
1053         if (skb_dst(skb))
1054                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1055         if (skb->len > mtu) {
1056                 *pmtu = mtu;
1057                 err = -EMSGSIZE;
1058                 goto tx_err_dst_release;
1059         }
1060
1061         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1062
1063         /*
1064          * Okay, now see if we can stuff it in the buffer as-is.
1065          */
1066         max_headroom += LL_RESERVED_SPACE(tdev);
1067
1068         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1069             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1070                 struct sk_buff *new_skb;
1071
1072                 new_skb = skb_realloc_headroom(skb, max_headroom);
1073                 if (!new_skb)
1074                         goto tx_err_dst_release;
1075
1076                 if (skb->sk)
1077                         skb_set_owner_w(new_skb, skb->sk);
1078                 consume_skb(skb);
1079                 skb = new_skb;
1080         }
1081         if (fl6->flowi6_mark) {
1082                 skb_dst_set(skb, dst);
1083                 ndst = NULL;
1084         } else {
1085                 skb_dst_set_noref(skb, dst);
1086         }
1087         skb->transport_header = skb->network_header;
1088
1089         proto = fl6->flowi6_proto;
1090         if (encap_limit >= 0) {
1091                 init_tel_txopt(&opt, encap_limit);
1092                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1093         }
1094
1095         if (likely(!skb->encapsulation)) {
1096                 skb_reset_inner_headers(skb);
1097                 skb->encapsulation = 1;
1098         }
1099
1100         skb_push(skb, sizeof(struct ipv6hdr));
1101         skb_reset_network_header(skb);
1102         ipv6h = ipv6_hdr(skb);
1103         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1104                      ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
1105         ipv6h->hop_limit = t->parms.hop_limit;
1106         ipv6h->nexthdr = proto;
1107         ipv6h->saddr = fl6->saddr;
1108         ipv6h->daddr = fl6->daddr;
1109         ip6tunnel_xmit(skb, dev);
1110         if (ndst)
1111                 ip6_tnl_dst_store(t, ndst);
1112         return 0;
1113 tx_err_link_failure:
1114         stats->tx_carrier_errors++;
1115         dst_link_failure(skb);
1116 tx_err_dst_release:
1117         dst_release(ndst);
1118         return err;
1119 }
1120
1121 static inline int
1122 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1123 {
1124         struct ip6_tnl *t = netdev_priv(dev);
1125         const struct iphdr  *iph = ip_hdr(skb);
1126         int encap_limit = -1;
1127         struct flowi6 fl6;
1128         __u8 dsfield;
1129         __u32 mtu;
1130         u8 tproto;
1131         int err;
1132
1133         tproto = ACCESS_ONCE(t->parms.proto);
1134         if (tproto != IPPROTO_IPIP && tproto != 0)
1135                 return -1;
1136
1137         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1138                 encap_limit = t->parms.encap_limit;
1139
1140         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1141         fl6.flowi6_proto = IPPROTO_IPIP;
1142
1143         dsfield = ipv4_get_dsfield(iph);
1144
1145         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1146                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1147                                           & IPV6_TCLASS_MASK;
1148         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1149                 fl6.flowi6_mark = skb->mark;
1150
1151         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1152         if (err != 0) {
1153                 /* XXX: send ICMP error even if DF is not set. */
1154                 if (err == -EMSGSIZE)
1155                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1156                                   htonl(mtu));
1157                 return -1;
1158         }
1159
1160         return 0;
1161 }
1162
1163 static inline int
1164 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1165 {
1166         struct ip6_tnl *t = netdev_priv(dev);
1167         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1168         int encap_limit = -1;
1169         __u16 offset;
1170         struct flowi6 fl6;
1171         __u8 dsfield;
1172         __u32 mtu;
1173         u8 tproto;
1174         int err;
1175
1176         tproto = ACCESS_ONCE(t->parms.proto);
1177         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1178             ip6_tnl_addr_conflict(t, ipv6h))
1179                 return -1;
1180
1181         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1182         if (offset > 0) {
1183                 struct ipv6_tlv_tnl_enc_lim *tel;
1184                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1185                 if (tel->encap_limit == 0) {
1186                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1187                                     ICMPV6_HDR_FIELD, offset + 2);
