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 #include <net/dst_metadata.h>
  61
  62 MODULE_AUTHOR("Ville Nuorvala");
  63 MODULE_DESCRIPTION("IPv6 tunneling device");
  64 MODULE_LICENSE("GPL");
  65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  66 MODULE_ALIAS_NETDEV("ip6tnl0");
  67
  68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
  69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
  70
  71 static bool log_ecn_error = true;
  72 module_param(log_ecn_error, bool, 0644);
  73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  74
  75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  76 {
  77         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  78
  79         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
  80 }
  81
  82 static int ip6_tnl_dev_init(struct net_device *dev);
  83 static void ip6_tnl_dev_setup(struct net_device *dev);
  84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  85
  86 static int ip6_tnl_net_id __read_mostly;
  87 struct ip6_tnl_net {
  88         /* the IPv6 tunnel fallback device */
  89         struct net_device *fb_tnl_dev;
  90         /* lists for storing tunnels in use */
  91         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
  92         struct ip6_tnl __rcu *tnls_wc[1];
  93         struct ip6_tnl __rcu **tnls[2];
  94         struct ip6_tnl __rcu *collect_md_tun;
  95 };
  96
  97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  98 {
  99         struct pcpu_sw_netstats tmp, sum = { 0 };
 100         int i;
 101
 102         for_each_possible_cpu(i) {
 103                 unsigned int start;
 104                 const struct pcpu_sw_netstats *tstats =
 105                                                    per_cpu_ptr(dev->tstats, i);
 106
 107                 do {
 108                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 109                         tmp.rx_packets = tstats->rx_packets;
 110                         tmp.rx_bytes = tstats->rx_bytes;
 111                         tmp.tx_packets = tstats->tx_packets;
 112                         tmp.tx_bytes =  tstats->tx_bytes;
 113                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 114
 115                 sum.rx_packets += tmp.rx_packets;
 116                 sum.rx_bytes   += tmp.rx_bytes;
 117                 sum.tx_packets += tmp.tx_packets;
 118                 sum.tx_bytes   += tmp.tx_bytes;
 119         }
 120         dev->stats.rx_packets = sum.rx_packets;
 121         dev->stats.rx_bytes   = sum.rx_bytes;
 122         dev->stats.tx_packets = sum.tx_packets;
 123         dev->stats.tx_bytes   = sum.tx_bytes;
 124         return &dev->stats;
 125 }
 126
 127 /**
 128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 129  *   @remote: the address of the tunnel exit-point
 130  *   @local: the address of the tunnel entry-point
 131  *
 132  * Return:
 133  *   tunnel matching given end-points if found,
 134  *   else fallback tunnel if its device is up,
 135  *   else %NULL
 136  **/
 137
 138 #define for_each_ip6_tunnel_rcu(start) \
 139         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 140
 141 static struct ip6_tnl *
 142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 143 {
 144         unsigned int hash = HASH(remote, local);
 145         struct ip6_tnl *t;
 146         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 147         struct in6_addr any;
 148
 149         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 150                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 151                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 152                     (t->dev->flags & IFF_UP))
 153                         return t;
 154         }
 155
 156         memset(&any, 0, sizeof(any));
 157         hash = HASH(&any, local);
 158         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 159                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 160                     ipv6_addr_any(&t->parms.raddr) &&
 161                     (t->dev->flags & IFF_UP))
 162                         return t;
 163         }
 164
 165         hash = HASH(remote, &any);
 166         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 167                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 168                     ipv6_addr_any(&t->parms.laddr) &&
 169                     (t->dev->flags & IFF_UP))
 170                         return t;
 171         }
 172
 173         t = rcu_dereference(ip6n->collect_md_tun);
 174         if (t)
 175                 return t;
 176
 177         t = rcu_dereference(ip6n->tnls_wc[0]);
 178         if (t && (t->dev->flags & IFF_UP))
 179                 return t;
 180
 181         return NULL;
 182 }
 183
 184 /**
 185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 186  *   @p: parameters containing tunnel end-points
 187  *
 188  * Description:
 189  *   ip6_tnl_bucket() returns the head of the list matching the
 190  *   &struct in6_addr entries laddr and raddr in @p.
 191  *
 192  * Return: head of IPv6 tunnel list
 193  **/
 194
 195 static struct ip6_tnl __rcu **
 196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 197 {
 198         const struct in6_addr *remote = &p->raddr;
 199         const struct in6_addr *local = &p->laddr;
 200         unsigned int h = 0;
 201         int prio = 0;
 202
 203         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 204                 prio = 1;
 205                 h = HASH(remote, local);
 206         }
 207         return &ip6n->tnls[prio][h];
 208 }
 209
 210 /**
 211  * ip6_tnl_link - add tunnel to hash table
 212  *   @t: tunnel to be added
 213  **/
 214
 215 static void
 216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 217 {
 218         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 219
 220         if (t->parms.collect_md)
 221                 rcu_assign_pointer(ip6n->collect_md_tun, t);
 222         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 223         rcu_assign_pointer(*tp, t);
 224 }
 225
 226 /**
 227  * ip6_tnl_unlink - remove tunnel from hash table
 228  *   @t: tunnel to be removed
 229  **/
 230
 231 static void
 232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 233 {
 234         struct ip6_tnl __rcu **tp;
 235         struct ip6_tnl *iter;
 236
 237         if (t->parms.collect_md)
 238                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 239
 240         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 241              (iter = rtnl_dereference(*tp)) != NULL;
 242              tp = &iter->next) {
 243                 if (t == iter) {
 244                         rcu_assign_pointer(*tp, t->next);
 245                         break;
 246                 }
 247         }
 248 }
 249
 250 static void ip6_dev_free(struct net_device *dev)
 251 {
 252         struct ip6_tnl *t = netdev_priv(dev);
 253
 254         gro_cells_destroy(&t->gro_cells);
 255         dst_cache_destroy(&t->dst_cache);
 256         free_percpu(dev->tstats);
 257         free_netdev(dev);
 258 }
 259
 260 static int ip6_tnl_create2(struct net_device *dev)
 261 {
 262         struct ip6_tnl *t = netdev_priv(dev);
 263         struct net *net = dev_net(dev);
 264         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 265         int err;
 266
 267         t = netdev_priv(dev);
 268
 269         dev->rtnl_link_ops = &ip6_link_ops;
 270         err = register_netdevice(dev);
 271         if (err < 0)
 272                 goto out;
 273
 274         strcpy(t->parms.name, dev->name);
 275
 276         ip6_tnl_link(ip6n, t);
 277         return 0;
 278
 279 out:
 280         return err;
 281 }
 282
 283 /**
 284  * ip6_tnl_create - create a new tunnel
 285  *   @p: tunnel parameters
 286  *   @pt: pointer to new tunnel
 287  *
 288  * Description:
 289  *   Create tunnel matching given parameters.
 290  *
 291  * Return:
 292  *   created tunnel or error pointer
 293  **/
 294
 295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 296 {
 297         struct net_device *dev;
 298         struct ip6_tnl *t;
 299         char name[IFNAMSIZ];
 300         int err = -E2BIG;
 301
 302         if (p->name[0]) {
 303                 if (!dev_valid_name(p->name))
 304                         goto failed;
 305                 strlcpy(name, p->name, IFNAMSIZ);
 306         } else {
 307                 sprintf(name, "ip6tnl%%d");
 308         }
 309         err = -ENOMEM;
 310         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 311                            ip6_tnl_dev_setup);
 312         if (!dev)
 313                 goto failed;
 314
 315         dev_net_set(dev, net);
 316
 317         t = netdev_priv(dev);
 318         t->parms = *p;
 319         t->net = dev_net(dev);
 320         err = ip6_tnl_create2(dev);
 321         if (err < 0)
 322                 goto failed_free;
 323
 324         return t;
 325
 326 failed_free:
 327         ip6_dev_free(dev);
 328 failed:
 329         return ERR_PTR(err);
 330 }
 331
 332 /**
 333  * ip6_tnl_locate - find or create tunnel matching given parameters
 334  *   @p: tunnel parameters
 335  *   @create: != 0 if allowed to create new tunnel if no match found
 336  *
 337  * Description:
 338  *   ip6_tnl_locate() first tries to locate an existing tunnel
 339  *   based on @parms. If this is unsuccessful, but @create is set a new
 340  *   tunnel device is created and registered for use.
