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 && t->dev->flags & IFF_UP)
 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         ip6_tnl_link(ip6n, t);
 276         return 0;
 277
 278 out:
 279         return err;
 280 }
 281
 282 /**
 283  * ip6_tnl_create - create a new tunnel
 284  *   @p: tunnel parameters
 285  *   @pt: pointer to new tunnel
 286  *
 287  * Description:
 288  *   Create tunnel matching given parameters.
 289  *
 290  * Return:
 291  *   created tunnel or error pointer
 292  **/
 293
 294 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 295 {
 296         struct net_device *dev;
 297         struct ip6_tnl *t;
 298         char name[IFNAMSIZ];
 299         int err = -E2BIG;
 300
 301         if (p->name[0]) {
 302                 if (!dev_valid_name(p->name))
 303                         goto failed;
 304                 strlcpy(name, p->name, IFNAMSIZ);
 305         } else {
 306                 sprintf(name, "ip6tnl%%d");
 307         }
 308         err = -ENOMEM;
 309         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 310                            ip6_tnl_dev_setup);
 311         if (!dev)
 312                 goto failed;
 313
 314         dev_net_set(dev, net);
 315
 316         t = netdev_priv(dev);
 317         t->parms = *p;
 318         t->net = dev_net(dev);
 319         err = ip6_tnl_create2(dev);
 320         if (err < 0)
 321                 goto failed_free;
 322
 323         return t;
 324
 325 failed_free:
 326         free_netdev(dev);
 327 failed:
 328         return ERR_PTR(err);
 329 }
 330
 331 /**
 332  * ip6_tnl_locate - find or create tunnel matching given parameters
 333  *   @p: tunnel parameters
 334  *   @create: != 0 if allowed to create new tunnel if no match found
 335  *
 336  * Description:
 337  *   ip6_tnl_locate() first tries to locate an existing tunnel
 338  *   based on @parms. If this is unsuccessful, but @create is set a new
 339  *   tunnel device is created and registered for use.
 340  *
 341  * Return:
 342  *   matching tunnel or error pointer
 343  **/
 344
 345 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 346                 struct __ip6_tnl_parm *p, int create)
 347 {
 348         const struct in6_addr *remote = &p->raddr;
 349         const struct in6_addr *local = &p->laddr;
 350         struct ip6_tnl __rcu **tp;
 351         struct ip6_tnl *t;
 352         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 353
 354         for (tp = ip6_tnl_bucket(ip6n, p);
 355              (t = rtnl_dereference(*tp)) != NULL;
 356              tp = &t->next) {
 357                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 358                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 359                         if (create)
 360                                 return ERR_PTR(-EEXIST);
 361
 362                         return t;
 363                 }
 364         }
 365         if (!create)
 366                 return ERR_PTR(-ENODEV);
 367         return ip6_tnl_create(net, p);
 368 }
 369
 370 /**
 371  * ip6_tnl_dev_uninit - tunnel device uninitializer
 372  *   @dev: the device to be destroyed
 373  *
 374  * Description:
 375  *   ip6_tnl_dev_uninit() removes tunnel from its list
 376  **/
 377
 378 static void
 379 ip6_tnl_dev_uninit(struct net_device *dev)
 380 {
 381         struct ip6_tnl *t = netdev_priv(dev);
 382         struct net *net = t->net;
 383         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 384
 385         if (dev == ip6n->fb_tnl_dev)
 386                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 387         else
 388                 ip6_tnl_unlink(ip6n, t);
 389         dst_cache_reset(&t->dst_cache);
 390         dev_put(dev);
 391 }
 392
 393 /**
 394  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 395  *   @skb: received socket buffer
 396  *
 397  * Return:
 398  *   0 if none was found,
 399  *   else index to encapsulation limit
 400  **/
 401
 402 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 403 {
 404         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 405         unsigned int nhoff = raw - skb->data;
 406         unsigned int off = nhoff + sizeof(*ipv6h);
 407         u8 next, nexthdr = ipv6h->nexthdr;
 408
 409         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 410                 struct ipv6_opt_hdr *hdr;
 411                 u16 optlen;
 412
 413                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 414                         break;
 415
 416                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 417                 if (nexthdr == NEXTHDR_FRAGMENT) {
 418                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 419                         if (frag_hdr->frag_off)
 420                                 break;
 421                         optlen = 8;
 422                 } else if (nexthdr == NEXTHDR_AUTH) {
 423                         optlen = (hdr->hdrlen + 2) << 2;
 424                 } else {
 425                         optlen = ipv6_optlen(hdr);
 426                 }
 427                 /* cache hdr->nexthdr, since pskb_may_pull() might
 428                  * invalidate hdr
 429                  */
 430                 next = hdr->nexthdr;
 431                 if (nexthdr == NEXTHDR_DEST) {
 432                         u16 i = 2;
 433
 434                         /* Remember : hdr is no longer valid at this point. */
 435                         if (!pskb_may_pull(skb, off + optlen))
 436                                 break;
 437
 438                         while (1) {
 439                                 struct ipv6_tlv_tnl_enc_lim *tel;
 440
 441                                 /* No more room for encapsulation limit */
 442                                 if (i + sizeof(*tel) > optlen)
 443                                         break;
 444
 445                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 446                                 /* return index of option if found and valid */
 447                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 448                                     tel->length == 1)
 449                                         return i + off - nhoff;
 450                                 /* else jump to next option */
 451                                 if (tel->type)
 452                                         i += tel->length + 2;
 453                                 else
 454                                         i++;
 455                         }
 456                 }
 457                 nexthdr = next;
 458                 off += optlen;
 459         }
 460         return 0;
 461 }
 462 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 463
 464 /**
 465  * ip6_tnl_err - tunnel error handler
 466  *
 467  * Description:
 468  *   ip6_tnl_err() should handle errors in the tunnel according
 469  *   to the specifications in RFC 2473.
