net/ipv6/ip6_tunnel.c

   1 /*
   2  *      IPv6 tunneling device
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
   7  *      Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
   8  *
   9  *      Based on:
  10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  11  *
  12  *      RFC 2473
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  */
  20
  21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  22
  23 #include <linux/module.h>
  24 #include <linux/capability.h>
  25 #include <linux/errno.h>
  26 #include <linux/types.h>
  27 #include <linux/sockios.h>
  28 #include <linux/icmp.h>
  29 #include <linux/if.h>
  30 #include <linux/in.h>
  31 #include <linux/ip.h>
  32 #include <linux/net.h>
  33 #include <linux/in6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/icmpv6.h>
  37 #include <linux/init.h>
  38 #include <linux/route.h>
  39 #include <linux/rtnetlink.h>
  40 #include <linux/netfilter_ipv6.h>
  41 #include <linux/slab.h>
  42 #include <linux/hash.h>
  43 #include <linux/etherdevice.h>
  44
  45 #include <linux/uaccess.h>
  46 #include <linux/atomic.h>
  47
  48 #include <net/icmp.h>
  49 #include <net/ip.h>
  50 #include <net/ip_tunnels.h>
  51 #include <net/ipv6.h>
  52 #include <net/ip6_route.h>
  53 #include <net/addrconf.h>
  54 #include <net/ip6_tunnel.h>
  55 #include <net/xfrm.h>
  56 #include <net/dsfield.h>
  57 #include <net/inet_ecn.h>
  58 #include <net/net_namespace.h>
  59 #include <net/netns/generic.h>
  60 #include <net/dst_metadata.h>
  61
  62 MODULE_AUTHOR("Ville Nuorvala");
  63 MODULE_DESCRIPTION("IPv6 tunneling device");
  64 MODULE_LICENSE("GPL");
  65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  66 MODULE_ALIAS_NETDEV("ip6tnl0");
  67
  68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
  69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
  70
  71 static bool log_ecn_error = true;
  72 module_param(log_ecn_error, bool, 0644);
  73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  74
  75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  76 {
  77         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  78
  79         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
  80 }
  81
  82 static int ip6_tnl_dev_init(struct net_device *dev);
  83 static void ip6_tnl_dev_setup(struct net_device *dev);
  84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  85
  86 static unsigned int ip6_tnl_net_id __read_mostly;
  87 struct ip6_tnl_net {
  88         /* the IPv6 tunnel fallback device */
  89         struct net_device *fb_tnl_dev;
  90         /* lists for storing tunnels in use */
  91         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
  92         struct ip6_tnl __rcu *tnls_wc[1];
  93         struct ip6_tnl __rcu **tnls[2];
  94         struct ip6_tnl __rcu *collect_md_tun;
  95 };
  96
  97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  98 {
  99         struct pcpu_sw_netstats tmp, sum = { 0 };
 100         int i;
 101
 102         for_each_possible_cpu(i) {
 103                 unsigned int start;
 104                 const struct pcpu_sw_netstats *tstats =
 105                                                    per_cpu_ptr(dev->tstats, i);
 106
 107                 do {
 108                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 109                         tmp.rx_packets = tstats->rx_packets;
 110                         tmp.rx_bytes = tstats->rx_bytes;
 111                         tmp.tx_packets = tstats->tx_packets;
 112                         tmp.tx_bytes =  tstats->tx_bytes;
 113                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 114
 115                 sum.rx_packets += tmp.rx_packets;
 116                 sum.rx_bytes   += tmp.rx_bytes;
 117                 sum.tx_packets += tmp.tx_packets;
 118                 sum.tx_bytes   += tmp.tx_bytes;
 119         }
 120         dev->stats.rx_packets = sum.rx_packets;
 121         dev->stats.rx_bytes   = sum.rx_bytes;
 122         dev->stats.tx_packets = sum.tx_packets;
 123         dev->stats.tx_bytes   = sum.tx_bytes;
 124         return &dev->stats;
 125 }
 126
 127 /**
 128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 129  *   @remote: the address of the tunnel exit-point
 130  *   @local: the address of the tunnel entry-point
 131  *
 132  * Return:
 133  *   tunnel matching given end-points if found,
 134  *   else fallback tunnel if its device is up,
 135  *   else %NULL
 136  **/
 137
 138 #define for_each_ip6_tunnel_rcu(start) \
 139         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 140
 141 static struct ip6_tnl *
 142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 143 {
 144         unsigned int hash = HASH(remote, local);
 145         struct ip6_tnl *t;
 146         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 147         struct in6_addr any;
 148
 149         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 150                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 151                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 152                     (t->dev->flags & IFF_UP))
 153                         return t;
 154         }
 155
 156         memset(&any, 0, sizeof(any));
 157         hash = HASH(&any, local);
 158         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 159                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 160                     ipv6_addr_any(&t->parms.raddr) &&
 161                     (t->dev->flags & IFF_UP))
 162                         return t;
 163         }
 164
 165         hash = HASH(remote, &any);
 166         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 167                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 168                     ipv6_addr_any(&t->parms.laddr) &&
 169                     (t->dev->flags & IFF_UP))
 170                         return t;
 171         }
 172
 173         t = rcu_dereference(ip6n->collect_md_tun);
 174         if (t)
 175                 return t;
 176
 177         t = rcu_dereference(ip6n->tnls_wc[0]);
 178         if (t && (t->dev->flags & IFF_UP))
 179                 return t;
 180
 181         return NULL;
 182 }
 183
 184 /**
 185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 186  *   @p: parameters containing tunnel end-points
 187  *
 188  * Description:
 189  *   ip6_tnl_bucket() returns the head of the list matching the
 190  *   &struct in6_addr entries laddr and raddr in @p.
 191  *
 192  * Return: head of IPv6 tunnel list
 193  **/
 194
 195 static struct ip6_tnl __rcu **
 196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 197 {
 198         const struct in6_addr *remote = &p->raddr;
 199         const struct in6_addr *local = &p->laddr;
 200         unsigned int h = 0;
 201         int prio = 0;
 202
 203         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 204                 prio = 1;
 205                 h = HASH(remote, local);
 206         }
 207         return &ip6n->tnls[prio][h];
 208 }
 209
 210 /**
 211  * ip6_tnl_link - add tunnel to hash table
 212  *   @t: tunnel to be added
 213  **/
 214
 215 static void
 216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 217 {
 218         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 219
 220         if (t->parms.collect_md)
 221                 rcu_assign_pointer(ip6n->collect_md_tun, t);
 222         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 223         rcu_assign_pointer(*tp, t);
 224 }
 225
 226 /**
 227  * ip6_tnl_unlink - remove tunnel from hash table
 228  *   @t: tunnel to be removed
 229  **/
 230
 231 static void
 232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 233 {
 234         struct ip6_tnl __rcu **tp;
 235         struct ip6_tnl *iter;
 236
 237         if (t->parms.collect_md)
 238                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 239
 240         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 241              (iter = rtnl_dereference(*tp)) != NULL;
 242              tp = &iter->next) {
 243                 if (t == iter) {
 244                         rcu_assign_pointer(*tp, t->next);
 245                         break;
 246                 }
 247         }
 248 }
 249
 250 static void ip6_dev_free(struct net_device *dev)
 251 {
 252         struct ip6_tnl *t = netdev_priv(dev);
 253
 254         gro_cells_destroy(&t->gro_cells);
 255         dst_cache_destroy(&t->dst_cache);
 256         free_percpu(dev->tstats);
 257 }
 258
 259 static int ip6_tnl_create2(struct net_device *dev)
 260 {
 261         struct ip6_tnl *t = netdev_priv(dev);
 262         struct net *net = dev_net(dev);
 263         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 264         int err;
 265
 266         t = netdev_priv(dev);
 267
 268         dev->rtnl_link_ops = &ip6_link_ops;
 269         err = register_netdevice(dev);
 270         if (err < 0)
 271                 goto out;
 272
 273         strcpy(t->parms.name, dev->name);
 274
 275         dev_hold(dev);
 276         ip6_tnl_link(ip6n, t);
 277         return 0;
 278
 279 out:
 280         return err;
 281 }
 282
 283 /**
 284  * ip6_tnl_create - create a new tunnel
 285  *   @p: tunnel parameters
 286  *   @pt: pointer to new tunnel
 287  *
 288  * Description:
 289  *   Create tunnel matching given parameters.