 341  *
 342  * Return:
 343  *   matching tunnel or error pointer
 344  **/
 345
 346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 347                 struct __ip6_tnl_parm *p, int create)
 348 {
 349         const struct in6_addr *remote = &p->raddr;
 350         const struct in6_addr *local = &p->laddr;
 351         struct ip6_tnl __rcu **tp;
 352         struct ip6_tnl *t;
 353         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 354
 355         for (tp = ip6_tnl_bucket(ip6n, p);
 356              (t = rtnl_dereference(*tp)) != NULL;
 357              tp = &t->next) {
 358                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 359                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 360                         if (create)
 361                                 return ERR_PTR(-EEXIST);
 362
 363                         return t;
 364                 }
 365         }
 366         if (!create)
 367                 return ERR_PTR(-ENODEV);
 368         return ip6_tnl_create(net, p);
 369 }
 370
 371 /**
 372  * ip6_tnl_dev_uninit - tunnel device uninitializer
 373  *   @dev: the device to be destroyed
 374  *
 375  * Description:
 376  *   ip6_tnl_dev_uninit() removes tunnel from its list
 377  **/
 378
 379 static void
 380 ip6_tnl_dev_uninit(struct net_device *dev)
 381 {
 382         struct ip6_tnl *t = netdev_priv(dev);
 383         struct net *net = t->net;
 384         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 385
 386         if (dev == ip6n->fb_tnl_dev)
 387                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 388         else
 389                 ip6_tnl_unlink(ip6n, t);
 390         dst_cache_reset(&t->dst_cache);
 391         dev_put(dev);
 392 }
 393
 394 /**
 395  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 396  *   @skb: received socket buffer
 397  *
 398  * Return:
 399  *   0 if none was found,
 400  *   else index to encapsulation limit
 401  **/
 402
 403 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 404 {
 405         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 406         unsigned int nhoff = raw - skb->data;
 407         unsigned int off = nhoff + sizeof(*ipv6h);
 408         u8 next, nexthdr = ipv6h->nexthdr;
 409
 410         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 411                 struct ipv6_opt_hdr *hdr;
 412                 u16 optlen;
 413
 414                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 415                         break;
 416
 417                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 418                 if (nexthdr == NEXTHDR_FRAGMENT) {
 419                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 420                         if (frag_hdr->frag_off)
 421                                 break;
 422                         optlen = 8;
 423                 } else if (nexthdr == NEXTHDR_AUTH) {
 424                         optlen = (hdr->hdrlen + 2) << 2;
 425                 } else {
 426                         optlen = ipv6_optlen(hdr);
 427                 }
 428                 /* cache hdr->nexthdr, since pskb_may_pull() might
 429                  * invalidate hdr
 430                  */
 431                 next = hdr->nexthdr;
 432                 if (nexthdr == NEXTHDR_DEST) {
 433                         u16 i = 2;
 434
 435                         /* Remember : hdr is no longer valid at this point. */
 436                         if (!pskb_may_pull(skb, off + optlen))
 437                                 break;
 438
 439                         while (1) {
 440                                 struct ipv6_tlv_tnl_enc_lim *tel;
 441
 442                                 /* No more room for encapsulation limit */
 443                                 if (i + sizeof(*tel) > optlen)
 444                                         break;
 445
 446                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 447                                 /* return index of option if found and valid */
 448                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 449                                     tel->length == 1)
 450                                         return i + off - nhoff;
 451                                 /* else jump to next option */
 452                                 if (tel->type)
 453                                         i += tel->length + 2;
 454                                 else
 455                                         i++;
 456                         }
 457                 }
 458                 nexthdr = next;
 459                 off += optlen;
 460         }
 461         return 0;
 462 }
 463 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 464
 465 /**
 466  * ip6_tnl_err - tunnel error handler
 467  *
 468  * Description:
 469  *   ip6_tnl_err() should handle errors in the tunnel according
 470  *   to the specifications in RFC 2473.
 471  **/
 472
 473 static int
 474 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 475             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 476 {
 477         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 478         struct ip6_tnl *t;
 479         int rel_msg = 0;
 480         u8 rel_type = ICMPV6_DEST_UNREACH;
 481         u8 rel_code = ICMPV6_ADDR_UNREACH;
 482         u8 tproto;
 483         __u32 rel_info = 0;
 484         __u16 len;
 485         int err = -ENOENT;
 486
 487         /* If the packet doesn't contain the original IPv6 header we are
 488            in trouble since we might need the source address for further
 489            processing of the error. */
 490
 491         rcu_read_lock();
 492         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 493         if (!t)
 494                 goto out;
 495
 496         tproto = ACCESS_ONCE(t->parms.proto);
 497         if (tproto != ipproto && tproto != 0)
 498                 goto out;
 499
 500         err = 0;
 501
 502         switch (*type) {
 503                 __u32 teli;
 504                 struct ipv6_tlv_tnl_enc_lim *tel;
 505                 __u32 mtu;
 506         case ICMPV6_DEST_UNREACH:
 507                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 508                                     t->parms.name);
 509                 rel_msg = 1;
 510                 break;
 511         case ICMPV6_TIME_EXCEED:
 512                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 513                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 514                                             t->parms.name);
 515                         rel_msg = 1;
 516                 }
 517                 break;
 518         case ICMPV6_PARAMPROB:
 519                 teli = 0;
 520                 if ((*code) == ICMPV6_HDR_FIELD)
 521                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 522
 523                 if (teli && teli == *info - 2) {
 524                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 525                         if (tel->encap_limit == 0) {
 526                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 527                                                     t->parms.name);
 528                                 rel_msg = 1;
 529                         }
 530                 } else {
 531                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 532                                             t->parms.name);
 533                 }
 534                 break;
 535         case ICMPV6_PKT_TOOBIG:
 536                 mtu = *info - offset;
 537                 if (mtu < IPV6_MIN_MTU)
 538                         mtu = IPV6_MIN_MTU;
 539                 t->dev->mtu = mtu;
 540
 541                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 542                 if (len > mtu) {
 543                         rel_type = ICMPV6_PKT_TOOBIG;
 544                         rel_code = 0;
 545                         rel_info = mtu;
 546                         rel_msg = 1;
 547                 }
 548                 break;
 549         }
 550
 551         *type = rel_type;
 552         *code = rel_code;
 553         *info = rel_info;
 554         *msg = rel_msg;
 555
 556 out:
 557         rcu_read_unlock();
 558         return err;
 559 }
 560
 561 static int
 562 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 563            u8 type, u8 code, int offset, __be32 info)
 564 {
 565         int rel_msg = 0;
 566         u8 rel_type = type;
 567         u8 rel_code = code;
 568         __u32 rel_info = ntohl(info);
 569         int err;
 570         struct sk_buff *skb2;
 571         const struct iphdr *eiph;
 572         struct rtable *rt;
 573         struct flowi4 fl4;
 574
 575         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 576                           &rel_msg, &rel_info, offset);
 577         if (err < 0)
 578                 return err;
 579
 580         if (rel_msg == 0)
 581                 return 0;
 582
 583         switch (rel_type) {
 584         case ICMPV6_DEST_UNREACH:
 585                 if (rel_code != ICMPV6_ADDR_UNREACH)
 586                         return 0;
 587                 rel_type = ICMP_DEST_UNREACH;
 588                 rel_code = ICMP_HOST_UNREACH;
 589                 break;
 590         case ICMPV6_PKT_TOOBIG:
 591                 if (rel_code != 0)
 592                         return 0;
 593                 rel_type = ICMP_DEST_UNREACH;
 594                 rel_code = ICMP_FRAG_NEEDED;
 595                 break;
 596         case NDISC_REDIRECT:
 597                 rel_type = ICMP_REDIRECT;
 598                 rel_code = ICMP_REDIR_HOST;
 599         default:
 600                 return 0;
 601         }
 602
 603         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 604                 return 0;
 605
 606         skb2 = skb_clone(skb, GFP_ATOMIC);
 607         if (!skb2)
 608                 return 0;
 609
 610         skb_dst_drop(skb2);
 611
 612         skb_pull(skb2, offset);
 613         skb_reset_network_header(skb2);
 614         eiph = ip_hdr(skb2);
 615
 616         /* Try to guess incoming interface */
 617         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 618                                    eiph->saddr, 0,
 619                                    0, 0,
 620                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 621         if (IS_ERR(rt))
 622                 goto out;
 623
 624         skb2->dev = rt->dst.dev;
 625
 626         /* route "incoming" packet */
 627         if (rt->rt_flags & RTCF_LOCAL) {
 628                 ip_rt_put(rt);
 629                 rt = NULL;
 630                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 631                                            eiph->daddr, eiph->saddr,
 632                                            0, 0,
 633                                            IPPROTO_IPIP,
 634                                            RT_TOS(eiph->tos), 0);
 635                 if (IS_ERR(rt) ||
 636                     rt->dst.dev->type != ARPHRD_TUNNEL6) {
 637                         if (!IS_ERR(rt))
 638                                 ip_rt_put(rt);
 639                         goto out;
 640                 }
 641                 skb_dst_set(skb2, &rt->dst);
 642         } else {
 643                 ip_rt_put(rt);
 644                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 645                                    skb2->dev) ||
 646                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 647                         goto out;
 648         }
 649
 650         /* change mtu on this route */
 651         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 652                 if (rel_info > dst_mtu(skb_dst(skb2)))
 653                         goto out;
 654
 655                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 656         }
 657         if (rel_type == ICMP_REDIRECT)
 658                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 659
 660         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 661
