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  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 127  *   @remote: the address of the tunnel exit-point
 128  *   @local: the address of the tunnel entry-point
 129  *
 130  * Return:
 131  *   tunnel matching given end-points if found,
 132  *   else fallback tunnel if its device is up,
 133  *   else %NULL
 134  **/
 135
 136 #define for_each_ip6_tunnel_rcu(start) \
 137         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 138
 139 static struct ip6_tnl *
 140 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 141 {
 142         unsigned int hash = HASH(remote, local);
 143         struct ip6_tnl *t;
 144         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 145         struct in6_addr any;
 146
 147         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 148                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 149                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 150                     (t->dev->flags & IFF_UP))
 151                         return t;
 152         }
 153
 154         memset(&any, 0, sizeof(any));
 155         hash = HASH(&any, local);
 156         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 157                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 158                     (t->dev->flags & IFF_UP))
 159                         return t;
 160         }
 161
 162         hash = HASH(remote, &any);
 163         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 164                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 165                     (t->dev->flags & IFF_UP))
 166                         return t;
 167         }
 168
 169         t = rcu_dereference(ip6n->tnls_wc[0]);
 170         if (t && (t->dev->flags & IFF_UP))
 171                 return t;
 172
 173         return NULL;
 174 }
 175
 176 /**
 177  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 178  *   @p: parameters containing tunnel end-points
 179  *
 180  * Description:
 181  *   ip6_tnl_bucket() returns the head of the list matching the
 182  *   &struct in6_addr entries laddr and raddr in @p.
 183  *
 184  * Return: head of IPv6 tunnel list
 185  **/
 186
 187 static struct ip6_tnl __rcu **
 188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 189 {
 190         const struct in6_addr *remote = &p->raddr;
 191         const struct in6_addr *local = &p->laddr;
 192         unsigned int h = 0;
 193         int prio = 0;
 194
 195         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 196                 prio = 1;
 197                 h = HASH(remote, local);
 198         }
 199         return &ip6n->tnls[prio][h];
 200 }
 201
 202 /**
 203  * ip6_tnl_link - add tunnel to hash table
 204  *   @t: tunnel to be added
 205  **/
 206
 207 static void
 208 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 209 {
 210         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 211
 212         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 213         rcu_assign_pointer(*tp, t);
 214 }
 215
 216 /**
 217  * ip6_tnl_unlink - remove tunnel from hash table
 218  *   @t: tunnel to be removed
 219  **/
 220
 221 static void
 222 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 223 {
 224         struct ip6_tnl __rcu **tp;
 225         struct ip6_tnl *iter;
 226
 227         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 228              (iter = rtnl_dereference(*tp)) != NULL;
 229              tp = &iter->next) {
 230                 if (t == iter) {
 231                         rcu_assign_pointer(*tp, t->next);
 232                         break;
 233                 }
 234         }
 235 }
 236
 237 static void ip6_dev_free(struct net_device *dev)
 238 {
 239         struct ip6_tnl *t = netdev_priv(dev);
 240
 241         gro_cells_destroy(&t->gro_cells);
 242         dst_cache_destroy(&t->dst_cache);
 243         free_percpu(dev->tstats);
 244         free_netdev(dev);
 245 }
 246
 247 static int ip6_tnl_create2(struct net_device *dev)
 248 {
 249         struct ip6_tnl *t = netdev_priv(dev);
 250         struct net *net = dev_net(dev);
 251         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 252         int err;
 253
 254         t = netdev_priv(dev);
 255
 256         dev->rtnl_link_ops = &ip6_link_ops;
 257         err = register_netdevice(dev);
 258         if (err < 0)
 259                 goto out;
 260
 261         strcpy(t->parms.name, dev->name);
 262
 263         dev_hold(dev);
 264         ip6_tnl_link(ip6n, t);
 265         return 0;
 266
 267 out:
 268         return err;
 269 }
 270
 271 /**
 272  * ip6_tnl_create - create a new tunnel
 273  *   @p: tunnel parameters
 274  *   @pt: pointer to new tunnel
 275  *
 276  * Description:
 277  *   Create tunnel matching given parameters.
 278  *
 279  * Return:
 280  *   created tunnel or error pointer
 281  **/
 282
 283 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 284 {
 285         struct net_device *dev;
 286         struct ip6_tnl *t;
 287         char name[IFNAMSIZ];
 288         int err = -ENOMEM;
 289
 290         if (p->name[0])
 291                 strlcpy(name, p->name, IFNAMSIZ);
 292         else
 293                 sprintf(name, "ip6tnl%%d");
 294
 295         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 296                            ip6_tnl_dev_setup);
 297         if (!dev)
 298                 goto failed;
 299
 300         dev_net_set(dev, net);
 301
 302         t = netdev_priv(dev);
 303         t->parms = *p;
 304         t->net = dev_net(dev);
 305         err = ip6_tnl_create2(dev);
 306         if (err < 0)
 307                 goto failed_free;
 308
 309         return t;
 310
 311 failed_free:
 312         ip6_dev_free(dev);
 313 failed:
 314         return ERR_PTR(err);
 315 }
 316
 317 /**
 318  * ip6_tnl_locate - find or create tunnel matching given parameters
 319  *   @p: tunnel parameters
 320  *   @create: != 0 if allowed to create new tunnel if no match found
 321  *
 322  * Description:
 323  *   ip6_tnl_locate() first tries to locate an existing tunnel
 324  *   based on @parms. If this is unsuccessful, but @create is set a new
 325  *   tunnel device is created and registered for use.
