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          <[email protected]>
   8  *      Yasuyuki Kozakai        <[email protected]>
   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         dev->rtnl_link_ops = &ip6_link_ops;
 261         err = register_netdevice(dev);
 262         if (err < 0)
 263                 goto out;
 264
 265         strcpy(t->parms.name, dev->name);
 266
 267         ip6_tnl_link(ip6n, t);
 268         return 0;
 269
 270 out:
 271         return err;
 272 }
 273
 274 /**
 275  * ip6_tnl_create - create a new tunnel
 276  *   @net: network namespace
 277  *   @p: tunnel parameters
 278  *
 279  * Description:
 280  *   Create tunnel matching given parameters.
 281  *
 282  * Return:
 283  *   created tunnel or error pointer
 284  **/
 285
 286 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 287 {
 288         struct net_device *dev;
 289         struct ip6_tnl *t;
 290         char name[IFNAMSIZ];
 291         int err = -E2BIG;
 292
 293         if (p->name[0]) {
 294                 if (!dev_valid_name(p->name))
 295                         goto failed;
 296                 strscpy(name, p->name, IFNAMSIZ);
 297         } else {
 298                 sprintf(name, "ip6tnl%%d");
 299         }
 300         err = -ENOMEM;
 301         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 302                            ip6_tnl_dev_setup);
 303         if (!dev)
 304                 goto failed;
 305
 306         dev_net_set(dev, net);
 307
 308         t = netdev_priv(dev);
 309         t->parms = *p;
 310         t->net = dev_net(dev);
 311         err = ip6_tnl_create2(dev);
 312         if (err < 0)
 313                 goto failed_free;
 314
 315         return t;
 316
 317 failed_free:
 318         free_netdev(dev);
 319 failed:
 320         return ERR_PTR(err);
 321 }
 322
 323 /**
 324  * ip6_tnl_locate - find or create tunnel matching given parameters
 325  *   @net: network namespace
 326  *   @p: tunnel parameters
 327  *   @create: != 0 if allowed to create new tunnel if no match found
 328  *
 329  * Description:
 330  *   ip6_tnl_locate() first tries to locate an existing tunnel
 331  *   based on @parms. If this is unsuccessful, but @create is set a new
 332  *   tunnel device is created and registered for use.
 333  *
 334  * Return:
 335  *   matching tunnel or error pointer
 336  **/
 337
 338 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 339                 struct __ip6_tnl_parm *p, int create)
 340 {
 341         const struct in6_addr *remote = &p->raddr;
 342         const struct in6_addr *local = &p->laddr;
 343         struct ip6_tnl __rcu **tp;
 344         struct ip6_tnl *t;
 345         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 346
 347         for (tp = ip6_tnl_bucket(ip6n, p);
 348              (t = rtnl_dereference(*tp)) != NULL;
 349              tp = &t->next) {
 350                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 351                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 352                     p->link == t->parms.link) {
 353                         if (create)
 354                                 return ERR_PTR(-EEXIST);
 355
 356                         return t;
 357                 }
 358         }
 359         if (!create)
 360                 return ERR_PTR(-ENODEV);
 361         return ip6_tnl_create(net, p);
 362 }
 363
 364 /**
 365  * ip6_tnl_dev_uninit - tunnel device uninitializer
 366  *   @dev: the device to be destroyed
 367  *
 368  * Description:
 369  *   ip6_tnl_dev_uninit() removes tunnel from its list
 370  **/
 371
 372 static void
 373 ip6_tnl_dev_uninit(struct net_device *dev)
 374 {
 375         struct ip6_tnl *t = netdev_priv(dev);
 376         struct net *net = t->net;
 377         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 378
 379         if (dev == ip6n->fb_tnl_dev)
 380                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 381         else
 382                 ip6_tnl_unlink(ip6n, t);
 383         dst_cache_reset(&t->dst_cache);
 384         netdev_put(dev, &t->dev_tracker);
 385 }
 386
 387 /**
 388  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
 389  *   @skb: received socket buffer
 390  *   @raw: the ICMPv6 error message data
 391  *
 392  * Return:
 393  *   0 if none was found,
 394  *   else index to encapsulation limit
 395  **/
 396
 397 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 398 {
 399         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 400         unsigned int nhoff = raw - skb->data;
 401         unsigned int off = nhoff + sizeof(*ipv6h);
 402         u8 next, nexthdr = ipv6h->nexthdr;
 403
 404         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 405                 struct ipv6_opt_hdr *hdr;
 406                 u16 optlen;
 407
 408                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 409                         break;
 410
 411                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 412                 if (nexthdr == NEXTHDR_FRAGMENT) {
 413                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 414                         if (frag_hdr->frag_off)
 415                                 break;
 416                         optlen = 8;
 417                 } else if (nexthdr == NEXTHDR_AUTH) {
 418                         optlen = ipv6_authlen(hdr);
 419                 } else {
 420                         optlen = ipv6_optlen(hdr);
 421                 }
 422                 /* cache hdr->nexthdr, since pskb_may_pull() might
 423                  * invalidate hdr
 424                  */
 425                 next = hdr->nexthdr;
 426                 if (nexthdr == NEXTHDR_DEST) {
 427                         u16 i = 2;
 428
 429                         /* Remember : hdr is no longer valid at this point. */
 430                         if (!pskb_may_pull(skb, off + optlen))
 431                                 break;
 432
 433                         while (1) {
 434                                 struct ipv6_tlv_tnl_enc_lim *tel;
 435
 436                                 /* No more room for encapsulation limit */
 437                                 if (i + sizeof(*tel) > optlen)
 438                                         break;
 439
 440                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 441                                 /* return index of option if found and valid */
 442                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 443                                     tel->length == 1)
 444                                         return i + off - nhoff;
 445                                 /* else jump to next option */
 446                                 if (tel->type)
 447                                         i += tel->length + 2;
 448                                 else
 449                                         i++;
 450                         }
 451                 }
 452                 nexthdr = next;
 453                 off += optlen;
 454         }
 455         return 0;
 456 }
 457 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 458
 459 /* ip6_tnl_err() should handle errors in the tunnel according to the
 460  * specifications in RFC 2473.