 662 out:
 663         kfree_skb(skb2);
 664         return 0;
 665 }
 666
 667 static int
 668 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 669            u8 type, u8 code, int offset, __be32 info)
 670 {
 671         int rel_msg = 0;
 672         u8 rel_type = type;
 673         u8 rel_code = code;
 674         __u32 rel_info = ntohl(info);
 675         int err;
 676
 677         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 678                           &rel_msg, &rel_info, offset);
 679         if (err < 0)
 680                 return err;
 681
 682         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 683                 struct rt6_info *rt;
 684                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 685
 686                 if (!skb2)
 687                         return 0;
 688
 689                 skb_dst_drop(skb2);
 690                 skb_pull(skb2, offset);
 691                 skb_reset_network_header(skb2);
 692
 693                 /* Try to guess incoming interface */
 694                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 695                                 NULL, 0, 0);
 696
 697                 if (rt && rt->dst.dev)
 698                         skb2->dev = rt->dst.dev;
 699
 700                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 701
 702                 ip6_rt_put(rt);
 703
 704                 kfree_skb(skb2);
 705         }
 706
 707         return 0;
 708 }
 709
 710 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 711                                        const struct ipv6hdr *ipv6h,
 712                                        struct sk_buff *skb)
 713 {
 714         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 715
 716         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 717                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 718
 719         return IP6_ECN_decapsulate(ipv6h, skb);
 720 }
 721
 722 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 723                                        const struct ipv6hdr *ipv6h,
 724                                        struct sk_buff *skb)
 725 {
 726         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 727                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 728
 729         return IP6_ECN_decapsulate(ipv6h, skb);
 730 }
 731
 732 __u32 ip6_tnl_get_cap(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 ltype = ipv6_addr_type(laddr);
 738         int rtype = ipv6_addr_type(raddr);
 739         __u32 flags = 0;
 740
 741         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 742                 flags = IP6_TNL_F_CAP_PER_PACKET;
 743         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 744                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 745                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 746                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 747                 if (ltype&IPV6_ADDR_UNICAST)
 748                         flags |= IP6_TNL_F_CAP_XMIT;
 749                 if (rtype&IPV6_ADDR_UNICAST)
 750                         flags |= IP6_TNL_F_CAP_RCV;
 751         }
 752         return flags;
 753 }
 754 EXPORT_SYMBOL(ip6_tnl_get_cap);
 755
 756 /* called with rcu_read_lock() */
 757 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 758                                   const struct in6_addr *laddr,
 759                                   const struct in6_addr *raddr)
 760 {
 761         struct __ip6_tnl_parm *p = &t->parms;
 762         int ret = 0;
 763         struct net *net = t->net;
 764
 765         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 766             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 767              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 768                 struct net_device *ldev = NULL;
 769
 770                 if (p->link)
 771                         ldev = dev_get_by_index_rcu(net, p->link);
 772
 773                 if ((ipv6_addr_is_multicast(laddr) ||
 774                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 775                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 776                         ret = 1;
 777         }
 778         return ret;
 779 }
 780 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 781
 782 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 783                          const struct tnl_ptk_info *tpi,
 784                          struct metadata_dst *tun_dst,
 785                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 786                                                 const struct ipv6hdr *ipv6h,
 787                                                 struct sk_buff *skb),
 788                          bool log_ecn_err)
 789 {
 790         struct pcpu_sw_netstats *tstats;
 791         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 792         int err;
 793
 794         if ((!(tpi->flags & TUNNEL_CSUM) &&
 795              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 796             ((tpi->flags & TUNNEL_CSUM) &&
 797              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 798                 tunnel->dev->stats.rx_crc_errors++;
 799                 tunnel->dev->stats.rx_errors++;
 800                 goto drop;
 801         }
 802
 803         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 804                 if (!(tpi->flags & TUNNEL_SEQ) ||
 805                     (tunnel->i_seqno &&
 806                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 807                         tunnel->dev->stats.rx_fifo_errors++;
 808                         tunnel->dev->stats.rx_errors++;
 809                         goto drop;
 810                 }
 811                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 812         }
 813
 814         skb->protocol = tpi->proto;
 815
 816         /* Warning: All skb pointers will be invalidated! */
 817         if (tunnel->dev->type == ARPHRD_ETHER) {
 818                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 819                         tunnel->dev->stats.rx_length_errors++;
 820                         tunnel->dev->stats.rx_errors++;
 821                         goto drop;
 822                 }
 823
 824                 ipv6h = ipv6_hdr(skb);
 825                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 826                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 827         } else {
 828                 skb->dev = tunnel->dev;
 829         }
 830
 831         skb_reset_network_header(skb);
 832         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 833
 834         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 835
 836         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 837         if (unlikely(err)) {
 838                 if (log_ecn_err)
 839                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 840                                              &ipv6h->saddr,
 841                                              ipv6_get_dsfield(ipv6h));
 842                 if (err > 1) {
 843                         ++tunnel->dev->stats.rx_frame_errors;
 844                         ++tunnel->dev->stats.rx_errors;
 845                         goto drop;
 846                 }
 847         }
 848
 849         tstats = this_cpu_ptr(tunnel->dev->tstats);
 850         u64_stats_update_begin(&tstats->syncp);
 851         tstats->rx_packets++;
 852         tstats->rx_bytes += skb->len;
 853         u64_stats_update_end(&tstats->syncp);
 854
 855         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 856
 857         if (tun_dst)
 858                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 859
 860         gro_cells_receive(&tunnel->gro_cells, skb);
 861         return 0;
 862
 863 drop:
 864         kfree_skb(skb);
 865         return 0;
 866 }
 867
 868 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 869                 const struct tnl_ptk_info *tpi,
 870                 struct metadata_dst *tun_dst,
 871                 bool log_ecn_err)
 872 {
 873         int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 874                                     const struct ipv6hdr *ipv6h,
 875                                     struct sk_buff *skb);
 876
 877         dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
 878         if (tpi->proto == htons(ETH_P_IP))
 879                 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
 880
 881         return __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
 882                              log_ecn_err);
 883 }
 884 EXPORT_SYMBOL(ip6_tnl_rcv);
 885
 886 static const struct tnl_ptk_info tpi_v6 = {
 887         /* no tunnel info required for ipxip6. */
 888         .proto = htons(ETH_P_IPV6),
 889 };
 890
 891 static const struct tnl_ptk_info tpi_v4 = {
 892         /* no tunnel info required for ipxip6. */
 893         .proto = htons(ETH_P_IP),
 894 };
 895
 896 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 897                       const struct tnl_ptk_info *tpi,
 898                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 899                                                   const struct ipv6hdr *ipv6h,
 900                                                   struct sk_buff *skb))
 901 {
 902         struct ip6_tnl *t;
 903         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 904         struct metadata_dst *tun_dst = NULL;
 905         int ret = -1;
 906
 907         rcu_read_lock();
 908         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 909
 910         if (t) {
 911                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 912
 913                 if (tproto != ipproto && tproto != 0)
 914                         goto drop;
 915                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 916                         goto drop;
 917                 ipv6h = ipv6_hdr(skb);
 918                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 919                         goto drop;
 920                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 921                         goto drop;
 922                 if (t->parms.collect_md) {
 923                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 924                         if (!tun_dst)
 925                                 goto drop;
 926                 }
 927                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 928                                     log_ecn_error);
 929         }
 930
 931         rcu_read_unlock();
 932
 933         return ret;
 934
 935 drop:
 936         rcu_read_unlock();
 937         kfree_skb(skb);
 938         return 0;
 939 }
 940
 941 static int ip4ip6_rcv(struct sk_buff *skb)
 942 {
 943         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 944                           ip4ip6_dscp_ecn_decapsulate);
 945 }
 946
 947 static int ip6ip6_rcv(struct sk_buff *skb)
 948 {
 949         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 950                           ip6ip6_dscp_ecn_decapsulate);
 951 }
 952
 953 struct ipv6_tel_txoption {
 954         struct ipv6_txoptions ops;
 955         __u8 dst_opt[8];
 956 };
 957
 958 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 959 {
 960         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 961
 962         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 963         opt->dst_opt[3] = 1;
 964         opt->dst_opt[4] = encap_limit;
 965         opt->dst_opt[5] = IPV6_TLV_PADN;
 966         opt->dst_opt[6] = 1;
 967
 968         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 969         opt->ops.opt_nflen = 8;
 970 }
 971
 972 /**
 973  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 974  *   @t: the outgoing tunnel device
 975  *   @hdr: IPv6 header from the incoming packet
 976  *
 977  * Description:
 978  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 979  *   doesn't match source of incoming packet.
