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 && t->dev->flags & IFF_UP)
 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 }
 258
 259 static int ip6_tnl_create2(struct net_device *dev)
 260 {
 261         struct ip6_tnl *t = netdev_priv(dev);
 262         struct net *net = dev_net(dev);
 263         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 264         int err;
 265
 266         t = netdev_priv(dev);
 267
 268         dev->rtnl_link_ops = &ip6_link_ops;
 269         err = register_netdevice(dev);
 270         if (err < 0)
 271                 goto out;
 272
 273         strcpy(t->parms.name, dev->name);
 274
 275         ip6_tnl_link(ip6n, t);
 276         return 0;
 277
 278 out:
 279         return err;
 280 }
 281
 282 /**
 283  * ip6_tnl_create - create a new tunnel
 284  *   @p: tunnel parameters
 285  *   @pt: pointer to new tunnel
 286  *
 287  * Description:
 288  *   Create tunnel matching given parameters.
 289  *
 290  * Return:
 291  *   created tunnel or error pointer
 292  **/
 293
 294 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 295 {
 296         struct net_device *dev;
 297         struct ip6_tnl *t;
 298         char name[IFNAMSIZ];
 299         int err = -E2BIG;
 300
 301         if (p->name[0]) {
 302                 if (!dev_valid_name(p->name))
 303                         goto failed;
 304                 strlcpy(name, p->name, IFNAMSIZ);
 305         } else {
 306                 sprintf(name, "ip6tnl%%d");
 307         }
 308         err = -ENOMEM;
 309         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 310                            ip6_tnl_dev_setup);
 311         if (!dev)
 312                 goto failed;
 313
 314         dev_net_set(dev, net);
 315
 316         t = netdev_priv(dev);
 317         t->parms = *p;
 318         t->net = dev_net(dev);
 319         err = ip6_tnl_create2(dev);
 320         if (err < 0)
 321                 goto failed_free;
 322
 323         return t;
 324
 325 failed_free:
 326         free_netdev(dev);
 327 failed:
 328         return ERR_PTR(err);
 329 }
 330
 331 /**
 332  * ip6_tnl_locate - find or create tunnel matching given parameters
 333  *   @p: tunnel parameters
 334  *   @create: != 0 if allowed to create new tunnel if no match found
 335  *
 336  * Description:
 337  *   ip6_tnl_locate() first tries to locate an existing tunnel
 338  *   based on @parms. If this is unsuccessful, but @create is set a new
 339  *   tunnel device is created and registered for use.
 340  *
 341  * Return:
 342  *   matching tunnel or error pointer
 343  **/
 344
 345 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 346                 struct __ip6_tnl_parm *p, int create)
 347 {
 348         const struct in6_addr *remote = &p->raddr;
 349         const struct in6_addr *local = &p->laddr;
 350         struct ip6_tnl __rcu **tp;
 351         struct ip6_tnl *t;
 352         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 353
 354         for (tp = ip6_tnl_bucket(ip6n, p);
 355              (t = rtnl_dereference(*tp)) != NULL;
 356              tp = &t->next) {
 357                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 358                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 359                         if (create)
 360                                 return ERR_PTR(-EEXIST);
 361
 362                         return t;
 363                 }
 364         }
 365         if (!create)
 366                 return ERR_PTR(-ENODEV);
 367         return ip6_tnl_create(net, p);
 368 }
 369
 370 /**
 371  * ip6_tnl_dev_uninit - tunnel device uninitializer
 372  *   @dev: the device to be destroyed
 373  *
 374  * Description:
 375  *   ip6_tnl_dev_uninit() removes tunnel from its list
 376  **/
 377
 378 static void
 379 ip6_tnl_dev_uninit(struct net_device *dev)
 380 {
 381         struct ip6_tnl *t = netdev_priv(dev);
 382         struct net *net = t->net;
 383         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 384
 385         if (dev == ip6n->fb_tnl_dev)
 386                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 387         else
 388                 ip6_tnl_unlink(ip6n, t);
 389         dst_cache_reset(&t->dst_cache);
 390         dev_put(dev);
 391 }
 392
 393 /**
 394  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 395  *   @skb: received socket buffer
 396  *
 397  * Return:
 398  *   0 if none was found,
 399  *   else index to encapsulation limit
 400  **/
 401
 402 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 403 {
 404         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 405         unsigned int nhoff = raw - skb->data;
 406         unsigned int off = nhoff + sizeof(*ipv6h);
 407         u8 next, nexthdr = ipv6h->nexthdr;
 408
 409         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 410                 struct ipv6_opt_hdr *hdr;
 411                 u16 optlen;
 412
 413                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 414                         break;
 415
 416                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 417                 if (nexthdr == NEXTHDR_FRAGMENT) {
 418                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 419                         if (frag_hdr->frag_off)
 420                                 break;
 421                         optlen = 8;
 422                 } else if (nexthdr == NEXTHDR_AUTH) {
 423                         optlen = (hdr->hdrlen + 2) << 2;
 424                 } else {
 425                         optlen = ipv6_optlen(hdr);
 426                 }
 427                 /* cache hdr->nexthdr, since pskb_may_pull() might
 428                  * invalidate hdr
 429                  */
 430                 next = hdr->nexthdr;
 431                 if (nexthdr == NEXTHDR_DEST) {
 432                         u16 i = 2;
 433
 434                         /* Remember : hdr is no longer valid at this point. */
 435                         if (!pskb_may_pull(skb, off + optlen))
 436                                 break;
 437
 438                         while (1) {
 439                                 struct ipv6_tlv_tnl_enc_lim *tel;
 440
 441                                 /* No more room for encapsulation limit */
 442                                 if (i + sizeof(*tel) > optlen)
 443                                         break;
 444
 445                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 446                                 /* return index of option if found and valid */
 447                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 448                                     tel->length == 1)
 449                                         return i + off - nhoff;
 450                                 /* else jump to next option */
 451                                 if (tel->type)
 452                                         i += tel->length + 2;
 453                                 else
 454                                         i++;
 455                         }
 456                 }
 457                 nexthdr = next;
 458                 off += optlen;
 459         }
 460         return 0;
 461 }
 462 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 463
 464 /**
 465  * ip6_tnl_err - tunnel error handler
 466  *
 467  * Description:
 468  *   ip6_tnl_err() should handle errors in the tunnel according
 469  *   to the specifications in RFC 2473.