 290  *
 291  * Return:
 292  *   created tunnel or error pointer
 293  **/
 294
 295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 296 {
 297         struct net_device *dev;
 298         struct ip6_tnl *t;
 299         char name[IFNAMSIZ];
 300         int err = -ENOMEM;
 301
 302         if (p->name[0])
 303                 strlcpy(name, p->name, IFNAMSIZ);
 304         else
 305                 sprintf(name, "ip6tnl%%d");
 306
 307         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 308                            ip6_tnl_dev_setup);
 309         if (!dev)
 310                 goto failed;
 311
 312         dev_net_set(dev, net);
 313
 314         t = netdev_priv(dev);
 315         t->parms = *p;
 316         t->net = dev_net(dev);
 317         err = ip6_tnl_create2(dev);
 318         if (err < 0)
 319                 goto failed_free;
 320
 321         return t;
 322
 323 failed_free:
 324         free_netdev(dev);
 325 failed:
 326         return ERR_PTR(err);
 327 }
 328
 329 /**
 330  * ip6_tnl_locate - find or create tunnel matching given parameters
 331  *   @p: tunnel parameters
 332  *   @create: != 0 if allowed to create new tunnel if no match found
 333  *
 334  * Description:
 335  *   ip6_tnl_locate() first tries to locate an existing tunnel
 336  *   based on @parms. If this is unsuccessful, but @create is set a new
 337  *   tunnel device is created and registered for use.
 338  *
 339  * Return:
 340  *   matching tunnel or error pointer
 341  **/
 342
 343 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 344                 struct __ip6_tnl_parm *p, int create)
 345 {
 346         const struct in6_addr *remote = &p->raddr;
 347         const struct in6_addr *local = &p->laddr;
 348         struct ip6_tnl __rcu **tp;
 349         struct ip6_tnl *t;
 350         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 351
 352         for (tp = ip6_tnl_bucket(ip6n, p);
 353              (t = rtnl_dereference(*tp)) != NULL;
 354              tp = &t->next) {
 355                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 356                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 357                         if (create)
 358                                 return ERR_PTR(-EEXIST);
 359
 360                         return t;
 361                 }
 362         }
 363         if (!create)
 364                 return ERR_PTR(-ENODEV);
 365         return ip6_tnl_create(net, p);
 366 }
 367
 368 /**
 369  * ip6_tnl_dev_uninit - tunnel device uninitializer
 370  *   @dev: the device to be destroyed
 371  *
 372  * Description:
 373  *   ip6_tnl_dev_uninit() removes tunnel from its list
 374  **/
 375
 376 static void
 377 ip6_tnl_dev_uninit(struct net_device *dev)
 378 {
 379         struct ip6_tnl *t = netdev_priv(dev);
 380         struct net *net = t->net;
 381         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 382
 383         if (dev == ip6n->fb_tnl_dev)
 384                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 385         else
 386                 ip6_tnl_unlink(ip6n, t);
 387         dst_cache_reset(&t->dst_cache);
 388         dev_put(dev);
 389 }
 390
 391 /**
 392  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 393  *   @skb: received socket buffer
 394  *
 395  * Return:
 396  *   0 if none was found,
 397  *   else index to encapsulation limit
 398  **/
 399
 400 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 401 {
 402         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 403         unsigned int nhoff = raw - skb->data;
 404         unsigned int off = nhoff + sizeof(*ipv6h);
 405         u8 next, nexthdr = ipv6h->nexthdr;
 406
 407         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 408                 struct ipv6_opt_hdr *hdr;
 409                 u16 optlen;
 410
 411                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 412                         break;
 413
 414                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 415                 if (nexthdr == NEXTHDR_FRAGMENT) {
 416                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 417                         if (frag_hdr->frag_off)
 418                                 break;
 419                         optlen = 8;
 420                 } else if (nexthdr == NEXTHDR_AUTH) {
 421                         optlen = (hdr->hdrlen + 2) << 2;
 422                 } else {
 423                         optlen = ipv6_optlen(hdr);
 424                 }
 425                 /* cache hdr->nexthdr, since pskb_may_pull() might
 426                  * invalidate hdr
 427                  */
 428                 next = hdr->nexthdr;
 429                 if (nexthdr == NEXTHDR_DEST) {
 430                         u16 i = 2;
 431
 432                         /* Remember : hdr is no longer valid at this point. */
 433                         if (!pskb_may_pull(skb, off + optlen))
 434                                 break;
 435
 436                         while (1) {
 437                                 struct ipv6_tlv_tnl_enc_lim *tel;
 438
 439                                 /* No more room for encapsulation limit */
 440                                 if (i + sizeof(*tel) > optlen)
 441                                         break;
 442
 443                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 444                                 /* return index of option if found and valid */
 445                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 446                                     tel->length == 1)
 447                                         return i + off - nhoff;
 448                                 /* else jump to next option */
 449                                 if (tel->type)
 450                                         i += tel->length + 2;
 451                                 else
 452                                         i++;
 453                         }
 454                 }
 455                 nexthdr = next;
 456                 off += optlen;
 457         }
 458         return 0;
 459 }
 460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 461
 462 /**
 463  * ip6_tnl_err - tunnel error handler
 464  *
 465  * Description:
 466  *   ip6_tnl_err() should handle errors in the tunnel according
 467  *   to the specifications in RFC 2473.
 468  **/
 469
 470 static int
 471 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 472             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 473 {
 474         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 475         struct ip6_tnl *t;
 476         int rel_msg = 0;
 477         u8 rel_type = ICMPV6_DEST_UNREACH;
 478         u8 rel_code = ICMPV6_ADDR_UNREACH;
 479         u8 tproto;
 480         __u32 rel_info = 0;
 481         __u16 len;
 482         int err = -ENOENT;
 483
 484         /* If the packet doesn't contain the original IPv6 header we are
 485            in trouble since we might need the source address for further
 486            processing of the error. */
 487
 488         rcu_read_lock();
 489         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 490         if (!t)
 491                 goto out;
 492
 493         tproto = ACCESS_ONCE(t->parms.proto);
 494         if (tproto != ipproto && tproto != 0)
 495                 goto out;
 496
 497         err = 0;
 498
 499         switch (*type) {
 500                 __u32 teli;
 501                 struct ipv6_tlv_tnl_enc_lim *tel;
 502                 __u32 mtu;
 503         case ICMPV6_DEST_UNREACH:
 504                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 505                                     t->parms.name);
 506                 rel_msg = 1;
 507                 break;
 508         case ICMPV6_TIME_EXCEED:
 509                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 510                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 511                                             t->parms.name);
 512                         rel_msg = 1;
 513                 }
 514                 break;
 515         case ICMPV6_PARAMPROB:
 516                 teli = 0;
 517                 if ((*code) == ICMPV6_HDR_FIELD)
 518                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 519
 520                 if (teli && teli == *info - 2) {
 521                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 522                         if (tel->encap_limit == 0) {
 523                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 524                                                     t->parms.name);
 525                                 rel_msg = 1;
 526                         }
 527                 } else {
 528                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 529                                             t->parms.name);
 530                 }
 531                 break;
 532         case ICMPV6_PKT_TOOBIG:
 533                 mtu = *info - offset;
 534                 if (mtu < IPV6_MIN_MTU)
 535                         mtu = IPV6_MIN_MTU;
 536                 t->dev->mtu = mtu;
 537
 538                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 539                 if (len > mtu) {
 540                         rel_type = ICMPV6_PKT_TOOBIG;
 541                         rel_code = 0;
 542                         rel_info = mtu;
 543                         rel_msg = 1;
 544                 }
 545                 break;
 546         }
 547
 548         *type = rel_type;
 549         *code = rel_code;
 550         *info = rel_info;
 551         *msg = rel_msg;
 552
 553 out:
 554         rcu_read_unlock();
 555         return err;
 556 }
 557
 558 static int
 559 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 560            u8 type, u8 code, int offset, __be32 info)
 561 {
 562         int rel_msg = 0;
 563         u8 rel_type = type;
 564         u8 rel_code = code;
 565         __u32 rel_info = ntohl(info);
 566         int err;
 567         struct sk_buff *skb2;
 568         const struct iphdr *eiph;
 569         struct rtable *rt;
 570         struct flowi4 fl4;
 571
 572         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 573                           &rel_msg, &rel_info, offset);
 574         if (err < 0)
 575                 return err;
 576
 577         if (rel_msg == 0)
 578                 return 0;
 579
 580         switch (rel_type) {
 581         case ICMPV6_DEST_UNREACH:
 582                 if (rel_code != ICMPV6_ADDR_UNREACH)
 583                         return 0;
 584                 rel_type = ICMP_DEST_UNREACH;
 585                 rel_code = ICMP_HOST_UNREACH;
 586                 break;
 587         case ICMPV6_PKT_TOOBIG:
 588                 if (rel_code != 0)
 589                         return 0;
 590                 rel_type = ICMP_DEST_UNREACH;
