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 <linux/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 unsigned 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         dev_hold(dev);
 277         ip6_tnl_link(ip6n, t);
 278         return 0;
 279
 280 out:
 281         return err;
 282 }
 283
 284 /**
 285  * ip6_tnl_create - create a new tunnel
 286  *   @p: tunnel parameters
 287  *   @pt: pointer to new tunnel
 288  *
 289  * Description:
 290  *   Create tunnel matching given parameters.
 291  *
 292  * Return:
 293  *   created tunnel or error pointer
 294  **/
 295
 296 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 297 {
 298         struct net_device *dev;
 299         struct ip6_tnl *t;
 300         char name[IFNAMSIZ];
 301         int err = -ENOMEM;
 302
 303         if (p->name[0])
 304                 strlcpy(name, p->name, IFNAMSIZ);
 305         else
 306                 sprintf(name, "ip6tnl%%d");
 307
 308         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 309                            ip6_tnl_dev_setup);
 310         if (!dev)
 311                 goto failed;
 312
 313         dev_net_set(dev, net);
 314
 315         t = netdev_priv(dev);
 316         t->parms = *p;
 317         t->net = dev_net(dev);
 318         err = ip6_tnl_create2(dev);
 319         if (err < 0)
 320                 goto failed_free;
 321
 322         return t;
 323
 324 failed_free:
 325         ip6_dev_free(dev);
 326 failed:
 327         return ERR_PTR(err);
 328 }
 329
 330 /**
 331  * ip6_tnl_locate - find or create tunnel matching given parameters
 332  *   @p: tunnel parameters
 333  *   @create: != 0 if allowed to create new tunnel if no match found
 334  *
 335  * Description:
 336  *   ip6_tnl_locate() first tries to locate an existing tunnel
 337  *   based on @parms. If this is unsuccessful, but @create is set a new
 338  *   tunnel device is created and registered for use.
 339  *
 340  * Return:
 341  *   matching tunnel or error pointer
 342  **/
 343
 344 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 345                 struct __ip6_tnl_parm *p, int create)
 346 {
 347         const struct in6_addr *remote = &p->raddr;
 348         const struct in6_addr *local = &p->laddr;
 349         struct ip6_tnl __rcu **tp;
 350         struct ip6_tnl *t;
 351         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 352
 353         for (tp = ip6_tnl_bucket(ip6n, p);
 354              (t = rtnl_dereference(*tp)) != NULL;
 355              tp = &t->next) {
 356                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 357                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 358                         if (create)
 359                                 return ERR_PTR(-EEXIST);
 360
 361                         return t;
 362                 }
 363         }
 364         if (!create)
 365                 return ERR_PTR(-ENODEV);
 366         return ip6_tnl_create(net, p);
 367 }
 368
 369 /**
 370  * ip6_tnl_dev_uninit - tunnel device uninitializer
 371  *   @dev: the device to be destroyed
 372  *
 373  * Description:
 374  *   ip6_tnl_dev_uninit() removes tunnel from its list
 375  **/
 376
 377 static void
 378 ip6_tnl_dev_uninit(struct net_device *dev)
 379 {
 380         struct ip6_tnl *t = netdev_priv(dev);
 381         struct net *net = t->net;
 382         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 383
 384         if (dev == ip6n->fb_tnl_dev)
 385                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 386         else
 387                 ip6_tnl_unlink(ip6n, t);
 388         dst_cache_reset(&t->dst_cache);
 389         dev_put(dev);
 390 }
 391
 392 /**
 393  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 394  *   @skb: received socket buffer
 395  *
 396  * Return:
 397  *   0 if none was found,
 398  *   else index to encapsulation limit
 399  **/
 400
 401 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 402 {
 403         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 404         unsigned int nhoff = raw - skb->data;
 405         unsigned int off = nhoff + sizeof(*ipv6h);
 406         u8 next, nexthdr = ipv6h->nexthdr;
 407
 408         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 409                 struct ipv6_opt_hdr *hdr;
 410                 u16 optlen;
 411
 412                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 413                         break;
 414
 415                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 416                 if (nexthdr == NEXTHDR_FRAGMENT) {
 417                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 418                         if (frag_hdr->frag_off)
 419                                 break;
 420                         optlen = 8;
 421                 } else if (nexthdr == NEXTHDR_AUTH) {
 422                         optlen = (hdr->hdrlen + 2) << 2;
 423                 } else {
 424                         optlen = ipv6_optlen(hdr);
 425                 }
 426                 /* cache hdr->nexthdr, since pskb_may_pull() might
 427                  * invalidate hdr
 428                  */
 429                 next = hdr->nexthdr;
 430                 if (nexthdr == NEXTHDR_DEST) {
 431                         u16 i = 2;
 432
 433                         /* Remember : hdr is no longer valid at this point. */
 434                         if (!pskb_may_pull(skb, off + optlen))
 435                                 break;
 436
 437                         while (1) {
 438                                 struct ipv6_tlv_tnl_enc_lim *tel;
 439
 440                                 /* No more room for encapsulation limit */
 441                                 if (i + sizeof(*tel) > optlen)
 442                                         break;
 443
 444                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 445                                 /* return index of option if found and valid */
 446                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 447                                     tel->length == 1)
 448                                         return i + off - nhoff;
 449                                 /* else jump to next option */
 450                                 if (tel->type)
 451                                         i += tel->length + 2;
 452                                 else
 453                                         i++;
 454                         }
 455                 }
 456                 nexthdr = next;
 457                 off += optlen;
 458         }
 459         return 0;
 460 }
 461 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 462
 463 /**
 464  * ip6_tnl_err - tunnel error handler
 465  *
 466  * Description:
 467  *   ip6_tnl_err() should handle errors in the tunnel according
 468  *   to the specifications in RFC 2473.
