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 <asm/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
  61 MODULE_AUTHOR("Ville Nuorvala");
  62 MODULE_DESCRIPTION("IPv6 tunneling device");
  63 MODULE_LICENSE("GPL");
  64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  65 MODULE_ALIAS_NETDEV("ip6tnl0");
  66
  67 #define HASH_SIZE_SHIFT  5
  68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
  69
  70 static bool log_ecn_error = true;
  71 module_param(log_ecn_error, bool, 0644);
  72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  73
  74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  75 {
  76         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  77
  78         return hash_32(hash, HASH_SIZE_SHIFT);
  79 }
  80
  81 static int ip6_tnl_dev_init(struct net_device *dev);
  82 static void ip6_tnl_dev_setup(struct net_device *dev);
  83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  84
  85 static int ip6_tnl_net_id __read_mostly;
  86 struct ip6_tnl_net {
  87         /* the IPv6 tunnel fallback device */
  88         struct net_device *fb_tnl_dev;
  89         /* lists for storing tunnels in use */
  90         struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
  91         struct ip6_tnl __rcu *tnls_wc[1];
  92         struct ip6_tnl __rcu **tnls[2];
  93 };
  94
  95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  96 {
  97         struct pcpu_sw_netstats tmp, sum = { 0 };
  98         int i;
  99
 100         for_each_possible_cpu(i) {
 101                 unsigned int start;
 102                 const struct pcpu_sw_netstats *tstats =
 103                                                    per_cpu_ptr(dev->tstats, i);
 104
 105                 do {
 106                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 107                         tmp.rx_packets = tstats->rx_packets;
 108                         tmp.rx_bytes = tstats->rx_bytes;
 109                         tmp.tx_packets = tstats->tx_packets;
 110                         tmp.tx_bytes =  tstats->tx_bytes;
 111                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 112
 113                 sum.rx_packets += tmp.rx_packets;
 114                 sum.rx_bytes   += tmp.rx_bytes;
 115                 sum.tx_packets += tmp.tx_packets;
 116                 sum.tx_bytes   += tmp.tx_bytes;
 117         }
 118         dev->stats.rx_packets = sum.rx_packets;
 119         dev->stats.rx_bytes   = sum.rx_bytes;
 120         dev->stats.tx_packets = sum.tx_packets;
 121         dev->stats.tx_bytes   = sum.tx_bytes;
 122         return &dev->stats;
 123 }
 124
 125 /**
 126  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 127  *   @remote: the address of the tunnel exit-point
 128  *   @local: the address of the tunnel entry-point
 129  *
 130  * Return:
 131  *   tunnel matching given end-points if found,
 132  *   else fallback tunnel if its device is up,
 133  *   else %NULL
 134  **/
 135
 136 #define for_each_ip6_tunnel_rcu(start) \
 137         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 138
 139 static struct ip6_tnl *
 140 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 141 {
 142         unsigned int hash = HASH(remote, local);
 143         struct ip6_tnl *t;
 144         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 145         struct in6_addr any;
 146
 147         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 148                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 149                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 150                     (t->dev->flags & IFF_UP))
 151                         return t;
 152         }
 153
 154         memset(&any, 0, sizeof(any));
 155         hash = HASH(&any, local);
 156         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 157                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 158                     (t->dev->flags & IFF_UP))
 159                         return t;
 160         }
 161
 162         hash = HASH(remote, &any);
 163         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 164                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 165                     (t->dev->flags & IFF_UP))
 166                         return t;
 167         }
 168
 169         t = rcu_dereference(ip6n->tnls_wc[0]);
 170         if (t && (t->dev->flags & IFF_UP))
 171                 return t;
 172
 173         return NULL;
 174 }
 175
 176 /**
 177  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 178  *   @p: parameters containing tunnel end-points
 179  *
 180  * Description:
 181  *   ip6_tnl_bucket() returns the head of the list matching the
 182  *   &struct in6_addr entries laddr and raddr in @p.
 183  *
 184  * Return: head of IPv6 tunnel list
 185  **/
 186
 187 static struct ip6_tnl __rcu **
 188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 189 {
 190         const struct in6_addr *remote = &p->raddr;
 191         const struct in6_addr *local = &p->laddr;
 192         unsigned int h = 0;
 193         int prio = 0;
 194
 195         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 196                 prio = 1;
 197                 h = HASH(remote, local);
 198         }
 199         return &ip6n->tnls[prio][h];
 200 }
 201
 202 /**
 203  * ip6_tnl_link - add tunnel to hash table
 204  *   @t: tunnel to be added
 205  **/
 206
 207 static void
 208 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 209 {
 210         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 211
 212         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 213         rcu_assign_pointer(*tp, t);
 214 }
 215
 216 /**
 217  * ip6_tnl_unlink - remove tunnel from hash table
 218  *   @t: tunnel to be removed
 219  **/
 220
 221 static void
 222 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 223 {
 224         struct ip6_tnl __rcu **tp;
 225         struct ip6_tnl *iter;
 226
 227         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 228              (iter = rtnl_dereference(*tp)) != NULL;
 229              tp = &iter->next) {
 230                 if (t == iter) {
 231                         rcu_assign_pointer(*tp, t->next);
 232                         break;
 233                 }
 234         }
 235 }
 236
 237 static void ip6_dev_free(struct net_device *dev)
 238 {
 239         struct ip6_tnl *t = netdev_priv(dev);
 240
 241         dst_cache_destroy(&t->dst_cache);
 242         free_percpu(dev->tstats);
 243         free_netdev(dev);
 244 }
 245
 246 static int ip6_tnl_create2(struct net_device *dev)
 247 {
 248         struct ip6_tnl *t = netdev_priv(dev);
 249         struct net *net = dev_net(dev);
 250         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 251         int err;
 252
 253         t = netdev_priv(dev);
 254
 255         err = register_netdevice(dev);
 256         if (err < 0)
 257                 goto out;
 258
 259         strcpy(t->parms.name, dev->name);
 260         dev->rtnl_link_ops = &ip6_link_ops;
 261
 262         dev_hold(dev);
 263         ip6_tnl_link(ip6n, t);
 264         return 0;
 265
 266 out:
 267         return err;
 268 }
 269
 270 /**
 271  * ip6_tnl_create - create a new tunnel
 272  *   @p: tunnel parameters
 273  *   @pt: pointer to new tunnel
 274  *
 275  * Description:
 276  *   Create tunnel matching given parameters.
