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 struct net_device_stats *ip6_get_stats(struct net_device *dev)
  93 {
  94         struct pcpu_sw_netstats tmp, sum = { 0 };
  95         int i;
  96
  97         for_each_possible_cpu(i) {
  98                 unsigned int start;
  99                 const struct pcpu_sw_netstats *tstats =
 100                                                    per_cpu_ptr(dev->tstats, i);
 101
 102                 do {
 103                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 104                         tmp.rx_packets = tstats->rx_packets;
 105                         tmp.rx_bytes = tstats->rx_bytes;
 106                         tmp.tx_packets = tstats->tx_packets;
 107                         tmp.tx_bytes =  tstats->tx_bytes;
 108                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 109
 110                 sum.rx_packets += tmp.rx_packets;
 111                 sum.rx_bytes   += tmp.rx_bytes;
 112                 sum.tx_packets += tmp.tx_packets;
 113                 sum.tx_bytes   += tmp.tx_bytes;
 114         }
 115         dev->stats.rx_packets = sum.rx_packets;
 116         dev->stats.rx_bytes   = sum.rx_bytes;
 117         dev->stats.tx_packets = sum.tx_packets;
 118         dev->stats.tx_bytes   = sum.tx_bytes;
 119         return &dev->stats;
 120 }
 121
 122 /**
 123  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 124  *   @link: ifindex of underlying interface
 125  *   @remote: the address of the tunnel exit-point
 126  *   @local: the address of the tunnel entry-point
 127  *
 128  * Return:
 129  *   tunnel matching given end-points if found,
 130  *   else fallback tunnel if its device is up,
 131  *   else %NULL
 132  **/
 133
 134 #define for_each_ip6_tunnel_rcu(start) \
 135         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 136
 137 static struct ip6_tnl *
 138 ip6_tnl_lookup(struct net *net, int link,
 139                const struct in6_addr *remote, const struct in6_addr *local)
 140 {
 141         unsigned int hash = HASH(remote, local);
 142         struct ip6_tnl *t, *cand = NULL;
 143         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 144         struct in6_addr any;
 145
 146         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 147                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 148                     !ipv6_addr_equal(remote, &t->parms.raddr) ||
 149                     !(t->dev->flags & IFF_UP))
 150                         continue;
 151
 152                 if (link == t->parms.link)
 153                         return t;
 154                 else
 155                         cand = t;
 156         }
 157
 158         memset(&any, 0, sizeof(any));
 159         hash = HASH(&any, local);
 160         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 161                 if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 162                     !ipv6_addr_any(&t->parms.raddr) ||
 163                     !(t->dev->flags & IFF_UP))
 164                         continue;
 165
 166                 if (link == t->parms.link)
 167                         return t;
 168                 else if (!cand)
 169                         cand = t;
 170         }
 171
 172         hash = HASH(remote, &any);
 173         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 174                 if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 175                     !ipv6_addr_any(&t->parms.laddr) ||
 176                     !(t->dev->flags & IFF_UP))
 177                         continue;
 178
 179                 if (link == t->parms.link)
 180                         return t;
 181                 else if (!cand)
 182                         cand = t;
 183         }
 184
 185         if (cand)
 186                 return cand;
 187
 188         t = rcu_dereference(ip6n->collect_md_tun);
 189         if (t && t->dev->flags & IFF_UP)
 190                 return t;
 191
 192         t = rcu_dereference(ip6n->tnls_wc[0]);
 193         if (t && (t->dev->flags & IFF_UP))
 194                 return t;
 195
 196         return NULL;
 197 }
 198
 199 /**
 200  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 201  *   @p: parameters containing tunnel end-points
 202  *
 203  * Description:
 204  *   ip6_tnl_bucket() returns the head of the list matching the
 205  *   &struct in6_addr entries laddr and raddr in @p.
 206  *
 207  * Return: head of IPv6 tunnel list
 208  **/
 209
 210 static struct ip6_tnl __rcu **
 211 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 212 {
 213         const struct in6_addr *remote = &p->raddr;
 214         const struct in6_addr *local = &p->laddr;
 215         unsigned int h = 0;
 216         int prio = 0;
 217
 218         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 219                 prio = 1;
 220                 h = HASH(remote, local);
 221         }
 222         return &ip6n->tnls[prio][h];
 223 }
 224
 225 /**
 226  * ip6_tnl_link - add tunnel to hash table
 227  *   @t: tunnel to be added
 228  **/
 229
 230 static void
 231 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 232 {
 233         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 234
 235         if (t->parms.collect_md)
 236                 rcu_assign_pointer(ip6n->collect_md_tun, t);
 237         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 238         rcu_assign_pointer(*tp, t);
 239 }
 240
 241 /**
 242  * ip6_tnl_unlink - remove tunnel from hash table
 243  *   @t: tunnel to be removed
 244  **/
 245
 246 static void
 247 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 248 {
 249         struct ip6_tnl __rcu **tp;
 250         struct ip6_tnl *iter;
 251
 252         if (t->parms.collect_md)
 253                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 254
 255         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 256              (iter = rtnl_dereference(*tp)) != NULL;
 257              tp = &iter->next) {
 258                 if (t == iter) {
 259                         rcu_assign_pointer(*tp, t->next);
 260                         break;
 261                 }
 262         }
 263 }
 264
 265 static void ip6_dev_free(struct net_device *dev)
 266 {
 267         struct ip6_tnl *t = netdev_priv(dev);
 268
 269         gro_cells_destroy(&t->gro_cells);
 270         dst_cache_destroy(&t->dst_cache);
 271         free_percpu(dev->tstats);
 272 }
 273
 274 static int ip6_tnl_create2(struct net_device *dev)
 275 {
 276         struct ip6_tnl *t = netdev_priv(dev);
 277         struct net *net = dev_net(dev);
 278         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 279         int err;
 280
 281         t = netdev_priv(dev);
 282
 283         dev->rtnl_link_ops = &ip6_link_ops;
 284         err = register_netdevice(dev);
 285         if (err < 0)
 286                 goto out;
 287
 288         strcpy(t->parms.name, dev->name);
 289
 290         dev_hold(dev);
 291         ip6_tnl_link(ip6n, t);
 292         return 0;
 293
 294 out:
 295         return err;
 296 }
 297
 298 /**
 299  * ip6_tnl_create - create a new tunnel
 300  *   @p: tunnel parameters
 301  *   @pt: pointer to new tunnel
 302  *
 303  * Description:
 304  *   Create tunnel matching given parameters.
 305  *
 306  * Return:
 307  *   created tunnel or error pointer
 308  **/
 309
 310 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 311 {
 312         struct net_device *dev;
 313         struct ip6_tnl *t;
 314         char name[IFNAMSIZ];
 315         int err = -E2BIG;
 316
 317         if (p->name[0]) {
 318                 if (!dev_valid_name(p->name))
 319                         goto failed;
 320                 strlcpy(name, p->name, IFNAMSIZ);
 321         } else {
 322                 sprintf(name, "ip6tnl%%d");
 323         }
 324         err = -ENOMEM;
 325         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 326                            ip6_tnl_dev_setup);
 327         if (!dev)
 328                 goto failed;
 329
 330         dev_net_set(dev, net);
 331
 332         t = netdev_priv(dev);
 333         t->parms = *p;
 334         t->net = dev_net(dev);
 335         err = ip6_tnl_create2(dev);
 336         if (err < 0)
 337                 goto failed_free;
 338
 339         return t;
 340
 341 failed_free:
 342         free_netdev(dev);
 343 failed:
 344         return ERR_PTR(err);
 345 }
 346
 347 /**
 348  * ip6_tnl_locate - find or create tunnel matching given parameters
 349  *   @p: tunnel parameters
 350  *   @create: != 0 if allowed to create new tunnel if no match found
 351  *
 352  * Description:
 353  *   ip6_tnl_locate() first tries to locate an existing tunnel
 354  *   based on @parms. If this is unsuccessful, but @create is set a new
 355  *   tunnel device is created and registered for use.