 470  **/
 471
 472 static int
 473 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 474             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 475 {
 476         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 477         struct net *net = dev_net(skb->dev);
 478         u8 rel_type = ICMPV6_DEST_UNREACH;
 479         u8 rel_code = ICMPV6_ADDR_UNREACH;
 480         __u32 rel_info = 0;
 481         struct ip6_tnl *t;
 482         int err = -ENOENT;
 483         int rel_msg = 0;
 484         u8 tproto;
 485         __u16 len;
 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 = READ_ONCE(t->parms.proto);
 497         if (tproto != ipproto && tproto != 0)
 498                 goto out;
 499
 500         err = 0;
 501
 502         switch (*type) {
 503                 struct ipv6_tlv_tnl_enc_lim *tel;
 504                 __u32 mtu, teli;
 505         case ICMPV6_DEST_UNREACH:
 506                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 507                                     t->parms.name);
 508                 rel_msg = 1;
 509                 break;
 510         case ICMPV6_TIME_EXCEED:
 511                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 512                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 513                                             t->parms.name);
 514                         rel_msg = 1;
 515                 }
 516                 break;
 517         case ICMPV6_PARAMPROB:
 518                 teli = 0;
 519                 if ((*code) == ICMPV6_HDR_FIELD)
 520                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 521
 522                 if (teli && teli == *info - 2) {
 523                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 524                         if (tel->encap_limit == 0) {
 525                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 526                                                     t->parms.name);
 527                                 rel_msg = 1;
 528                         }
 529                 } else {
 530                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 531                                             t->parms.name);
 532                 }
 533                 break;
 534         case ICMPV6_PKT_TOOBIG:
 535                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 536                                 sock_net_uid(net, NULL));
 537                 mtu = *info - offset;
 538                 if (mtu < IPV6_MIN_MTU)
 539                         mtu = IPV6_MIN_MTU;
 540                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 541                 if (len > mtu) {
 542                         rel_type = ICMPV6_PKT_TOOBIG;
 543                         rel_code = 0;
 544                         rel_info = mtu;
 545                         rel_msg = 1;
 546                 }
 547                 break;
 548         case NDISC_REDIRECT:
 549                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 550                              sock_net_uid(net, NULL));
 551                 break;
 552         }
 553
 554         *type = rel_type;
 555         *code = rel_code;
 556         *info = rel_info;
 557         *msg = rel_msg;
 558
 559 out:
 560         rcu_read_unlock();
 561         return err;
 562 }
 563
 564 static int
 565 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 566            u8 type, u8 code, int offset, __be32 info)
 567 {
 568         __u32 rel_info = ntohl(info);
 569         const struct iphdr *eiph;
 570         struct sk_buff *skb2;
 571         int err, rel_msg = 0;
 572         u8 rel_type = type;
 573         u8 rel_code = code;
 574         struct rtable *rt;
 575         struct flowi4 fl4;
 576
 577         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 578                           &rel_msg, &rel_info, offset);
 579         if (err < 0)
 580                 return err;
 581
 582         if (rel_msg == 0)
 583                 return 0;
 584
 585         switch (rel_type) {
 586         case ICMPV6_DEST_UNREACH:
 587                 if (rel_code != ICMPV6_ADDR_UNREACH)
 588                         return 0;
 589                 rel_type = ICMP_DEST_UNREACH;
 590                 rel_code = ICMP_HOST_UNREACH;
 591                 break;
 592         case ICMPV6_PKT_TOOBIG:
 593                 if (rel_code != 0)
 594                         return 0;
 595                 rel_type = ICMP_DEST_UNREACH;
 596                 rel_code = ICMP_FRAG_NEEDED;
 597                 break;
 598         default:
 599                 return 0;
 600         }
 601
 602         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 603                 return 0;
 604
 605         skb2 = skb_clone(skb, GFP_ATOMIC);
 606         if (!skb2)
 607                 return 0;
 608
 609         skb_dst_drop(skb2);
 610
 611         skb_pull(skb2, offset);
 612         skb_reset_network_header(skb2);
 613         eiph = ip_hdr(skb2);
 614
 615         /* Try to guess incoming interface */
 616         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 617                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 618         if (IS_ERR(rt))
 619                 goto out;
 620
 621         skb2->dev = rt->dst.dev;
 622         ip_rt_put(rt);
 623
 624         /* route "incoming" packet */
 625         if (rt->rt_flags & RTCF_LOCAL) {
 626                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 627                                            eiph->daddr, eiph->saddr, 0, 0,
 628                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 629                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
 630                         if (!IS_ERR(rt))
 631                                 ip_rt_put(rt);
 632                         goto out;
 633                 }
 634                 skb_dst_set(skb2, &rt->dst);
 635         } else {
 636                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 637                                    skb2->dev) ||
 638                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 639                         goto out;
 640         }
 641
 642         /* change mtu on this route */
 643         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 644                 if (rel_info > dst_mtu(skb_dst(skb2)))
 645                         goto out;
 646
 647                 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
 648         }
 649
 650         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 651
 652 out:
 653         kfree_skb(skb2);
 654         return 0;
 655 }
 656
 657 static int
 658 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 659            u8 type, u8 code, int offset, __be32 info)
 660 {
 661         __u32 rel_info = ntohl(info);
 662         int err, rel_msg = 0;
 663         u8 rel_type = type;
 664         u8 rel_code = code;
 665
 666         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 667                           &rel_msg, &rel_info, offset);
 668         if (err < 0)
 669                 return err;
 670
 671         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 672                 struct rt6_info *rt;
 673                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 674
 675                 if (!skb2)
 676                         return 0;
 677
 678                 skb_dst_drop(skb2);
 679                 skb_pull(skb2, offset);
 680                 skb_reset_network_header(skb2);
 681
 682                 /* Try to guess incoming interface */
 683                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 684                                 NULL, 0, skb2, 0);
 685
 686                 if (rt && rt->dst.dev)
 687                         skb2->dev = rt->dst.dev;
 688
 689                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 690
 691                 ip6_rt_put(rt);
 692
 693                 kfree_skb(skb2);
 694         }
 695
 696         return 0;
 697 }
 698
 699 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 700                                        const struct ipv6hdr *ipv6h,
 701                                        struct sk_buff *skb)
 702 {
 703         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 704
 705         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 706                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 707
 708         return IP6_ECN_decapsulate(ipv6h, skb);
 709 }
 710
 711 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 712                                        const struct ipv6hdr *ipv6h,
 713                                        struct sk_buff *skb)
 714 {
 715         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 716                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 717
 718         return IP6_ECN_decapsulate(ipv6h, skb);
 719 }
 720
 721 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 722                              const struct in6_addr *laddr,
 723                              const struct in6_addr *raddr)
 724 {
 725         struct __ip6_tnl_parm *p = &t->parms;
 726         int ltype = ipv6_addr_type(laddr);
 727         int rtype = ipv6_addr_type(raddr);
 728         __u32 flags = 0;
 729
 730         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 731                 flags = IP6_TNL_F_CAP_PER_PACKET;
 732         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 733                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 734                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 735                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 736                 if (ltype&IPV6_ADDR_UNICAST)
 737                         flags |= IP6_TNL_F_CAP_XMIT;
 738                 if (rtype&IPV6_ADDR_UNICAST)
 739                         flags |= IP6_TNL_F_CAP_RCV;
 740         }
 741         return flags;
 742 }
 743 EXPORT_SYMBOL(ip6_tnl_get_cap);
 744
 745 /* called with rcu_read_lock() */
 746 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 747                                   const struct in6_addr *laddr,
 748                                   const struct in6_addr *raddr)
 749 {
 750         struct __ip6_tnl_parm *p = &t->parms;
 751         int ret = 0;
 752         struct net *net = t->net;
 753
 754         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 755             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 756              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 757                 struct net_device *ldev = NULL;
 758
 759                 if (p->link)
 760                         ldev = dev_get_by_index_rcu(net, p->link);
 761
 762                 if ((ipv6_addr_is_multicast(laddr) ||
 763                      likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