 591                 rel_code = ICMP_FRAG_NEEDED;
 592                 break;
 593         case NDISC_REDIRECT:
 594                 rel_type = ICMP_REDIRECT;
 595                 rel_code = ICMP_REDIR_HOST;
 596         default:
 597                 return 0;
 598         }
 599
 600         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 601                 return 0;
 602
 603         skb2 = skb_clone(skb, GFP_ATOMIC);
 604         if (!skb2)
 605                 return 0;
 606
 607         skb_dst_drop(skb2);
 608
 609         skb_pull(skb2, offset);
 610         skb_reset_network_header(skb2);
 611         eiph = ip_hdr(skb2);
 612
 613         /* Try to guess incoming interface */
 614         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 615                                    eiph->saddr, 0,
 616                                    0, 0,
 617                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 618         if (IS_ERR(rt))
 619                 goto out;
 620
 621         skb2->dev = rt->dst.dev;
 622
 623         /* route "incoming" packet */
 624         if (rt->rt_flags & RTCF_LOCAL) {
 625                 ip_rt_put(rt);
 626                 rt = NULL;
 627                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 628                                            eiph->daddr, eiph->saddr,
 629                                            0, 0,
 630                                            IPPROTO_IPIP,
 631                                            RT_TOS(eiph->tos), 0);
 632                 if (IS_ERR(rt) ||
 633                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 634                         if (!IS_ERR(rt))
 635                                 ip_rt_put(rt);
 636                         goto out;
 637                 }
 638                 skb_dst_set(skb2, &rt->dst);
 639         } else {
 640                 ip_rt_put(rt);
 641                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 642                                    skb2->dev) ||
 643                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 644                         goto out;
 645         }
 646
 647         /* change mtu on this route */
 648         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 649                 if (rel_info > dst_mtu(skb_dst(skb2)))
 650                         goto out;
 651
 652                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 653         }
 654         if (rel_type == ICMP_REDIRECT)
 655                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 656
 657         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 658
 659 out:
 660         kfree_skb(skb2);
 661         return 0;
 662 }
 663
 664 static int
 665 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 666            u8 type, u8 code, int offset, __be32 info)
 667 {
 668         int rel_msg = 0;
 669         u8 rel_type = type;
 670         u8 rel_code = code;
 671         __u32 rel_info = ntohl(info);
 672         int err;
 673
 674         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 675                           &rel_msg, &rel_info, offset);
 676         if (err < 0)
 677                 return err;
 678
 679         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 680                 struct rt6_info *rt;
 681                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 682
 683                 if (!skb2)
 684                         return 0;
 685
 686                 skb_dst_drop(skb2);
 687                 skb_pull(skb2, offset);
 688                 skb_reset_network_header(skb2);
 689
 690                 /* Try to guess incoming interface */
 691                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 692                                 NULL, 0, 0);
 693
 694                 if (rt && rt->dst.dev)
 695                         skb2->dev = rt->dst.dev;
 696
 697                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 698
 699                 ip6_rt_put(rt);
 700
 701                 kfree_skb(skb2);
 702         }
 703
 704         return 0;
 705 }
 706
 707 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 708                                        const struct ipv6hdr *ipv6h,
 709                                        struct sk_buff *skb)
 710 {
 711         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 712
 713         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 714                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 715
 716         return IP6_ECN_decapsulate(ipv6h, skb);
 717 }
 718
 719 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 720                                        const struct ipv6hdr *ipv6h,
 721                                        struct sk_buff *skb)
 722 {
 723         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 724                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 725
 726         return IP6_ECN_decapsulate(ipv6h, skb);
 727 }
 728
 729 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 730                              const struct in6_addr *laddr,
 731                              const struct in6_addr *raddr)
 732 {
 733         struct __ip6_tnl_parm *p = &t->parms;
 734         int ltype = ipv6_addr_type(laddr);
 735         int rtype = ipv6_addr_type(raddr);
 736         __u32 flags = 0;
 737
 738         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 739                 flags = IP6_TNL_F_CAP_PER_PACKET;
 740         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 741                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 742                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 743                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 744                 if (ltype&IPV6_ADDR_UNICAST)
 745                         flags |= IP6_TNL_F_CAP_XMIT;
 746                 if (rtype&IPV6_ADDR_UNICAST)
 747                         flags |= IP6_TNL_F_CAP_RCV;
 748         }
 749         return flags;
 750 }
 751 EXPORT_SYMBOL(ip6_tnl_get_cap);
 752
 753 /* called with rcu_read_lock() */
 754 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 755                                   const struct in6_addr *laddr,
 756                                   const struct in6_addr *raddr)
 757 {
 758         struct __ip6_tnl_parm *p = &t->parms;
 759         int ret = 0;
 760         struct net *net = t->net;
 761
 762         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 763             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 764              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 765                 struct net_device *ldev = NULL;
 766
 767                 if (p->link)
 768                         ldev = dev_get_by_index_rcu(net, p->link);
 769
 770                 if ((ipv6_addr_is_multicast(laddr) ||
 771                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 772                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 773                         ret = 1;
 774         }
 775         return ret;
 776 }
 777 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 778
 779 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 780                          const struct tnl_ptk_info *tpi,
 781                          struct metadata_dst *tun_dst,
 782                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 783                                                 const struct ipv6hdr *ipv6h,
 784                                                 struct sk_buff *skb),
 785                          bool log_ecn_err)
 786 {
 787         struct pcpu_sw_netstats *tstats;
 788         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 789         int err;
 790
 791         if ((!(tpi->flags & TUNNEL_CSUM) &&
 792              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 793             ((tpi->flags & TUNNEL_CSUM) &&
 794              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 795                 tunnel->dev->stats.rx_crc_errors++;
 796                 tunnel->dev->stats.rx_errors++;
 797                 goto drop;
 798         }
 799
 800         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 801                 if (!(tpi->flags & TUNNEL_SEQ) ||
 802                     (tunnel->i_seqno &&
 803                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 804                         tunnel->dev->stats.rx_fifo_errors++;
 805                         tunnel->dev->stats.rx_errors++;
 806                         goto drop;
 807                 }
 808                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 809         }
 810
 811         skb->protocol = tpi->proto;
 812
 813         /* Warning: All skb pointers will be invalidated! */
 814         if (tunnel->dev->type == ARPHRD_ETHER) {
 815                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 816                         tunnel->dev->stats.rx_length_errors++;
 817                         tunnel->dev->stats.rx_errors++;
 818                         goto drop;
 819                 }
 820
 821                 ipv6h = ipv6_hdr(skb);
 822                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 823                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 824         } else {
 825                 skb->dev = tunnel->dev;
 826         }
 827
 828         skb_reset_network_header(skb);
 829         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 830
 831         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 832
 833         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 834         if (unlikely(err)) {
 835                 if (log_ecn_err)
 836                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 837                                              &ipv6h->saddr,
 838                                              ipv6_get_dsfield(ipv6h));
 839                 if (err > 1) {
 840                         ++tunnel->dev->stats.rx_frame_errors;
 841                         ++tunnel->dev->stats.rx_errors;
 842                         goto drop;
 843                 }
 844         }
 845
 846         tstats = this_cpu_ptr(tunnel->dev->tstats);
 847         u64_stats_update_begin(&tstats->syncp);
 848         tstats->rx_packets++;
 849         tstats->rx_bytes += skb->len;
 850         u64_stats_update_end(&tstats->syncp);
 851
 852         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 853
 854         if (tun_dst)
 855                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 856
 857         gro_cells_receive(&tunnel->gro_cells, skb);
 858         return 0;
 859
 860 drop:
 861         if (tun_dst)
 862                 dst_release((struct dst_entry *)tun_dst);
 863         kfree_skb(skb);
 864         return 0;
 865 }
 866