 469  **/
 470
 471 static int
 472 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 473             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 474 {
 475         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 476         struct ip6_tnl *t;
 477         int rel_msg = 0;
 478         u8 rel_type = ICMPV6_DEST_UNREACH;
 479         u8 rel_code = ICMPV6_ADDR_UNREACH;
 480         u8 tproto;
 481         __u32 rel_info = 0;
 482         __u16 len;
 483         int err = -ENOENT;
 484
 485         /* If the packet doesn't contain the original IPv6 header we are
 486            in trouble since we might need the source address for further
 487            processing of the error. */
 488
 489         rcu_read_lock();
 490         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 491         if (!t)
 492                 goto out;
 493
 494         tproto = ACCESS_ONCE(t->parms.proto);
 495         if (tproto != ipproto && tproto != 0)
 496                 goto out;
 497
 498         err = 0;
 499
 500         switch (*type) {
 501                 __u32 teli;
 502                 struct ipv6_tlv_tnl_enc_lim *tel;
 503                 __u32 mtu;
 504         case ICMPV6_DEST_UNREACH:
 505                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 506                                     t->parms.name);
 507                 rel_msg = 1;
 508                 break;
 509         case ICMPV6_TIME_EXCEED:
 510                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 511                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 512                                             t->parms.name);
 513                         rel_msg = 1;
 514                 }
 515                 break;
 516         case ICMPV6_PARAMPROB:
 517                 teli = 0;
 518                 if ((*code) == ICMPV6_HDR_FIELD)
 519                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 520
 521                 if (teli && teli == *info - 2) {
 522                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 523                         if (tel->encap_limit == 0) {
 524                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 525                                                     t->parms.name);
 526                                 rel_msg = 1;
 527                         }
 528                 } else {
 529                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 530                                             t->parms.name);
 531                 }
 532                 break;
 533         case ICMPV6_PKT_TOOBIG:
 534                 mtu = *info - offset;
 535                 if (mtu < IPV6_MIN_MTU)
 536                         mtu = IPV6_MIN_MTU;
 537                 t->dev->mtu = mtu;
 538
 539                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 540                 if (len > mtu) {
 541                         rel_type = ICMPV6_PKT_TOOBIG;
 542                         rel_code = 0;
 543                         rel_info = mtu;
 544                         rel_msg = 1;
 545                 }
 546                 break;
 547         }
 548
 549         *type = rel_type;
 550         *code = rel_code;
 551         *info = rel_info;
 552         *msg = rel_msg;
 553
 554 out:
 555         rcu_read_unlock();
 556         return err;
 557 }
 558
 559 static int
 560 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 561            u8 type, u8 code, int offset, __be32 info)
 562 {
 563         int rel_msg = 0;
 564         u8 rel_type = type;
 565         u8 rel_code = code;
 566         __u32 rel_info = ntohl(info);
 567         int err;
 568         struct sk_buff *skb2;
 569         const struct iphdr *eiph;
 570         struct rtable *rt;
 571         struct flowi4 fl4;
 572
 573         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 574                           &rel_msg, &rel_info, offset);
 575         if (err < 0)
 576                 return err;
 577
 578         if (rel_msg == 0)
 579                 return 0;
 580
 581         switch (rel_type) {
 582         case ICMPV6_DEST_UNREACH:
 583                 if (rel_code != ICMPV6_ADDR_UNREACH)
 584                         return 0;
 585                 rel_type = ICMP_DEST_UNREACH;
 586                 rel_code = ICMP_HOST_UNREACH;
 587                 break;
 588         case ICMPV6_PKT_TOOBIG:
 589                 if (rel_code != 0)
 590                         return 0;
 591                 rel_type = ICMP_DEST_UNREACH;
 592                 rel_code = ICMP_FRAG_NEEDED;
 593                 break;
 594         case NDISC_REDIRECT:
 595                 rel_type = ICMP_REDIRECT;
 596                 rel_code = ICMP_REDIR_HOST;
 597         default:
 598                 return 0;
 599         }
 600
 601         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 602                 return 0;
 603
 604         skb2 = skb_clone(skb, GFP_ATOMIC);
 605         if (!skb2)
 606                 return 0;
 607
 608         skb_dst_drop(skb2);
 609
 610         skb_pull(skb2, offset);
 611         skb_reset_network_header(skb2);
 612         eiph = ip_hdr(skb2);
 613
 614         /* Try to guess incoming interface */
 615         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 616                                    eiph->saddr, 0,
 617                                    0, 0,
 618                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 619         if (IS_ERR(rt))
 620                 goto out;
 621
 622         skb2->dev = rt->dst.dev;
 623
 624         /* route "incoming" packet */
 625         if (rt->rt_flags & RTCF_LOCAL) {
 626                 ip_rt_put(rt);
 627                 rt = NULL;
 628                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 629                                            eiph->daddr, eiph->saddr,
 630                                            0, 0,
 631                                            IPPROTO_IPIP,
 632                                            RT_TOS(eiph->tos), 0);
 633                 if (IS_ERR(rt) ||
 634                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 635                         if (!IS_ERR(rt))
 636                                 ip_rt_put(rt);
 637                         goto out;
 638                 }
 639                 skb_dst_set(skb2, &rt->dst);
 640         } else {
 641                 ip_rt_put(rt);
 642                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 643                                    skb2->dev) ||
 644                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 645                         goto out;
 646         }
 647
 648         /* change mtu on this route */
 649         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 650                 if (rel_info > dst_mtu(skb_dst(skb2)))
 651                         goto out;
 652
 653                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 654         }
 655         if (rel_type == ICMP_REDIRECT)
 656                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 657
 658         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 659
 660 out:
 661         kfree_skb(skb2);
 662         return 0;
 663 }
 664
 665 static int
 666 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 667            u8 type, u8 code, int offset, __be32 info)
 668 {
 669         int rel_msg = 0;
 670         u8 rel_type = type;
 671         u8 rel_code = code;
 672         __u32 rel_info = ntohl(info);
 673         int err;
 674
 675         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 676                           &rel_msg, &rel_info, offset);
 677         if (err < 0)
 678                 return err;
 679
 680         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 681                 struct rt6_info *rt;
 682                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 683
 684                 if (!skb2)
 685                         return 0;
 686
 687                 skb_dst_drop(skb2);
 688                 skb_pull(skb2, offset);
 689                 skb_reset_network_header(skb2);
 690
 691                 /* Try to guess incoming interface */
 692                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 693                                 NULL, 0, 0);
 694
 695                 if (rt && rt->dst.dev)
 696                         skb2->dev = rt->dst.dev;
 697
 698                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 699
 700                 ip6_rt_put(rt);
 701
 702                 kfree_skb(skb2);
 703         }
 704
 705         return 0;
 706 }
 707
 708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 709                                        const struct ipv6hdr *ipv6h,
 710                                        struct sk_buff *skb)
 711 {
 712         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 713
 714         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 715                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 716
 717         return IP6_ECN_decapsulate(ipv6h, skb);
 718 }
 719
 720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 721                                        const struct ipv6hdr *ipv6h,
 722                                        struct sk_buff *skb)
 723 {
 724         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 725                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 726
 727         return IP6_ECN_decapsulate(ipv6h, skb);
 728 }
 729
 730 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 731                              const struct in6_addr *laddr,
 732                              const struct in6_addr *raddr)
 733 {
 734         struct __ip6_tnl_parm *p = &t->parms;
 735         int ltype = ipv6_addr_type(laddr);
 736         int rtype = ipv6_addr_type(raddr);
 737         __u32 flags = 0;
 738
 739         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 740                 flags = IP6_TNL_F_CAP_PER_PACKET;
 741         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 742                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 743                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 744                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 745                 if (ltype&IPV6_ADDR_UNICAST)
 746                         flags |= IP6_TNL_F_CAP_XMIT;
 747                 if (rtype&IPV6_ADDR_UNICAST)
 748                         flags |= IP6_TNL_F_CAP_RCV;
 749         }
 750         return flags;
 751 }
 752 EXPORT_SYMBOL(ip6_tnl_get_cap);