 326  *
 327  * Return:
 328  *   matching tunnel or error pointer
 329  **/
 330
 331 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 332                 struct __ip6_tnl_parm *p, int create)
 333 {
 334         const struct in6_addr *remote = &p->raddr;
 335         const struct in6_addr *local = &p->laddr;
 336         struct ip6_tnl __rcu **tp;
 337         struct ip6_tnl *t;
 338         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 339
 340         for (tp = ip6_tnl_bucket(ip6n, p);
 341              (t = rtnl_dereference(*tp)) != NULL;
 342              tp = &t->next) {
 343                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 344                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 345                         if (create)
 346                                 return ERR_PTR(-EEXIST);
 347
 348                         return t;
 349                 }
 350         }
 351         if (!create)
 352                 return ERR_PTR(-ENODEV);
 353         return ip6_tnl_create(net, p);
 354 }
 355
 356 /**
 357  * ip6_tnl_dev_uninit - tunnel device uninitializer
 358  *   @dev: the device to be destroyed
 359  *
 360  * Description:
 361  *   ip6_tnl_dev_uninit() removes tunnel from its list
 362  **/
 363
 364 static void
 365 ip6_tnl_dev_uninit(struct net_device *dev)
 366 {
 367         struct ip6_tnl *t = netdev_priv(dev);
 368         struct net *net = t->net;
 369         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 370
 371         if (dev == ip6n->fb_tnl_dev)
 372                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 373         else
 374                 ip6_tnl_unlink(ip6n, t);
 375         dst_cache_reset(&t->dst_cache);
 376         dev_put(dev);
 377 }
 378
 379 /**
 380  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 381  *   @skb: received socket buffer
 382  *
 383  * Return:
 384  *   0 if none was found,
 385  *   else index to encapsulation limit
 386  **/
 387
 388 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 389 {
 390         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 391         __u8 nexthdr = ipv6h->nexthdr;
 392         __u16 off = sizeof(*ipv6h);
 393
 394         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 395                 __u16 optlen = 0;
 396                 struct ipv6_opt_hdr *hdr;
 397                 if (raw + off + sizeof(*hdr) > skb->data &&
 398                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 399                         break;
 400
 401                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 402                 if (nexthdr == NEXTHDR_FRAGMENT) {
 403                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 404                         if (frag_hdr->frag_off)
 405                                 break;
 406                         optlen = 8;
 407                 } else if (nexthdr == NEXTHDR_AUTH) {
 408                         optlen = (hdr->hdrlen + 2) << 2;
 409                 } else {
 410                         optlen = ipv6_optlen(hdr);
 411                 }
 412                 if (nexthdr == NEXTHDR_DEST) {
 413                         __u16 i = off + 2;
 414                         while (1) {
 415                                 struct ipv6_tlv_tnl_enc_lim *tel;
 416
 417                                 /* No more room for encapsulation limit */
 418                                 if (i + sizeof (*tel) > off + optlen)
 419                                         break;
 420
 421                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 422                                 /* return index of option if found and valid */
 423                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 424                                     tel->length == 1)
 425                                         return i;
 426                                 /* else jump to next option */
 427                                 if (tel->type)
 428                                         i += tel->length + 2;
 429                                 else
 430                                         i++;
 431                         }
 432                 }
 433                 nexthdr = hdr->nexthdr;
 434                 off += optlen;
 435         }
 436         return 0;
 437 }
 438 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 439
 440 /**
 441  * ip6_tnl_err - tunnel error handler
 442  *
 443  * Description:
 444  *   ip6_tnl_err() should handle errors in the tunnel according
 445  *   to the specifications in RFC 2473.
 446  **/
 447
 448 static int
 449 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 450             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 451 {
 452         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 453         struct ip6_tnl *t;
 454         int rel_msg = 0;
 455         u8 rel_type = ICMPV6_DEST_UNREACH;
 456         u8 rel_code = ICMPV6_ADDR_UNREACH;
 457         u8 tproto;
 458         __u32 rel_info = 0;
 459         __u16 len;
 460         int err = -ENOENT;
 461
 462         /* If the packet doesn't contain the original IPv6 header we are
 463            in trouble since we might need the source address for further
 464            processing of the error. */
 465
 466         rcu_read_lock();
 467         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 468         if (!t)
 469                 goto out;
 470
 471         tproto = ACCESS_ONCE(t->parms.proto);
 472         if (tproto != ipproto && tproto != 0)
 473                 goto out;
 474
 475         err = 0;
 476
 477         switch (*type) {
 478                 __u32 teli;
 479                 struct ipv6_tlv_tnl_enc_lim *tel;
 480                 __u32 mtu;
 481         case ICMPV6_DEST_UNREACH:
 482                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 483                                     t->parms.name);
 484                 rel_msg = 1;
 485                 break;
 486         case ICMPV6_TIME_EXCEED:
 487                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 488                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 489                                             t->parms.name);
 490                         rel_msg = 1;
 491                 }
 492                 break;
 493         case ICMPV6_PARAMPROB:
 494                 teli = 0;
 495                 if ((*code) == ICMPV6_HDR_FIELD)
 496                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 497
 498                 if (teli && teli == *info - 2) {
 499                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 500                         if (tel->encap_limit == 0) {
 501                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 502                                                     t->parms.name);
 503                                 rel_msg = 1;
 504                         }
 505                 } else {
 506                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 507                                             t->parms.name);
 508                 }
 509                 break;
 510         case ICMPV6_PKT_TOOBIG:
 511                 mtu = *info - offset;
 512                 if (mtu < IPV6_MIN_MTU)
 513                         mtu = IPV6_MIN_MTU;
 514                 t->dev->mtu = mtu;
 515
 516                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 517                 if (len > mtu) {
 518                         rel_type = ICMPV6_PKT_TOOBIG;
 519                         rel_code = 0;
 520                         rel_info = mtu;
 521                         rel_msg = 1;
 522                 }
 523                 break;
 524         }
 525
 526         *type = rel_type;
 527         *code = rel_code;
 528         *info = rel_info;
 529         *msg = rel_msg;
 530
 531 out:
 532         rcu_read_unlock();
 533         return err;
 534 }
 535
 536 static int
 537 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 538            u8 type, u8 code, int offset, __be32 info)
 539 {
 540         int rel_msg = 0;
 541         u8 rel_type = type;
 542         u8 rel_code = code;
 543         __u32 rel_info = ntohl(info);
 544         int err;
 545         struct sk_buff *skb2;
 546         const struct iphdr *eiph;
 547         struct rtable *rt;
 548         struct flowi4 fl4;
 549
 550         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 551                           &rel_msg, &rel_info, offset);
 552         if (err < 0)
 553                 return err;
 554
 555         if (rel_msg == 0)
 556                 return 0;
 557
 558         switch (rel_type) {
 559         case ICMPV6_DEST_UNREACH:
 560                 if (rel_code != ICMPV6_ADDR_UNREACH)
 561                         return 0;
 562                 rel_type = ICMP_DEST_UNREACH;
 563                 rel_code = ICMP_HOST_UNREACH;
 564                 break;
 565         case ICMPV6_PKT_TOOBIG:
 566                 if (rel_code != 0)
 567                         return 0;
 568                 rel_type = ICMP_DEST_UNREACH;
 569                 rel_code = ICMP_FRAG_NEEDED;
 570                 break;
 571         case NDISC_REDIRECT:
 572                 rel_type = ICMP_REDIRECT;
 573                 rel_code = ICMP_REDIR_HOST;
 574         default:
 575                 return 0;
 576         }
 577
 578         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 579                 return 0;
 580
 581         skb2 = skb_clone(skb, GFP_ATOMIC);
 582         if (!skb2)
 583                 return 0;
 584