 461  */
 462 static int
 463 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 464             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 465 {
 466         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 467         struct net *net = dev_net(skb->dev);
 468         u8 rel_type = ICMPV6_DEST_UNREACH;
 469         u8 rel_code = ICMPV6_ADDR_UNREACH;
 470         __u32 rel_info = 0;
 471         struct ip6_tnl *t;
 472         int err = -ENOENT;
 473         int rel_msg = 0;
 474         u8 tproto;
 475         __u16 len;
 476
 477         /* If the packet doesn't contain the original IPv6 header we are
 478            in trouble since we might need the source address for further
 479            processing of the error. */
 480
 481         rcu_read_lock();
 482         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
 483         if (!t)
 484                 goto out;
 485
 486         tproto = READ_ONCE(t->parms.proto);
 487         if (tproto != ipproto && tproto != 0)
 488                 goto out;
 489
 490         err = 0;
 491
 492         switch (*type) {
 493         case ICMPV6_DEST_UNREACH:
 494                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 495                                     t->parms.name);
 496                 rel_msg = 1;
 497                 break;
 498         case ICMPV6_TIME_EXCEED:
 499                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 500                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 501                                             t->parms.name);
 502                         rel_msg = 1;
 503                 }
 504                 break;
 505         case ICMPV6_PARAMPROB: {
 506                 struct ipv6_tlv_tnl_enc_lim *tel;
 507                 __u32 teli;
 508
 509                 teli = 0;
 510                 if ((*code) == ICMPV6_HDR_FIELD)
 511                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 512
 513                 if (teli && teli == *info - 2) {
 514                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 515                         if (tel->encap_limit == 0) {
 516                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 517                                                     t->parms.name);
 518                                 rel_msg = 1;
 519                         }
 520                 } else {
 521                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 522                                             t->parms.name);
 523                 }
 524                 break;
 525         }
 526         case ICMPV6_PKT_TOOBIG: {
 527                 __u32 mtu;
 528
 529                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 530                                 sock_net_uid(net, NULL));
 531                 mtu = *info - offset;
 532                 if (mtu < IPV6_MIN_MTU)
 533                         mtu = IPV6_MIN_MTU;
 534                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 535                 if (len > mtu) {
 536                         rel_type = ICMPV6_PKT_TOOBIG;
 537                         rel_code = 0;
 538                         rel_info = mtu;
 539                         rel_msg = 1;
 540                 }
 541                 break;
 542         }
 543         case NDISC_REDIRECT:
 544                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 545                              sock_net_uid(net, NULL));
 546                 break;
 547         }
 548
 549         *type = rel_type;
 550         *code = rel_code;
 551         *info = rel_info;
 552         *msg = rel_msg;
 553
 554 out:
 555         rcu_read_unlock();
 556         return err;
 557 }
 558
 559 static int
 560 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 561            u8 type, u8 code, int offset, __be32 info)
 562 {
 563         __u32 rel_info = ntohl(info);
 564         const struct iphdr *eiph;
 565         struct sk_buff *skb2;
 566         int err, rel_msg = 0;
 567         u8 rel_type = type;
 568         u8 rel_code = code;
 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         default:
 594                 return 0;
 595         }
 596
 597         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 598                 return 0;
 599
 600         skb2 = skb_clone(skb, GFP_ATOMIC);
 601         if (!skb2)
 602                 return 0;
 603
 604         skb_dst_drop(skb2);
 605
 606         skb_pull(skb2, offset);
 607         skb_reset_network_header(skb2);
 608         eiph = ip_hdr(skb2);
 609
 610         /* Try to guess incoming interface */
 611         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 612                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 613         if (IS_ERR(rt))
 614                 goto out;
 615
 616         skb2->dev = rt->dst.dev;
 617         ip_rt_put(rt);
 618
 619         /* route "incoming" packet */
 620         if (rt->rt_flags & RTCF_LOCAL) {
 621                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 622                                            eiph->daddr, eiph->saddr, 0, 0,
 623                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 624                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
 625                         if (!IS_ERR(rt))
 626                                 ip_rt_put(rt);
 627                         goto out;
 628                 }
 629                 skb_dst_set(skb2, &rt->dst);
 630         } else {
 631                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 632                                    skb2->dev) ||
 633                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 634                         goto out;
 635         }
 636
 637         /* change mtu on this route */
 638         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 639                 if (rel_info > dst_mtu(skb_dst(skb2)))
 640                         goto out;
 641
 642                 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
 643         }
 644
 645         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 646
 647 out:
 648         kfree_skb(skb2);
 649         return 0;
 650 }
 651
 652 static int
 653 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 654            u8 type, u8 code, int offset, __be32 info)
 655 {
 656         __u32 rel_info = ntohl(info);
 657         int err, rel_msg = 0;
 658         u8 rel_type = type;
 659         u8 rel_code = code;
 660
 661         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 662                           &rel_msg, &rel_info, offset);
 663         if (err < 0)
 664                 return err;
 665
 666         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 667                 struct rt6_info *rt;
 668                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 669
 670                 if (!skb2)
 671                         return 0;
 672
 673                 skb_dst_drop(skb2);
 674                 skb_pull(skb2, offset);
 675                 skb_reset_network_header(skb2);
 676
 677                 /* Try to guess incoming interface */
 678                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 679                                 NULL, 0, skb2, 0);
 680
 681                 if (rt && rt->dst.dev)
 682                         skb2->dev = rt->dst.dev;
 683
 684                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 685
 686                 ip6_rt_put(rt);
 687
 688                 kfree_skb(skb2);
 689         }
 690
 691         return 0;
 692 }
 693
 694 static int
 695 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 696             u8 type, u8 code, int offset, __be32 info)
 697 {
 698         __u32 rel_info = ntohl(info);
 699         int err, rel_msg = 0;
 700         u8 rel_type = type;
 701         u8 rel_code = code;
 702
 703         err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
 704                           &rel_msg, &rel_info, offset);
 705         return err;
 706 }
 707
 708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 709                                        const struct ipv6hdr *ipv6h,
 710                                        struct sk_buff *skb)
 711 {
 712         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 713
 714         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 715                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 716
 717         return IP6_ECN_decapsulate(ipv6h, skb);
 718 }
 719
 720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 721                                        const struct ipv6hdr *ipv6h,
 722                                        struct sk_buff *skb)
 723 {
 724         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 725                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 726
 727         return IP6_ECN_decapsulate(ipv6h, skb);
 728 }
 729
 730 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 731                                                const struct ipv6hdr *ipv6h,
 732                                                struct sk_buff *skb)
 733 {
 734         /* ECN is not supported in AF_MPLS */
 735         return 0;
 736 }
 737
 738 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 739                              const struct in6_addr *laddr,
 740                              const struct in6_addr *raddr)
 741 {
 742         struct __ip6_tnl_parm *p = &t->parms;
 743         int ltype = ipv6_addr_type(laddr);
 744         int rtype = ipv6_addr_type(raddr);
 745         __u32 flags = 0;
 746
 747         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 748                 flags = IP6_TNL_F_CAP_PER_PACKET;
 749         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 750                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 751                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 752                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 753                 if (ltype&IPV6_ADDR_UNICAST)
 754                         flags |= IP6_TNL_F_CAP_XMIT;
 755                 if (rtype&IPV6_ADDR_UNICAST)
 756                         flags |= IP6_TNL_F_CAP_RCV;
 757         }
 758         return flags;
 759 }
 760 EXPORT_SYMBOL(ip6_tnl_get_cap);
 761
 762 /* called with rcu_read_lock() */
 763 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 764                                   const struct in6_addr *laddr,
 765                                   const struct in6_addr *raddr)
 766 {
 767         struct __ip6_tnl_parm *p = &t->parms;
 768         int ret = 0;
 769         struct net *net = t->net;
 770
 771         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 772             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 773              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 774                 struct net_device *ldev = NULL;
 775
 776                 if (p->link)
 777                         ldev = dev_get_by_index_rcu(net, p->link);
 778
 779                 if ((ipv6_addr_is_multicast(laddr) ||