 753
 754 /* called with rcu_read_lock() */
 755 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 756                                   const struct in6_addr *laddr,
 757                                   const struct in6_addr *raddr)
 758 {
 759         struct __ip6_tnl_parm *p = &t->parms;
 760         int ret = 0;
 761         struct net *net = t->net;
 762
 763         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 764             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 765              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 766                 struct net_device *ldev = NULL;
 767
 768                 if (p->link)
 769                         ldev = dev_get_by_index_rcu(net, p->link);
 770
 771                 if ((ipv6_addr_is_multicast(laddr) ||
 772                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 773                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 774                         ret = 1;
 775         }
 776         return ret;
 777 }
 778 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 779
 780 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 781                          const struct tnl_ptk_info *tpi,
 782                          struct metadata_dst *tun_dst,
 783                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 784                                                 const struct ipv6hdr *ipv6h,
 785                                                 struct sk_buff *skb),
 786                          bool log_ecn_err)
 787 {
 788         struct pcpu_sw_netstats *tstats;
 789         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 790         int err;
 791
 792         if ((!(tpi->flags & TUNNEL_CSUM) &&
 793              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 794             ((tpi->flags & TUNNEL_CSUM) &&
 795              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 796                 tunnel->dev->stats.rx_crc_errors++;
 797                 tunnel->dev->stats.rx_errors++;
 798                 goto drop;
 799         }
 800
 801         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 802                 if (!(tpi->flags & TUNNEL_SEQ) ||
 803                     (tunnel->i_seqno &&
 804                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 805                         tunnel->dev->stats.rx_fifo_errors++;
 806                         tunnel->dev->stats.rx_errors++;
 807                         goto drop;
 808                 }
 809                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 810         }
 811
 812         skb->protocol = tpi->proto;
 813
 814         /* Warning: All skb pointers will be invalidated! */
 815         if (tunnel->dev->type == ARPHRD_ETHER) {
 816                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 817                         tunnel->dev->stats.rx_length_errors++;
 818                         tunnel->dev->stats.rx_errors++;
 819                         goto drop;
 820                 }
 821
 822                 ipv6h = ipv6_hdr(skb);
 823                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 824                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 825         } else {
 826                 skb->dev = tunnel->dev;
 827         }
 828
 829         skb_reset_network_header(skb);
 830         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 831
 832         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 833
 834         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 835         if (unlikely(err)) {
 836                 if (log_ecn_err)
 837                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 838                                              &ipv6h->saddr,
 839                                              ipv6_get_dsfield(ipv6h));
 840                 if (err > 1) {
 841                         ++tunnel->dev->stats.rx_frame_errors;
 842                         ++tunnel->dev->stats.rx_errors;
 843                         goto drop;
 844                 }
 845         }
 846
 847         tstats = this_cpu_ptr(tunnel->dev->tstats);
 848         u64_stats_update_begin(&tstats->syncp);
 849         tstats->rx_packets++;
 850         tstats->rx_bytes += skb->len;
 851         u64_stats_update_end(&tstats->syncp);
 852
 853         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 854
 855         if (tun_dst)
 856                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 857
 858         gro_cells_receive(&tunnel->gro_cells, skb);
 859         return 0;
 860
 861 drop:
 862         kfree_skb(skb);
 863         return 0;
 864 }
 865
 866 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 867                 const struct tnl_ptk_info *tpi,
 868                 struct metadata_dst *tun_dst,
 869                 bool log_ecn_err)
 870 {
 871         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
 872                              log_ecn_err);
 873 }
 874 EXPORT_SYMBOL(ip6_tnl_rcv);
 875
 876 static const struct tnl_ptk_info tpi_v6 = {
 877         /* no tunnel info required for ipxip6. */
 878         .proto = htons(ETH_P_IPV6),
 879 };
 880
 881 static const struct tnl_ptk_info tpi_v4 = {
 882         /* no tunnel info required for ipxip6. */
 883         .proto = htons(ETH_P_IP),
 884 };
 885
 886 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 887                       const struct tnl_ptk_info *tpi,
 888                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 889                                                   const struct ipv6hdr *ipv6h,
 890                                                   struct sk_buff *skb))
 891 {
 892         struct ip6_tnl *t;
 893         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 894         struct metadata_dst *tun_dst = NULL;
 895         int ret = -1;
 896
 897         rcu_read_lock();
 898         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 899
 900         if (t) {
 901                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 902
 903                 if (tproto != ipproto && tproto != 0)
 904                         goto drop;
 905                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 906                         goto drop;
 907                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 908                         goto drop;
 909                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 910                         goto drop;
 911                 if (t->parms.collect_md) {
 912                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 913                         if (!tun_dst)
 914                                 return 0;
 915                 }
 916                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 917                                     log_ecn_error);
 918         }
 919
 920         rcu_read_unlock();
 921
 922         return ret;
 923
 924 drop:
 925         rcu_read_unlock();
 926         kfree_skb(skb);
 927         return 0;
 928 }
 929
 930 static int ip4ip6_rcv(struct sk_buff *skb)
 931 {
 932         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 933                           ip4ip6_dscp_ecn_decapsulate);
 934 }
 935
 936 static int ip6ip6_rcv(struct sk_buff *skb)
 937 {
 938         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 939                           ip6ip6_dscp_ecn_decapsulate);
 940 }
 941
 942 struct ipv6_tel_txoption {
 943         struct ipv6_txoptions ops;
 944         __u8 dst_opt[8];
 945 };
 946
 947 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 948 {
 949         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 950
 951         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 952         opt->dst_opt[3] = 1;
 953         opt->dst_opt[4] = encap_limit;
 954         opt->dst_opt[5] = IPV6_TLV_PADN;
 955         opt->dst_opt[6] = 1;
 956
 957         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 958         opt->ops.opt_nflen = 8;
 959 }
 960
 961 /**
 962  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 963  *   @t: the outgoing tunnel device
 964  *   @hdr: IPv6 header from the incoming packet
 965  *
 966  * Description:
 967  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 968  *   doesn't match source of incoming packet.
 969  *
 970  * Return:
 971  *   1 if conflict,
 972  *   0 else
 973  **/
 974
 975 static inline bool
 976 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 977 {
 978         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 979 }
 980
 981 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 982                      const struct in6_addr *laddr,
 983                      const struct in6_addr *raddr)
 984 {
 985         struct __ip6_tnl_parm *p = &t->parms;
 986         int ret = 0;
 987         struct net *net = t->net;
 988
 989         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 990             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 991              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 992                 struct net_device *ldev = NULL;
 993
 994                 rcu_read_lock();
 995                 if (p->link)
 996                         ldev = dev_get_by_index_rcu(net, p->link);
 997
 998                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 999                         pr_warn("%s xmit: Local address not yet configured!\n",
1000                                 p->name);
1001                 else if (!ipv6_addr_is_multicast(raddr) &&
1002                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1003                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1004                                 p->name);
1005                 else
1006                         ret = 1;
1007                 rcu_read_unlock();
1008         }
1009         return ret;
1010 }
1011 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1012
1013 /**
1014  * ip6_tnl_xmit - encapsulate packet and send
1015  *   @skb: the outgoing socket buffer
1016  *   @dev: the outgoing tunnel device
1017  *   @dsfield: dscp code for outer header
1018  *   @fl6: flow of tunneled packet
1019  *   @encap_limit: encapsulation limit
1020  *   @pmtu: Path MTU is stored if packet is too big
1021  *   @proto: next header value
1022  *
1023  * Description:
1024  *   Build new header and do some sanity checks on the packet before sending
1025  *   it.
1026  *
1027  * Return:
1028  *   0 on success
1029  *   -1 fail
1030  *   %-EMSGSIZE message too big. return mtu in this case.