 356  *
 357  * Return:
 358  *   matching tunnel or error pointer
 359  **/
 360
 361 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 362                 struct __ip6_tnl_parm *p, int create)
 363 {
 364         const struct in6_addr *remote = &p->raddr;
 365         const struct in6_addr *local = &p->laddr;
 366         struct ip6_tnl __rcu **tp;
 367         struct ip6_tnl *t;
 368         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 369
 370         for (tp = ip6_tnl_bucket(ip6n, p);
 371              (t = rtnl_dereference(*tp)) != NULL;
 372              tp = &t->next) {
 373                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 374                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 375                     p->link == t->parms.link) {
 376                         if (create)
 377                                 return ERR_PTR(-EEXIST);
 378
 379                         return t;
 380                 }
 381         }
 382         if (!create)
 383                 return ERR_PTR(-ENODEV);
 384         return ip6_tnl_create(net, p);
 385 }
 386
 387 /**
 388  * ip6_tnl_dev_uninit - tunnel device uninitializer
 389  *   @dev: the device to be destroyed
 390  *
 391  * Description:
 392  *   ip6_tnl_dev_uninit() removes tunnel from its list
 393  **/
 394
 395 static void
 396 ip6_tnl_dev_uninit(struct net_device *dev)
 397 {
 398         struct ip6_tnl *t = netdev_priv(dev);
 399         struct net *net = t->net;
 400         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 401
 402         if (dev == ip6n->fb_tnl_dev)
 403                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 404         else
 405                 ip6_tnl_unlink(ip6n, t);
 406         dst_cache_reset(&t->dst_cache);
 407         dev_put(dev);
 408 }
 409
 410 /**
 411  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 412  *   @skb: received socket buffer
 413  *
 414  * Return:
 415  *   0 if none was found,
 416  *   else index to encapsulation limit
 417  **/
 418
 419 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 420 {
 421         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 422         unsigned int nhoff = raw - skb->data;
 423         unsigned int off = nhoff + sizeof(*ipv6h);
 424         u8 next, nexthdr = ipv6h->nexthdr;
 425
 426         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 427                 struct ipv6_opt_hdr *hdr;
 428                 u16 optlen;
 429
 430                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 431                         break;
 432
 433                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 434                 if (nexthdr == NEXTHDR_FRAGMENT) {
 435                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 436                         if (frag_hdr->frag_off)
 437                                 break;
 438                         optlen = 8;
 439                 } else if (nexthdr == NEXTHDR_AUTH) {
 440                         optlen = ipv6_authlen(hdr);
 441                 } else {
 442                         optlen = ipv6_optlen(hdr);
 443                 }
 444                 /* cache hdr->nexthdr, since pskb_may_pull() might
 445                  * invalidate hdr
 446                  */
 447                 next = hdr->nexthdr;
 448                 if (nexthdr == NEXTHDR_DEST) {
 449                         u16 i = 2;
 450
 451                         /* Remember : hdr is no longer valid at this point. */
 452                         if (!pskb_may_pull(skb, off + optlen))
 453                                 break;
 454
 455                         while (1) {
 456                                 struct ipv6_tlv_tnl_enc_lim *tel;
 457
 458                                 /* No more room for encapsulation limit */
 459                                 if (i + sizeof(*tel) > optlen)
 460                                         break;
 461
 462                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 463                                 /* return index of option if found and valid */
 464                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 465                                     tel->length == 1)
 466                                         return i + off - nhoff;
 467                                 /* else jump to next option */
 468                                 if (tel->type)
 469                                         i += tel->length + 2;
 470                                 else
 471                                         i++;
 472                         }
 473                 }
 474                 nexthdr = next;
 475                 off += optlen;
 476         }
 477         return 0;
 478 }
 479 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 480
 481 /**
 482  * ip6_tnl_err - tunnel error handler
 483  *
 484  * Description:
 485  *   ip6_tnl_err() should handle errors in the tunnel according
 486  *   to the specifications in RFC 2473.
 487  **/
 488
 489 static int
 490 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 491             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 492 {
 493         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 494         struct net *net = dev_net(skb->dev);
 495         u8 rel_type = ICMPV6_DEST_UNREACH;
 496         u8 rel_code = ICMPV6_ADDR_UNREACH;
 497         __u32 rel_info = 0;
 498         struct ip6_tnl *t;
 499         int err = -ENOENT;
 500         int rel_msg = 0;
 501         u8 tproto;
 502         __u16 len;
 503
 504         /* If the packet doesn't contain the original IPv6 header we are
 505            in trouble since we might need the source address for further
 506            processing of the error. */
 507
 508         rcu_read_lock();
 509         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
 510         if (!t)
 511                 goto out;
 512
 513         tproto = READ_ONCE(t->parms.proto);
 514         if (tproto != ipproto && tproto != 0)
 515                 goto out;
 516
 517         err = 0;
 518
 519         switch (*type) {
 520                 struct ipv6_tlv_tnl_enc_lim *tel;
 521                 __u32 mtu, teli;
 522         case ICMPV6_DEST_UNREACH:
 523                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 524                                     t->parms.name);
 525                 rel_msg = 1;
 526                 break;
 527         case ICMPV6_TIME_EXCEED:
 528                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 529                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 530                                             t->parms.name);
 531                         rel_msg = 1;
 532                 }
 533                 break;
 534         case ICMPV6_PARAMPROB:
 535                 teli = 0;
 536                 if ((*code) == ICMPV6_HDR_FIELD)
 537                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 538
 539                 if (teli && teli == *info - 2) {
 540                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 541                         if (tel->encap_limit == 0) {
 542                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 543                                                     t->parms.name);
 544                                 rel_msg = 1;
 545                         }
 546                 } else {
 547                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 548                                             t->parms.name);
 549                 }
 550                 break;
 551         case ICMPV6_PKT_TOOBIG:
 552                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 553                                 sock_net_uid(net, NULL));
 554                 mtu = *info - offset;
 555                 if (mtu < IPV6_MIN_MTU)
 556                         mtu = IPV6_MIN_MTU;
 557                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 558                 if (len > mtu) {
 559                         rel_type = ICMPV6_PKT_TOOBIG;
 560                         rel_code = 0;
 561                         rel_info = mtu;
 562                         rel_msg = 1;
 563                 }
 564                 break;
 565         case NDISC_REDIRECT:
 566                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 567                              sock_net_uid(net, NULL));
 568                 break;
 569         }
 570
 571         *type = rel_type;
 572         *code = rel_code;
 573         *info = rel_info;
 574         *msg = rel_msg;
 575
 576 out:
 577         rcu_read_unlock();
 578         return err;
 579 }
 580
 581 static int
 582 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 583            u8 type, u8 code, int offset, __be32 info)
 584 {
 585         __u32 rel_info = ntohl(info);
 586         const struct iphdr *eiph;
 587         struct sk_buff *skb2;
 588         int err, rel_msg = 0;
 589         u8 rel_type = type;
 590         u8 rel_code = code;
 591         struct rtable *rt;
 592         struct flowi4 fl4;
 593
 594         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 595                           &rel_msg, &rel_info, offset);
 596         if (err < 0)
 597                 return err;
 598
 599         if (rel_msg == 0)
 600                 return 0;
 601
 602         switch (rel_type) {
 603         case ICMPV6_DEST_UNREACH:
 604                 if (rel_code != ICMPV6_ADDR_UNREACH)
 605                         return 0;
 606                 rel_type = ICMP_DEST_UNREACH;
 607                 rel_code = ICMP_HOST_UNREACH;
 608                 break;
 609         case ICMPV6_PKT_TOOBIG:
 610                 if (rel_code != 0)
 611                         return 0;
 612                 rel_type = ICMP_DEST_UNREACH;
 613                 rel_code = ICMP_FRAG_NEEDED;
 614                 break;
 615         default:
 616                 return 0;
 617         }
 618
 619         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 620                 return 0;
 621
 622         skb2 = skb_clone(skb, GFP_ATOMIC);
 623         if (!skb2)
 624                 return 0;
 625
 626         skb_dst_drop(skb2);
 627
 628         skb_pull(skb2, offset);
 629         skb_reset_network_header(skb2);
 630         eiph = ip_hdr(skb2);
 631