 470  **/
 471
 472 static int
 473 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 474             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 475 {
 476         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 477         struct ip6_tnl *t;
 478         int rel_msg = 0;
 479         u8 rel_type = ICMPV6_DEST_UNREACH;
 480         u8 rel_code = ICMPV6_ADDR_UNREACH;
 481         u8 tproto;
 482         __u32 rel_info = 0;
 483         __u16 len;
 484         int err = -ENOENT;
 485
 486         /* If the packet doesn't contain the original IPv6 header we are
 487            in trouble since we might need the source address for further
 488            processing of the error. */
 489
 490         rcu_read_lock();
 491         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 492         if (!t)
 493                 goto out;
 494
 495         tproto = ACCESS_ONCE(t->parms.proto);
 496         if (tproto != ipproto && tproto != 0)
 497                 goto out;
 498
 499         err = 0;
 500
 501         switch (*type) {
 502                 __u32 teli;
 503                 struct ipv6_tlv_tnl_enc_lim *tel;
 504                 __u32 mtu;
 505         case ICMPV6_DEST_UNREACH:
 506                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 507                                     t->parms.name);
 508                 rel_msg = 1;
 509                 break;
 510         case ICMPV6_TIME_EXCEED:
 511                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 512                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 513                                             t->parms.name);
 514                         rel_msg = 1;
 515                 }
 516                 break;
 517         case ICMPV6_PARAMPROB:
 518                 teli = 0;
 519                 if ((*code) == ICMPV6_HDR_FIELD)
 520                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 521
 522                 if (teli && teli == *info - 2) {
 523                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 524                         if (tel->encap_limit == 0) {
 525                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 526                                                     t->parms.name);
 527                                 rel_msg = 1;
 528                         }
 529                 } else {
 530                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 531                                             t->parms.name);
 532                 }
 533                 break;
 534         case ICMPV6_PKT_TOOBIG:
 535                 mtu = *info - offset;
 536                 if (mtu < IPV6_MIN_MTU)
 537                         mtu = IPV6_MIN_MTU;
 538                 t->dev->mtu = mtu;
 539
 540                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 541                 if (len > mtu) {
 542                         rel_type = ICMPV6_PKT_TOOBIG;
 543                         rel_code = 0;
 544                         rel_info = mtu;
 545                         rel_msg = 1;
 546                 }
 547                 break;
 548         }
 549
 550         *type = rel_type;
 551         *code = rel_code;
 552         *info = rel_info;
 553         *msg = rel_msg;
 554
 555 out:
 556         rcu_read_unlock();
 557         return err;
 558 }
 559
 560 static int
 561 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 562            u8 type, u8 code, int offset, __be32 info)
 563 {
 564         int rel_msg = 0;
 565         u8 rel_type = type;
 566         u8 rel_code = code;
 567         __u32 rel_info = ntohl(info);
 568         int err;
 569         struct sk_buff *skb2;
 570         const struct iphdr *eiph;
 571         struct rtable *rt;
 572         struct flowi4 fl4;
 573
 574         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 575                           &rel_msg, &rel_info, offset);
 576         if (err < 0)
 577                 return err;
 578
 579         if (rel_msg == 0)
 580                 return 0;
 581
 582         switch (rel_type) {
 583         case ICMPV6_DEST_UNREACH:
 584                 if (rel_code != ICMPV6_ADDR_UNREACH)
 585                         return 0;
 586                 rel_type = ICMP_DEST_UNREACH;
 587                 rel_code = ICMP_HOST_UNREACH;
 588                 break;
 589         case ICMPV6_PKT_TOOBIG:
 590                 if (rel_code != 0)
 591                         return 0;
 592                 rel_type = ICMP_DEST_UNREACH;
 593                 rel_code = ICMP_FRAG_NEEDED;
 594                 break;
 595         case NDISC_REDIRECT:
 596                 rel_type = ICMP_REDIRECT;
 597                 rel_code = ICMP_REDIR_HOST;
 598         default:
 599                 return 0;
 600         }
 601
 602         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 603                 return 0;
 604
 605         skb2 = skb_clone(skb, GFP_ATOMIC);
 606         if (!skb2)
 607                 return 0;
 608
 609         skb_dst_drop(skb2);
 610
 611         skb_pull(skb2, offset);
 612         skb_reset_network_header(skb2);
 613         eiph = ip_hdr(skb2);
 614
 615         /* Try to guess incoming interface */
 616         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 617                                    eiph->saddr, 0,
 618                                    0, 0,
 619                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 620         if (IS_ERR(rt))
 621                 goto out;
 622
 623         skb2->dev = rt->dst.dev;
 624
 625         /* route "incoming" packet */
 626         if (rt->rt_flags & RTCF_LOCAL) {
 627                 ip_rt_put(rt);
 628                 rt = NULL;
 629                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 630                                            eiph->daddr, eiph->saddr,
 631                                            0, 0,
 632                                            IPPROTO_IPIP,
 633                                            RT_TOS(eiph->tos), 0);
 634                 if (IS_ERR(rt) ||
 635                     rt->dst.dev->type != ARPHRD_TUNNEL6) {
 636                         if (!IS_ERR(rt))
 637                                 ip_rt_put(rt);
 638                         goto out;
 639                 }
 640                 skb_dst_set(skb2, &rt->dst);
 641         } else {
 642                 ip_rt_put(rt);
 643                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 644                                    skb2->dev) ||
 645                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 646                         goto out;
 647         }
 648
 649         /* change mtu on this route */
 650         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 651                 if (rel_info > dst_mtu(skb_dst(skb2)))
 652                         goto out;
 653
 654                 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
 655         }
 656         if (rel_type == ICMP_REDIRECT)
 657                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 658
 659         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 660
 661 out:
 662         kfree_skb(skb2);
 663         return 0;
 664 }
 665
 666 static int
 667 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 668            u8 type, u8 code, int offset, __be32 info)
 669 {
 670         int rel_msg = 0;
 671         u8 rel_type = type;
 672         u8 rel_code = code;
 673         __u32 rel_info = ntohl(info);
 674         int err;
 675
 676         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 677                           &rel_msg, &rel_info, offset);
 678         if (err < 0)
 679                 return err;
 680
 681         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 682                 struct rt6_info *rt;
 683                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 684
 685                 if (!skb2)
 686                         return 0;
 687
 688                 skb_dst_drop(skb2);
 689                 skb_pull(skb2, offset);
 690                 skb_reset_network_header(skb2);
 691
 692                 /* Try to guess incoming interface */
 693                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 694                                 NULL, 0, 0);
 695
 696                 if (rt && rt->dst.dev)
 697                         skb2->dev = rt->dst.dev;
 698
 699                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 700
 701                 ip6_rt_put(rt);
 702
 703                 kfree_skb(skb2);
 704         }
 705
 706         return 0;
 707 }
 708
 709 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 710                                        const struct ipv6hdr *ipv6h,
 711                                        struct sk_buff *skb)
 712 {
 713         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 714
 715         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 716                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 717
 718         return IP6_ECN_decapsulate(ipv6h, skb);
 719 }
 720
 721 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 722                                        const struct ipv6hdr *ipv6h,
 723                                        struct sk_buff *skb)
 724 {
 725         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 726                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 727
 728         return IP6_ECN_decapsulate(ipv6h, skb);
 729 }
 730
 731 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 732                              const struct in6_addr *laddr,
 733                              const struct in6_addr *raddr)
 734 {
 735         struct __ip6_tnl_parm *p = &t->parms;
 736         int ltype = ipv6_addr_type(laddr);
 737         int rtype = ipv6_addr_type(raddr);
 738         __u32 flags = 0;
 739
 740         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 741                 flags = IP6_TNL_F_CAP_PER_PACKET;
 742         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 743                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 744                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 745                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 746                 if (ltype&IPV6_ADDR_UNICAST)
 747                         flags |= IP6_TNL_F_CAP_XMIT;
 748                 if (rtype&IPV6_ADDR_UNICAST)
 749                         flags |= IP6_TNL_F_CAP_RCV;
 750         }
 751         return flags;
 752 }
 753 EXPORT_SYMBOL(ip6_tnl_get_cap);
 754
 755 /* called with rcu_read_lock() */
 756 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 757                                   const struct in6_addr *laddr,
 758                                   const struct in6_addr *raddr)
 759 {
 760         struct __ip6_tnl_parm *p = &t->parms;