 585         skb_dst_drop(skb2);
 586
 587         skb_pull(skb2, offset);
 588         skb_reset_network_header(skb2);
 589         eiph = ip_hdr(skb2);
 590
 591         /* Try to guess incoming interface */
 592         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 593                                    eiph->saddr, 0,
 594                                    0, 0,
 595                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 596         if (IS_ERR(rt))
 597                 goto out;
 598
 599         skb2->dev = rt->dst.dev;
 600
 601         /* route "incoming" packet */
 602         if (rt->rt_flags & RTCF_LOCAL) {
 603                 ip_rt_put(rt);
 604                 rt = NULL;
 605                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 606                                            eiph->daddr, eiph->saddr,
 607                                            0, 0,
 608                                            IPPROTO_IPIP,
 609                                            RT_TOS(eiph->tos), 0);
 610                 if (IS_ERR(rt) ||
 611                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 612                         if (!IS_ERR(rt))
 613                                 ip_rt_put(rt);
 614                         goto out;
 615                 }
 616                 skb_dst_set(skb2, &rt->dst);
 617         } else {
 618                 ip_rt_put(rt);
 619                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 620                                    skb2->dev) ||
 621                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 622                         goto out;
 623         }
 624
 625         /* change mtu on this route */
 626         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 627                 if (rel_info > dst_mtu(skb_dst(skb2)))
 628                         goto out;
 629
 630                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 631         }
 632         if (rel_type == ICMP_REDIRECT)
 633                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 634
 635         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 636
 637 out:
 638         kfree_skb(skb2);
 639         return 0;
 640 }
 641
 642 static int
 643 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 644            u8 type, u8 code, int offset, __be32 info)
 645 {
 646         int rel_msg = 0;
 647         u8 rel_type = type;
 648         u8 rel_code = code;
 649         __u32 rel_info = ntohl(info);
 650         int err;
 651
 652         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 653                           &rel_msg, &rel_info, offset);
 654         if (err < 0)
 655                 return err;
 656
 657         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 658                 struct rt6_info *rt;
 659                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 660
 661                 if (!skb2)
 662                         return 0;
 663
 664                 skb_dst_drop(skb2);
 665                 skb_pull(skb2, offset);
 666                 skb_reset_network_header(skb2);
 667
 668                 /* Try to guess incoming interface */
 669                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 670                                 NULL, 0, 0);
 671
 672                 if (rt && rt->dst.dev)
 673                         skb2->dev = rt->dst.dev;
 674
 675                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 676
 677                 ip6_rt_put(rt);
 678
 679                 kfree_skb(skb2);
 680         }
 681
 682         return 0;
 683 }
 684
 685 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 686                                        const struct ipv6hdr *ipv6h,
 687                                        struct sk_buff *skb)
 688 {
 689         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 690
 691         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 692                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 693
 694         return IP6_ECN_decapsulate(ipv6h, skb);
 695 }
 696
 697 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 698                                        const struct ipv6hdr *ipv6h,
 699                                        struct sk_buff *skb)
 700 {
 701         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 702                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 703
 704         return IP6_ECN_decapsulate(ipv6h, skb);
 705 }
 706
 707 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 708                              const struct in6_addr *laddr,
 709                              const struct in6_addr *raddr)
 710 {
 711         struct __ip6_tnl_parm *p = &t->parms;
 712         int ltype = ipv6_addr_type(laddr);
 713         int rtype = ipv6_addr_type(raddr);
 714         __u32 flags = 0;
 715
 716         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 717                 flags = IP6_TNL_F_CAP_PER_PACKET;
 718         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 719                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 720                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 721                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 722                 if (ltype&IPV6_ADDR_UNICAST)
 723                         flags |= IP6_TNL_F_CAP_XMIT;
 724                 if (rtype&IPV6_ADDR_UNICAST)
 725                         flags |= IP6_TNL_F_CAP_RCV;
 726         }
 727         return flags;
 728 }
 729 EXPORT_SYMBOL(ip6_tnl_get_cap);
 730
 731 /* called with rcu_read_lock() */
 732 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 733                                   const struct in6_addr *laddr,
 734                                   const struct in6_addr *raddr)
 735 {
 736         struct __ip6_tnl_parm *p = &t->parms;
 737         int ret = 0;
 738         struct net *net = t->net;
 739
 740         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 741             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 742              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 743                 struct net_device *ldev = NULL;
 744
 745                 if (p->link)
 746                         ldev = dev_get_by_index_rcu(net, p->link);
 747
 748                 if ((ipv6_addr_is_multicast(laddr) ||
 749                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 750                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 751                         ret = 1;
 752         }
 753         return ret;
 754 }
 755 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 756
 757 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 758                          const struct tnl_ptk_info *tpi,
 759                          struct metadata_dst *tun_dst,
 760                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 761                                                 const struct ipv6hdr *ipv6h,
 762                                                 struct sk_buff *skb),
 763                          bool log_ecn_err)
 764 {
 765         struct pcpu_sw_netstats *tstats;
 766         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 767         int err;
 768
 769         if ((!(tpi->flags & TUNNEL_CSUM) &&
 770              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 771             ((tpi->flags & TUNNEL_CSUM) &&
 772              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 773                 tunnel->dev->stats.rx_crc_errors++;
 774                 tunnel->dev->stats.rx_errors++;
 775                 goto drop;
 776         }
 777
 778         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 779                 if (!(tpi->flags & TUNNEL_SEQ) ||
 780                     (tunnel->i_seqno &&
 781                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 782                         tunnel->dev->stats.rx_fifo_errors++;
 783                         tunnel->dev->stats.rx_errors++;
 784                         goto drop;
 785                 }
 786                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 787         }
 788
 789         skb->protocol = tpi->proto;
 790
 791         /* Warning: All skb pointers will be invalidated! */
 792         if (tunnel->dev->type == ARPHRD_ETHER) {
 793                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 794                         tunnel->dev->stats.rx_length_errors++;
 795                         tunnel->dev->stats.rx_errors++;
 796                         goto drop;
 797                 }
 798
 799                 ipv6h = ipv6_hdr(skb);
 800                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 801                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 802         } else {
 803                 skb->dev = tunnel->dev;
 804         }
 805
 806         skb_reset_network_header(skb);
 807         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 808
 809         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 810
 811         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 812         if (unlikely(err)) {
 813                 if (log_ecn_err)
 814                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 815                                              &ipv6h->saddr,
 816                                              ipv6_get_dsfield(ipv6h));
 817                 if (err > 1) {
 818                         ++tunnel->dev->stats.rx_frame_errors;
 819                         ++tunnel->dev->stats.rx_errors;