1188                         return -1;
1189                 }
1190                 encap_limit = tel->encap_limit - 1;
1191         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1192                 encap_limit = t->parms.encap_limit;
1193
1194         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1195         fl6.flowi6_proto = IPPROTO_IPV6;
1196
1197         dsfield = ipv6_get_dsfield(ipv6h);
1198         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1199                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1200         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1201                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1202         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1203                 fl6.flowi6_mark = skb->mark;
1204
1205         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1206         if (err != 0) {
1207                 if (err == -EMSGSIZE)
1208                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1209                 return -1;
1210         }
1211
1212         return 0;
1213 }
1214
1215 static netdev_tx_t
1216 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1217 {
1218         struct ip6_tnl *t = netdev_priv(dev);
1219         struct net_device_stats *stats = &t->dev->stats;
1220         int ret;
1221
1222         switch (skb->protocol) {
1223         case htons(ETH_P_IP):
1224                 ret = ip4ip6_tnl_xmit(skb, dev);
1225                 break;
1226         case htons(ETH_P_IPV6):
1227                 ret = ip6ip6_tnl_xmit(skb, dev);
1228                 break;
1229         default:
1230                 goto tx_err;
1231         }
1232
1233         if (ret < 0)
1234                 goto tx_err;
1235
1236         return NETDEV_TX_OK;
1237
1238 tx_err:
1239         stats->tx_errors++;
1240         stats->tx_dropped++;
1241         kfree_skb(skb);
1242         return NETDEV_TX_OK;
1243 }
1244
1245 static void ip6_tnl_link_config(struct ip6_tnl *t)
1246 {
1247         struct net_device *dev = t->dev;
1248         struct __ip6_tnl_parm *p = &t->parms;
1249         struct flowi6 *fl6 = &t->fl.u.ip6;
1250
1251         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1252         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1253
1254         /* Set up flowi template */
1255         fl6->saddr = p->laddr;
1256         fl6->daddr = p->raddr;
1257         fl6->flowi6_oif = p->link;
1258         fl6->flowlabel = 0;
1259
1260         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1261                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1262         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1263                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1264
1265         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1266         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1267
1268         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1269                 dev->flags |= IFF_POINTOPOINT;
1270         else
1271                 dev->flags &= ~IFF_POINTOPOINT;
1272
1273         dev->iflink = p->link;
1274
1275         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1276                 int strict = (ipv6_addr_type(&p->raddr) &
1277                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1278
1279                 struct rt6_info *rt = rt6_lookup(t->net,
1280                                                  &p->raddr, &p->laddr,
1281                                                  p->link, strict);
1282
1283                 if (rt == NULL)
1284                         return;
1285
1286                 if (rt->dst.dev) {
1287                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1288                                 sizeof(struct ipv6hdr);
1289
1290                         dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1291                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1292                                 dev->mtu -= 8;
1293
1294                         if (dev->mtu < IPV6_MIN_MTU)
1295                                 dev->mtu = IPV6_MIN_MTU;
1296                 }
1297                 ip6_rt_put(rt);
1298         }
1299 }
1300
1301 /**
1302  * ip6_tnl_change - update the tunnel parameters
1303  *   @t: tunnel to be changed
1304  *   @p: tunnel configuration parameters
1305  *
1306  * Description:
1307  *   ip6_tnl_change() updates the tunnel parameters
1308  **/
1309
1310 static int
1311 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1312 {
1313         t->parms.laddr = p->laddr;
1314         t->parms.raddr = p->raddr;
1315         t->parms.flags = p->flags;
1316         t->parms.hop_limit = p->hop_limit;
1317         t->parms.encap_limit = p->encap_limit;
1318         t->parms.flowinfo = p->flowinfo;
1319         t->parms.link = p->link;
1320         t->parms.proto = p->proto;
