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 }
 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         dev_hold(dev);
 276         ip6_tnl_link(ip6n, t);
 277         return 0;
 278
 279 out:
 280         return err;
 281 }
 282
 283 /**
 284  * ip6_tnl_create - create a new tunnel
 285  *   @p: tunnel parameters
 286  *   @pt: pointer to new tunnel
 287  *
 288  * Description:
 289  *   Create tunnel matching given parameters.
 290  *
 291  * Return:
 292  *   created tunnel or error pointer
 293  **/
 294
 295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 296 {
 297         struct net_device *dev;
 298         struct ip6_tnl *t;
 299         char name[IFNAMSIZ];
 300         int err = -ENOMEM;
 301
 302         if (p->name[0])
 303                 strlcpy(name, p->name, IFNAMSIZ);
 304         else
 305                 sprintf(name, "ip6tnl%%d");
 306
 307         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 308                            ip6_tnl_dev_setup);
 309         if (!dev)
 310                 goto failed;
 311
 312         dev_net_set(dev, net);
 313
 314         t = netdev_priv(dev);
 315         t->parms = *p;
 316         t->net = dev_net(dev);
 317         err = ip6_tnl_create2(dev);
 318         if (err < 0)
 319                 goto failed_free;
 320
 321         return t;
 322
 323 failed_free:
 324         free_netdev(dev);
 325 failed:
 326         return ERR_PTR(err);
 327 }
 328
 329 /**
 330  * ip6_tnl_locate - find or create tunnel matching given parameters
 331  *   @p: tunnel parameters
 332  *   @create: != 0 if allowed to create new tunnel if no match found
 333  *
 334  * Description:
 335  *   ip6_tnl_locate() first tries to locate an existing tunnel
 336  *   based on @parms. If this is unsuccessful, but @create is set a new
 337  *   tunnel device is created and registered for use.
 338  *
 339  * Return:
 340  *   matching tunnel or error pointer
 341  **/
 342
 343 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 344                 struct __ip6_tnl_parm *p, int create)
 345 {
 346         const struct in6_addr *remote = &p->raddr;
 347         const struct in6_addr *local = &p->laddr;
 348         struct ip6_tnl __rcu **tp;
 349         struct ip6_tnl *t;
 350         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 351
 352         for (tp = ip6_tnl_bucket(ip6n, p);
 353              (t = rtnl_dereference(*tp)) != NULL;
 354              tp = &t->next) {
 355                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 356                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 357                         if (create)
 358                                 return ERR_PTR(-EEXIST);
 359
 360                         return t;
 361                 }
 362         }
 363         if (!create)
 364                 return ERR_PTR(-ENODEV);
 365         return ip6_tnl_create(net, p);
 366 }
 367
 368 /**
 369  * ip6_tnl_dev_uninit - tunnel device uninitializer
 370  *   @dev: the device to be destroyed
 371  *
 372  * Description:
 373  *   ip6_tnl_dev_uninit() removes tunnel from its list
 374  **/
 375
 376 static void
 377 ip6_tnl_dev_uninit(struct net_device *dev)
 378 {
 379         struct ip6_tnl *t = netdev_priv(dev);
 380         struct net *net = t->net;
 381         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 382
 383         if (dev == ip6n->fb_tnl_dev)
 384                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 385         else
 386                 ip6_tnl_unlink(ip6n, t);
 387         dst_cache_reset(&t->dst_cache);
 388         dev_put(dev);
 389 }
 390
 391 /**
 392  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 393  *   @skb: received socket buffer
 394  *
 395  * Return:
 396  *   0 if none was found,
 397  *   else index to encapsulation limit
 398  **/
 399
 400 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 401 {
 402         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 403         unsigned int nhoff = raw - skb->data;
 404         unsigned int off = nhoff + sizeof(*ipv6h);
 405         u8 next, nexthdr = ipv6h->nexthdr;
 406
 407         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 408                 struct ipv6_opt_hdr *hdr;
 409                 u16 optlen;
 410
 411                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 412                         break;
 413
 414                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 415                 if (nexthdr == NEXTHDR_FRAGMENT) {
 416                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 417                         if (frag_hdr->frag_off)
 418                                 break;
 419                         optlen = 8;
 420                 } else if (nexthdr == NEXTHDR_AUTH) {
 421                         optlen = (hdr->hdrlen + 2) << 2;
 422                 } else {
 423                         optlen = ipv6_optlen(hdr);
 424                 }
 425                 /* cache hdr->nexthdr, since pskb_may_pull() might
 426                  * invalidate hdr
 427                  */
 428                 next = hdr->nexthdr;
 429                 if (nexthdr == NEXTHDR_DEST) {
 430                         u16 i = 2;
 431
 432                         /* Remember : hdr is no longer valid at this point. */
 433                         if (!pskb_may_pull(skb, off + optlen))
 434                                 break;
 435
 436                         while (1) {
 437                                 struct ipv6_tlv_tnl_enc_lim *tel;
 438
 439                                 /* No more room for encapsulation limit */
 440                                 if (i + sizeof(*tel) > optlen)
 441                                         break;
 442
 443                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 444                                 /* return index of option if found and valid */
 445                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 446                                     tel->length == 1)
 447                                         return i + off - nhoff;
 448                                 /* else jump to next option */
 449                                 if (tel->type)
 450                                         i += tel->length + 2;
 451                                 else
 452                                         i++;
 453                         }
 454                 }
 455                 nexthdr = next;
 456                 off += optlen;
 457         }
 458         return 0;
 459 }
 460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 461
 462 /**
 463  * ip6_tnl_err - tunnel error handler
 464  *
 465  * Description:
 466  *   ip6_tnl_err() should handle errors in the tunnel according
 467  *   to the specifications in RFC 2473.