 277  *
 278  * Return:
 279  *   created tunnel or error pointer
 280  **/
 281
 282 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 283 {
 284         struct net_device *dev;
 285         struct ip6_tnl *t;
 286         char name[IFNAMSIZ];
 287         int err = -ENOMEM;
 288
 289         if (p->name[0])
 290                 strlcpy(name, p->name, IFNAMSIZ);
 291         else
 292                 sprintf(name, "ip6tnl%%d");
 293
 294         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 295                            ip6_tnl_dev_setup);
 296         if (!dev)
 297                 goto failed;
 298
 299         dev_net_set(dev, net);
 300
 301         t = netdev_priv(dev);
 302         t->parms = *p;
 303         t->net = dev_net(dev);
 304         err = ip6_tnl_create2(dev);
 305         if (err < 0)
 306                 goto failed_free;
 307
 308         return t;
 309
 310 failed_free:
 311         ip6_dev_free(dev);
 312 failed:
 313         return ERR_PTR(err);
 314 }
 315
 316 /**
 317  * ip6_tnl_locate - find or create tunnel matching given parameters
 318  *   @p: tunnel parameters
 319  *   @create: != 0 if allowed to create new tunnel if no match found
 320  *
 321  * Description:
 322  *   ip6_tnl_locate() first tries to locate an existing tunnel
 323  *   based on @parms. If this is unsuccessful, but @create is set a new
 324  *   tunnel device is created and registered for use.
 325  *
 326  * Return:
 327  *   matching tunnel or error pointer
 328  **/
 329
 330 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 331                 struct __ip6_tnl_parm *p, int create)
 332 {
 333         const struct in6_addr *remote = &p->raddr;
 334         const struct in6_addr *local = &p->laddr;
 335         struct ip6_tnl __rcu **tp;
 336         struct ip6_tnl *t;
 337         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 338
 339         for (tp = ip6_tnl_bucket(ip6n, p);
 340              (t = rtnl_dereference(*tp)) != NULL;
 341              tp = &t->next) {
 342                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 343                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 344                         if (create)
 345                                 return ERR_PTR(-EEXIST);
 346
 347                         return t;
 348                 }
 349         }
 350         if (!create)
 351                 return ERR_PTR(-ENODEV);
 352         return ip6_tnl_create(net, p);
 353 }
 354
 355 /**
 356  * ip6_tnl_dev_uninit - tunnel device uninitializer
 357  *   @dev: the device to be destroyed
 358  *
 359  * Description:
 360  *   ip6_tnl_dev_uninit() removes tunnel from its list
 361  **/
 362
 363 static void
 364 ip6_tnl_dev_uninit(struct net_device *dev)
 365 {
 366         struct ip6_tnl *t = netdev_priv(dev);
 367         struct net *net = t->net;
 368         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 369
 370         if (dev == ip6n->fb_tnl_dev)
 371                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 372         else
 373                 ip6_tnl_unlink(ip6n, t);
 374         dst_cache_reset(&t->dst_cache);
 375         dev_put(dev);
 376 }
 377
 378 /**
 379  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 380  *   @skb: received socket buffer
 381  *
 382  * Return:
 383  *   0 if none was found,
 384  *   else index to encapsulation limit
 385  **/
 386
 387 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 388 {
 389         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 390         __u8 nexthdr = ipv6h->nexthdr;
 391         __u16 off = sizeof(*ipv6h);
 392
 393         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 394                 __u16 optlen = 0;
 395                 struct ipv6_opt_hdr *hdr;
 396                 if (raw + off + sizeof(*hdr) > skb->data &&
 397                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 398                         break;
 399
 400                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 401                 if (nexthdr == NEXTHDR_FRAGMENT) {
 402                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 403                         if (frag_hdr->frag_off)
 404                                 break;
 405                         optlen = 8;
 406                 } else if (nexthdr == NEXTHDR_AUTH) {
 407                         optlen = (hdr->hdrlen + 2) << 2;
 408                 } else {
 409                         optlen = ipv6_optlen(hdr);
 410                 }
 411                 if (nexthdr == NEXTHDR_DEST) {
 412                         __u16 i = off + 2;
 413                         while (1) {
 414                                 struct ipv6_tlv_tnl_enc_lim *tel;
 415
 416                                 /* No more room for encapsulation limit */
 417                                 if (i + sizeof (*tel) > off + optlen)
 418                                         break;
 419
 420                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 421                                 /* return index of option if found and valid */
 422                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 423                                     tel->length == 1)
 424                                         return i;
 425                                 /* else jump to next option */
 426                                 if (tel->type)
 427                                         i += tel->length + 2;
 428                                 else
 429                                         i++;
 430                         }
 431                 }
 432                 nexthdr = hdr->nexthdr;
 433                 off += optlen;
 434         }
 435         return 0;
 436 }
 437 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 438
 439 /**
 440  * ip6_tnl_err - tunnel error handler
 441  *
 442  * Description:
 443  *   ip6_tnl_err() should handle errors in the tunnel according
 444  *   to the specifications in RFC 2473.
 445  **/
 446
 447 static int
 448 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 449             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 450 {
 451         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 452         struct ip6_tnl *t;
 453         int rel_msg = 0;
 454         u8 rel_type = ICMPV6_DEST_UNREACH;
 455         u8 rel_code = ICMPV6_ADDR_UNREACH;
 456         u8 tproto;
 457         __u32 rel_info = 0;
 458         __u16 len;
 459         int err = -ENOENT;
 460
 461         /* If the packet doesn't contain the original IPv6 header we are
 462            in trouble since we might need the source address for further
 463            processing of the error. */
 464
 465         rcu_read_lock();
 466         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 467         if (!t)
 468                 goto out;
 469
 470         tproto = ACCESS_ONCE(t->parms.proto);
 471         if (tproto != ipproto && tproto != 0)
 472                 goto out;
 473
 474         err = 0;
 475
 476         switch (*type) {
 477                 __u32 teli;
 478                 struct ipv6_tlv_tnl_enc_lim *tel;
 479                 __u32 mtu;
 480         case ICMPV6_DEST_UNREACH:
 481                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 482                                     t->parms.name);
 483                 rel_msg = 1;
 484                 break;
 485         case ICMPV6_TIME_EXCEED:
 486                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 487                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 488                                             t->parms.name);
 489                         rel_msg = 1;
 490                 }
 491                 break;
 492         case ICMPV6_PARAMPROB:
 493                 teli = 0;
 494                 if ((*code) == ICMPV6_HDR_FIELD)
 495                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 496
 497                 if (teli && teli == *info - 2) {
 498                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 499                         if (tel->encap_limit == 0) {
 500                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 501                                                     t->parms.name);
 502                                 rel_msg = 1;
 503                         }
 504                 } else {
 505                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 506                                             t->parms.name);
 507                 }
 508                 break;
 509         case ICMPV6_PKT_TOOBIG:
 510                 mtu = *info - offset;
 511                 if (mtu < IPV6_MIN_MTU)
 512                         mtu = IPV6_MIN_MTU;
 513                 t->dev->mtu = mtu;
 514
 515                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 516                 if (len > mtu) {
 517                         rel_type = ICMPV6_PKT_TOOBIG;
 518                         rel_code = 0;
 519                         rel_info = mtu;
 520                         rel_msg = 1;
 521                 }
 522                 break;
 523         }
 524
 525         *type = rel_type;
 526         *code = rel_code;