 780                      likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
 781                                                     0, IFA_F_TENTATIVE))) &&
 782                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
 783                      likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
 784                                                      0, IFA_F_TENTATIVE))))
 785                         ret = 1;
 786         }
 787         return ret;
 788 }
 789 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 790
 791 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 792                          const struct tnl_ptk_info *tpi,
 793                          struct metadata_dst *tun_dst,
 794                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 795                                                 const struct ipv6hdr *ipv6h,
 796                                                 struct sk_buff *skb),
 797                          bool log_ecn_err)
 798 {
 799         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 800         int err;
 801
 802         if ((!(tpi->flags & TUNNEL_CSUM) &&
 803              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 804             ((tpi->flags & TUNNEL_CSUM) &&
 805              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 806                 DEV_STATS_INC(tunnel->dev, rx_crc_errors);
 807                 DEV_STATS_INC(tunnel->dev, rx_errors);
 808                 goto drop;
 809         }
 810
 811         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 812                 if (!(tpi->flags & TUNNEL_SEQ) ||
 813                     (tunnel->i_seqno &&
 814                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 815                         DEV_STATS_INC(tunnel->dev, rx_fifo_errors);
 816                         DEV_STATS_INC(tunnel->dev, rx_errors);
 817                         goto drop;
 818                 }
 819                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 820         }
 821
 822         skb->protocol = tpi->proto;
 823
 824         /* Warning: All skb pointers will be invalidated! */
 825         if (tunnel->dev->type == ARPHRD_ETHER) {
 826                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 827                         DEV_STATS_INC(tunnel->dev, rx_length_errors);
 828                         DEV_STATS_INC(tunnel->dev, rx_errors);
 829                         goto drop;
 830                 }
 831
 832                 ipv6h = ipv6_hdr(skb);
 833                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 834                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 835         } else {
 836                 skb->dev = tunnel->dev;
 837                 skb_reset_mac_header(skb);
 838         }
 839
 840         skb_reset_network_header(skb);
 841         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 842
 843         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 844
 845         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 846         if (unlikely(err)) {
 847                 if (log_ecn_err)
 848                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 849                                              &ipv6h->saddr,
 850                                              ipv6_get_dsfield(ipv6h));
 851                 if (err > 1) {
 852                         DEV_STATS_INC(tunnel->dev, rx_frame_errors);
 853                         DEV_STATS_INC(tunnel->dev, rx_errors);
 854                         goto drop;
 855                 }
 856         }
 857
 858         dev_sw_netstats_rx_add(tunnel->dev, skb->len);
 859
 860         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 861
 862         if (tun_dst)
 863                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 864
 865         gro_cells_receive(&tunnel->gro_cells, skb);
 866         return 0;
 867
 868 drop:
 869         if (tun_dst)
 870                 dst_release((struct dst_entry *)tun_dst);
 871         kfree_skb(skb);
 872         return 0;
 873 }
 874
 875 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 876                 const struct tnl_ptk_info *tpi,
 877                 struct metadata_dst *tun_dst,
 878                 bool log_ecn_err)
 879 {
 880         int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 881                                     const struct ipv6hdr *ipv6h,
 882                                     struct sk_buff *skb);
 883
 884         dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
 885         if (tpi->proto == htons(ETH_P_IP))
 886                 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
 887
 888         return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 889                              log_ecn_err);
 890 }
 891 EXPORT_SYMBOL(ip6_tnl_rcv);
 892
 893 static const struct tnl_ptk_info tpi_v6 = {
 894         /* no tunnel info required for ipxip6. */
 895         .proto = htons(ETH_P_IPV6),
 896 };
 897
 898 static const struct tnl_ptk_info tpi_v4 = {
 899         /* no tunnel info required for ipxip6. */
 900         .proto = htons(ETH_P_IP),
 901 };
 902
 903 static const struct tnl_ptk_info tpi_mpls = {
 904         /* no tunnel info required for mplsip6. */
 905         .proto = htons(ETH_P_MPLS_UC),
 906 };
 907
 908 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 909                       const struct tnl_ptk_info *tpi,
 910                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 911                                                   const struct ipv6hdr *ipv6h,
 912                                                   struct sk_buff *skb))
 913 {
 914         struct ip6_tnl *t;
 915         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 916         struct metadata_dst *tun_dst = NULL;
 917         int ret = -1;
 918
 919         rcu_read_lock();
 920         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
 921
 922         if (t) {
 923                 u8 tproto = READ_ONCE(t->parms.proto);
 924
 925                 if (tproto != ipproto && tproto != 0)
 926                         goto drop;
 927                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 928                         goto drop;
 929                 ipv6h = ipv6_hdr(skb);
 930                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 931                         goto drop;
 932                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 933                         goto drop;
 934                 if (t->parms.collect_md) {
 935                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 936                         if (!tun_dst)
 937                                 goto drop;
 938                 }
 939                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 940                                     log_ecn_error);
 941         }
 942
 943         rcu_read_unlock();
 944
 945         return ret;
 946
 947 drop:
 948         rcu_read_unlock();
 949         kfree_skb(skb);
 950         return 0;
 951 }
 952
 953 static int ip4ip6_rcv(struct sk_buff *skb)
 954 {
 955         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 956                           ip4ip6_dscp_ecn_decapsulate);
 957 }
 958
 959 static int ip6ip6_rcv(struct sk_buff *skb)
 960 {
 961         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 962                           ip6ip6_dscp_ecn_decapsulate);
 963 }
 964
 965 static int mplsip6_rcv(struct sk_buff *skb)
 966 {
 967         return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
 968                           mplsip6_dscp_ecn_decapsulate);
 969 }
 970
 971 struct ipv6_tel_txoption {
 972         struct ipv6_txoptions ops;
 973         __u8 dst_opt[8];
 974 };
 975
 976 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 977 {
 978         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 979
 980         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 981         opt->dst_opt[3] = 1;
 982         opt->dst_opt[4] = encap_limit;
 983         opt->dst_opt[5] = IPV6_TLV_PADN;
 984         opt->dst_opt[6] = 1;
 985
 986         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 987         opt->ops.opt_nflen = 8;
 988 }
 989
 990 /**
 991  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 992  *   @t: the outgoing tunnel device
 993  *   @hdr: IPv6 header from the incoming packet
 994  *
 995  * Description:
 996  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 997  *   doesn't match source of incoming packet.
 998  *
 999  * Return:
1000  *   1 if conflict,
1001  *   0 else
1002  **/
1003
1004 static inline bool
1005 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1006 {
1007         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1008 }
1009
1010 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1011                      const struct in6_addr *laddr,
1012                      const struct in6_addr *raddr)
1013 {
1014         struct __ip6_tnl_parm *p = &t->parms;
1015         int ret = 0;
1016         struct net *net = t->net;
1017
1018         if (t->parms.collect_md)
1019                 return 1;
1020
1021         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1022             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1023              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1024                 struct net_device *ldev = NULL;
1025
1026                 rcu_read_lock();
1027                 if (p->link)
1028                         ldev = dev_get_by_index_rcu(net, p->link);
1029
1030                 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1031                                                       0, IFA_F_TENTATIVE)))
1032                         pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1033                                             p->name);
1034                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1035                          !ipv6_addr_is_multicast(raddr) &&
1036                          unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1037                                                           true, 0, IFA_F_TENTATIVE)))
1038                         pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1039                                             p->name);
1040                 else
1041                         ret = 1;
1042                 rcu_read_unlock();
1043         }
1044         return ret;
1045 }
1046 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1047
1048 /**
1049  * ip6_tnl_xmit - encapsulate packet and send
1050  *   @skb: the outgoing socket buffer
1051  *   @dev: the outgoing tunnel device
1052  *   @dsfield: dscp code for outer header
1053  *   @fl6: flow of tunneled packet
1054  *   @encap_limit: encapsulation limit
1055  *   @pmtu: Path MTU is stored if packet is too big
1056  *   @proto: next header value
1057  *
1058  * Description:
1059  *   Build new header and do some sanity checks on the packet before sending
1060  *   it.
1061  *
1062  * Return:
1063  *   0 on success
1064  *   -1 fail
1065  *   %-EMSGSIZE message too big. return mtu in this case.