 632         /* Try to guess incoming interface */
 633         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 634                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 635         if (IS_ERR(rt))
 636                 goto out;
 637
 638         skb2->dev = rt->dst.dev;
 639         ip_rt_put(rt);
 640
 641         /* route "incoming" packet */
 642         if (rt->rt_flags & RTCF_LOCAL) {
 643                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 644                                            eiph->daddr, eiph->saddr, 0, 0,
 645                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 646                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
 647                         if (!IS_ERR(rt))
 648                                 ip_rt_put(rt);
 649                         goto out;
 650                 }
 651                 skb_dst_set(skb2, &rt->dst);
 652         } else {
 653                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 654                                    skb2->dev) ||
 655                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 656                         goto out;
 657         }
 658
 659         /* change mtu on this route */
 660         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 661                 if (rel_info > dst_mtu(skb_dst(skb2)))
 662                         goto out;
 663
 664                 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
 665         }
 666
 667         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 668
 669 out:
 670         kfree_skb(skb2);
 671         return 0;
 672 }
 673
 674 static int
 675 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 676            u8 type, u8 code, int offset, __be32 info)
 677 {
 678         __u32 rel_info = ntohl(info);
 679         int err, rel_msg = 0;
 680         u8 rel_type = type;
 681         u8 rel_code = code;
 682
 683         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 684                           &rel_msg, &rel_info, offset);
 685         if (err < 0)
 686                 return err;
 687
 688         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 689                 struct rt6_info *rt;
 690                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 691
 692                 if (!skb2)
 693                         return 0;
 694
 695                 skb_dst_drop(skb2);
 696                 skb_pull(skb2, offset);
 697                 skb_reset_network_header(skb2);
 698
 699                 /* Try to guess incoming interface */
 700                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 701                                 NULL, 0, skb2, 0);
 702
 703                 if (rt && rt->dst.dev)
 704                         skb2->dev = rt->dst.dev;
 705
 706                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 707
 708                 ip6_rt_put(rt);
 709
 710                 kfree_skb(skb2);
 711         }
 712
 713         return 0;
 714 }
 715
 716 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 717                                        const struct ipv6hdr *ipv6h,
 718                                        struct sk_buff *skb)
 719 {
 720         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 721
 722         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 723                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 724
 725         return IP6_ECN_decapsulate(ipv6h, skb);
 726 }
 727
 728 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 729                                        const struct ipv6hdr *ipv6h,
 730                                        struct sk_buff *skb)
 731 {
 732         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 733                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 734
 735         return IP6_ECN_decapsulate(ipv6h, skb);
 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         struct pcpu_sw_netstats *tstats;
 800         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 801         int err;
 802
 803         if ((!(tpi->flags & TUNNEL_CSUM) &&
 804              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 805             ((tpi->flags & TUNNEL_CSUM) &&
 806              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 807                 tunnel->dev->stats.rx_crc_errors++;
 808                 tunnel->dev->stats.rx_errors++;
 809                 goto drop;
 810         }
 811
 812         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 813                 if (!(tpi->flags & TUNNEL_SEQ) ||
 814                     (tunnel->i_seqno &&
 815                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 816                         tunnel->dev->stats.rx_fifo_errors++;
 817                         tunnel->dev->stats.rx_errors++;
 818                         goto drop;
 819                 }
 820                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 821         }
 822
 823         skb->protocol = tpi->proto;
 824
 825         /* Warning: All skb pointers will be invalidated! */
 826         if (tunnel->dev->type == ARPHRD_ETHER) {
 827                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 828                         tunnel->dev->stats.rx_length_errors++;
 829                         tunnel->dev->stats.rx_errors++;
 830                         goto drop;
 831                 }
 832
 833                 ipv6h = ipv6_hdr(skb);
 834                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 835                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 836         } else {
 837                 skb->dev = tunnel->dev;
 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                         ++tunnel->dev->stats.rx_frame_errors;
 853                         ++tunnel->dev->stats.rx_errors;
 854                         goto drop;
 855                 }
 856         }
 857
 858         tstats = this_cpu_ptr(tunnel->dev->tstats);
 859         u64_stats_update_begin(&tstats->syncp);
 860         tstats->rx_packets++;
 861         tstats->rx_bytes += skb->len;
 862         u64_stats_update_end(&tstats->syncp);
 863
 864         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 865
 866         if (tun_dst)
 867                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 868
 869         gro_cells_receive(&tunnel->gro_cells, skb);
 870         return 0;
 871
 872 drop:
 873         if (tun_dst)
 874                 dst_release((struct dst_entry *)tun_dst);
 875         kfree_skb(skb);
 876         return 0;
 877 }
 878
 879 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 880                 const struct tnl_ptk_info *tpi,
 881                 struct metadata_dst *tun_dst,
 882                 bool log_ecn_err)
 883 {
 884         return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
 885                              log_ecn_err);
 886 }
 887 EXPORT_SYMBOL(ip6_tnl_rcv);
 888
 889 static const struct tnl_ptk_info tpi_v6 = {
 890         /* no tunnel info required for ipxip6. */
 891         .proto = htons(ETH_P_IPV6),
 892 };
 893
 894 static const struct tnl_ptk_info tpi_v4 = {
 895         /* no tunnel info required for ipxip6. */
 896         .proto = htons(ETH_P_IP),
 897 };
 898
 899 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 900                       const struct tnl_ptk_info *tpi,
 901                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 902                                                   const struct ipv6hdr *ipv6h,
 903                                                   struct sk_buff *skb))
 904 {
 905         struct ip6_tnl *t;
 906         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 907         struct metadata_dst *tun_dst = NULL;
 908         int ret = -1;
 909
 910         rcu_read_lock();
 911         t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
 912
 913         if (t) {
 914                 u8 tproto = READ_ONCE(t->parms.proto);
 915
 916                 if (tproto != ipproto && tproto != 0)
 917                         goto drop;
 918                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 919                         goto drop;
 920                 ipv6h = ipv6_hdr(skb);
 921                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 922                         goto drop;
 923                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 924                         goto drop;
 925                 if (t->parms.collect_md) {
 926                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 927                         if (!tun_dst)
 928                                 goto drop;
 929                 }
 930                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 931                                     log_ecn_error);
 932         }
 933
 934         rcu_read_unlock();
 935
 936         return ret;
 937
 938 drop:
 939         rcu_read_unlock();
 940         kfree_skb(skb);
 941         return 0;
 942 }
 943
 944 static int ip4ip6_rcv(struct sk_buff *skb)
 945 {
 946         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 947                           ip4ip6_dscp_ecn_decapsulate);
 948 }
 949
 950 static int ip6ip6_rcv(struct sk_buff *skb)
 951 {
 952         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 953                           ip6ip6_dscp_ecn_decapsulate);
 954 }
 955
 956 struct ipv6_tel_txoption {
 957         struct ipv6_txoptions ops;
 958         __u8 dst_opt[8];
 959 };
 960
 961 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 962 {
 963         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 964
 965         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 966         opt->dst_opt[3] = 1;
 967         opt->dst_opt[4] = encap_limit;
 968         opt->dst_opt[5] = IPV6_TLV_PADN;
 969         opt->dst_opt[6] = 1;
 970
 971         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 972         opt->ops.opt_nflen = 8;
 973 }
 974
 975 /**
 976  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 977  *   @t: the outgoing tunnel device
 978  *   @hdr: IPv6 header from the incoming packet
 979  *
 980  * Description:
 981  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 982  *   doesn't match source of incoming packet.