 761         int ret = 0;
 762         struct net *net = t->net;
 763
 764         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 765             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 766              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 767                 struct net_device *ldev = NULL;
 768
 769                 if (p->link)
 770                         ldev = dev_get_by_index_rcu(net, p->link);
 771
 772                 if ((ipv6_addr_is_multicast(laddr) ||
 773                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 774                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 775                         ret = 1;
 776         }
 777         return ret;
 778 }
 779 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 780
 781 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 782                          const struct tnl_ptk_info *tpi,
 783                          struct metadata_dst *tun_dst,
 784                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 785                                                 const struct ipv6hdr *ipv6h,
 786                                                 struct sk_buff *skb),
 787                          bool log_ecn_err)
 788 {
 789         struct pcpu_sw_netstats *tstats;
 790         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 791         int err;
 792
 793         if ((!(tpi->flags & TUNNEL_CSUM) &&
 794              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 795             ((tpi->flags & TUNNEL_CSUM) &&
 796              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 797                 tunnel->dev->stats.rx_crc_errors++;
 798                 tunnel->dev->stats.rx_errors++;
 799                 goto drop;
 800         }
 801
 802         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 803                 if (!(tpi->flags & TUNNEL_SEQ) ||
 804                     (tunnel->i_seqno &&
 805                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 806                         tunnel->dev->stats.rx_fifo_errors++;
 807                         tunnel->dev->stats.rx_errors++;
 808                         goto drop;
 809                 }
 810                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 811         }
 812
 813         skb->protocol = tpi->proto;
 814
 815         /* Warning: All skb pointers will be invalidated! */
 816         if (tunnel->dev->type == ARPHRD_ETHER) {
 817                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 818                         tunnel->dev->stats.rx_length_errors++;
 819                         tunnel->dev->stats.rx_errors++;
 820                         goto drop;
 821                 }
 822
 823                 ipv6h = ipv6_hdr(skb);
 824                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 825                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 826         } else {
 827                 skb->dev = tunnel->dev;
 828         }
 829
 830         skb_reset_network_header(skb);
 831         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 832
 833         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 834
 835         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 836         if (unlikely(err)) {
 837                 if (log_ecn_err)
 838                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 839                                              &ipv6h->saddr,
 840                                              ipv6_get_dsfield(ipv6h));
 841                 if (err > 1) {
 842                         ++tunnel->dev->stats.rx_frame_errors;
 843                         ++tunnel->dev->stats.rx_errors;
 844                         goto drop;
 845                 }
 846         }
 847
 848         tstats = this_cpu_ptr(tunnel->dev->tstats);
 849         u64_stats_update_begin(&tstats->syncp);
 850         tstats->rx_packets++;
 851         tstats->rx_bytes += skb->len;
 852         u64_stats_update_end(&tstats->syncp);
 853
 854         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 855
 856         if (tun_dst)
 857                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 858
 859         gro_cells_receive(&tunnel->gro_cells, skb);
 860         return 0;
 861
 862 drop:
 863         if (tun_dst)
 864                 dst_release((struct dst_entry *)tun_dst);
 865         kfree_skb(skb);
 866         return 0;
 867 }
 868
 869 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 870                 const struct tnl_ptk_info *tpi,
 871                 struct metadata_dst *tun_dst,
 872                 bool log_ecn_err)
 873 {
 874         int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 875                                     const struct ipv6hdr *ipv6h,
 876                                     struct sk_buff *skb);
 877
 878         dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
 879         if (tpi->proto == htons(ETH_P_IP))
 880                 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
 881
 882         return __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
 883                              log_ecn_err);
 884 }
 885 EXPORT_SYMBOL(ip6_tnl_rcv);
 886
 887 static const struct tnl_ptk_info tpi_v6 = {
 888         /* no tunnel info required for ipxip6. */
 889         .proto = htons(ETH_P_IPV6),
 890 };
 891
 892 static const struct tnl_ptk_info tpi_v4 = {
 893         /* no tunnel info required for ipxip6. */
 894         .proto = htons(ETH_P_IP),
 895 };
 896
 897 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 898                       const struct tnl_ptk_info *tpi,
 899                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 900                                                   const struct ipv6hdr *ipv6h,
 901                                                   struct sk_buff *skb))
 902 {
 903         struct ip6_tnl *t;
 904         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 905         struct metadata_dst *tun_dst = NULL;
 906         int ret = -1;
 907
 908         rcu_read_lock();
 909         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 910
 911         if (t) {
 912                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 913
 914                 if (tproto != ipproto && tproto != 0)
 915                         goto drop;
 916                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 917                         goto drop;
 918                 ipv6h = ipv6_hdr(skb);
 919                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 920                         goto drop;
 921                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 922                         goto drop;
 923                 if (t->parms.collect_md) {
 924                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 925                         if (!tun_dst)
 926                                 goto drop;
 927                 }
 928                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 929                                     log_ecn_error);
 930         }
 931
 932         rcu_read_unlock();
 933
 934         return ret;
 935
 936 drop:
 937         rcu_read_unlock();
 938         kfree_skb(skb);
 939         return 0;
 940 }
 941
 942 static int ip4ip6_rcv(struct sk_buff *skb)
 943 {
 944         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 945                           ip4ip6_dscp_ecn_decapsulate);
 946 }
 947
 948 static int ip6ip6_rcv(struct sk_buff *skb)
 949 {
 950         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 951                           ip6ip6_dscp_ecn_decapsulate);
 952 }
 953
 954 struct ipv6_tel_txoption {
 955         struct ipv6_txoptions ops;
 956         __u8 dst_opt[8];
 957 };
 958
 959 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 960 {
 961         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 962
 963         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 964         opt->dst_opt[3] = 1;
 965         opt->dst_opt[4] = encap_limit;
 966         opt->dst_opt[5] = IPV6_TLV_PADN;
 967         opt->dst_opt[6] = 1;
 968
 969         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 970         opt->ops.opt_nflen = 8;
 971 }
 972
 973 /**
 974  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 975  *   @t: the outgoing tunnel device
 976  *   @hdr: IPv6 header from the incoming packet
 977  *
 978  * Description:
 979  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 980  *   doesn't match source of incoming packet.
 981  *
 982  * Return:
 983  *   1 if conflict,
 984  *   0 else
 985  **/
 986
 987 static inline bool
 988 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 989 {
 990         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 991 }
 992
 993 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 994                      const struct in6_addr *laddr,
 995                      const struct in6_addr *raddr)
 996 {
 997         struct __ip6_tnl_parm *p = &t->parms;
 998         int ret = 0;
 999         struct net *net = t->net;
1000
1001         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1002             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1003              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1004                 struct net_device *ldev = NULL;
1005
1006                 rcu_read_lock();
1007                 if (p->link)
1008                         ldev = dev_get_by_index_rcu(net, p->link);
1009
1010                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1011                         pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1012                                             p->name);
1013                 else if (!ipv6_addr_is_multicast(raddr) &&
1014                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1015                         pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1016                                             p->name);
1017                 else
1018                         ret = 1;
1019                 rcu_read_unlock();
1020         }
1021         return ret;
1022 }
1023 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1024
1025 /**
1026  * ip6_tnl_xmit - encapsulate packet and send
1027  *   @skb: the outgoing socket buffer
1028  *   @dev: the outgoing tunnel device
1029  *   @dsfield: dscp code for outer header
1030  *   @fl6: flow of tunneled packet
1031  *   @encap_limit: encapsulation limit
1032  *   @pmtu: Path MTU is stored if packet is too big
1033  *   @proto: next header value
1034  *
1035  * Description:
1036  *   Build new header and do some sanity checks on the packet before sending
1037  *   it.
1038  *
1039  * Return:
1040  *   0 on success
1041  *   -1 fail
1042  *   %-EMSGSIZE message too big. return mtu in this case.