1321         ip6_tnl_dst_reset(t);
1322         ip6_tnl_link_config(t);
1323         return 0;
1324 }
1325
1326 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1327 {
1328         struct net *net = t->net;
1329         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1330         int err;
1331
1332         ip6_tnl_unlink(ip6n, t);
1333         synchronize_net();
1334         err = ip6_tnl_change(t, p);
1335         ip6_tnl_link(ip6n, t);
1336         netdev_state_change(t->dev);
1337         return err;
1338 }
1339
1340 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1341 {
1342         /* for default tnl0 device allow to change only the proto */
1343         t->parms.proto = p->proto;
1344         netdev_state_change(t->dev);
1345         return 0;
1346 }
1347
1348 static void
1349 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1350 {
1351         p->laddr = u->laddr;
1352         p->raddr = u->raddr;
1353         p->flags = u->flags;
1354         p->hop_limit = u->hop_limit;
1355         p->encap_limit = u->encap_limit;
1356         p->flowinfo = u->flowinfo;
1357         p->link = u->link;
1358         p->proto = u->proto;
1359         memcpy(p->name, u->name, sizeof(u->name));
1360 }
1361
1362 static void
1363 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1364 {
1365         u->laddr = p->laddr;
1366         u->raddr = p->raddr;
1367         u->flags = p->flags;
1368         u->hop_limit = p->hop_limit;
1369         u->encap_limit = p->encap_limit;
1370         u->flowinfo = p->flowinfo;
1371         u->link = p->link;
1372         u->proto = p->proto;
1373         memcpy(u->name, p->name, sizeof(u->name));
1374 }
1375
1376 /**
1377  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1378  *   @dev: virtual device associated with tunnel
1379  *   @ifr: parameters passed from userspace
1380  *   @cmd: command to be performed
1381  *
1382  * Description:
1383  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1384  *   from userspace.
1385  *
1386  *   The possible commands are the following:
1387  *     %SIOCGETTUNNEL: get tunnel parameters for device
1388  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1389  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1390  *     %SIOCDELTUNNEL: delete tunnel
1391  *
1392  *   The fallback device "ip6tnl0", created during module
1393  *   initialization, can be used for creating other tunnel devices.
1394  *
1395  * Return:
1396  *   0 on success,
1397  *   %-EFAULT if unable to copy data to or from userspace,
1398  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1399  *   %-EINVAL if passed tunnel parameters are invalid,
1400  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1401  *   %-ENODEV if attempting to change or delete a nonexisting device
1402  **/
1403
1404 static int
1405 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1406 {
1407         int err = 0;
1408         struct ip6_tnl_parm p;
1409         struct __ip6_tnl_parm p1;
1410         struct ip6_tnl *t = netdev_priv(dev);
1411         struct net *net = t->net;
1412         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1413
1414         switch (cmd) {
1415         case SIOCGETTUNNEL:
1416                 if (dev == ip6n->fb_tnl_dev) {
1417                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1418                                 err = -EFAULT;
1419                                 break;
1420                         }
1421                         ip6_tnl_parm_from_user(&p1, &p);
1422                         t = ip6_tnl_locate(net, &p1, 0);
1423                         if (IS_ERR(t))
1424                                 t = netdev_priv(dev);
1425                 } else {
1426                         memset(&p, 0, sizeof(p));
1427                 }
1428                 ip6_tnl_parm_to_user(&p, &t->parms);
1429                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1430                         err = -EFAULT;
1431                 }
1432                 break;
1433         case SIOCADDTUNNEL:
1434         case SIOCCHGTUNNEL:
1435                 err = -EPERM;
1436                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1437                         break;
1438                 err = -EFAULT;
1439                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1440                         break;
1441                 err = -EINVAL;
1442                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1443                     p.proto != 0)
1444                         break;
1445                 ip6_tnl_parm_from_user(&p1, &p);
1446                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1447                 if (cmd == SIOCCHGTUNNEL) {
1448                         if (!IS_ERR(t)) {
1449                                 if (t->dev != dev) {
1450                                         err = -EEXIST;
1451                                         break;
1452                                 }