 983  *
 984  * Return:
 985  *   1 if conflict,
 986  *   0 else
 987  **/
 988
 989 static inline bool
 990 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 991 {
 992         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 993 }
 994
 995 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 996                      const struct in6_addr *laddr,
 997                      const struct in6_addr *raddr)
 998 {
 999         struct __ip6_tnl_parm *p = &t->parms;
1000         int ret = 0;
1001         struct net *net = t->net;
1002
1003         if (t->parms.collect_md)
1004                 return 1;
1005
1006         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1007             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1008              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1009                 struct net_device *ldev = NULL;
1010
1011                 rcu_read_lock();
1012                 if (p->link)
1013                         ldev = dev_get_by_index_rcu(net, p->link);
1014
1015                 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1016                                                       0, IFA_F_TENTATIVE)))
1017                         pr_warn("%s xmit: Local address not yet configured!\n",
1018                                 p->name);
1019                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1020                          !ipv6_addr_is_multicast(raddr) &&
1021                          unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1022                                                           true, 0, IFA_F_TENTATIVE)))
1023                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1024                                 p->name);
1025                 else
1026                         ret = 1;
1027                 rcu_read_unlock();
1028         }
1029         return ret;
1030 }
1031 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1032
1033 /**
1034  * ip6_tnl_xmit - encapsulate packet and send
1035  *   @skb: the outgoing socket buffer
1036  *   @dev: the outgoing tunnel device
1037  *   @dsfield: dscp code for outer header
1038  *   @fl6: flow of tunneled packet
1039  *   @encap_limit: encapsulation limit
1040  *   @pmtu: Path MTU is stored if packet is too big
1041  *   @proto: next header value
1042  *
1043  * Description:
1044  *   Build new header and do some sanity checks on the packet before sending
1045  *   it.
1046  *
1047  * Return:
1048  *   0 on success
1049  *   -1 fail
1050  *   %-EMSGSIZE message too big. return mtu in this case.
1051  **/
1052
1053 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1054                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1055                  __u8 proto)
1056 {
1057         struct ip6_tnl *t = netdev_priv(dev);
1058         struct net *net = t->net;
1059         struct net_device_stats *stats = &t->dev->stats;
1060         struct ipv6hdr *ipv6h;
1061         struct ipv6_tel_txoption opt;
1062         struct dst_entry *dst = NULL, *ndst = NULL;
1063         struct net_device *tdev;
1064         int mtu;
1065         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1066         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1067         unsigned int max_headroom = psh_hlen;
1068         bool use_cache = false;
1069         u8 hop_limit;
1070         int err = -1;
1071
1072         if (t->parms.collect_md) {
1073                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1074                 goto route_lookup;
1075         } else {
1076                 hop_limit = t->parms.hop_limit;
1077         }
1078
1079         /* NBMA tunnel */
1080         if (ipv6_addr_any(&t->parms.raddr)) {
1081                 if (skb->protocol == htons(ETH_P_IPV6)) {
1082                         struct in6_addr *addr6;
1083                         struct neighbour *neigh;
1084                         int addr_type;
1085
1086                         if (!skb_dst(skb))
1087                                 goto tx_err_link_failure;
1088
1089                         neigh = dst_neigh_lookup(skb_dst(skb),
1090                                                  &ipv6_hdr(skb)->daddr);
1091                         if (!neigh)
1092                                 goto tx_err_link_failure;
1093
1094                         addr6 = (struct in6_addr *)&neigh->primary_key;
1095                         addr_type = ipv6_addr_type(addr6);
1096
1097                         if (addr_type == IPV6_ADDR_ANY)
1098                                 addr6 = &ipv6_hdr(skb)->daddr;
1099
1100                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1101                         neigh_release(neigh);
1102                 }
1103         } else if (t->parms.proto != 0 && !(t->parms.flags &
1104                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1105                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1106                 /* enable the cache only if neither the outer protocol nor the
1107                  * routing decision depends on the current inner header value
1108                  */
1109                 use_cache = true;
1110         }
1111
1112         if (use_cache)
1113                 dst = dst_cache_get(&t->dst_cache);
1114
1115         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1116                 goto tx_err_link_failure;
1117
1118         if (!dst) {
1119 route_lookup:
1120                 /* add dsfield to flowlabel for route lookup */
1121                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1122
1123                 dst = ip6_route_output(net, NULL, fl6);
1124
1125                 if (dst->error)
1126                         goto tx_err_link_failure;
1127                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1128                 if (IS_ERR(dst)) {
1129                         err = PTR_ERR(dst);
1130                         dst = NULL;
1131                         goto tx_err_link_failure;
1132                 }
1133                 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1134                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1135                                        &fl6->daddr, 0, &fl6->saddr))
1136                         goto tx_err_link_failure;
1137                 ndst = dst;
1138         }
1139
1140         tdev = dst->dev;
1141
1142         if (tdev == dev) {
1143                 stats->collisions++;
1144                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1145                                      t->parms.name);
1146                 goto tx_err_dst_release;
1147         }
1148         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1149         if (encap_limit >= 0) {
1150                 max_headroom += 8;
1151                 mtu -= 8;
1152         }
1153         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1154                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1155
1156         skb_dst_update_pmtu_no_confirm(skb, mtu);
1157         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1158                 *pmtu = mtu;
1159                 err = -EMSGSIZE;
1160                 goto tx_err_dst_release;
1161         }
1162
1163         if (t->err_count > 0) {
1164                 if (time_before(jiffies,
1165                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1166                         t->err_count--;
1167
1168                         dst_link_failure(skb);
1169                 } else {
1170                         t->err_count = 0;
1171                 }
1172         }
1173
1174         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1175
1176         /*
1177          * Okay, now see if we can stuff it in the buffer as-is.
1178          */
1179         max_headroom += LL_RESERVED_SPACE(tdev);
1180
1181         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1182             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1183                 struct sk_buff *new_skb;
1184
1185                 new_skb = skb_realloc_headroom(skb, max_headroom);
1186                 if (!new_skb)
1187                         goto tx_err_dst_release;
1188
1189                 if (skb->sk)
1190                         skb_set_owner_w(new_skb, skb->sk);
1191                 consume_skb(skb);
1192                 skb = new_skb;
1193         }
1194
1195         if (t->parms.collect_md) {
1196                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1197                         goto tx_err_dst_release;
1198         } else {
1199                 if (use_cache && ndst)
1200                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1201         }
1202         skb_dst_set(skb, dst);
1203
1204         if (hop_limit == 0) {
1205                 if (skb->protocol == htons(ETH_P_IP))
1206                         hop_limit = ip_hdr(skb)->ttl;
1207                 else if (skb->protocol == htons(ETH_P_IPV6))
1208                         hop_limit = ipv6_hdr(skb)->hop_limit;
1209                 else
1210                         hop_limit = ip6_dst_hoplimit(dst);
1211         }
1212
1213         /* Calculate max headroom for all the headers and adjust
1214          * needed_headroom if necessary.