 468  **/
 469
 470 static int
 471 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 472             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 473 {
 474         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 475         struct ip6_tnl *t;
 476         int rel_msg = 0;
 477         u8 rel_type = ICMPV6_DEST_UNREACH;
 478         u8 rel_code = ICMPV6_ADDR_UNREACH;
 479         u8 tproto;
 480         __u32 rel_info = 0;
 481         __u16 len;
 482         int err = -ENOENT;
 483
 484         /* If the packet doesn't contain the original IPv6 header we are
 485            in trouble since we might need the source address for further
 486            processing of the error. */
 487
 488         rcu_read_lock();
 489         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 490         if (!t)
 491                 goto out;
 492
 493         tproto = ACCESS_ONCE(t->parms.proto);
 494         if (tproto != ipproto && tproto != 0)
 495                 goto out;
 496
 497         err = 0;
 498
 499         switch (*type) {
 500                 __u32 teli;
 501                 struct ipv6_tlv_tnl_enc_lim *tel;
 502                 __u32 mtu;
 503         case ICMPV6_DEST_UNREACH:
 504                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 505                                     t->parms.name);
 506                 rel_msg = 1;
 507                 break;
 508         case ICMPV6_TIME_EXCEED:
 509                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 510                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 511                                             t->parms.name);
 512                         rel_msg = 1;
 513                 }
 514                 break;
 515         case ICMPV6_PARAMPROB:
 516                 teli = 0;
 517                 if ((*code) == ICMPV6_HDR_FIELD)
 518                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 519
 520                 if (teli && teli == *info - 2) {
 521                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 522                         if (tel->encap_limit == 0) {
 523                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 524                                                     t->parms.name);
 525                                 rel_msg = 1;
 526                         }
 527                 } else {
 528                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 529                                             t->parms.name);
 530                 }
 531                 break;
 532         case ICMPV6_PKT_TOOBIG:
 533                 mtu = *info - offset;
 534                 if (mtu < IPV6_MIN_MTU)
 535                         mtu = IPV6_MIN_MTU;
 536                 t->dev->mtu = mtu;
 537
 538                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 539                 if (len > mtu) {
 540                         rel_type = ICMPV6_PKT_TOOBIG;
 541                         rel_code = 0;
 542                         rel_info = mtu;
 543                         rel_msg = 1;
 544                 }
 545                 break;
 546         }
 547
 548         *type = rel_type;
 549         *code = rel_code;
 550         *info = rel_info;
 551         *msg = rel_msg;
 552
 553 out:
 554         rcu_read_unlock();
 555         return err;
 556 }
 557
 558 static int
 559 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 560            u8 type, u8 code, int offset, __be32 info)
 561 {
 562         int rel_msg = 0;
 563         u8 rel_type = type;
 564         u8 rel_code = code;
 565         __u32 rel_info = ntohl(info);
 566         int err;
 567         struct sk_buff *skb2;
 568         const struct iphdr *eiph;
 569         struct rtable *rt;
 570         struct flowi4 fl4;
 571
 572         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 573                           &rel_msg, &rel_info, offset);
 574         if (err < 0)
 575                 return err;
 576
 577         if (rel_msg == 0)
 578                 return 0;
 579
 580         switch (rel_type) {
 581         case ICMPV6_DEST_UNREACH:
 582                 if (rel_code != ICMPV6_ADDR_UNREACH)
 583                         return 0;
 584                 rel_type = ICMP_DEST_UNREACH;
 585                 rel_code = ICMP_HOST_UNREACH;
 586                 break;
 587         case ICMPV6_PKT_TOOBIG:
 588                 if (rel_code != 0)
 589                         return 0;
 590                 rel_type = ICMP_DEST_UNREACH;
 591                 rel_code = ICMP_FRAG_NEEDED;
 592                 break;
 593         case NDISC_REDIRECT:
 594                 rel_type = ICMP_REDIRECT;
 595                 rel_code = ICMP_REDIR_HOST;
 596         default:
 597                 return 0;
 598         }
 599
 600         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 601                 return 0;
 602
 603         skb2 = skb_clone(skb, GFP_ATOMIC);
 604         if (!skb2)
 605                 return 0;
 606
 607         skb_dst_drop(skb2);
 608
 609         skb_pull(skb2, offset);
 610         skb_reset_network_header(skb2);
 611         eiph = ip_hdr(skb2);
 612
 613         /* Try to guess incoming interface */
 614         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 615                                    eiph->saddr, 0,
 616                                    0, 0,
 617                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 618         if (IS_ERR(rt))
 619                 goto out;
 620
 621         skb2->dev = rt->dst.dev;
 622
 623         /* route "incoming" packet */
 624         if (rt->rt_flags & RTCF_LOCAL) {
 625                 ip_rt_put(rt);
 626                 rt = NULL;
 627                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 628                                            eiph->daddr, eiph->saddr,
 629                                            0, 0,
 630                                            IPPROTO_IPIP,
 631                                            RT_TOS(eiph->tos), 0);
 632                 if (IS_ERR(rt) ||
 633                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 634                         if (!IS_ERR(rt))
 635                                 ip_rt_put(rt);
 636                         goto out;
 637                 }
 638                 skb_dst_set(skb2, &rt->dst);
 639         } else {
 640                 ip_rt_put(rt);
 641                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 642                                    skb2->dev) ||
 643                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 644                         goto out;
 645         }
 646
 647         /* change mtu on this route */
 648         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 649                 if (rel_info > dst_mtu(skb_dst(skb2)))
 650                         goto out;
 651
 652                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 653         }
 654         if (rel_type == ICMP_REDIRECT)
 655                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 656
 657         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 658
 659 out:
 660         kfree_skb(skb2);
 661         return 0;
 662 }
 663
 664 static int
 665 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 666            u8 type, u8 code, int offset, __be32 info)
 667 {
 668         int rel_msg = 0;
 669         u8 rel_type = type;
 670         u8 rel_code = code;
 671         __u32 rel_info = ntohl(info);
 672         int err;
 673
 674         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 675                           &rel_msg, &rel_info, offset);
 676         if (err < 0)
 677                 return err;
 678
 679         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 680                 struct rt6_info *rt;
 681                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 682
 683                 if (!skb2)
 684                         return 0;
 685
 686                 skb_dst_drop(skb2);
 687                 skb_pull(skb2, offset);
 688                 skb_reset_network_header(skb2);
 689
 690                 /* Try to guess incoming interface */
 691                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 692                                 NULL, 0, 0);
 693
 694                 if (rt && rt->dst.dev)
 695                         skb2->dev = rt->dst.dev;
 696
 697                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 698
 699                 ip6_rt_put(rt);
 700
 701                 kfree_skb(skb2);
 702         }
 703
 704         return 0;
 705 }
 706
 707 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 708                                        const struct ipv6hdr *ipv6h,
 709                                        struct sk_buff *skb)
 710 {
 711         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 712
 713         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 714                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 715
 716         return IP6_ECN_decapsulate(ipv6h, skb);
 717 }
 718
 719 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 720                                        const struct ipv6hdr *ipv6h,
 721                                        struct sk_buff *skb)
 722 {
 723         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 724                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 725
 726         return IP6_ECN_decapsulate(ipv6h, skb);
 727 }
 728
 729 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 730                              const struct in6_addr *laddr,
 731                              const struct in6_addr *raddr)
 732 {
 733         struct __ip6_tnl_parm *p = &t->parms;
 734         int ltype = ipv6_addr_type(laddr);
 735         int rtype = ipv6_addr_type(raddr);
 736         __u32 flags = 0;
 737
 738         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 739                 flags = IP6_TNL_F_CAP_PER_PACKET;
 740         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 741                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 742                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 743                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 744                 if (ltype&IPV6_ADDR_UNICAST)
 745                         flags |= IP6_TNL_F_CAP_XMIT;
 746                 if (rtype&IPV6_ADDR_UNICAST)
 747                         flags |= IP6_TNL_F_CAP_RCV;
 748         }
 749         return flags;
 750 }
 751 EXPORT_SYMBOL(ip6_tnl_get_cap);