 820                         goto drop;
 821                 }
 822         }
 823
 824         tstats = this_cpu_ptr(tunnel->dev->tstats);
 825         u64_stats_update_begin(&tstats->syncp);
 826         tstats->rx_packets++;
 827         tstats->rx_bytes += skb->len;
 828         u64_stats_update_end(&tstats->syncp);
 829
 830         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 831
 832         gro_cells_receive(&tunnel->gro_cells, skb);
 833         return 0;
 834
 835 drop:
 836         kfree_skb(skb);
 837         return 0;
 838 }
 839
 840 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 841                 const struct tnl_ptk_info *tpi,
 842                 struct metadata_dst *tun_dst,
 843                 bool log_ecn_err)
 844 {
 845         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
 846                              log_ecn_err);
 847 }
 848 EXPORT_SYMBOL(ip6_tnl_rcv);
 849
 850 static const struct tnl_ptk_info tpi_v6 = {
 851         /* no tunnel info required for ipxip6. */
 852         .proto = htons(ETH_P_IPV6),
 853 };
 854
 855 static const struct tnl_ptk_info tpi_v4 = {
 856         /* no tunnel info required for ipxip6. */
 857         .proto = htons(ETH_P_IP),
 858 };
 859
 860 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 861                       const struct tnl_ptk_info *tpi,
 862                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 863                                                   const struct ipv6hdr *ipv6h,
 864                                                   struct sk_buff *skb))
 865 {
 866         struct ip6_tnl *t;
 867         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 868         int ret = -1;
 869
 870         rcu_read_lock();
 871         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 872
 873         if (t) {
 874                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 875
 876                 if (tproto != ipproto && tproto != 0)
 877                         goto drop;
 878                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 879                         goto drop;
 880                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 881                         goto drop;
 882                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 883                         goto drop;
 884                 ret = __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
 885                                     log_ecn_error);
 886         }
 887
 888         rcu_read_unlock();
 889
 890         return ret;
 891
 892 drop:
 893         rcu_read_unlock();
 894         kfree_skb(skb);
 895         return 0;
 896 }
 897
 898 static int ip4ip6_rcv(struct sk_buff *skb)
 899 {
 900         return ipxip6_rcv(skb, IPPROTO_IP, &tpi_v4,
 901                           ip4ip6_dscp_ecn_decapsulate);
 902 }
 903
 904 static int ip6ip6_rcv(struct sk_buff *skb)
 905 {
 906         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 907                           ip6ip6_dscp_ecn_decapsulate);
 908 }
 909
 910 struct ipv6_tel_txoption {
 911         struct ipv6_txoptions ops;
 912         __u8 dst_opt[8];
 913 };
 914
 915 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 916 {
 917         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 918
 919         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 920         opt->dst_opt[3] = 1;
 921         opt->dst_opt[4] = encap_limit;
 922         opt->dst_opt[5] = IPV6_TLV_PADN;
 923         opt->dst_opt[6] = 1;
 924
 925         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 926         opt->ops.opt_nflen = 8;
 927 }
 928
 929 /**
 930  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 931  *   @t: the outgoing tunnel device
 932  *   @hdr: IPv6 header from the incoming packet
 933  *
 934  * Description:
 935  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 936  *   doesn't match source of incoming packet.
 937  *
 938  * Return:
 939  *   1 if conflict,
 940  *   0 else
 941  **/
 942
 943 static inline bool
 944 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 945 {
 946         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 947 }
 948
 949 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 950                      const struct in6_addr *laddr,
 951                      const struct in6_addr *raddr)
 952 {
 953         struct __ip6_tnl_parm *p = &t->parms;
 954         int ret = 0;
 955         struct net *net = t->net;
 956
 957         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 958             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 959              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 960                 struct net_device *ldev = NULL;
 961
 962                 rcu_read_lock();
 963                 if (p->link)
 964                         ldev = dev_get_by_index_rcu(net, p->link);
 965
 966                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 967                         pr_warn("%s xmit: Local address not yet configured!\n",
 968                                 p->name);
 969                 else if (!ipv6_addr_is_multicast(raddr) &&
 970                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
 971                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 972                                 p->name);
 973                 else
 974                         ret = 1;
 975                 rcu_read_unlock();
 976         }
 977         return ret;
 978 }
 979 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
 980
 981 /**
 982  * ip6_tnl_xmit2 - encapsulate packet and send
 983  *   @skb: the outgoing socket buffer
 984  *   @dev: the outgoing tunnel device
 985  *   @dsfield: dscp code for outer header
 986  *   @fl: flow of tunneled packet
 987  *   @encap_limit: encapsulation limit
 988  *   @pmtu: Path MTU is stored if packet is too big
 989  *
 990  * Description:
 991  *   Build new header and do some sanity checks on the packet before sending
 992  *   it.
 993  *
 994  * Return:
 995  *   0 on success
 996  *   -1 fail
 997  *   %-EMSGSIZE message too big. return mtu in this case.
 998  **/
 999
1000 static int ip6_tnl_xmit2(struct sk_buff *skb,
1001                          struct net_device *dev,
1002                          __u8 dsfield,
1003                          struct flowi6 *fl6,
1004                          int encap_limit,
1005                          __u32 *pmtu)
1006 {
1007         struct ip6_tnl *t = netdev_priv(dev);
1008         struct net *net = t->net;
1009         struct net_device_stats *stats = &t->dev->stats;
1010         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1011         struct ipv6_tel_txoption opt;
1012         struct dst_entry *dst = NULL, *ndst = NULL;
1013         struct net_device *tdev;
1014         int mtu;
1015         unsigned int max_headroom = sizeof(struct ipv6hdr);
1016         u8 proto;
1017         int err = -1;
1018
1019         /* NBMA tunnel */
1020         if (ipv6_addr_any(&t->parms.raddr)) {
1021                 struct in6_addr *addr6;
1022                 struct neighbour *neigh;
1023                 int addr_type;
1024
1025                 if (!skb_dst(skb))
1026                         goto tx_err_link_failure;
1027
1028                 neigh = dst_neigh_lookup(skb_dst(skb),
1029                                          &ipv6_hdr(skb)->daddr);
1030                 if (!neigh)
1031                         goto tx_err_link_failure;
1032
1033                 addr6 = (struct in6_addr *)&neigh->primary_key;
1034                 addr_type = ipv6_addr_type(addr6);
1035
1036                 if (addr_type == IPV6_ADDR_ANY)
1037                         addr6 = &ipv6_hdr(skb)->daddr;
1038
1039                 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1040                 neigh_release(neigh);
1041         } else if (!fl6->flowi6_mark)
1042                 dst = dst_cache_get(&t->dst_cache);
1043
1044         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1045                 goto tx_err_link_failure;
1046
1047         if (!dst) {
1048                 dst = ip6_route_output(net, NULL, fl6);
1049
1050                 if (dst->error)
1051                         goto tx_err_link_failure;
1052                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1053                 if (IS_ERR(dst)) {
1054                         err = PTR_ERR(dst);
1055                         dst = NULL;
1056                         goto tx_err_link_failure;
1057                 }
1058                 ndst = dst;
1059         }
1060
1061         tdev = dst->dev;
1062
1063         if (tdev == dev) {
1064                 stats->collisions++;
1065                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1066                                      t->parms.name);
1067                 goto tx_err_dst_release;
1068         }
1069         mtu = dst_mtu(dst) - sizeof(*ipv6h);
1070         if (encap_limit >= 0) {
1071                 max_headroom += 8;
1072                 mtu -= 8;
1073         }
1074         if (mtu < IPV6_MIN_MTU)
1075                 mtu = IPV6_MIN_MTU;
1076         if (skb_dst(skb))
1077                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1078         if (skb->len > mtu) {
1079                 *pmtu = mtu;
1080                 err = -EMSGSIZE;
1081                 goto tx_err_dst_release;
1082         }
1083
1084         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1085
1086         /*
1087          * Okay, now see if we can stuff it in the buffer as-is.