1043  **/
1044
1045 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1046                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1047                  __u8 proto)
1048 {
1049         struct ip6_tnl *t = netdev_priv(dev);
1050         struct net *net = t->net;
1051         struct net_device_stats *stats = &t->dev->stats;
1052         struct ipv6hdr *ipv6h;
1053         struct ipv6_tel_txoption opt;
1054         struct dst_entry *dst = NULL, *ndst = NULL;
1055         struct net_device *tdev;
1056         int mtu;
1057         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1058         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1059         unsigned int max_headroom = psh_hlen;
1060         bool use_cache = false;
1061         u8 hop_limit;
1062         int err = -1;
1063
1064         if (t->parms.collect_md) {
1065                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1066                 goto route_lookup;
1067         } else {
1068                 hop_limit = t->parms.hop_limit;
1069         }
1070
1071         /* NBMA tunnel */
1072         if (ipv6_addr_any(&t->parms.raddr)) {
1073                 if (skb->protocol == htons(ETH_P_IPV6)) {
1074                         struct in6_addr *addr6;
1075                         struct neighbour *neigh;
1076                         int addr_type;
1077
1078                         if (!skb_dst(skb))
1079                                 goto tx_err_link_failure;
1080
1081                         neigh = dst_neigh_lookup(skb_dst(skb),
1082                                                  &ipv6_hdr(skb)->daddr);
1083                         if (!neigh)
1084                                 goto tx_err_link_failure;
1085
1086                         addr6 = (struct in6_addr *)&neigh->primary_key;
1087                         addr_type = ipv6_addr_type(addr6);
1088
1089                         if (addr_type == IPV6_ADDR_ANY)
1090                                 addr6 = &ipv6_hdr(skb)->daddr;
1091
1092                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1093                         neigh_release(neigh);
1094                 }
1095         } else if (t->parms.proto != 0 && !(t->parms.flags &
1096                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1097                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1098                 /* enable the cache only if neither the outer protocol nor the
1099                  * routing decision depends on the current inner header value
1100                  */
1101                 use_cache = true;
1102         }
1103
1104         if (use_cache)
1105                 dst = dst_cache_get(&t->dst_cache);
1106
1107         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1108                 goto tx_err_link_failure;
1109
1110         if (!dst) {
1111 route_lookup:
1112                 /* add dsfield to flowlabel for route lookup */
1113                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1114
1115                 dst = ip6_route_output(net, NULL, fl6);
1116
1117                 if (dst->error)
1118                         goto tx_err_link_failure;
1119                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1120                 if (IS_ERR(dst)) {
1121                         err = PTR_ERR(dst);
1122                         dst = NULL;
1123                         goto tx_err_link_failure;
1124                 }
1125                 if (t->parms.collect_md &&
1126                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1127                                        &fl6->daddr, 0, &fl6->saddr))
1128                         goto tx_err_link_failure;
1129                 ndst = dst;
1130         }
1131
1132         tdev = dst->dev;
1133
1134         if (tdev == dev) {
1135                 stats->collisions++;
1136                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1137                                      t->parms.name);
1138                 goto tx_err_dst_release;
1139         }
1140         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1141         if (encap_limit >= 0) {
1142                 max_headroom += 8;
1143                 mtu -= 8;
1144         }
1145         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1146                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1147
1148         skb_dst_update_pmtu_no_confirm(skb, mtu);
1149         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1150                 *pmtu = mtu;
1151                 err = -EMSGSIZE;
1152                 goto tx_err_dst_release;
1153         }
1154
1155         if (t->err_count > 0) {
1156                 if (time_before(jiffies,
1157                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1158                         t->err_count--;
1159
1160                         dst_link_failure(skb);
1161                 } else {
1162                         t->err_count = 0;
1163                 }
1164         }
1165
1166         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1167
1168         /*
1169          * Okay, now see if we can stuff it in the buffer as-is.
1170          */
1171         max_headroom += LL_RESERVED_SPACE(tdev);
1172
1173         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1174             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1175                 struct sk_buff *new_skb;
1176
1177                 new_skb = skb_realloc_headroom(skb, max_headroom);
1178                 if (!new_skb)
1179                         goto tx_err_dst_release;
1180
1181                 if (skb->sk)
1182                         skb_set_owner_w(new_skb, skb->sk);
1183                 consume_skb(skb);
1184                 skb = new_skb;
1185         }
1186
1187         if (t->parms.collect_md) {
1188                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1189                         goto tx_err_dst_release;
1190         } else {
1191                 if (use_cache && ndst)
1192                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1193         }
1194         skb_dst_set(skb, dst);
1195
1196         hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1197
1198         /* Calculate max headroom for all the headers and adjust
1199          * needed_headroom if necessary.
1200          */
1201         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1202                         + dst->header_len + t->hlen;
1203         if (max_headroom > dev->needed_headroom)
1204                 dev->needed_headroom = max_headroom;
1205
1206         err = ip6_tnl_encap(skb, t, &proto, fl6);
1207         if (err)
1208                 return err;
1209
1210         if (encap_limit >= 0) {
1211                 init_tel_txopt(&opt, encap_limit);
1212                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1213         }
1214
1215         skb_push(skb, sizeof(struct ipv6hdr));
1216         skb_reset_network_header(skb);
1217         ipv6h = ipv6_hdr(skb);
1218         ip6_flow_hdr(ipv6h, dsfield,
1219                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1220         ipv6h->hop_limit = hop_limit;
1221         ipv6h->nexthdr = proto;
1222         ipv6h->saddr = fl6->saddr;
1223         ipv6h->daddr = fl6->daddr;
1224         ip6tunnel_xmit(NULL, skb, dev);
1225         return 0;
1226 tx_err_link_failure:
1227         stats->tx_carrier_errors++;
1228         dst_link_failure(skb);
1229 tx_err_dst_release:
1230         dst_release(dst);
1231         return err;
1232 }
1233 EXPORT_SYMBOL(ip6_tnl_xmit);
1234
1235 static inline int
1236 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1237 {
1238         struct ip6_tnl *t = netdev_priv(dev);
1239         const struct iphdr  *iph;
1240         int encap_limit = -1;
1241         struct flowi6 fl6;
1242         __u8 dsfield;
1243         __u32 mtu;
1244         u8 tproto;
1245         int err;
1246
1247         /* ensure we can access the full inner ip header */
1248         if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1249                 return -1;
1250
1251         iph = ip_hdr(skb);
1252         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1253
1254         tproto = ACCESS_ONCE(t->parms.proto);
1255         if (tproto != IPPROTO_IPIP && tproto != 0)
1256                 return -1;
1257
1258         if (t->parms.collect_md) {
1259                 struct ip_tunnel_info *tun_info;
1260                 const struct ip_tunnel_key *key;
1261
1262                 tun_info = skb_tunnel_info(skb);
1263                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1264                              ip_tunnel_info_af(tun_info) != AF_INET6))
1265                         return -1;
1266                 key = &tun_info->key;
1267                 memset(&fl6, 0, sizeof(fl6));
1268                 fl6.flowi6_proto = IPPROTO_IPIP;
1269                 fl6.daddr = key->u.ipv6.dst;
1270                 fl6.flowlabel = key->label;
1271                 dsfield =  key->tos;
1272         } else {
1273                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1274                         encap_limit = t->parms.encap_limit;
1275
1276                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1277                 fl6.flowi6_proto = IPPROTO_IPIP;
1278
1279                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1280                         dsfield = ipv4_get_dsfield(iph);
1281                 else
1282                         dsfield = ip6_tclass(t->parms.flowinfo);
1283                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1284                         fl6.flowi6_mark = skb->mark;
1285                 else
1286                         fl6.flowi6_mark = t->parms.fwmark;
1287         }
1288
1289         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1290         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1291
1292         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1293                 return -1;
1294
1295         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1296