1066  **/
1067
1068 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1069                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1070                  __u8 proto)
1071 {
1072         struct ip6_tnl *t = netdev_priv(dev);
1073         struct net *net = t->net;
1074         struct ipv6hdr *ipv6h;
1075         struct ipv6_tel_txoption opt;
1076         struct dst_entry *dst = NULL, *ndst = NULL;
1077         struct net_device *tdev;
1078         int mtu;
1079         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1080         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1081         unsigned int max_headroom = psh_hlen;
1082         __be16 payload_protocol;
1083         bool use_cache = false;
1084         u8 hop_limit;
1085         int err = -1;
1086
1087         payload_protocol = skb_protocol(skb, true);
1088
1089         if (t->parms.collect_md) {
1090                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1091                 goto route_lookup;
1092         } else {
1093                 hop_limit = t->parms.hop_limit;
1094         }
1095
1096         /* NBMA tunnel */
1097         if (ipv6_addr_any(&t->parms.raddr)) {
1098                 if (payload_protocol == htons(ETH_P_IPV6)) {
1099                         struct in6_addr *addr6;
1100                         struct neighbour *neigh;
1101                         int addr_type;
1102
1103                         if (!skb_dst(skb))
1104                                 goto tx_err_link_failure;
1105
1106                         neigh = dst_neigh_lookup(skb_dst(skb),
1107                                                  &ipv6_hdr(skb)->daddr);
1108                         if (!neigh)
1109                                 goto tx_err_link_failure;
1110
1111                         addr6 = (struct in6_addr *)&neigh->primary_key;
1112                         addr_type = ipv6_addr_type(addr6);
1113
1114                         if (addr_type == IPV6_ADDR_ANY)
1115                                 addr6 = &ipv6_hdr(skb)->daddr;
1116
1117                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1118                         neigh_release(neigh);
1119                 } else if (payload_protocol == htons(ETH_P_IP)) {
1120                         const struct rtable *rt = skb_rtable(skb);
1121
1122                         if (!rt)
1123                                 goto tx_err_link_failure;
1124
1125                         if (rt->rt_gw_family == AF_INET6)
1126                                 memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1127                 }
1128         } else if (t->parms.proto != 0 && !(t->parms.flags &
1129                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1130                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1131                 /* enable the cache only if neither the outer protocol nor the
1132                  * routing decision depends on the current inner header value
1133                  */
1134                 use_cache = true;
1135         }
1136
1137         if (use_cache)
1138                 dst = dst_cache_get(&t->dst_cache);
1139
1140         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1141                 goto tx_err_link_failure;
1142
1143         if (!dst) {
1144 route_lookup:
1145                 /* add dsfield to flowlabel for route lookup */
1146                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1147
1148                 dst = ip6_route_output(net, NULL, fl6);
1149
1150                 if (dst->error)
1151                         goto tx_err_link_failure;
1152                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1153                 if (IS_ERR(dst)) {
1154                         err = PTR_ERR(dst);
1155                         dst = NULL;
1156                         goto tx_err_link_failure;
1157                 }
1158                 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1159                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1160                                        &fl6->daddr, 0, &fl6->saddr))
1161                         goto tx_err_link_failure;
1162                 ndst = dst;
1163         }
1164
1165         tdev = dst->dev;
1166
1167         if (tdev == dev) {
1168                 DEV_STATS_INC(dev, collisions);
1169                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1170                                      t->parms.name);
1171                 goto tx_err_dst_release;
1172         }
1173         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1174         if (encap_limit >= 0) {
1175                 max_headroom += 8;
1176                 mtu -= 8;
1177         }
1178         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1179                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1180
1181         skb_dst_update_pmtu_no_confirm(skb, mtu);
1182         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1183                 *pmtu = mtu;
1184                 err = -EMSGSIZE;
1185                 goto tx_err_dst_release;
1186         }
1187
1188         if (t->err_count > 0) {
1189                 if (time_before(jiffies,
1190                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1191                         t->err_count--;
1192
1193                         dst_link_failure(skb);
1194                 } else {
1195                         t->err_count = 0;
1196                 }
1197         }
1198
1199         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1200
1201         /*
1202          * Okay, now see if we can stuff it in the buffer as-is.
1203          */
1204         max_headroom += LL_RESERVED_SPACE(tdev);
1205
1206         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1207             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1208                 struct sk_buff *new_skb;
1209
1210                 new_skb = skb_realloc_headroom(skb, max_headroom);
1211                 if (!new_skb)
1212                         goto tx_err_dst_release;
1213
1214                 if (skb->sk)
1215                         skb_set_owner_w(new_skb, skb->sk);
1216                 consume_skb(skb);
1217                 skb = new_skb;
1218         }
1219
1220         if (t->parms.collect_md) {
1221                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1222                         goto tx_err_dst_release;
1223         } else {
1224                 if (use_cache && ndst)
1225                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1226         }
1227         skb_dst_set(skb, dst);
1228
1229         if (hop_limit == 0) {
1230                 if (payload_protocol == htons(ETH_P_IP))
1231                         hop_limit = ip_hdr(skb)->ttl;
1232                 else if (payload_protocol == htons(ETH_P_IPV6))
1233                         hop_limit = ipv6_hdr(skb)->hop_limit;
1234                 else
1235                         hop_limit = ip6_dst_hoplimit(dst);
1236         }
1237
1238         /* Calculate max headroom for all the headers and adjust
1239          * needed_headroom if necessary.
1240          */
1241         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1242                         + dst->header_len + t->hlen;
1243         if (max_headroom > READ_ONCE(dev->needed_headroom))
1244                 WRITE_ONCE(dev->needed_headroom, max_headroom);
1245
1246         err = ip6_tnl_encap(skb, t, &proto, fl6);
1247         if (err)
1248                 return err;
1249
1250         if (encap_limit >= 0) {
1251                 init_tel_txopt(&opt, encap_limit);
1252                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1253         }
1254
1255         skb_push(skb, sizeof(struct ipv6hdr));
1256         skb_reset_network_header(skb);
1257         ipv6h = ipv6_hdr(skb);
1258         ip6_flow_hdr(ipv6h, dsfield,
1259                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1260         ipv6h->hop_limit = hop_limit;
1261         ipv6h->nexthdr = proto;
1262         ipv6h->saddr = fl6->saddr;
1263         ipv6h->daddr = fl6->daddr;
1264         ip6tunnel_xmit(NULL, skb, dev);
1265         return 0;
1266 tx_err_link_failure:
1267         DEV_STATS_INC(dev, tx_carrier_errors);
1268         dst_link_failure(skb);
1269 tx_err_dst_release:
1270         dst_release(dst);
1271         return err;
1272 }
1273 EXPORT_SYMBOL(ip6_tnl_xmit);
1274
1275 static inline int
1276 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1277                 u8 protocol)
1278 {
1279         struct ip6_tnl *t = netdev_priv(dev);
1280         struct ipv6hdr *ipv6h;
1281         const struct iphdr  *iph;
1282         int encap_limit = -1;
1283         __u16 offset;
1284         struct flowi6 fl6;
1285         __u8 dsfield, orig_dsfield;
1286         __u32 mtu;
1287         u8 tproto;
1288         int err;
1289
1290         tproto = READ_ONCE(t->parms.proto);
1291         if (tproto != protocol && tproto != 0)
1292                 return -1;
1293
1294         if (t->parms.collect_md) {
1295                 struct ip_tunnel_info *tun_info;
1296                 const struct ip_tunnel_key *key;
1297
1298                 tun_info = skb_tunnel_info(skb);
1299                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1300                              ip_tunnel_info_af(tun_info) != AF_INET6))
1301                         return -1;
1302                 key = &tun_info->key;
1303                 memset(&fl6, 0, sizeof(fl6));
1304                 fl6.flowi6_proto = protocol;
1305                 fl6.saddr = key->u.ipv6.src;
1306                 fl6.daddr = key->u.ipv6.dst;
1307                 fl6.flowlabel = key->label;
1308                 dsfield =  key->tos;
1309                 switch (protocol) {
1310                 case IPPROTO_IPIP:
1311                         iph = ip_hdr(skb);
1312                         orig_dsfield = ipv4_get_dsfield(iph);
1313                         break;
1314                 case IPPROTO_IPV6:
1315                         ipv6h = ipv6_hdr(skb);
1316                         orig_dsfield = ipv6_get_dsfield(ipv6h);
1317                         break;
1318                 default:
1319                         orig_dsfield = dsfield;
1320                         break;
1321                 }
1322         } else {
1323                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1324                         encap_limit = t->parms.encap_limit;
1325                 if (protocol == IPPROTO_IPV6) {
1326                         offset = ip6_tnl_parse_tlv_enc_lim(skb,
1327                                                 skb_network_header(skb));
1328                         /* ip6_tnl_parse_tlv_enc_lim() might have
1329                          * reallocated skb->head
1330                          */
1331                         if (offset > 0) {
1332                                 struct ipv6_tlv_tnl_enc_lim *tel;
1333
1334                                 tel = (void *)&skb_network_header(skb)[offset];
1335                                 if (tel->encap_limit == 0) {
1336                                         icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1337                                                         ICMPV6_HDR_FIELD, offset + 2);
1338                                         return -1;
1339                                 }
1340                                 encap_limit = tel->encap_limit - 1;
1341                         }
1342                 }
1343
1344                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1345                 fl6.flowi6_proto = protocol;
1346
1347                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1348                         fl6.flowi6_mark = skb->mark;
1349                 else
1350                         fl6.flowi6_mark = t->parms.fwmark;
1351                 switch (protocol) {
1352                 case IPPROTO_IPIP:
1353                         iph = ip_hdr(skb);
1354                         orig_dsfield = ipv4_get_dsfield(iph);
1355                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1356                                 dsfield = orig_dsfield;
1357                         else
1358                                 dsfield = ip6_tclass(t->parms.flowinfo);
1359                         break;
1360                 case IPPROTO_IPV6:
1361                         ipv6h = ipv6_hdr(skb);
1362                         orig_dsfield = ipv6_get_dsfield(ipv6h);
1363                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1364                                 dsfield = orig_dsfield;
1365                         else
1366                                 dsfield = ip6_tclass(t->parms.flowinfo);
1367                         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1368                                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1369                         break;
1370                 default:
1371                         orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1372                         break;
1373                 }
1374         }
1375
1376         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1377         dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1378
1379         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1380                 return -1;
1381
1382         skb_set_inner_ipproto(skb, protocol);
1383
1384         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1385                            protocol);
1386         if (err != 0) {
1387                 /* XXX: send ICMP error even if DF is not set. */
1388                 if (err == -EMSGSIZE)
1389                         switch (protocol) {
1390                         case IPPROTO_IPIP:
1391                                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1392                                               ICMP_FRAG_NEEDED, htonl(mtu));
1393                                 break;
1394                         case IPPROTO_IPV6:
1395                                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1396                                 break;
1397                         default:
1398                                 break;
1399                         }
1400                 return -1;
1401         }
1402
1403         return 0;
1404 }
1405
1406 static netdev_tx_t
1407 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1408 {
1409         struct ip6_tnl *t = netdev_priv(dev);
1410         u8 ipproto;
1411         int ret;
1412
1413         if (!pskb_inet_may_pull(skb))
1414                 goto tx_err;
1415
1416         switch (skb->protocol) {
1417         case htons(ETH_P_IP):
1418                 ipproto = IPPROTO_IPIP;
1419                 break;
1420         case htons(ETH_P_IPV6):
1421                 if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1422                         goto tx_err;
1423                 ipproto = IPPROTO_IPV6;
1424                 break;
1425         case htons(ETH_P_MPLS_UC):
1426                 ipproto = IPPROTO_MPLS;
1427                 break;
1428         default:
1429                 goto tx_err;
1430         }
1431
1432         ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1433         if (ret < 0)
1434                 goto tx_err;
1435
1436         return NETDEV_TX_OK;
1437
1438 tx_err:
1439         DEV_STATS_INC(dev, tx_errors);
1440         DEV_STATS_INC(dev, tx_dropped);
1441         kfree_skb(skb);
1442         return NETDEV_TX_OK;
1443 }
1444
1445 static void ip6_tnl_link_config(struct ip6_tnl *t)
1446 {
1447         struct net_device *dev = t->dev;
1448         struct net_device *tdev = NULL;
1449         struct __ip6_tnl_parm *p = &t->parms;
1450         struct flowi6 *fl6 = &t->fl.u.ip6;
1451         int t_hlen;
1452         int mtu;
1453
1454         __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1455         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1456
1457         /* Set up flowi template */
1458         fl6->saddr = p->laddr;
1459         fl6->daddr = p->raddr;
1460         fl6->flowi6_oif = p->link;
1461         fl6->flowlabel = 0;
1462
1463         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1464                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1465         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1466                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1467
1468         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1469         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1470
1471         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1472                 dev->flags |= IFF_POINTOPOINT;
1473         else
1474                 dev->flags &= ~IFF_POINTOPOINT;
1475
1476         t->tun_hlen = 0;
1477         t->hlen = t->encap_hlen + t->tun_hlen;
1478         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1479
1480         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1481                 int strict = (ipv6_addr_type(&p->raddr) &
1482                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1483
1484                 struct rt6_info *rt = rt6_lookup(t->net,
1485                                                  &p->raddr, &p->laddr,
1486                                                  p->link, NULL, strict);
1487                 if (rt) {
1488                         tdev = rt->dst.dev;
1489                         ip6_rt_put(rt);
1490                 }
1491
1492                 if (!tdev && p->link)
1493                         tdev = __dev_get_by_index(t->net, p->link);
1494
1495                 if (tdev) {
1496                         dev->hard_header_len = tdev->hard_header_len + t_hlen;
1497                         mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1498
1499                         mtu = mtu - t_hlen;
1500                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1501                                 mtu -= 8;
1502
1503                         if (mtu < IPV6_MIN_MTU)
1504                                 mtu = IPV6_MIN_MTU;
1505                         WRITE_ONCE(dev->mtu, mtu);
1506                 }
1507         }
1508 }
1509
1510 /**
1511  * ip6_tnl_change - update the tunnel parameters
1512  *   @t: tunnel to be changed
1513  *   @p: tunnel configuration parameters
1514  *
1515  * Description:
1516  *   ip6_tnl_change() updates the tunnel parameters
1517  **/
1518
1519 static void
1520 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1521 {
1522         t->parms.laddr = p->laddr;
1523         t->parms.raddr = p->raddr;
1524         t->parms.flags = p->flags;
1525         t->parms.hop_limit = p->hop_limit;
1526         t->parms.encap_limit = p->encap_limit;
1527         t->parms.flowinfo = p->flowinfo;
1528         t->parms.link = p->link;
1529         t->parms.proto = p->proto;
1530         t->parms.fwmark = p->fwmark;
1531         dst_cache_reset(&t->dst_cache);
1532         ip6_tnl_link_config(t);
1533 }
1534
1535 static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1536 {
1537         struct net *net = t->net;
1538         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1539
1540         ip6_tnl_unlink(ip6n, t);
1541         synchronize_net();
1542         ip6_tnl_change(t, p);
1543         ip6_tnl_link(ip6n, t);
1544         netdev_state_change(t->dev);
1545 }
1546
1547 static void ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1548 {
1549         /* for default tnl0 device allow to change only the proto */
1550         t->parms.proto = p->proto;
1551         netdev_state_change(t->dev);
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_siocdevprivate - configure ipv6 tunnels from userspace
1584  *   @dev: virtual device associated with tunnel
1585  *   @ifr: unused
1586  *   @data: parameters passed from userspace
1587  *   @cmd: command to be performed
1588  *
1589  * Description:
1590  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1591  *   from userspace.
1592  *
1593  *   The possible commands are the following:
1594  *     %SIOCGETTUNNEL: get tunnel parameters for device
1595  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1596  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1597  *     %SIOCDELTUNNEL: delete tunnel
1598  *
1599  *   The fallback device "ip6tnl0", created during module
1600  *   initialization, can be used for creating other tunnel devices.