 752
 753 /* called with rcu_read_lock() */
 754 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 755                                   const struct in6_addr *laddr,
 756                                   const struct in6_addr *raddr)
 757 {
 758         struct __ip6_tnl_parm *p = &t->parms;
 759         int ret = 0;
 760         struct net *net = t->net;
 761
 762         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 763             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 764              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 765                 struct net_device *ldev = NULL;
 766
 767                 if (p->link)
 768                         ldev = dev_get_by_index_rcu(net, p->link);
 769
 770                 if ((ipv6_addr_is_multicast(laddr) ||
 771                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 772                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 773                         ret = 1;
 774         }
 775         return ret;
 776 }
 777 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 778
 779 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 780                          const struct tnl_ptk_info *tpi,
 781                          struct metadata_dst *tun_dst,
 782                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 783                                                 const struct ipv6hdr *ipv6h,
 784                                                 struct sk_buff *skb),
 785                          bool log_ecn_err)
 786 {
 787         struct pcpu_sw_netstats *tstats;
 788         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 789         int err;
 790
 791         if ((!(tpi->flags & TUNNEL_CSUM) &&
 792              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 793             ((tpi->flags & TUNNEL_CSUM) &&
 794              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 795                 tunnel->dev->stats.rx_crc_errors++;
 796                 tunnel->dev->stats.rx_errors++;
 797                 goto drop;
 798         }
 799
 800         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 801                 if (!(tpi->flags & TUNNEL_SEQ) ||
 802                     (tunnel->i_seqno &&
 803                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 804                         tunnel->dev->stats.rx_fifo_errors++;
 805                         tunnel->dev->stats.rx_errors++;
 806                         goto drop;
 807                 }
 808                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 809         }
 810
 811         skb->protocol = tpi->proto;
 812
 813         /* Warning: All skb pointers will be invalidated! */
 814         if (tunnel->dev->type == ARPHRD_ETHER) {
 815                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 816                         tunnel->dev->stats.rx_length_errors++;
 817                         tunnel->dev->stats.rx_errors++;
 818                         goto drop;
 819                 }
 820
 821                 ipv6h = ipv6_hdr(skb);
 822                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 823                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 824         } else {
 825                 skb->dev = tunnel->dev;
 826         }
 827
 828         skb_reset_network_header(skb);
 829         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 830
 831         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 832
 833         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 834         if (unlikely(err)) {
 835                 if (log_ecn_err)
 836                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 837                                              &ipv6h->saddr,
 838                                              ipv6_get_dsfield(ipv6h));
 839                 if (err > 1) {
 840                         ++tunnel->dev->stats.rx_frame_errors;
 841                         ++tunnel->dev->stats.rx_errors;
 842                         goto drop;
 843                 }
 844         }
 845
 846         tstats = this_cpu_ptr(tunnel->dev->tstats);
 847         u64_stats_update_begin(&tstats->syncp);
 848         tstats->rx_packets++;
 849         tstats->rx_bytes += skb->len;
 850         u64_stats_update_end(&tstats->syncp);
 851
 852         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 853
 854         if (tun_dst)
 855                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 856
 857         gro_cells_receive(&tunnel->gro_cells, skb);
 858         return 0;
 859
 860 drop:
 861         kfree_skb(skb);
 862         return 0;
 863 }
 864
 865 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 866                 const struct tnl_ptk_info *tpi,
 867                 struct metadata_dst *tun_dst,
 868                 bool log_ecn_err)
 869 {
 870         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
 871                              log_ecn_err);
 872 }
 873 EXPORT_SYMBOL(ip6_tnl_rcv);
 874
 875 static const struct tnl_ptk_info tpi_v6 = {
 876         /* no tunnel info required for ipxip6. */
 877         .proto = htons(ETH_P_IPV6),
 878 };
 879
 880 static const struct tnl_ptk_info tpi_v4 = {
 881         /* no tunnel info required for ipxip6. */
 882         .proto = htons(ETH_P_IP),
 883 };
 884
 885 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 886                       const struct tnl_ptk_info *tpi,
 887                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 888                                                   const struct ipv6hdr *ipv6h,
 889                                                   struct sk_buff *skb))
 890 {
 891         struct ip6_tnl *t;
 892         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 893         struct metadata_dst *tun_dst = NULL;
 894         int ret = -1;
 895
 896         rcu_read_lock();
 897         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 898
 899         if (t) {
 900                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 901
 902                 if (tproto != ipproto && tproto != 0)
 903                         goto drop;
 904                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 905                         goto drop;
 906                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 907                         goto drop;
 908                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 909                         goto drop;
 910                 if (t->parms.collect_md) {
 911                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 912                         if (!tun_dst)
 913                                 return 0;
 914                 }
 915                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 916                                     log_ecn_error);
 917         }
 918
 919         rcu_read_unlock();
 920
 921         return ret;
 922
 923 drop:
 924         rcu_read_unlock();
 925         kfree_skb(skb);
 926         return 0;
 927 }
 928
 929 static int ip4ip6_rcv(struct sk_buff *skb)
 930 {
 931         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 932                           ip4ip6_dscp_ecn_decapsulate);
 933 }
 934
 935 static int ip6ip6_rcv(struct sk_buff *skb)
 936 {
 937         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 938                           ip6ip6_dscp_ecn_decapsulate);
 939 }
 940
 941 struct ipv6_tel_txoption {
 942         struct ipv6_txoptions ops;
 943         __u8 dst_opt[8];
 944 };
 945
 946 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 947 {
 948         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 949
 950         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 951         opt->dst_opt[3] = 1;
 952         opt->dst_opt[4] = encap_limit;
 953         opt->dst_opt[5] = IPV6_TLV_PADN;
 954         opt->dst_opt[6] = 1;
 955
 956         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 957         opt->ops.opt_nflen = 8;
 958 }
 959
 960 /**
 961  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 962  *   @t: the outgoing tunnel device
 963  *   @hdr: IPv6 header from the incoming packet
 964  *
 965  * Description:
 966  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 967  *   doesn't match source of incoming packet.
 968  *
 969  * Return:
 970  *   1 if conflict,
 971  *   0 else
 972  **/
 973
 974 static inline bool
 975 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 976 {
 977         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 978 }
 979
 980 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 981                      const struct in6_addr *laddr,
 982                      const struct in6_addr *raddr)
 983 {
 984         struct __ip6_tnl_parm *p = &t->parms;
 985         int ret = 0;
 986         struct net *net = t->net;
 987
 988         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 989             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 990              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 991                 struct net_device *ldev = NULL;
 992
 993                 rcu_read_lock();
 994                 if (p->link)
 995                         ldev = dev_get_by_index_rcu(net, p->link);
 996
 997                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 998                         pr_warn("%s xmit: Local address not yet configured!\n",
 999                                 p->name);
1000                 else if (!ipv6_addr_is_multicast(raddr) &&
1001                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1002                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1003                                 p->name);
1004                 else
1005                         ret = 1;
1006                 rcu_read_unlock();
1007         }
1008         return ret;
1009 }
1010 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1011
1012 /**
1013  * ip6_tnl_xmit - encapsulate packet and send
1014  *   @skb: the outgoing socket buffer
1015  *   @dev: the outgoing tunnel device
1016  *   @dsfield: dscp code for outer header
1017  *   @fl6: flow of tunneled packet
1018  *   @encap_limit: encapsulation limit
1019  *   @pmtu: Path MTU is stored if packet is too big
1020  *   @proto: next header value
1021  *
1022  * Description:
1023  *   Build new header and do some sanity checks on the packet before sending
1024  *   it.
1025  *
1026  * Return:
1027  *   0 on success
1028  *   -1 fail
1029  *   %-EMSGSIZE message too big. return mtu in this case.