 867 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 868                 const struct tnl_ptk_info *tpi,
 869                 struct metadata_dst *tun_dst,
 870                 bool log_ecn_err)
 871 {
 872         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
 873                              log_ecn_err);
 874 }
 875 EXPORT_SYMBOL(ip6_tnl_rcv);
 876
 877 static const struct tnl_ptk_info tpi_v6 = {
 878         /* no tunnel info required for ipxip6. */
 879         .proto = htons(ETH_P_IPV6),
 880 };
 881
 882 static const struct tnl_ptk_info tpi_v4 = {
 883         /* no tunnel info required for ipxip6. */
 884         .proto = htons(ETH_P_IP),
 885 };
 886
 887 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 888                       const struct tnl_ptk_info *tpi,
 889                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 890                                                   const struct ipv6hdr *ipv6h,
 891                                                   struct sk_buff *skb))
 892 {
 893         struct ip6_tnl *t;
 894         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 895         struct metadata_dst *tun_dst = NULL;
 896         int ret = -1;
 897
 898         rcu_read_lock();
 899         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 900
 901         if (t) {
 902                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 903
 904                 if (tproto != ipproto && tproto != 0)
 905                         goto drop;
 906                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 907                         goto drop;
 908                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 909                         goto drop;
 910                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 911                         goto drop;
 912                 if (t->parms.collect_md) {
 913                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 914                         if (!tun_dst)
 915                                 return 0;
 916                 }
 917                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 918                                     log_ecn_error);
 919         }
 920
 921         rcu_read_unlock();
 922
 923         return ret;
 924
 925 drop:
 926         rcu_read_unlock();
 927         kfree_skb(skb);
 928         return 0;
 929 }
 930
 931 static int ip4ip6_rcv(struct sk_buff *skb)
 932 {
 933         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 934                           ip4ip6_dscp_ecn_decapsulate);
 935 }
 936
 937 static int ip6ip6_rcv(struct sk_buff *skb)
 938 {
 939         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 940                           ip6ip6_dscp_ecn_decapsulate);
 941 }
 942
 943 struct ipv6_tel_txoption {
 944         struct ipv6_txoptions ops;
 945         __u8 dst_opt[8];
 946 };
 947
 948 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 949 {
 950         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 951
 952         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 953         opt->dst_opt[3] = 1;
 954         opt->dst_opt[4] = encap_limit;
 955         opt->dst_opt[5] = IPV6_TLV_PADN;
 956         opt->dst_opt[6] = 1;
 957
 958         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 959         opt->ops.opt_nflen = 8;
 960 }
 961
 962 /**
 963  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 964  *   @t: the outgoing tunnel device
 965  *   @hdr: IPv6 header from the incoming packet
 966  *
 967  * Description:
 968  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 969  *   doesn't match source of incoming packet.
 970  *
 971  * Return:
 972  *   1 if conflict,
 973  *   0 else
 974  **/
 975
 976 static inline bool
 977 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 978 {
 979         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 980 }
 981
 982 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 983                      const struct in6_addr *laddr,
 984                      const struct in6_addr *raddr)
 985 {
 986         struct __ip6_tnl_parm *p = &t->parms;
 987         int ret = 0;
 988         struct net *net = t->net;
 989
 990         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 991             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 992              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 993                 struct net_device *ldev = NULL;
 994
 995                 rcu_read_lock();
 996                 if (p->link)
 997                         ldev = dev_get_by_index_rcu(net, p->link);
 998
 999                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1000                         pr_warn("%s xmit: Local address not yet configured!\n",
1001                                 p->name);
1002                 else if (!ipv6_addr_is_multicast(raddr) &&
1003                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1004                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1005                                 p->name);
1006                 else
1007                         ret = 1;
1008                 rcu_read_unlock();
1009         }
1010         return ret;
1011 }
1012 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1013
1014 /**
1015  * ip6_tnl_xmit - encapsulate packet and send
1016  *   @skb: the outgoing socket buffer
1017  *   @dev: the outgoing tunnel device
1018  *   @dsfield: dscp code for outer header
1019  *   @fl6: flow of tunneled packet
1020  *   @encap_limit: encapsulation limit
1021  *   @pmtu: Path MTU is stored if packet is too big
1022  *   @proto: next header value
1023  *
1024  * Description:
1025  *   Build new header and do some sanity checks on the packet before sending
1026  *   it.
1027  *
1028  * Return:
1029  *   0 on success
1030  *   -1 fail
1031  *   %-EMSGSIZE message too big. return mtu in this case.
1032  **/
1033
1034 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1035                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1036                  __u8 proto)
1037 {
1038         struct ip6_tnl *t = netdev_priv(dev);
1039         struct net *net = t->net;
1040         struct net_device_stats *stats = &t->dev->stats;
1041         struct ipv6hdr *ipv6h;
1042         struct ipv6_tel_txoption opt;
1043         struct dst_entry *dst = NULL, *ndst = NULL;
1044         struct net_device *tdev;
1045         int mtu;
1046         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1047         unsigned int max_headroom = psh_hlen;
1048         bool use_cache = false;
1049         u8 hop_limit;
1050         int err = -1;
1051
1052         if (t->parms.collect_md) {
1053                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1054                 goto route_lookup;
1055         } else {
1056                 hop_limit = t->parms.hop_limit;
1057         }
1058
1059         /* NBMA tunnel */
1060         if (ipv6_addr_any(&t->parms.raddr)) {
1061                 if (skb->protocol == htons(ETH_P_IPV6)) {
1062                         struct in6_addr *addr6;
1063                         struct neighbour *neigh;
1064                         int addr_type;
1065
1066                         if (!skb_dst(skb))
1067                                 goto tx_err_link_failure;
1068
1069                         neigh = dst_neigh_lookup(skb_dst(skb),
1070                                                  &ipv6_hdr(skb)->daddr);
1071                         if (!neigh)
1072                                 goto tx_err_link_failure;
1073
1074                         addr6 = (struct in6_addr *)&neigh->primary_key;
1075                         addr_type = ipv6_addr_type(addr6);
1076
1077                         if (addr_type == IPV6_ADDR_ANY)
1078                                 addr6 = &ipv6_hdr(skb)->daddr;
1079
1080                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1081                         neigh_release(neigh);
1082                 }
1083         } else if (!(t->parms.flags &
1084                      (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1085                 /* enable the cache only only if the routing decision does
1086                  * not depend on the current inner header value
1087                  */
1088                 use_cache = true;
1089         }
1090
1091         if (use_cache)
1092                 dst = dst_cache_get(&t->dst_cache);
1093
1094         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1095                 goto tx_err_link_failure;
1096
1097         if (!dst) {
1098 route_lookup:
1099                 /* add dsfield to flowlabel for route lookup */
1100                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1101
1102                 dst = ip6_route_output(net, NULL, fl6);
1103
1104                 if (dst->error)
1105                         goto tx_err_link_failure;
1106                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1107                 if (IS_ERR(dst)) {
1108                         err = PTR_ERR(dst);
1109                         dst = NULL;
1110                         goto tx_err_link_failure;
1111                 }
1112                 if (t->parms.collect_md &&
1113                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1114                                        &fl6->daddr, 0, &fl6->saddr))
1115                         goto tx_err_link_failure;
1116                 ndst = dst;
1117         }
1118
1119         tdev = dst->dev;
1120
1121         if (tdev == dev) {
1122                 stats->collisions++;
1123                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1124                                      t->parms.name);
1125                 goto tx_err_dst_release;
1126         }
1127         mtu = dst_mtu(dst) - psh_hlen - t->tun_hlen;
1128         if (encap_limit >= 0) {
1129                 max_headroom += 8;
1130                 mtu -= 8;
1131         }
1132         if (mtu < IPV6_MIN_MTU)
1133                 mtu = IPV6_MIN_MTU;
1134         if (skb_dst(skb) && !t->parms.collect_md)
1135                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1136         if (skb->len - t->tun_hlen > mtu && !skb_is_gso(skb)) {
1137                 *pmtu = mtu;
1138                 err = -EMSGSIZE;
1139                 goto tx_err_dst_release;
1140         }
1141
1142         if (t->err_count > 0) {
1143                 if (time_before(jiffies,
1144                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1145                         t->err_count--;
1146
1147                         dst_link_failure(skb);
1148                 } else {
1149                         t->err_count = 0;
1150                 }
1151         }
1152
1153         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1154
1155         /*
1156          * Okay, now see if we can stuff it in the buffer as-is.