 980  *
 981  * Return:
 982  *   1 if conflict,
 983  *   0 else
 984  **/
 985
 986 static inline bool
 987 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 988 {
 989         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 990 }
 991
 992 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 993                      const struct in6_addr *laddr,
 994                      const struct in6_addr *raddr)
 995 {
 996         struct __ip6_tnl_parm *p = &t->parms;
 997         int ret = 0;
 998         struct net *net = t->net;
 999
1000         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1001             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1002              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1003                 struct net_device *ldev = NULL;
1004
1005                 rcu_read_lock();
1006                 if (p->link)
1007                         ldev = dev_get_by_index_rcu(net, p->link);
1008
1009                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1010                         pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1011                                             p->name);
1012                 else if (!ipv6_addr_is_multicast(raddr) &&
1013                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1014                         pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1015                                             p->name);
1016                 else
1017                         ret = 1;
1018                 rcu_read_unlock();
1019         }
1020         return ret;
1021 }
1022 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1023
1024 /**
1025  * ip6_tnl_xmit - encapsulate packet and send
1026  *   @skb: the outgoing socket buffer
1027  *   @dev: the outgoing tunnel device
1028  *   @dsfield: dscp code for outer header
1029  *   @fl6: flow of tunneled packet
1030  *   @encap_limit: encapsulation limit
1031  *   @pmtu: Path MTU is stored if packet is too big
1032  *   @proto: next header value
1033  *
1034  * Description:
1035  *   Build new header and do some sanity checks on the packet before sending
1036  *   it.
1037  *
1038  * Return:
1039  *   0 on success
1040  *   -1 fail
1041  *   %-EMSGSIZE message too big. return mtu in this case.
1042  **/
1043
1044 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1045                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1046                  __u8 proto)
1047 {
1048         struct ip6_tnl *t = netdev_priv(dev);
1049         struct net *net = t->net;
1050         struct net_device_stats *stats = &t->dev->stats;
1051         struct ipv6hdr *ipv6h;
1052         struct ipv6_tel_txoption opt;
1053         struct dst_entry *dst = NULL, *ndst = NULL;
1054         struct net_device *tdev;
1055         int mtu;
1056         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1057         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1058         unsigned int max_headroom = psh_hlen;
1059         bool use_cache = false;
1060         u8 hop_limit;
1061         int err = -1;
1062
1063         if (t->parms.collect_md) {
1064                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1065                 goto route_lookup;
1066         } else {
1067                 hop_limit = t->parms.hop_limit;
1068         }
1069
1070         /* NBMA tunnel */
1071         if (ipv6_addr_any(&t->parms.raddr)) {
1072                 if (skb->protocol == htons(ETH_P_IPV6)) {
1073                         struct in6_addr *addr6;
1074                         struct neighbour *neigh;
1075                         int addr_type;
1076
1077                         if (!skb_dst(skb))
1078                                 goto tx_err_link_failure;
1079
1080                         neigh = dst_neigh_lookup(skb_dst(skb),
1081                                                  &ipv6_hdr(skb)->daddr);
1082                         if (!neigh)
1083                                 goto tx_err_link_failure;
1084
1085                         addr6 = (struct in6_addr *)&neigh->primary_key;
1086                         addr_type = ipv6_addr_type(addr6);
1087
1088                         if (addr_type == IPV6_ADDR_ANY)
1089                                 addr6 = &ipv6_hdr(skb)->daddr;
1090
1091                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1092                         neigh_release(neigh);
1093                 }
1094         } else if (t->parms.proto != 0 && !(t->parms.flags &
1095                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1096                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1097                 /* enable the cache only if neither the outer protocol nor the
1098                  * routing decision depends on the current inner header value
1099                  */
1100                 use_cache = true;
1101         }
1102
1103         if (use_cache)
1104                 dst = dst_cache_get(&t->dst_cache);
1105
1106         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1107                 goto tx_err_link_failure;
1108
1109         if (!dst) {
1110 route_lookup:
1111                 /* add dsfield to flowlabel for route lookup */
1112                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1113
1114                 dst = ip6_route_output(net, NULL, fl6);
1115
1116                 if (dst->error)
1117                         goto tx_err_link_failure;
1118                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1119                 if (IS_ERR(dst)) {
1120                         err = PTR_ERR(dst);
1121                         dst = NULL;
1122                         goto tx_err_link_failure;
1123                 }
1124                 if (t->parms.collect_md &&
1125                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1126                                        &fl6->daddr, 0, &fl6->saddr))
1127                         goto tx_err_link_failure;
1128                 ndst = dst;
1129         }
1130
1131         tdev = dst->dev;
1132
1133         if (tdev == dev) {
1134                 stats->collisions++;
1135                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1136                                      t->parms.name);
1137                 goto tx_err_dst_release;
1138         }
1139         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1140         if (encap_limit >= 0) {
1141                 max_headroom += 8;
1142                 mtu -= 8;
1143         }
1144         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1145                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1146
1147         if (skb_dst(skb) && !t->parms.collect_md)
1148                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1149         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1150                 *pmtu = mtu;
1151                 err = -EMSGSIZE;
1152                 goto tx_err_dst_release;
1153         }
1154
1155         if (t->err_count > 0) {
1156                 if (time_before(jiffies,
1157                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1158                         t->err_count--;
1159
1160                         dst_link_failure(skb);
1161                 } else {
1162                         t->err_count = 0;
1163                 }
1164         }
1165
1166         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1167
1168         /*
1169          * Okay, now see if we can stuff it in the buffer as-is.
1170          */
1171         max_headroom += LL_RESERVED_SPACE(tdev);
1172
1173         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1174             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1175                 struct sk_buff *new_skb;
1176
1177                 new_skb = skb_realloc_headroom(skb, max_headroom);
1178                 if (!new_skb)
1179                         goto tx_err_dst_release;
1180
1181                 if (skb->sk)
1182                         skb_set_owner_w(new_skb, skb->sk);
1183                 consume_skb(skb);
1184                 skb = new_skb;
1185         }
1186
1187         if (t->parms.collect_md) {
1188                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1189                         goto tx_err_dst_release;
1190         } else {
1191                 if (use_cache && ndst)
1192                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1193         }
1194         skb_dst_set(skb, dst);
1195
1196         /* Calculate max headroom for all the headers and adjust
1197          * needed_headroom if necessary.