1297         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1298                            IPPROTO_IPIP);
1299         if (err != 0) {
1300                 /* XXX: send ICMP error even if DF is not set. */
1301                 if (err == -EMSGSIZE)
1302                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1303                                   htonl(mtu));
1304                 return -1;
1305         }
1306
1307         return 0;
1308 }
1309
1310 static inline int
1311 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1312 {
1313         struct ip6_tnl *t = netdev_priv(dev);
1314         struct ipv6hdr *ipv6h;
1315         int encap_limit = -1;
1316         __u16 offset;
1317         struct flowi6 fl6;
1318         __u8 dsfield;
1319         __u32 mtu;
1320         u8 tproto;
1321         int err;
1322
1323         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1324                 return -1;
1325
1326         ipv6h = ipv6_hdr(skb);
1327         tproto = ACCESS_ONCE(t->parms.proto);
1328         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1329             ip6_tnl_addr_conflict(t, ipv6h))
1330                 return -1;
1331
1332         if (t->parms.collect_md) {
1333                 struct ip_tunnel_info *tun_info;
1334                 const struct ip_tunnel_key *key;
1335
1336                 tun_info = skb_tunnel_info(skb);
1337                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1338                              ip_tunnel_info_af(tun_info) != AF_INET6))
1339                         return -1;
1340                 key = &tun_info->key;
1341                 memset(&fl6, 0, sizeof(fl6));
1342                 fl6.flowi6_proto = IPPROTO_IPV6;
1343                 fl6.daddr = key->u.ipv6.dst;
1344                 fl6.flowlabel = key->label;
1345                 dsfield = key->tos;
1346         } else {
1347                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1348                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1349                 ipv6h = ipv6_hdr(skb);
1350                 if (offset > 0) {
1351                         struct ipv6_tlv_tnl_enc_lim *tel;
1352
1353                         tel = (void *)&skb_network_header(skb)[offset];
1354                         if (tel->encap_limit == 0) {
1355                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1356                                             ICMPV6_HDR_FIELD, offset + 2);
1357                                 return -1;
1358                         }
1359                         encap_limit = tel->encap_limit - 1;
1360                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1361                         encap_limit = t->parms.encap_limit;
1362                 }
1363
1364                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1365                 fl6.flowi6_proto = IPPROTO_IPV6;
1366
1367                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1368                         dsfield = ipv6_get_dsfield(ipv6h);
1369                 else
1370                         dsfield = ip6_tclass(t->parms.flowinfo);
1371                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1372                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1373                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1374                         fl6.flowi6_mark = skb->mark;
1375                 else
1376                         fl6.flowi6_mark = t->parms.fwmark;
1377         }
1378
1379         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1380         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1381
1382         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1383                 return -1;
1384
1385         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1386
1387         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1388                            IPPROTO_IPV6);
1389         if (err != 0) {
1390                 if (err == -EMSGSIZE)
1391                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1392                 return -1;
1393         }
1394
1395         return 0;
1396 }
1397
1398 static netdev_tx_t
1399 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1400 {
1401         struct ip6_tnl *t = netdev_priv(dev);
1402         struct net_device_stats *stats = &t->dev->stats;
1403         int ret;
1404
1405         switch (skb->protocol) {
1406         case htons(ETH_P_IP):
1407                 ret = ip4ip6_tnl_xmit(skb, dev);
1408                 break;
1409         case htons(ETH_P_IPV6):
1410                 ret = ip6ip6_tnl_xmit(skb, dev);
1411                 break;
1412         default:
1413                 goto tx_err;
1414         }
1415
1416         if (ret < 0)
1417                 goto tx_err;
1418
1419         return NETDEV_TX_OK;
1420
1421 tx_err:
1422         stats->tx_errors++;
1423         stats->tx_dropped++;
1424         kfree_skb(skb);
1425         return NETDEV_TX_OK;
1426 }
1427
1428 static void ip6_tnl_link_config(struct ip6_tnl *t)
1429 {
1430         struct net_device *dev = t->dev;
1431         struct __ip6_tnl_parm *p = &t->parms;
1432         struct flowi6 *fl6 = &t->fl.u.ip6;
1433         int t_hlen;
1434
1435         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1436         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1437
1438         /* Set up flowi template */
1439         fl6->saddr = p->laddr;
1440         fl6->daddr = p->raddr;
1441         fl6->flowi6_oif = p->link;
1442         fl6->flowlabel = 0;
1443
1444         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1445                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1446         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1447                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1448
1449         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1450         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1451
1452         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1453                 dev->flags |= IFF_POINTOPOINT;
1454         else
1455                 dev->flags &= ~IFF_POINTOPOINT;
1456
1457         t->tun_hlen = 0;
1458         t->hlen = t->encap_hlen + t->tun_hlen;
1459         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1460
1461         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1462                 int strict = (ipv6_addr_type(&p->raddr) &
1463                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1464
1465                 struct rt6_info *rt = rt6_lookup(t->net,
1466                                                  &p->raddr, &p->laddr,
1467                                                  p->link, strict);
1468
1469                 if (!rt)
1470                         return;
1471
1472                 if (rt->dst.dev) {
1473                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1474                                 t_hlen;
1475
1476                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1477                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1478                                 dev->mtu -= 8;
1479
1480                         if (dev->mtu < IPV6_MIN_MTU)
1481                                 dev->mtu = IPV6_MIN_MTU;
1482                 }
1483                 ip6_rt_put(rt);
1484         }
1485 }
1486
1487 /**
1488  * ip6_tnl_change - update the tunnel parameters
1489  *   @t: tunnel to be changed
1490  *   @p: tunnel configuration parameters
1491  *
1492  * Description:
1493  *   ip6_tnl_change() updates the tunnel parameters
1494  **/
1495
1496 static int
1497 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1498 {
1499         t->parms.laddr = p->laddr;
1500         t->parms.raddr = p->raddr;
1501         t->parms.flags = p->flags;
1502         t->parms.hop_limit = p->hop_limit;
1503         t->parms.encap_limit = p->encap_limit;
1504         t->parms.flowinfo = p->flowinfo;
1505         t->parms.link = p->link;
1506         t->parms.proto = p->proto;
1507         t->parms.fwmark = p->fwmark;
1508         dst_cache_reset(&t->dst_cache);
1509         ip6_tnl_link_config(t);
1510         return 0;
1511 }
1512
1513 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1514 {
1515         struct net *net = t->net;
1516         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1517         int err;
1518
1519         ip6_tnl_unlink(ip6n, t);
1520         synchronize_net();
1521         err = ip6_tnl_change(t, p);
1522         ip6_tnl_link(ip6n, t);
1523         netdev_state_change(t->dev);
1524         return err;
1525 }
1526
1527 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1528 {
1529         /* for default tnl0 device allow to change only the proto */
1530         t->parms.proto = p->proto;
1531         netdev_state_change(t->dev);
1532         return 0;
1533 }
1534
1535 static void
1536 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1537 {
1538         p->laddr = u->laddr;
1539         p->raddr = u->raddr;
1540         p->flags = u->flags;
1541         p->hop_limit = u->hop_limit;
1542         p->encap_limit = u->encap_limit;
1543         p->flowinfo = u->flowinfo;
1544         p->link = u->link;
1545         p->proto = u->proto;
1546         memcpy(p->name, u->name, sizeof(u->name));
1547 }
1548
1549 static void
1550 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1551 {
1552         u->laddr = p->laddr;
1553         u->raddr = p->raddr;
1554         u->flags = p->flags;
1555         u->hop_limit = p->hop_limit;
1556         u->encap_limit = p->encap_limit;
1557         u->flowinfo = p->flowinfo;
1558         u->link = p->link;
1559         u->proto = p->proto;
1560         memcpy(u->name, p->name, sizeof(u->name));
1561 }
1562
1563 /**
1564  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1565  *   @dev: virtual device associated with tunnel
1566  *   @ifr: parameters passed from userspace
1567  *   @cmd: command to be performed
1568  *
1569  * Description:
1570  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1571  *   from userspace.