1030  **/
1031
1032 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1033                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1034                  __u8 proto)
1035 {
1036         struct ip6_tnl *t = netdev_priv(dev);
1037         struct net *net = t->net;
1038         struct net_device_stats *stats = &t->dev->stats;
1039         struct ipv6hdr *ipv6h;
1040         struct ipv6_tel_txoption opt;
1041         struct dst_entry *dst = NULL, *ndst = NULL;
1042         struct net_device *tdev;
1043         int mtu;
1044         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1045         unsigned int max_headroom = psh_hlen;
1046         bool use_cache = false;
1047         u8 hop_limit;
1048         int err = -1;
1049
1050         if (t->parms.collect_md) {
1051                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1052                 goto route_lookup;
1053         } else {
1054                 hop_limit = t->parms.hop_limit;
1055         }
1056
1057         /* NBMA tunnel */
1058         if (ipv6_addr_any(&t->parms.raddr)) {
1059                 if (skb->protocol == htons(ETH_P_IPV6)) {
1060                         struct in6_addr *addr6;
1061                         struct neighbour *neigh;
1062                         int addr_type;
1063
1064                         if (!skb_dst(skb))
1065                                 goto tx_err_link_failure;
1066
1067                         neigh = dst_neigh_lookup(skb_dst(skb),
1068                                                  &ipv6_hdr(skb)->daddr);
1069                         if (!neigh)
1070                                 goto tx_err_link_failure;
1071
1072                         addr6 = (struct in6_addr *)&neigh->primary_key;
1073                         addr_type = ipv6_addr_type(addr6);
1074
1075                         if (addr_type == IPV6_ADDR_ANY)
1076                                 addr6 = &ipv6_hdr(skb)->daddr;
1077
1078                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1079                         neigh_release(neigh);
1080                 }
1081         } else if (!(t->parms.flags &
1082                      (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1083                 /* enable the cache only only if the routing decision does
1084                  * not depend on the current inner header value
1085                  */
1086                 use_cache = true;
1087         }
1088
1089         if (use_cache)
1090                 dst = dst_cache_get(&t->dst_cache);
1091
1092         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1093                 goto tx_err_link_failure;
1094
1095         if (!dst) {
1096 route_lookup:
1097                 /* add dsfield to flowlabel for route lookup */
1098                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1099
1100                 dst = ip6_route_output(net, NULL, fl6);
1101
1102                 if (dst->error)
1103                         goto tx_err_link_failure;
1104                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1105                 if (IS_ERR(dst)) {
1106                         err = PTR_ERR(dst);
1107                         dst = NULL;
1108                         goto tx_err_link_failure;
1109                 }
1110                 if (t->parms.collect_md &&
1111                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1112                                        &fl6->daddr, 0, &fl6->saddr))
1113                         goto tx_err_link_failure;
1114                 ndst = dst;
1115         }
1116
1117         tdev = dst->dev;
1118
1119         if (tdev == dev) {
1120                 stats->collisions++;
1121                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1122                                      t->parms.name);
1123                 goto tx_err_dst_release;
1124         }
1125         mtu = dst_mtu(dst) - psh_hlen - t->tun_hlen;
1126         if (encap_limit >= 0) {
1127                 max_headroom += 8;
1128                 mtu -= 8;
1129         }
1130         if (mtu < IPV6_MIN_MTU)
1131                 mtu = IPV6_MIN_MTU;
1132         if (skb_dst(skb) && !t->parms.collect_md)
1133                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1134         if (skb->len - t->tun_hlen > mtu && !skb_is_gso(skb)) {
1135                 *pmtu = mtu;
1136                 err = -EMSGSIZE;
1137                 goto tx_err_dst_release;
1138         }
1139
1140         if (t->err_count > 0) {
1141                 if (time_before(jiffies,
1142                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1143                         t->err_count--;
1144
1145                         dst_link_failure(skb);
1146                 } else {
1147                         t->err_count = 0;
1148                 }
1149         }
1150
1151         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1152
1153         /*
1154          * Okay, now see if we can stuff it in the buffer as-is.
1155          */
1156         max_headroom += LL_RESERVED_SPACE(tdev);
1157
1158         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1159             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1160                 struct sk_buff *new_skb;
1161
1162                 new_skb = skb_realloc_headroom(skb, max_headroom);
1163                 if (!new_skb)
1164                         goto tx_err_dst_release;
1165
1166                 if (skb->sk)
1167                         skb_set_owner_w(new_skb, skb->sk);
1168                 consume_skb(skb);
1169                 skb = new_skb;
1170         }
1171
1172         if (t->parms.collect_md) {
1173                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1174                         goto tx_err_dst_release;
1175         } else {
1176                 if (use_cache && ndst)
1177                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1178         }
1179         skb_dst_set(skb, dst);
1180
1181         if (encap_limit >= 0) {
1182                 init_tel_txopt(&opt, encap_limit);
1183                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1184         }
1185
1186         /* Calculate max headroom for all the headers and adjust
1187          * needed_headroom if necessary.
1188          */
1189         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1190                         + dst->header_len + t->hlen;
1191         if (max_headroom > dev->needed_headroom)
1192                 dev->needed_headroom = max_headroom;
1193
1194         err = ip6_tnl_encap(skb, t, &proto, fl6);
1195         if (err)
1196                 return err;
1197
1198         skb_push(skb, sizeof(struct ipv6hdr));
1199         skb_reset_network_header(skb);
1200         ipv6h = ipv6_hdr(skb);
1201         ip6_flow_hdr(ipv6h, dsfield,
1202                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1203         ipv6h->hop_limit = hop_limit;
1204         ipv6h->nexthdr = proto;
1205         ipv6h->saddr = fl6->saddr;
1206         ipv6h->daddr = fl6->daddr;
1207         ip6tunnel_xmit(NULL, skb, dev);
1208         return 0;
1209 tx_err_link_failure:
1210         stats->tx_carrier_errors++;
1211         dst_link_failure(skb);
1212 tx_err_dst_release:
1213         dst_release(dst);
1214         return err;
1215 }
1216 EXPORT_SYMBOL(ip6_tnl_xmit);
1217
1218 static inline int
1219 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1220 {
1221         struct ip6_tnl *t = netdev_priv(dev);
1222         const struct iphdr  *iph = ip_hdr(skb);
1223         int encap_limit = -1;
1224         struct flowi6 fl6;
1225         __u8 dsfield;
1226         __u32 mtu;
1227         u8 tproto;
1228         int err;
1229
1230         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1231
1232         tproto = ACCESS_ONCE(t->parms.proto);
1233         if (tproto != IPPROTO_IPIP && tproto != 0)
1234                 return -1;
1235
1236         if (t->parms.collect_md) {
1237                 struct ip_tunnel_info *tun_info;
1238                 const struct ip_tunnel_key *key;
1239
1240                 tun_info = skb_tunnel_info(skb);
1241                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1242                              ip_tunnel_info_af(tun_info) != AF_INET6))
1243                         return -1;
1244                 key = &tun_info->key;
1245                 memset(&fl6, 0, sizeof(fl6));
1246                 fl6.flowi6_proto = IPPROTO_IPIP;
1247                 fl6.daddr = key->u.ipv6.dst;
1248                 fl6.flowlabel = key->label;
1249                 dsfield = ip6_tclass(key->label);
1250         } else {
1251                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1252                         encap_limit = t->parms.encap_limit;
1253
1254                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1255                 fl6.flowi6_proto = IPPROTO_IPIP;
1256
1257                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1258                         dsfield = ipv4_get_dsfield(iph);
1259                 else
1260                         dsfield = ip6_tclass(t->parms.flowinfo);
1261                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1262                         fl6.flowi6_mark = skb->mark;
1263                 else
1264                         fl6.flowi6_mark = t->parms.fwmark;
1265         }
1266
1267         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1268
1269         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1270                 return -1;
1271
1272         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1273
1274         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1275
1276         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1277                            IPPROTO_IPIP);
1278         if (err != 0) {
