net/ipv6/ip6_tunnel.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 tunneling device
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
   8  *      Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
   9  *
  10  *      Based on:
  11  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  12  *
  13  *      RFC 2473
  14  */
  15
  16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18 #include <linux/module.h>
  19 #include <linux/capability.h>
  20 #include <linux/errno.h>
  21 #include <linux/types.h>
  22 #include <linux/sockios.h>
  23 #include <linux/icmp.h>
  24 #include <linux/if.h>
  25 #include <linux/in.h>
  26 #include <linux/ip.h>
  27 #include <linux/net.h>
  28 #include <linux/in6.h>
  29 #include <linux/netdevice.h>
  30 #include <linux/if_arp.h>
  31 #include <linux/icmpv6.h>
  32 #include <linux/init.h>
  33 #include <linux/route.h>
  34 #include <linux/rtnetlink.h>
  35 #include <linux/netfilter_ipv6.h>
  36 #include <linux/slab.h>
  37 #include <linux/hash.h>
  38 #include <linux/etherdevice.h>
  39
  40 #include <linux/uaccess.h>
  41 #include <linux/atomic.h>
  42
  43 #include <net/icmp.h>
  44 #include <net/ip.h>
  45 #include <net/ip_tunnels.h>
  46 #include <net/ipv6.h>
  47 #include <net/ip6_route.h>
  48 #include <net/addrconf.h>
  49 #include <net/ip6_tunnel.h>
  50 #include <net/xfrm.h>
  51 #include <net/dsfield.h>
  52 #include <net/inet_ecn.h>
  53 #include <net/net_namespace.h>
  54 #include <net/netns/generic.h>
  55 #include <net/dst_metadata.h>
  56
  57 MODULE_AUTHOR("Ville Nuorvala");
  58 MODULE_DESCRIPTION("IPv6 tunneling device");
  59 MODULE_LICENSE("GPL");
  60 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  61 MODULE_ALIAS_NETDEV("ip6tnl0");
  62
  63 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
  64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
  65
  66 static bool log_ecn_error = true;
  67 module_param(log_ecn_error, bool, 0644);
  68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  69
  70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  71 {
  72         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  73
  74         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
  75 }
  76
  77 static int ip6_tnl_dev_init(struct net_device *dev);
  78 static void ip6_tnl_dev_setup(struct net_device *dev);
  79 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  80
  81 static unsigned int ip6_tnl_net_id __read_mostly;
  82 struct ip6_tnl_net {
  83         /* the IPv6 tunnel fallback device */
  84         struct net_device *fb_tnl_dev;
  85         /* lists for storing tunnels in use */
  86         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
  87         struct ip6_tnl __rcu *tnls_wc[1];
  88         struct ip6_tnl __rcu **tnls[2];
  89         struct ip6_tnl __rcu *collect_md_tun;
  90 };
  91
  92 static inline int ip6_tnl_mpls_supported(void)
  93 {
  94         return IS_ENABLED(CONFIG_MPLS);
  95 }
  96
  97 #define for_each_ip6_tunnel_rcu(start) \
  98         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
  99
 100 /**
 101  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 102  *   @net: network namespace
 103  *   @link: ifindex of underlying interface
 104  *   @remote: the address of the tunnel exit-point
 105  *   @local: the address of the tunnel entry-point
 106  *
 107  * Return:
 108  *   tunnel matching given end-points if found,
 109  *   else fallback tunnel if its device is up,
 110  *   else %NULL
 111  **/
 112
 113 static struct ip6_tnl *
 114 ip6_tnl_lookup(struct net *net, int link,
 115                const struct in6_addr *remote, const struct in6_addr *local)
 116 {
 117         unsigned int hash = HASH(remote, local);
 118         struct ip6_tnl *t, *cand = NULL;
 119         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 120         struct in6_addr any;
 121
 122         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 123                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 124                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 125                     !(t->dev->flags & IFF_UP))
 126                         continue;
 127
 128                 if (link == t->parms.link)
 129                         return t;
 130                 else
 131                         cand = t;
 132         }
 133
 134         memset(&any, 0, sizeof(any));
 135         hash = HASH(&any, local);
 136         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 137                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 138                     !ipv6_addr_any(&t->parms.raddr) ||
 139                     !(t->dev->flags & IFF_UP))
 140                         continue;
 141
 142                 if (link == t->parms.link)
 143                         return t;
 144                 else if (!cand)
 145                         cand = t;
 146         }
 147
 148         hash = HASH(remote, &any);
 149         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 150                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 151                     !ipv6_addr_any(&t->parms.laddr) ||
 152                     !(t->dev->flags & IFF_UP))
 153                         continue;
 154
 155                 if (link == t->parms.link)
 156                         return t;
 157                 else if (!cand)
 158                         cand = t;
 159         }
 160
 161         if (cand)
 162                 return cand;
 163
 164         t = rcu_dereference(ip6n->collect_md_tun);
 165         if (t && t->dev->flags & IFF_UP)
 166                 return t;
 167
 168         t = rcu_dereference(ip6n->tnls_wc[0]);
 169         if (t && (t->dev->flags & IFF_UP))
 170                 return t;
 171
 172         return NULL;
 173 }
 174
 175 /**
 176  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 177  *   @ip6n: the private data for ip6_vti in the netns
 178  *   @p: parameters containing tunnel end-points
 179  *
 180  * Description:
 181  *   ip6_tnl_bucket() returns the head of the list matching the
 182  *   &struct in6_addr entries laddr and raddr in @p.
 183  *
 184  * Return: head of IPv6 tunnel list
 185  **/
 186
 187 static struct ip6_tnl __rcu **
 188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 189 {
 190         const struct in6_addr *remote = &p->raddr;
 191         const struct in6_addr *local = &p->laddr;
 192         unsigned int h = 0;
 193         int prio = 0;
 194
 195         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 196                 prio = 1;
 197                 h = HASH(remote, local);
 198         }
 199         return &ip6n->tnls[prio][h];
 200 }
 201
 202 /**
 203  * ip6_tnl_link - add tunnel to hash table
 204  *   @ip6n: the private data for ip6_vti in the netns
 205  *   @t: tunnel to be added
 206  **/
 207
 208 static void
 209 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 210 {
 211         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 212
 213         if (t->parms.collect_md)
 214                 rcu_assign_pointer(ip6n->collect_md_tun, t);
 215         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 216         rcu_assign_pointer(*tp, t);
 217 }
 218
 219 /**
 220  * ip6_tnl_unlink - remove tunnel from hash table
 221  *   @ip6n: the private data for ip6_vti in the netns
 222  *   @t: tunnel to be removed
 223  **/
 224
 225 static void
 226 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 227 {
 228         struct ip6_tnl __rcu **tp;
 229         struct ip6_tnl *iter;
 230
 231         if (t->parms.collect_md)
 232                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 233
 234         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 235              (iter = rtnl_dereference(*tp)) != NULL;
 236              tp = &iter->next) {
 237                 if (t == iter) {
 238                         rcu_assign_pointer(*tp, t->next);
 239                         break;
 240                 }
 241         }
 242 }
 243
 244 static void ip6_dev_free(struct net_device *dev)
 245 {
 246         struct ip6_tnl *t = netdev_priv(dev);
 247
 248         gro_cells_destroy(&t->gro_cells);
 249         dst_cache_destroy(&t->dst_cache);
 250         free_percpu(dev->tstats);
 251 }
 252
 253 static int ip6_tnl_create2(struct net_device *dev)
 254 {
 255         struct ip6_tnl *t = netdev_priv(dev);
 256         struct net *net = dev_net(dev);
 257         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 258         int err;
 259
 260         t = netdev_priv(dev);
 261
 262         dev->rtnl_link_ops = &ip6_link_ops;
 263         err = register_netdevice(dev);
 264         if (err < 0)
 265                 goto out;
 266
 267         strcpy(t->parms.name, dev->name);
 268
 269         ip6_tnl_link(ip6n, t);
 270         return 0;
 271
 272 out:
 273         return err;
 274 }
 275
 276 /**
 277  * ip6_tnl_create - create a new tunnel
 278  *   @net: network namespace
 279  *   @p: tunnel parameters
 280  *
 281  * Description:
 282  *   Create tunnel matching given parameters.