1572  *
1573  *   The possible commands are the following:
1574  *     %SIOCGETTUNNEL: get tunnel parameters for device
1575  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1576  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1577  *     %SIOCDELTUNNEL: delete tunnel
1578  *
1579  *   The fallback device "ip6tnl0", created during module
1580  *   initialization, can be used for creating other tunnel devices.
1581  *
1582  * Return:
1583  *   0 on success,
1584  *   %-EFAULT if unable to copy data to or from userspace,
1585  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1586  *   %-EINVAL if passed tunnel parameters are invalid,
1587  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1588  *   %-ENODEV if attempting to change or delete a nonexisting device
1589  **/
1590
1591 static int
1592 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1593 {
1594         int err = 0;
1595         struct ip6_tnl_parm p;
1596         struct __ip6_tnl_parm p1;
1597         struct ip6_tnl *t = netdev_priv(dev);
1598         struct net *net = t->net;
1599         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1600
1601         memset(&p1, 0, sizeof(p1));
1602
1603         switch (cmd) {
1604         case SIOCGETTUNNEL:
1605                 if (dev == ip6n->fb_tnl_dev) {
1606                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1607                                 err = -EFAULT;
1608                                 break;
1609                         }
1610                         ip6_tnl_parm_from_user(&p1, &p);
1611                         t = ip6_tnl_locate(net, &p1, 0);
1612                         if (IS_ERR(t))
1613                                 t = netdev_priv(dev);
1614                 } else {
1615                         memset(&p, 0, sizeof(p));
1616                 }
1617                 ip6_tnl_parm_to_user(&p, &t->parms);
1618                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1619                         err = -EFAULT;
1620                 }
1621                 break;
1622         case SIOCADDTUNNEL:
1623         case SIOCCHGTUNNEL:
1624                 err = -EPERM;
1625                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1626                         break;
1627                 err = -EFAULT;
1628                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1629                         break;
1630                 err = -EINVAL;
1631                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1632                     p.proto != 0)
1633                         break;
1634                 ip6_tnl_parm_from_user(&p1, &p);
1635                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1636                 if (cmd == SIOCCHGTUNNEL) {
1637                         if (!IS_ERR(t)) {
1638                                 if (t->dev != dev) {
1639                                         err = -EEXIST;
1640                                         break;
1641                                 }
1642                         } else
1643                                 t = netdev_priv(dev);
1644                         if (dev == ip6n->fb_tnl_dev)
1645                                 err = ip6_tnl0_update(t, &p1);
1646                         else
1647                                 err = ip6_tnl_update(t, &p1);
1648                 }
1649                 if (!IS_ERR(t)) {
1650                         err = 0;
1651                         ip6_tnl_parm_to_user(&p, &t->parms);
1652                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1653                                 err = -EFAULT;
1654
1655                 } else {
1656                         err = PTR_ERR(t);
1657                 }
1658                 break;
1659         case SIOCDELTUNNEL:
1660                 err = -EPERM;
1661                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1662                         break;
1663
1664                 if (dev == ip6n->fb_tnl_dev) {
1665                         err = -EFAULT;
1666                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1667                                 break;
1668                         err = -ENOENT;
1669                         ip6_tnl_parm_from_user(&p1, &p);
1670                         t = ip6_tnl_locate(net, &p1, 0);
1671                         if (IS_ERR(t))
1672                                 break;
1673                         err = -EPERM;
1674                         if (t->dev == ip6n->fb_tnl_dev)
1675                                 break;
1676                         dev = t->dev;
1677                 }
1678                 err = 0;
1679                 unregister_netdevice(dev);
1680                 break;
1681         default:
1682                 err = -EINVAL;
1683         }
1684         return err;
1685 }
1686
1687 /**
1688  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1689  *   @dev: virtual device associated with tunnel
1690  *   @new_mtu: the new mtu
1691  *
1692  * Return:
1693  *   0 on success,
1694  *   %-EINVAL if mtu too small
1695  **/
1696
1697 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1698 {
1699         struct ip6_tnl *tnl = netdev_priv(dev);
1700
1701         if (tnl->parms.proto == IPPROTO_IPV6) {
1702                 if (new_mtu < IPV6_MIN_MTU)
1703                         return -EINVAL;
1704         } else {
1705                 if (new_mtu < ETH_MIN_MTU)
1706                         return -EINVAL;
1707         }
1708         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1709                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1710                         return -EINVAL;
1711         } else {
1712                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1713                         return -EINVAL;
1714         }
1715         dev->mtu = new_mtu;
1716         return 0;
1717 }
1718 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1719
1720 int ip6_tnl_get_iflink(const struct net_device *dev)
1721 {
1722         struct ip6_tnl *t = netdev_priv(dev);
1723
1724         return t->parms.link;
1725 }
1726 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1727
1728 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1729                           unsigned int num)
1730 {
1731         if (num >= MAX_IPTUN_ENCAP_OPS)
1732                 return -ERANGE;
1733
1734         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1735                         &ip6tun_encaps[num],
1736                         NULL, ops) ? 0 : -1;
1737 }
1738 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1739
1740 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1741                           unsigned int num)
1742 {
1743         int ret;
1744
1745         if (num >= MAX_IPTUN_ENCAP_OPS)
1746                 return -ERANGE;
1747
1748         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1749                        &ip6tun_encaps[num],
1750                        ops, NULL) == ops) ? 0 : -1;
1751
1752         synchronize_net();
1753
1754         return ret;
1755 }
1756 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1757
1758 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1759                         struct ip_tunnel_encap *ipencap)
1760 {
1761         int hlen;
1762
1763         memset(&t->encap, 0, sizeof(t->encap));
1764
1765         hlen = ip6_encap_hlen(ipencap);
1766         if (hlen < 0)
1767                 return hlen;
1768
1769         t->encap.type = ipencap->type;
1770         t->encap.sport = ipencap->sport;
1771         t->encap.dport = ipencap->dport;
1772         t->encap.flags = ipencap->flags;
1773
1774         t->encap_hlen = hlen;
1775         t->hlen = t->encap_hlen + t->tun_hlen;
1776
1777         return 0;
1778 }
1779 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1780
1781 static const struct net_device_ops ip6_tnl_netdev_ops = {
1782         .ndo_init       = ip6_tnl_dev_init,
1783         .ndo_uninit     = ip6_tnl_dev_uninit,
1784         .ndo_start_xmit = ip6_tnl_start_xmit,
1785         .ndo_do_ioctl   = ip6_tnl_ioctl,
1786         .ndo_change_mtu = ip6_tnl_change_mtu,
1787         .ndo_get_stats  = ip6_get_stats,
1788         .ndo_get_iflink = ip6_tnl_get_iflink,
1789 };
1790
1791 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1792                          NETIF_F_FRAGLIST |     \
1793                          NETIF_F_HIGHDMA |      \
1794                          NETIF_F_GSO_SOFTWARE | \
1795                          NETIF_F_HW_CSUM)
1796
1797 /**
1798  * ip6_tnl_dev_setup - setup virtual tunnel device
1799  *   @dev: virtual device associated with tunnel
1800  *
1801  * Description:
1802  *   Initialize function pointers and device parameters
1803  **/
1804
1805 static void ip6_tnl_dev_setup(struct net_device *dev)
1806 {
1807         dev->netdev_ops = &ip6_tnl_netdev_ops;
1808         dev->needs_free_netdev = true;
1809         dev->priv_destructor = ip6_dev_free;
1810
1811         dev->type = ARPHRD_TUNNEL6;
1812         dev->flags |= IFF_NOARP;
1813         dev->addr_len = sizeof(struct in6_addr);
1814         dev->features |= NETIF_F_LLTX;
1815         netif_keep_dst(dev);
1816
1817         dev->features           |= IPXIPX_FEATURES;