1279                 /* XXX: send ICMP error even if DF is not set. */
1280                 if (err == -EMSGSIZE)
1281                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1282                                   htonl(mtu));
1283                 return -1;
1284         }
1285
1286         return 0;
1287 }
1288
1289 static inline int
1290 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1291 {
1292         struct ip6_tnl *t = netdev_priv(dev);
1293         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1294         int encap_limit = -1;
1295         __u16 offset;
1296         struct flowi6 fl6;
1297         __u8 dsfield;
1298         __u32 mtu;
1299         u8 tproto;
1300         int err;
1301
1302         tproto = ACCESS_ONCE(t->parms.proto);
1303         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1304             ip6_tnl_addr_conflict(t, ipv6h))
1305                 return -1;
1306
1307         if (t->parms.collect_md) {
1308                 struct ip_tunnel_info *tun_info;
1309                 const struct ip_tunnel_key *key;
1310
1311                 tun_info = skb_tunnel_info(skb);
1312                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1313                              ip_tunnel_info_af(tun_info) != AF_INET6))
1314                         return -1;
1315                 key = &tun_info->key;
1316                 memset(&fl6, 0, sizeof(fl6));
1317                 fl6.flowi6_proto = IPPROTO_IPV6;
1318                 fl6.daddr = key->u.ipv6.dst;
1319                 fl6.flowlabel = key->label;
1320                 dsfield = ip6_tclass(key->label);
1321         } else {
1322                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1323                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1324                 ipv6h = ipv6_hdr(skb);
1325                 if (offset > 0) {
1326                         struct ipv6_tlv_tnl_enc_lim *tel;
1327
1328                         tel = (void *)&skb_network_header(skb)[offset];
1329                         if (tel->encap_limit == 0) {
1330                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1331                                             ICMPV6_HDR_FIELD, offset + 2);
1332                                 return -1;
1333                         }
1334                         encap_limit = tel->encap_limit - 1;
1335                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1336                         encap_limit = t->parms.encap_limit;
1337                 }
1338
1339                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1340                 fl6.flowi6_proto = IPPROTO_IPV6;
1341
1342                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1343                         dsfield = ipv6_get_dsfield(ipv6h);
1344                 else
1345                         dsfield = ip6_tclass(t->parms.flowinfo);
1346                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1347                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1348                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1349                         fl6.flowi6_mark = skb->mark;
1350                 else
1351                         fl6.flowi6_mark = t->parms.fwmark;
1352         }
1353
1354         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1355
1356         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1357                 return -1;
1358
1359         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1360
1361         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1362
1363         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1364                            IPPROTO_IPV6);
1365         if (err != 0) {
1366                 if (err == -EMSGSIZE)
1367                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1368                 return -1;
1369         }
1370
1371         return 0;
1372 }
1373
1374 static netdev_tx_t
1375 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1376 {
1377         struct ip6_tnl *t = netdev_priv(dev);
1378         struct net_device_stats *stats = &t->dev->stats;
1379         int ret;
1380
1381         switch (skb->protocol) {
1382         case htons(ETH_P_IP):
1383                 ret = ip4ip6_tnl_xmit(skb, dev);
1384                 break;
1385         case htons(ETH_P_IPV6):
1386                 ret = ip6ip6_tnl_xmit(skb, dev);
1387                 break;
1388         default:
1389                 goto tx_err;
1390         }
1391
1392         if (ret < 0)
1393                 goto tx_err;
1394
1395         return NETDEV_TX_OK;
1396
1397 tx_err:
1398         stats->tx_errors++;
1399         stats->tx_dropped++;
1400         kfree_skb(skb);
1401         return NETDEV_TX_OK;
1402 }
1403
1404 static void ip6_tnl_link_config(struct ip6_tnl *t)
1405 {
1406         struct net_device *dev = t->dev;
1407         struct __ip6_tnl_parm *p = &t->parms;
1408         struct flowi6 *fl6 = &t->fl.u.ip6;
1409         int t_hlen;
1410
1411         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1412         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1413
1414         /* Set up flowi template */
1415         fl6->saddr = p->laddr;
1416         fl6->daddr = p->raddr;
1417         fl6->flowi6_oif = p->link;
1418         fl6->flowlabel = 0;
1419
1420         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1421                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1422         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1423                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1424
1425         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1426         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1427
1428         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1429                 dev->flags |= IFF_POINTOPOINT;
1430         else
1431                 dev->flags &= ~IFF_POINTOPOINT;
1432
1433         t->tun_hlen = 0;
1434         t->hlen = t->encap_hlen + t->tun_hlen;
1435         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1436
1437         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1438                 int strict = (ipv6_addr_type(&p->raddr) &
1439                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1440
1441                 struct rt6_info *rt = rt6_lookup(t->net,
1442                                                  &p->raddr, &p->laddr,
1443                                                  p->link, strict);
1444
1445                 if (!rt)
1446                         return;
1447
1448                 if (rt->dst.dev) {
1449                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1450                                 t_hlen;
1451
1452                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1453                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1454                                 dev->mtu -= 8;
1455
1456                         if (dev->mtu < IPV6_MIN_MTU)
1457                                 dev->mtu = IPV6_MIN_MTU;
1458                 }
1459                 ip6_rt_put(rt);
1460         }
1461 }
1462
1463 /**
1464  * ip6_tnl_change - update the tunnel parameters
1465  *   @t: tunnel to be changed
1466  *   @p: tunnel configuration parameters
1467  *
1468  * Description:
1469  *   ip6_tnl_change() updates the tunnel parameters
1470  **/
1471
1472 static int
1473 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1474 {
1475         t->parms.laddr = p->laddr;
1476         t->parms.raddr = p->raddr;
1477         t->parms.flags = p->flags;
1478         t->parms.hop_limit = p->hop_limit;
1479         t->parms.encap_limit = p->encap_limit;
1480         t->parms.flowinfo = p->flowinfo;
1481         t->parms.link = p->link;
1482         t->parms.proto = p->proto;
1483         t->parms.fwmark = p->fwmark;
1484         dst_cache_reset(&t->dst_cache);
1485         ip6_tnl_link_config(t);
1486         return 0;
1487 }
1488
1489 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1490 {
1491         struct net *net = t->net;
1492         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1493         int err;
1494
1495         ip6_tnl_unlink(ip6n, t);
1496         synchronize_net();
1497         err = ip6_tnl_change(t, p);
1498         ip6_tnl_link(ip6n, t);
1499         netdev_state_change(t->dev);
1500         return err;
1501 }
1502
1503 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1504 {
1505         /* for default tnl0 device allow to change only the proto */
1506         t->parms.proto = p->proto;
1507         netdev_state_change(t->dev);
1508         return 0;
1509 }
1510
1511 static void
1512 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1513 {
1514         p->laddr = u->laddr;
1515         p->raddr = u->raddr;
1516         p->flags = u->flags;
1517         p->hop_limit = u->hop_limit;
1518         p->encap_limit = u->encap_limit;
1519         p->flowinfo = u->flowinfo;
1520         p->link = u->link;
1521         p->proto = u->proto;
1522         memcpy(p->name, u->name, sizeof(u->name));
1523 }
1524
1525 static void
1526 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1527 {
1528         u->laddr = p->laddr;
1529         u->raddr = p->raddr;
1530         u->flags = p->flags;
1531         u->hop_limit = p->hop_limit;
1532         u->encap_limit = p->encap_limit;
1533         u->flowinfo = p->flowinfo;
1534         u->link = p->link;
1535         u->proto = p->proto;
1536         memcpy(u->name, p->name, sizeof(u->name));
1537 }
1538
1539 /**
1540  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1541  *   @dev: virtual device associated with tunnel
1542  *   @ifr: parameters passed from userspace
1543  *   @cmd: command to be performed
1544  *
1545  * Description:
1546  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1547  *   from userspace.