 283  *
 284  * Return:
 285  *   created tunnel or error pointer
 286  **/
 287
 288 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 289 {
 290         struct net_device *dev;
 291         struct ip6_tnl *t;
 292         char name[IFNAMSIZ];
 293         int err = -E2BIG;
 294
 295         if (p->name[0]) {
 296                 if (!dev_valid_name(p->name))
 297                         goto failed;
 298                 strlcpy(name, p->name, IFNAMSIZ);
 299         } else {
 300                 sprintf(name, "ip6tnl%%d");
 301         }
 302         err = -ENOMEM;
 303         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 304                            ip6_tnl_dev_setup);
 305         if (!dev)
 306                 goto failed;
 307
 308         dev_net_set(dev, net);
 309
 310         t = netdev_priv(dev);
 311         t->parms = *p;
 312         t->net = dev_net(dev);
 313         err = ip6_tnl_create2(dev);
 314         if (err < 0)
 315                 goto failed_free;
 316
 317         return t;
 318
 319 failed_free:
 320         free_netdev(dev);
 321 failed:
 322         return ERR_PTR(err);
 323 }
 324
 325 /**
 326  * ip6_tnl_locate - find or create tunnel matching given parameters
 327  *   @net: network namespace
 328  *   @p: tunnel parameters
 329  *   @create: != 0 if allowed to create new tunnel if no match found
 330  *
 331  * Description:
 332  *   ip6_tnl_locate() first tries to locate an existing tunnel
 333  *   based on @parms. If this is unsuccessful, but @create is set a new
 334  *   tunnel device is created and registered for use.
 335  *
 336  * Return:
 337  *   matching tunnel or error pointer
 338  **/
 339
 340 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 341                 struct __ip6_tnl_parm *p, int create)
 342 {
 343         const struct in6_addr *remote = &p->raddr;
 344         const struct in6_addr *local = &p->laddr;
 345         struct ip6_tnl __rcu **tp;
 346         struct ip6_tnl *t;
 347         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 348
 349         for (tp = ip6_tnl_bucket(ip6n, p);
 350              (t = rtnl_dereference(*tp)) != NULL;
 351              tp = &t->next) {
 352                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 353                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 354                     p->link == t->parms.link) {
 355                         if (create)
 356                                 return ERR_PTR(-EEXIST);
 357
 358                         return t;
 359                 }
 360         }
 361         if (!create)
 362                 return ERR_PTR(-ENODEV);
 363         return ip6_tnl_create(net, p);
 364 }
 365
 366 /**
 367  * ip6_tnl_dev_uninit - tunnel device uninitializer
 368  *   @dev: the device to be destroyed
 369  *
 370  * Description:
 371  *   ip6_tnl_dev_uninit() removes tunnel from its list
 372  **/
 373
 374 static void
 375 ip6_tnl_dev_uninit(struct net_device *dev)
 376 {
 377         struct ip6_tnl *t = netdev_priv(dev);
 378         struct net *net = t->net;
 379         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 380
 381         if (dev == ip6n->fb_tnl_dev)
 382                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 383         else
 384                 ip6_tnl_unlink(ip6n, t);
 385         dst_cache_reset(&t->dst_cache);
 386         dev_put(dev);
 387 }
 388
 389 /**
 390  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
 391  *   @skb: received socket buffer
 392  *   @raw: the ICMPv6 error message data
 393  *
 394  * Return:
 395  *   0 if none was found,
 396  *   else index to encapsulation limit
 397  **/
 398
 399 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 400 {
 401         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 402         unsigned int nhoff = raw - skb->data;
 403         unsigned int off = nhoff + sizeof(*ipv6h);
 404         u8 next, nexthdr = ipv6h->nexthdr;
 405
 406         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 407                 struct ipv6_opt_hdr *hdr;
 408                 u16 optlen;
 409
 410                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 411                         break;
 412
 413                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 414                 if (nexthdr == NEXTHDR_FRAGMENT) {
 415                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 416                         if (frag_hdr->frag_off)
 417                                 break;
 418                         optlen = 8;
 419                 } else if (nexthdr == NEXTHDR_AUTH) {
 420                         optlen = ipv6_authlen(hdr);
 421                 } else {
 422                         optlen = ipv6_optlen(hdr);
 423                 }
 424                 /* cache hdr->nexthdr, since pskb_may_pull() might
 425                  * invalidate hdr
 426                  */
 427                 next = hdr->nexthdr;
 428                 if (nexthdr == NEXTHDR_DEST) {
 429                         u16 i = 2;
 430
 431                         /* Remember : hdr is no longer valid at this point. */
 432                         if (!pskb_may_pull(skb, off + optlen))
 433                                 break;
 434
 435                         while (1) {
 436                                 struct ipv6_tlv_tnl_enc_lim *tel;
 437
 438                                 /* No more room for encapsulation limit */
 439                                 if (i + sizeof(*tel) > optlen)
 440                                         break;
 441
 442                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 443                                 /* return index of option if found and valid */
 444                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 445                                     tel->length == 1)
 446                                         return i + off - nhoff;
 447                                 /* else jump to next option */
 448                                 if (tel->type)
 449                                         i += tel->length + 2;
 450                                 else
 451                                         i++;
 452                         }
 453                 }
 454                 nexthdr = next;
 455                 off += optlen;
 456         }
 457         return 0;
 458 }
 459 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 460
 461 /* ip6_tnl_err() should handle errors in the tunnel according to the
 462  * specifications in RFC 2473.
 463  */
 464 static int
 465 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 466             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 467 {
 468         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 469         struct net *net = dev_net(skb->dev);
 470         u8 rel_type = ICMPV6_DEST_UNREACH;
 471         u8 rel_code = ICMPV6_ADDR_UNREACH;
 472         __u32 rel_info = 0;
 473         struct ip6_tnl *t;
 474         int err = -ENOENT;
 475         int rel_msg = 0;
 476         u8 tproto;
 477         __u16 len;
 478
 479         /* If the packet doesn't contain the original IPv6 header we are
 480            in trouble since we might need the source address for further
 481            processing of the error. */
 482
 483         rcu_read_lock();
 484         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
 485         if (!t)
 486                 goto out;
 487
 488         tproto = READ_ONCE(t->parms.proto);
 489         if (tproto != ipproto && tproto != 0)
 490                 goto out;
 491
 492         err = 0;
 493
 494         switch (*type) {
 495         case ICMPV6_DEST_UNREACH:
 496                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 497                                     t->parms.name);
 498                 rel_msg = 1;
 499                 break;
 500         case ICMPV6_TIME_EXCEED:
 501                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 502                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 503                                             t->parms.name);
 504                         rel_msg = 1;
 505                 }
 506                 break;
 507         case ICMPV6_PARAMPROB: {
 508                 struct ipv6_tlv_tnl_enc_lim *tel;
 509                 __u32 teli;
 510
 511                 teli = 0;
 512                 if ((*code) == ICMPV6_HDR_FIELD)
 513                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 514
 515                 if (teli && teli == *info - 2) {
 516                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 517                         if (tel->encap_limit == 0) {
 518                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 519                                                     t->parms.name);
 520                                 rel_msg = 1;
 521                         }
 522                 } else {
 523                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 524                                             t->parms.name);
 525                 }
 526                 break;
 527         }
 528         case ICMPV6_PKT_TOOBIG: {
 529                 __u32 mtu;
 530
 531                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 532                                 sock_net_uid(net, NULL));
 533                 mtu = *info - offset;
 534                 if (mtu < IPV6_MIN_MTU)
 535                         mtu = IPV6_MIN_MTU;
 536                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 537                 if (len > mtu) {
 538                         rel_type = ICMPV6_PKT_TOOBIG;
 539                         rel_code = 0;
 540                         rel_info = mtu;
 541                         rel_msg = 1;
 542                 }
 543                 break;
 544         }
 545         case NDISC_REDIRECT:
 546                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 547                              sock_net_uid(net, NULL));
 548                 break;
 549         }
 550
 551         *type = rel_type;
 552         *code = rel_code;
 553         *info = rel_info;
 554         *msg = rel_msg;
 555
 556 out:
 557         rcu_read_unlock();
 558         return err;
 559 }
 560
 561 static int
 562 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 563            u8 type, u8 code, int offset, __be32 info)
 564 {
 565         __u32 rel_info = ntohl(info);
 566         const struct iphdr *eiph;
 567         struct sk_buff *skb2;
 568         int err, rel_msg = 0;
 569         u8 rel_type = type;
 570         u8 rel_code = code;
 571         struct rtable *rt;
 572         struct flowi4 fl4;
 573
 574         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 575                           &rel_msg, &rel_info, offset);
 576         if (err < 0)
 577                 return err;
 578
 579         if (rel_msg == 0)
 580                 return 0;
 581
 582         switch (rel_type) {
 583         case ICMPV6_DEST_UNREACH:
 584                 if (rel_code != ICMPV6_ADDR_UNREACH)
 585                         return 0;
 586                 rel_type = ICMP_DEST_UNREACH;
 587                 rel_code = ICMP_HOST_UNREACH;
 588                 break;
 589         case ICMPV6_PKT_TOOBIG:
 590                 if (rel_code != 0)
 591                         return 0;
 592                 rel_type = ICMP_DEST_UNREACH;
 593                 rel_code = ICMP_FRAG_NEEDED;
 594                 break;
 595         default:
 596                 return 0;
 597         }
 598
 599         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 600                 return 0;
 601
 602         skb2 = skb_clone(skb, GFP_ATOMIC);
 603         if (!skb2)
 604                 return 0;
 605
 606         skb_dst_drop(skb2);
 607
 608         skb_pull(skb2, offset);
 609         skb_reset_network_header(skb2);
 610         eiph = ip_hdr(skb2);
 611
 612         /* Try to guess incoming interface */
 613         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 614                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 615         if (IS_ERR(rt))
 616                 goto out;
 617
 618         skb2->dev = rt->dst.dev;
 619         ip_rt_put(rt);
 620
 621         /* route "incoming" packet */
 622         if (rt->rt_flags & RTCF_LOCAL) {
 623                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 624                                            eiph->daddr, eiph->saddr, 0, 0,
 625                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 626                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
 627                         if (!IS_ERR(rt))
 628                                 ip_rt_put(rt);
 629                         goto out;
 630                 }
 631                 skb_dst_set(skb2, &rt->dst);
 632         } else {
 633                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 634                                    skb2->dev) ||
 635                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 636                         goto out;
 637         }
 638
 639         /* change mtu on this route */
 640         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 641                 if (rel_info > dst_mtu(skb_dst(skb2)))
 642                         goto out;
 643
 644                 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
 645         }
 646
 647         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 648
 649 out:
 650         kfree_skb(skb2);
 651         return 0;
 652 }
 653
 654 static int