1031  **/
1032
1033 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1034                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1035                  __u8 proto)
1036 {
1037         struct ip6_tnl *t = netdev_priv(dev);
1038         struct net *net = t->net;
1039         struct net_device_stats *stats = &t->dev->stats;
1040         struct ipv6hdr *ipv6h;
1041         struct ipv6_tel_txoption opt;
1042         struct dst_entry *dst = NULL, *ndst = NULL;
1043         struct net_device *tdev;
1044         int mtu;
1045         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1046         unsigned int max_headroom = psh_hlen;
1047         bool use_cache = false;
1048         u8 hop_limit;
1049         int err = -1;
1050
1051         if (t->parms.collect_md) {
1052                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1053                 goto route_lookup;
1054         } else {
1055                 hop_limit = t->parms.hop_limit;
1056         }
1057
1058         /* NBMA tunnel */
1059         if (ipv6_addr_any(&t->parms.raddr)) {
1060                 if (skb->protocol == htons(ETH_P_IPV6)) {
1061                         struct in6_addr *addr6;
1062                         struct neighbour *neigh;
1063                         int addr_type;
1064
1065                         if (!skb_dst(skb))
1066                                 goto tx_err_link_failure;
1067
1068                         neigh = dst_neigh_lookup(skb_dst(skb),
1069                                                  &ipv6_hdr(skb)->daddr);
1070                         if (!neigh)
1071                                 goto tx_err_link_failure;
1072
1073                         addr6 = (struct in6_addr *)&neigh->primary_key;
1074                         addr_type = ipv6_addr_type(addr6);
1075
1076                         if (addr_type == IPV6_ADDR_ANY)
1077                                 addr6 = &ipv6_hdr(skb)->daddr;
1078
1079                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1080                         neigh_release(neigh);
1081                 }
1082         } else if (!(t->parms.flags &
1083                      (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1084                 /* enable the cache only only if the routing decision does
1085                  * not depend on the current inner header value
1086                  */
1087                 use_cache = true;
1088         }
1089
1090         if (use_cache)
1091                 dst = dst_cache_get(&t->dst_cache);
1092
1093         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1094                 goto tx_err_link_failure;
1095
1096         if (!dst) {
1097 route_lookup:
1098                 dst = ip6_route_output(net, NULL, fl6);
1099
1100                 if (dst->error)
1101                         goto tx_err_link_failure;
1102                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1103                 if (IS_ERR(dst)) {
1104                         err = PTR_ERR(dst);
1105                         dst = NULL;
1106                         goto tx_err_link_failure;
1107                 }
1108                 if (t->parms.collect_md &&
1109                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1110                                        &fl6->daddr, 0, &fl6->saddr))
1111                         goto tx_err_link_failure;
1112                 ndst = dst;
1113         }
1114
1115         tdev = dst->dev;
1116
1117         if (tdev == dev) {
1118                 stats->collisions++;
1119                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1120                                      t->parms.name);
1121                 goto tx_err_dst_release;
1122         }
1123         mtu = dst_mtu(dst) - psh_hlen - t->tun_hlen;
1124         if (encap_limit >= 0) {
1125                 max_headroom += 8;
1126                 mtu -= 8;
1127         }
1128         if (mtu < IPV6_MIN_MTU)
1129                 mtu = IPV6_MIN_MTU;
1130         if (skb_dst(skb) && !t->parms.collect_md)
1131                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1132         if (skb->len - t->tun_hlen > mtu && !skb_is_gso(skb)) {
1133                 *pmtu = mtu;
1134                 err = -EMSGSIZE;
1135                 goto tx_err_dst_release;
1136         }
1137
1138         if (t->err_count > 0) {
1139                 if (time_before(jiffies,
1140                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1141                         t->err_count--;
1142
1143                         dst_link_failure(skb);
1144                 } else {
1145                         t->err_count = 0;
1146                 }
1147         }
1148
1149         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1150
1151         /*
1152          * Okay, now see if we can stuff it in the buffer as-is.
1153          */
1154         max_headroom += LL_RESERVED_SPACE(tdev);
1155
1156         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1157             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1158                 struct sk_buff *new_skb;
1159
1160                 new_skb = skb_realloc_headroom(skb, max_headroom);
1161                 if (!new_skb)
1162                         goto tx_err_dst_release;
1163
1164                 if (skb->sk)
1165                         skb_set_owner_w(new_skb, skb->sk);
1166                 consume_skb(skb);
1167                 skb = new_skb;
1168         }
1169
1170         if (t->parms.collect_md) {
1171                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1172                         goto tx_err_dst_release;
1173         } else {
1174                 if (use_cache && ndst)
1175                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1176         }
1177         skb_dst_set(skb, dst);
1178
1179         if (encap_limit >= 0) {
1180                 init_tel_txopt(&opt, encap_limit);
1181                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1182         }
1183
1184         /* Calculate max headroom for all the headers and adjust
1185          * needed_headroom if necessary.
1186          */
1187         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1188                         + dst->header_len + t->hlen;
1189         if (max_headroom > dev->needed_headroom)
1190                 dev->needed_headroom = max_headroom;
1191
1192         err = ip6_tnl_encap(skb, t, &proto, fl6);
1193         if (err)
1194                 return err;
1195
1196         skb_push(skb, sizeof(struct ipv6hdr));
1197         skb_reset_network_header(skb);
1198         ipv6h = ipv6_hdr(skb);
1199         ip6_flow_hdr(ipv6h, dsfield,
1200                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1201         ipv6h->hop_limit = hop_limit;
1202         ipv6h->nexthdr = proto;
1203         ipv6h->saddr = fl6->saddr;
1204         ipv6h->daddr = fl6->daddr;
1205         ip6tunnel_xmit(NULL, skb, dev);
1206         return 0;
1207 tx_err_link_failure:
1208         stats->tx_carrier_errors++;
1209         dst_link_failure(skb);
1210 tx_err_dst_release:
1211         dst_release(dst);
1212         return err;
1213 }
1214 EXPORT_SYMBOL(ip6_tnl_xmit);
1215
1216 static inline int
1217 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1218 {
1219         struct ip6_tnl *t = netdev_priv(dev);
1220         const struct iphdr  *iph = ip_hdr(skb);
1221         int encap_limit = -1;
1222         struct flowi6 fl6;
1223         __u8 dsfield;
1224         __u32 mtu;
1225         u8 tproto;
1226         int err;
1227
1228         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1229
1230         tproto = ACCESS_ONCE(t->parms.proto);
1231         if (tproto != IPPROTO_IPIP && tproto != 0)
1232                 return -1;
1233
1234         if (t->parms.collect_md) {
1235                 struct ip_tunnel_info *tun_info;
1236                 const struct ip_tunnel_key *key;
1237
1238                 tun_info = skb_tunnel_info(skb);
1239                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1240                              ip_tunnel_info_af(tun_info) != AF_INET6))
1241                         return -1;
1242                 key = &tun_info->key;
1243                 memset(&fl6, 0, sizeof(fl6));
1244                 fl6.flowi6_proto = IPPROTO_IPIP;
1245                 fl6.daddr = key->u.ipv6.dst;
1246                 fl6.flowlabel = key->label;
1247                 dsfield = ip6_tclass(key->label);
1248         } else {
1249                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1250                         encap_limit = t->parms.encap_limit;
1251
1252                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1253                 fl6.flowi6_proto = IPPROTO_IPIP;
1254
1255                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1256                         dsfield = ipv4_get_dsfield(iph);
1257                 else
1258                         dsfield = ip6_tclass(t->parms.flowinfo);
1259                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1260                         fl6.flowi6_mark = skb->mark;
1261                 else
1262                         fl6.flowi6_mark = t->parms.fwmark;
1263         }
1264
1265         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1266
1267         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1268                 return -1;
1269
1270         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1271
1272         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1273
1274         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1275                            IPPROTO_IPIP);
1276         if (err != 0) {
1277                 /* XXX: send ICMP error even if DF is not set. */