 527         *info = rel_info;
 528         *msg = rel_msg;
 529
 530 out:
 531         rcu_read_unlock();
 532         return err;
 533 }
 534
 535 static int
 536 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 537            u8 type, u8 code, int offset, __be32 info)
 538 {
 539         int rel_msg = 0;
 540         u8 rel_type = type;
 541         u8 rel_code = code;
 542         __u32 rel_info = ntohl(info);
 543         int err;
 544         struct sk_buff *skb2;
 545         const struct iphdr *eiph;
 546         struct rtable *rt;
 547         struct flowi4 fl4;
 548
 549         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 550                           &rel_msg, &rel_info, offset);
 551         if (err < 0)
 552                 return err;
 553
 554         if (rel_msg == 0)
 555                 return 0;
 556
 557         switch (rel_type) {
 558         case ICMPV6_DEST_UNREACH:
 559                 if (rel_code != ICMPV6_ADDR_UNREACH)
 560                         return 0;
 561                 rel_type = ICMP_DEST_UNREACH;
 562                 rel_code = ICMP_HOST_UNREACH;
 563                 break;
 564         case ICMPV6_PKT_TOOBIG:
 565                 if (rel_code != 0)
 566                         return 0;
 567                 rel_type = ICMP_DEST_UNREACH;
 568                 rel_code = ICMP_FRAG_NEEDED;
 569                 break;
 570         case NDISC_REDIRECT:
 571                 rel_type = ICMP_REDIRECT;
 572                 rel_code = ICMP_REDIR_HOST;
 573         default:
 574                 return 0;
 575         }
 576
 577         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 578                 return 0;
 579
 580         skb2 = skb_clone(skb, GFP_ATOMIC);
 581         if (!skb2)
 582                 return 0;
 583
 584         skb_dst_drop(skb2);
 585
 586         skb_pull(skb2, offset);
 587         skb_reset_network_header(skb2);
 588         eiph = ip_hdr(skb2);
 589
 590         /* Try to guess incoming interface */
 591         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 592                                    eiph->saddr, 0,
 593                                    0, 0,
 594                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 595         if (IS_ERR(rt))
 596                 goto out;
 597
 598         skb2->dev = rt->dst.dev;
 599
 600         /* route "incoming" packet */
 601         if (rt->rt_flags & RTCF_LOCAL) {
 602                 ip_rt_put(rt);
 603                 rt = NULL;
 604                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 605                                            eiph->daddr, eiph->saddr,
 606                                            0, 0,
 607                                            IPPROTO_IPIP,
 608                                            RT_TOS(eiph->tos), 0);
 609                 if (IS_ERR(rt) ||
 610                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 611                         if (!IS_ERR(rt))
 612                                 ip_rt_put(rt);
 613                         goto out;
 614                 }
 615                 skb_dst_set(skb2, &rt->dst);
 616         } else {
 617                 ip_rt_put(rt);
 618                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 619                                    skb2->dev) ||
 620                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 621                         goto out;
 622         }
 623
 624         /* change mtu on this route */
 625         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 626                 if (rel_info > dst_mtu(skb_dst(skb2)))
 627                         goto out;
 628
 629                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 630         }
 631         if (rel_type == ICMP_REDIRECT)
 632                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 633
 634         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 635
 636 out:
 637         kfree_skb(skb2);
 638         return 0;
 639 }
 640
 641 static int
 642 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 643            u8 type, u8 code, int offset, __be32 info)
 644 {
 645         int rel_msg = 0;
 646         u8 rel_type = type;
 647         u8 rel_code = code;
 648         __u32 rel_info = ntohl(info);
 649         int err;
 650
 651         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 652                           &rel_msg, &rel_info, offset);
 653         if (err < 0)
 654                 return err;
 655
 656         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 657                 struct rt6_info *rt;
 658                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 659
 660                 if (!skb2)
 661                         return 0;
 662
 663                 skb_dst_drop(skb2);
 664                 skb_pull(skb2, offset);
 665                 skb_reset_network_header(skb2);
 666
 667                 /* Try to guess incoming interface */
 668                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 669                                 NULL, 0, 0);
 670
 671                 if (rt && rt->dst.dev)
 672                         skb2->dev = rt->dst.dev;
 673
 674                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 675
 676                 ip6_rt_put(rt);
 677
 678                 kfree_skb(skb2);
 679         }
 680
 681         return 0;
 682 }
 683
 684 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 685                                        const struct ipv6hdr *ipv6h,
 686                                        struct sk_buff *skb)
 687 {
 688         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 689
 690         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 691                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 692
 693         return IP6_ECN_decapsulate(ipv6h, skb);
 694 }
 695
 696 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 697                                        const struct ipv6hdr *ipv6h,
 698                                        struct sk_buff *skb)
 699 {
 700         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 701                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 702
 703         return IP6_ECN_decapsulate(ipv6h, skb);
 704 }
 705
 706 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 707                              const struct in6_addr *laddr,
 708                              const struct in6_addr *raddr)
 709 {
 710         struct __ip6_tnl_parm *p = &t->parms;
 711         int ltype = ipv6_addr_type(laddr);
 712         int rtype = ipv6_addr_type(raddr);
 713         __u32 flags = 0;
 714
 715         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 716                 flags = IP6_TNL_F_CAP_PER_PACKET;
 717         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 718                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 719                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 720                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 721                 if (ltype&IPV6_ADDR_UNICAST)
 722                         flags |= IP6_TNL_F_CAP_XMIT;
 723                 if (rtype&IPV6_ADDR_UNICAST)
 724                         flags |= IP6_TNL_F_CAP_RCV;
 725         }
 726         return flags;
 727 }
 728 EXPORT_SYMBOL(ip6_tnl_get_cap);
 729
 730 /* called with rcu_read_lock() */
 731 int ip6_tnl_rcv_ctl(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 ret = 0;
 737         struct net *net = t->net;
 738
 739         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 740             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 741              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 742                 struct net_device *ldev = NULL;
 743
 744                 if (p->link)
 745                         ldev = dev_get_by_index_rcu(net, p->link);
 746
 747                 if ((ipv6_addr_is_multicast(laddr) ||
 748                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 749                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 750                         ret = 1;
 751         }
 752         return ret;
 753 }
 754 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 755
 756 /**
 757  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
 758  *   @skb: received socket buffer
 759  *   @protocol: ethernet protocol ID
 760  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
 761  *
 762  * Return: 0
 763  **/
 764
 765 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 766                        __u8 ipproto,
 767                        int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 768                                                    const struct ipv6hdr *ipv6h,
 769                                                    struct sk_buff *skb))
 770 {
 771         struct ip6_tnl *t;
 772         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 773         u8 tproto;
 774         int err;
 775
 776         rcu_read_lock();
 777         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 778         if (t) {
 779                 struct pcpu_sw_netstats *tstats;