1198          */
1199         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1200                         + dst->header_len + t->hlen;
1201         if (max_headroom > dev->needed_headroom)
1202                 dev->needed_headroom = max_headroom;
1203
1204         err = ip6_tnl_encap(skb, t, &proto, fl6);
1205         if (err)
1206                 return err;
1207
1208         if (encap_limit >= 0) {
1209                 init_tel_txopt(&opt, encap_limit);
1210                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1211         }
1212
1213         skb_push(skb, sizeof(struct ipv6hdr));
1214         skb_reset_network_header(skb);
1215         ipv6h = ipv6_hdr(skb);
1216         ip6_flow_hdr(ipv6h, dsfield,
1217                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1218         ipv6h->hop_limit = hop_limit;
1219         ipv6h->nexthdr = proto;
1220         ipv6h->saddr = fl6->saddr;
1221         ipv6h->daddr = fl6->daddr;
1222         ip6tunnel_xmit(NULL, skb, dev);
1223         return 0;
1224 tx_err_link_failure:
1225         stats->tx_carrier_errors++;
1226         dst_link_failure(skb);
1227 tx_err_dst_release:
1228         dst_release(dst);
1229         return err;
1230 }
1231 EXPORT_SYMBOL(ip6_tnl_xmit);
1232
1233 static inline int
1234 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1235 {
1236         struct ip6_tnl *t = netdev_priv(dev);
1237         const struct iphdr  *iph;
1238         int encap_limit = -1;
1239         struct flowi6 fl6;
1240         __u8 dsfield;
1241         __u32 mtu;
1242         u8 tproto;
1243         int err;
1244
1245         /* ensure we can access the full inner ip header */
1246         if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1247                 return -1;
1248
1249         iph = ip_hdr(skb);
1250         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1251
1252         tproto = ACCESS_ONCE(t->parms.proto);
1253         if (tproto != IPPROTO_IPIP && tproto != 0)
1254                 return -1;
1255
1256         if (t->parms.collect_md) {
1257                 struct ip_tunnel_info *tun_info;
1258                 const struct ip_tunnel_key *key;
1259
1260                 tun_info = skb_tunnel_info(skb);
1261                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1262                              ip_tunnel_info_af(tun_info) != AF_INET6))
1263                         return -1;
1264                 key = &tun_info->key;
1265                 memset(&fl6, 0, sizeof(fl6));
1266                 fl6.flowi6_proto = IPPROTO_IPIP;
1267                 fl6.daddr = key->u.ipv6.dst;
1268                 fl6.flowlabel = key->label;
1269                 dsfield =  key->tos;
1270         } else {
1271                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1272                         encap_limit = t->parms.encap_limit;
1273
1274                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1275                 fl6.flowi6_proto = IPPROTO_IPIP;
1276
1277                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1278                         dsfield = ipv4_get_dsfield(iph);
1279                 else
1280                         dsfield = ip6_tclass(t->parms.flowinfo);
1281                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1282                         fl6.flowi6_mark = skb->mark;
1283         }
1284
1285         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1286
1287         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1288                 return -1;
1289
1290         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1291
1292         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1293                            IPPROTO_IPIP);
1294         if (err != 0) {
1295                 /* XXX: send ICMP error even if DF is not set. */
1296                 if (err == -EMSGSIZE)
1297                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1298                                   htonl(mtu));
1299                 return -1;
1300         }
1301
1302         return 0;
1303 }
1304
1305 static inline int
1306 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1307 {
1308         struct ip6_tnl *t = netdev_priv(dev);
1309         struct ipv6hdr *ipv6h;
1310         int encap_limit = -1;
1311         __u16 offset;
1312         struct flowi6 fl6;
1313         __u8 dsfield;
1314         __u32 mtu;
1315         u8 tproto;
1316         int err;
1317
1318         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1319                 return -1;
1320
1321         ipv6h = ipv6_hdr(skb);
1322         tproto = ACCESS_ONCE(t->parms.proto);
1323         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1324             ip6_tnl_addr_conflict(t, ipv6h))
1325                 return -1;
1326
1327         if (t->parms.collect_md) {
1328                 struct ip_tunnel_info *tun_info;
1329                 const struct ip_tunnel_key *key;
1330
1331                 tun_info = skb_tunnel_info(skb);
1332                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1333                              ip_tunnel_info_af(tun_info) != AF_INET6))
1334                         return -1;
1335                 key = &tun_info->key;
1336                 memset(&fl6, 0, sizeof(fl6));
1337                 fl6.flowi6_proto = IPPROTO_IPV6;
1338                 fl6.daddr = key->u.ipv6.dst;
1339                 fl6.flowlabel = key->label;
1340                 dsfield = key->tos;
1341         } else {
1342                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1343                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1344                 ipv6h = ipv6_hdr(skb);
1345                 if (offset > 0) {
1346                         struct ipv6_tlv_tnl_enc_lim *tel;
1347
1348                         tel = (void *)&skb_network_header(skb)[offset];
1349                         if (tel->encap_limit == 0) {
1350                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1351                                             ICMPV6_HDR_FIELD, offset + 2);
1352                                 return -1;
1353                         }
1354                         encap_limit = tel->encap_limit - 1;
1355                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1356                         encap_limit = t->parms.encap_limit;
1357                 }
1358
1359                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1360                 fl6.flowi6_proto = IPPROTO_IPV6;
1361
1362                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1363                         dsfield = ipv6_get_dsfield(ipv6h);
1364                 else
1365                         dsfield = ip6_tclass(t->parms.flowinfo);
1366                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1367                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1368                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1369                         fl6.flowi6_mark = skb->mark;
1370         }
1371
1372         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1373
1374         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1375                 return -1;
1376
1377         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1378
1379         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1380                            IPPROTO_IPV6);
1381         if (err != 0) {
1382                 if (err == -EMSGSIZE)
1383                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1384                 return -1;
1385         }
1386
1387         return 0;
1388 }
1389
1390 static netdev_tx_t
1391 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1392 {
1393         struct ip6_tnl *t = netdev_priv(dev);
1394         struct net_device_stats *stats = &t->dev->stats;
1395         int ret;
1396
1397         switch (skb->protocol) {
1398         case htons(ETH_P_IP):
1399                 ret = ip4ip6_tnl_xmit(skb, dev);
1400                 break;
1401         case htons(ETH_P_IPV6):
1402                 ret = ip6ip6_tnl_xmit(skb, dev);
1403                 break;
1404         default:
1405                 goto tx_err;
1406         }
1407
1408         if (ret < 0)
1409                 goto tx_err;
1410
1411         return NETDEV_TX_OK;
1412
1413 tx_err:
1414         stats->tx_errors++;
1415         stats->tx_dropped++;
1416         kfree_skb(skb);
1417         return NETDEV_TX_OK;
1418 }
1419
1420 static void ip6_tnl_link_config(struct ip6_tnl *t)
1421 {
1422         struct net_device *dev = t->dev;
1423         struct __ip6_tnl_parm *p = &t->parms;
1424         struct flowi6 *fl6 = &t->fl.u.ip6;
1425         int t_hlen;
1426
1427         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1428         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1429
1430         /* Set up flowi template */
1431         fl6->saddr = p->laddr;
1432         fl6->daddr = p->raddr;
1433         fl6->flowi6_oif = p->link;
1434         fl6->flowlabel = 0;
1435
1436         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1437                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1438         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1439                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1440
1441         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1442         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1443
1444         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1445                 dev->flags |= IFF_POINTOPOINT;
1446         else
1447                 dev->flags &= ~IFF_POINTOPOINT;
1448
1449         t->tun_hlen = 0;
1450         t->hlen = t->encap_hlen + t->tun_hlen;
1451         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1452
1453         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1454                 int strict = (ipv6_addr_type(&p->raddr) &
1455                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1456
1457                 struct rt6_info *rt = rt6_lookup(t->net,