1278                 if (err == -EMSGSIZE)
1279                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1280                                   htonl(mtu));
1281                 return -1;
1282         }
1283
1284         return 0;
1285 }
1286
1287 static inline int
1288 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1289 {
1290         struct ip6_tnl *t = netdev_priv(dev);
1291         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1292         int encap_limit = -1;
1293         __u16 offset;
1294         struct flowi6 fl6;
1295         __u8 dsfield;
1296         __u32 mtu;
1297         u8 tproto;
1298         int err;
1299
1300         tproto = ACCESS_ONCE(t->parms.proto);
1301         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1302             ip6_tnl_addr_conflict(t, ipv6h))
1303                 return -1;
1304
1305         if (t->parms.collect_md) {
1306                 struct ip_tunnel_info *tun_info;
1307                 const struct ip_tunnel_key *key;
1308
1309                 tun_info = skb_tunnel_info(skb);
1310                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1311                              ip_tunnel_info_af(tun_info) != AF_INET6))
1312                         return -1;
1313                 key = &tun_info->key;
1314                 memset(&fl6, 0, sizeof(fl6));
1315                 fl6.flowi6_proto = IPPROTO_IPV6;
1316                 fl6.daddr = key->u.ipv6.dst;
1317                 fl6.flowlabel = key->label;
1318                 dsfield = ip6_tclass(key->label);
1319         } else {
1320                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1321                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1322                 ipv6h = ipv6_hdr(skb);
1323                 if (offset > 0) {
1324                         struct ipv6_tlv_tnl_enc_lim *tel;
1325
1326                         tel = (void *)&skb_network_header(skb)[offset];
1327                         if (tel->encap_limit == 0) {
1328                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1329                                             ICMPV6_HDR_FIELD, offset + 2);
1330                                 return -1;
1331                         }
1332                         encap_limit = tel->encap_limit - 1;
1333                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1334                         encap_limit = t->parms.encap_limit;
1335                 }
1336
1337                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1338                 fl6.flowi6_proto = IPPROTO_IPV6;
1339
1340                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1341                         dsfield = ipv6_get_dsfield(ipv6h);
1342                 else
1343                         dsfield = ip6_tclass(t->parms.flowinfo);
1344                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1345                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1346                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1347                         fl6.flowi6_mark = skb->mark;
1348                 else
1349                         fl6.flowi6_mark = t->parms.fwmark;
1350         }
1351
1352         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1353
1354         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1355                 return -1;
1356
1357         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1358
1359         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1360
1361         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1362                            IPPROTO_IPV6);
1363         if (err != 0) {
1364                 if (err == -EMSGSIZE)
1365                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1366                 return -1;
1367         }
1368
1369         return 0;
1370 }
1371
1372 static netdev_tx_t
1373 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1374 {
1375         struct ip6_tnl *t = netdev_priv(dev);
1376         struct net_device_stats *stats = &t->dev->stats;
1377         int ret;
1378
1379         switch (skb->protocol) {
1380         case htons(ETH_P_IP):
1381                 ret = ip4ip6_tnl_xmit(skb, dev);
1382                 break;
1383         case htons(ETH_P_IPV6):
1384                 ret = ip6ip6_tnl_xmit(skb, dev);
1385                 break;
1386         default:
1387                 goto tx_err;
1388         }
1389
1390         if (ret < 0)
1391                 goto tx_err;
1392
1393         return NETDEV_TX_OK;
1394
1395 tx_err:
1396         stats->tx_errors++;
1397         stats->tx_dropped++;
1398         kfree_skb(skb);
1399         return NETDEV_TX_OK;
1400 }
1401
1402 static void ip6_tnl_link_config(struct ip6_tnl *t)
1403 {
1404         struct net_device *dev = t->dev;
1405         struct __ip6_tnl_parm *p = &t->parms;
1406         struct flowi6 *fl6 = &t->fl.u.ip6;
1407         int t_hlen;
1408
1409         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1410         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1411
1412         /* Set up flowi template */
1413         fl6->saddr = p->laddr;
1414         fl6->daddr = p->raddr;
1415         fl6->flowi6_oif = p->link;
1416         fl6->flowlabel = 0;
1417
1418         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1419                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1420         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1421                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1422
1423         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1424         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1425
1426         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1427                 dev->flags |= IFF_POINTOPOINT;
1428         else
1429                 dev->flags &= ~IFF_POINTOPOINT;
1430
1431         t->tun_hlen = 0;
1432         t->hlen = t->encap_hlen + t->tun_hlen;
1433         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1434
1435         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1436                 int strict = (ipv6_addr_type(&p->raddr) &
1437                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1438
1439                 struct rt6_info *rt = rt6_lookup(t->net,
1440                                                  &p->raddr, &p->laddr,
1441                                                  p->link, strict);
1442
1443                 if (!rt)
1444                         return;
1445
1446                 if (rt->dst.dev) {
1447                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1448                                 t_hlen;
1449
1450                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1451                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1452                                 dev->mtu -= 8;
1453
1454                         if (dev->mtu < IPV6_MIN_MTU)
1455                                 dev->mtu = IPV6_MIN_MTU;
1456                 }
1457                 ip6_rt_put(rt);
1458         }
1459 }
1460
1461 /**
1462  * ip6_tnl_change - update the tunnel parameters
1463  *   @t: tunnel to be changed
1464  *   @p: tunnel configuration parameters
1465  *
1466  * Description:
1467  *   ip6_tnl_change() updates the tunnel parameters
1468  **/
1469
1470 static int
1471 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1472 {
1473         t->parms.laddr = p->laddr;
1474         t->parms.raddr = p->raddr;
1475         t->parms.flags = p->flags;
1476         t->parms.hop_limit = p->hop_limit;
1477         t->parms.encap_limit = p->encap_limit;
1478         t->parms.flowinfo = p->flowinfo;
1479         t->parms.link = p->link;
1480         t->parms.proto = p->proto;
1481         t->parms.fwmark = p->fwmark;
1482         dst_cache_reset(&t->dst_cache);
1483         ip6_tnl_link_config(t);
1484         return 0;
1485 }
1486
1487 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1488 {
1489         struct net *net = t->net;
1490         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1491         int err;
1492
1493         ip6_tnl_unlink(ip6n, t);
1494         synchronize_net();
1495         err = ip6_tnl_change(t, p);
1496         ip6_tnl_link(ip6n, t);
1497         netdev_state_change(t->dev);
1498         return err;
1499 }
1500
1501 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1502 {
1503         /* for default tnl0 device allow to change only the proto */
1504         t->parms.proto = p->proto;
1505         netdev_state_change(t->dev);
1506         return 0;
1507 }
1508
1509 static void
1510 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1511 {
1512         p->laddr = u->laddr;
1513         p->raddr = u->raddr;
1514         p->flags = u->flags;
1515         p->hop_limit = u->hop_limit;
1516         p->encap_limit = u->encap_limit;
1517         p->flowinfo = u->flowinfo;
1518         p->link = u->link;
1519         p->proto = u->proto;
1520         memcpy(p->name, u->name, sizeof(u->name));
1521 }
1522
1523 static void
1524 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1525 {
1526         u->laddr = p->laddr;
1527         u->raddr = p->raddr;
1528         u->flags = p->flags;
1529         u->hop_limit = p->hop_limit;
1530         u->encap_limit = p->encap_limit;
1531         u->flowinfo = p->flowinfo;
1532         u->link = p->link;
1533         u->proto = p->proto;
1534         memcpy(u->name, p->name, sizeof(u->name));
1535 }
1536
1537 /**
1538  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1539  *   @dev: virtual device associated with tunnel
1540  *   @ifr: parameters passed from userspace
1541  *   @cmd: command to be performed
1542  *
1543  * Description:
1544  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1545  *   from userspace.