 764                                                     0, IFA_F_TENTATIVE))) &&
 765                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
 766                      likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
 767                                                      0, IFA_F_TENTATIVE))))
 768                         ret = 1;
 769         }
 770         return ret;
 771 }
 772 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 773
 774 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 775                          const struct tnl_ptk_info *tpi,
 776                          struct metadata_dst *tun_dst,
 777                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 778                                                 const struct ipv6hdr *ipv6h,
 779                                                 struct sk_buff *skb),
 780                          bool log_ecn_err)
 781 {
 782         struct pcpu_sw_netstats *tstats;
 783         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 784         int err;
 785
 786         if ((!(tpi->flags & TUNNEL_CSUM) &&
 787              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 788             ((tpi->flags & TUNNEL_CSUM) &&
 789              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 790                 tunnel->dev->stats.rx_crc_errors++;
 791                 tunnel->dev->stats.rx_errors++;
 792                 goto drop;
 793         }
 794
 795         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 796                 if (!(tpi->flags & TUNNEL_SEQ) ||
 797                     (tunnel->i_seqno &&
 798                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 799                         tunnel->dev->stats.rx_fifo_errors++;
 800                         tunnel->dev->stats.rx_errors++;
 801                         goto drop;
 802                 }
 803                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 804         }
 805
 806         skb->protocol = tpi->proto;
 807
 808         /* Warning: All skb pointers will be invalidated! */
 809         if (tunnel->dev->type == ARPHRD_ETHER) {
 810                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 811                         tunnel->dev->stats.rx_length_errors++;
 812                         tunnel->dev->stats.rx_errors++;
 813                         goto drop;
 814                 }
 815
 816                 ipv6h = ipv6_hdr(skb);
 817                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 818                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 819         } else {
 820                 skb->dev = tunnel->dev;
 821         }
 822
 823         skb_reset_network_header(skb);
 824         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 825
 826         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 827
 828         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 829         if (unlikely(err)) {
 830                 if (log_ecn_err)
 831                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 832                                              &ipv6h->saddr,
 833                                              ipv6_get_dsfield(ipv6h));
 834                 if (err > 1) {
 835                         ++tunnel->dev->stats.rx_frame_errors;
 836                         ++tunnel->dev->stats.rx_errors;
 837                         goto drop;
 838                 }
 839         }
 840
 841         tstats = this_cpu_ptr(tunnel->dev->tstats);
 842         u64_stats_update_begin(&tstats->syncp);
 843         tstats->rx_packets++;
 844         tstats->rx_bytes += skb->len;
 845         u64_stats_update_end(&tstats->syncp);
 846
 847         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 848
 849         if (tun_dst)
 850                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 851
 852         gro_cells_receive(&tunnel->gro_cells, skb);
 853         return 0;
 854
 855 drop:
 856         if (tun_dst)
 857                 dst_release((struct dst_entry *)tun_dst);
 858         kfree_skb(skb);
 859         return 0;
 860 }
 861
 862 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 863                 const struct tnl_ptk_info *tpi,
 864                 struct metadata_dst *tun_dst,
 865                 bool log_ecn_err)
 866 {
 867         int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 868                                     const struct ipv6hdr *ipv6h,
 869                                     struct sk_buff *skb);
 870
 871         dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
 872         if (tpi->proto == htons(ETH_P_IP))
 873                 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
 874
 875         return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 876                              log_ecn_err);
 877 }
 878 EXPORT_SYMBOL(ip6_tnl_rcv);
 879
 880 static const struct tnl_ptk_info tpi_v6 = {
 881         /* no tunnel info required for ipxip6. */
 882         .proto = htons(ETH_P_IPV6),
 883 };
 884
 885 static const struct tnl_ptk_info tpi_v4 = {
 886         /* no tunnel info required for ipxip6. */
 887         .proto = htons(ETH_P_IP),
 888 };
 889
 890 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 891                       const struct tnl_ptk_info *tpi,
 892                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 893                                                   const struct ipv6hdr *ipv6h,
 894                                                   struct sk_buff *skb))
 895 {
 896         struct ip6_tnl *t;
 897         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 898         struct metadata_dst *tun_dst = NULL;
 899         int ret = -1;
 900
 901         rcu_read_lock();
 902         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 903
 904         if (t) {
 905                 u8 tproto = READ_ONCE(t->parms.proto);
 906
 907                 if (tproto != ipproto && tproto != 0)
 908                         goto drop;
 909                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 910                         goto drop;
 911                 ipv6h = ipv6_hdr(skb);
 912                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 913                         goto drop;
 914                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 915                         goto drop;
 916                 if (t->parms.collect_md) {
 917                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 918                         if (!tun_dst)
 919                                 goto drop;
 920                 }
 921                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 922                                     log_ecn_error);
 923         }
 924
 925         rcu_read_unlock();
 926
 927         return ret;
 928
 929 drop:
 930         rcu_read_unlock();
 931         kfree_skb(skb);
 932         return 0;
 933 }
 934
 935 static int ip4ip6_rcv(struct sk_buff *skb)
 936 {
 937         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 938                           ip4ip6_dscp_ecn_decapsulate);
 939 }
 940
 941 static int ip6ip6_rcv(struct sk_buff *skb)
 942 {
 943         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 944                           ip6ip6_dscp_ecn_decapsulate);
 945 }
 946
 947 struct ipv6_tel_txoption {
 948         struct ipv6_txoptions ops;
 949         __u8 dst_opt[8];
 950 };
 951
 952 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 953 {
 954         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 955
 956         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 957         opt->dst_opt[3] = 1;
 958         opt->dst_opt[4] = encap_limit;
 959         opt->dst_opt[5] = IPV6_TLV_PADN;
 960         opt->dst_opt[6] = 1;
 961
 962         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 963         opt->ops.opt_nflen = 8;
 964 }
 965
 966 /**
 967  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 968  *   @t: the outgoing tunnel device
 969  *   @hdr: IPv6 header from the incoming packet
 970  *
 971  * Description:
 972  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 973  *   doesn't match source of incoming packet.
 974  *
 975  * Return:
 976  *   1 if conflict,
 977  *   0 else
 978  **/
 979
 980 static inline bool
 981 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 982 {
 983         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 984 }
 985
 986 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 987                      const struct in6_addr *laddr,
 988                      const struct in6_addr *raddr)
 989 {
 990         struct __ip6_tnl_parm *p = &t->parms;
 991         int ret = 0;
 992         struct net *net = t->net;
 993
 994         if (t->parms.collect_md)
 995                 return 1;
 996
 997         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 998             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 999              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1000                 struct net_device *ldev = NULL;
1001
1002                 rcu_read_lock();
1003                 if (p->link)
1004                         ldev = dev_get_by_index_rcu(net, p->link);
1005
1006                 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1007                                                       0, IFA_F_TENTATIVE)))
1008                         pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1009                                             p->name);
1010                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1011                          !ipv6_addr_is_multicast(raddr) &&
1012                          unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1013                                                           true, 0, IFA_F_TENTATIVE)))
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 && ipv6_addr_any(&fl6->saddr) &&
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         skb_dst_update_pmtu_no_confirm(skb, mtu);
1148         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1149                 *pmtu = mtu;
1150                 err = -EMSGSIZE;
1151                 goto tx_err_dst_release;
1152         }
1153
1154         if (t->err_count > 0) {
1155                 if (time_before(jiffies,
1156                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1157                         t->err_count--;
1158
1159                         dst_link_failure(skb);
1160                 } else {
1161                         t->err_count = 0;
1162                 }
1163         }
1164
1165         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1166
1167         /*
1168          * Okay, now see if we can stuff it in the buffer as-is.