1453                         } else
1454                                 t = netdev_priv(dev);
1455                         if (dev == ip6n->fb_tnl_dev)
1456                                 err = ip6_tnl0_update(t, &p1);
1457                         else
1458                                 err = ip6_tnl_update(t, &p1);
1459                 }
1460                 if (!IS_ERR(t)) {
1461                         err = 0;
1462                         ip6_tnl_parm_to_user(&p, &t->parms);
1463                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1464                                 err = -EFAULT;
1465
1466                 } else {
1467                         err = PTR_ERR(t);
1468                 }
1469                 break;
1470         case SIOCDELTUNNEL:
1471                 err = -EPERM;
1472                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1473                         break;
1474
1475                 if (dev == ip6n->fb_tnl_dev) {
1476                         err = -EFAULT;
1477                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1478                                 break;
1479                         err = -ENOENT;
1480                         ip6_tnl_parm_from_user(&p1, &p);
1481                         t = ip6_tnl_locate(net, &p1, 0);
1482                         if (IS_ERR(t))
1483                                 break;
1484                         err = -EPERM;
1485                         if (t->dev == ip6n->fb_tnl_dev)
1486                                 break;
1487                         dev = t->dev;
1488                 }
1489                 err = 0;
1490                 unregister_netdevice(dev);
1491                 break;
1492         default:
1493                 err = -EINVAL;
1494         }
1495         return err;
1496 }
1497
1498 /**
1499  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1500  *   @dev: virtual device associated with tunnel
1501  *   @new_mtu: the new mtu
1502  *
1503  * Return:
1504  *   0 on success,
1505  *   %-EINVAL if mtu too small
1506  **/
1507
1508 static int
1509 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1510 {
1511         struct ip6_tnl *tnl = netdev_priv(dev);
1512
1513         if (tnl->parms.proto == IPPROTO_IPIP) {
1514                 if (new_mtu < 68)
1515                         return -EINVAL;
1516         } else {
1517                 if (new_mtu < IPV6_MIN_MTU)
1518                         return -EINVAL;
1519         }
1520         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1521                 return -EINVAL;
1522         dev->mtu = new_mtu;
1523         return 0;
1524 }
1525
1526
1527 static const struct net_device_ops ip6_tnl_netdev_ops = {
1528         .ndo_init       = ip6_tnl_dev_init,
1529         .ndo_uninit     = ip6_tnl_dev_uninit,
1530         .ndo_start_xmit = ip6_tnl_xmit,
1531         .ndo_do_ioctl   = ip6_tnl_ioctl,
1532         .ndo_change_mtu = ip6_tnl_change_mtu,
1533         .ndo_get_stats  = ip6_get_stats,
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                 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1650                            sizeof(struct in6_addr));
1651
1652         if (data[IFLA_IPTUN_REMOTE])
1653                 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1654                            sizeof(struct in6_addr));
1655
1656         if (data[IFLA_IPTUN_TTL])
1657                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1658
1659         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1660                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1661
1662         if (data[IFLA_IPTUN_FLOWINFO])
1663                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1664
1665         if (data[IFLA_IPTUN_FLAGS])
1666                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1667
1668         if (data[IFLA_IPTUN_PROTO])
1669                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1670 }
1671
1672 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1673                            struct nlattr *tb[], struct nlattr *data[])
1674 {
1675         struct net *net = dev_net(dev);
1676         struct ip6_tnl *nt, *t;
1677
1678         nt = netdev_priv(dev);
1679         ip6_tnl_netlink_parms(data, &nt->parms);
1680
1681         t = ip6_tnl_locate(net, &nt->parms, 0);
1682         if (!IS_ERR(t))
1683                 return -EEXIST;
1684
1685         return ip6_tnl_create2(dev);
1686 }
1687
1688 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1689                               struct nlattr *data[])
1690 {
1691         struct ip6_tnl *t = netdev_priv(dev);
1692         struct __ip6_tnl_parm p;
1693         struct net *net = t->net;
1694         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1695
1696         if (dev == ip6n->fb_tnl_dev)
1697                 return -EINVAL;
1698
1699         ip6_tnl_netlink_parms(data, &p);
1700
1701         t = ip6_tnl_locate(net, &p, 0);