1546  *
1547  *   The possible commands are the following:
1548  *     %SIOCGETTUNNEL: get tunnel parameters for device
1549  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1550  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1551  *     %SIOCDELTUNNEL: delete tunnel
1552  *
1553  *   The fallback device "ip6tnl0", created during module
1554  *   initialization, can be used for creating other tunnel devices.
1555  *
1556  * Return:
1557  *   0 on success,
1558  *   %-EFAULT if unable to copy data to or from userspace,
1559  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1560  *   %-EINVAL if passed tunnel parameters are invalid,
1561  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1562  *   %-ENODEV if attempting to change or delete a nonexisting device
1563  **/
1564
1565 static int
1566 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1567 {
1568         int err = 0;
1569         struct ip6_tnl_parm p;
1570         struct __ip6_tnl_parm p1;
1571         struct ip6_tnl *t = netdev_priv(dev);
1572         struct net *net = t->net;
1573         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1574
1575         memset(&p1, 0, sizeof(p1));
1576
1577         switch (cmd) {
1578         case SIOCGETTUNNEL:
1579                 if (dev == ip6n->fb_tnl_dev) {
1580                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1581                                 err = -EFAULT;
1582                                 break;
1583                         }
1584                         ip6_tnl_parm_from_user(&p1, &p);
1585                         t = ip6_tnl_locate(net, &p1, 0);
1586                         if (IS_ERR(t))
1587                                 t = netdev_priv(dev);
1588                 } else {
1589                         memset(&p, 0, sizeof(p));
1590                 }
1591                 ip6_tnl_parm_to_user(&p, &t->parms);
1592                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1593                         err = -EFAULT;
1594                 }
1595                 break;
1596         case SIOCADDTUNNEL:
1597         case SIOCCHGTUNNEL:
1598                 err = -EPERM;
1599                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1600                         break;
1601                 err = -EFAULT;
1602                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1603                         break;
1604                 err = -EINVAL;
1605                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1606                     p.proto != 0)
1607                         break;
1608                 ip6_tnl_parm_from_user(&p1, &p);
1609                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1610                 if (cmd == SIOCCHGTUNNEL) {
1611                         if (!IS_ERR(t)) {
1612                                 if (t->dev != dev) {
1613                                         err = -EEXIST;
1614                                         break;
1615                                 }
1616                         } else
1617                                 t = netdev_priv(dev);
1618                         if (dev == ip6n->fb_tnl_dev)
1619                                 err = ip6_tnl0_update(t, &p1);
1620                         else
1621                                 err = ip6_tnl_update(t, &p1);
1622                 }
1623                 if (!IS_ERR(t)) {
1624                         err = 0;
1625                         ip6_tnl_parm_to_user(&p, &t->parms);
1626                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1627                                 err = -EFAULT;
1628
1629                 } else {
1630                         err = PTR_ERR(t);
1631                 }
1632                 break;
1633         case SIOCDELTUNNEL:
1634                 err = -EPERM;
1635                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1636                         break;
1637
1638                 if (dev == ip6n->fb_tnl_dev) {
1639                         err = -EFAULT;
1640                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1641                                 break;
1642                         err = -ENOENT;
1643                         ip6_tnl_parm_from_user(&p1, &p);
1644                         t = ip6_tnl_locate(net, &p1, 0);
1645                         if (IS_ERR(t))
1646                                 break;
1647                         err = -EPERM;
1648                         if (t->dev == ip6n->fb_tnl_dev)
1649                                 break;
1650                         dev = t->dev;
1651                 }
1652                 err = 0;
1653                 unregister_netdevice(dev);
1654                 break;
1655         default:
1656                 err = -EINVAL;
1657         }
1658         return err;
1659 }
1660
1661 /**
1662  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1663  *   @dev: virtual device associated with tunnel
1664  *   @new_mtu: the new mtu
1665  *
1666  * Return:
1667  *   0 on success,
1668  *   %-EINVAL if mtu too small
1669  **/
1670
1671 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1672 {
1673         struct ip6_tnl *tnl = netdev_priv(dev);
1674
1675         if (tnl->parms.proto == IPPROTO_IPIP) {
1676                 if (new_mtu < ETH_MIN_MTU)
1677                         return -EINVAL;
1678         } else {
1679                 if (new_mtu < IPV6_MIN_MTU)
1680                         return -EINVAL;
1681         }
1682         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1683                 return -EINVAL;
1684         dev->mtu = new_mtu;
1685         return 0;
1686 }
1687 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1688
1689 int ip6_tnl_get_iflink(const struct net_device *dev)
1690 {
1691         struct ip6_tnl *t = netdev_priv(dev);
1692
1693         return t->parms.link;
1694 }
1695 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1696
1697 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1698                           unsigned int num)
1699 {
1700         if (num >= MAX_IPTUN_ENCAP_OPS)
1701                 return -ERANGE;
1702
1703         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1704                         &ip6tun_encaps[num],
1705                         NULL, ops) ? 0 : -1;
1706 }
1707 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1708
1709 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1710                           unsigned int num)
1711 {
1712         int ret;
1713
1714         if (num >= MAX_IPTUN_ENCAP_OPS)
1715                 return -ERANGE;
1716
1717         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1718                        &ip6tun_encaps[num],
1719                        ops, NULL) == ops) ? 0 : -1;
1720
1721         synchronize_net();
1722
1723         return ret;
1724 }
1725 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1726
1727 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1728                         struct ip_tunnel_encap *ipencap)
1729 {
1730         int hlen;
1731
1732         memset(&t->encap, 0, sizeof(t->encap));
1733
1734         hlen = ip6_encap_hlen(ipencap);
1735         if (hlen < 0)
1736                 return hlen;
1737
1738         t->encap.type = ipencap->type;
1739         t->encap.sport = ipencap->sport;
1740         t->encap.dport = ipencap->dport;
1741         t->encap.flags = ipencap->flags;
1742
1743         t->encap_hlen = hlen;
1744         t->hlen = t->encap_hlen + t->tun_hlen;
1745
1746         return 0;
1747 }
1748 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1749
1750 static const struct net_device_ops ip6_tnl_netdev_ops = {
1751         .ndo_init       = ip6_tnl_dev_init,
1752         .ndo_uninit     = ip6_tnl_dev_uninit,
1753         .ndo_start_xmit = ip6_tnl_start_xmit,
1754         .ndo_do_ioctl   = ip6_tnl_ioctl,
1755         .ndo_change_mtu = ip6_tnl_change_mtu,
1756         .ndo_get_stats  = ip6_get_stats,
1757         .ndo_get_iflink = ip6_tnl_get_iflink,
1758 };
1759
1760 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1761                          NETIF_F_FRAGLIST |     \
1762                          NETIF_F_HIGHDMA |      \
1763                          NETIF_F_GSO_SOFTWARE | \
1764                          NETIF_F_HW_CSUM)
1765
1766 /**
1767  * ip6_tnl_dev_setup - setup virtual tunnel device
1768  *   @dev: virtual device associated with tunnel
1769  *
1770  * Description:
1771  *   Initialize function pointers and device parameters
1772  **/
1773
1774 static void ip6_tnl_dev_setup(struct net_device *dev)
1775 {
1776         dev->netdev_ops = &ip6_tnl_netdev_ops;
1777         dev->destructor = ip6_dev_free;
1778
1779         dev->type = ARPHRD_TUNNEL6;
1780         dev->flags |= IFF_NOARP;
1781         dev->addr_len = sizeof(struct in6_addr);
1782         dev->features |= NETIF_F_LLTX;
1783         netif_keep_dst(dev);
1784
1785         dev->features           |= IPXIPX_FEATURES;
1786         dev->hw_features        |= IPXIPX_FEATURES;
1787