1169          */
1170         max_headroom += LL_RESERVED_SPACE(tdev);
1171
1172         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1173             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1174                 struct sk_buff *new_skb;
1175
1176                 new_skb = skb_realloc_headroom(skb, max_headroom);
1177                 if (!new_skb)
1178                         goto tx_err_dst_release;
1179
1180                 if (skb->sk)
1181                         skb_set_owner_w(new_skb, skb->sk);
1182                 consume_skb(skb);
1183                 skb = new_skb;
1184         }
1185
1186         if (t->parms.collect_md) {
1187                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1188                         goto tx_err_dst_release;
1189         } else {
1190                 if (use_cache && ndst)
1191                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1192         }
1193         skb_dst_set(skb, dst);
1194
1195         if (hop_limit == 0) {
1196                 if (skb->protocol == htons(ETH_P_IP))
1197                         hop_limit = ip_hdr(skb)->ttl;
1198                 else if (skb->protocol == htons(ETH_P_IPV6))
1199                         hop_limit = ipv6_hdr(skb)->hop_limit;
1200                 else
1201                         hop_limit = ip6_dst_hoplimit(dst);
1202         }
1203
1204         /* Calculate max headroom for all the headers and adjust
1205          * needed_headroom if necessary.
1206          */
1207         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1208                         + dst->header_len + t->hlen;
1209         if (max_headroom > dev->needed_headroom)
1210                 dev->needed_headroom = max_headroom;
1211
1212         err = ip6_tnl_encap(skb, t, &proto, fl6);
1213         if (err)
1214                 return err;
1215
1216         if (encap_limit >= 0) {
1217                 init_tel_txopt(&opt, encap_limit);
1218                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1219         }
1220
1221         skb_push(skb, sizeof(struct ipv6hdr));
1222         skb_reset_network_header(skb);
1223         ipv6h = ipv6_hdr(skb);
1224         ip6_flow_hdr(ipv6h, dsfield,
1225                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1226         ipv6h->hop_limit = hop_limit;
1227         ipv6h->nexthdr = proto;
1228         ipv6h->saddr = fl6->saddr;
1229         ipv6h->daddr = fl6->daddr;
1230         ip6tunnel_xmit(NULL, skb, dev);
1231         return 0;
1232 tx_err_link_failure:
1233         stats->tx_carrier_errors++;
1234         dst_link_failure(skb);
1235 tx_err_dst_release:
1236         dst_release(dst);
1237         return err;
1238 }
1239 EXPORT_SYMBOL(ip6_tnl_xmit);
1240
1241 static inline int
1242 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1243 {
1244         struct ip6_tnl *t = netdev_priv(dev);
1245         const struct iphdr  *iph;
1246         int encap_limit = -1;
1247         struct flowi6 fl6;
1248         __u8 dsfield;
1249         __u32 mtu;
1250         u8 tproto;
1251         int err;
1252
1253         iph = ip_hdr(skb);
1254         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1255
1256         tproto = READ_ONCE(t->parms.proto);
1257         if (tproto != IPPROTO_IPIP && tproto != 0)
1258                 return -1;
1259
1260         if (t->parms.collect_md) {
1261                 struct ip_tunnel_info *tun_info;
1262                 const struct ip_tunnel_key *key;
1263
1264                 tun_info = skb_tunnel_info(skb);
1265                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1266                              ip_tunnel_info_af(tun_info) != AF_INET6))
1267                         return -1;
1268                 key = &tun_info->key;
1269                 memset(&fl6, 0, sizeof(fl6));
1270                 fl6.flowi6_proto = IPPROTO_IPIP;
1271                 fl6.saddr = key->u.ipv6.src;
1272                 fl6.daddr = key->u.ipv6.dst;
1273                 fl6.flowlabel = key->label;
1274                 dsfield =  key->tos;
1275         } else {
1276                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1277                         encap_limit = t->parms.encap_limit;
1278
1279                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1280                 fl6.flowi6_proto = IPPROTO_IPIP;
1281
1282                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1283                         dsfield = ipv4_get_dsfield(iph);
1284                 else
1285                         dsfield = ip6_tclass(t->parms.flowinfo);
1286                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1287                         fl6.flowi6_mark = skb->mark;
1288                 else
1289                         fl6.flowi6_mark = t->parms.fwmark;
1290         }
1291
1292         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1293         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1294
1295         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1296                 return -1;
1297
1298         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1299
1300         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1301                            IPPROTO_IPIP);
1302         if (err != 0) {
1303                 /* XXX: send ICMP error even if DF is not set. */
1304                 if (err == -EMSGSIZE)
1305                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1306                                   htonl(mtu));
1307                 return -1;
1308         }
1309
1310         return 0;
1311 }
1312
1313 static inline int
1314 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1315 {
1316         struct ip6_tnl *t = netdev_priv(dev);
1317         struct ipv6hdr *ipv6h;
1318         int encap_limit = -1;
1319         __u16 offset;
1320         struct flowi6 fl6;
1321         __u8 dsfield;
1322         __u32 mtu;
1323         u8 tproto;
1324         int err;
1325
1326         ipv6h = ipv6_hdr(skb);
1327         tproto = READ_ONCE(t->parms.proto);
1328         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1329             ip6_tnl_addr_conflict(t, ipv6h))
1330                 return -1;
1331
1332         if (t->parms.collect_md) {
1333                 struct ip_tunnel_info *tun_info;
1334                 const struct ip_tunnel_key *key;
1335
1336                 tun_info = skb_tunnel_info(skb);
1337                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1338                              ip_tunnel_info_af(tun_info) != AF_INET6))
1339                         return -1;
1340                 key = &tun_info->key;
1341                 memset(&fl6, 0, sizeof(fl6));
1342                 fl6.flowi6_proto = IPPROTO_IPV6;
1343                 fl6.saddr = key->u.ipv6.src;
1344                 fl6.daddr = key->u.ipv6.dst;
1345                 fl6.flowlabel = key->label;
1346                 dsfield = key->tos;
1347         } else {
1348                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1349                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1350                 ipv6h = ipv6_hdr(skb);
1351                 if (offset > 0) {
1352                         struct ipv6_tlv_tnl_enc_lim *tel;
1353
1354                         tel = (void *)&skb_network_header(skb)[offset];
1355                         if (tel->encap_limit == 0) {
1356                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1357                                             ICMPV6_HDR_FIELD, offset + 2);
1358                                 return -1;
1359                         }
1360                         encap_limit = tel->encap_limit - 1;
1361                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1362                         encap_limit = t->parms.encap_limit;
1363                 }