1157          */
1158         max_headroom += LL_RESERVED_SPACE(tdev);
1159
1160         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1161             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1162                 struct sk_buff *new_skb;
1163
1164                 new_skb = skb_realloc_headroom(skb, max_headroom);
1165                 if (!new_skb)
1166                         goto tx_err_dst_release;
1167
1168                 if (skb->sk)
1169                         skb_set_owner_w(new_skb, skb->sk);
1170                 consume_skb(skb);
1171                 skb = new_skb;
1172         }
1173
1174         if (t->parms.collect_md) {
1175                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1176                         goto tx_err_dst_release;
1177         } else {
1178                 if (use_cache && ndst)
1179                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1180         }
1181         skb_dst_set(skb, dst);
1182
1183         if (encap_limit >= 0) {
1184                 init_tel_txopt(&opt, encap_limit);
1185                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1186         }
1187
1188         /* Calculate max headroom for all the headers and adjust
1189          * needed_headroom if necessary.
1190          */
1191         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1192                         + dst->header_len + t->hlen;
1193         if (max_headroom > dev->needed_headroom)
1194                 dev->needed_headroom = max_headroom;
1195
1196         err = ip6_tnl_encap(skb, t, &proto, fl6);
1197         if (err)
1198                 return err;
1199
1200         skb_push(skb, sizeof(struct ipv6hdr));
1201         skb_reset_network_header(skb);
1202         ipv6h = ipv6_hdr(skb);
1203         ip6_flow_hdr(ipv6h, dsfield,
1204                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1205         ipv6h->hop_limit = hop_limit;
1206         ipv6h->nexthdr = proto;
1207         ipv6h->saddr = fl6->saddr;
1208         ipv6h->daddr = fl6->daddr;
1209         ip6tunnel_xmit(NULL, skb, dev);
1210         return 0;
1211 tx_err_link_failure:
1212         stats->tx_carrier_errors++;
1213         dst_link_failure(skb);
1214 tx_err_dst_release:
1215         dst_release(dst);
1216         return err;
1217 }
1218 EXPORT_SYMBOL(ip6_tnl_xmit);
1219
1220 static inline int
1221 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1222 {
1223         struct ip6_tnl *t = netdev_priv(dev);
1224         const struct iphdr  *iph = ip_hdr(skb);
1225         int encap_limit = -1;
1226         struct flowi6 fl6;
1227         __u8 dsfield;
1228         __u32 mtu;
1229         u8 tproto;
1230         int err;
1231
1232         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1233
1234         tproto = ACCESS_ONCE(t->parms.proto);
1235         if (tproto != IPPROTO_IPIP && tproto != 0)
1236                 return -1;
1237
1238         if (t->parms.collect_md) {
1239                 struct ip_tunnel_info *tun_info;
1240                 const struct ip_tunnel_key *key;
1241
1242                 tun_info = skb_tunnel_info(skb);
1243                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1244                              ip_tunnel_info_af(tun_info) != AF_INET6))
1245                         return -1;
1246                 key = &tun_info->key;
1247                 memset(&fl6, 0, sizeof(fl6));
1248                 fl6.flowi6_proto = IPPROTO_IPIP;
1249                 fl6.daddr = key->u.ipv6.dst;
1250                 fl6.flowlabel = key->label;
1251                 dsfield =  key->tos;
1252         } else {
1253                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1254                         encap_limit = t->parms.encap_limit;
1255
1256                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1257                 fl6.flowi6_proto = IPPROTO_IPIP;
1258
1259                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1260                         dsfield = ipv4_get_dsfield(iph);
1261                 else
1262                         dsfield = ip6_tclass(t->parms.flowinfo);
1263                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1264                         fl6.flowi6_mark = skb->mark;
1265                 else
1266                         fl6.flowi6_mark = t->parms.fwmark;
1267         }
1268
1269         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1270
1271         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1272                 return -1;
1273
1274         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1275
1276         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1277
1278         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1279                            IPPROTO_IPIP);
1280         if (err != 0) {
1281                 /* XXX: send ICMP error even if DF is not set. */
1282                 if (err == -EMSGSIZE)
1283                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1284                                   htonl(mtu));
1285                 return -1;
1286         }
1287
1288         return 0;
1289 }
1290
1291 static inline int
1292 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1293 {
1294         struct ip6_tnl *t = netdev_priv(dev);
1295         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1296         int encap_limit = -1;
1297         __u16 offset;
1298         struct flowi6 fl6;
1299         __u8 dsfield;
1300         __u32 mtu;
1301         u8 tproto;
1302         int err;
1303
1304         tproto = ACCESS_ONCE(t->parms.proto);
1305         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1306             ip6_tnl_addr_conflict(t, ipv6h))
1307                 return -1;
1308
1309         if (t->parms.collect_md) {
1310                 struct ip_tunnel_info *tun_info;
1311                 const struct ip_tunnel_key *key;
1312
1313                 tun_info = skb_tunnel_info(skb);
1314                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1315                              ip_tunnel_info_af(tun_info) != AF_INET6))
1316                         return -1;
1317                 key = &tun_info->key;
1318                 memset(&fl6, 0, sizeof(fl6));
1319                 fl6.flowi6_proto = IPPROTO_IPV6;
1320                 fl6.daddr = key->u.ipv6.dst;
1321                 fl6.flowlabel = key->label;
1322                 dsfield = key->tos;
1323         } else {
1324                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1325                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1326                 ipv6h = ipv6_hdr(skb);
1327                 if (offset > 0) {
1328                         struct ipv6_tlv_tnl_enc_lim *tel;
1329
1330                         tel = (void *)&skb_network_header(skb)[offset];
1331                         if (tel->encap_limit == 0) {
1332                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1333                                             ICMPV6_HDR_FIELD, offset + 2);
1334                                 return -1;
1335                         }
1336                         encap_limit = tel->encap_limit - 1;
1337                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1338                         encap_limit = t->parms.encap_limit;
1339                 }
1340
1341                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1342                 fl6.flowi6_proto = IPPROTO_IPV6;
1343
1344                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1345                         dsfield = ipv6_get_dsfield(ipv6h);
1346                 else
1347                         dsfield = ip6_tclass(t->parms.flowinfo);
1348                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1349                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1350                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1351                         fl6.flowi6_mark = skb->mark;
1352                 else
1353                         fl6.flowi6_mark = t->parms.fwmark;
1354         }
1355
1356         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1357
1358         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1359                 return -1;
1360
1361         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1362
1363         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1364
1365         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1366                            IPPROTO_IPV6);
1367         if (err != 0) {
1368                 if (err == -EMSGSIZE)
1369                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1370                 return -1;
1371         }
1372
1373         return 0;
1374 }
1375
1376 static netdev_tx_t
1377 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1378 {
1379         struct ip6_tnl *t = netdev_priv(dev);
1380         struct net_device_stats *stats = &t->dev->stats;
1381         int ret;
1382
1383         switch (skb->protocol) {
1384         case htons(ETH_P_IP):
1385                 ret = ip4ip6_tnl_xmit(skb, dev);
1386                 break;
1387         case htons(ETH_P_IPV6):
1388                 ret = ip6ip6_tnl_xmit(skb, dev);
1389                 break;
1390         default:
1391                 goto tx_err;
1392         }
1393
1394         if (ret < 0)
1395                 goto tx_err;
1396
1397         return NETDEV_TX_OK;
1398
1399 tx_err:
1400         stats->tx_errors++;
1401         stats->tx_dropped++;
1402         kfree_skb(skb);
1403         return NETDEV_TX_OK;
1404 }
1405
1406 static void ip6_tnl_link_config(struct ip6_tnl *t)
1407 {
1408         struct net_device *dev = t->dev;
1409         struct __ip6_tnl_parm *p = &t->parms;
1410         struct flowi6 *fl6 = &t->fl.u.ip6;
1411         int t_hlen;
1412
1413         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1414         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1415
1416         /* Set up flowi template */
1417         fl6->saddr = p->laddr;
1418         fl6->daddr = p->raddr;
1419         fl6->flowi6_oif = p->link;
1420         fl6->flowlabel = 0;
1421
1422         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1423                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1424         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1425                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1426