 655 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 656            u8 type, u8 code, int offset, __be32 info)
 657 {
 658         __u32 rel_info = ntohl(info);
 659         int err, rel_msg = 0;
 660         u8 rel_type = type;
 661         u8 rel_code = code;
 662
 663         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 664                           &rel_msg, &rel_info, offset);
 665         if (err < 0)
 666                 return err;
 667
 668         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 669                 struct rt6_info *rt;
 670                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 671
 672                 if (!skb2)
 673                         return 0;
 674
 675                 skb_dst_drop(skb2);
 676                 skb_pull(skb2, offset);
 677                 skb_reset_network_header(skb2);
 678
 679                 /* Try to guess incoming interface */
 680                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 681                                 NULL, 0, skb2, 0);
 682
 683                 if (rt && rt->dst.dev)
 684                         skb2->dev = rt->dst.dev;
 685
 686                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 687
 688                 ip6_rt_put(rt);
 689
 690                 kfree_skb(skb2);
 691         }
 692
 693         return 0;
 694 }
 695
 696 static int
 697 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 698             u8 type, u8 code, int offset, __be32 info)
 699 {
 700         __u32 rel_info = ntohl(info);
 701         int err, rel_msg = 0;
 702         u8 rel_type = type;
 703         u8 rel_code = code;
 704
 705         err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
 706                           &rel_msg, &rel_info, offset);
 707         return err;
 708 }
 709
 710 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 711                                        const struct ipv6hdr *ipv6h,
 712                                        struct sk_buff *skb)
 713 {
 714         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 715
 716         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 717                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 718
 719         return IP6_ECN_decapsulate(ipv6h, skb);
 720 }
 721
 722 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 723                                        const struct ipv6hdr *ipv6h,
 724                                        struct sk_buff *skb)
 725 {
 726         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 727                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 728
 729         return IP6_ECN_decapsulate(ipv6h, skb);
 730 }
 731
 732 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 733                                                const struct ipv6hdr *ipv6h,
 734                                                struct sk_buff *skb)
 735 {
 736         /* ECN is not supported in AF_MPLS */
 737         return 0;
 738 }
 739
 740 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 741                              const struct in6_addr *laddr,
 742                              const struct in6_addr *raddr)
 743 {
 744         struct __ip6_tnl_parm *p = &t->parms;
 745         int ltype = ipv6_addr_type(laddr);
 746         int rtype = ipv6_addr_type(raddr);
 747         __u32 flags = 0;
 748
 749         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 750                 flags = IP6_TNL_F_CAP_PER_PACKET;
 751         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 752                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 753                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 754                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 755                 if (ltype&IPV6_ADDR_UNICAST)
 756                         flags |= IP6_TNL_F_CAP_XMIT;
 757                 if (rtype&IPV6_ADDR_UNICAST)
 758                         flags |= IP6_TNL_F_CAP_RCV;
 759         }
 760         return flags;
 761 }
 762 EXPORT_SYMBOL(ip6_tnl_get_cap);
 763
 764 /* called with rcu_read_lock() */
 765 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 766                                   const struct in6_addr *laddr,
 767                                   const struct in6_addr *raddr)
 768 {
 769         struct __ip6_tnl_parm *p = &t->parms;
 770         int ret = 0;
 771         struct net *net = t->net;
 772
 773         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 774             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 775              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 776                 struct net_device *ldev = NULL;
 777
 778                 if (p->link)
 779                         ldev = dev_get_by_index_rcu(net, p->link);
 780
 781                 if ((ipv6_addr_is_multicast(laddr) ||
 782                      likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
 783                                                     0, IFA_F_TENTATIVE))) &&
 784                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
 785                      likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
 786                                                      0, IFA_F_TENTATIVE))))
 787                         ret = 1;
 788         }
 789         return ret;
 790 }
 791 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 792
 793 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 794                          const struct tnl_ptk_info *tpi,
 795                          struct metadata_dst *tun_dst,
 796                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 797                                                 const struct ipv6hdr *ipv6h,
 798                                                 struct sk_buff *skb),
 799                          bool log_ecn_err)
 800 {
 801         struct pcpu_sw_netstats *tstats;
 802         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 803         int err;
 804
 805         if ((!(tpi->flags & TUNNEL_CSUM) &&
 806              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 807             ((tpi->flags & TUNNEL_CSUM) &&
 808              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 809                 tunnel->dev->stats.rx_crc_errors++;
 810                 tunnel->dev->stats.rx_errors++;
 811                 goto drop;
 812         }
 813
 814         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 815                 if (!(tpi->flags & TUNNEL_SEQ) ||
 816                     (tunnel->i_seqno &&
 817                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 818                         tunnel->dev->stats.rx_fifo_errors++;
 819                         tunnel->dev->stats.rx_errors++;
 820                         goto drop;
 821                 }
 822                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 823         }
 824
 825         skb->protocol = tpi->proto;
 826
 827         /* Warning: All skb pointers will be invalidated! */
 828         if (tunnel->dev->type == ARPHRD_ETHER) {
 829                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 830                         tunnel->dev->stats.rx_length_errors++;
 831                         tunnel->dev->stats.rx_errors++;
 832                         goto drop;
 833                 }
 834
 835                 ipv6h = ipv6_hdr(skb);
 836                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 837                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 838         } else {
 839                 skb->dev = tunnel->dev;
 840         }
 841
 842         skb_reset_network_header(skb);
 843         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 844
 845         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 846
 847         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 848         if (unlikely(err)) {
 849                 if (log_ecn_err)
 850                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 851                                              &ipv6h->saddr,
 852                                              ipv6_get_dsfield(ipv6h));
 853                 if (err > 1) {
 854                         ++tunnel->dev->stats.rx_frame_errors;
 855                         ++tunnel->dev->stats.rx_errors;
 856                         goto drop;
 857                 }
 858         }
 859
 860         tstats = this_cpu_ptr(tunnel->dev->tstats);
 861         u64_stats_update_begin(&tstats->syncp);
 862         tstats->rx_packets++;
 863         tstats->rx_bytes += skb->len;
 864         u64_stats_update_end(&tstats->syncp);
 865
 866         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 867
 868         if (tun_dst)
 869                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 870
 871         gro_cells_receive(&tunnel->gro_cells, skb);
 872         return 0;
 873
 874 drop:
 875         if (tun_dst)
 876                 dst_release((struct dst_entry *)tun_dst);
 877         kfree_skb(skb);
 878         return 0;
 879 }
 880
 881 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 882                 const struct tnl_ptk_info *tpi,
 883                 struct metadata_dst *tun_dst,
 884                 bool log_ecn_err)
 885 {
 886         int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 887                                     const struct ipv6hdr *ipv6h,
 888                                     struct sk_buff *skb);
 889
 890         dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
 891         if (tpi->proto == htons(ETH_P_IP))
 892                 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
 893
 894         return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 895                              log_ecn_err);
 896 }
 897 EXPORT_SYMBOL(ip6_tnl_rcv);
 898
 899 static const struct tnl_ptk_info tpi_v6 = {
 900         /* no tunnel info required for ipxip6. */
 901         .proto = htons(ETH_P_IPV6),
 902 };
 903
 904 static const struct tnl_ptk_info tpi_v4 = {
 905         /* no tunnel info required for ipxip6. */
 906         .proto = htons(ETH_P_IP),
 907 };
 908
 909 static const struct tnl_ptk_info tpi_mpls = {
 910         /* no tunnel info required for mplsip6. */
 911         .proto = htons(ETH_P_MPLS_UC),
 912 };
 913
 914 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 915                       const struct tnl_ptk_info *tpi,
 916                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 917                                                   const struct ipv6hdr *ipv6h,
 918                                                   struct sk_buff *skb))
 919 {
 920         struct ip6_tnl *t;
 921         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 922         struct metadata_dst *tun_dst = NULL;
 923         int ret = -1;
 924
 925         rcu_read_lock();
 926         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
 927
 928         if (t) {
 929                 u8 tproto = READ_ONCE(t->parms.proto);
 930
 931                 if (tproto != ipproto && tproto != 0)
 932                         goto drop;
 933                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 934                         goto drop;
 935                 ipv6h = ipv6_hdr(skb);
 936                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 937                         goto drop;
 938                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 939                         goto drop;
 940                 if (t->parms.collect_md) {
 941                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 942                         if (!tun_dst)
 943                                 goto drop;
 944                 }
 945                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 946                                     log_ecn_error);
 947         }
 948
 949         rcu_read_unlock();
 950
 951         return ret;
 952
 953 drop:
 954         rcu_read_unlock();
 955         kfree_skb(skb);
 956         return 0;
 957 }
 958
 959 static int ip4ip6_rcv(struct sk_buff *skb)
 960 {
 961         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 962                           ip4ip6_dscp_ecn_decapsulate);
 963 }
 964
 965 static int ip6ip6_rcv(struct sk_buff *skb)
 966 {
 967         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 968                           ip6ip6_dscp_ecn_decapsulate);
 969 }
 970
 971 static int mplsip6_rcv(struct sk_buff *skb)
 972 {
 973         return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
 974                           mplsip6_dscp_ecn_decapsulate);
 975 }
 976
 977 struct ipv6_tel_txoption {
 978         struct ipv6_txoptions ops;
 979         __u8 dst_opt[8];
 980 };
 981
 982 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 983 {
 984         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 985
 986         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 987         opt->dst_opt[3] = 1;
 988         opt->dst_opt[4] = encap_limit;
 989         opt->dst_opt[5] = IPV6_TLV_PADN;
 990         opt->dst_opt[6] = 1;
 991
 992         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 993         opt->ops.opt_nflen = 8;
 994 }
 995
 996 /**
 997  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 998  *   @t: the outgoing tunnel device
 999  *   @hdr: IPv6 header from the incoming packet
1000  *
1001  * Description:
1002  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1003  *   doesn't match source of incoming packet.