1458                                                  &p->raddr, &p->laddr,
1459                                                  p->link, strict);
1460
1461                 if (!rt)
1462                         return;
1463
1464                 if (rt->dst.dev) {
1465                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1466                                 t_hlen;
1467
1468                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1469                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1470                                 dev->mtu -= 8;
1471
1472                         if (dev->mtu < IPV6_MIN_MTU)
1473                                 dev->mtu = IPV6_MIN_MTU;
1474                 }
1475                 ip6_rt_put(rt);
1476         }
1477 }
1478
1479 /**
1480  * ip6_tnl_change - update the tunnel parameters
1481  *   @t: tunnel to be changed
1482  *   @p: tunnel configuration parameters
1483  *
1484  * Description:
1485  *   ip6_tnl_change() updates the tunnel parameters
1486  **/
1487
1488 static int
1489 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1490 {
1491         t->parms.laddr = p->laddr;
1492         t->parms.raddr = p->raddr;
1493         t->parms.flags = p->flags;
1494         t->parms.hop_limit = p->hop_limit;
1495         t->parms.encap_limit = p->encap_limit;
1496         t->parms.flowinfo = p->flowinfo;
1497         t->parms.link = p->link;
1498         t->parms.proto = p->proto;
1499         dst_cache_reset(&t->dst_cache);
1500         ip6_tnl_link_config(t);
1501         return 0;
1502 }
1503
1504 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1505 {
1506         struct net *net = t->net;
1507         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1508         int err;
1509
1510         ip6_tnl_unlink(ip6n, t);
1511         synchronize_net();
1512         err = ip6_tnl_change(t, p);
1513         ip6_tnl_link(ip6n, t);
1514         netdev_state_change(t->dev);
1515         return err;
1516 }
1517
1518 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1519 {
1520         /* for default tnl0 device allow to change only the proto */
1521         t->parms.proto = p->proto;
1522         netdev_state_change(t->dev);
1523         return 0;
1524 }
1525
1526 static void
1527 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1528 {
1529         p->laddr = u->laddr;
1530         p->raddr = u->raddr;
1531         p->flags = u->flags;
1532         p->hop_limit = u->hop_limit;
1533         p->encap_limit = u->encap_limit;
1534         p->flowinfo = u->flowinfo;
1535         p->link = u->link;
1536         p->proto = u->proto;
1537         memcpy(p->name, u->name, sizeof(u->name));
1538 }
1539
1540 static void
1541 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1542 {
1543         u->laddr = p->laddr;
1544         u->raddr = p->raddr;
1545         u->flags = p->flags;
1546         u->hop_limit = p->hop_limit;
1547         u->encap_limit = p->encap_limit;
1548         u->flowinfo = p->flowinfo;
1549         u->link = p->link;
1550         u->proto = p->proto;
1551         memcpy(u->name, p->name, sizeof(u->name));
1552 }
1553
1554 /**
1555  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1556  *   @dev: virtual device associated with tunnel
1557  *   @ifr: parameters passed from userspace
1558  *   @cmd: command to be performed
1559  *
1560  * Description:
1561  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1562  *   from userspace.
1563  *
1564  *   The possible commands are the following:
1565  *     %SIOCGETTUNNEL: get tunnel parameters for device
1566  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1567  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1568  *     %SIOCDELTUNNEL: delete tunnel
1569  *
1570  *   The fallback device "ip6tnl0", created during module
1571  *   initialization, can be used for creating other tunnel devices.
1572  *
1573  * Return:
1574  *   0 on success,
1575  *   %-EFAULT if unable to copy data to or from userspace,
1576  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1577  *   %-EINVAL if passed tunnel parameters are invalid,
1578  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1579  *   %-ENODEV if attempting to change or delete a nonexisting device
1580  **/
1581
1582 static int
1583 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1584 {
1585         int err = 0;
1586         struct ip6_tnl_parm p;
1587         struct __ip6_tnl_parm p1;
1588         struct ip6_tnl *t = netdev_priv(dev);
1589         struct net *net = t->net;
1590         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1591
1592         memset(&p1, 0, sizeof(p1));
1593
1594         switch (cmd) {
1595         case SIOCGETTUNNEL:
1596                 if (dev == ip6n->fb_tnl_dev) {
1597                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1598                                 err = -EFAULT;
1599                                 break;
1600                         }
1601                         ip6_tnl_parm_from_user(&p1, &p);
1602                         t = ip6_tnl_locate(net, &p1, 0);
1603                         if (IS_ERR(t))
1604                                 t = netdev_priv(dev);
1605                 } else {
1606                         memset(&p, 0, sizeof(p));
1607                 }
1608                 ip6_tnl_parm_to_user(&p, &t->parms);
1609                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1610                         err = -EFAULT;
1611                 }
1612                 break;
1613         case SIOCADDTUNNEL:
1614         case SIOCCHGTUNNEL:
1615                 err = -EPERM;
1616                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1617                         break;
1618                 err = -EFAULT;
1619                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1620                         break;
1621                 err = -EINVAL;
1622                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1623                     p.proto != 0)
1624                         break;
1625                 ip6_tnl_parm_from_user(&p1, &p);
1626                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1627                 if (cmd == SIOCCHGTUNNEL) {
1628                         if (!IS_ERR(t)) {
1629                                 if (t->dev != dev) {
1630                                         err = -EEXIST;
1631                                         break;
1632                                 }
1633                         } else
1634                                 t = netdev_priv(dev);
1635                         if (dev == ip6n->fb_tnl_dev)
1636                                 err = ip6_tnl0_update(t, &p1);
1637                         else
1638                                 err = ip6_tnl_update(t, &p1);
1639                 }
1640                 if (!IS_ERR(t)) {
1641                         err = 0;
1642                         ip6_tnl_parm_to_user(&p, &t->parms);
1643                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1644                                 err = -EFAULT;
1645
1646                 } else {
1647                         err = PTR_ERR(t);
1648                 }
1649                 break;
1650         case SIOCDELTUNNEL:
1651                 err = -EPERM;
1652                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1653                         break;
1654
1655                 if (dev == ip6n->fb_tnl_dev) {
1656                         err = -EFAULT;
1657                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1658                                 break;
1659                         err = -ENOENT;
1660                         ip6_tnl_parm_from_user(&p1, &p);
1661                         t = ip6_tnl_locate(net, &p1, 0);
1662                         if (IS_ERR(t))
1663                                 break;
1664                         err = -EPERM;
1665                         if (t->dev == ip6n->fb_tnl_dev)
1666                                 break;
1667                         dev = t->dev;
1668                 }
1669                 err = 0;
1670                 unregister_netdevice(dev);
1671                 break;
1672         default:
1673                 err = -EINVAL;
1674         }
1675         return err;
1676 }
1677
1678 /**
1679  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1680  *   @dev: virtual device associated with tunnel
1681  *   @new_mtu: the new mtu
1682  *
1683  * Return:
1684  *   0 on success,
1685  *   %-EINVAL if mtu too small
1686  **/
1687
1688 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1689 {
1690         struct ip6_tnl *tnl = netdev_priv(dev);
1691
1692         if (tnl->parms.proto == IPPROTO_IPIP) {
1693                 if (new_mtu < 68)
1694                         return -EINVAL;
1695         } else {
1696                 if (new_mtu < IPV6_MIN_MTU)
1697                         return -EINVAL;
1698         }
1699         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1700                 return -EINVAL;
1701         dev->mtu = new_mtu;
1702         return 0;
1703 }
1704 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1705
1706 int ip6_tnl_get_iflink(const struct net_device *dev)
1707 {
1708         struct ip6_tnl *t = netdev_priv(dev);
1709
1710         return t->parms.link;
1711 }
1712 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1713
1714 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1715                           unsigned int num)
1716 {
1717         if (num >= MAX_IPTUN_ENCAP_OPS)
1718                 return -ERANGE;
1719
1720         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1721                         &ip6tun_encaps[num],
1722                         NULL, ops) ? 0 : -1;
1723 }
1724 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1725
1726 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1727                           unsigned int num)
1728 {
1729         int ret;
1730
1731         if (num >= MAX_IPTUN_ENCAP_OPS)
1732                 return -ERANGE;
1733
1734         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1735                        &ip6tun_encaps[num],