1427         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1428         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1429
1430         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1431                 dev->flags |= IFF_POINTOPOINT;
1432         else
1433                 dev->flags &= ~IFF_POINTOPOINT;
1434
1435         t->tun_hlen = 0;
1436         t->hlen = t->encap_hlen + t->tun_hlen;
1437         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1438
1439         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1440                 int strict = (ipv6_addr_type(&p->raddr) &
1441                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1442
1443                 struct rt6_info *rt = rt6_lookup(t->net,
1444                                                  &p->raddr, &p->laddr,
1445                                                  p->link, strict);
1446
1447                 if (!rt)
1448                         return;
1449
1450                 if (rt->dst.dev) {
1451                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1452                                 t_hlen;
1453
1454                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1455                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1456                                 dev->mtu -= 8;
1457
1458                         if (dev->mtu < IPV6_MIN_MTU)
1459                                 dev->mtu = IPV6_MIN_MTU;
1460                 }
1461                 ip6_rt_put(rt);
1462         }
1463 }
1464
1465 /**
1466  * ip6_tnl_change - update the tunnel parameters
1467  *   @t: tunnel to be changed
1468  *   @p: tunnel configuration parameters
1469  *
1470  * Description:
1471  *   ip6_tnl_change() updates the tunnel parameters
1472  **/
1473
1474 static int
1475 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1476 {
1477         t->parms.laddr = p->laddr;
1478         t->parms.raddr = p->raddr;
1479         t->parms.flags = p->flags;
1480         t->parms.hop_limit = p->hop_limit;
1481         t->parms.encap_limit = p->encap_limit;
1482         t->parms.flowinfo = p->flowinfo;
1483         t->parms.link = p->link;
1484         t->parms.proto = p->proto;
1485         t->parms.fwmark = p->fwmark;
1486         dst_cache_reset(&t->dst_cache);
1487         ip6_tnl_link_config(t);
1488         return 0;
1489 }
1490
1491 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1492 {
1493         struct net *net = t->net;
1494         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1495         int err;
1496
1497         ip6_tnl_unlink(ip6n, t);
1498         synchronize_net();
1499         err = ip6_tnl_change(t, p);
1500         ip6_tnl_link(ip6n, t);
1501         netdev_state_change(t->dev);
1502         return err;
1503 }
1504
1505 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1506 {
1507         /* for default tnl0 device allow to change only the proto */
1508         t->parms.proto = p->proto;
1509         netdev_state_change(t->dev);
1510         return 0;
1511 }
1512
1513 static void
1514 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1515 {
1516         p->laddr = u->laddr;
1517         p->raddr = u->raddr;
1518         p->flags = u->flags;
1519         p->hop_limit = u->hop_limit;
1520         p->encap_limit = u->encap_limit;
1521         p->flowinfo = u->flowinfo;
1522         p->link = u->link;
1523         p->proto = u->proto;
1524         memcpy(p->name, u->name, sizeof(u->name));
1525 }
1526
1527 static void
1528 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1529 {
1530         u->laddr = p->laddr;
1531         u->raddr = p->raddr;
1532         u->flags = p->flags;
1533         u->hop_limit = p->hop_limit;
1534         u->encap_limit = p->encap_limit;
1535         u->flowinfo = p->flowinfo;
1536         u->link = p->link;
1537         u->proto = p->proto;
1538         memcpy(u->name, p->name, sizeof(u->name));
1539 }
1540
1541 /**
1542  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1543  *   @dev: virtual device associated with tunnel
1544  *   @ifr: parameters passed from userspace
1545  *   @cmd: command to be performed
1546  *
1547  * Description:
1548  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1549  *   from userspace.
1550  *
1551  *   The possible commands are the following:
1552  *     %SIOCGETTUNNEL: get tunnel parameters for device
1553  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1554  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1555  *     %SIOCDELTUNNEL: delete tunnel
1556  *
1557  *   The fallback device "ip6tnl0", created during module
1558  *   initialization, can be used for creating other tunnel devices.
1559  *
1560  * Return:
1561  *   0 on success,
1562  *   %-EFAULT if unable to copy data to or from userspace,
1563  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1564  *   %-EINVAL if passed tunnel parameters are invalid,
1565  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1566  *   %-ENODEV if attempting to change or delete a nonexisting device
1567  **/
1568
1569 static int
1570 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1571 {
1572         int err = 0;
1573         struct ip6_tnl_parm p;
1574         struct __ip6_tnl_parm p1;
1575         struct ip6_tnl *t = netdev_priv(dev);
1576         struct net *net = t->net;
1577         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1578
1579         memset(&p1, 0, sizeof(p1));
1580
1581         switch (cmd) {
1582         case SIOCGETTUNNEL:
1583                 if (dev == ip6n->fb_tnl_dev) {
1584                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1585                                 err = -EFAULT;
1586                                 break;
1587                         }
1588                         ip6_tnl_parm_from_user(&p1, &p);
1589                         t = ip6_tnl_locate(net, &p1, 0);
1590                         if (IS_ERR(t))
1591                                 t = netdev_priv(dev);
1592                 } else {
1593                         memset(&p, 0, sizeof(p));
1594                 }
1595                 ip6_tnl_parm_to_user(&p, &t->parms);
1596                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1597                         err = -EFAULT;
1598                 }
1599                 break;
1600         case SIOCADDTUNNEL:
1601         case SIOCCHGTUNNEL:
1602                 err = -EPERM;
1603                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1604                         break;
1605                 err = -EFAULT;
1606                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1607                         break;
1608                 err = -EINVAL;
1609                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1610                     p.proto != 0)
1611                         break;
1612                 ip6_tnl_parm_from_user(&p1, &p);
1613                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1614                 if (cmd == SIOCCHGTUNNEL) {
1615                         if (!IS_ERR(t)) {
1616                                 if (t->dev != dev) {
1617                                         err = -EEXIST;
1618                                         break;
1619                                 }
1620                         } else
1621                                 t = netdev_priv(dev);
1622                         if (dev == ip6n->fb_tnl_dev)
1623                                 err = ip6_tnl0_update(t, &p1);
1624                         else
1625                                 err = ip6_tnl_update(t, &p1);
1626                 }
1627                 if (!IS_ERR(t)) {
1628                         err = 0;
1629                         ip6_tnl_parm_to_user(&p, &t->parms);
1630                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1631                                 err = -EFAULT;
1632
1633                 } else {
1634                         err = PTR_ERR(t);
1635                 }
1636                 break;
1637         case SIOCDELTUNNEL:
1638                 err = -EPERM;
1639                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1640                         break;
1641
1642                 if (dev == ip6n->fb_tnl_dev) {
1643                         err = -EFAULT;
1644                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1645                                 break;
1646                         err = -ENOENT;
1647                         ip6_tnl_parm_from_user(&p1, &p);
1648                         t = ip6_tnl_locate(net, &p1, 0);
1649                         if (IS_ERR(t))
1650                                 break;
1651                         err = -EPERM;
1652                         if (t->dev == ip6n->fb_tnl_dev)
1653                                 break;
1654                         dev = t->dev;
1655                 }
1656                 err = 0;
1657                 unregister_netdevice(dev);
1658                 break;
1659         default:
1660                 err = -EINVAL;
1661         }
1662         return err;
1663 }
1664
1665 /**
1666  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1667  *   @dev: virtual device associated with tunnel
1668  *   @new_mtu: the new mtu
1669  *
1670  * Return:
1671  *   0 on success,
1672  *   %-EINVAL if mtu too small
1673  **/
1674
1675 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1676 {
1677         struct ip6_tnl *tnl = netdev_priv(dev);
1678
1679         if (tnl->parms.proto == IPPROTO_IPIP) {
1680                 if (new_mtu < ETH_MIN_MTU)
1681                         return -EINVAL;
1682         } else {
1683                 if (new_mtu < IPV6_MIN_MTU)
1684                         return -EINVAL;
1685         }
1686         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1687                 return -EINVAL;
1688         dev->mtu = new_mtu;
1689         return 0;
1690 }
1691 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1692
1693 int ip6_tnl_get_iflink(const struct net_device *dev)
1694 {
1695         struct ip6_tnl *t = netdev_priv(dev);
1696
1697         return t->parms.link;
1698 }
1699 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1700
1701 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1702                           unsigned int num)
1703 {
1704         if (num >= MAX_IPTUN_ENCAP_OPS)
1705                 return -ERANGE;
1706
1707         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1708                         &ip6tun_encaps[num],