1004  *
1005  * Return:
1006  *   1 if conflict,
1007  *   0 else
1008  **/
1009
1010 static inline bool
1011 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1012 {
1013         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1014 }
1015
1016 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1017                      const struct in6_addr *laddr,
1018                      const struct in6_addr *raddr)
1019 {
1020         struct __ip6_tnl_parm *p = &t->parms;
1021         int ret = 0;
1022         struct net *net = t->net;
1023
1024         if (t->parms.collect_md)
1025                 return 1;
1026
1027         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1028             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1029              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1030                 struct net_device *ldev = NULL;
1031
1032                 rcu_read_lock();
1033                 if (p->link)
1034                         ldev = dev_get_by_index_rcu(net, p->link);
1035
1036                 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1037                                                       0, IFA_F_TENTATIVE)))
1038                         pr_warn("%s xmit: Local address not yet configured!\n",
1039                                 p->name);
1040                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1041                          !ipv6_addr_is_multicast(raddr) &&
1042                          unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1043                                                           true, 0, IFA_F_TENTATIVE)))
1044                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1045                                 p->name);
1046                 else
1047                         ret = 1;
1048                 rcu_read_unlock();
1049         }
1050         return ret;
1051 }
1052 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1053
1054 /**
1055  * ip6_tnl_xmit - encapsulate packet and send
1056  *   @skb: the outgoing socket buffer
1057  *   @dev: the outgoing tunnel device
1058  *   @dsfield: dscp code for outer header
1059  *   @fl6: flow of tunneled packet
1060  *   @encap_limit: encapsulation limit
1061  *   @pmtu: Path MTU is stored if packet is too big
1062  *   @proto: next header value
1063  *
1064  * Description:
1065  *   Build new header and do some sanity checks on the packet before sending
1066  *   it.
1067  *
1068  * Return:
1069  *   0 on success
1070  *   -1 fail
1071  *   %-EMSGSIZE message too big. return mtu in this case.
1072  **/
1073
1074 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1075                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1076                  __u8 proto)
1077 {
1078         struct ip6_tnl *t = netdev_priv(dev);
1079         struct net *net = t->net;
1080         struct net_device_stats *stats = &t->dev->stats;
1081         struct ipv6hdr *ipv6h;
1082         struct ipv6_tel_txoption opt;
1083         struct dst_entry *dst = NULL, *ndst = NULL;
1084         struct net_device *tdev;
1085         int mtu;
1086         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1087         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1088         unsigned int max_headroom = psh_hlen;
1089         bool use_cache = false;
1090         u8 hop_limit;
1091         int err = -1;
1092
1093         if (t->parms.collect_md) {
1094                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1095                 goto route_lookup;
1096         } else {
1097                 hop_limit = t->parms.hop_limit;
1098         }
1099
1100         /* NBMA tunnel */
1101         if (ipv6_addr_any(&t->parms.raddr)) {
1102                 if (skb->protocol == htons(ETH_P_IPV6)) {
1103                         struct in6_addr *addr6;
1104                         struct neighbour *neigh;
1105                         int addr_type;
1106
1107                         if (!skb_dst(skb))
1108                                 goto tx_err_link_failure;
1109
1110                         neigh = dst_neigh_lookup(skb_dst(skb),
1111                                                  &ipv6_hdr(skb)->daddr);
1112                         if (!neigh)
1113                                 goto tx_err_link_failure;
1114
1115                         addr6 = (struct in6_addr *)&neigh->primary_key;
1116                         addr_type = ipv6_addr_type(addr6);
1117
1118                         if (addr_type == IPV6_ADDR_ANY)
1119                                 addr6 = &ipv6_hdr(skb)->daddr;
1120
1121                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1122                         neigh_release(neigh);
1123                 }
1124         } else if (t->parms.proto != 0 && !(t->parms.flags &
1125                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1126                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1127                 /* enable the cache only if neither the outer protocol nor the
1128                  * routing decision depends on the current inner header value
1129                  */
1130                 use_cache = true;
1131         }
1132
1133         if (use_cache)
1134                 dst = dst_cache_get(&t->dst_cache);
1135
1136         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1137                 goto tx_err_link_failure;
1138
1139         if (!dst) {
1140 route_lookup:
1141                 /* add dsfield to flowlabel for route lookup */
1142                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1143
1144                 dst = ip6_route_output(net, NULL, fl6);
1145
1146                 if (dst->error)
1147                         goto tx_err_link_failure;
1148                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1149                 if (IS_ERR(dst)) {
1150                         err = PTR_ERR(dst);
1151                         dst = NULL;
1152                         goto tx_err_link_failure;
1153                 }
1154                 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1155                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1156                                        &fl6->daddr, 0, &fl6->saddr))
1157                         goto tx_err_link_failure;
1158                 ndst = dst;
1159         }
1160
1161         tdev = dst->dev;
1162
1163         if (tdev == dev) {
1164                 stats->collisions++;
1165                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1166                                      t->parms.name);
1167                 goto tx_err_dst_release;
1168         }
1169         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1170         if (encap_limit >= 0) {
1171                 max_headroom += 8;
1172                 mtu -= 8;
1173         }
1174         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1175                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1176
1177         skb_dst_update_pmtu_no_confirm(skb, mtu);
1178         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1179                 *pmtu = mtu;
1180                 err = -EMSGSIZE;
1181                 goto tx_err_dst_release;
1182         }
1183
1184         if (t->err_count > 0) {
1185                 if (time_before(jiffies,
1186                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1187                         t->err_count--;
1188
1189                         dst_link_failure(skb);
1190                 } else {
1191                         t->err_count = 0;
1192                 }
1193         }
1194
1195         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1196
1197         /*
1198          * Okay, now see if we can stuff it in the buffer as-is.