1088          */
1089         max_headroom += LL_RESERVED_SPACE(tdev);
1090
1091         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1092             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1093                 struct sk_buff *new_skb;
1094
1095                 new_skb = skb_realloc_headroom(skb, max_headroom);
1096                 if (!new_skb)
1097                         goto tx_err_dst_release;
1098
1099                 if (skb->sk)
1100                         skb_set_owner_w(new_skb, skb->sk);
1101                 consume_skb(skb);
1102                 skb = new_skb;
1103         }
1104
1105         if (!fl6->flowi6_mark && ndst)
1106                 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1107         skb_dst_set(skb, dst);
1108
1109         skb->transport_header = skb->network_header;
1110
1111         proto = fl6->flowi6_proto;
1112         if (encap_limit >= 0) {
1113                 init_tel_txopt(&opt, encap_limit);
1114                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1115         }
1116
1117         if (likely(!skb->encapsulation)) {
1118                 skb_reset_inner_headers(skb);
1119                 skb->encapsulation = 1;
1120         }
1121
1122         skb_push(skb, sizeof(struct ipv6hdr));
1123         skb_reset_network_header(skb);
1124         ipv6h = ipv6_hdr(skb);
1125         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1126                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1127         ipv6h->hop_limit = t->parms.hop_limit;
1128         ipv6h->nexthdr = proto;
1129         ipv6h->saddr = fl6->saddr;
1130         ipv6h->daddr = fl6->daddr;
1131         ip6tunnel_xmit(NULL, skb, dev);
1132         return 0;
1133 tx_err_link_failure:
1134         stats->tx_carrier_errors++;
1135         dst_link_failure(skb);
1136 tx_err_dst_release:
1137         dst_release(dst);
1138         return err;
1139 }
1140
1141 static inline int
1142 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1143 {
1144         struct ip6_tnl *t = netdev_priv(dev);
1145         const struct iphdr  *iph = ip_hdr(skb);
1146         int encap_limit = -1;
1147         struct flowi6 fl6;
1148         __u8 dsfield;
1149         __u32 mtu;
1150         u8 tproto;
1151         int err;
1152
1153         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1154
1155         tproto = ACCESS_ONCE(t->parms.proto);
1156         if (tproto != IPPROTO_IPIP && tproto != 0)
1157                 return -1;
1158
1159         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1160                 encap_limit = t->parms.encap_limit;
1161
1162         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1163         fl6.flowi6_proto = IPPROTO_IPIP;
1164
1165         dsfield = ipv4_get_dsfield(iph);
1166
1167         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1168                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1169                                           & IPV6_TCLASS_MASK;
1170         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1171                 fl6.flowi6_mark = skb->mark;
1172
1173         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1174         if (err != 0) {
1175                 /* XXX: send ICMP error even if DF is not set. */
1176                 if (err == -EMSGSIZE)
1177                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1178                                   htonl(mtu));
1179                 return -1;
1180         }
1181
1182         return 0;
1183 }
1184
1185 static inline int
1186 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1187 {
1188         struct ip6_tnl *t = netdev_priv(dev);
1189         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1190         int encap_limit = -1;
1191         __u16 offset;
1192         struct flowi6 fl6;
1193         __u8 dsfield;
1194         __u32 mtu;
1195         u8 tproto;
1196         int err;
1197
1198         tproto = ACCESS_ONCE(t->parms.proto);
1199         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1200             ip6_tnl_addr_conflict(t, ipv6h))
1201                 return -1;
1202
1203         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1204         if (offset > 0) {
1205                 struct ipv6_tlv_tnl_enc_lim *tel;
1206                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1207                 if (tel->encap_limit == 0) {
1208                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1209                                     ICMPV6_HDR_FIELD, offset + 2);
1210                         return -1;
1211                 }
1212                 encap_limit = tel->encap_limit - 1;
1213         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1214                 encap_limit = t->parms.encap_limit;
1215
1216         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1217         fl6.flowi6_proto = IPPROTO_IPV6;
1218
1219         dsfield = ipv6_get_dsfield(ipv6h);
1220         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1221                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1222         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1223                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1224         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1225                 fl6.flowi6_mark = skb->mark;
1226
1227         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1228         if (err != 0) {
1229                 if (err == -EMSGSIZE)
1230                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1231                 return -1;
1232         }
1233
1234         return 0;
1235 }
1236
1237 static netdev_tx_t
1238 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1239 {
1240         struct ip6_tnl *t = netdev_priv(dev);
1241         struct net_device_stats *stats = &t->dev->stats;
1242         int ret;
1243
1244         switch (skb->protocol) {
1245         case htons(ETH_P_IP):
1246                 ret = ip4ip6_tnl_xmit(skb, dev);
1247                 break;
1248         case htons(ETH_P_IPV6):
1249                 ret = ip6ip6_tnl_xmit(skb, dev);
1250                 break;
1251         default:
1252                 goto tx_err;
1253         }
1254
1255         if (ret < 0)
1256                 goto tx_err;
1257
1258         return NETDEV_TX_OK;
1259
1260 tx_err:
1261         stats->tx_errors++;
1262         stats->tx_dropped++;
1263         kfree_skb(skb);
1264         return NETDEV_TX_OK;
1265 }
1266
1267 static void ip6_tnl_link_config(struct ip6_tnl *t)
1268 {
1269         struct net_device *dev = t->dev;
1270         struct __ip6_tnl_parm *p = &t->parms;
1271         struct flowi6 *fl6 = &t->fl.u.ip6;
1272
1273         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1274         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1275
1276         /* Set up flowi template */
1277         fl6->saddr = p->laddr;
1278         fl6->daddr = p->raddr;
1279         fl6->flowi6_oif = p->link;
1280         fl6->flowlabel = 0;
1281
1282         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1283                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1284         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1285                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1286
1287         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1288         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1289
1290         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1291                 dev->flags |= IFF_POINTOPOINT;
1292         else
1293                 dev->flags &= ~IFF_POINTOPOINT;
1294
1295         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1296                 int strict = (ipv6_addr_type(&p->raddr) &
1297                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1298
1299                 struct rt6_info *rt = rt6_lookup(t->net,
1300                                                  &p->raddr, &p->laddr,
1301                                                  p->link, strict);
1302