1736                        ops, NULL) == ops) ? 0 : -1;
1737
1738         synchronize_net();
1739
1740         return ret;
1741 }
1742 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1743
1744 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1745                         struct ip_tunnel_encap *ipencap)
1746 {
1747         int hlen;
1748
1749         memset(&t->encap, 0, sizeof(t->encap));
1750
1751         hlen = ip6_encap_hlen(ipencap);
1752         if (hlen < 0)
1753                 return hlen;
1754
1755         t->encap.type = ipencap->type;
1756         t->encap.sport = ipencap->sport;
1757         t->encap.dport = ipencap->dport;
1758         t->encap.flags = ipencap->flags;
1759
1760         t->encap_hlen = hlen;
1761         t->hlen = t->encap_hlen + t->tun_hlen;
1762
1763         return 0;
1764 }
1765 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1766
1767 static const struct net_device_ops ip6_tnl_netdev_ops = {
1768         .ndo_init       = ip6_tnl_dev_init,
1769         .ndo_uninit     = ip6_tnl_dev_uninit,
1770         .ndo_start_xmit = ip6_tnl_start_xmit,
1771         .ndo_do_ioctl   = ip6_tnl_ioctl,
1772         .ndo_change_mtu = ip6_tnl_change_mtu,
1773         .ndo_get_stats  = ip6_get_stats,
1774         .ndo_get_iflink = ip6_tnl_get_iflink,
1775 };
1776
1777 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1778                          NETIF_F_FRAGLIST |     \
1779                          NETIF_F_HIGHDMA |      \
1780                          NETIF_F_GSO_SOFTWARE | \
1781                          NETIF_F_HW_CSUM)
1782
1783 /**
1784  * ip6_tnl_dev_setup - setup virtual tunnel device
1785  *   @dev: virtual device associated with tunnel
1786  *
1787  * Description:
1788  *   Initialize function pointers and device parameters
1789  **/
1790
1791 static void ip6_tnl_dev_setup(struct net_device *dev)
1792 {
1793         dev->netdev_ops = &ip6_tnl_netdev_ops;
1794         dev->destructor = ip6_dev_free;
1795
1796         dev->type = ARPHRD_TUNNEL6;
1797         dev->flags |= IFF_NOARP;
1798         dev->addr_len = sizeof(struct in6_addr);
1799         dev->features |= NETIF_F_LLTX;
1800         netif_keep_dst(dev);
1801
1802         dev->features           |= IPXIPX_FEATURES;
1803         dev->hw_features        |= IPXIPX_FEATURES;
1804
1805         /* This perm addr will be used as interface identifier by IPv6 */
1806         dev->addr_assign_type = NET_ADDR_RANDOM;
1807         eth_random_addr(dev->perm_addr);
1808 }
1809
1810
1811 /**
1812  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1813  *   @dev: virtual device associated with tunnel
1814  **/
1815
1816 static inline int
1817 ip6_tnl_dev_init_gen(struct net_device *dev)
1818 {
1819         struct ip6_tnl *t = netdev_priv(dev);
1820         int ret;
1821         int t_hlen;
1822
1823         t->dev = dev;
1824         t->net = dev_net(dev);
1825         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1826         if (!dev->tstats)
1827                 return -ENOMEM;
1828
1829         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1830         if (ret)
1831                 goto free_stats;
1832
1833         ret = gro_cells_init(&t->gro_cells, dev);
1834         if (ret)
1835                 goto destroy_dst;
1836
1837         t->tun_hlen = 0;
1838         t->hlen = t->encap_hlen + t->tun_hlen;
1839         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1840
1841         dev->type = ARPHRD_TUNNEL6;
1842         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1843         dev->mtu = ETH_DATA_LEN - t_hlen;
1844         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1845                 dev->mtu -= 8;
1846
1847         dev_hold(dev);
1848         return 0;
1849
1850 destroy_dst:
1851         dst_cache_destroy(&t->dst_cache);
1852 free_stats:
1853         free_percpu(dev->tstats);
1854         dev->tstats = NULL;
1855
1856         return ret;
1857 }
1858
1859 /**
1860  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1861  *   @dev: virtual device associated with tunnel
1862  **/
1863
1864 static int ip6_tnl_dev_init(struct net_device *dev)
1865 {
1866         struct ip6_tnl *t = netdev_priv(dev);
1867         int err = ip6_tnl_dev_init_gen(dev);
1868
1869         if (err)
1870                 return err;
1871         ip6_tnl_link_config(t);
1872         if (t->parms.collect_md)
1873                 netif_keep_dst(dev);
1874         return 0;
1875 }
1876
1877 /**
1878  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1879  *   @dev: fallback device
1880  *
1881  * Return: 0
1882  **/
1883
1884 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1885 {
1886         struct ip6_tnl *t = netdev_priv(dev);
1887         struct net *net = dev_net(dev);
1888         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1889
1890         t->parms.proto = IPPROTO_IPV6;
1891
1892         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1893         return 0;
1894 }
1895
1896 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1897 {
1898         u8 proto;
1899
1900         if (!data || !data[IFLA_IPTUN_PROTO])
1901                 return 0;
1902
1903         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1904         if (proto != IPPROTO_IPV6 &&
1905             proto != IPPROTO_IPIP &&
1906             proto != 0)
1907                 return -EINVAL;
1908
1909         return 0;
1910 }
1911
1912 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1913                                   struct __ip6_tnl_parm *parms)
1914 {
1915         memset(parms, 0, sizeof(*parms));
1916
1917         if (!data)
1918                 return;
1919
1920         if (data[IFLA_IPTUN_LINK])
1921                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1922
1923         if (data[IFLA_IPTUN_LOCAL])
1924                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1925
1926         if (data[IFLA_IPTUN_REMOTE])
1927                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1928
1929         if (data[IFLA_IPTUN_TTL])
1930                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1931
1932         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1933                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1934
1935         if (data[IFLA_IPTUN_FLOWINFO])
1936                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1937
1938         if (data[IFLA_IPTUN_FLAGS])
1939                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1940
1941         if (data[IFLA_IPTUN_PROTO])
1942                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1943
1944         if (data[IFLA_IPTUN_COLLECT_METADATA])
1945                 parms->collect_md = true;
1946 }
1947
1948 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1949                                         struct ip_tunnel_encap *ipencap)
1950 {
1951         bool ret = false;
1952
1953         memset(ipencap, 0, sizeof(*ipencap));
1954
1955         if (!data)
1956                 return ret;
1957
1958         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1959                 ret = true;
1960                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1961         }
1962
1963         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1964                 ret = true;
1965                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1966         }
1967
1968         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1969                 ret = true;
1970                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1971         }
1972
1973         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1974                 ret = true;
1975                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1976         }
1977
1978         return ret;
1979 }
1980
1981 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1982                            struct nlattr *tb[], struct nlattr *data[])
1983 {
1984         struct net *net = dev_net(dev);
1985         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1986         struct ip_tunnel_encap ipencap;
1987         struct ip6_tnl *nt, *t;
1988         int err;
1989
1990         nt = netdev_priv(dev);
1991
1992         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1993                 err = ip6_tnl_encap_setup(nt, &ipencap);
1994                 if (err < 0)
1995                         return err;
1996         }
1997
1998         ip6_tnl_netlink_parms(data, &nt->parms);
1999
2000         if (nt->parms.collect_md) {
2001                 if (rtnl_dereference(ip6n->collect_md_tun))
2002                         return -EEXIST;
2003         } else {
2004                 t = ip6_tnl_locate(net, &nt->parms, 0);
2005                 if (!IS_ERR(t))
2006                         return -EEXIST;
2007         }
2008
2009         err = ip6_tnl_create2(dev);
2010         if (!err && tb[IFLA_MTU])
2011                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2012
2013         return err;
2014 }
2015
2016 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2017                               struct nlattr *data[])
2018 {
2019         struct ip6_tnl *t = netdev_priv(dev);
2020         struct __ip6_tnl_parm p;
2021         struct net *net = t->net;
2022         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2023         struct ip_tunnel_encap ipencap;
2024
2025         if (dev == ip6n->fb_tnl_dev)
2026                 return -EINVAL;
2027
2028         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2029                 int err = ip6_tnl_encap_setup(t, &ipencap);
2030
2031                 if (err < 0)
2032                         return err;
2033         }
2034         ip6_tnl_netlink_parms(data, &p);
2035         if (p.collect_md)
2036                 return -EINVAL;
2037
2038         t = ip6_tnl_locate(net, &p, 0);
2039         if (!IS_ERR(t)) {
2040                 if (t->dev != dev)
2041                         return -EEXIST;
2042         } else
2043                 t = netdev_priv(dev);
2044