 780
 781                 tproto = ACCESS_ONCE(t->parms.proto);
 782                 if (tproto != ipproto && tproto != 0) {
 783                         rcu_read_unlock();
 784                         goto discard;
 785                 }
 786
 787                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 788                         rcu_read_unlock();
 789                         goto discard;
 790                 }
 791
 792                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
 793                         t->dev->stats.rx_dropped++;
 794                         rcu_read_unlock();
 795                         goto discard;
 796                 }
 797                 skb->mac_header = skb->network_header;
 798                 skb_reset_network_header(skb);
 799                 skb->protocol = htons(protocol);
 800                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 801
 802                 __skb_tunnel_rx(skb, t->dev, t->net);
 803
 804                 err = dscp_ecn_decapsulate(t, ipv6h, skb);
 805                 if (unlikely(err)) {
 806                         if (log_ecn_error)
 807                                 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
 808                                                      &ipv6h->saddr,
 809                                                      ipv6_get_dsfield(ipv6h));
 810                         if (err > 1) {
 811                                 ++t->dev->stats.rx_frame_errors;
 812                                 ++t->dev->stats.rx_errors;
 813                                 rcu_read_unlock();
 814                                 goto discard;
 815                         }
 816                 }
 817
 818                 tstats = this_cpu_ptr(t->dev->tstats);
 819                 u64_stats_update_begin(&tstats->syncp);
 820                 tstats->rx_packets++;
 821                 tstats->rx_bytes += skb->len;
 822                 u64_stats_update_end(&tstats->syncp);
 823
 824                 netif_rx(skb);
 825
 826                 rcu_read_unlock();
 827                 return 0;
 828         }
 829         rcu_read_unlock();
 830         return 1;
 831
 832 discard:
 833         kfree_skb(skb);
 834         return 0;
 835 }
 836
 837 static int ip4ip6_rcv(struct sk_buff *skb)
 838 {
 839         return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
 840                            ip4ip6_dscp_ecn_decapsulate);
 841 }
 842
 843 static int ip6ip6_rcv(struct sk_buff *skb)
 844 {
 845         return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
 846                            ip6ip6_dscp_ecn_decapsulate);
 847 }
 848
 849 struct ipv6_tel_txoption {
 850         struct ipv6_txoptions ops;
 851         __u8 dst_opt[8];
 852 };
 853
 854 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 855 {
 856         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 857
 858         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 859         opt->dst_opt[3] = 1;
 860         opt->dst_opt[4] = encap_limit;
 861         opt->dst_opt[5] = IPV6_TLV_PADN;
 862         opt->dst_opt[6] = 1;
 863
 864         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 865         opt->ops.opt_nflen = 8;
 866 }
 867
 868 /**
 869  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 870  *   @t: the outgoing tunnel device
 871  *   @hdr: IPv6 header from the incoming packet
 872  *
 873  * Description:
 874  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 875  *   doesn't match source of incoming packet.
 876  *
 877  * Return:
 878  *   1 if conflict,
 879  *   0 else
 880  **/
 881
 882 static inline bool
 883 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 884 {
 885         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 886 }
 887
 888 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 889                      const struct in6_addr *laddr,
 890                      const struct in6_addr *raddr)
 891 {
 892         struct __ip6_tnl_parm *p = &t->parms;
 893         int ret = 0;
 894         struct net *net = t->net;
 895
 896         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 897             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 898              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 899                 struct net_device *ldev = NULL;
 900
 901                 rcu_read_lock();
 902                 if (p->link)
 903                         ldev = dev_get_by_index_rcu(net, p->link);
 904
 905                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 906                         pr_warn("%s xmit: Local address not yet configured!\n",
 907                                 p->name);
 908                 else if (!ipv6_addr_is_multicast(raddr) &&
 909                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
 910                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 911                                 p->name);
 912                 else
 913                         ret = 1;
 914                 rcu_read_unlock();
 915         }
 916         return ret;
 917 }
 918 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
 919
 920 /**
 921  * ip6_tnl_xmit2 - encapsulate packet and send
 922  *   @skb: the outgoing socket buffer
 923  *   @dev: the outgoing tunnel device
 924  *   @dsfield: dscp code for outer header
 925  *   @fl: flow of tunneled packet
 926  *   @encap_limit: encapsulation limit
 927  *   @pmtu: Path MTU is stored if packet is too big
 928  *
 929  * Description:
 930  *   Build new header and do some sanity checks on the packet before sending
 931  *   it.
 932  *
 933  * Return:
 934  *   0 on success
 935  *   -1 fail
 936  *   %-EMSGSIZE message too big. return mtu in this case.
 937  **/
 938
 939 static int ip6_tnl_xmit2(struct sk_buff *skb,
 940                          struct net_device *dev,
 941                          __u8 dsfield,
 942                          struct flowi6 *fl6,
 943                          int encap_limit,
 944                          __u32 *pmtu)
 945 {
 946         struct ip6_tnl *t = netdev_priv(dev);
 947         struct net *net = t->net;
 948         struct net_device_stats *stats = &t->dev->stats;
 949         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 950         struct ipv6_tel_txoption opt;
 951         struct dst_entry *dst = NULL, *ndst = NULL;
 952         struct net_device *tdev;
 953         int mtu;
 954         unsigned int max_headroom = sizeof(struct ipv6hdr);
 955         u8 proto;
 956         int err = -1;
 957
 958         /* NBMA tunnel */
 959         if (ipv6_addr_any(&t->parms.raddr)) {
 960                 struct in6_addr *addr6;
 961                 struct neighbour *neigh;
 962                 int addr_type;
 963
 964                 if (!skb_dst(skb))
 965                         goto tx_err_link_failure;
 966
 967                 neigh = dst_neigh_lookup(skb_dst(skb),
 968                                          &ipv6_hdr(skb)->daddr);
 969                 if (!neigh)
 970                         goto tx_err_link_failure;
 971
 972                 addr6 = (struct in6_addr *)&neigh->primary_key;
 973                 addr_type = ipv6_addr_type(addr6);
 974
 975                 if (addr_type == IPV6_ADDR_ANY)
 976                         addr6 = &ipv6_hdr(skb)->daddr;
 977
 978                 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
 979                 neigh_release(neigh);
 980         } else if (!fl6->flowi6_mark)
 981                 dst = dst_cache_get(&t->dst_cache);
 982
 983         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
 984                 goto tx_err_link_failure;
 985
 986         if (!dst) {
 987                 dst = ip6_route_output(net, NULL, fl6);
 988
 989                 if (dst->error)
 990                         goto tx_err_link_failure;
 991                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
 992                 if (IS_ERR(dst)) {
 993                         err = PTR_ERR(dst);
 994                         dst = NULL;
 995                         goto tx_err_link_failure;
 996                 }
 997                 ndst = dst;
 998         }
 999
1000         tdev = dst->dev;
1001
1002         if (tdev == dev) {
1003                 stats->collisions++;
1004                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1005                                      t->parms.name);
1006                 goto tx_err_dst_release;
1007         }
1008         mtu = dst_mtu(dst) - sizeof(*ipv6h);
1009         if (encap_limit >= 0) {
1010                 max_headroom += 8;
1011                 mtu -= 8;
1012         }
1013         if (mtu < IPV6_MIN_MTU)
1014                 mtu = IPV6_MIN_MTU;
1015         if (skb_dst(skb))
1016                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1017         if (skb->len > mtu) {
1018                 *pmtu = mtu;
1019                 err = -EMSGSIZE;
1020                 goto tx_err_dst_release;
1021         }
1022
1023         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1024
1025         /*
1026          * Okay, now see if we can stuff it in the buffer as-is.