1199          */
1200         max_headroom += LL_RESERVED_SPACE(tdev);
1201
1202         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1203             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1204                 struct sk_buff *new_skb;
1205
1206                 new_skb = skb_realloc_headroom(skb, max_headroom);
1207                 if (!new_skb)
1208                         goto tx_err_dst_release;
1209
1210                 if (skb->sk)
1211                         skb_set_owner_w(new_skb, skb->sk);
1212                 consume_skb(skb);
1213                 skb = new_skb;
1214         }
1215
1216         if (t->parms.collect_md) {
1217                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1218                         goto tx_err_dst_release;
1219         } else {
1220                 if (use_cache && ndst)
1221                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1222         }
1223         skb_dst_set(skb, dst);
1224
1225         if (hop_limit == 0) {
1226                 if (skb->protocol == htons(ETH_P_IP))
1227                         hop_limit = ip_hdr(skb)->ttl;
1228                 else if (skb->protocol == htons(ETH_P_IPV6))
1229                         hop_limit = ipv6_hdr(skb)->hop_limit;
1230                 else
1231                         hop_limit = ip6_dst_hoplimit(dst);
1232         }
1233
1234         /* Calculate max headroom for all the headers and adjust
1235          * needed_headroom if necessary.
1236          */
1237         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1238                         + dst->header_len + t->hlen;
1239         if (max_headroom > dev->needed_headroom)
1240                 dev->needed_headroom = max_headroom;
1241
1242         skb_set_inner_ipproto(skb, proto);
1243
1244         err = ip6_tnl_encap(skb, t, &proto, fl6);
1245         if (err)
1246                 return err;
1247
1248         if (encap_limit >= 0) {
1249                 init_tel_txopt(&opt, encap_limit);
1250                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1251         }
1252
1253         skb_push(skb, sizeof(struct ipv6hdr));
1254         skb_reset_network_header(skb);
1255         ipv6h = ipv6_hdr(skb);
1256         ip6_flow_hdr(ipv6h, dsfield,
1257                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1258         ipv6h->hop_limit = hop_limit;
1259         ipv6h->nexthdr = proto;
1260         ipv6h->saddr = fl6->saddr;
1261         ipv6h->daddr = fl6->daddr;
1262         ip6tunnel_xmit(NULL, skb, dev);
1263         return 0;
1264 tx_err_link_failure:
1265         stats->tx_carrier_errors++;
1266         dst_link_failure(skb);
1267 tx_err_dst_release:
1268         dst_release(dst);
1269         return err;
1270 }
1271 EXPORT_SYMBOL(ip6_tnl_xmit);
1272
1273 static inline int
1274 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1275                 u8 protocol)
1276 {
1277         struct ip6_tnl *t = netdev_priv(dev);
1278         struct ipv6hdr *ipv6h;
1279         const struct iphdr  *iph;
1280         int encap_limit = -1;
1281         __u16 offset;
1282         struct flowi6 fl6;
1283         __u8 dsfield, orig_dsfield;
1284         __u32 mtu;
1285         u8 tproto;
1286         int err;
1287
1288         tproto = READ_ONCE(t->parms.proto);
1289         if (tproto != protocol && tproto != 0)
1290                 return -1;
1291
1292         if (t->parms.collect_md) {
1293                 struct ip_tunnel_info *tun_info;
1294                 const struct ip_tunnel_key *key;
1295
1296                 tun_info = skb_tunnel_info(skb);
1297                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1298                              ip_tunnel_info_af(tun_info) != AF_INET6))
1299                         return -1;
1300                 key = &tun_info->key;
1301                 memset(&fl6, 0, sizeof(fl6));
1302                 fl6.flowi6_proto = protocol;
1303                 fl6.saddr = key->u.ipv6.src;
1304                 fl6.daddr = key->u.ipv6.dst;
1305                 fl6.flowlabel = key->label;
1306                 dsfield =  key->tos;
1307                 switch (protocol) {
1308                 case IPPROTO_IPIP:
1309                         iph = ip_hdr(skb);
1310                         orig_dsfield = ipv4_get_dsfield(iph);
1311                         break;
1312                 case IPPROTO_IPV6:
1313                         ipv6h = ipv6_hdr(skb);
1314                         orig_dsfield = ipv6_get_dsfield(ipv6h);
1315                         break;
1316                 default:
1317                         orig_dsfield = dsfield;
1318                         break;
1319                 }
1320         } else {
1321                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1322                         encap_limit = t->parms.encap_limit;
1323                 if (protocol == IPPROTO_IPV6) {
1324                         offset = ip6_tnl_parse_tlv_enc_lim(skb,
1325                                                 skb_network_header(skb));
1326                         /* ip6_tnl_parse_tlv_enc_lim() might have
1327                          * reallocated skb->head
1328                          */
1329                         if (offset > 0) {
1330                                 struct ipv6_tlv_tnl_enc_lim *tel;
1331
1332                                 tel = (void *)&skb_network_header(skb)[offset];
1333                                 if (tel->encap_limit == 0) {
1334                                         icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1335                                                         ICMPV6_HDR_FIELD, offset + 2);
1336                                         return -1;
1337                                 }
1338                                 encap_limit = tel->encap_limit - 1;
1339                         }
1340                 }
1341
1342                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1343                 fl6.flowi6_proto = protocol;
1344
1345                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1346                         fl6.flowi6_mark = skb->mark;
1347                 else
1348                         fl6.flowi6_mark = t->parms.fwmark;
1349                 switch (protocol) {
1350                 case IPPROTO_IPIP:
1351                         iph = ip_hdr(skb);
1352                         orig_dsfield = ipv4_get_dsfield(iph);
1353                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1354                                 dsfield = orig_dsfield;
1355                         else
1356                                 dsfield = ip6_tclass(t->parms.flowinfo);
1357                         break;
1358                 case IPPROTO_IPV6:
1359                         ipv6h = ipv6_hdr(skb);
1360                         orig_dsfield = ipv6_get_dsfield(ipv6h);
1361                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1362                                 dsfield = orig_dsfield;
1363                         else
1364                                 dsfield = ip6_tclass(t->parms.flowinfo);
1365                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1366                                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1367                         break;
1368                 default:
1369                         orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1370                         break;
1371                 }
1372         }
1373
1374         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1375         dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1376
1377         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1378                 return -1;
1379
1380         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1381                            protocol);
1382         if (err != 0) {
1383                 /* XXX: send ICMP error even if DF is not set. */
1384                 if (err == -EMSGSIZE)
1385                         switch (protocol) {
1386                         case IPPROTO_IPIP:
1387                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1388                                               ICMP_FRAG_NEEDED, htonl(mtu));
1389                                 break;
1390                         case IPPROTO_IPV6:
1391                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1392                                 break;
1393                         default:
1394                                 break;
1395                         }
1396                 return -1;
1397         }
1398
1399         return 0;
1400 }
1401
1402 static netdev_tx_t
1403 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1404 {
1405         struct ip6_tnl *t = netdev_priv(dev);
1406         struct net_device_stats *stats = &t->dev->stats;
1407         u8 ipproto;
1408         int ret;
1409
1410         if (!pskb_inet_may_pull(skb))
1411                 goto tx_err;
1412
1413         switch (skb->protocol) {
1414         case htons(ETH_P_IP):
1415                 ipproto = IPPROTO_IPIP;
1416                 break;
1417         case htons(ETH_P_IPV6):
1418                 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1419                         goto tx_err;
1420                 ipproto = IPPROTO_IPV6;
1421                 break;
1422         case htons(ETH_P_MPLS_UC):
1423                 ipproto = IPPROTO_MPLS;
1424                 break;
1425         default:
1426                 goto tx_err;
1427         }
1428
1429         ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1430         if (ret < 0)
1431                 goto tx_err;
1432
1433         return NETDEV_TX_OK;
1434
1435 tx_err:
1436         stats->tx_errors++;
1437         stats->tx_dropped++;
1438         kfree_skb(skb);
1439         return NETDEV_TX_OK;
1440 }
1441
1442 static void ip6_tnl_link_config(struct ip6_tnl *t)
1443 {
1444         struct net_device *dev = t->dev;
1445         struct net_device *tdev = NULL;
1446         struct __ip6_tnl_parm *p = &t->parms;
1447         struct flowi6 *fl6 = &t->fl.u.ip6;
1448         unsigned int mtu;
1449         int t_hlen;
1450
1451         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1452         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1453
1454         /* Set up flowi template */
1455         fl6->saddr = p->laddr;
1456         fl6->daddr = p->raddr;
1457         fl6->flowi6_oif = p->link;
1458         fl6->flowlabel = 0;
1459
1460         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1461                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1462         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1463                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1464
1465         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1466         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1467
1468         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1469                 dev->flags |= IFF_POINTOPOINT;