1364
1365                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1366                 fl6.flowi6_proto = IPPROTO_IPV6;
1367
1368                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1369                         dsfield = ipv6_get_dsfield(ipv6h);
1370                 else
1371                         dsfield = ip6_tclass(t->parms.flowinfo);
1372                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1373                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1374                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1375                         fl6.flowi6_mark = skb->mark;
1376                 else
1377                         fl6.flowi6_mark = t->parms.fwmark;
1378         }
1379
1380         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1381         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1382
1383         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1384                 return -1;
1385
1386         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1387
1388         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1389                            IPPROTO_IPV6);
1390         if (err != 0) {
1391                 if (err == -EMSGSIZE)
1392                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1393                 return -1;
1394         }
1395
1396         return 0;
1397 }
1398
1399 static netdev_tx_t
1400 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1401 {
1402         struct ip6_tnl *t = netdev_priv(dev);
1403         struct net_device_stats *stats = &t->dev->stats;
1404         int ret;
1405
1406         if (!pskb_inet_may_pull(skb))
1407                 goto tx_err;
1408
1409         switch (skb->protocol) {
1410         case htons(ETH_P_IP):
1411                 ret = ip4ip6_tnl_xmit(skb, dev);
1412                 break;
1413         case htons(ETH_P_IPV6):
1414                 ret = ip6ip6_tnl_xmit(skb, dev);
1415                 break;
1416         default:
1417                 goto tx_err;
1418         }
1419
1420         if (ret < 0)
1421                 goto tx_err;
1422
1423         return NETDEV_TX_OK;
1424
1425 tx_err:
1426         stats->tx_errors++;
1427         stats->tx_dropped++;
1428         kfree_skb(skb);
1429         return NETDEV_TX_OK;
1430 }
1431
1432 static void ip6_tnl_link_config(struct ip6_tnl *t)
1433 {
1434         struct net_device *dev = t->dev;
1435         struct __ip6_tnl_parm *p = &t->parms;
1436         struct flowi6 *fl6 = &t->fl.u.ip6;
1437         int t_hlen;
1438
1439         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1440         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1441
1442         /* Set up flowi template */
1443         fl6->saddr = p->laddr;
1444         fl6->daddr = p->raddr;
1445         fl6->flowi6_oif = p->link;
1446         fl6->flowlabel = 0;
1447
1448         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1449                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1450         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1451                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1452
1453         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1454         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1455
1456         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1457                 dev->flags |= IFF_POINTOPOINT;
1458         else
1459                 dev->flags &= ~IFF_POINTOPOINT;
1460
1461         t->tun_hlen = 0;
1462         t->hlen = t->encap_hlen + t->tun_hlen;
1463         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1464
1465         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1466                 int strict = (ipv6_addr_type(&p->raddr) &
1467                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1468
1469                 struct rt6_info *rt = rt6_lookup(t->net,
1470                                                  &p->raddr, &p->laddr,
1471                                                  p->link, NULL, strict);
1472
1473                 if (!rt)
1474                         return;
1475
1476                 if (rt->dst.dev) {
1477                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1478                                 t_hlen;
1479
1480                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1481                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1482                                 dev->mtu -= 8;
1483
1484                         if (dev->mtu < IPV6_MIN_MTU)
1485                                 dev->mtu = IPV6_MIN_MTU;
1486                 }
1487                 ip6_rt_put(rt);
1488         }
1489 }
1490
1491 /**
1492  * ip6_tnl_change - update the tunnel parameters
1493  *   @t: tunnel to be changed
1494  *   @p: tunnel configuration parameters
1495  *
1496  * Description:
1497  *   ip6_tnl_change() updates the tunnel parameters
1498  **/
1499
1500 static int
1501 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1502 {
1503         t->parms.laddr = p->laddr;
1504         t->parms.raddr = p->raddr;
1505         t->parms.flags = p->flags;
1506         t->parms.hop_limit = p->hop_limit;
1507         t->parms.encap_limit = p->encap_limit;
1508         t->parms.flowinfo = p->flowinfo;
1509         t->parms.link = p->link;
1510         t->parms.proto = p->proto;
1511         t->parms.fwmark = p->fwmark;
1512         dst_cache_reset(&t->dst_cache);
1513         ip6_tnl_link_config(t);
1514         return 0;
1515 }
1516
1517 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1518 {
1519         struct net *net = t->net;
1520         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1521         int err;
1522
1523         ip6_tnl_unlink(ip6n, t);
1524         synchronize_net();
1525         err = ip6_tnl_change(t, p);
1526         ip6_tnl_link(ip6n, t);
1527         netdev_state_change(t->dev);
1528         return err;
1529 }
1530
1531 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1532 {
1533         /* for default tnl0 device allow to change only the proto */
1534         t->parms.proto = p->proto;
1535         netdev_state_change(t->dev);
1536         return 0;
1537 }
1538
1539 static void
1540 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1541 {
1542         p->laddr = u->laddr;
1543         p->raddr = u->raddr;
1544         p->flags = u->flags;
1545         p->hop_limit = u->hop_limit;
1546         p->encap_limit = u->encap_limit;
1547         p->flowinfo = u->flowinfo;
1548         p->link = u->link;
1549         p->proto = u->proto;
1550         memcpy(p->name, u->name, sizeof(u->name));
1551 }
1552
1553 static void
1554 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1555 {
1556         u->laddr = p->laddr;
1557         u->raddr = p->raddr;
1558         u->flags = p->flags;
1559         u->hop_limit = p->hop_limit;
1560         u->encap_limit = p->encap_limit;
1561         u->flowinfo = p->flowinfo;
1562         u->link = p->link;
1563         u->proto = p->proto;
1564         memcpy(u->name, p->name, sizeof(u->name));
1565 }
1566
1567 /**
1568  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1569  *   @dev: virtual device associated with tunnel
1570  *   @ifr: parameters passed from userspace
1571  *   @cmd: command to be performed
1572  *
1573  * Description:
1574  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1575  *   from userspace.
1576  *
1577  *   The possible commands are the following:
1578  *     %SIOCGETTUNNEL: get tunnel parameters for device
1579  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1580  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1581  *     %SIOCDELTUNNEL: delete tunnel
1582  *
1583  *   The fallback device "ip6tnl0", created during module
1584  *   initialization, can be used for creating other tunnel devices.