1027          */
1028         max_headroom += LL_RESERVED_SPACE(tdev);
1029
1030         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1031             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1032                 struct sk_buff *new_skb;
1033
1034                 new_skb = skb_realloc_headroom(skb, max_headroom);
1035                 if (!new_skb)
1036                         goto tx_err_dst_release;
1037
1038                 if (skb->sk)
1039                         skb_set_owner_w(new_skb, skb->sk);
1040                 consume_skb(skb);
1041                 skb = new_skb;
1042         }
1043
1044         if (!fl6->flowi6_mark && ndst)
1045                 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1046         skb_dst_set(skb, dst);
1047
1048         skb->transport_header = skb->network_header;
1049
1050         proto = fl6->flowi6_proto;
1051         if (encap_limit >= 0) {
1052                 init_tel_txopt(&opt, encap_limit);
1053                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1054         }
1055
1056         if (likely(!skb->encapsulation)) {
1057                 skb_reset_inner_headers(skb);
1058                 skb->encapsulation = 1;
1059         }
1060
1061         skb_push(skb, sizeof(struct ipv6hdr));
1062         skb_reset_network_header(skb);
1063         ipv6h = ipv6_hdr(skb);
1064         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1065                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1066         ipv6h->hop_limit = t->parms.hop_limit;
1067         ipv6h->nexthdr = proto;
1068         ipv6h->saddr = fl6->saddr;
1069         ipv6h->daddr = fl6->daddr;
1070         ip6tunnel_xmit(NULL, skb, dev);
1071         return 0;
1072 tx_err_link_failure:
1073         stats->tx_carrier_errors++;
1074         dst_link_failure(skb);
1075 tx_err_dst_release:
1076         dst_release(dst);
1077         return err;
1078 }
1079
1080 static inline int
1081 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1082 {
1083         struct ip6_tnl *t = netdev_priv(dev);
1084         const struct iphdr  *iph = ip_hdr(skb);
1085         int encap_limit = -1;
1086         struct flowi6 fl6;
1087         __u8 dsfield;
1088         __u32 mtu;
1089         u8 tproto;
1090         int err;
1091
1092         tproto = ACCESS_ONCE(t->parms.proto);
1093         if (tproto != IPPROTO_IPIP && tproto != 0)
1094                 return -1;
1095
1096         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1097                 encap_limit = t->parms.encap_limit;
1098
1099         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1100         fl6.flowi6_proto = IPPROTO_IPIP;
1101
1102         dsfield = ipv4_get_dsfield(iph);
1103
1104         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1105                 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1106                                           & IPV6_TCLASS_MASK;
1107         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1108                 fl6.flowi6_mark = skb->mark;
1109
1110         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1111         if (err != 0) {
1112                 /* XXX: send ICMP error even if DF is not set. */
1113                 if (err == -EMSGSIZE)
1114                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1115                                   htonl(mtu));
1116                 return -1;
1117         }
1118
1119         return 0;
1120 }
1121
1122 static inline int
1123 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1124 {
1125         struct ip6_tnl *t = netdev_priv(dev);
1126         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1127         int encap_limit = -1;
1128         __u16 offset;
1129         struct flowi6 fl6;
1130         __u8 dsfield;
1131         __u32 mtu;
1132         u8 tproto;
1133         int err;
1134
1135         tproto = ACCESS_ONCE(t->parms.proto);
1136         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1137             ip6_tnl_addr_conflict(t, ipv6h))
1138                 return -1;
1139
1140         offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1141         if (offset > 0) {
1142                 struct ipv6_tlv_tnl_enc_lim *tel;
1143                 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1144                 if (tel->encap_limit == 0) {
1145                         icmpv6_send(skb, ICMPV6_PARAMPROB,
1146                                     ICMPV6_HDR_FIELD, offset + 2);
1147                         return -1;
1148                 }
1149                 encap_limit = tel->encap_limit - 1;
1150         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1151                 encap_limit = t->parms.encap_limit;
1152
1153         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1154         fl6.flowi6_proto = IPPROTO_IPV6;
1155
1156         dsfield = ipv6_get_dsfield(ipv6h);
1157         if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1158                 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1159         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1160                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1161         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1162                 fl6.flowi6_mark = skb->mark;
1163
1164         err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1165         if (err != 0) {
1166                 if (err == -EMSGSIZE)
1167                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1168                 return -1;
1169         }
1170
1171         return 0;
1172 }
1173
1174 static netdev_tx_t
1175 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1176 {
1177         struct ip6_tnl *t = netdev_priv(dev);
1178         struct net_device_stats *stats = &t->dev->stats;
1179         int ret;
1180
1181         switch (skb->protocol) {
1182         case htons(ETH_P_IP):
1183                 ret = ip4ip6_tnl_xmit(skb, dev);
1184                 break;
1185         case htons(ETH_P_IPV6):
1186                 ret = ip6ip6_tnl_xmit(skb, dev);
1187                 break;
1188         default:
1189                 goto tx_err;
1190         }
1191
1192         if (ret < 0)
1193                 goto tx_err;
1194
1195         return NETDEV_TX_OK;
1196
1197 tx_err:
1198         stats->tx_errors++;
1199         stats->tx_dropped++;
1200         kfree_skb(skb);
1201         return NETDEV_TX_OK;
1202 }
1203
1204 static void ip6_tnl_link_config(struct ip6_tnl *t)
1205 {
1206         struct net_device *dev = t->dev;
1207         struct __ip6_tnl_parm *p = &t->parms;
1208         struct flowi6 *fl6 = &t->fl.u.ip6;
1209
1210         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1211         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1212
1213         /* Set up flowi template */
1214         fl6->saddr = p->laddr;
1215         fl6->daddr = p->raddr;
1216         fl6->flowi6_oif = p->link;
1217         fl6->flowlabel = 0;
1218
1219         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1220                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1221         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1222                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1223
1224         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1225         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1226
1227         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1228                 dev->flags |= IFF_POINTOPOINT;
1229         else
1230                 dev->flags &= ~IFF_POINTOPOINT;
1231
1232         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1233                 int strict = (ipv6_addr_type(&p->raddr) &