1470         else
1471                 dev->flags &= ~IFF_POINTOPOINT;
1472
1473         t->tun_hlen = 0;
1474         t->hlen = t->encap_hlen + t->tun_hlen;
1475         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1476
1477         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1478                 int strict = (ipv6_addr_type(&p->raddr) &
1479                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1480
1481                 struct rt6_info *rt = rt6_lookup(t->net,
1482                                                  &p->raddr, &p->laddr,
1483                                                  p->link, NULL, strict);
1484                 if (rt) {
1485                         tdev = rt->dst.dev;
1486                         ip6_rt_put(rt);
1487                 }
1488
1489                 if (!tdev && p->link)
1490                         tdev = __dev_get_by_index(t->net, p->link);
1491
1492                 if (tdev) {
1493                         dev->hard_header_len = tdev->hard_header_len + t_hlen;
1494                         mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1495
1496                         dev->mtu = mtu - t_hlen;
1497                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1498                                 dev->mtu -= 8;
1499
1500                         if (dev->mtu < IPV6_MIN_MTU)
1501                                 dev->mtu = IPV6_MIN_MTU;
1502                 }
1503         }
1504 }
1505
1506 /**
1507  * ip6_tnl_change - update the tunnel parameters
1508  *   @t: tunnel to be changed
1509  *   @p: tunnel configuration parameters
1510  *
1511  * Description:
1512  *   ip6_tnl_change() updates the tunnel parameters
1513  **/
1514
1515 static int
1516 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1517 {
1518         t->parms.laddr = p->laddr;
1519         t->parms.raddr = p->raddr;
1520         t->parms.flags = p->flags;
1521         t->parms.hop_limit = p->hop_limit;
1522         t->parms.encap_limit = p->encap_limit;
1523         t->parms.flowinfo = p->flowinfo;
1524         t->parms.link = p->link;
1525         t->parms.proto = p->proto;
1526         t->parms.fwmark = p->fwmark;
1527         dst_cache_reset(&t->dst_cache);
1528         ip6_tnl_link_config(t);
1529         return 0;
1530 }
1531
1532 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1533 {
1534         struct net *net = t->net;
1535         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1536         int err;
1537
1538         ip6_tnl_unlink(ip6n, t);
1539         synchronize_net();
1540         err = ip6_tnl_change(t, p);
1541         ip6_tnl_link(ip6n, t);
1542         netdev_state_change(t->dev);
1543         return err;
1544 }
1545
1546 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1547 {
1548         /* for default tnl0 device allow to change only the proto */
1549         t->parms.proto = p->proto;
1550         netdev_state_change(t->dev);
1551         return 0;
1552 }
1553
1554 static void
1555 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1556 {
1557         p->laddr = u->laddr;
1558         p->raddr = u->raddr;
1559         p->flags = u->flags;
1560         p->hop_limit = u->hop_limit;
1561         p->encap_limit = u->encap_limit;
1562         p->flowinfo = u->flowinfo;
1563         p->link = u->link;
1564         p->proto = u->proto;
1565         memcpy(p->name, u->name, sizeof(u->name));
1566 }
1567
1568 static void
1569 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1570 {
1571         u->laddr = p->laddr;
1572         u->raddr = p->raddr;
1573         u->flags = p->flags;
1574         u->hop_limit = p->hop_limit;
1575         u->encap_limit = p->encap_limit;
1576         u->flowinfo = p->flowinfo;
1577         u->link = p->link;
1578         u->proto = p->proto;
1579         memcpy(u->name, p->name, sizeof(u->name));
1580 }
1581
1582 /**
1583  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1584  *   @dev: virtual device associated with tunnel
1585  *   @ifr: parameters passed from userspace
1586  *   @cmd: command to be performed
1587  *
1588  * Description:
1589  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1590  *   from userspace.
1591  *
1592  *   The possible commands are the following:
1593  *     %SIOCGETTUNNEL: get tunnel parameters for device
1594  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1595  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1596  *     %SIOCDELTUNNEL: delete tunnel
1597  *
1598  *   The fallback device "ip6tnl0", created during module
1599  *   initialization, can be used for creating other tunnel devices.
1600  *
1601  * Return:
1602  *   0 on success,
1603  *   %-EFAULT if unable to copy data to or from userspace,
1604  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1605  *   %-EINVAL if passed tunnel parameters are invalid,
1606  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1607  *   %-ENODEV if attempting to change or delete a nonexisting device
1608  **/
1609
1610 static int
1611 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1612 {
1613         int err = 0;
1614         struct ip6_tnl_parm p;
1615         struct __ip6_tnl_parm p1;
1616         struct ip6_tnl *t = netdev_priv(dev);
1617         struct net *net = t->net;
1618         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1619
1620         memset(&p1, 0, sizeof(p1));
1621
1622         switch (cmd) {
1623         case SIOCGETTUNNEL:
1624                 if (dev == ip6n->fb_tnl_dev) {
1625                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1626                                 err = -EFAULT;
1627                                 break;
1628                         }
1629                         ip6_tnl_parm_from_user(&p1, &p);
1630                         t = ip6_tnl_locate(net, &p1, 0);
1631                         if (IS_ERR(t))
1632                                 t = netdev_priv(dev);
1633                 } else {
1634                         memset(&p, 0, sizeof(p));
1635                 }
1636                 ip6_tnl_parm_to_user(&p, &t->parms);
1637                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1638                         err = -EFAULT;
1639                 }
1640                 break;
1641         case SIOCADDTUNNEL:
1642         case SIOCCHGTUNNEL:
1643                 err = -EPERM;
1644                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1645                         break;
1646                 err = -EFAULT;
1647                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1648                         break;
1649                 err = -EINVAL;
1650                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1651                     p.proto != 0)
1652                         break;
1653                 ip6_tnl_parm_from_user(&p1, &p);
1654                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1655                 if (cmd == SIOCCHGTUNNEL) {
1656                         if (!IS_ERR(t)) {
1657                                 if (t->dev != dev) {
1658                                         err = -EEXIST;
1659                                         break;
1660                                 }
1661                         } else
1662                                 t = netdev_priv(dev);
1663                         if (dev == ip6n->fb_tnl_dev)
1664                                 err = ip6_tnl0_update(t, &p1);
1665                         else
1666                                 err = ip6_tnl_update(t, &p1);
1667                 }
1668                 if (!IS_ERR(t)) {
1669                         err = 0;
1670                         ip6_tnl_parm_to_user(&p, &t->parms);
1671                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1672                                 err = -EFAULT;
1673
1674                 } else {
1675                         err = PTR_ERR(t);
1676                 }
1677                 break;
1678         case SIOCDELTUNNEL:
1679                 err = -EPERM;
1680                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1681                         break;
1682
1683                 if (dev == ip6n->fb_tnl_dev) {
1684                         err = -EFAULT;
1685                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1686                                 break;
1687                         err = -ENOENT;
1688                         ip6_tnl_parm_from_user(&p1, &p);
1689                         t = ip6_tnl_locate(net, &p1, 0);
1690                         if (IS_ERR(t))
1691                                 break;
1692                         err = -EPERM;
1693                         if (t->dev == ip6n->fb_tnl_dev)
1694                                 break;
1695                         dev = t->dev;
1696                 }
1697                 err = 0;
1698                 unregister_netdevice(dev);
1699                 break;
1700         default:
1701                 err = -EINVAL;
1702         }
1703         return err;
1704 }
1705
1706 /**
1707  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1708  *   @dev: virtual device associated with tunnel
1709  *   @new_mtu: the new mtu
1710  *
1711  * Return:
1712  *   0 on success,
1713  *   %-EINVAL if mtu too small
1714  **/
1715
1716 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1717 {
1718         struct ip6_tnl *tnl = netdev_priv(dev);
1719
1720         if (tnl->parms.proto == IPPROTO_IPV6) {
1721                 if (new_mtu < IPV6_MIN_MTU)
1722                         return -EINVAL;
1723         } else {
1724                 if (new_mtu < ETH_MIN_MTU)
1725                         return -EINVAL;
1726         }
1727         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1728                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1729                         return -EINVAL;
1730         } else {
1731                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1732                         return -EINVAL;
1733         }
1734         dev->mtu = new_mtu;
1735         return 0;
1736 }
1737 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1738
1739 int ip6_tnl_get_iflink(const struct net_device *dev)
1740 {
1741         struct ip6_tnl *t = netdev_priv(dev);
1742
1743         return t->parms.link;
1744 }
1745 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1746
1747 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1748                           unsigned int num)
1749 {
1750         if (num >= MAX_IPTUN_ENCAP_OPS)
1751                 return -ERANGE;
1752
1753         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1754                         &ip6tun_encaps[num],
1755                         NULL, ops) ? 0 : -1;
1756 }
1757 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1758
1759 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1760                           unsigned int num)
1761 {
1762         int ret;
1763
1764         if (num >= MAX_IPTUN_ENCAP_OPS)
1765                 return -ERANGE;