1585  *
1586  * Return:
1587  *   0 on success,
1588  *   %-EFAULT if unable to copy data to or from userspace,
1589  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1590  *   %-EINVAL if passed tunnel parameters are invalid,
1591  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1592  *   %-ENODEV if attempting to change or delete a nonexisting device
1593  **/
1594
1595 static int
1596 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1597 {
1598         int err = 0;
1599         struct ip6_tnl_parm p;
1600         struct __ip6_tnl_parm p1;
1601         struct ip6_tnl *t = netdev_priv(dev);
1602         struct net *net = t->net;
1603         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1604
1605         memset(&p1, 0, sizeof(p1));
1606
1607         switch (cmd) {
1608         case SIOCGETTUNNEL:
1609                 if (dev == ip6n->fb_tnl_dev) {
1610                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1611                                 err = -EFAULT;
1612                                 break;
1613                         }
1614                         ip6_tnl_parm_from_user(&p1, &p);
1615                         t = ip6_tnl_locate(net, &p1, 0);
1616                         if (IS_ERR(t))
1617                                 t = netdev_priv(dev);
1618                 } else {
1619                         memset(&p, 0, sizeof(p));
1620                 }
1621                 ip6_tnl_parm_to_user(&p, &t->parms);
1622                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1623                         err = -EFAULT;
1624                 }
1625                 break;
1626         case SIOCADDTUNNEL:
1627         case SIOCCHGTUNNEL:
1628                 err = -EPERM;
1629                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1630                         break;
1631                 err = -EFAULT;
1632                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1633                         break;
1634                 err = -EINVAL;
1635                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1636                     p.proto != 0)
1637                         break;
1638                 ip6_tnl_parm_from_user(&p1, &p);
1639                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1640                 if (cmd == SIOCCHGTUNNEL) {
1641                         if (!IS_ERR(t)) {
1642                                 if (t->dev != dev) {
1643                                         err = -EEXIST;
1644                                         break;
1645                                 }
1646                         } else
1647                                 t = netdev_priv(dev);
1648                         if (dev == ip6n->fb_tnl_dev)
1649                                 err = ip6_tnl0_update(t, &p1);
1650                         else
1651                                 err = ip6_tnl_update(t, &p1);
1652                 }
1653                 if (!IS_ERR(t)) {
1654                         err = 0;
1655                         ip6_tnl_parm_to_user(&p, &t->parms);
1656                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1657                                 err = -EFAULT;
1658
1659                 } else {
1660                         err = PTR_ERR(t);
1661                 }
1662                 break;
1663         case SIOCDELTUNNEL:
1664                 err = -EPERM;
1665                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1666                         break;
1667
1668                 if (dev == ip6n->fb_tnl_dev) {
1669                         err = -EFAULT;
1670                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1671                                 break;
1672                         err = -ENOENT;
1673                         ip6_tnl_parm_from_user(&p1, &p);
1674                         t = ip6_tnl_locate(net, &p1, 0);
1675                         if (IS_ERR(t))
1676                                 break;
1677                         err = -EPERM;
1678                         if (t->dev == ip6n->fb_tnl_dev)
1679                                 break;
1680                         dev = t->dev;
1681                 }
1682                 err = 0;
1683                 unregister_netdevice(dev);
1684                 break;
1685         default:
1686                 err = -EINVAL;
1687         }
1688         return err;
1689 }
1690
1691 /**
1692  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1693  *   @dev: virtual device associated with tunnel
1694  *   @new_mtu: the new mtu
1695  *
1696  * Return:
1697  *   0 on success,
1698  *   %-EINVAL if mtu too small
1699  **/
1700
1701 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1702 {
1703         struct ip6_tnl *tnl = netdev_priv(dev);
1704
1705         if (tnl->parms.proto == IPPROTO_IPV6) {
1706                 if (new_mtu < IPV6_MIN_MTU)
1707                         return -EINVAL;
1708         } else {
1709                 if (new_mtu < ETH_MIN_MTU)
1710                         return -EINVAL;
1711         }
1712         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1713                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1714                         return -EINVAL;
1715         } else {
1716                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1717                         return -EINVAL;
1718         }
1719         dev->mtu = new_mtu;
1720         return 0;
1721 }
1722 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1723
1724 int ip6_tnl_get_iflink(const struct net_device *dev)
1725 {
1726         struct ip6_tnl *t = netdev_priv(dev);
1727
1728         return t->parms.link;
1729 }
1730 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1731
1732 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1733                           unsigned int num)
1734 {
1735         if (num >= MAX_IPTUN_ENCAP_OPS)
1736                 return -ERANGE;
1737
1738         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1739                         &ip6tun_encaps[num],
1740                         NULL, ops) ? 0 : -1;
1741 }
1742 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1743
1744 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1745                           unsigned int num)
1746 {
1747         int ret;
1748
1749         if (num >= MAX_IPTUN_ENCAP_OPS)
1750                 return -ERANGE;
1751
1752         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1753                        &ip6tun_encaps[num],
1754                        ops, NULL) == ops) ? 0 : -1;
1755
1756         synchronize_net();
1757
1758         return ret;
1759 }
1760 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1761
1762 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1763                         struct ip_tunnel_encap *ipencap)
1764 {
1765         int hlen;
1766
1767         memset(&t->encap, 0, sizeof(t->encap));
1768
1769         hlen = ip6_encap_hlen(ipencap);
1770         if (hlen < 0)
1771                 return hlen;
1772
1773         t->encap.type = ipencap->type;
1774         t->encap.sport = ipencap->sport;
1775         t->encap.dport = ipencap->dport;
1776         t->encap.flags = ipencap->flags;
1777
1778         t->encap_hlen = hlen;
1779         t->hlen = t->encap_hlen + t->tun_hlen;
1780
1781         return 0;
1782 }
1783 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1784
1785 static const struct net_device_ops ip6_tnl_netdev_ops = {
1786         .ndo_init       = ip6_tnl_dev_init,
1787         .ndo_uninit     = ip6_tnl_dev_uninit,
1788         .ndo_start_xmit = ip6_tnl_start_xmit,
1789         .ndo_do_ioctl   = ip6_tnl_ioctl,
1790         .ndo_change_mtu = ip6_tnl_change_mtu,
1791         .ndo_get_stats  = ip6_get_stats,
1792         .ndo_get_iflink = ip6_tnl_get_iflink,
1793 };
1794
1795 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1796                          NETIF_F_FRAGLIST |     \
1797                          NETIF_F_HIGHDMA |      \
1798                          NETIF_F_GSO_SOFTWARE | \
1799                          NETIF_F_HW_CSUM)
1800
1801 /**
1802  * ip6_tnl_dev_setup - setup virtual tunnel device
1803  *   @dev: virtual device associated with tunnel
1804  *
1805  * Description:
1806  *   Initialize function pointers and device parameters
1807  **/
1808
1809 static void ip6_tnl_dev_setup(struct net_device *dev)
1810 {
1811         dev->netdev_ops = &ip6_tnl_netdev_ops;
1812         dev->needs_free_netdev = true;
1813         dev->priv_destructor = ip6_dev_free;
1814
1815         dev->type = ARPHRD_TUNNEL6;
1816         dev->flags |= IFF_NOARP;
1817         dev->addr_len = sizeof(struct in6_addr);
1818         dev->features |= NETIF_F_LLTX;
1819         netif_keep_dst(dev);
1820
1821         dev->features           |= IPXIPX_FEATURES;
1822         dev->hw_features        |= IPXIPX_FEATURES;
1823
1824         /* This perm addr will be used as interface identifier by IPv6 */
1825         dev->addr_assign_type = NET_ADDR_RANDOM;
1826         eth_random_addr(dev->perm_addr);
1827 }
1828
1829
1830 /**
1831  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1832  *   @dev: virtual device associated with tunnel
1833  **/
1834
1835 static inline int
1836 ip6_tnl_dev_init_gen(struct net_device *dev)
1837 {
1838         struct ip6_tnl *t = netdev_priv(dev);