1234                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1235
1236                 struct rt6_info *rt = rt6_lookup(t->net,
1237                                                  &p->raddr, &p->laddr,
1238                                                  p->link, strict);
1239
1240                 if (!rt)
1241                         return;
1242
1243                 if (rt->dst.dev) {
1244                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1245                                 sizeof(struct ipv6hdr);
1246
1247                         dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1248                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1249                                 dev->mtu -= 8;
1250
1251                         if (dev->mtu < IPV6_MIN_MTU)
1252                                 dev->mtu = IPV6_MIN_MTU;
1253                 }
1254                 ip6_rt_put(rt);
1255         }
1256 }
1257
1258 /**
1259  * ip6_tnl_change - update the tunnel parameters
1260  *   @t: tunnel to be changed
1261  *   @p: tunnel configuration parameters
1262  *
1263  * Description:
1264  *   ip6_tnl_change() updates the tunnel parameters
1265  **/
1266
1267 static int
1268 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1269 {
1270         t->parms.laddr = p->laddr;
1271         t->parms.raddr = p->raddr;
1272         t->parms.flags = p->flags;
1273         t->parms.hop_limit = p->hop_limit;
1274         t->parms.encap_limit = p->encap_limit;
1275         t->parms.flowinfo = p->flowinfo;
1276         t->parms.link = p->link;
1277         t->parms.proto = p->proto;
1278         dst_cache_reset(&t->dst_cache);
1279         ip6_tnl_link_config(t);
1280         return 0;
1281 }
1282
1283 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1284 {
1285         struct net *net = t->net;
1286         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1287         int err;
1288
1289         ip6_tnl_unlink(ip6n, t);
1290         synchronize_net();
1291         err = ip6_tnl_change(t, p);
1292         ip6_tnl_link(ip6n, t);
1293         netdev_state_change(t->dev);
1294         return err;
1295 }
1296
1297 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1298 {
1299         /* for default tnl0 device allow to change only the proto */
1300         t->parms.proto = p->proto;
1301         netdev_state_change(t->dev);
1302         return 0;
1303 }
1304
1305 static void
1306 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1307 {
1308         p->laddr = u->laddr;
1309         p->raddr = u->raddr;
1310         p->flags = u->flags;
1311         p->hop_limit = u->hop_limit;
1312         p->encap_limit = u->encap_limit;
1313         p->flowinfo = u->flowinfo;
1314         p->link = u->link;
1315         p->proto = u->proto;
1316         memcpy(p->name, u->name, sizeof(u->name));
1317 }
1318
1319 static void
1320 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1321 {
1322         u->laddr = p->laddr;
1323         u->raddr = p->raddr;
1324         u->flags = p->flags;
1325         u->hop_limit = p->hop_limit;
1326         u->encap_limit = p->encap_limit;
1327         u->flowinfo = p->flowinfo;
1328         u->link = p->link;
1329         u->proto = p->proto;
1330         memcpy(u->name, p->name, sizeof(u->name));
1331 }
1332
1333 /**
1334  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1335  *   @dev: virtual device associated with tunnel
1336  *   @ifr: parameters passed from userspace
1337  *   @cmd: command to be performed
1338  *
1339  * Description:
1340  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1341  *   from userspace.
1342  *
1343  *   The possible commands are the following:
1344  *     %SIOCGETTUNNEL: get tunnel parameters for device
1345  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1346  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1347  *     %SIOCDELTUNNEL: delete tunnel
1348  *
1349  *   The fallback device "ip6tnl0", created during module
1350  *   initialization, can be used for creating other tunnel devices.
1351  *
1352  * Return:
1353  *   0 on success,
1354  *   %-EFAULT if unable to copy data to or from userspace,
1355  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1356  *   %-EINVAL if passed tunnel parameters are invalid,
1357  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1358  *   %-ENODEV if attempting to change or delete a nonexisting device
1359  **/
1360
1361 static int
1362 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1363 {
1364         int err = 0;
1365         struct ip6_tnl_parm p;
1366         struct __ip6_tnl_parm p1;
1367         struct ip6_tnl *t = netdev_priv(dev);
1368         struct net *net = t->net;
1369         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1370
1371         switch (cmd) {
1372         case SIOCGETTUNNEL:
1373                 if (dev == ip6n->fb_tnl_dev) {
1374                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1375                                 err = -EFAULT;
1376                                 break;
1377                         }
1378                         ip6_tnl_parm_from_user(&p1, &p);
1379                         t = ip6_tnl_locate(net, &p1, 0);
1380                         if (IS_ERR(t))
1381                                 t = netdev_priv(dev);
1382                 } else {
1383                         memset(&p, 0, sizeof(p));
1384                 }
1385                 ip6_tnl_parm_to_user(&p, &t->parms);
1386                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1387                         err = -EFAULT;
1388                 }
1389                 break;
1390         case SIOCADDTUNNEL:
1391         case SIOCCHGTUNNEL:
1392                 err = -EPERM;
1393                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1394                         break;
1395                 err = -EFAULT;
1396                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1397                         break;
1398                 err = -EINVAL;
1399                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1400                     p.proto != 0)
1401                         break;
1402                 ip6_tnl_parm_from_user(&p1, &p);
1403                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1404                 if (cmd == SIOCCHGTUNNEL) {
1405                         if (!IS_ERR(t)) {
1406                                 if (t->dev != dev) {
1407                                         err = -EEXIST;
1408                                         break;
1409                                 }
1410                         } else
1411                                 t = netdev_priv(dev);
1412                         if (dev == ip6n->fb_tnl_dev)
1413                                 err = ip6_tnl0_update(t, &p1);
1414                         else
1415                                 err = ip6_tnl_update(t, &p1);
1416                 }
1417                 if (!IS_ERR(t)) {
1418                         err = 0;
1419                         ip6_tnl_parm_to_user(&p, &t->parms);
1420                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1421                                 err = -EFAULT;
1422
1423                 } else {
1424                         err = PTR_ERR(t);
1425                 }
1426                 break;
1427         case SIOCDELTUNNEL:
1428                 err = -EPERM;
1429                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1430                         break;
1431
1432                 if (dev == ip6n->fb_tnl_dev) {
1433                         err = -EFAULT;
1434                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1435                                 break;