1215          */
1216         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1217                         + dst->header_len + t->hlen;
1218         if (max_headroom > dev->needed_headroom)
1219                 dev->needed_headroom = max_headroom;
1220
1221         err = ip6_tnl_encap(skb, t, &proto, fl6);
1222         if (err)
1223                 return err;
1224
1225         if (encap_limit >= 0) {
1226                 init_tel_txopt(&opt, encap_limit);
1227                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1228         }
1229
1230         skb_push(skb, sizeof(struct ipv6hdr));
1231         skb_reset_network_header(skb);
1232         ipv6h = ipv6_hdr(skb);
1233         ip6_flow_hdr(ipv6h, dsfield,
1234                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1235         ipv6h->hop_limit = hop_limit;
1236         ipv6h->nexthdr = proto;
1237         ipv6h->saddr = fl6->saddr;
1238         ipv6h->daddr = fl6->daddr;
1239         ip6tunnel_xmit(NULL, skb, dev);
1240         return 0;
1241 tx_err_link_failure:
1242         stats->tx_carrier_errors++;
1243         dst_link_failure(skb);
1244 tx_err_dst_release:
1245         dst_release(dst);
1246         return err;
1247 }
1248 EXPORT_SYMBOL(ip6_tnl_xmit);
1249
1250 static inline int
1251 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1252 {
1253         struct ip6_tnl *t = netdev_priv(dev);
1254         const struct iphdr  *iph;
1255         int encap_limit = -1;
1256         struct flowi6 fl6;
1257         __u8 dsfield;
1258         __u32 mtu;
1259         u8 tproto;
1260         int err;
1261
1262         iph = ip_hdr(skb);
1263         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1264
1265         tproto = READ_ONCE(t->parms.proto);
1266         if (tproto != IPPROTO_IPIP && tproto != 0)
1267                 return -1;
1268
1269         if (t->parms.collect_md) {
1270                 struct ip_tunnel_info *tun_info;
1271                 const struct ip_tunnel_key *key;
1272
1273                 tun_info = skb_tunnel_info(skb);
1274                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1275                              ip_tunnel_info_af(tun_info) != AF_INET6))
1276                         return -1;
1277                 key = &tun_info->key;
1278                 memset(&fl6, 0, sizeof(fl6));
1279                 fl6.flowi6_proto = IPPROTO_IPIP;
1280                 fl6.saddr = key->u.ipv6.src;
1281                 fl6.daddr = key->u.ipv6.dst;
1282                 fl6.flowlabel = key->label;
1283                 dsfield =  key->tos;
1284         } else {
1285                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1286                         encap_limit = t->parms.encap_limit;
1287
1288                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1289                 fl6.flowi6_proto = IPPROTO_IPIP;
1290
1291                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1292                         dsfield = ipv4_get_dsfield(iph);
1293                 else
1294                         dsfield = ip6_tclass(t->parms.flowinfo);
1295                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1296                         fl6.flowi6_mark = skb->mark;
1297                 else
1298                         fl6.flowi6_mark = t->parms.fwmark;
1299         }
1300
1301         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1302         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1303
1304         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1305                 return -1;
1306
1307         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1308
1309         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1310                            IPPROTO_IPIP);
1311         if (err != 0) {
1312                 /* XXX: send ICMP error even if DF is not set. */
1313                 if (err == -EMSGSIZE)
1314                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1315                                   htonl(mtu));
1316                 return -1;
1317         }
1318
1319         return 0;
1320 }
1321
1322 static inline int
1323 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1324 {
1325         struct ip6_tnl *t = netdev_priv(dev);
1326         struct ipv6hdr *ipv6h;
1327         int encap_limit = -1;
1328         __u16 offset;
1329         struct flowi6 fl6;
1330         __u8 dsfield;
1331         __u32 mtu;
1332         u8 tproto;
1333         int err;
1334
1335         ipv6h = ipv6_hdr(skb);
1336         tproto = READ_ONCE(t->parms.proto);
1337         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1338             ip6_tnl_addr_conflict(t, ipv6h))
1339                 return -1;
1340
1341         if (t->parms.collect_md) {
1342                 struct ip_tunnel_info *tun_info;
1343                 const struct ip_tunnel_key *key;
1344
1345                 tun_info = skb_tunnel_info(skb);
1346                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1347                              ip_tunnel_info_af(tun_info) != AF_INET6))
1348                         return -1;
1349                 key = &tun_info->key;
1350                 memset(&fl6, 0, sizeof(fl6));
1351                 fl6.flowi6_proto = IPPROTO_IPV6;
1352                 fl6.saddr = key->u.ipv6.src;
1353                 fl6.daddr = key->u.ipv6.dst;
1354                 fl6.flowlabel = key->label;
1355                 dsfield = key->tos;
1356         } else {
1357                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1358                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1359                 ipv6h = ipv6_hdr(skb);
1360                 if (offset > 0) {
1361                         struct ipv6_tlv_tnl_enc_lim *tel;
1362
1363                         tel = (void *)&skb_network_header(skb)[offset];
1364                         if (tel->encap_limit == 0) {
1365                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1366                                             ICMPV6_HDR_FIELD, offset + 2);
1367                                 return -1;
1368                         }
1369                         encap_limit = tel->encap_limit - 1;
1370                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1371                         encap_limit = t->parms.encap_limit;
1372                 }
1373
1374                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1375                 fl6.flowi6_proto = IPPROTO_IPV6;
1376
1377                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1378                         dsfield = ipv6_get_dsfield(ipv6h);
1379                 else
1380                         dsfield = ip6_tclass(t->parms.flowinfo);
1381                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1382                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1383                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1384                         fl6.flowi6_mark = skb->mark;
1385                 else
1386                         fl6.flowi6_mark = t->parms.fwmark;
1387         }
1388
1389         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1390         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1391
1392         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1393                 return -1;
1394
1395         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1396
1397         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1398                            IPPROTO_IPV6);
1399         if (err != 0) {
1400                 if (err == -EMSGSIZE)
1401                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1402                 return -1;
1403         }
1404
1405         return 0;
1406 }
1407
1408 static netdev_tx_t
1409 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1410 {
1411         struct ip6_tnl *t = netdev_priv(dev);
1412         struct net_device_stats *stats = &t->dev->stats;
1413         int ret;
1414
1415         if (!pskb_inet_may_pull(skb))
1416                 goto tx_err;
1417
1418         switch (skb->protocol) {
1419         case htons(ETH_P_IP):
1420                 ret = ip4ip6_tnl_xmit(skb, dev);
1421                 break;
1422         case htons(ETH_P_IPV6):
1423                 ret = ip6ip6_tnl_xmit(skb, dev);
1424                 break;
1425         default:
1426                 goto tx_err;
1427         }
1428
1429         if (ret < 0)
1430                 goto tx_err;
1431
1432         return NETDEV_TX_OK;
1433
1434 tx_err:
1435         stats->tx_errors++;
1436         stats->tx_dropped++;
1437         kfree_skb(skb);
1438         return NETDEV_TX_OK;
1439 }
1440
1441 static void ip6_tnl_link_config(struct ip6_tnl *t)
1442 {
1443         struct net_device *dev = t->dev;
1444         struct net_device *tdev = NULL;
1445         struct __ip6_tnl_parm *p = &t->parms;
1446         struct flowi6 *fl6 = &t->fl.u.ip6;
1447         unsigned int mtu;
1448         int t_hlen;
1449
1450         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1451         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1452
1453         /* Set up flowi template */
1454         fl6->saddr = p->laddr;
1455         fl6->daddr = p->raddr;
1456         fl6->flowi6_oif = p->link;
1457         fl6->flowlabel = 0;
1458
1459         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1460                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1461         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1462                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1463
1464         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1465         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1466
1467         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1468                 dev->flags |= IFF_POINTOPOINT;
1469         else
1470                 dev->flags &= ~IFF_POINTOPOINT;
1471
1472         t->tun_hlen = 0;
1473         t->hlen = t->encap_hlen + t->tun_hlen;
1474         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1475
1476         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1477                 int strict = (ipv6_addr_type(&p->raddr) &
1478                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1479
1480                 struct rt6_info *rt = rt6_lookup(t->net,
1481                                                  &p->raddr, &p->laddr,
1482                                                  p->link, NULL, strict);
1483                 if (rt) {
1484                         tdev = rt->dst.dev;
1485                         ip6_rt_put(rt);
1486                 }
1487
1488                 if (!tdev && p->link)
1489                         tdev = __dev_get_by_index(t->net, p->link);
1490
1491                 if (tdev) {
1492                         dev->hard_header_len = tdev->hard_header_len + t_hlen;
1493                         mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1494
1495                         dev->mtu = mtu - t_hlen;
1496                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1497                                 dev->mtu -= 8;
1498
1499                         if (dev->mtu < IPV6_MIN_MTU)
1500                                 dev->mtu = IPV6_MIN_MTU;
1501                 }
1502         }
1503 }
1504
1505 /**
1506  * ip6_tnl_change - update the tunnel parameters
1507  *   @t: tunnel to be changed
1508  *   @p: tunnel configuration parameters
1509  *
1510  * Description:
1511  *   ip6_tnl_change() updates the tunnel parameters
1512  **/
1513
1514 static int
1515 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1516 {
1517         t->parms.laddr = p->laddr;
1518         t->parms.raddr = p->raddr;
1519         t->parms.flags = p->flags;
1520         t->parms.hop_limit = p->hop_limit;
1521         t->parms.encap_limit = p->encap_limit;
1522         t->parms.flowinfo = p->flowinfo;
1523         t->parms.link = p->link;
1524         t->parms.proto = p->proto;
1525         t->parms.fwmark = p->fwmark;
1526         dst_cache_reset(&t->dst_cache);
1527         ip6_tnl_link_config(t);
1528         return 0;
1529 }
1530
1531 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1532 {
1533         struct net *net = t->net;
1534         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1535         int err;
1536
1537         ip6_tnl_unlink(ip6n, t);
1538         synchronize_net();
1539         err = ip6_tnl_change(t, p);
1540         ip6_tnl_link(ip6n, t);
1541         netdev_state_change(t->dev);
1542         return err;
1543 }
1544
1545 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1546 {
1547         /* for default tnl0 device allow to change only the proto */
1548         t->parms.proto = p->proto;
1549         netdev_state_change(t->dev);
1550         return 0;
1551 }
1552
1553 static void
1554 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1555 {
1556         p->laddr = u->laddr;
1557         p->raddr = u->raddr;
1558         p->flags = u->flags;
1559         p->hop_limit = u->hop_limit;
1560         p->encap_limit = u->encap_limit;
1561         p->flowinfo = u->flowinfo;
1562         p->link = u->link;
1563         p->proto = u->proto;
1564         memcpy(p->name, u->name, sizeof(u->name));
1565 }
1566
1567 static void
1568 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1569 {
1570         u->laddr = p->laddr;
1571         u->raddr = p->raddr;
1572         u->flags = p->flags;
1573         u->hop_limit = p->hop_limit;
1574         u->encap_limit = p->encap_limit;
1575         u->flowinfo = p->flowinfo;
1576         u->link = p->link;
1577         u->proto = p->proto;
1578         memcpy(u->name, p->name, sizeof(u->name));
1579 }
1580
1581 /**
1582  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1583  *   @dev: virtual device associated with tunnel
1584  *   @ifr: parameters passed from userspace
1585  *   @cmd: command to be performed
1586  *
1587  * Description:
1588  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1589  *   from userspace.