1839         int ret;
1840         int t_hlen;
1841
1842         t->dev = dev;
1843         t->net = dev_net(dev);
1844         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1845         if (!dev->tstats)
1846                 return -ENOMEM;
1847
1848         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1849         if (ret)
1850                 goto free_stats;
1851
1852         ret = gro_cells_init(&t->gro_cells, dev);
1853         if (ret)
1854                 goto destroy_dst;
1855
1856         t->tun_hlen = 0;
1857         t->hlen = t->encap_hlen + t->tun_hlen;
1858         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1859
1860         dev->type = ARPHRD_TUNNEL6;
1861         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1862         dev->mtu = ETH_DATA_LEN - t_hlen;
1863         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1864                 dev->mtu -= 8;
1865         dev->min_mtu = ETH_MIN_MTU;
1866         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1867
1868         dev_hold(dev);
1869         return 0;
1870
1871 destroy_dst:
1872         dst_cache_destroy(&t->dst_cache);
1873 free_stats:
1874         free_percpu(dev->tstats);
1875         dev->tstats = NULL;
1876
1877         return ret;
1878 }
1879
1880 /**
1881  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1882  *   @dev: virtual device associated with tunnel
1883  **/
1884
1885 static int ip6_tnl_dev_init(struct net_device *dev)
1886 {
1887         struct ip6_tnl *t = netdev_priv(dev);
1888         int err = ip6_tnl_dev_init_gen(dev);
1889
1890         if (err)
1891                 return err;
1892         ip6_tnl_link_config(t);
1893         if (t->parms.collect_md)
1894                 netif_keep_dst(dev);
1895         return 0;
1896 }
1897
1898 /**
1899  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1900  *   @dev: fallback device
1901  *
1902  * Return: 0
1903  **/
1904
1905 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1906 {
1907         struct ip6_tnl *t = netdev_priv(dev);
1908         struct net *net = dev_net(dev);
1909         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1910
1911         t->parms.proto = IPPROTO_IPV6;
1912
1913         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1914         return 0;
1915 }
1916
1917 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1918                             struct netlink_ext_ack *extack)
1919 {
1920         u8 proto;
1921
1922         if (!data || !data[IFLA_IPTUN_PROTO])
1923                 return 0;
1924
1925         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1926         if (proto != IPPROTO_IPV6 &&
1927             proto != IPPROTO_IPIP &&
1928             proto != 0)
1929                 return -EINVAL;
1930
1931         return 0;
1932 }
1933
1934 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1935                                   struct __ip6_tnl_parm *parms)
1936 {
1937         memset(parms, 0, sizeof(*parms));
1938
1939         if (!data)
1940                 return;
1941
1942         if (data[IFLA_IPTUN_LINK])
1943                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1944
1945         if (data[IFLA_IPTUN_LOCAL])
1946                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1947
1948         if (data[IFLA_IPTUN_REMOTE])
1949                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1950
1951         if (data[IFLA_IPTUN_TTL])
1952                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1953
1954         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1955                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1956
1957         if (data[IFLA_IPTUN_FLOWINFO])
1958                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1959
1960         if (data[IFLA_IPTUN_FLAGS])
1961                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1962
1963         if (data[IFLA_IPTUN_PROTO])
1964                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1965
1966         if (data[IFLA_IPTUN_COLLECT_METADATA])
1967                 parms->collect_md = true;
1968
1969         if (data[IFLA_IPTUN_FWMARK])
1970                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1971 }
1972
1973 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1974                                         struct ip_tunnel_encap *ipencap)
1975 {
1976         bool ret = false;
1977
1978         memset(ipencap, 0, sizeof(*ipencap));
1979
1980         if (!data)
1981                 return ret;
1982
1983         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1984                 ret = true;
1985                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1986         }
1987
1988         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1989                 ret = true;
1990                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1991         }
1992
1993         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1994                 ret = true;
1995                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1996         }
1997
1998         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1999                 ret = true;
2000                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2001         }
2002
2003         return ret;
2004 }
2005
2006 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2007                            struct nlattr *tb[], struct nlattr *data[],
2008                            struct netlink_ext_ack *extack)
2009 {
2010         struct net *net = dev_net(dev);
2011         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2012         struct ip_tunnel_encap ipencap;
2013         struct ip6_tnl *nt, *t;
2014         int err;
2015
2016         nt = netdev_priv(dev);
2017
2018         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2019                 err = ip6_tnl_encap_setup(nt, &ipencap);
2020                 if (err < 0)
2021                         return err;
2022         }
2023
2024         ip6_tnl_netlink_parms(data, &nt->parms);
2025
2026         if (nt->parms.collect_md) {
2027                 if (rtnl_dereference(ip6n->collect_md_tun))
2028                         return -EEXIST;
2029         } else {
2030                 t = ip6_tnl_locate(net, &nt->parms, 0);
2031                 if (!IS_ERR(t))
2032                         return -EEXIST;
2033         }
2034
2035         err = ip6_tnl_create2(dev);
2036         if (!err && tb[IFLA_MTU])
2037                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2038
2039         return err;
2040 }
2041
2042 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2043                               struct nlattr *data[],
2044                               struct netlink_ext_ack *extack)
2045 {
2046         struct ip6_tnl *t = netdev_priv(dev);
2047         struct __ip6_tnl_parm p;
2048         struct net *net = t->net;
2049         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2050         struct ip_tunnel_encap ipencap;
2051
2052         if (dev == ip6n->fb_tnl_dev)
2053                 return -EINVAL;
2054
2055         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2056                 int err = ip6_tnl_encap_setup(t, &ipencap);
2057
2058                 if (err < 0)
2059                         return err;
2060         }
2061         ip6_tnl_netlink_parms(data, &p);
2062         if (p.collect_md)
2063                 return -EINVAL;
2064
2065         t = ip6_tnl_locate(net, &p, 0);
2066         if (!IS_ERR(t)) {
2067                 if (t->dev != dev)
2068                         return -EEXIST;
2069         } else
2070                 t = netdev_priv(dev);
2071
2072         return ip6_tnl_update(t, &p);
2073 }
2074
2075 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2076 {
2077         struct net *net = dev_net(dev);
2078         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2079
2080         if (dev != ip6n->fb_tnl_dev)
2081                 unregister_netdevice_queue(dev, head);
2082 }
2083
2084 static size_t ip6_tnl_get_size(const struct net_device *dev)
2085 {
2086         return
2087                 /* IFLA_IPTUN_LINK */
2088                 nla_total_size(4) +
2089                 /* IFLA_IPTUN_LOCAL */
2090                 nla_total_size(sizeof(struct in6_addr)) +
2091                 /* IFLA_IPTUN_REMOTE */
2092                 nla_total_size(sizeof(struct in6_addr)) +
2093                 /* IFLA_IPTUN_TTL */
2094                 nla_total_size(1) +
2095                 /* IFLA_IPTUN_ENCAP_LIMIT */
2096                 nla_total_size(1) +
2097                 /* IFLA_IPTUN_FLOWINFO */
2098                 nla_total_size(4) +
2099                 /* IFLA_IPTUN_FLAGS */
2100                 nla_total_size(4) +
2101                 /* IFLA_IPTUN_PROTO */
2102                 nla_total_size(1) +
2103                 /* IFLA_IPTUN_ENCAP_TYPE */
2104                 nla_total_size(2) +
2105                 /* IFLA_IPTUN_ENCAP_FLAGS */
2106                 nla_total_size(2) +
2107                 /* IFLA_IPTUN_ENCAP_SPORT */
2108                 nla_total_size(2) +
2109                 /* IFLA_IPTUN_ENCAP_DPORT */
2110                 nla_total_size(2) +
2111                 /* IFLA_IPTUN_COLLECT_METADATA */
2112                 nla_total_size(0) +