1436                         err = -ENOENT;
1437                         ip6_tnl_parm_from_user(&p1, &p);
1438                         t = ip6_tnl_locate(net, &p1, 0);
1439                         if (IS_ERR(t))
1440                                 break;
1441                         err = -EPERM;
1442                         if (t->dev == ip6n->fb_tnl_dev)
1443                                 break;
1444                         dev = t->dev;
1445                 }
1446                 err = 0;
1447                 unregister_netdevice(dev);
1448                 break;
1449         default:
1450                 err = -EINVAL;
1451         }
1452         return err;
1453 }
1454
1455 /**
1456  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1457  *   @dev: virtual device associated with tunnel
1458  *   @new_mtu: the new mtu
1459  *
1460  * Return:
1461  *   0 on success,
1462  *   %-EINVAL if mtu too small
1463  **/
1464
1465 static int
1466 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1467 {
1468         struct ip6_tnl *tnl = netdev_priv(dev);
1469
1470         if (tnl->parms.proto == IPPROTO_IPIP) {
1471                 if (new_mtu < 68)
1472                         return -EINVAL;
1473         } else {
1474                 if (new_mtu < IPV6_MIN_MTU)
1475                         return -EINVAL;
1476         }
1477         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1478                 return -EINVAL;
1479         dev->mtu = new_mtu;
1480         return 0;
1481 }
1482
1483 int ip6_tnl_get_iflink(const struct net_device *dev)
1484 {
1485         struct ip6_tnl *t = netdev_priv(dev);
1486
1487         return t->parms.link;
1488 }
1489 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1490
1491 static const struct net_device_ops ip6_tnl_netdev_ops = {
1492         .ndo_init       = ip6_tnl_dev_init,
1493         .ndo_uninit     = ip6_tnl_dev_uninit,
1494         .ndo_start_xmit = ip6_tnl_xmit,
1495         .ndo_do_ioctl   = ip6_tnl_ioctl,
1496         .ndo_change_mtu = ip6_tnl_change_mtu,
1497         .ndo_get_stats  = ip6_get_stats,
1498         .ndo_get_iflink = ip6_tnl_get_iflink,
1499 };
1500
1501
1502 /**
1503  * ip6_tnl_dev_setup - setup virtual tunnel device
1504  *   @dev: virtual device associated with tunnel
1505  *
1506  * Description:
1507  *   Initialize function pointers and device parameters
1508  **/
1509
1510 static void ip6_tnl_dev_setup(struct net_device *dev)
1511 {
1512         struct ip6_tnl *t;
1513
1514         dev->netdev_ops = &ip6_tnl_netdev_ops;
1515         dev->destructor = ip6_dev_free;
1516
1517         dev->type = ARPHRD_TUNNEL6;
1518         dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1519         dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1520         t = netdev_priv(dev);
1521         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1522                 dev->mtu -= 8;
1523         dev->flags |= IFF_NOARP;
1524         dev->addr_len = sizeof(struct in6_addr);
1525         netif_keep_dst(dev);
1526         /* This perm addr will be used as interface identifier by IPv6 */
1527         dev->addr_assign_type = NET_ADDR_RANDOM;
1528         eth_random_addr(dev->perm_addr);
1529 }
1530
1531
1532 /**
1533  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1534  *   @dev: virtual device associated with tunnel
1535  **/
1536
1537 static inline int
1538 ip6_tnl_dev_init_gen(struct net_device *dev)
1539 {
1540         struct ip6_tnl *t = netdev_priv(dev);
1541         int ret;
1542
1543         t->dev = dev;
1544         t->net = dev_net(dev);
1545         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1546         if (!dev->tstats)
1547                 return -ENOMEM;
1548
1549         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1550         if (ret) {
1551                 free_percpu(dev->tstats);
1552                 dev->tstats = NULL;
1553                 return ret;
1554         }
1555
1556         return 0;
1557 }
1558
1559 /**
1560  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1561  *   @dev: virtual device associated with tunnel
1562  **/
1563
1564 static int ip6_tnl_dev_init(struct net_device *dev)
1565 {
1566         struct ip6_tnl *t = netdev_priv(dev);
1567         int err = ip6_tnl_dev_init_gen(dev);
1568
1569         if (err)
1570                 return err;
1571         ip6_tnl_link_config(t);
1572         return 0;
1573 }
1574
1575 /**
1576  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1577  *   @dev: fallback device
1578  *
1579  * Return: 0
1580  **/
1581
1582 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1583 {
1584         struct ip6_tnl *t = netdev_priv(dev);
1585         struct net *net = dev_net(dev);
1586         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1587
1588         t->parms.proto = IPPROTO_IPV6;
1589         dev_hold(dev);
1590
1591         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1592         return 0;
1593 }
1594
1595 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1596 {
1597         u8 proto;
1598
1599         if (!data || !data[IFLA_IPTUN_PROTO])
1600                 return 0;
1601
1602         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1603         if (proto != IPPROTO_IPV6 &&
1604             proto != IPPROTO_IPIP &&
1605             proto != 0)
1606                 return -EINVAL;
1607
1608         return 0;
1609 }
1610
1611 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1612                                   struct __ip6_tnl_parm *parms)
1613 {
1614         memset(parms, 0, sizeof(*parms));
1615
1616         if (!data)
1617                 return;
1618
1619         if (data[IFLA_IPTUN_LINK])
1620                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1621
1622         if (data[IFLA_IPTUN_LOCAL])
1623                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1624
1625         if (data[IFLA_IPTUN_REMOTE])
1626                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1627
1628         if (data[IFLA_IPTUN_TTL])
1629                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1630
1631         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1632                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1633
1634         if (data[IFLA_IPTUN_FLOWINFO])
1635                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1636
1637         if (data[IFLA_IPTUN_FLAGS])
1638                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1639
1640         if (data[IFLA_IPTUN_PROTO])
1641                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1642 }
1643
1644 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1645                            struct nlattr *tb[], struct nlattr *data[])
1646 {
1647         struct net *net = dev_net(dev);
1648         struct ip6_tnl *nt, *t;
1649
1650         nt = netdev_priv(dev);
1651         ip6_tnl_netlink_parms(data, &nt->parms);
1652
1653         t = ip6_tnl_locate(net, &nt->parms, 0);
1654         if (!IS_ERR(t))
1655                 return -EEXIST;
1656
1657         return ip6_tnl_create2(dev);
1658 }
1659
1660 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1661                               struct nlattr *data[])
1662 {
1663         struct ip6_tnl *t = netdev_priv(dev);
1664         struct __ip6_tnl_parm p;
1665         struct net *net = t->net;
1666         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1667
1668         if (dev == ip6n->fb_tnl_dev)
1669                 return -EINVAL;
1670
1671         ip6_tnl_netlink_parms(data, &p);
1672
1673         t = ip6_tnl_locate(net, &p, 0);
1674         if (!IS_ERR(t)) {
1675                 if (t->dev != dev)
1676                         return -EEXIST;
1677         } else