1766
1767         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1768                        &ip6tun_encaps[num],
1769                        ops, NULL) == ops) ? 0 : -1;
1770
1771         synchronize_net();
1772
1773         return ret;
1774 }
1775 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1776
1777 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1778                         struct ip_tunnel_encap *ipencap)
1779 {
1780         int hlen;
1781
1782         memset(&t->encap, 0, sizeof(t->encap));
1783
1784         hlen = ip6_encap_hlen(ipencap);
1785         if (hlen < 0)
1786                 return hlen;
1787
1788         t->encap.type = ipencap->type;
1789         t->encap.sport = ipencap->sport;
1790         t->encap.dport = ipencap->dport;
1791         t->encap.flags = ipencap->flags;
1792
1793         t->encap_hlen = hlen;
1794         t->hlen = t->encap_hlen + t->tun_hlen;
1795
1796         return 0;
1797 }
1798 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1799
1800 static const struct net_device_ops ip6_tnl_netdev_ops = {
1801         .ndo_init       = ip6_tnl_dev_init,
1802         .ndo_uninit     = ip6_tnl_dev_uninit,
1803         .ndo_start_xmit = ip6_tnl_start_xmit,
1804         .ndo_do_ioctl   = ip6_tnl_ioctl,
1805         .ndo_change_mtu = ip6_tnl_change_mtu,
1806         .ndo_get_stats64 = dev_get_tstats64,
1807         .ndo_get_iflink = ip6_tnl_get_iflink,
1808 };
1809
1810 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1811                          NETIF_F_FRAGLIST |     \
1812                          NETIF_F_HIGHDMA |      \
1813                          NETIF_F_GSO_SOFTWARE | \
1814                          NETIF_F_HW_CSUM)
1815
1816 /**
1817  * ip6_tnl_dev_setup - setup virtual tunnel device
1818  *   @dev: virtual device associated with tunnel
1819  *
1820  * Description:
1821  *   Initialize function pointers and device parameters
1822  **/
1823
1824 static void ip6_tnl_dev_setup(struct net_device *dev)
1825 {
1826         dev->netdev_ops = &ip6_tnl_netdev_ops;
1827         dev->header_ops = &ip_tunnel_header_ops;
1828         dev->needs_free_netdev = true;
1829         dev->priv_destructor = ip6_dev_free;
1830
1831         dev->type = ARPHRD_TUNNEL6;
1832         dev->flags |= IFF_NOARP;
1833         dev->addr_len = sizeof(struct in6_addr);
1834         dev->features |= NETIF_F_LLTX;
1835         netif_keep_dst(dev);
1836
1837         dev->features           |= IPXIPX_FEATURES;
1838         dev->hw_features        |= IPXIPX_FEATURES;
1839
1840         /* This perm addr will be used as interface identifier by IPv6 */
1841         dev->addr_assign_type = NET_ADDR_RANDOM;
1842         eth_random_addr(dev->perm_addr);
1843 }
1844
1845
1846 /**
1847  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1848  *   @dev: virtual device associated with tunnel
1849  **/
1850
1851 static inline int
1852 ip6_tnl_dev_init_gen(struct net_device *dev)
1853 {
1854         struct ip6_tnl *t = netdev_priv(dev);
1855         int ret;
1856         int t_hlen;
1857
1858         t->dev = dev;
1859         t->net = dev_net(dev);
1860         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1861         if (!dev->tstats)
1862                 return -ENOMEM;
1863
1864         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1865         if (ret)
1866                 goto free_stats;
1867
1868         ret = gro_cells_init(&t->gro_cells, dev);
1869         if (ret)
1870                 goto destroy_dst;
1871
1872         t->tun_hlen = 0;
1873         t->hlen = t->encap_hlen + t->tun_hlen;
1874         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1875
1876         dev->type = ARPHRD_TUNNEL6;
1877         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1878         dev->mtu = ETH_DATA_LEN - t_hlen;
1879         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1880                 dev->mtu -= 8;
1881         dev->min_mtu = ETH_MIN_MTU;
1882         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1883
1884         dev_hold(dev);
1885         return 0;
1886
1887 destroy_dst:
1888         dst_cache_destroy(&t->dst_cache);
1889 free_stats:
1890         free_percpu(dev->tstats);
1891         dev->tstats = NULL;
1892
1893         return ret;
1894 }
1895
1896 /**
1897  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1898  *   @dev: virtual device associated with tunnel
1899  **/
1900
1901 static int ip6_tnl_dev_init(struct net_device *dev)
1902 {
1903         struct ip6_tnl *t = netdev_priv(dev);
1904         int err = ip6_tnl_dev_init_gen(dev);
1905
1906         if (err)
1907                 return err;
1908         ip6_tnl_link_config(t);
1909         if (t->parms.collect_md)
1910                 netif_keep_dst(dev);
1911         return 0;
1912 }
1913
1914 /**
1915  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1916  *   @dev: fallback device
1917  *
1918  * Return: 0
1919  **/
1920
1921 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1922 {
1923         struct ip6_tnl *t = netdev_priv(dev);
1924         struct net *net = dev_net(dev);
1925         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1926
1927         t->parms.proto = IPPROTO_IPV6;
1928
1929         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1930         return 0;
1931 }
1932
1933 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1934                             struct netlink_ext_ack *extack)
1935 {
1936         u8 proto;
1937
1938         if (!data || !data[IFLA_IPTUN_PROTO])
1939                 return 0;
1940
1941         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1942         if (proto != IPPROTO_IPV6 &&
1943             proto != IPPROTO_IPIP &&
1944             proto != 0)
1945                 return -EINVAL;
1946
1947         return 0;
1948 }
1949
1950 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1951                                   struct __ip6_tnl_parm *parms)
1952 {
1953         memset(parms, 0, sizeof(*parms));
1954
1955         if (!data)
1956                 return;
1957
1958         if (data[IFLA_IPTUN_LINK])
1959                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1960
1961         if (data[IFLA_IPTUN_LOCAL])
1962                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1963
1964         if (data[IFLA_IPTUN_REMOTE])
1965                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1966
1967         if (data[IFLA_IPTUN_TTL])
1968                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1969
1970         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1971                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1972
1973         if (data[IFLA_IPTUN_FLOWINFO])
1974                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1975
1976         if (data[IFLA_IPTUN_FLAGS])
1977                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1978
1979         if (data[IFLA_IPTUN_PROTO])
1980                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1981
1982         if (data[IFLA_IPTUN_COLLECT_METADATA])
1983                 parms->collect_md = true;
1984
1985         if (data[IFLA_IPTUN_FWMARK])
1986                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1987 }
1988
1989 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1990                                         struct ip_tunnel_encap *ipencap)
1991 {
1992         bool ret = false;
1993
1994         memset(ipencap, 0, sizeof(*ipencap));
1995
1996         if (!data)
1997                 return ret;
1998
1999         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
2000                 ret = true;
2001                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2002         }
2003
2004         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2005                 ret = true;
2006                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2007         }
2008
2009         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2010                 ret = true;
2011                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2012         }
2013
2014         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2015                 ret = true;
2016                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2017         }
2018
2019         return ret;
2020 }
2021
2022 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2023                            struct nlattr *tb[], struct nlattr *data[],
2024                            struct netlink_ext_ack *extack)
2025 {
2026         struct net *net = dev_net(dev);
2027         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2028         struct ip_tunnel_encap ipencap;
2029         struct ip6_tnl *nt, *t;
2030         int err;
2031
2032         nt = netdev_priv(dev);
2033
2034         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2035                 err = ip6_tnl_encap_setup(nt, &ipencap);
2036                 if (err < 0)
2037                         return err;
2038         }
2039
2040         ip6_tnl_netlink_parms(data, &nt->parms);
2041
2042         if (nt->parms.collect_md) {
2043                 if (rtnl_dereference(ip6n->collect_md_tun))
2044                         return -EEXIST;
2045         } else {
2046                 t = ip6_tnl_locate(net, &nt->parms, 0);
2047                 if (!IS_ERR(t))
2048                         return -EEXIST;
2049         }
2050
2051         err = ip6_tnl_create2(dev);
2052         if (!err && tb[IFLA_MTU])
2053                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2054
2055         return err;
2056 }
2057
2058 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2059                               struct nlattr *data[],
2060                               struct netlink_ext_ack *extack)
2061 {
2062         struct ip6_tnl *t = netdev_priv(dev);
2063         struct __ip6_tnl_parm p;
2064         struct net *net = t->net;
2065         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2066         struct ip_tunnel_encap ipencap;
2067
2068         if (dev == ip6n->fb_tnl_dev)
2069                 return -EINVAL;
2070
2071         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2072                 int err = ip6_tnl_encap_setup(t, &ipencap);
2073
2074                 if (err < 0)
2075                         return err;
2076         }
2077         ip6_tnl_netlink_parms(data, &p);
2078         if (p.collect_md)
2079                 return -EINVAL;
2080
2081         t = ip6_tnl_locate(net, &p, 0);
2082         if (!IS_ERR(t)) {
2083                 if (t->dev != dev)
2084                         return -EEXIST;
2085         } else
2086                 t = netdev_priv(dev);
2087
2088         return ip6_tnl_update(t, &p);
2089 }
2090
2091 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2092 {
2093         struct net *net = dev_net(dev);
2094         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2095
2096         if (dev != ip6n->fb_tnl_dev)
2097                 unregister_netdevice_queue(dev, head);