2113                 /* IFLA_IPTUN_FWMARK */
2114                 nla_total_size(4) +
2115                 0;
2116 }
2117
2118 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2119 {
2120         struct ip6_tnl *tunnel = netdev_priv(dev);
2121         struct __ip6_tnl_parm *parm = &tunnel->parms;
2122
2123         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2124             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2125             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2126             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2127             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2128             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2129             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2130             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2131             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2132                 goto nla_put_failure;
2133
2134         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2135             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2136             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2137             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2138                 goto nla_put_failure;
2139
2140         if (parm->collect_md)
2141                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2142                         goto nla_put_failure;
2143
2144         return 0;
2145
2146 nla_put_failure:
2147         return -EMSGSIZE;
2148 }
2149
2150 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2151 {
2152         struct ip6_tnl *tunnel = netdev_priv(dev);
2153
2154         return tunnel->net;
2155 }
2156 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2157
2158 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2159         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2160         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2161         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2162         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2163         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2164         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2165         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2166         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2167         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2168         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2169         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2170         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2171         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2172         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2173 };
2174
2175 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2176         .kind           = "ip6tnl",
2177         .maxtype        = IFLA_IPTUN_MAX,
2178         .policy         = ip6_tnl_policy,
2179         .priv_size      = sizeof(struct ip6_tnl),
2180         .setup          = ip6_tnl_dev_setup,
2181         .validate       = ip6_tnl_validate,
2182         .newlink        = ip6_tnl_newlink,
2183         .changelink     = ip6_tnl_changelink,
2184         .dellink        = ip6_tnl_dellink,
2185         .get_size       = ip6_tnl_get_size,
2186         .fill_info      = ip6_tnl_fill_info,
2187         .get_link_net   = ip6_tnl_get_link_net,
2188 };
2189
2190 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2191         .handler        = ip4ip6_rcv,
2192         .err_handler    = ip4ip6_err,
2193         .priority       =       1,
2194 };
2195
2196 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2197         .handler        = ip6ip6_rcv,
2198         .err_handler    = ip6ip6_err,
2199         .priority       =       1,
2200 };
2201
2202 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2203 {
2204         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2205         struct net_device *dev, *aux;
2206         int h;
2207         struct ip6_tnl *t;
2208
2209         for_each_netdev_safe(net, dev, aux)
2210                 if (dev->rtnl_link_ops == &ip6_link_ops)
2211                         unregister_netdevice_queue(dev, list);
2212
2213         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2214                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2215                 while (t) {
2216                         /* If dev is in the same netns, it has already
2217                          * been added to the list by the previous loop.
2218                          */
2219                         if (!net_eq(dev_net(t->dev), net))
2220                                 unregister_netdevice_queue(t->dev, list);
2221                         t = rtnl_dereference(t->next);
2222                 }
2223         }
2224
2225         t = rtnl_dereference(ip6n->tnls_wc[0]);
2226         while (t) {
2227                 /* If dev is in the same netns, it has already
2228                  * been added to the list by the previous loop.
2229                  */
2230                 if (!net_eq(dev_net(t->dev), net))
2231                         unregister_netdevice_queue(t->dev, list);
2232                 t = rtnl_dereference(t->next);
2233         }
2234 }
2235
2236 static int __net_init ip6_tnl_init_net(struct net *net)
2237 {
2238         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2239         struct ip6_tnl *t = NULL;
2240         int err;
2241
2242         ip6n->tnls[0] = ip6n->tnls_wc;
2243         ip6n->tnls[1] = ip6n->tnls_r_l;
2244
2245         if (!net_has_fallback_tunnels(net))
2246                 return 0;
2247         err = -ENOMEM;
2248         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2249                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2250
2251         if (!ip6n->fb_tnl_dev)
2252                 goto err_alloc_dev;
2253         dev_net_set(ip6n->fb_tnl_dev, net);
2254         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2255         /* FB netdevice is special: we have one, and only one per netns.
2256          * Allowing to move it to another netns is clearly unsafe.
2257          */
2258         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2259
2260         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2261         if (err < 0)
2262                 goto err_register;
2263
2264         err = register_netdev(ip6n->fb_tnl_dev);
2265         if (err < 0)
2266                 goto err_register;
2267
2268         t = netdev_priv(ip6n->fb_tnl_dev);
2269
2270         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2271         return 0;
2272
2273 err_register:
2274         free_netdev(ip6n->fb_tnl_dev);
2275 err_alloc_dev:
2276         return err;
2277 }
2278
2279 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2280 {
2281         struct net *net;
2282         LIST_HEAD(list);
2283
2284         rtnl_lock();
2285         list_for_each_entry(net, net_list, exit_list)
2286                 ip6_tnl_destroy_tunnels(net, &list);
2287         unregister_netdevice_many(&list);
2288         rtnl_unlock();
2289 }
2290
2291 static struct pernet_operations ip6_tnl_net_ops = {
2292         .init = ip6_tnl_init_net,
2293         .exit_batch = ip6_tnl_exit_batch_net,
2294         .id   = &ip6_tnl_net_id,
2295         .size = sizeof(struct ip6_tnl_net),
2296 };
2297
2298 /**
2299  * ip6_tunnel_init - register protocol and reserve needed resources
2300  *
2301  * Return: 0 on success
2302  **/
2303
2304 static int __init ip6_tunnel_init(void)
2305 {
2306         int  err;
2307
2308         if (!ipv6_mod_enabled())
2309                 return -EOPNOTSUPP;
2310
2311         err = register_pernet_device(&ip6_tnl_net_ops);
2312         if (err < 0)
2313                 goto out_pernet;
2314
2315         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2316         if (err < 0) {
2317                 pr_err("%s: can't register ip4ip6\n", __func__);
2318                 goto out_ip4ip6;
2319         }
2320
2321         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2322         if (err < 0) {
2323                 pr_err("%s: can't register ip6ip6\n", __func__);
2324                 goto out_ip6ip6;
2325         }
2326         err = rtnl_link_register(&ip6_link_ops);
2327         if (err < 0)
2328                 goto rtnl_link_failed;
2329
2330         return 0;
2331
2332 rtnl_link_failed:
2333         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2334 out_ip6ip6:
2335         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2336 out_ip4ip6:
2337         unregister_pernet_device(&ip6_tnl_net_ops);
2338 out_pernet:
2339         return err;
2340 }
2341
2342 /**
2343  * ip6_tunnel_cleanup - free resources and unregister protocol
2344  **/
2345
2346 static void __exit ip6_tunnel_cleanup(void)
2347 {
2348         rtnl_link_unregister(&ip6_link_ops);
2349         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2350                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2351
2352         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2353                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2354
2355         unregister_pernet_device(&ip6_tnl_net_ops);
2356 }
2357
2358 module_init(ip6_tunnel_init);
2359 module_exit(ip6_tunnel_cleanup);