1788         /* This perm addr will be used as interface identifier by IPv6 */
1789         dev->addr_assign_type = NET_ADDR_RANDOM;
1790         eth_random_addr(dev->perm_addr);
1791 }
1792
1793
1794 /**
1795  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1796  *   @dev: virtual device associated with tunnel
1797  **/
1798
1799 static inline int
1800 ip6_tnl_dev_init_gen(struct net_device *dev)
1801 {
1802         struct ip6_tnl *t = netdev_priv(dev);
1803         int ret;
1804         int t_hlen;
1805
1806         t->dev = dev;
1807         t->net = dev_net(dev);
1808         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1809         if (!dev->tstats)
1810                 return -ENOMEM;
1811
1812         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1813         if (ret)
1814                 goto free_stats;
1815
1816         ret = gro_cells_init(&t->gro_cells, dev);
1817         if (ret)
1818                 goto destroy_dst;
1819
1820         t->tun_hlen = 0;
1821         t->hlen = t->encap_hlen + t->tun_hlen;
1822         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1823
1824         dev->type = ARPHRD_TUNNEL6;
1825         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1826         dev->mtu = ETH_DATA_LEN - t_hlen;
1827         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1828                 dev->mtu -= 8;
1829         dev->min_mtu = ETH_MIN_MTU;
1830         dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1831
1832         return 0;
1833
1834 destroy_dst:
1835         dst_cache_destroy(&t->dst_cache);
1836 free_stats:
1837         free_percpu(dev->tstats);
1838         dev->tstats = NULL;
1839
1840         return ret;
1841 }
1842
1843 /**
1844  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1845  *   @dev: virtual device associated with tunnel
1846  **/
1847
1848 static int ip6_tnl_dev_init(struct net_device *dev)
1849 {
1850         struct ip6_tnl *t = netdev_priv(dev);
1851         int err = ip6_tnl_dev_init_gen(dev);
1852
1853         if (err)
1854                 return err;
1855         ip6_tnl_link_config(t);
1856         if (t->parms.collect_md) {
1857                 dev->features |= NETIF_F_NETNS_LOCAL;
1858                 netif_keep_dst(dev);
1859         }
1860         return 0;
1861 }
1862
1863 /**
1864  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1865  *   @dev: fallback device
1866  *
1867  * Return: 0
1868  **/
1869
1870 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1871 {
1872         struct ip6_tnl *t = netdev_priv(dev);
1873         struct net *net = dev_net(dev);
1874         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1875
1876         t->parms.proto = IPPROTO_IPV6;
1877         dev_hold(dev);
1878
1879         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1880         return 0;
1881 }
1882
1883 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1884 {
1885         u8 proto;
1886
1887         if (!data || !data[IFLA_IPTUN_PROTO])
1888                 return 0;
1889
1890         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1891         if (proto != IPPROTO_IPV6 &&
1892             proto != IPPROTO_IPIP &&
1893             proto != 0)
1894                 return -EINVAL;
1895
1896         return 0;
1897 }
1898
1899 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1900                                   struct __ip6_tnl_parm *parms)
1901 {
1902         memset(parms, 0, sizeof(*parms));
1903
1904         if (!data)
1905                 return;
1906
1907         if (data[IFLA_IPTUN_LINK])
1908                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1909
1910         if (data[IFLA_IPTUN_LOCAL])
1911                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1912
1913         if (data[IFLA_IPTUN_REMOTE])
1914                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1915
1916         if (data[IFLA_IPTUN_TTL])
1917                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1918
1919         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1920                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1921
1922         if (data[IFLA_IPTUN_FLOWINFO])
1923                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1924
1925         if (data[IFLA_IPTUN_FLAGS])
1926                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1927
1928         if (data[IFLA_IPTUN_PROTO])
1929                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1930
1931         if (data[IFLA_IPTUN_COLLECT_METADATA])
1932                 parms->collect_md = true;
1933
1934         if (data[IFLA_IPTUN_FWMARK])
1935                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1936 }
1937
1938 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1939                                         struct ip_tunnel_encap *ipencap)
1940 {
1941         bool ret = false;
1942
1943         memset(ipencap, 0, sizeof(*ipencap));
1944
1945         if (!data)
1946                 return ret;
1947
1948         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1949                 ret = true;
1950                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1951         }
1952
1953         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1954                 ret = true;
1955                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1956         }
1957
1958         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1959                 ret = true;
1960                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1961         }
1962
1963         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1964                 ret = true;
1965                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1966         }
1967
1968         return ret;
1969 }
1970
1971 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1972                            struct nlattr *tb[], struct nlattr *data[])
1973 {
1974         struct net *net = dev_net(dev);
1975         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1976         struct ip6_tnl *nt, *t;
1977         struct ip_tunnel_encap ipencap;
1978
1979         nt = netdev_priv(dev);
1980
1981         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1982                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1983
1984                 if (err < 0)
1985                         return err;
1986         }
1987
1988         ip6_tnl_netlink_parms(data, &nt->parms);
1989
1990         if (nt->parms.collect_md) {
1991                 if (rtnl_dereference(ip6n->collect_md_tun))
1992                         return -EEXIST;
1993         } else {
1994                 t = ip6_tnl_locate(net, &nt->parms, 0);
1995                 if (!IS_ERR(t))
1996                         return -EEXIST;
1997         }
1998
1999         return ip6_tnl_create2(dev);
2000 }
2001
2002 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2003                               struct nlattr *data[])
2004 {
2005         struct ip6_tnl *t = netdev_priv(dev);
2006         struct __ip6_tnl_parm p;
2007         struct net *net = t->net;
2008         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2009         struct ip_tunnel_encap ipencap;
2010
2011         if (dev == ip6n->fb_tnl_dev)
2012                 return -EINVAL;
2013
2014         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2015                 int err = ip6_tnl_encap_setup(t, &ipencap);
2016
2017                 if (err < 0)
2018                         return err;
2019         }
2020         ip6_tnl_netlink_parms(data, &p);
2021         if (p.collect_md)
2022                 return -EINVAL;
2023
2024         t = ip6_tnl_locate(net, &p, 0);
2025         if (!IS_ERR(t)) {
2026                 if (t->dev != dev)
2027                         return -EEXIST;
2028         } else
2029                 t = netdev_priv(dev);
2030
2031         return ip6_tnl_update(t, &p);
2032 }
2033
2034 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2035 {
2036         struct net *net = dev_net(dev);
2037         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2038
2039         if (dev != ip6n->fb_tnl_dev)
2040                 unregister_netdevice_queue(dev, head);
2041 }
2042
2043 static size_t ip6_tnl_get_size(const struct net_device *dev)
2044 {
2045         return
2046                 /* IFLA_IPTUN_LINK */
2047                 nla_total_size(4) +
2048                 /* IFLA_IPTUN_LOCAL */
2049                 nla_total_size(sizeof(struct in6_addr)) +
2050                 /* IFLA_IPTUN_REMOTE */
2051                 nla_total_size(sizeof(struct in6_addr)) +
2052                 /* IFLA_IPTUN_TTL */
2053                 nla_total_size(1) +
2054                 /* IFLA_IPTUN_ENCAP_LIMIT */
2055                 nla_total_size(1) +
2056                 /* IFLA_IPTUN_FLOWINFO */
2057                 nla_total_size(4) +
2058                 /* IFLA_IPTUN_FLAGS */
2059                 nla_total_size(4) +
2060                 /* IFLA_IPTUN_PROTO */
2061                 nla_total_size(1) +