1548  *
1549  *   The possible commands are the following:
1550  *     %SIOCGETTUNNEL: get tunnel parameters for device
1551  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1552  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1553  *     %SIOCDELTUNNEL: delete tunnel
1554  *
1555  *   The fallback device "ip6tnl0", created during module
1556  *   initialization, can be used for creating other tunnel devices.
1557  *
1558  * Return:
1559  *   0 on success,
1560  *   %-EFAULT if unable to copy data to or from userspace,
1561  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1562  *   %-EINVAL if passed tunnel parameters are invalid,
1563  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1564  *   %-ENODEV if attempting to change or delete a nonexisting device
1565  **/
1566
1567 static int
1568 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1569 {
1570         int err = 0;
1571         struct ip6_tnl_parm p;
1572         struct __ip6_tnl_parm p1;
1573         struct ip6_tnl *t = netdev_priv(dev);
1574         struct net *net = t->net;
1575         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1576
1577         memset(&p1, 0, sizeof(p1));
1578
1579         switch (cmd) {
1580         case SIOCGETTUNNEL:
1581                 if (dev == ip6n->fb_tnl_dev) {
1582                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1583                                 err = -EFAULT;
1584                                 break;
1585                         }
1586                         ip6_tnl_parm_from_user(&p1, &p);
1587                         t = ip6_tnl_locate(net, &p1, 0);
1588                         if (IS_ERR(t))
1589                                 t = netdev_priv(dev);
1590                 } else {
1591                         memset(&p, 0, sizeof(p));
1592                 }
1593                 ip6_tnl_parm_to_user(&p, &t->parms);
1594                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1595                         err = -EFAULT;
1596                 }
1597                 break;
1598         case SIOCADDTUNNEL:
1599         case SIOCCHGTUNNEL:
1600                 err = -EPERM;
1601                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1602                         break;
1603                 err = -EFAULT;
1604                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1605                         break;
1606                 err = -EINVAL;
1607                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1608                     p.proto != 0)
1609                         break;
1610                 ip6_tnl_parm_from_user(&p1, &p);
1611                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1612                 if (cmd == SIOCCHGTUNNEL) {
1613                         if (!IS_ERR(t)) {
1614                                 if (t->dev != dev) {
1615                                         err = -EEXIST;
1616                                         break;
1617                                 }
1618                         } else
1619                                 t = netdev_priv(dev);
1620                         if (dev == ip6n->fb_tnl_dev)
1621                                 err = ip6_tnl0_update(t, &p1);
1622                         else
1623                                 err = ip6_tnl_update(t, &p1);
1624                 }
1625                 if (!IS_ERR(t)) {
1626                         err = 0;
1627                         ip6_tnl_parm_to_user(&p, &t->parms);
1628                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1629                                 err = -EFAULT;
1630
1631                 } else {
1632                         err = PTR_ERR(t);
1633                 }
1634                 break;
1635         case SIOCDELTUNNEL:
1636                 err = -EPERM;
1637                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1638                         break;
1639
1640                 if (dev == ip6n->fb_tnl_dev) {
1641                         err = -EFAULT;
1642                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1643                                 break;
1644                         err = -ENOENT;
1645                         ip6_tnl_parm_from_user(&p1, &p);
1646                         t = ip6_tnl_locate(net, &p1, 0);
1647                         if (IS_ERR(t))
1648                                 break;
1649                         err = -EPERM;
1650                         if (t->dev == ip6n->fb_tnl_dev)
1651                                 break;
1652                         dev = t->dev;
1653                 }
1654                 err = 0;
1655                 unregister_netdevice(dev);
1656                 break;
1657         default:
1658                 err = -EINVAL;
1659         }
1660         return err;
1661 }
1662
1663 /**
1664  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1665  *   @dev: virtual device associated with tunnel
1666  *   @new_mtu: the new mtu
1667  *
1668  * Return:
1669  *   0 on success,
1670  *   %-EINVAL if mtu too small
1671  **/
1672
1673 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1674 {
1675         struct ip6_tnl *tnl = netdev_priv(dev);
1676
1677         if (tnl->parms.proto == IPPROTO_IPIP) {
1678                 if (new_mtu < ETH_MIN_MTU)
1679                         return -EINVAL;
1680         } else {
1681                 if (new_mtu < IPV6_MIN_MTU)
1682                         return -EINVAL;
1683         }
1684         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1685                 return -EINVAL;
1686         dev->mtu = new_mtu;
1687         return 0;
1688 }
1689 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1690
1691 int ip6_tnl_get_iflink(const struct net_device *dev)
1692 {
1693         struct ip6_tnl *t = netdev_priv(dev);
1694
1695         return t->parms.link;
1696 }
1697 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1698
1699 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1700                           unsigned int num)
1701 {
1702         if (num >= MAX_IPTUN_ENCAP_OPS)
1703                 return -ERANGE;
1704
1705         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1706                         &ip6tun_encaps[num],
1707                         NULL, ops) ? 0 : -1;
1708 }
1709 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1710
1711 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1712                           unsigned int num)
1713 {
1714         int ret;
1715
1716         if (num >= MAX_IPTUN_ENCAP_OPS)
1717                 return -ERANGE;
1718
1719         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1720                        &ip6tun_encaps[num],
1721                        ops, NULL) == ops) ? 0 : -1;
1722
1723         synchronize_net();
1724
1725         return ret;
1726 }
1727 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1728
1729 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1730                         struct ip_tunnel_encap *ipencap)
1731 {
1732         int hlen;
1733
1734         memset(&t->encap, 0, sizeof(t->encap));
1735
1736         hlen = ip6_encap_hlen(ipencap);
1737         if (hlen < 0)
1738                 return hlen;
1739
1740         t->encap.type = ipencap->type;
1741         t->encap.sport = ipencap->sport;
1742         t->encap.dport = ipencap->dport;
1743         t->encap.flags = ipencap->flags;
1744
1745         t->encap_hlen = hlen;
1746         t->hlen = t->encap_hlen + t->tun_hlen;
1747
1748         return 0;
1749 }
1750 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1751
1752 static const struct net_device_ops ip6_tnl_netdev_ops = {
1753         .ndo_init       = ip6_tnl_dev_init,
1754         .ndo_uninit     = ip6_tnl_dev_uninit,
1755         .ndo_start_xmit = ip6_tnl_start_xmit,
1756         .ndo_do_ioctl   = ip6_tnl_ioctl,
1757         .ndo_change_mtu = ip6_tnl_change_mtu,
1758         .ndo_get_stats  = ip6_get_stats,
1759         .ndo_get_iflink = ip6_tnl_get_iflink,
1760 };
1761
1762 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1763                          NETIF_F_FRAGLIST |     \
1764                          NETIF_F_HIGHDMA |      \
1765                          NETIF_F_GSO_SOFTWARE | \
1766                          NETIF_F_HW_CSUM)
1767
1768 /**
1769  * ip6_tnl_dev_setup - setup virtual tunnel device
1770  *   @dev: virtual device associated with tunnel
1771  *
1772  * Description:
1773  *   Initialize function pointers and device parameters
1774  **/
1775
1776 static void ip6_tnl_dev_setup(struct net_device *dev)
1777 {
1778         dev->netdev_ops = &ip6_tnl_netdev_ops;
1779         dev->needs_free_netdev = true;
1780         dev->priv_destructor = ip6_dev_free;
1781
1782         dev->type = ARPHRD_TUNNEL6;
1783         dev->flags |= IFF_NOARP;
1784         dev->addr_len = sizeof(struct in6_addr);
1785         dev->features |= NETIF_F_LLTX;
1786         netif_keep_dst(dev);
1787
1788         dev->features           |= IPXIPX_FEATURES;
1789         dev->hw_features        |= IPXIPX_FEATURES;
1790