1590  *
1591  *   The possible commands are the following:
1592  *     %SIOCGETTUNNEL: get tunnel parameters for device
1593  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1594  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1595  *     %SIOCDELTUNNEL: delete tunnel
1596  *
1597  *   The fallback device "ip6tnl0", created during module
1598  *   initialization, can be used for creating other tunnel devices.
1599  *
1600  * Return:
1601  *   0 on success,
1602  *   %-EFAULT if unable to copy data to or from userspace,
1603  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1604  *   %-EINVAL if passed tunnel parameters are invalid,
1605  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1606  *   %-ENODEV if attempting to change or delete a nonexisting device
1607  **/
1608
1609 static int
1610 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1611 {
1612         int err = 0;
1613         struct ip6_tnl_parm p;
1614         struct __ip6_tnl_parm p1;
1615         struct ip6_tnl *t = netdev_priv(dev);
1616         struct net *net = t->net;
1617         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1618
1619         memset(&p1, 0, sizeof(p1));
1620
1621         switch (cmd) {
1622         case SIOCGETTUNNEL:
1623                 if (dev == ip6n->fb_tnl_dev) {
1624                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1625                                 err = -EFAULT;
1626                                 break;
1627                         }
1628                         ip6_tnl_parm_from_user(&p1, &p);
1629                         t = ip6_tnl_locate(net, &p1, 0);
1630                         if (IS_ERR(t))
1631                                 t = netdev_priv(dev);
1632                 } else {
1633                         memset(&p, 0, sizeof(p));
1634                 }
1635                 ip6_tnl_parm_to_user(&p, &t->parms);
1636                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1637                         err = -EFAULT;
1638                 }
1639                 break;
1640         case SIOCADDTUNNEL:
1641         case SIOCCHGTUNNEL:
1642                 err = -EPERM;
1643                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1644                         break;
1645                 err = -EFAULT;
1646                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1647                         break;
1648                 err = -EINVAL;
1649                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1650                     p.proto != 0)
1651                         break;
1652                 ip6_tnl_parm_from_user(&p1, &p);
1653                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1654                 if (cmd == SIOCCHGTUNNEL) {
1655                         if (!IS_ERR(t)) {
1656                                 if (t->dev != dev) {
1657                                         err = -EEXIST;
1658                                         break;
1659                                 }
1660                         } else
1661                                 t = netdev_priv(dev);
1662                         if (dev == ip6n->fb_tnl_dev)
1663                                 err = ip6_tnl0_update(t, &p1);
1664                         else
1665                                 err = ip6_tnl_update(t, &p1);
1666                 }
1667                 if (!IS_ERR(t)) {
1668                         err = 0;
1669                         ip6_tnl_parm_to_user(&p, &t->parms);
1670                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1671                                 err = -EFAULT;
1672
1673                 } else {
1674                         err = PTR_ERR(t);
1675                 }
1676                 break;
1677         case SIOCDELTUNNEL:
1678                 err = -EPERM;
1679                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1680                         break;
1681
1682                 if (dev == ip6n->fb_tnl_dev) {
1683                         err = -EFAULT;
1684                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1685                                 break;
1686                         err = -ENOENT;
1687                         ip6_tnl_parm_from_user(&p1, &p);
1688                         t = ip6_tnl_locate(net, &p1, 0);
1689                         if (IS_ERR(t))
1690                                 break;
1691                         err = -EPERM;
1692                         if (t->dev == ip6n->fb_tnl_dev)
1693                                 break;
1694                         dev = t->dev;
1695                 }
1696                 err = 0;
1697                 unregister_netdevice(dev);
1698                 break;
1699         default:
1700                 err = -EINVAL;
1701         }
1702         return err;
1703 }
1704
1705 /**
1706  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1707  *   @dev: virtual device associated with tunnel
1708  *   @new_mtu: the new mtu
1709  *
1710  * Return:
1711  *   0 on success,
1712  *   %-EINVAL if mtu too small
1713  **/
1714
1715 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1716 {
1717         struct ip6_tnl *tnl = netdev_priv(dev);
1718
1719         if (tnl->parms.proto == IPPROTO_IPV6) {
1720                 if (new_mtu < IPV6_MIN_MTU)
1721                         return -EINVAL;
1722         } else {
1723                 if (new_mtu < ETH_MIN_MTU)
1724                         return -EINVAL;
1725         }
1726         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1727                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1728                         return -EINVAL;
1729         } else {
1730                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1731                         return -EINVAL;
1732         }
1733         dev->mtu = new_mtu;
1734         return 0;
1735 }
1736 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1737
1738 int ip6_tnl_get_iflink(const struct net_device *dev)
1739 {
1740         struct ip6_tnl *t = netdev_priv(dev);
1741
1742         return t->parms.link;
1743 }
1744 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1745
1746 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1747                           unsigned int num)
1748 {
1749         if (num >= MAX_IPTUN_ENCAP_OPS)
1750                 return -ERANGE;
1751
1752         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1753                         &ip6tun_encaps[num],
1754                         NULL, ops) ? 0 : -1;
1755 }
1756 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1757
1758 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1759                           unsigned int num)
1760 {
1761         int ret;
1762
1763         if (num >= MAX_IPTUN_ENCAP_OPS)
1764                 return -ERANGE;
1765
1766         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1767                        &ip6tun_encaps[num],
1768                        ops, NULL) == ops) ? 0 : -1;
1769
1770         synchronize_net();
1771
1772         return ret;
1773 }
1774 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1775
1776 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1777                         struct ip_tunnel_encap *ipencap)
1778 {
1779         int hlen;
1780
1781         memset(&t->encap, 0, sizeof(t->encap));
1782
1783         hlen = ip6_encap_hlen(ipencap);
1784         if (hlen < 0)
1785                 return hlen;
1786
1787         t->encap.type = ipencap->type;
1788         t->encap.sport = ipencap->sport;
1789         t->encap.dport = ipencap->dport;
1790         t->encap.flags = ipencap->flags;
1791
1792         t->encap_hlen = hlen;
1793         t->hlen = t->encap_hlen + t->tun_hlen;
1794
1795         return 0;
1796 }
1797 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1798
1799 static const struct net_device_ops ip6_tnl_netdev_ops = {
1800         .ndo_init       = ip6_tnl_dev_init,
1801         .ndo_uninit     = ip6_tnl_dev_uninit,
1802         .ndo_start_xmit = ip6_tnl_start_xmit,
1803         .ndo_do_ioctl   = ip6_tnl_ioctl,
1804         .ndo_change_mtu = ip6_tnl_change_mtu,
1805         .ndo_get_stats  = ip6_get_stats,
1806         .ndo_get_iflink = ip6_tnl_get_iflink,
1807 };
1808
1809 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1810                          NETIF_F_FRAGLIST |     \
1811                          NETIF_F_HIGHDMA |      \
1812                          NETIF_F_GSO_SOFTWARE | \
1813                          NETIF_F_HW_CSUM)
1814
1815 /**
1816  * ip6_tnl_dev_setup - setup virtual tunnel device
1817  *   @dev: virtual device associated with tunnel
1818  *
1819  * Description:
1820  *   Initialize function pointers and device parameters
1821  **/
1822
1823 static void ip6_tnl_dev_setup(struct net_device *dev)
1824 {
1825         dev->netdev_ops = &ip6_tnl_netdev_ops;
1826         dev->needs_free_netdev = true;
1827         dev->priv_destructor = ip6_dev_free;
1828
1829         dev->type = ARPHRD_TUNNEL6;
1830         dev->flags |= IFF_NOARP;
1831         dev->addr_len = sizeof(struct in6_addr);
1832         dev->features |= NETIF_F_LLTX;
1833         netif_keep_dst(dev);
1834
1835         dev->features           |= IPXIPX_FEATURES;
1836         dev->hw_features        |= IPXIPX_FEATURES;
1837