1303                 if (!rt)
1304                         return;
1305
1306                 if (rt->dst.dev) {
1307                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1308                                 sizeof(struct ipv6hdr);
1309
1310                         dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1311                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1312                                 dev->mtu -= 8;
1313
1314                         if (dev->mtu < IPV6_MIN_MTU)
1315                                 dev->mtu = IPV6_MIN_MTU;
1316                 }
1317                 ip6_rt_put(rt);
1318         }
1319 }
1320
1321 /**
1322  * ip6_tnl_change - update the tunnel parameters
1323  *   @t: tunnel to be changed
1324  *   @p: tunnel configuration parameters
1325  *
1326  * Description:
1327  *   ip6_tnl_change() updates the tunnel parameters
1328  **/
1329
1330 static int
1331 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1332 {
1333         t->parms.laddr = p->laddr;
1334         t->parms.raddr = p->raddr;
1335         t->parms.flags = p->flags;
1336         t->parms.hop_limit = p->hop_limit;
1337         t->parms.encap_limit = p->encap_limit;
1338         t->parms.flowinfo = p->flowinfo;
1339         t->parms.link = p->link;
1340         t->parms.proto = p->proto;
1341         dst_cache_reset(&t->dst_cache);
1342         ip6_tnl_link_config(t);
1343         return 0;
1344 }
1345
1346 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1347 {
1348         struct net *net = t->net;
1349         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1350         int err;
1351
1352         ip6_tnl_unlink(ip6n, t);
1353         synchronize_net();
1354         err = ip6_tnl_change(t, p);
1355         ip6_tnl_link(ip6n, t);
1356         netdev_state_change(t->dev);
1357         return err;
1358 }
1359
1360 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1361 {
1362         /* for default tnl0 device allow to change only the proto */
1363         t->parms.proto = p->proto;
1364         netdev_state_change(t->dev);
1365         return 0;
1366 }
1367
1368 static void
1369 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1370 {
1371         p->laddr = u->laddr;
1372         p->raddr = u->raddr;
1373         p->flags = u->flags;
1374         p->hop_limit = u->hop_limit;
1375         p->encap_limit = u->encap_limit;
1376         p->flowinfo = u->flowinfo;
1377         p->link = u->link;
1378         p->proto = u->proto;
1379         memcpy(p->name, u->name, sizeof(u->name));
1380 }
1381
1382 static void
1383 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1384 {
1385         u->laddr = p->laddr;
1386         u->raddr = p->raddr;
1387         u->flags = p->flags;
1388         u->hop_limit = p->hop_limit;
1389         u->encap_limit = p->encap_limit;
1390         u->flowinfo = p->flowinfo;
1391         u->link = p->link;
1392         u->proto = p->proto;
1393         memcpy(u->name, p->name, sizeof(u->name));
1394 }
1395
1396 /**
1397  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1398  *   @dev: virtual device associated with tunnel
1399  *   @ifr: parameters passed from userspace
1400  *   @cmd: command to be performed
1401  *
1402  * Description:
1403  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1404  *   from userspace.
1405  *
1406  *   The possible commands are the following:
1407  *     %SIOCGETTUNNEL: get tunnel parameters for device
1408  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1409  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1410  *     %SIOCDELTUNNEL: delete tunnel
1411  *
1412  *   The fallback device "ip6tnl0", created during module
1413  *   initialization, can be used for creating other tunnel devices.
1414  *
1415  * Return:
1416  *   0 on success,
1417  *   %-EFAULT if unable to copy data to or from userspace,
1418  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1419  *   %-EINVAL if passed tunnel parameters are invalid,
1420  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1421  *   %-ENODEV if attempting to change or delete a nonexisting device
1422  **/
1423
1424 static int
1425 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1426 {
1427         int err = 0;
1428         struct ip6_tnl_parm p;
1429         struct __ip6_tnl_parm p1;
1430         struct ip6_tnl *t = netdev_priv(dev);
1431         struct net *net = t->net;
1432         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1433
1434         memset(&p1, 0, sizeof(p1));
1435
1436         switch (cmd) {
1437         case SIOCGETTUNNEL:
1438                 if (dev == ip6n->fb_tnl_dev) {
1439                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1440                                 err = -EFAULT;
1441                                 break;
1442                         }
1443                         ip6_tnl_parm_from_user(&p1, &p);
1444                         t = ip6_tnl_locate(net, &p1, 0);
1445                         if (IS_ERR(t))
1446                                 t = netdev_priv(dev);
1447                 } else {
1448                         memset(&p, 0, sizeof(p));
1449                 }
1450                 ip6_tnl_parm_to_user(&p, &t->parms);
1451                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1452                         err = -EFAULT;
1453                 }
1454                 break;
1455         case SIOCADDTUNNEL:
1456         case SIOCCHGTUNNEL:
1457                 err = -EPERM;
1458                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1459                         break;
1460                 err = -EFAULT;
1461                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1462                         break;
1463                 err = -EINVAL;
1464                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1465                     p.proto != 0)
1466                         break;
1467                 ip6_tnl_parm_from_user(&p1, &p);
1468                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1469                 if (cmd == SIOCCHGTUNNEL) {
1470                         if (!IS_ERR(t)) {
1471                                 if (t->dev != dev) {
1472                                         err = -EEXIST;
1473                                         break;
1474                                 }
1475                         } else
1476                                 t = netdev_priv(dev);
1477                         if (dev == ip6n->fb_tnl_dev)
1478                                 err = ip6_tnl0_update(t, &p1);
1479                         else
1480                                 err = ip6_tnl_update(t, &p1);
1481                 }
1482                 if (!IS_ERR(t)) {
1483                         err = 0;
1484                         ip6_tnl_parm_to_user(&p, &t->parms);
1485                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1486                                 err = -EFAULT;
1487
1488                 } else {
1489                         err = PTR_ERR(t);
1490                 }
1491                 break;
1492         case SIOCDELTUNNEL:
1493                 err = -EPERM;
1494                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1495                         break;
1496
1497                 if (dev == ip6n->fb_tnl_dev) {
1498                         err = -EFAULT;
1499                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1500                                 break;
1501                         err = -ENOENT;
1502                         ip6_tnl_parm_from_user(&p1, &p);
1503                         t = ip6_tnl_locate(net, &p1, 0);
1504                         if (IS_ERR(t))
1505                                 break;
1506                         err = -EPERM;
1507                         if (t->dev == ip6n->fb_tnl_dev)
1508                                 break;
1509                         dev = t->dev;
1510                 }
1511                 err = 0;
1512                 unregister_netdevice(dev);
1513                 break;