2098 }
2099
2100 static size_t ip6_tnl_get_size(const struct net_device *dev)
2101 {
2102         return
2103                 /* IFLA_IPTUN_LINK */
2104                 nla_total_size(4) +
2105                 /* IFLA_IPTUN_LOCAL */
2106                 nla_total_size(sizeof(struct in6_addr)) +
2107                 /* IFLA_IPTUN_REMOTE */
2108                 nla_total_size(sizeof(struct in6_addr)) +
2109                 /* IFLA_IPTUN_TTL */
2110                 nla_total_size(1) +
2111                 /* IFLA_IPTUN_ENCAP_LIMIT */
2112                 nla_total_size(1) +
2113                 /* IFLA_IPTUN_FLOWINFO */
2114                 nla_total_size(4) +
2115                 /* IFLA_IPTUN_FLAGS */
2116                 nla_total_size(4) +
2117                 /* IFLA_IPTUN_PROTO */
2118                 nla_total_size(1) +
2119                 /* IFLA_IPTUN_ENCAP_TYPE */
2120                 nla_total_size(2) +
2121                 /* IFLA_IPTUN_ENCAP_FLAGS */
2122                 nla_total_size(2) +
2123                 /* IFLA_IPTUN_ENCAP_SPORT */
2124                 nla_total_size(2) +
2125                 /* IFLA_IPTUN_ENCAP_DPORT */
2126                 nla_total_size(2) +
2127                 /* IFLA_IPTUN_COLLECT_METADATA */
2128                 nla_total_size(0) +
2129                 /* IFLA_IPTUN_FWMARK */
2130                 nla_total_size(4) +
2131                 0;
2132 }
2133
2134 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2135 {
2136         struct ip6_tnl *tunnel = netdev_priv(dev);
2137         struct __ip6_tnl_parm *parm = &tunnel->parms;
2138
2139         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2140             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2141             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2142             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2143             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2144             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2145             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2146             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2147             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2148                 goto nla_put_failure;
2149
2150         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2151             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2152             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2153             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2154                 goto nla_put_failure;
2155
2156         if (parm->collect_md)
2157                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2158                         goto nla_put_failure;
2159
2160         return 0;
2161
2162 nla_put_failure:
2163         return -EMSGSIZE;
2164 }
2165
2166 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2167 {
2168         struct ip6_tnl *tunnel = netdev_priv(dev);
2169
2170         return tunnel->net;
2171 }
2172 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2173
2174 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2175         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2176         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2177         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2178         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2179         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2180         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2181         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2182         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2183         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2184         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2185         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2186         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2187         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2188         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2189 };
2190
2191 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2192         .kind           = "ip6tnl",
2193         .maxtype        = IFLA_IPTUN_MAX,
2194         .policy         = ip6_tnl_policy,
2195         .priv_size      = sizeof(struct ip6_tnl),
2196         .setup          = ip6_tnl_dev_setup,
2197         .validate       = ip6_tnl_validate,
2198         .newlink        = ip6_tnl_newlink,
2199         .changelink     = ip6_tnl_changelink,
2200         .dellink        = ip6_tnl_dellink,
2201         .get_size       = ip6_tnl_get_size,
2202         .fill_info      = ip6_tnl_fill_info,
2203         .get_link_net   = ip6_tnl_get_link_net,
2204 };
2205
2206 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2207         .handler        = ip4ip6_rcv,
2208         .err_handler    = ip4ip6_err,
2209         .priority       =       1,
2210 };
2211
2212 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2213         .handler        = ip6ip6_rcv,
2214         .err_handler    = ip6ip6_err,
2215         .priority       =       1,
2216 };
2217
2218 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2219         .handler        = mplsip6_rcv,
2220         .err_handler    = mplsip6_err,
2221         .priority       =       1,
2222 };
2223
2224 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2225 {
2226         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2227         struct net_device *dev, *aux;
2228         int h;
2229         struct ip6_tnl *t;
2230
2231         for_each_netdev_safe(net, dev, aux)
2232                 if (dev->rtnl_link_ops == &ip6_link_ops)
2233                         unregister_netdevice_queue(dev, list);
2234
2235         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2236                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2237                 while (t) {
2238                         /* If dev is in the same netns, it has already
2239                          * been added to the list by the previous loop.
2240                          */
2241                         if (!net_eq(dev_net(t->dev), net))
2242                                 unregister_netdevice_queue(t->dev, list);
2243                         t = rtnl_dereference(t->next);
2244                 }
2245         }
2246 }
2247
2248 static int __net_init ip6_tnl_init_net(struct net *net)
2249 {
2250         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2251         struct ip6_tnl *t = NULL;
2252         int err;
2253
2254         ip6n->tnls[0] = ip6n->tnls_wc;
2255         ip6n->tnls[1] = ip6n->tnls_r_l;
2256
2257         if (!net_has_fallback_tunnels(net))
2258                 return 0;
2259         err = -ENOMEM;
2260         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2261                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2262
2263         if (!ip6n->fb_tnl_dev)
2264                 goto err_alloc_dev;
2265         dev_net_set(ip6n->fb_tnl_dev, net);
2266         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2267         /* FB netdevice is special: we have one, and only one per netns.
2268          * Allowing to move it to another netns is clearly unsafe.
2269          */
2270         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2271
2272         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2273         if (err < 0)
2274                 goto err_register;
2275
2276         err = register_netdev(ip6n->fb_tnl_dev);
2277         if (err < 0)
2278                 goto err_register;
2279
2280         t = netdev_priv(ip6n->fb_tnl_dev);
2281
2282         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2283         return 0;
2284
2285 err_register:
2286         free_netdev(ip6n->fb_tnl_dev);
2287 err_alloc_dev:
2288         return err;
2289 }
2290
2291 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2292 {
2293         struct net *net;
2294         LIST_HEAD(list);
2295
2296         rtnl_lock();
2297         list_for_each_entry(net, net_list, exit_list)
2298                 ip6_tnl_destroy_tunnels(net, &list);
2299         unregister_netdevice_many(&list);
2300         rtnl_unlock();
2301 }
2302
2303 static struct pernet_operations ip6_tnl_net_ops = {
2304         .init = ip6_tnl_init_net,
2305         .exit_batch = ip6_tnl_exit_batch_net,
2306         .id   = &ip6_tnl_net_id,
2307         .size = sizeof(struct ip6_tnl_net),
2308 };
2309
2310 /**
2311  * ip6_tunnel_init - register protocol and reserve needed resources
2312  *
2313  * Return: 0 on success
2314  **/
2315
2316 static int __init ip6_tunnel_init(void)
2317 {
2318         int  err;
2319
2320         if (!ipv6_mod_enabled())
2321                 return -EOPNOTSUPP;
2322
2323         err = register_pernet_device(&ip6_tnl_net_ops);
2324         if (err < 0)
2325                 goto out_pernet;
2326
2327         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2328         if (err < 0) {
2329                 pr_err("%s: can't register ip4ip6\n", __func__);
2330                 goto out_ip4ip6;
2331         }
2332
2333         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2334         if (err < 0) {
2335                 pr_err("%s: can't register ip6ip6\n", __func__);
2336                 goto out_ip6ip6;
2337         }
2338
2339         if (ip6_tnl_mpls_supported()) {
2340                 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2341                 if (err < 0) {
2342                         pr_err("%s: can't register mplsip6\n", __func__);
2343                         goto out_mplsip6;
2344                 }
2345         }
2346
2347         err = rtnl_link_register(&ip6_link_ops);
2348         if (err < 0)
2349                 goto rtnl_link_failed;
2350
2351         return 0;
2352
2353 rtnl_link_failed:
2354         if (ip6_tnl_mpls_supported())
2355                 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2356 out_mplsip6:
2357         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2358 out_ip6ip6:
2359         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2360 out_ip4ip6:
2361         unregister_pernet_device(&ip6_tnl_net_ops);
2362 out_pernet:
2363         return err;
2364 }
2365
2366 /**
2367  * ip6_tunnel_cleanup - free resources and unregister protocol
2368  **/
2369
2370 static void __exit ip6_tunnel_cleanup(void)
2371 {
2372         rtnl_link_unregister(&ip6_link_ops);
2373         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2374                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2375
2376         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2377                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2378
2379         if (ip6_tnl_mpls_supported() &&
2380             xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2381                 pr_info("%s: can't deregister mplsip6\n", __func__);
2382         unregister_pernet_device(&ip6_tnl_net_ops);
2383 }
2384
2385 module_init(ip6_tunnel_init);
2386 module_exit(ip6_tunnel_cleanup);