1601  *
1602  * Return:
1603  *   0 on success,
1604  *   %-EFAULT if unable to copy data to or from userspace,
1605  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1606  *   %-EINVAL if passed tunnel parameters are invalid,
1607  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1608  *   %-ENODEV if attempting to change or delete a nonexisting device
1609  **/
1610
1611 static int
1612 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1613                        void __user *data, int cmd)
1614 {
1615         int err = 0;
1616         struct ip6_tnl_parm p;
1617         struct __ip6_tnl_parm p1;
1618         struct ip6_tnl *t = netdev_priv(dev);
1619         struct net *net = t->net;
1620         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1621
1622         memset(&p1, 0, sizeof(p1));
1623
1624         switch (cmd) {
1625         case SIOCGETTUNNEL:
1626                 if (dev == ip6n->fb_tnl_dev) {
1627                         if (copy_from_user(&p, data, sizeof(p))) {
1628                                 err = -EFAULT;
1629                                 break;
1630                         }
1631                         ip6_tnl_parm_from_user(&p1, &p);
1632                         t = ip6_tnl_locate(net, &p1, 0);
1633                         if (IS_ERR(t))
1634                                 t = netdev_priv(dev);
1635                 } else {
1636                         memset(&p, 0, sizeof(p));
1637                 }
1638                 ip6_tnl_parm_to_user(&p, &t->parms);
1639                 if (copy_to_user(data, &p, sizeof(p)))
1640                         err = -EFAULT;
1641                 break;
1642         case SIOCADDTUNNEL:
1643         case SIOCCHGTUNNEL:
1644                 err = -EPERM;
1645                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1646                         break;
1647                 err = -EFAULT;
1648                 if (copy_from_user(&p, data, sizeof(p)))
1649                         break;
1650                 err = -EINVAL;
1651                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1652                     p.proto != 0)
1653                         break;
1654                 ip6_tnl_parm_from_user(&p1, &p);
1655                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1656                 if (cmd == SIOCCHGTUNNEL) {
1657                         if (!IS_ERR(t)) {
1658                                 if (t->dev != dev) {
1659                                         err = -EEXIST;
1660                                         break;
1661                                 }
1662                         } else
1663                                 t = netdev_priv(dev);
1664                         if (dev == ip6n->fb_tnl_dev)
1665                                 ip6_tnl0_update(t, &p1);
1666                         else
1667                                 ip6_tnl_update(t, &p1);
1668                 }
1669                 if (!IS_ERR(t)) {
1670                         err = 0;
1671                         ip6_tnl_parm_to_user(&p, &t->parms);
1672                         if (copy_to_user(data, &p, sizeof(p)))
1673                                 err = -EFAULT;
1674
1675                 } else {
1676                         err = PTR_ERR(t);
1677                 }
1678                 break;
1679         case SIOCDELTUNNEL:
1680                 err = -EPERM;
1681                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1682                         break;
1683
1684                 if (dev == ip6n->fb_tnl_dev) {
1685                         err = -EFAULT;
1686                         if (copy_from_user(&p, data, sizeof(p)))
1687                                 break;
1688                         err = -ENOENT;
1689                         ip6_tnl_parm_from_user(&p1, &p);
1690                         t = ip6_tnl_locate(net, &p1, 0);
1691                         if (IS_ERR(t))
1692                                 break;
1693                         err = -EPERM;
1694                         if (t->dev == ip6n->fb_tnl_dev)
1695                                 break;
1696                         dev = t->dev;
1697                 }
1698                 err = 0;
1699                 unregister_netdevice(dev);
1700                 break;
1701         default:
1702                 err = -EINVAL;
1703         }
1704         return err;
1705 }
1706
1707 /**
1708  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1709  *   @dev: virtual device associated with tunnel
1710  *   @new_mtu: the new mtu
1711  *
1712  * Return:
1713  *   0 on success,
1714  *   %-EINVAL if mtu too small
1715  **/
1716
1717 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1718 {
1719         struct ip6_tnl *tnl = netdev_priv(dev);
1720
1721         if (tnl->parms.proto == IPPROTO_IPV6) {
1722                 if (new_mtu < IPV6_MIN_MTU)
1723                         return -EINVAL;
1724         } else {
1725                 if (new_mtu < ETH_MIN_MTU)
1726                         return -EINVAL;
1727         }
1728         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1729                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1730                         return -EINVAL;
1731         } else {
1732                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1733                         return -EINVAL;
1734         }
1735         dev->mtu = new_mtu;
1736         return 0;
1737 }
1738 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1739
1740 int ip6_tnl_get_iflink(const struct net_device *dev)
1741 {
1742         struct ip6_tnl *t = netdev_priv(dev);
1743
1744         return t->parms.link;
1745 }
1746 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1747
1748 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1749                           unsigned int num)
1750 {
1751         if (num >= MAX_IPTUN_ENCAP_OPS)
1752                 return -ERANGE;
1753
1754         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1755                         &ip6tun_encaps[num],
1756                         NULL, ops) ? 0 : -1;
1757 }
1758 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1759
1760 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1761                           unsigned int num)
1762 {
1763         int ret;
1764
1765         if (num >= MAX_IPTUN_ENCAP_OPS)
1766                 return -ERANGE;
1767
1768         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1769                        &ip6tun_encaps[num],
1770                        ops, NULL) == ops) ? 0 : -1;
1771
1772         synchronize_net();
1773
1774         return ret;
1775 }
1776 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1777
1778 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1779                         struct ip_tunnel_encap *ipencap)
1780 {
1781         int hlen;
1782
1783         memset(&t->encap, 0, sizeof(t->encap));
1784
1785         hlen = ip6_encap_hlen(ipencap);
1786         if (hlen < 0)
1787                 return hlen;
1788
1789         t->encap.type = ipencap->type;
1790         t->encap.sport = ipencap->sport;
1791         t->encap.dport = ipencap->dport;
1792         t->encap.flags = ipencap->flags;
1793
1794         t->encap_hlen = hlen;
1795         t->hlen = t->encap_hlen + t->tun_hlen;
1796
1797         return 0;
1798 }
1799 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1800
1801 static const struct net_device_ops ip6_tnl_netdev_ops = {
1802         .ndo_init       = ip6_tnl_dev_init,
1803         .ndo_uninit     = ip6_tnl_dev_uninit,
1804         .ndo_start_xmit = ip6_tnl_start_xmit,
1805         .ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1806         .ndo_change_mtu = ip6_tnl_change_mtu,
1807         .ndo_get_stats64 = dev_get_tstats64,
1808         .ndo_get_iflink = ip6_tnl_get_iflink,
1809 };
1810
1811 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1812                          NETIF_F_FRAGLIST |     \
1813                          NETIF_F_HIGHDMA |      \
1814                          NETIF_F_GSO_SOFTWARE | \
1815                          NETIF_F_HW_CSUM)
1816
1817 /**
1818  * ip6_tnl_dev_setup - setup virtual tunnel device
1819  *   @dev: virtual device associated with tunnel
1820  *
1821  * Description:
1822  *   Initialize function pointers and device parameters
1823  **/
1824
1825 static void ip6_tnl_dev_setup(struct net_device *dev)
1826 {
1827         dev->netdev_ops = &ip6_tnl_netdev_ops;
1828         dev->header_ops = &ip_tunnel_header_ops;
1829         dev->needs_free_netdev = true;
1830         dev->priv_destructor = ip6_dev_free;
1831
1832         dev->type = ARPHRD_TUNNEL6;
1833         dev->flags |= IFF_NOARP;
1834         dev->addr_len = sizeof(struct in6_addr);
1835         dev->features |= NETIF_F_LLTX;
1836         netif_keep_dst(dev);
1837
1838         dev->features           |= IPXIPX_FEATURES;
1839         dev->hw_features        |= IPXIPX_FEATURES;
1840
1841         /* This perm addr will be used as interface identifier by IPv6 */
1842         dev->addr_assign_type = NET_ADDR_RANDOM;
1843         eth_random_addr(dev->perm_addr);
1844 }
1845
1846
1847 /**
1848  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1849  *   @dev: virtual device associated with tunnel
1850  **/
1851
1852 static inline int