1838         /* This perm addr will be used as interface identifier by IPv6 */
1839         dev->addr_assign_type = NET_ADDR_RANDOM;
1840         eth_random_addr(dev->perm_addr);
1841 }
1842
1843
1844 /**
1845  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1846  *   @dev: virtual device associated with tunnel
1847  **/
1848
1849 static inline int
1850 ip6_tnl_dev_init_gen(struct net_device *dev)
1851 {
1852         struct ip6_tnl *t = netdev_priv(dev);
1853         int ret;
1854         int t_hlen;
1855
1856         t->dev = dev;
1857         t->net = dev_net(dev);
1858         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1859         if (!dev->tstats)
1860                 return -ENOMEM;
1861
1862         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1863         if (ret)
1864                 goto free_stats;
1865
1866         ret = gro_cells_init(&t->gro_cells, dev);
1867         if (ret)
1868                 goto destroy_dst;
1869
1870         t->tun_hlen = 0;
1871         t->hlen = t->encap_hlen + t->tun_hlen;
1872         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1873
1874         dev->type = ARPHRD_TUNNEL6;
1875         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1876         dev->mtu = ETH_DATA_LEN - t_hlen;
1877         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1878                 dev->mtu -= 8;
1879         dev->min_mtu = ETH_MIN_MTU;
1880         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1881
1882         return 0;
1883
1884 destroy_dst:
1885         dst_cache_destroy(&t->dst_cache);
1886 free_stats:
1887         free_percpu(dev->tstats);
1888         dev->tstats = NULL;
1889
1890         return ret;
1891 }
1892
1893 /**
1894  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1895  *   @dev: virtual device associated with tunnel
1896  **/
1897
1898 static int ip6_tnl_dev_init(struct net_device *dev)
1899 {
1900         struct ip6_tnl *t = netdev_priv(dev);
1901         int err = ip6_tnl_dev_init_gen(dev);
1902
1903         if (err)
1904                 return err;
1905         ip6_tnl_link_config(t);
1906         if (t->parms.collect_md)
1907                 netif_keep_dst(dev);
1908         return 0;
1909 }
1910
1911 /**
1912  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1913  *   @dev: fallback device
1914  *
1915  * Return: 0
1916  **/
1917
1918 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1919 {
1920         struct ip6_tnl *t = netdev_priv(dev);
1921         struct net *net = dev_net(dev);
1922         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1923
1924         t->parms.proto = IPPROTO_IPV6;
1925         dev_hold(dev);
1926
1927         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1928         return 0;
1929 }
1930
1931 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1932                             struct netlink_ext_ack *extack)
1933 {
1934         u8 proto;
1935
1936         if (!data || !data[IFLA_IPTUN_PROTO])
1937                 return 0;
1938
1939         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1940         if (proto != IPPROTO_IPV6 &&
1941             proto != IPPROTO_IPIP &&
1942             proto != 0)
1943                 return -EINVAL;
1944
1945         return 0;
1946 }
1947
1948 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1949                                   struct __ip6_tnl_parm *parms)
1950 {
1951         memset(parms, 0, sizeof(*parms));
1952
1953         if (!data)
1954                 return;
1955
1956         if (data[IFLA_IPTUN_LINK])
1957                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1958
1959         if (data[IFLA_IPTUN_LOCAL])
1960                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1961
1962         if (data[IFLA_IPTUN_REMOTE])
1963                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1964
1965         if (data[IFLA_IPTUN_TTL])
1966                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1967
1968         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1969                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1970
1971         if (data[IFLA_IPTUN_FLOWINFO])
1972                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1973
1974         if (data[IFLA_IPTUN_FLAGS])
1975                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1976
1977         if (data[IFLA_IPTUN_PROTO])
1978                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1979
1980         if (data[IFLA_IPTUN_COLLECT_METADATA])
1981                 parms->collect_md = true;
1982
1983         if (data[IFLA_IPTUN_FWMARK])
1984                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1985 }
1986
1987 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1988                                         struct ip_tunnel_encap *ipencap)
1989 {
1990         bool ret = false;
1991
1992         memset(ipencap, 0, sizeof(*ipencap));
1993
1994         if (!data)
1995                 return ret;
1996
1997         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1998                 ret = true;
1999                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2000         }
2001
2002         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2003                 ret = true;
2004                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2005         }
2006
2007         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2008                 ret = true;
2009                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2010         }
2011
2012         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2013                 ret = true;
2014                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2015         }
2016
2017         return ret;
2018 }
2019
2020 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2021                            struct nlattr *tb[], struct nlattr *data[],
2022                            struct netlink_ext_ack *extack)
2023 {
2024         struct net *net = dev_net(dev);
2025         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2026         struct ip_tunnel_encap ipencap;
2027         struct ip6_tnl *nt, *t;
2028         int err;
2029
2030         nt = netdev_priv(dev);
2031
2032         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2033                 err = ip6_tnl_encap_setup(nt, &ipencap);
2034                 if (err < 0)
2035                         return err;
2036         }
2037
2038         ip6_tnl_netlink_parms(data, &nt->parms);
2039
2040         if (nt->parms.collect_md) {
2041                 if (rtnl_dereference(ip6n->collect_md_tun))
2042                         return -EEXIST;
2043         } else {
2044                 t = ip6_tnl_locate(net, &nt->parms, 0);
2045                 if (!IS_ERR(t))
2046                         return -EEXIST;
2047         }
2048
2049         err = ip6_tnl_create2(dev);
2050         if (!err && tb[IFLA_MTU])
2051                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2052
2053         return err;
2054 }
2055
2056 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2057                               struct nlattr *data[],
2058                               struct netlink_ext_ack *extack)
2059 {
2060         struct ip6_tnl *t = netdev_priv(dev);
2061         struct __ip6_tnl_parm p;
2062         struct net *net = t->net;
2063         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2064         struct ip_tunnel_encap ipencap;
2065
2066         if (dev == ip6n->fb_tnl_dev)
2067                 return -EINVAL;
2068
2069         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2070                 int err = ip6_tnl_encap_setup(t, &ipencap);
2071
2072                 if (err < 0)
2073                         return err;
2074         }
2075         ip6_tnl_netlink_parms(data, &p);
2076         if (p.collect_md)
2077                 return -EINVAL;
2078
2079         t = ip6_tnl_locate(net, &p, 0);
2080         if (!IS_ERR(t)) {
2081                 if (t->dev != dev)
2082                         return -EEXIST;
2083         } else
2084                 t = netdev_priv(dev);
2085
2086         return ip6_tnl_update(t, &p);
2087 }
2088
2089 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2090 {
2091         struct net *net = dev_net(dev);
2092         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2093
2094         if (dev != ip6n->fb_tnl_dev)
2095                 unregister_netdevice_queue(dev, head);
2096 }
2097
2098 static size_t ip6_tnl_get_size(const struct net_device *dev)
2099 {
2100         return
2101                 /* IFLA_IPTUN_LINK */
2102                 nla_total_size(4) +
2103                 /* IFLA_IPTUN_LOCAL */
2104                 nla_total_size(sizeof(struct in6_addr)) +
2105                 /* IFLA_IPTUN_REMOTE */
2106                 nla_total_size(sizeof(struct in6_addr)) +
2107                 /* IFLA_IPTUN_TTL */
2108                 nla_total_size(1) +
2109                 /* IFLA_IPTUN_ENCAP_LIMIT */
2110                 nla_total_size(1) +
2111                 /* IFLA_IPTUN_FLOWINFO */
2112                 nla_total_size(4) +
2113                 /* IFLA_IPTUN_FLAGS */
2114                 nla_total_size(4) +
2115                 /* IFLA_IPTUN_PROTO */
2116                 nla_total_size(1) +