1702         if (!IS_ERR(t)) {
1703                 if (t->dev != dev)
1704                         return -EEXIST;
1705         } else
1706                 t = netdev_priv(dev);
1707
1708         return ip6_tnl_update(t, &p);
1709 }
1710
1711 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1712 {
1713         struct net *net = dev_net(dev);
1714         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1715
1716         if (dev != ip6n->fb_tnl_dev)
1717                 unregister_netdevice_queue(dev, head);
1718 }
1719
1720 static size_t ip6_tnl_get_size(const struct net_device *dev)
1721 {
1722         return
1723                 /* IFLA_IPTUN_LINK */
1724                 nla_total_size(4) +
1725                 /* IFLA_IPTUN_LOCAL */
1726                 nla_total_size(sizeof(struct in6_addr)) +
1727                 /* IFLA_IPTUN_REMOTE */
1728                 nla_total_size(sizeof(struct in6_addr)) +
1729                 /* IFLA_IPTUN_TTL */
1730                 nla_total_size(1) +
1731                 /* IFLA_IPTUN_ENCAP_LIMIT */
1732                 nla_total_size(1) +
1733                 /* IFLA_IPTUN_FLOWINFO */
1734                 nla_total_size(4) +
1735                 /* IFLA_IPTUN_FLAGS */
1736                 nla_total_size(4) +
1737                 /* IFLA_IPTUN_PROTO */
1738                 nla_total_size(1) +
1739                 0;
1740 }
1741
1742 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1743 {
1744         struct ip6_tnl *tunnel = netdev_priv(dev);
1745         struct __ip6_tnl_parm *parm = &tunnel->parms;
1746
1747         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1748             nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1749                     &parm->laddr) ||
1750             nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1751                     &parm->raddr) ||
1752             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1753             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1754             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1755             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1756             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1757                 goto nla_put_failure;
1758         return 0;
1759
1760 nla_put_failure:
1761         return -EMSGSIZE;
1762 }
1763
1764 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1765 {
1766         struct ip6_tnl *tunnel = netdev_priv(dev);
1767
1768         return tunnel->net;
1769 }
1770 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1771
1772 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1773         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1774         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1775         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1776         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1777         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1778         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1779         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1780         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1781 };
1782
1783 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1784         .kind           = "ip6tnl",
1785         .maxtype        = IFLA_IPTUN_MAX,
1786         .policy         = ip6_tnl_policy,
1787         .priv_size      = sizeof(struct ip6_tnl),
1788         .setup          = ip6_tnl_dev_setup,
1789         .validate       = ip6_tnl_validate,
1790         .newlink        = ip6_tnl_newlink,
1791         .changelink     = ip6_tnl_changelink,
1792         .dellink        = ip6_tnl_dellink,
1793         .get_size       = ip6_tnl_get_size,
1794         .fill_info      = ip6_tnl_fill_info,
1795         .get_link_net   = ip6_tnl_get_link_net,
1796 };
1797
1798 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1799         .handler        = ip4ip6_rcv,
1800         .err_handler    = ip4ip6_err,
1801         .priority       =       1,
1802 };
1803
1804 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1805         .handler        = ip6ip6_rcv,
1806         .err_handler    = ip6ip6_err,
1807         .priority       =       1,
1808 };
1809
1810 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1811 {
1812         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1813         struct net_device *dev, *aux;
1814         int h;
1815         struct ip6_tnl *t;
1816         LIST_HEAD(list);
1817
1818         for_each_netdev_safe(net, dev, aux)
1819                 if (dev->rtnl_link_ops == &ip6_link_ops)
1820                         unregister_netdevice_queue(dev, &list);
1821
1822         for (h = 0; h < HASH_SIZE; h++) {
1823                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1824                 while (t != NULL) {
1825                         /* If dev is in the same netns, it has already
1826                          * been added to the list by the previous loop.