1853 ip6_tnl_dev_init_gen(struct net_device *dev)
1854 {
1855         struct ip6_tnl *t = netdev_priv(dev);
1856         int ret;
1857         int t_hlen;
1858
1859         t->dev = dev;
1860         t->net = dev_net(dev);
1861         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1862         if (!dev->tstats)
1863                 return -ENOMEM;
1864
1865         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1866         if (ret)
1867                 goto free_stats;
1868
1869         ret = gro_cells_init(&t->gro_cells, dev);
1870         if (ret)
1871                 goto destroy_dst;
1872
1873         t->tun_hlen = 0;
1874         t->hlen = t->encap_hlen + t->tun_hlen;
1875         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1876
1877         dev->type = ARPHRD_TUNNEL6;
1878         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1879         dev->mtu = ETH_DATA_LEN - t_hlen;
1880         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1881                 dev->mtu -= 8;
1882         dev->min_mtu = ETH_MIN_MTU;
1883         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1884
1885         netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
1886         return 0;
1887
1888 destroy_dst:
1889         dst_cache_destroy(&t->dst_cache);
1890 free_stats:
1891         free_percpu(dev->tstats);
1892         dev->tstats = NULL;
1893
1894         return ret;
1895 }
1896
1897 /**
1898  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1899  *   @dev: virtual device associated with tunnel
1900  **/
1901
1902 static int ip6_tnl_dev_init(struct net_device *dev)
1903 {
1904         struct ip6_tnl *t = netdev_priv(dev);
1905         int err = ip6_tnl_dev_init_gen(dev);
1906
1907         if (err)
1908                 return err;
1909         ip6_tnl_link_config(t);
1910         if (t->parms.collect_md)
1911                 netif_keep_dst(dev);
1912         return 0;
1913 }
1914
1915 /**
1916  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1917  *   @dev: fallback device
1918  *
1919  * Return: 0
1920  **/
1921
1922 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1923 {
1924         struct ip6_tnl *t = netdev_priv(dev);
1925         struct net *net = dev_net(dev);
1926         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1927
1928         t->parms.proto = IPPROTO_IPV6;
1929
1930         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1931         return 0;
1932 }
1933
1934 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1935                             struct netlink_ext_ack *extack)
1936 {
1937         u8 proto;
1938
1939         if (!data || !data[IFLA_IPTUN_PROTO])
1940                 return 0;
1941
1942         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1943         if (proto != IPPROTO_IPV6 &&
1944             proto != IPPROTO_IPIP &&
1945             proto != 0)
1946                 return -EINVAL;
1947
1948         return 0;
1949 }
1950
1951 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1952                                   struct __ip6_tnl_parm *parms)
1953 {
1954         memset(parms, 0, sizeof(*parms));
1955
1956         if (!data)
1957                 return;
1958
1959         if (data[IFLA_IPTUN_LINK])
1960                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1961
1962         if (data[IFLA_IPTUN_LOCAL])
1963                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1964
1965         if (data[IFLA_IPTUN_REMOTE])
1966                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1967
1968         if (data[IFLA_IPTUN_TTL])
1969                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1970
1971         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1972                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1973
1974         if (data[IFLA_IPTUN_FLOWINFO])
1975                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1976
1977         if (data[IFLA_IPTUN_FLAGS])
1978                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1979
1980         if (data[IFLA_IPTUN_PROTO])
1981                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1982
1983         if (data[IFLA_IPTUN_COLLECT_METADATA])
1984                 parms->collect_md = true;
1985
1986         if (data[IFLA_IPTUN_FWMARK])
1987                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1988 }
1989
1990 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1991                            struct nlattr *tb[], struct nlattr *data[],
1992                            struct netlink_ext_ack *extack)
1993 {
1994         struct net *net = dev_net(dev);
1995         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1996         struct ip_tunnel_encap ipencap;
1997         struct ip6_tnl *nt, *t;
1998         int err;
1999
2000         nt = netdev_priv(dev);
2001
2002         if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2003                 err = ip6_tnl_encap_setup(nt, &ipencap);
2004                 if (err < 0)
2005                         return err;
2006         }
2007
2008         ip6_tnl_netlink_parms(data, &nt->parms);
2009
2010         if (nt->parms.collect_md) {
2011                 if (rtnl_dereference(ip6n->collect_md_tun))
2012                         return -EEXIST;
2013         } else {
2014                 t = ip6_tnl_locate(net, &nt->parms, 0);
2015                 if (!IS_ERR(t))
2016                         return -EEXIST;
2017         }
2018
2019         err = ip6_tnl_create2(dev);
2020         if (!err && tb[IFLA_MTU])
2021                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2022
2023         return err;
2024 }
2025
2026 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2027                               struct nlattr *data[],
2028                               struct netlink_ext_ack *extack)
2029 {
2030         struct ip6_tnl *t = netdev_priv(dev);
2031         struct __ip6_tnl_parm p;
2032         struct net *net = t->net;
2033         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2034         struct ip_tunnel_encap ipencap;
2035
2036         if (dev == ip6n->fb_tnl_dev)
2037                 return -EINVAL;
2038
2039         if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2040                 int err = ip6_tnl_encap_setup(t, &ipencap);
2041
2042                 if (err < 0)
2043                         return err;
2044         }
2045         ip6_tnl_netlink_parms(data, &p);
2046         if (p.collect_md)
2047                 return -EINVAL;
2048
2049         t = ip6_tnl_locate(net, &p, 0);
2050         if (!IS_ERR(t)) {
2051                 if (t->dev != dev)
2052                         return -EEXIST;
2053         } else
2054                 t = netdev_priv(dev);
2055
2056         ip6_tnl_update(t, &p);
2057         return 0;
2058 }
2059
2060 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2061 {
2062         struct net *net = dev_net(dev);
2063         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2064
2065         if (dev != ip6n->fb_tnl_dev)
2066                 unregister_netdevice_queue(dev, head);
2067 }
2068
2069 static size_t ip6_tnl_get_size(const struct net_device *dev)
2070 {
2071         return
2072                 /* IFLA_IPTUN_LINK */
2073                 nla_total_size(4) +
2074                 /* IFLA_IPTUN_LOCAL */
2075                 nla_total_size(sizeof(struct in6_addr)) +
2076                 /* IFLA_IPTUN_REMOTE */
2077                 nla_total_size(sizeof(struct in6_addr)) +
2078                 /* IFLA_IPTUN_TTL */
2079                 nla_total_size(1) +
2080                 /* IFLA_IPTUN_ENCAP_LIMIT */
2081                 nla_total_size(1) +
2082                 /* IFLA_IPTUN_FLOWINFO */
2083                 nla_total_size(4) +
2084                 /* IFLA_IPTUN_FLAGS */
2085                 nla_total_size(4) +
2086                 /* IFLA_IPTUN_PROTO */
2087                 nla_total_size(1) +
2088                 /* IFLA_IPTUN_ENCAP_TYPE */
2089                 nla_total_size(2) +
2090                 /* IFLA_IPTUN_ENCAP_FLAGS */
2091                 nla_total_size(2) +
2092                 /* IFLA_IPTUN_ENCAP_SPORT */
2093                 nla_total_size(2) +
2094                 /* IFLA_IPTUN_ENCAP_DPORT */
2095                 nla_total_size(2) +
2096                 /* IFLA_IPTUN_COLLECT_METADATA */
2097                 nla_total_size(0) +
2098                 /* IFLA_IPTUN_FWMARK */
2099                 nla_total_size(4) +
2100                 0;
2101 }
2102
2103 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2104 {
2105         struct ip6_tnl *tunnel = netdev_priv(dev);
2106         struct __ip6_tnl_parm *parm = &tunnel->parms;
2107
2108         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2109             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2110             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2111             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2112             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2113             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2114             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2115             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2116             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2117                 goto nla_put_failure;