1791         /* This perm addr will be used as interface identifier by IPv6 */
1792         dev->addr_assign_type = NET_ADDR_RANDOM;
1793         eth_random_addr(dev->perm_addr);
1794 }
1795
1796
1797 /**
1798  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1799  *   @dev: virtual device associated with tunnel
1800  **/
1801
1802 static inline int
1803 ip6_tnl_dev_init_gen(struct net_device *dev)
1804 {
1805         struct ip6_tnl *t = netdev_priv(dev);
1806         int ret;
1807         int t_hlen;
1808
1809         t->dev = dev;
1810         t->net = dev_net(dev);
1811         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1812         if (!dev->tstats)
1813                 return -ENOMEM;
1814
1815         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1816         if (ret)
1817                 goto free_stats;
1818
1819         ret = gro_cells_init(&t->gro_cells, dev);
1820         if (ret)
1821                 goto destroy_dst;
1822
1823         t->tun_hlen = 0;
1824         t->hlen = t->encap_hlen + t->tun_hlen;
1825         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1826
1827         dev->type = ARPHRD_TUNNEL6;
1828         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1829         dev->mtu = ETH_DATA_LEN - t_hlen;
1830         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1831                 dev->mtu -= 8;
1832         dev->min_mtu = ETH_MIN_MTU;
1833         dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1834
1835         return 0;
1836
1837 destroy_dst:
1838         dst_cache_destroy(&t->dst_cache);
1839 free_stats:
1840         free_percpu(dev->tstats);
1841         dev->tstats = NULL;
1842
1843         return ret;
1844 }
1845
1846 /**
1847  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1848  *   @dev: virtual device associated with tunnel
1849  **/
1850
1851 static int ip6_tnl_dev_init(struct net_device *dev)
1852 {
1853         struct ip6_tnl *t = netdev_priv(dev);
1854         int err = ip6_tnl_dev_init_gen(dev);
1855
1856         if (err)
1857                 return err;
1858         ip6_tnl_link_config(t);
1859         if (t->parms.collect_md) {
1860                 dev->features |= NETIF_F_NETNS_LOCAL;
1861                 netif_keep_dst(dev);
1862         }
1863         return 0;
1864 }
1865
1866 /**
1867  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1868  *   @dev: fallback device
1869  *
1870  * Return: 0
1871  **/
1872
1873 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1874 {
1875         struct ip6_tnl *t = netdev_priv(dev);
1876         struct net *net = dev_net(dev);
1877         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1878
1879         t->parms.proto = IPPROTO_IPV6;
1880         dev_hold(dev);
1881
1882         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1883         return 0;
1884 }
1885
1886 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1887 {
1888         u8 proto;
1889
1890         if (!data || !data[IFLA_IPTUN_PROTO])
1891                 return 0;
1892
1893         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1894         if (proto != IPPROTO_IPV6 &&
1895             proto != IPPROTO_IPIP &&
1896             proto != 0)
1897                 return -EINVAL;
1898
1899         return 0;
1900 }
1901
1902 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1903                                   struct __ip6_tnl_parm *parms)
1904 {
1905         memset(parms, 0, sizeof(*parms));
1906
1907         if (!data)
1908                 return;
1909
1910         if (data[IFLA_IPTUN_LINK])
1911                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1912
1913         if (data[IFLA_IPTUN_LOCAL])
1914                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1915
1916         if (data[IFLA_IPTUN_REMOTE])
1917                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1918
1919         if (data[IFLA_IPTUN_TTL])
1920                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1921
1922         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1923                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1924
1925         if (data[IFLA_IPTUN_FLOWINFO])
1926                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1927
1928         if (data[IFLA_IPTUN_FLAGS])
1929                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1930
1931         if (data[IFLA_IPTUN_PROTO])
1932                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1933
1934         if (data[IFLA_IPTUN_COLLECT_METADATA])
1935                 parms->collect_md = true;
1936
1937         if (data[IFLA_IPTUN_FWMARK])
1938                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1939 }
1940
1941 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1942                                         struct ip_tunnel_encap *ipencap)
1943 {
1944         bool ret = false;
1945
1946         memset(ipencap, 0, sizeof(*ipencap));
1947
1948         if (!data)
1949                 return ret;
1950
1951         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1952                 ret = true;
1953                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1954         }
1955
1956         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1957                 ret = true;
1958                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1959         }
1960
1961         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1962                 ret = true;
1963                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1964         }
1965
1966         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1967                 ret = true;
1968                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1969         }
1970
1971         return ret;
1972 }
1973
1974 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1975                            struct nlattr *tb[], struct nlattr *data[])
1976 {
1977         struct net *net = dev_net(dev);
1978         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1979         struct ip6_tnl *nt, *t;
1980         struct ip_tunnel_encap ipencap;
1981
1982         nt = netdev_priv(dev);
1983
1984         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1985                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1986
1987                 if (err < 0)
1988                         return err;
1989         }
1990
1991         ip6_tnl_netlink_parms(data, &nt->parms);
1992
1993         if (nt->parms.collect_md) {
1994                 if (rtnl_dereference(ip6n->collect_md_tun))
1995                         return -EEXIST;
1996         } else {
1997                 t = ip6_tnl_locate(net, &nt->parms, 0);
1998                 if (!IS_ERR(t))
1999                         return -EEXIST;
2000         }
2001
2002         return ip6_tnl_create2(dev);
2003 }
2004
2005 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2006                               struct nlattr *data[])
2007 {
2008         struct ip6_tnl *t = netdev_priv(dev);
2009         struct __ip6_tnl_parm p;
2010         struct net *net = t->net;
2011         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2012         struct ip_tunnel_encap ipencap;
2013
2014         if (dev == ip6n->fb_tnl_dev)
2015                 return -EINVAL;
2016
2017         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2018                 int err = ip6_tnl_encap_setup(t, &ipencap);
2019
2020                 if (err < 0)
2021                         return err;
2022         }
2023         ip6_tnl_netlink_parms(data, &p);
2024         if (p.collect_md)
2025                 return -EINVAL;
2026
2027         t = ip6_tnl_locate(net, &p, 0);
2028         if (!IS_ERR(t)) {
2029                 if (t->dev != dev)
2030                         return -EEXIST;
2031         } else
2032                 t = netdev_priv(dev);
2033
2034         return ip6_tnl_update(t, &p);
2035 }
2036
2037 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2038 {
2039         struct net *net = dev_net(dev);
2040         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2041
2042         if (dev != ip6n->fb_tnl_dev)
2043                 unregister_netdevice_queue(dev, head);
2044 }
2045
2046 static size_t ip6_tnl_get_size(const struct net_device *dev)
2047 {
2048         return
2049                 /* IFLA_IPTUN_LINK */
2050                 nla_total_size(4) +
2051                 /* IFLA_IPTUN_LOCAL */
2052                 nla_total_size(sizeof(struct in6_addr)) +
2053                 /* IFLA_IPTUN_REMOTE */
2054                 nla_total_size(sizeof(struct in6_addr)) +
2055                 /* IFLA_IPTUN_TTL */
2056                 nla_total_size(1) +
2057                 /* IFLA_IPTUN_ENCAP_LIMIT */
2058                 nla_total_size(1) +
2059                 /* IFLA_IPTUN_FLOWINFO */
2060                 nla_total_size(4) +
2061                 /* IFLA_IPTUN_FLAGS */
2062                 nla_total_size(4) +
2063                 /* IFLA_IPTUN_PROTO */
2064                 nla_total_size(1) +