1709                         NULL, ops) ? 0 : -1;
1710 }
1711 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1712
1713 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1714                           unsigned int num)
1715 {
1716         int ret;
1717
1718         if (num >= MAX_IPTUN_ENCAP_OPS)
1719                 return -ERANGE;
1720
1721         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1722                        &ip6tun_encaps[num],
1723                        ops, NULL) == ops) ? 0 : -1;
1724
1725         synchronize_net();
1726
1727         return ret;
1728 }
1729 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1730
1731 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1732                         struct ip_tunnel_encap *ipencap)
1733 {
1734         int hlen;
1735
1736         memset(&t->encap, 0, sizeof(t->encap));
1737
1738         hlen = ip6_encap_hlen(ipencap);
1739         if (hlen < 0)
1740                 return hlen;
1741
1742         t->encap.type = ipencap->type;
1743         t->encap.sport = ipencap->sport;
1744         t->encap.dport = ipencap->dport;
1745         t->encap.flags = ipencap->flags;
1746
1747         t->encap_hlen = hlen;
1748         t->hlen = t->encap_hlen + t->tun_hlen;
1749
1750         return 0;
1751 }
1752 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1753
1754 static const struct net_device_ops ip6_tnl_netdev_ops = {
1755         .ndo_init       = ip6_tnl_dev_init,
1756         .ndo_uninit     = ip6_tnl_dev_uninit,
1757         .ndo_start_xmit = ip6_tnl_start_xmit,
1758         .ndo_do_ioctl   = ip6_tnl_ioctl,
1759         .ndo_change_mtu = ip6_tnl_change_mtu,
1760         .ndo_get_stats  = ip6_get_stats,
1761         .ndo_get_iflink = ip6_tnl_get_iflink,
1762 };
1763
1764 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1765                          NETIF_F_FRAGLIST |     \
1766                          NETIF_F_HIGHDMA |      \
1767                          NETIF_F_GSO_SOFTWARE | \
1768                          NETIF_F_HW_CSUM)
1769
1770 /**
1771  * ip6_tnl_dev_setup - setup virtual tunnel device
1772  *   @dev: virtual device associated with tunnel
1773  *
1774  * Description:
1775  *   Initialize function pointers and device parameters
1776  **/
1777
1778 static void ip6_tnl_dev_setup(struct net_device *dev)
1779 {
1780         dev->netdev_ops = &ip6_tnl_netdev_ops;
1781         dev->needs_free_netdev = true;
1782         dev->priv_destructor = ip6_dev_free;
1783
1784         dev->type = ARPHRD_TUNNEL6;
1785         dev->flags |= IFF_NOARP;
1786         dev->addr_len = sizeof(struct in6_addr);
1787         dev->features |= NETIF_F_LLTX;
1788         netif_keep_dst(dev);
1789
1790         dev->features           |= IPXIPX_FEATURES;
1791         dev->hw_features        |= IPXIPX_FEATURES;
1792
1793         /* This perm addr will be used as interface identifier by IPv6 */
1794         dev->addr_assign_type = NET_ADDR_RANDOM;
1795         eth_random_addr(dev->perm_addr);
1796 }
1797
1798
1799 /**
1800  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1801  *   @dev: virtual device associated with tunnel
1802  **/
1803
1804 static inline int
1805 ip6_tnl_dev_init_gen(struct net_device *dev)
1806 {
1807         struct ip6_tnl *t = netdev_priv(dev);
1808         int ret;
1809         int t_hlen;
1810
1811         t->dev = dev;
1812         t->net = dev_net(dev);
1813         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1814         if (!dev->tstats)
1815                 return -ENOMEM;
1816
1817         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1818         if (ret)
1819                 goto free_stats;
1820
1821         ret = gro_cells_init(&t->gro_cells, dev);
1822         if (ret)
1823                 goto destroy_dst;
1824
1825         t->tun_hlen = 0;
1826         t->hlen = t->encap_hlen + t->tun_hlen;
1827         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1828
1829         dev->type = ARPHRD_TUNNEL6;
1830         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1831         dev->mtu = ETH_DATA_LEN - t_hlen;
1832         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1833                 dev->mtu -= 8;
1834         dev->min_mtu = ETH_MIN_MTU;
1835         dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1836
1837         return 0;
1838
1839 destroy_dst:
1840         dst_cache_destroy(&t->dst_cache);
1841 free_stats:
1842         free_percpu(dev->tstats);
1843         dev->tstats = NULL;
1844
1845         return ret;
1846 }
1847
1848 /**
1849  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1850  *   @dev: virtual device associated with tunnel
1851  **/
1852
1853 static int ip6_tnl_dev_init(struct net_device *dev)
1854 {
1855         struct ip6_tnl *t = netdev_priv(dev);
1856         int err = ip6_tnl_dev_init_gen(dev);
1857
1858         if (err)
1859                 return err;
1860         ip6_tnl_link_config(t);
1861         if (t->parms.collect_md) {
1862                 dev->features |= NETIF_F_NETNS_LOCAL;
1863                 netif_keep_dst(dev);
1864         }
1865         return 0;
1866 }
1867
1868 /**
1869  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1870  *   @dev: fallback device
1871  *
1872  * Return: 0
1873  **/
1874
1875 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1876 {
1877         struct ip6_tnl *t = netdev_priv(dev);
1878         struct net *net = dev_net(dev);
1879         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1880
1881         t->parms.proto = IPPROTO_IPV6;
1882         dev_hold(dev);
1883
1884         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1885         return 0;
1886 }
1887
1888 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1889 {
1890         u8 proto;
1891
1892         if (!data || !data[IFLA_IPTUN_PROTO])
1893                 return 0;
1894
1895         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1896         if (proto != IPPROTO_IPV6 &&
1897             proto != IPPROTO_IPIP &&
1898             proto != 0)
1899                 return -EINVAL;
1900
1901         return 0;
1902 }
1903
1904 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1905                                   struct __ip6_tnl_parm *parms)
1906 {
1907         memset(parms, 0, sizeof(*parms));
1908
1909         if (!data)
1910                 return;
1911
1912         if (data[IFLA_IPTUN_LINK])
1913                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1914
1915         if (data[IFLA_IPTUN_LOCAL])
1916                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1917
1918         if (data[IFLA_IPTUN_REMOTE])
1919                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1920
1921         if (data[IFLA_IPTUN_TTL])
1922                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1923
1924         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1925                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1926
1927         if (data[IFLA_IPTUN_FLOWINFO])
1928                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1929
1930         if (data[IFLA_IPTUN_FLAGS])
1931                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1932
1933         if (data[IFLA_IPTUN_PROTO])
1934                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1935
1936         if (data[IFLA_IPTUN_COLLECT_METADATA])
1937                 parms->collect_md = true;
1938
1939         if (data[IFLA_IPTUN_FWMARK])
1940                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1941 }
1942
1943 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1944                                         struct ip_tunnel_encap *ipencap)
1945 {
1946         bool ret = false;
1947
1948         memset(ipencap, 0, sizeof(*ipencap));
1949
1950         if (!data)
1951                 return ret;
1952
1953         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1954                 ret = true;
1955                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1956         }
1957
1958         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1959                 ret = true;
1960                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1961         }
1962
1963         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1964                 ret = true;
1965                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1966         }
1967
1968         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1969                 ret = true;
1970                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1971         }
1972
1973         return ret;
1974 }
1975
1976 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1977                            struct nlattr *tb[], struct nlattr *data[])
1978 {
1979         struct net *net = dev_net(dev);
1980         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1981         struct ip6_tnl *nt, *t;
1982         struct ip_tunnel_encap ipencap;
1983
1984         nt = netdev_priv(dev);
1985
1986         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1987                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1988
1989                 if (err < 0)
1990                         return err;
1991         }
1992
1993         ip6_tnl_netlink_parms(data, &nt->parms);
1994
1995         if (nt->parms.collect_md) {
1996                 if (rtnl_dereference(ip6n->collect_md_tun))
1997                         return -EEXIST;
1998         } else {
1999                 t = ip6_tnl_locate(net, &nt->parms, 0);
2000                 if (!IS_ERR(t))
2001                         return -EEXIST;
2002         }
2003
2004         return ip6_tnl_create2(dev);
2005 }
2006
2007 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2008                               struct nlattr *data[])
2009 {
2010         struct ip6_tnl *t = netdev_priv(dev);
2011         struct __ip6_tnl_parm p;
2012         struct net *net = t->net;
2013         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2014         struct ip_tunnel_encap ipencap;
2015
2016         if (dev == ip6n->fb_tnl_dev)
2017                 return -EINVAL;
2018
2019         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2020                 int err = ip6_tnl_encap_setup(t, &ipencap);
2021
2022                 if (err < 0)
2023                         return err;
2024         }
2025         ip6_tnl_netlink_parms(data, &p);
2026         if (p.collect_md)
2027                 return -EINVAL;