1514         default:
1515                 err = -EINVAL;
1516         }
1517         return err;
1518 }
1519
1520 /**
1521  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1522  *   @dev: virtual device associated with tunnel
1523  *   @new_mtu: the new mtu
1524  *
1525  * Return:
1526  *   0 on success,
1527  *   %-EINVAL if mtu too small
1528  **/
1529
1530 static int
1531 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1532 {
1533         struct ip6_tnl *tnl = netdev_priv(dev);
1534
1535         if (tnl->parms.proto == IPPROTO_IPIP) {
1536                 if (new_mtu < 68)
1537                         return -EINVAL;
1538         } else {
1539                 if (new_mtu < IPV6_MIN_MTU)
1540                         return -EINVAL;
1541         }
1542         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1543                 return -EINVAL;
1544         dev->mtu = new_mtu;
1545         return 0;
1546 }
1547
1548 int ip6_tnl_get_iflink(const struct net_device *dev)
1549 {
1550         struct ip6_tnl *t = netdev_priv(dev);
1551
1552         return t->parms.link;
1553 }
1554 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1555
1556 static const struct net_device_ops ip6_tnl_netdev_ops = {
1557         .ndo_init       = ip6_tnl_dev_init,
1558         .ndo_uninit     = ip6_tnl_dev_uninit,
1559         .ndo_start_xmit = ip6_tnl_xmit,
1560         .ndo_do_ioctl   = ip6_tnl_ioctl,
1561         .ndo_change_mtu = ip6_tnl_change_mtu,
1562         .ndo_get_stats  = ip6_get_stats,
1563         .ndo_get_iflink = ip6_tnl_get_iflink,
1564 };
1565
1566
1567 /**
1568  * ip6_tnl_dev_setup - setup virtual tunnel device
1569  *   @dev: virtual device associated with tunnel
1570  *
1571  * Description:
1572  *   Initialize function pointers and device parameters
1573  **/
1574
1575 static void ip6_tnl_dev_setup(struct net_device *dev)
1576 {
1577         struct ip6_tnl *t;
1578
1579         dev->netdev_ops = &ip6_tnl_netdev_ops;
1580         dev->destructor = ip6_dev_free;
1581
1582         dev->type = ARPHRD_TUNNEL6;
1583         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1584         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1585         t = netdev_priv(dev);
1586         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1587                 dev->mtu -= 8;
1588         dev->flags |= IFF_NOARP;
1589         dev->addr_len = sizeof(struct in6_addr);
1590         netif_keep_dst(dev);
1591         /* This perm addr will be used as interface identifier by IPv6 */
1592         dev->addr_assign_type = NET_ADDR_RANDOM;
1593         eth_random_addr(dev->perm_addr);
1594 }
1595
1596
1597 /**
1598  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1599  *   @dev: virtual device associated with tunnel
1600  **/
1601
1602 static inline int
1603 ip6_tnl_dev_init_gen(struct net_device *dev)
1604 {
1605         struct ip6_tnl *t = netdev_priv(dev);
1606         int ret;
1607
1608         t->dev = dev;
1609         t->net = dev_net(dev);
1610         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1611         if (!dev->tstats)
1612                 return -ENOMEM;
1613
1614         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1615         if (ret)
1616                 goto free_stats;
1617
1618         ret = gro_cells_init(&t->gro_cells, dev);
1619         if (ret)
1620                 goto destroy_dst;
1621
1622         return 0;
1623
1624 destroy_dst:
1625         dst_cache_destroy(&t->dst_cache);
1626 free_stats:
1627         free_percpu(dev->tstats);
1628         dev->tstats = NULL;
1629
1630         return ret;
1631 }
1632
1633 /**
1634  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1635  *   @dev: virtual device associated with tunnel
1636  **/
1637
1638 static int ip6_tnl_dev_init(struct net_device *dev)
1639 {
1640         struct ip6_tnl *t = netdev_priv(dev);
1641         int err = ip6_tnl_dev_init_gen(dev);
1642
1643         if (err)
1644                 return err;
1645         ip6_tnl_link_config(t);
1646         return 0;
1647 }
1648
1649 /**
1650  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1651  *   @dev: fallback device
1652  *
1653  * Return: 0
1654  **/
1655
1656 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1657 {
1658         struct ip6_tnl *t = netdev_priv(dev);
1659         struct net *net = dev_net(dev);
1660         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1661
1662         t->parms.proto = IPPROTO_IPV6;
1663         dev_hold(dev);
1664
1665         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1666         return 0;
1667 }
1668
1669 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1670 {
1671         u8 proto;
1672
1673         if (!data || !data[IFLA_IPTUN_PROTO])
1674                 return 0;
1675
1676         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1677         if (proto != IPPROTO_IPV6 &&
1678             proto != IPPROTO_IPIP &&
1679             proto != 0)
1680                 return -EINVAL;
1681
1682         return 0;
1683 }
1684
1685 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1686                                   struct __ip6_tnl_parm *parms)
1687 {
1688         memset(parms, 0, sizeof(*parms));
1689
1690         if (!data)
1691                 return;
1692
1693         if (data[IFLA_IPTUN_LINK])
1694                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1695
1696         if (data[IFLA_IPTUN_LOCAL])
1697                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1698
1699         if (data[IFLA_IPTUN_REMOTE])
1700                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1701
1702         if (data[IFLA_IPTUN_TTL])
1703                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1704
1705         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1706                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1707
1708         if (data[IFLA_IPTUN_FLOWINFO])
1709                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1710
1711         if (data[IFLA_IPTUN_FLAGS])
1712                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1713
1714         if (data[IFLA_IPTUN_PROTO])
1715                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1716 }
1717
1718 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1719                            struct nlattr *tb[], struct nlattr *data[])
1720 {
1721         struct net *net = dev_net(dev);
1722         struct ip6_tnl *nt, *t;
1723
1724         nt = netdev_priv(dev);
1725         ip6_tnl_netlink_parms(data, &nt->parms);
1726
1727         t = ip6_tnl_locate(net, &nt->parms, 0);
1728         if (!IS_ERR(t))
1729                 return -EEXIST;
1730
1731         return ip6_tnl_create2(dev);
1732 }
1733
1734 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1735                               struct nlattr *data[])
1736 {
1737         struct ip6_tnl *t = netdev_priv(dev);
1738         struct __ip6_tnl_parm p;
1739         struct net *net = t->net;
1740         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1741
1742         if (dev == ip6n->fb_tnl_dev)
1743                 return -EINVAL;
1744
1745         ip6_tnl_netlink_parms(data, &p);
1746
1747         t = ip6_tnl_locate(net, &p, 0);
1748         if (!IS_ERR(t)) {
1749                 if (t->dev != dev)
1750                         return -EEXIST;
1751         } else
1752                 t = netdev_priv(dev);
1753
1754         return ip6_tnl_update(t, &p);
1755 }
1756
1757 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1758 {
1759         struct net *net = dev_net(dev);
1760         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1761
1762         if (dev != ip6n->fb_tnl_dev)
1763                 unregister_netdevice_queue(dev, head);
1764 }
1765
1766 static size_t ip6_tnl_get_size(const struct net_device *dev)
1767 {
1768         return
1769                 /* IFLA_IPTUN_LINK */
1770                 nla_total_size(4) +
1771                 /* IFLA_IPTUN_LOCAL */
1772                 nla_total_size(sizeof(struct in6_addr)) +