2117                 /* IFLA_IPTUN_ENCAP_TYPE */
2118                 nla_total_size(2) +
2119                 /* IFLA_IPTUN_ENCAP_FLAGS */
2120                 nla_total_size(2) +
2121                 /* IFLA_IPTUN_ENCAP_SPORT */
2122                 nla_total_size(2) +
2123                 /* IFLA_IPTUN_ENCAP_DPORT */
2124                 nla_total_size(2) +
2125                 /* IFLA_IPTUN_COLLECT_METADATA */
2126                 nla_total_size(0) +
2127                 /* IFLA_IPTUN_FWMARK */
2128                 nla_total_size(4) +
2129                 0;
2130 }
2131
2132 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2133 {
2134         struct ip6_tnl *tunnel = netdev_priv(dev);
2135         struct __ip6_tnl_parm *parm = &tunnel->parms;
2136
2137         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2138             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2139             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2140             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2141             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2142             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2143             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2144             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2145             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2146                 goto nla_put_failure;
2147
2148         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2149             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2150             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2151             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2152                 goto nla_put_failure;
2153
2154         if (parm->collect_md)
2155                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2156                         goto nla_put_failure;
2157
2158         return 0;
2159
2160 nla_put_failure:
2161         return -EMSGSIZE;
2162 }
2163
2164 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2165 {
2166         struct ip6_tnl *tunnel = netdev_priv(dev);
2167
2168         return tunnel->net;
2169 }
2170 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2171
2172 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2173         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2174         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2175         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2176         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2177         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2178         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2179         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2180         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2181         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2182         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2183         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2184         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2185         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2186         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2187 };
2188
2189 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2190         .kind           = "ip6tnl",
2191         .maxtype        = IFLA_IPTUN_MAX,
2192         .policy         = ip6_tnl_policy,
2193         .priv_size      = sizeof(struct ip6_tnl),
2194         .setup          = ip6_tnl_dev_setup,
2195         .validate       = ip6_tnl_validate,
2196         .newlink        = ip6_tnl_newlink,
2197         .changelink     = ip6_tnl_changelink,
2198         .dellink        = ip6_tnl_dellink,
2199         .get_size       = ip6_tnl_get_size,
2200         .fill_info      = ip6_tnl_fill_info,
2201         .get_link_net   = ip6_tnl_get_link_net,
2202 };
2203
2204 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2205         .handler        = ip4ip6_rcv,
2206         .err_handler    = ip4ip6_err,
2207         .priority       =       1,
2208 };
2209
2210 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2211         .handler        = ip6ip6_rcv,
2212         .err_handler    = ip6ip6_err,
2213         .priority       =       1,
2214 };
2215
2216 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2217 {
2218         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2219         struct net_device *dev, *aux;
2220         int h;
2221         struct ip6_tnl *t;
2222
2223         for_each_netdev_safe(net, dev, aux)
2224                 if (dev->rtnl_link_ops == &ip6_link_ops)
2225                         unregister_netdevice_queue(dev, list);
2226
2227         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2228                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2229                 while (t) {
2230                         /* If dev is in the same netns, it has already
2231                          * been added to the list by the previous loop.
2232                          */
2233                         if (!net_eq(dev_net(t->dev), net))
2234                                 unregister_netdevice_queue(t->dev, list);
2235                         t = rtnl_dereference(t->next);
2236                 }
2237         }
2238 }
2239
2240 static int __net_init ip6_tnl_init_net(struct net *net)
2241 {
2242         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2243         struct ip6_tnl *t = NULL;
2244         int err;
2245
2246         ip6n->tnls[0] = ip6n->tnls_wc;
2247         ip6n->tnls[1] = ip6n->tnls_r_l;
2248
2249         if (!net_has_fallback_tunnels(net))
2250                 return 0;
2251         err = -ENOMEM;
2252         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2253                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2254
2255         if (!ip6n->fb_tnl_dev)
2256                 goto err_alloc_dev;
2257         dev_net_set(ip6n->fb_tnl_dev, net);
2258         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2259         /* FB netdevice is special: we have one, and only one per netns.
2260          * Allowing to move it to another netns is clearly unsafe.
2261          */
2262         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2263
2264         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2265         if (err < 0)
2266                 goto err_register;
2267
2268         err = register_netdev(ip6n->fb_tnl_dev);
2269         if (err < 0)
2270                 goto err_register;
2271
2272         t = netdev_priv(ip6n->fb_tnl_dev);
2273
2274         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2275         return 0;
2276
2277 err_register:
2278         free_netdev(ip6n->fb_tnl_dev);
2279 err_alloc_dev:
2280         return err;
2281 }
2282
2283 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2284 {
2285         struct net *net;
2286         LIST_HEAD(list);
2287
2288         rtnl_lock();
2289         list_for_each_entry(net, net_list, exit_list)
2290                 ip6_tnl_destroy_tunnels(net, &list);
2291         unregister_netdevice_many(&list);
2292         rtnl_unlock();
2293 }
2294
2295 static struct pernet_operations ip6_tnl_net_ops = {
2296         .init = ip6_tnl_init_net,
2297         .exit_batch = ip6_tnl_exit_batch_net,
2298         .id   = &ip6_tnl_net_id,
2299         .size = sizeof(struct ip6_tnl_net),
2300 };
2301
2302 /**
2303  * ip6_tunnel_init - register protocol and reserve needed resources
2304  *
2305  * Return: 0 on success
2306  **/
2307
2308 static int __init ip6_tunnel_init(void)
2309 {
2310         int  err;
2311
2312         if (!ipv6_mod_enabled())
2313                 return -EOPNOTSUPP;
2314
2315         err = register_pernet_device(&ip6_tnl_net_ops);
2316         if (err < 0)
2317                 goto out_pernet;
2318
2319         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2320         if (err < 0) {
2321                 pr_err("%s: can't register ip4ip6\n", __func__);
2322                 goto out_ip4ip6;
2323         }
2324
2325         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2326         if (err < 0) {
2327                 pr_err("%s: can't register ip6ip6\n", __func__);
2328                 goto out_ip6ip6;
2329         }
2330         err = rtnl_link_register(&ip6_link_ops);
2331         if (err < 0)
2332                 goto rtnl_link_failed;
2333
2334         return 0;
2335
2336 rtnl_link_failed:
2337         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2338 out_ip6ip6:
2339         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2340 out_ip4ip6:
2341         unregister_pernet_device(&ip6_tnl_net_ops);
2342 out_pernet:
2343         return err;
2344 }
2345
2346 /**
2347  * ip6_tunnel_cleanup - free resources and unregister protocol
2348  **/
2349
2350 static void __exit ip6_tunnel_cleanup(void)
2351 {
2352         rtnl_link_unregister(&ip6_link_ops);
2353         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2354                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2355
2356         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2357                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2358
2359         unregister_pernet_device(&ip6_tnl_net_ops);
2360 }
2361
2362 module_init(ip6_tunnel_init);
2363 module_exit(ip6_tunnel_cleanup);