1827                          */
1828                         if (!net_eq(dev_net(t->dev), net))
1829                                 unregister_netdevice_queue(t->dev, &list);
1830                         t = rtnl_dereference(t->next);
1831                 }
1832         }
1833
1834         unregister_netdevice_many(&list);
1835 }
1836
1837 static int __net_init ip6_tnl_init_net(struct net *net)
1838 {
1839         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1840         struct ip6_tnl *t = NULL;
1841         int err;
1842
1843         ip6n->tnls[0] = ip6n->tnls_wc;
1844         ip6n->tnls[1] = ip6n->tnls_r_l;
1845
1846         err = -ENOMEM;
1847         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1848                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1849
1850         if (!ip6n->fb_tnl_dev)
1851                 goto err_alloc_dev;
1852         dev_net_set(ip6n->fb_tnl_dev, net);
1853         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1854         /* FB netdevice is special: we have one, and only one per netns.
1855          * Allowing to move it to another netns is clearly unsafe.
1856          */
1857         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1858
1859         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1860         if (err < 0)
1861                 goto err_register;
1862
1863         err = register_netdev(ip6n->fb_tnl_dev);
1864         if (err < 0)
1865                 goto err_register;
1866
1867         t = netdev_priv(ip6n->fb_tnl_dev);
1868
1869         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1870         return 0;
1871
1872 err_register:
1873         ip6_dev_free(ip6n->fb_tnl_dev);
1874 err_alloc_dev:
1875         return err;
1876 }
1877
1878 static void __net_exit ip6_tnl_exit_net(struct net *net)
1879 {
1880         rtnl_lock();
1881         ip6_tnl_destroy_tunnels(net);
1882         rtnl_unlock();
1883 }
1884
1885 static struct pernet_operations ip6_tnl_net_ops = {
1886         .init = ip6_tnl_init_net,
1887         .exit = ip6_tnl_exit_net,
1888         .id   = &ip6_tnl_net_id,
1889         .size = sizeof(struct ip6_tnl_net),
1890 };
1891
1892 /**
1893  * ip6_tunnel_init - register protocol and reserve needed resources
1894  *
1895  * Return: 0 on success
1896  **/
1897
1898 static int __init ip6_tunnel_init(void)
1899 {
1900         int  err;
1901
1902         err = register_pernet_device(&ip6_tnl_net_ops);
1903         if (err < 0)
1904                 goto out_pernet;
1905
1906         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1907         if (err < 0) {
1908                 pr_err("%s: can't register ip4ip6\n", __func__);
1909                 goto out_ip4ip6;
1910         }
1911
1912         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1913         if (err < 0) {
1914                 pr_err("%s: can't register ip6ip6\n", __func__);
1915                 goto out_ip6ip6;
1916         }
1917         err = rtnl_link_register(&ip6_link_ops);
1918         if (err < 0)
1919                 goto rtnl_link_failed;
1920
1921         return 0;
1922
1923 rtnl_link_failed:
1924         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1925 out_ip6ip6:
1926         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1927 out_ip4ip6:
1928         unregister_pernet_device(&ip6_tnl_net_ops);
1929 out_pernet:
1930         return err;
1931 }
1932
1933 /**
1934  * ip6_tunnel_cleanup - free resources and unregister protocol
1935  **/
1936
1937 static void __exit ip6_tunnel_cleanup(void)
1938 {
1939         rtnl_link_unregister(&ip6_link_ops);
1940         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1941                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1942
1943         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1944                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1945
1946         unregister_pernet_device(&ip6_tnl_net_ops);
1947 }
1948
1949 module_init(ip6_tunnel_init);
1950 module_exit(ip6_tunnel_cleanup);