1773                 /* IFLA_IPTUN_REMOTE */
1774                 nla_total_size(sizeof(struct in6_addr)) +
1775                 /* IFLA_IPTUN_TTL */
1776                 nla_total_size(1) +
1777                 /* IFLA_IPTUN_ENCAP_LIMIT */
1778                 nla_total_size(1) +
1779                 /* IFLA_IPTUN_FLOWINFO */
1780                 nla_total_size(4) +
1781                 /* IFLA_IPTUN_FLAGS */
1782                 nla_total_size(4) +
1783                 /* IFLA_IPTUN_PROTO */
1784                 nla_total_size(1) +
1785                 0;
1786 }
1787
1788 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1789 {
1790         struct ip6_tnl *tunnel = netdev_priv(dev);
1791         struct __ip6_tnl_parm *parm = &tunnel->parms;
1792
1793         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1794             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1795             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1796             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1797             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1798             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1799             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1800             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1801                 goto nla_put_failure;
1802         return 0;
1803
1804 nla_put_failure:
1805         return -EMSGSIZE;
1806 }
1807
1808 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1809 {
1810         struct ip6_tnl *tunnel = netdev_priv(dev);
1811
1812         return tunnel->net;
1813 }
1814 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1815
1816 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1817         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1818         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1819         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1820         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1821         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1822         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1823         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1824         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1825 };
1826
1827 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1828         .kind           = "ip6tnl",
1829         .maxtype        = IFLA_IPTUN_MAX,
1830         .policy         = ip6_tnl_policy,
1831         .priv_size      = sizeof(struct ip6_tnl),
1832         .setup          = ip6_tnl_dev_setup,
1833         .validate       = ip6_tnl_validate,
1834         .newlink        = ip6_tnl_newlink,
1835         .changelink     = ip6_tnl_changelink,
1836         .dellink        = ip6_tnl_dellink,
1837         .get_size       = ip6_tnl_get_size,
1838         .fill_info      = ip6_tnl_fill_info,
1839         .get_link_net   = ip6_tnl_get_link_net,
1840 };
1841
1842 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1843         .handler        = ip4ip6_rcv,
1844         .err_handler    = ip4ip6_err,
1845         .priority       =       1,
1846 };
1847
1848 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1849         .handler        = ip6ip6_rcv,
1850         .err_handler    = ip6ip6_err,
1851         .priority       =       1,
1852 };
1853
1854 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1855 {
1856         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1857         struct net_device *dev, *aux;
1858         int h;
1859         struct ip6_tnl *t;
1860         LIST_HEAD(list);
1861
1862         for_each_netdev_safe(net, dev, aux)
1863                 if (dev->rtnl_link_ops == &ip6_link_ops)
1864                         unregister_netdevice_queue(dev, &list);
1865
1866         for (h = 0; h < HASH_SIZE; h++) {
1867                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1868                 while (t) {
1869                         /* If dev is in the same netns, it has already
1870                          * been added to the list by the previous loop.
1871                          */
1872                         if (!net_eq(dev_net(t->dev), net))
1873                                 unregister_netdevice_queue(t->dev, &list);
1874                         t = rtnl_dereference(t->next);
1875                 }
1876         }
1877
1878         unregister_netdevice_many(&list);
1879 }
1880
1881 static int __net_init ip6_tnl_init_net(struct net *net)
1882 {
1883         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1884         struct ip6_tnl *t = NULL;
1885         int err;
1886
1887         ip6n->tnls[0] = ip6n->tnls_wc;
1888         ip6n->tnls[1] = ip6n->tnls_r_l;
1889
1890         err = -ENOMEM;
1891         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1892                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1893
1894         if (!ip6n->fb_tnl_dev)
1895                 goto err_alloc_dev;
1896         dev_net_set(ip6n->fb_tnl_dev, net);
1897         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1898         /* FB netdevice is special: we have one, and only one per netns.
1899          * Allowing to move it to another netns is clearly unsafe.
1900          */
1901         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1902
1903         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1904         if (err < 0)
1905                 goto err_register;
1906
1907         err = register_netdev(ip6n->fb_tnl_dev);
1908         if (err < 0)
1909                 goto err_register;
1910
1911         t = netdev_priv(ip6n->fb_tnl_dev);
1912
1913         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1914         return 0;
1915
1916 err_register:
1917         ip6_dev_free(ip6n->fb_tnl_dev);
1918 err_alloc_dev:
1919         return err;
1920 }
1921
1922 static void __net_exit ip6_tnl_exit_net(struct net *net)
1923 {
1924         rtnl_lock();
1925         ip6_tnl_destroy_tunnels(net);
1926         rtnl_unlock();
1927 }
1928
1929 static struct pernet_operations ip6_tnl_net_ops = {
1930         .init = ip6_tnl_init_net,
1931         .exit = ip6_tnl_exit_net,
1932         .id   = &ip6_tnl_net_id,
1933         .size = sizeof(struct ip6_tnl_net),
1934 };
1935
1936 /**
1937  * ip6_tunnel_init - register protocol and reserve needed resources
1938  *
1939  * Return: 0 on success
1940  **/
1941
1942 static int __init ip6_tunnel_init(void)
1943 {
1944         int  err;
1945
1946         err = register_pernet_device(&ip6_tnl_net_ops);
1947         if (err < 0)
1948                 goto out_pernet;
1949
1950         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1951         if (err < 0) {
1952                 pr_err("%s: can't register ip4ip6\n", __func__);
1953                 goto out_ip4ip6;
1954         }
1955
1956         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1957         if (err < 0) {
1958                 pr_err("%s: can't register ip6ip6\n", __func__);
1959                 goto out_ip6ip6;
1960         }
1961         err = rtnl_link_register(&ip6_link_ops);
1962         if (err < 0)
1963                 goto rtnl_link_failed;
1964
1965         return 0;
1966
1967 rtnl_link_failed:
1968         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1969 out_ip6ip6:
1970         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1971 out_ip4ip6:
1972         unregister_pernet_device(&ip6_tnl_net_ops);
1973 out_pernet:
1974         return err;
1975 }
1976
1977 /**
1978  * ip6_tunnel_cleanup - free resources and unregister protocol
1979  **/
1980
1981 static void __exit ip6_tunnel_cleanup(void)
1982 {
1983         rtnl_link_unregister(&ip6_link_ops);
1984         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1985                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1986
1987         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1988                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1989
1990         unregister_pernet_device(&ip6_tnl_net_ops);
1991 }
1992
1993 module_init(ip6_tunnel_init);
1994 module_exit(ip6_tunnel_cleanup);