2118
2119         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2120             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2121             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2122             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2123                 goto nla_put_failure;
2124
2125         if (parm->collect_md)
2126                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2127                         goto nla_put_failure;
2128
2129         return 0;
2130
2131 nla_put_failure:
2132         return -EMSGSIZE;
2133 }
2134
2135 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2136 {
2137         struct ip6_tnl *tunnel = netdev_priv(dev);
2138
2139         return tunnel->net;
2140 }
2141 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2142
2143 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2144         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2145         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2146         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2147         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2148         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2149         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2150         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2151         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2152         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2153         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2154         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2155         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2156         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2157         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2158 };
2159
2160 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2161         .kind           = "ip6tnl",
2162         .maxtype        = IFLA_IPTUN_MAX,
2163         .policy         = ip6_tnl_policy,
2164         .priv_size      = sizeof(struct ip6_tnl),
2165         .setup          = ip6_tnl_dev_setup,
2166         .validate       = ip6_tnl_validate,
2167         .newlink        = ip6_tnl_newlink,
2168         .changelink     = ip6_tnl_changelink,
2169         .dellink        = ip6_tnl_dellink,
2170         .get_size       = ip6_tnl_get_size,
2171         .fill_info      = ip6_tnl_fill_info,
2172         .get_link_net   = ip6_tnl_get_link_net,
2173 };
2174
2175 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2176         .handler        = ip4ip6_rcv,
2177         .err_handler    = ip4ip6_err,
2178         .priority       =       1,
2179 };
2180
2181 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2182         .handler        = ip6ip6_rcv,
2183         .err_handler    = ip6ip6_err,
2184         .priority       =       1,
2185 };
2186
2187 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2188         .handler        = mplsip6_rcv,
2189         .err_handler    = mplsip6_err,
2190         .priority       =       1,
2191 };
2192
2193 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2194 {
2195         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2196         struct net_device *dev, *aux;
2197         int h;
2198         struct ip6_tnl *t;
2199
2200         for_each_netdev_safe(net, dev, aux)
2201                 if (dev->rtnl_link_ops == &ip6_link_ops)
2202                         unregister_netdevice_queue(dev, list);
2203
2204         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2205                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2206                 while (t) {
2207                         /* If dev is in the same netns, it has already
2208                          * been added to the list by the previous loop.
2209                          */
2210                         if (!net_eq(dev_net(t->dev), net))
2211                                 unregister_netdevice_queue(t->dev, list);
2212                         t = rtnl_dereference(t->next);
2213                 }
2214         }
2215
2216         t = rtnl_dereference(ip6n->tnls_wc[0]);
2217         while (t) {
2218                 /* If dev is in the same netns, it has already
2219                  * been added to the list by the previous loop.
2220                  */
2221                 if (!net_eq(dev_net(t->dev), net))
2222                         unregister_netdevice_queue(t->dev, list);
2223                 t = rtnl_dereference(t->next);
2224         }
2225 }
2226
2227 static int __net_init ip6_tnl_init_net(struct net *net)
2228 {
2229         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2230         struct ip6_tnl *t = NULL;
2231         int err;
2232
2233         ip6n->tnls[0] = ip6n->tnls_wc;
2234         ip6n->tnls[1] = ip6n->tnls_r_l;
2235
2236         if (!net_has_fallback_tunnels(net))
2237                 return 0;
2238         err = -ENOMEM;
2239         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2240                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2241
2242         if (!ip6n->fb_tnl_dev)
2243                 goto err_alloc_dev;
2244         dev_net_set(ip6n->fb_tnl_dev, net);
2245         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2246         /* FB netdevice is special: we have one, and only one per netns.
2247          * Allowing to move it to another netns is clearly unsafe.
2248          */
2249         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2250
2251         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2252         if (err < 0)
2253                 goto err_register;
2254
2255         err = register_netdev(ip6n->fb_tnl_dev);
2256         if (err < 0)
2257                 goto err_register;
2258
2259         t = netdev_priv(ip6n->fb_tnl_dev);
2260
2261         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2262         return 0;
2263
2264 err_register:
2265         free_netdev(ip6n->fb_tnl_dev);
2266 err_alloc_dev:
2267         return err;
2268 }
2269
2270 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2271 {
2272         struct net *net;
2273         LIST_HEAD(list);
2274
2275         rtnl_lock();
2276         list_for_each_entry(net, net_list, exit_list)
2277                 ip6_tnl_destroy_tunnels(net, &list);
2278         unregister_netdevice_many(&list);
2279         rtnl_unlock();
2280 }
2281
2282 static struct pernet_operations ip6_tnl_net_ops = {
2283         .init = ip6_tnl_init_net,
2284         .exit_batch = ip6_tnl_exit_batch_net,
2285         .id   = &ip6_tnl_net_id,
2286         .size = sizeof(struct ip6_tnl_net),
2287 };
2288
2289 /**
2290  * ip6_tunnel_init - register protocol and reserve needed resources
2291  *
2292  * Return: 0 on success
2293  **/
2294
2295 static int __init ip6_tunnel_init(void)
2296 {
2297         int  err;
2298
2299         if (!ipv6_mod_enabled())
2300                 return -EOPNOTSUPP;
2301
2302         err = register_pernet_device(&ip6_tnl_net_ops);
2303         if (err < 0)
2304                 goto out_pernet;
2305
2306         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2307         if (err < 0) {
2308                 pr_err("%s: can't register ip4ip6\n", __func__);
2309                 goto out_ip4ip6;
2310         }
2311
2312         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2313         if (err < 0) {
2314                 pr_err("%s: can't register ip6ip6\n", __func__);
2315                 goto out_ip6ip6;
2316         }
2317
2318         if (ip6_tnl_mpls_supported()) {
2319                 err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2320                 if (err < 0) {
2321                         pr_err("%s: can't register mplsip6\n", __func__);
2322                         goto out_mplsip6;
2323                 }
2324         }
2325
2326         err = rtnl_link_register(&ip6_link_ops);
2327         if (err < 0)
2328                 goto rtnl_link_failed;
2329
2330         return 0;
2331
2332 rtnl_link_failed:
2333         if (ip6_tnl_mpls_supported())
2334                 xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2335 out_mplsip6:
2336         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2337 out_ip6ip6:
2338         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2339 out_ip4ip6:
2340         unregister_pernet_device(&ip6_tnl_net_ops);
2341 out_pernet:
2342         return err;
2343 }
2344
2345 /**
2346  * ip6_tunnel_cleanup - free resources and unregister protocol
2347  **/
2348
2349 static void __exit ip6_tunnel_cleanup(void)
2350 {
2351         rtnl_link_unregister(&ip6_link_ops);
2352         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2353                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2354
2355         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2356                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2357
2358         if (ip6_tnl_mpls_supported() &&
2359             xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2360                 pr_info("%s: can't deregister mplsip6\n", __func__);
2361         unregister_pernet_device(&ip6_tnl_net_ops);
2362 }
2363
2364 module_init(ip6_tunnel_init);
2365 module_exit(ip6_tunnel_cleanup);