2065                 /* IFLA_IPTUN_ENCAP_TYPE */
2066                 nla_total_size(2) +
2067                 /* IFLA_IPTUN_ENCAP_FLAGS */
2068                 nla_total_size(2) +
2069                 /* IFLA_IPTUN_ENCAP_SPORT */
2070                 nla_total_size(2) +
2071                 /* IFLA_IPTUN_ENCAP_DPORT */
2072                 nla_total_size(2) +
2073                 /* IFLA_IPTUN_COLLECT_METADATA */
2074                 nla_total_size(0) +
2075                 /* IFLA_IPTUN_FWMARK */
2076                 nla_total_size(4) +
2077                 0;
2078 }
2079
2080 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2081 {
2082         struct ip6_tnl *tunnel = netdev_priv(dev);
2083         struct __ip6_tnl_parm *parm = &tunnel->parms;
2084
2085         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2086             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2087             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2088             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2089             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2090             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2091             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2092             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2093             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2094                 goto nla_put_failure;
2095
2096         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2097             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2098             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2099             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2100                 goto nla_put_failure;
2101
2102         if (parm->collect_md)
2103                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2104                         goto nla_put_failure;
2105
2106         return 0;
2107
2108 nla_put_failure:
2109         return -EMSGSIZE;
2110 }
2111
2112 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2113 {
2114         struct ip6_tnl *tunnel = netdev_priv(dev);
2115
2116         return tunnel->net;
2117 }
2118 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2119
2120 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2121         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2122         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2123         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2124         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2125         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2126         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2127         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2128         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2129         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2130         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2131         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2132         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2133         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2134         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2135 };
2136
2137 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2138         .kind           = "ip6tnl",
2139         .maxtype        = IFLA_IPTUN_MAX,
2140         .policy         = ip6_tnl_policy,
2141         .priv_size      = sizeof(struct ip6_tnl),
2142         .setup          = ip6_tnl_dev_setup,
2143         .validate       = ip6_tnl_validate,
2144         .newlink        = ip6_tnl_newlink,
2145         .changelink     = ip6_tnl_changelink,
2146         .dellink        = ip6_tnl_dellink,
2147         .get_size       = ip6_tnl_get_size,
2148         .fill_info      = ip6_tnl_fill_info,
2149         .get_link_net   = ip6_tnl_get_link_net,
2150 };
2151
2152 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2153         .handler        = ip4ip6_rcv,
2154         .err_handler    = ip4ip6_err,
2155         .priority       =       1,
2156 };
2157
2158 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2159         .handler        = ip6ip6_rcv,
2160         .err_handler    = ip6ip6_err,
2161         .priority       =       1,
2162 };
2163
2164 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2165 {
2166         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2167         struct net_device *dev, *aux;
2168         int h;
2169         struct ip6_tnl *t;
2170         LIST_HEAD(list);
2171
2172         for_each_netdev_safe(net, dev, aux)
2173                 if (dev->rtnl_link_ops == &ip6_link_ops)
2174                         unregister_netdevice_queue(dev, &list);
2175
2176         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2177                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2178                 while (t) {
2179                         /* If dev is in the same netns, it has already
2180                          * been added to the list by the previous loop.
2181                          */
2182                         if (!net_eq(dev_net(t->dev), net))
2183                                 unregister_netdevice_queue(t->dev, &list);
2184                         t = rtnl_dereference(t->next);
2185                 }
2186         }
2187
2188         unregister_netdevice_many(&list);
2189 }
2190
2191 static int __net_init ip6_tnl_init_net(struct net *net)
2192 {
2193         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2194         struct ip6_tnl *t = NULL;
2195         int err;
2196
2197         ip6n->tnls[0] = ip6n->tnls_wc;
2198         ip6n->tnls[1] = ip6n->tnls_r_l;
2199
2200         err = -ENOMEM;
2201         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2202                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2203
2204         if (!ip6n->fb_tnl_dev)
2205                 goto err_alloc_dev;
2206         dev_net_set(ip6n->fb_tnl_dev, net);
2207         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2208         /* FB netdevice is special: we have one, and only one per netns.
2209          * Allowing to move it to another netns is clearly unsafe.
2210          */
2211         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2212
2213         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2214         if (err < 0)
2215                 goto err_register;
2216
2217         err = register_netdev(ip6n->fb_tnl_dev);
2218         if (err < 0)
2219                 goto err_register;
2220
2221         t = netdev_priv(ip6n->fb_tnl_dev);
2222
2223         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2224         return 0;
2225
2226 err_register:
2227         free_netdev(ip6n->fb_tnl_dev);
2228 err_alloc_dev:
2229         return err;
2230 }
2231
2232 static void __net_exit ip6_tnl_exit_net(struct net *net)
2233 {
2234         rtnl_lock();
2235         ip6_tnl_destroy_tunnels(net);
2236         rtnl_unlock();
2237 }
2238
2239 static struct pernet_operations ip6_tnl_net_ops = {
2240         .init = ip6_tnl_init_net,
2241         .exit = ip6_tnl_exit_net,
2242         .id   = &ip6_tnl_net_id,
2243         .size = sizeof(struct ip6_tnl_net),
2244 };
2245
2246 /**
2247  * ip6_tunnel_init - register protocol and reserve needed resources
2248  *
2249  * Return: 0 on success
2250  **/
2251
2252 static int __init ip6_tunnel_init(void)
2253 {
2254         int  err;
2255
2256         err = register_pernet_device(&ip6_tnl_net_ops);
2257         if (err < 0)
2258                 goto out_pernet;
2259
2260         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2261         if (err < 0) {
2262                 pr_err("%s: can't register ip4ip6\n", __func__);
2263                 goto out_ip4ip6;
2264         }
2265
2266         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2267         if (err < 0) {
2268                 pr_err("%s: can't register ip6ip6\n", __func__);
2269                 goto out_ip6ip6;
2270         }
2271         err = rtnl_link_register(&ip6_link_ops);
2272         if (err < 0)
2273                 goto rtnl_link_failed;
2274
2275         return 0;
2276
2277 rtnl_link_failed:
2278         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2279 out_ip6ip6:
2280         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2281 out_ip4ip6:
2282         unregister_pernet_device(&ip6_tnl_net_ops);
2283 out_pernet:
2284         return err;
2285 }
2286
2287 /**
2288  * ip6_tunnel_cleanup - free resources and unregister protocol
2289  **/
2290
2291 static void __exit ip6_tunnel_cleanup(void)
2292 {
2293         rtnl_link_unregister(&ip6_link_ops);
2294         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2295                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2296
2297         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2298                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2299
2300         unregister_pernet_device(&ip6_tnl_net_ops);
2301 }
2302
2303 module_init(ip6_tunnel_init);
2304 module_exit(ip6_tunnel_cleanup);