2062                 /* IFLA_IPTUN_ENCAP_TYPE */
2063                 nla_total_size(2) +
2064                 /* IFLA_IPTUN_ENCAP_FLAGS */
2065                 nla_total_size(2) +
2066                 /* IFLA_IPTUN_ENCAP_SPORT */
2067                 nla_total_size(2) +
2068                 /* IFLA_IPTUN_ENCAP_DPORT */
2069                 nla_total_size(2) +
2070                 /* IFLA_IPTUN_COLLECT_METADATA */
2071                 nla_total_size(0) +
2072                 /* IFLA_IPTUN_FWMARK */
2073                 nla_total_size(4) +
2074                 0;
2075 }
2076
2077 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2078 {
2079         struct ip6_tnl *tunnel = netdev_priv(dev);
2080         struct __ip6_tnl_parm *parm = &tunnel->parms;
2081
2082         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2083             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2084             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2085             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2086             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2087             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2088             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2089             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2090             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2091                 goto nla_put_failure;
2092
2093         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2094             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2095             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2096             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2097                 goto nla_put_failure;
2098
2099         if (parm->collect_md)
2100                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2101                         goto nla_put_failure;
2102
2103         return 0;
2104
2105 nla_put_failure:
2106         return -EMSGSIZE;
2107 }
2108
2109 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2110 {
2111         struct ip6_tnl *tunnel = netdev_priv(dev);
2112
2113         return tunnel->net;
2114 }
2115 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2116
2117 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2118         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2119         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2120         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2121         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2122         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2123         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2124         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2125         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2126         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2127         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2128         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2129         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2130         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2131         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2132 };
2133
2134 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2135         .kind           = "ip6tnl",
2136         .maxtype        = IFLA_IPTUN_MAX,
2137         .policy         = ip6_tnl_policy,
2138         .priv_size      = sizeof(struct ip6_tnl),
2139         .setup          = ip6_tnl_dev_setup,
2140         .validate       = ip6_tnl_validate,
2141         .newlink        = ip6_tnl_newlink,
2142         .changelink     = ip6_tnl_changelink,
2143         .dellink        = ip6_tnl_dellink,
2144         .get_size       = ip6_tnl_get_size,
2145         .fill_info      = ip6_tnl_fill_info,
2146         .get_link_net   = ip6_tnl_get_link_net,
2147 };
2148
2149 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2150         .handler        = ip4ip6_rcv,
2151         .err_handler    = ip4ip6_err,
2152         .priority       =       1,
2153 };
2154
2155 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2156         .handler        = ip6ip6_rcv,
2157         .err_handler    = ip6ip6_err,
2158         .priority       =       1,
2159 };
2160
2161 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2162 {
2163         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2164         struct net_device *dev, *aux;
2165         int h;
2166         struct ip6_tnl *t;
2167         LIST_HEAD(list);
2168
2169         for_each_netdev_safe(net, dev, aux)
2170                 if (dev->rtnl_link_ops == &ip6_link_ops)
2171                         unregister_netdevice_queue(dev, &list);
2172
2173         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2174                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2175                 while (t) {
2176                         /* If dev is in the same netns, it has already
2177                          * been added to the list by the previous loop.
2178                          */
2179                         if (!net_eq(dev_net(t->dev), net))
2180                                 unregister_netdevice_queue(t->dev, &list);
2181                         t = rtnl_dereference(t->next);
2182                 }
2183         }
2184
2185         unregister_netdevice_many(&list);
2186 }
2187
2188 static int __net_init ip6_tnl_init_net(struct net *net)
2189 {
2190         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2191         struct ip6_tnl *t = NULL;
2192         int err;
2193
2194         ip6n->tnls[0] = ip6n->tnls_wc;
2195         ip6n->tnls[1] = ip6n->tnls_r_l;
2196
2197         err = -ENOMEM;
2198         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2199                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2200
2201         if (!ip6n->fb_tnl_dev)
2202                 goto err_alloc_dev;
2203         dev_net_set(ip6n->fb_tnl_dev, net);
2204         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2205         /* FB netdevice is special: we have one, and only one per netns.
2206          * Allowing to move it to another netns is clearly unsafe.
2207          */
2208         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2209
2210         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2211         if (err < 0)
2212                 goto err_register;
2213
2214         err = register_netdev(ip6n->fb_tnl_dev);
2215         if (err < 0)
2216                 goto err_register;
2217
2218         t = netdev_priv(ip6n->fb_tnl_dev);
2219
2220         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2221         return 0;
2222
2223 err_register:
2224         ip6_dev_free(ip6n->fb_tnl_dev);
2225 err_alloc_dev:
2226         return err;
2227 }
2228
2229 static void __net_exit ip6_tnl_exit_net(struct net *net)
2230 {
2231         rtnl_lock();
2232         ip6_tnl_destroy_tunnels(net);
2233         rtnl_unlock();
2234 }
2235
2236 static struct pernet_operations ip6_tnl_net_ops = {
2237         .init = ip6_tnl_init_net,
2238         .exit = ip6_tnl_exit_net,
2239         .id   = &ip6_tnl_net_id,
2240         .size = sizeof(struct ip6_tnl_net),
2241 };
2242
2243 /**
2244  * ip6_tunnel_init - register protocol and reserve needed resources
2245  *
2246  * Return: 0 on success
2247  **/
2248
2249 static int __init ip6_tunnel_init(void)
2250 {
2251         int  err;
2252
2253         err = register_pernet_device(&ip6_tnl_net_ops);
2254         if (err < 0)
2255                 goto out_pernet;
2256
2257         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2258         if (err < 0) {
2259                 pr_err("%s: can't register ip4ip6\n", __func__);
2260                 goto out_ip4ip6;
2261         }
2262
2263         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2264         if (err < 0) {
2265                 pr_err("%s: can't register ip6ip6\n", __func__);
2266                 goto out_ip6ip6;
2267         }
2268         err = rtnl_link_register(&ip6_link_ops);
2269         if (err < 0)
2270                 goto rtnl_link_failed;
2271
2272         return 0;
2273
2274 rtnl_link_failed:
2275         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2276 out_ip6ip6:
2277         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2278 out_ip4ip6:
2279         unregister_pernet_device(&ip6_tnl_net_ops);
2280 out_pernet:
2281         return err;
2282 }
2283
2284 /**
2285  * ip6_tunnel_cleanup - free resources and unregister protocol
2286  **/
2287
2288 static void __exit ip6_tunnel_cleanup(void)
2289 {
2290         rtnl_link_unregister(&ip6_link_ops);
2291         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2292                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2293
2294         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2295                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2296
2297         unregister_pernet_device(&ip6_tnl_net_ops);
2298 }
2299
2300 module_init(ip6_tunnel_init);
2301 module_exit(ip6_tunnel_cleanup);