2028
2029         t = ip6_tnl_locate(net, &p, 0);
2030         if (!IS_ERR(t)) {
2031                 if (t->dev != dev)
2032                         return -EEXIST;
2033         } else
2034                 t = netdev_priv(dev);
2035
2036         return ip6_tnl_update(t, &p);
2037 }
2038
2039 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2040 {
2041         struct net *net = dev_net(dev);
2042         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2043
2044         if (dev != ip6n->fb_tnl_dev)
2045                 unregister_netdevice_queue(dev, head);
2046 }
2047
2048 static size_t ip6_tnl_get_size(const struct net_device *dev)
2049 {
2050         return
2051                 /* IFLA_IPTUN_LINK */
2052                 nla_total_size(4) +
2053                 /* IFLA_IPTUN_LOCAL */
2054                 nla_total_size(sizeof(struct in6_addr)) +
2055                 /* IFLA_IPTUN_REMOTE */
2056                 nla_total_size(sizeof(struct in6_addr)) +
2057                 /* IFLA_IPTUN_TTL */
2058                 nla_total_size(1) +
2059                 /* IFLA_IPTUN_ENCAP_LIMIT */
2060                 nla_total_size(1) +
2061                 /* IFLA_IPTUN_FLOWINFO */
2062                 nla_total_size(4) +
2063                 /* IFLA_IPTUN_FLAGS */
2064                 nla_total_size(4) +
2065                 /* IFLA_IPTUN_PROTO */
2066                 nla_total_size(1) +
2067                 /* IFLA_IPTUN_ENCAP_TYPE */
2068                 nla_total_size(2) +
2069                 /* IFLA_IPTUN_ENCAP_FLAGS */
2070                 nla_total_size(2) +
2071                 /* IFLA_IPTUN_ENCAP_SPORT */
2072                 nla_total_size(2) +
2073                 /* IFLA_IPTUN_ENCAP_DPORT */
2074                 nla_total_size(2) +
2075                 /* IFLA_IPTUN_COLLECT_METADATA */
2076                 nla_total_size(0) +
2077                 /* IFLA_IPTUN_FWMARK */
2078                 nla_total_size(4) +
2079                 0;
2080 }
2081
2082 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2083 {
2084         struct ip6_tnl *tunnel = netdev_priv(dev);
2085         struct __ip6_tnl_parm *parm = &tunnel->parms;
2086
2087         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2088             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2089             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2090             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2091             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2092             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2093             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2094             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2095             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2096                 goto nla_put_failure;
2097
2098         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2099             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2100             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2101             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2102                 goto nla_put_failure;
2103
2104         if (parm->collect_md)
2105                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2106                         goto nla_put_failure;
2107
2108         return 0;
2109
2110 nla_put_failure:
2111         return -EMSGSIZE;
2112 }
2113
2114 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2115 {
2116         struct ip6_tnl *tunnel = netdev_priv(dev);
2117
2118         return tunnel->net;
2119 }
2120 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2121
2122 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2123         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2124         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2125         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2126         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2127         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2128         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2129         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2130         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2131         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2132         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2133         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2134         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2135         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2136         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2137 };
2138
2139 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2140         .kind           = "ip6tnl",
2141         .maxtype        = IFLA_IPTUN_MAX,
2142         .policy         = ip6_tnl_policy,
2143         .priv_size      = sizeof(struct ip6_tnl),
2144         .setup          = ip6_tnl_dev_setup,
2145         .validate       = ip6_tnl_validate,
2146         .newlink        = ip6_tnl_newlink,
2147         .changelink     = ip6_tnl_changelink,
2148         .dellink        = ip6_tnl_dellink,
2149         .get_size       = ip6_tnl_get_size,
2150         .fill_info      = ip6_tnl_fill_info,
2151         .get_link_net   = ip6_tnl_get_link_net,
2152 };
2153
2154 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2155         .handler        = ip4ip6_rcv,
2156         .err_handler    = ip4ip6_err,
2157         .priority       =       1,
2158 };
2159
2160 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2161         .handler        = ip6ip6_rcv,
2162         .err_handler    = ip6ip6_err,
2163         .priority       =       1,
2164 };
2165
2166 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2167 {
2168         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2169         struct net_device *dev, *aux;
2170         int h;
2171         struct ip6_tnl *t;
2172         LIST_HEAD(list);
2173
2174         for_each_netdev_safe(net, dev, aux)
2175                 if (dev->rtnl_link_ops == &ip6_link_ops)
2176                         unregister_netdevice_queue(dev, &list);
2177
2178         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2179                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2180                 while (t) {
2181                         /* If dev is in the same netns, it has already
2182                          * been added to the list by the previous loop.
2183                          */
2184                         if (!net_eq(dev_net(t->dev), net))
2185                                 unregister_netdevice_queue(t->dev, &list);
2186                         t = rtnl_dereference(t->next);
2187                 }
2188         }
2189
2190         unregister_netdevice_many(&list);
2191 }
2192
2193 static int __net_init ip6_tnl_init_net(struct net *net)
2194 {
2195         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2196         struct ip6_tnl *t = NULL;
2197         int err;
2198
2199         ip6n->tnls[0] = ip6n->tnls_wc;
2200         ip6n->tnls[1] = ip6n->tnls_r_l;
2201
2202         err = -ENOMEM;
2203         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2204                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2205
2206         if (!ip6n->fb_tnl_dev)
2207                 goto err_alloc_dev;
2208         dev_net_set(ip6n->fb_tnl_dev, net);
2209         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2210         /* FB netdevice is special: we have one, and only one per netns.
2211          * Allowing to move it to another netns is clearly unsafe.
2212          */
2213         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2214
2215         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2216         if (err < 0)
2217                 goto err_register;
2218
2219         err = register_netdev(ip6n->fb_tnl_dev);
2220         if (err < 0)
2221                 goto err_register;
2222
2223         t = netdev_priv(ip6n->fb_tnl_dev);
2224
2225         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2226         return 0;
2227
2228 err_register:
2229         free_netdev(ip6n->fb_tnl_dev);
2230 err_alloc_dev:
2231         return err;
2232 }
2233
2234 static void __net_exit ip6_tnl_exit_net(struct net *net)
2235 {
2236         rtnl_lock();
2237         ip6_tnl_destroy_tunnels(net);
2238         rtnl_unlock();
2239 }
2240
2241 static struct pernet_operations ip6_tnl_net_ops = {
2242         .init = ip6_tnl_init_net,
2243         .exit = ip6_tnl_exit_net,
2244         .id   = &ip6_tnl_net_id,
2245         .size = sizeof(struct ip6_tnl_net),
2246 };
2247
2248 /**
2249  * ip6_tunnel_init - register protocol and reserve needed resources
2250  *
2251  * Return: 0 on success
2252  **/
2253
2254 static int __init ip6_tunnel_init(void)
2255 {
2256         int  err;
2257
2258         err = register_pernet_device(&ip6_tnl_net_ops);
2259         if (err < 0)
2260                 goto out_pernet;
2261
2262         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2263         if (err < 0) {
2264                 pr_err("%s: can't register ip4ip6\n", __func__);
2265                 goto out_ip4ip6;
2266         }
2267
2268         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2269         if (err < 0) {
2270                 pr_err("%s: can't register ip6ip6\n", __func__);
2271                 goto out_ip6ip6;
2272         }
2273         err = rtnl_link_register(&ip6_link_ops);
2274         if (err < 0)
2275                 goto rtnl_link_failed;
2276
2277         return 0;
2278
2279 rtnl_link_failed:
2280         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2281 out_ip6ip6:
2282         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2283 out_ip4ip6:
2284         unregister_pernet_device(&ip6_tnl_net_ops);
2285 out_pernet:
2286         return err;
2287 }
2288
2289 /**
2290  * ip6_tunnel_cleanup - free resources and unregister protocol
2291  **/
2292
2293 static void __exit ip6_tunnel_cleanup(void)
2294 {
2295         rtnl_link_unregister(&ip6_link_ops);
2296         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2297                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2298
2299         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2300                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2301
2302         unregister_pernet_device(&ip6_tnl_net_ops);
2303 }
2304
2305 module_init(ip6_tunnel_init);
2306 module_exit(ip6_tunnel_cleanup);