1818         dev->hw_features        |= IPXIPX_FEATURES;
1819
1820         /* This perm addr will be used as interface identifier by IPv6 */
1821         dev->addr_assign_type = NET_ADDR_RANDOM;
1822         eth_random_addr(dev->perm_addr);
1823 }
1824
1825
1826 /**
1827  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1828  *   @dev: virtual device associated with tunnel
1829  **/
1830
1831 static inline int
1832 ip6_tnl_dev_init_gen(struct net_device *dev)
1833 {
1834         struct ip6_tnl *t = netdev_priv(dev);
1835         int ret;
1836         int t_hlen;
1837
1838         t->dev = dev;
1839         t->net = dev_net(dev);
1840         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1841         if (!dev->tstats)
1842                 return -ENOMEM;
1843
1844         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1845         if (ret)
1846                 goto free_stats;
1847
1848         ret = gro_cells_init(&t->gro_cells, dev);
1849         if (ret)
1850                 goto destroy_dst;
1851
1852         t->tun_hlen = 0;
1853         t->hlen = t->encap_hlen + t->tun_hlen;
1854         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1855
1856         dev->type = ARPHRD_TUNNEL6;
1857         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1858         dev->mtu = ETH_DATA_LEN - t_hlen;
1859         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1860                 dev->mtu -= 8;
1861         dev->min_mtu = ETH_MIN_MTU;
1862         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1863
1864         dev_hold(dev);
1865         return 0;
1866
1867 destroy_dst:
1868         dst_cache_destroy(&t->dst_cache);
1869 free_stats:
1870         free_percpu(dev->tstats);
1871         dev->tstats = NULL;
1872
1873         return ret;
1874 }
1875
1876 /**
1877  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1878  *   @dev: virtual device associated with tunnel
1879  **/
1880
1881 static int ip6_tnl_dev_init(struct net_device *dev)
1882 {
1883         struct ip6_tnl *t = netdev_priv(dev);
1884         int err = ip6_tnl_dev_init_gen(dev);
1885
1886         if (err)
1887                 return err;
1888         ip6_tnl_link_config(t);
1889         if (t->parms.collect_md)
1890                 netif_keep_dst(dev);
1891         return 0;
1892 }
1893
1894 /**
1895  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1896  *   @dev: fallback device
1897  *
1898  * Return: 0
1899  **/
1900
1901 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1902 {
1903         struct ip6_tnl *t = netdev_priv(dev);
1904         struct net *net = dev_net(dev);
1905         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1906
1907         t->parms.proto = IPPROTO_IPV6;
1908
1909         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1910         return 0;
1911 }
1912
1913 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1914                             struct netlink_ext_ack *extack)
1915 {
1916         u8 proto;
1917
1918         if (!data || !data[IFLA_IPTUN_PROTO])
1919                 return 0;
1920
1921         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1922         if (proto != IPPROTO_IPV6 &&
1923             proto != IPPROTO_IPIP &&
1924             proto != 0)
1925                 return -EINVAL;
1926
1927         return 0;
1928 }
1929
1930 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1931                                   struct __ip6_tnl_parm *parms)
1932 {
1933         memset(parms, 0, sizeof(*parms));
1934
1935         if (!data)
1936                 return;
1937
1938         if (data[IFLA_IPTUN_LINK])
1939                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1940
1941         if (data[IFLA_IPTUN_LOCAL])
1942                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1943
1944         if (data[IFLA_IPTUN_REMOTE])
1945                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1946
1947         if (data[IFLA_IPTUN_TTL])
1948                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1949
1950         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1951                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1952
1953         if (data[IFLA_IPTUN_FLOWINFO])
1954                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1955
1956         if (data[IFLA_IPTUN_FLAGS])
1957                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1958
1959         if (data[IFLA_IPTUN_PROTO])
1960                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1961
1962         if (data[IFLA_IPTUN_COLLECT_METADATA])
1963                 parms->collect_md = true;
1964
1965         if (data[IFLA_IPTUN_FWMARK])
1966                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1967 }
1968
1969 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1970                                         struct ip_tunnel_encap *ipencap)
1971 {
1972         bool ret = false;
1973
1974         memset(ipencap, 0, sizeof(*ipencap));
1975
1976         if (!data)
1977                 return ret;
1978
1979         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1980                 ret = true;
1981                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1982         }
1983
1984         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1985                 ret = true;
1986                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1987         }
1988
1989         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1990                 ret = true;
1991                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1992         }
1993
1994         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1995                 ret = true;
1996                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1997         }
1998
1999         return ret;
2000 }
2001
2002 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2003                            struct nlattr *tb[], struct nlattr *data[],
2004                            struct netlink_ext_ack *extack)
2005 {
2006         struct net *net = dev_net(dev);
2007         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2008         struct ip_tunnel_encap ipencap;
2009         struct ip6_tnl *nt, *t;
2010         int err;
2011
2012         nt = netdev_priv(dev);
2013
2014         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2015                 err = ip6_tnl_encap_setup(nt, &ipencap);
2016                 if (err < 0)
2017                         return err;
2018         }
2019
2020         ip6_tnl_netlink_parms(data, &nt->parms);
2021
2022         if (nt->parms.collect_md) {
2023                 if (rtnl_dereference(ip6n->collect_md_tun))
2024                         return -EEXIST;
2025         } else {
2026                 t = ip6_tnl_locate(net, &nt->parms, 0);
2027                 if (!IS_ERR(t))
2028                         return -EEXIST;
2029         }
2030
2031         err = ip6_tnl_create2(dev);
2032         if (!err && tb[IFLA_MTU])
2033                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2034
2035         return err;
2036 }
2037
2038 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2039                               struct nlattr *data[],
2040                               struct netlink_ext_ack *extack)
2041 {
2042         struct ip6_tnl *t = netdev_priv(dev);
2043         struct __ip6_tnl_parm p;
2044         struct net *net = t->net;
2045         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2046         struct ip_tunnel_encap ipencap;
2047
2048         if (dev == ip6n->fb_tnl_dev)
2049                 return -EINVAL;
2050
2051         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2052                 int err = ip6_tnl_encap_setup(t, &ipencap);
2053
2054                 if (err < 0)
2055                         return err;
2056         }
2057         ip6_tnl_netlink_parms(data, &p);
2058         if (p.collect_md)
2059                 return -EINVAL;
2060
2061         t = ip6_tnl_locate(net, &p, 0);
2062         if (!IS_ERR(t)) {
2063                 if (t->dev != dev)
2064                         return -EEXIST;
2065         } else
2066                 t = netdev_priv(dev);
2067
2068         return ip6_tnl_update(t, &p);
2069 }
2070
2071 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2072 {
2073         struct net *net = dev_net(dev);
2074         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2075
2076         if (dev != ip6n->fb_tnl_dev)
2077                 unregister_netdevice_queue(dev, head);
2078 }
2079
2080 static size_t ip6_tnl_get_size(const struct net_device *dev)
2081 {
2082         return
2083                 /* IFLA_IPTUN_LINK */
2084                 nla_total_size(4) +
2085                 /* IFLA_IPTUN_LOCAL */
2086                 nla_total_size(sizeof(struct in6_addr)) +
2087                 /* IFLA_IPTUN_REMOTE */
2088                 nla_total_size(sizeof(struct in6_addr)) +
2089                 /* IFLA_IPTUN_TTL */
2090                 nla_total_size(1) +
2091                 /* IFLA_IPTUN_ENCAP_LIMIT */
2092                 nla_total_size(1) +
2093                 /* IFLA_IPTUN_FLOWINFO */
2094                 nla_total_size(4) +
2095                 /* IFLA_IPTUN_FLAGS */