2045         return ip6_tnl_update(t, &p);
2046 }
2047
2048 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2049 {
2050         struct net *net = dev_net(dev);
2051         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2052
2053         if (dev != ip6n->fb_tnl_dev)
2054                 unregister_netdevice_queue(dev, head);
2055 }
2056
2057 static size_t ip6_tnl_get_size(const struct net_device *dev)
2058 {
2059         return
2060                 /* IFLA_IPTUN_LINK */
2061                 nla_total_size(4) +
2062                 /* IFLA_IPTUN_LOCAL */
2063                 nla_total_size(sizeof(struct in6_addr)) +
2064                 /* IFLA_IPTUN_REMOTE */
2065                 nla_total_size(sizeof(struct in6_addr)) +
2066                 /* IFLA_IPTUN_TTL */
2067                 nla_total_size(1) +
2068                 /* IFLA_IPTUN_ENCAP_LIMIT */
2069                 nla_total_size(1) +
2070                 /* IFLA_IPTUN_FLOWINFO */
2071                 nla_total_size(4) +
2072                 /* IFLA_IPTUN_FLAGS */
2073                 nla_total_size(4) +
2074                 /* IFLA_IPTUN_PROTO */
2075                 nla_total_size(1) +
2076                 /* IFLA_IPTUN_ENCAP_TYPE */
2077                 nla_total_size(2) +
2078                 /* IFLA_IPTUN_ENCAP_FLAGS */
2079                 nla_total_size(2) +
2080                 /* IFLA_IPTUN_ENCAP_SPORT */
2081                 nla_total_size(2) +
2082                 /* IFLA_IPTUN_ENCAP_DPORT */
2083                 nla_total_size(2) +
2084                 /* IFLA_IPTUN_COLLECT_METADATA */
2085                 nla_total_size(0) +
2086                 0;
2087 }
2088
2089 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2090 {
2091         struct ip6_tnl *tunnel = netdev_priv(dev);
2092         struct __ip6_tnl_parm *parm = &tunnel->parms;
2093
2094         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2095             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2096             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2097             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2098             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2099             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2100             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2101             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
2102                 goto nla_put_failure;
2103
2104         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2105             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2106             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2107             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2108                 goto nla_put_failure;
2109
2110         if (parm->collect_md)
2111                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2112                         goto nla_put_failure;
2113         return 0;
2114
2115 nla_put_failure:
2116         return -EMSGSIZE;
2117 }
2118
2119 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2120 {
2121         struct ip6_tnl *tunnel = netdev_priv(dev);
2122
2123         return tunnel->net;
2124 }
2125 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2126
2127 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2128         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2129         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2130         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2131         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2132         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2133         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2134         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2135         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2136         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2137         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2138         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2139         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2140         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2141 };
2142
2143 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2144         .kind           = "ip6tnl",
2145         .maxtype        = IFLA_IPTUN_MAX,
2146         .policy         = ip6_tnl_policy,
2147         .priv_size      = sizeof(struct ip6_tnl),
2148         .setup          = ip6_tnl_dev_setup,
2149         .validate       = ip6_tnl_validate,
2150         .newlink        = ip6_tnl_newlink,
2151         .changelink     = ip6_tnl_changelink,
2152         .dellink        = ip6_tnl_dellink,
2153         .get_size       = ip6_tnl_get_size,
2154         .fill_info      = ip6_tnl_fill_info,
2155         .get_link_net   = ip6_tnl_get_link_net,
2156 };
2157
2158 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2159         .handler        = ip4ip6_rcv,
2160         .err_handler    = ip4ip6_err,
2161         .priority       =       1,
2162 };
2163
2164 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2165         .handler        = ip6ip6_rcv,
2166         .err_handler    = ip6ip6_err,
2167         .priority       =       1,
2168 };
2169
2170 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2171 {
2172         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2173         struct net_device *dev, *aux;
2174         int h;
2175         struct ip6_tnl *t;
2176         LIST_HEAD(list);
2177
2178         for_each_netdev_safe(net, dev, aux)
2179                 if (dev->rtnl_link_ops == &ip6_link_ops)
2180                         unregister_netdevice_queue(dev, &list);
2181
2182         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2183                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2184                 while (t) {
2185                         /* If dev is in the same netns, it has already
2186                          * been added to the list by the previous loop.
2187                          */
2188                         if (!net_eq(dev_net(t->dev), net))
2189                                 unregister_netdevice_queue(t->dev, &list);
2190                         t = rtnl_dereference(t->next);
2191                 }
2192         }
2193
2194         unregister_netdevice_many(&list);
2195 }
2196
2197 static int __net_init ip6_tnl_init_net(struct net *net)
2198 {
2199         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2200         struct ip6_tnl *t = NULL;
2201         int err;
2202
2203         ip6n->tnls[0] = ip6n->tnls_wc;
2204         ip6n->tnls[1] = ip6n->tnls_r_l;
2205
2206         err = -ENOMEM;
2207         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2208                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2209
2210         if (!ip6n->fb_tnl_dev)
2211                 goto err_alloc_dev;
2212         dev_net_set(ip6n->fb_tnl_dev, net);
2213         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2214         /* FB netdevice is special: we have one, and only one per netns.
2215          * Allowing to move it to another netns is clearly unsafe.
2216          */
2217         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2218
2219         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2220         if (err < 0)
2221                 goto err_register;
2222
2223         err = register_netdev(ip6n->fb_tnl_dev);
2224         if (err < 0)
2225                 goto err_register;
2226
2227         t = netdev_priv(ip6n->fb_tnl_dev);
2228
2229         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2230         return 0;
2231
2232 err_register:
2233         ip6_dev_free(ip6n->fb_tnl_dev);
2234 err_alloc_dev:
2235         return err;
2236 }
2237
2238 static void __net_exit ip6_tnl_exit_net(struct net *net)
2239 {
2240         rtnl_lock();
2241         ip6_tnl_destroy_tunnels(net);
2242         rtnl_unlock();
2243 }
2244
2245 static struct pernet_operations ip6_tnl_net_ops = {
2246         .init = ip6_tnl_init_net,
2247         .exit = ip6_tnl_exit_net,
2248         .id   = &ip6_tnl_net_id,
2249         .size = sizeof(struct ip6_tnl_net),
2250 };
2251
2252 /**
2253  * ip6_tunnel_init - register protocol and reserve needed resources
2254  *
2255  * Return: 0 on success
2256  **/
2257
2258 static int __init ip6_tunnel_init(void)
2259 {
2260         int  err;
2261
2262         if (!ipv6_mod_enabled())
2263                 return -EOPNOTSUPP;
2264
2265         err = register_pernet_device(&ip6_tnl_net_ops);
2266         if (err < 0)
2267                 goto out_pernet;
2268
2269         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2270         if (err < 0) {
2271                 pr_err("%s: can't register ip4ip6\n", __func__);
2272                 goto out_ip4ip6;
2273         }
2274
2275         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2276         if (err < 0) {
2277                 pr_err("%s: can't register ip6ip6\n", __func__);
2278                 goto out_ip6ip6;
2279         }
2280         err = rtnl_link_register(&ip6_link_ops);
2281         if (err < 0)
2282                 goto rtnl_link_failed;
2283
2284         return 0;
2285
2286 rtnl_link_failed:
2287         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2288 out_ip6ip6:
2289         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2290 out_ip4ip6:
2291         unregister_pernet_device(&ip6_tnl_net_ops);
2292 out_pernet:
2293         return err;
2294 }
2295
2296 /**
2297  * ip6_tunnel_cleanup - free resources and unregister protocol
2298  **/
2299
2300 static void __exit ip6_tunnel_cleanup(void)
2301 {
2302         rtnl_link_unregister(&ip6_link_ops);
2303         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2304                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2305
2306         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2307                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2308
2309         unregister_pernet_device(&ip6_tnl_net_ops);
2310 }
2311
2312 module_init(ip6_tunnel_init);
2313 module_exit(ip6_tunnel_cleanup);