2096                 nla_total_size(4) +
2097                 /* IFLA_IPTUN_PROTO */
2098                 nla_total_size(1) +
2099                 /* IFLA_IPTUN_ENCAP_TYPE */
2100                 nla_total_size(2) +
2101                 /* IFLA_IPTUN_ENCAP_FLAGS */
2102                 nla_total_size(2) +
2103                 /* IFLA_IPTUN_ENCAP_SPORT */
2104                 nla_total_size(2) +
2105                 /* IFLA_IPTUN_ENCAP_DPORT */
2106                 nla_total_size(2) +
2107                 /* IFLA_IPTUN_COLLECT_METADATA */
2108                 nla_total_size(0) +
2109                 /* IFLA_IPTUN_FWMARK */
2110                 nla_total_size(4) +
2111                 0;
2112 }
2113
2114 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2115 {
2116         struct ip6_tnl *tunnel = netdev_priv(dev);
2117         struct __ip6_tnl_parm *parm = &tunnel->parms;
2118
2119         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2120             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2121             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2122             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2123             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2124             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2125             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2126             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2127             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2128                 goto nla_put_failure;
2129
2130         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2131             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2132             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2133             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2134                 goto nla_put_failure;
2135
2136         if (parm->collect_md)
2137                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2138                         goto nla_put_failure;
2139
2140         return 0;
2141
2142 nla_put_failure:
2143         return -EMSGSIZE;
2144 }
2145
2146 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2147 {
2148         struct ip6_tnl *tunnel = netdev_priv(dev);
2149
2150         return tunnel->net;
2151 }
2152 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2153
2154 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2155         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2156         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2157         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2158         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2159         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2160         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2161         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2162         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2163         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2164         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2165         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2166         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2167         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2168         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2169 };
2170
2171 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2172         .kind           = "ip6tnl",
2173         .maxtype        = IFLA_IPTUN_MAX,
2174         .policy         = ip6_tnl_policy,
2175         .priv_size      = sizeof(struct ip6_tnl),
2176         .setup          = ip6_tnl_dev_setup,
2177         .validate       = ip6_tnl_validate,
2178         .newlink        = ip6_tnl_newlink,
2179         .changelink     = ip6_tnl_changelink,
2180         .dellink        = ip6_tnl_dellink,
2181         .get_size       = ip6_tnl_get_size,
2182         .fill_info      = ip6_tnl_fill_info,
2183         .get_link_net   = ip6_tnl_get_link_net,
2184 };
2185
2186 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2187         .handler        = ip4ip6_rcv,
2188         .err_handler    = ip4ip6_err,
2189         .priority       =       1,
2190 };
2191
2192 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2193         .handler        = ip6ip6_rcv,
2194         .err_handler    = ip6ip6_err,
2195         .priority       =       1,
2196 };
2197
2198 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2199 {
2200         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2201         struct net_device *dev, *aux;
2202         int h;
2203         struct ip6_tnl *t;
2204         LIST_HEAD(list);
2205
2206         for_each_netdev_safe(net, dev, aux)
2207                 if (dev->rtnl_link_ops == &ip6_link_ops)
2208                         unregister_netdevice_queue(dev, &list);
2209
2210         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2211                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2212                 while (t) {
2213                         /* If dev is in the same netns, it has already
2214                          * been added to the list by the previous loop.
2215                          */
2216                         if (!net_eq(dev_net(t->dev), net))
2217                                 unregister_netdevice_queue(t->dev, &list);
2218                         t = rtnl_dereference(t->next);
2219                 }
2220         }
2221
2222         unregister_netdevice_many(&list);
2223 }
2224
2225 static int __net_init ip6_tnl_init_net(struct net *net)
2226 {
2227         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2228         struct ip6_tnl *t = NULL;
2229         int err;
2230
2231         ip6n->tnls[0] = ip6n->tnls_wc;
2232         ip6n->tnls[1] = ip6n->tnls_r_l;
2233
2234         err = -ENOMEM;
2235         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2236                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2237
2238         if (!ip6n->fb_tnl_dev)
2239                 goto err_alloc_dev;
2240         dev_net_set(ip6n->fb_tnl_dev, net);
2241         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2242         /* FB netdevice is special: we have one, and only one per netns.
2243          * Allowing to move it to another netns is clearly unsafe.
2244          */
2245         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2246
2247         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2248         if (err < 0)
2249                 goto err_register;
2250
2251         err = register_netdev(ip6n->fb_tnl_dev);
2252         if (err < 0)
2253                 goto err_register;
2254
2255         t = netdev_priv(ip6n->fb_tnl_dev);
2256
2257         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2258         return 0;
2259
2260 err_register:
2261         free_netdev(ip6n->fb_tnl_dev);
2262 err_alloc_dev:
2263         return err;
2264 }
2265
2266 static void __net_exit ip6_tnl_exit_net(struct net *net)
2267 {
2268         rtnl_lock();
2269         ip6_tnl_destroy_tunnels(net);
2270         rtnl_unlock();
2271 }
2272
2273 static struct pernet_operations ip6_tnl_net_ops = {
2274         .init = ip6_tnl_init_net,
2275         .exit = ip6_tnl_exit_net,
2276         .id   = &ip6_tnl_net_id,
2277         .size = sizeof(struct ip6_tnl_net),
2278 };
2279
2280 /**
2281  * ip6_tunnel_init - register protocol and reserve needed resources
2282  *
2283  * Return: 0 on success
2284  **/
2285
2286 static int __init ip6_tunnel_init(void)
2287 {
2288         int  err;
2289
2290         if (!ipv6_mod_enabled())
2291                 return -EOPNOTSUPP;
2292
2293         err = register_pernet_device(&ip6_tnl_net_ops);
2294         if (err < 0)
2295                 goto out_pernet;
2296
2297         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2298         if (err < 0) {
2299                 pr_err("%s: can't register ip4ip6\n", __func__);
2300                 goto out_ip4ip6;
2301         }
2302
2303         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2304         if (err < 0) {
2305                 pr_err("%s: can't register ip6ip6\n", __func__);
2306                 goto out_ip6ip6;
2307         }
2308         err = rtnl_link_register(&ip6_link_ops);
2309         if (err < 0)
2310                 goto rtnl_link_failed;
2311
2312         return 0;
2313
2314 rtnl_link_failed:
2315         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2316 out_ip6ip6:
2317         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2318 out_ip4ip6:
2319         unregister_pernet_device(&ip6_tnl_net_ops);
2320 out_pernet:
2321         return err;
2322 }
2323
2324 /**
2325  * ip6_tunnel_cleanup - free resources and unregister protocol
2326  **/
2327
2328 static void __exit ip6_tunnel_cleanup(void)
2329 {
2330         rtnl_link_unregister(&ip6_link_ops);
2331         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2332                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2333
2334         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2335                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2336
2337         unregister_pernet_device(&ip6_tnl_net_ops);
2338 }
2339
2340 module_init(ip6_tunnel_init);
2341 module_exit(ip6_tunnel_cleanup);