1678                 t = netdev_priv(dev);
1679
1680         return ip6_tnl_update(t, &p);
1681 }
1682
1683 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1684 {
1685         struct net *net = dev_net(dev);
1686         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1687
1688         if (dev != ip6n->fb_tnl_dev)
1689                 unregister_netdevice_queue(dev, head);
1690 }
1691
1692 static size_t ip6_tnl_get_size(const struct net_device *dev)
1693 {
1694         return
1695                 /* IFLA_IPTUN_LINK */
1696                 nla_total_size(4) +
1697                 /* IFLA_IPTUN_LOCAL */
1698                 nla_total_size(sizeof(struct in6_addr)) +
1699                 /* IFLA_IPTUN_REMOTE */
1700                 nla_total_size(sizeof(struct in6_addr)) +
1701                 /* IFLA_IPTUN_TTL */
1702                 nla_total_size(1) +
1703                 /* IFLA_IPTUN_ENCAP_LIMIT */
1704                 nla_total_size(1) +
1705                 /* IFLA_IPTUN_FLOWINFO */
1706                 nla_total_size(4) +
1707                 /* IFLA_IPTUN_FLAGS */
1708                 nla_total_size(4) +
1709                 /* IFLA_IPTUN_PROTO */
1710                 nla_total_size(1) +
1711                 0;
1712 }
1713
1714 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1715 {
1716         struct ip6_tnl *tunnel = netdev_priv(dev);
1717         struct __ip6_tnl_parm *parm = &tunnel->parms;
1718
1719         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1720             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1721             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1722             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1723             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1724             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1725             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1726             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1727                 goto nla_put_failure;
1728         return 0;
1729
1730 nla_put_failure:
1731         return -EMSGSIZE;
1732 }
1733
1734 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1735 {
1736         struct ip6_tnl *tunnel = netdev_priv(dev);
1737
1738         return tunnel->net;
1739 }
1740 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1741
1742 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1743         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1744         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
1745         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
1746         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1747         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
1748         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
1749         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
1750         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1751 };
1752
1753 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1754         .kind           = "ip6tnl",
1755         .maxtype        = IFLA_IPTUN_MAX,
1756         .policy         = ip6_tnl_policy,
1757         .priv_size      = sizeof(struct ip6_tnl),
1758         .setup          = ip6_tnl_dev_setup,
1759         .validate       = ip6_tnl_validate,
1760         .newlink        = ip6_tnl_newlink,
1761         .changelink     = ip6_tnl_changelink,
1762         .dellink        = ip6_tnl_dellink,
1763         .get_size       = ip6_tnl_get_size,
1764         .fill_info      = ip6_tnl_fill_info,
1765         .get_link_net   = ip6_tnl_get_link_net,
1766 };
1767
1768 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1769         .handler        = ip4ip6_rcv,
1770         .err_handler    = ip4ip6_err,
1771         .priority       =       1,
1772 };
1773
1774 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1775         .handler        = ip6ip6_rcv,
1776         .err_handler    = ip6ip6_err,
1777         .priority       =       1,
1778 };
1779
1780 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1781 {
1782         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1783         struct net_device *dev, *aux;
1784         int h;
1785         struct ip6_tnl *t;
1786         LIST_HEAD(list);
1787
1788         for_each_netdev_safe(net, dev, aux)
1789                 if (dev->rtnl_link_ops == &ip6_link_ops)
1790                         unregister_netdevice_queue(dev, &list);
1791
1792         for (h = 0; h < HASH_SIZE; h++) {
1793                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1794                 while (t) {
1795                         /* If dev is in the same netns, it has already
1796                          * been added to the list by the previous loop.
1797                          */
1798                         if (!net_eq(dev_net(t->dev), net))
1799                                 unregister_netdevice_queue(t->dev, &list);
1800                         t = rtnl_dereference(t->next);
1801                 }
1802         }
1803
1804         unregister_netdevice_many(&list);
1805 }
1806
1807 static int __net_init ip6_tnl_init_net(struct net *net)
1808 {
1809         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1810         struct ip6_tnl *t = NULL;
1811         int err;
1812
1813         ip6n->tnls[0] = ip6n->tnls_wc;
1814         ip6n->tnls[1] = ip6n->tnls_r_l;
1815
1816         err = -ENOMEM;
1817         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1818                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1819
1820         if (!ip6n->fb_tnl_dev)
1821                 goto err_alloc_dev;
1822         dev_net_set(ip6n->fb_tnl_dev, net);
1823         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1824         /* FB netdevice is special: we have one, and only one per netns.
1825          * Allowing to move it to another netns is clearly unsafe.
1826          */
1827         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1828
1829         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1830         if (err < 0)
1831                 goto err_register;
1832
1833         err = register_netdev(ip6n->fb_tnl_dev);
1834         if (err < 0)
1835                 goto err_register;
1836
1837         t = netdev_priv(ip6n->fb_tnl_dev);
1838
1839         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1840         return 0;
1841
1842 err_register:
1843         ip6_dev_free(ip6n->fb_tnl_dev);
1844 err_alloc_dev:
1845         return err;
1846 }
1847
1848 static void __net_exit ip6_tnl_exit_net(struct net *net)
1849 {
1850         rtnl_lock();
1851         ip6_tnl_destroy_tunnels(net);
1852         rtnl_unlock();
1853 }
1854
1855 static struct pernet_operations ip6_tnl_net_ops = {
1856         .init = ip6_tnl_init_net,
1857         .exit = ip6_tnl_exit_net,
1858         .id   = &ip6_tnl_net_id,
1859         .size = sizeof(struct ip6_tnl_net),
1860 };
1861
1862 /**
1863  * ip6_tunnel_init - register protocol and reserve needed resources
1864  *
1865  * Return: 0 on success
1866  **/
1867
1868 static int __init ip6_tunnel_init(void)
1869 {
1870         int  err;
1871
1872         err = register_pernet_device(&ip6_tnl_net_ops);
1873         if (err < 0)
1874                 goto out_pernet;
1875
1876         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1877         if (err < 0) {
1878                 pr_err("%s: can't register ip4ip6\n", __func__);
1879                 goto out_ip4ip6;
1880         }
1881
1882         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1883         if (err < 0) {
1884                 pr_err("%s: can't register ip6ip6\n", __func__);
1885                 goto out_ip6ip6;
1886         }
1887         err = rtnl_link_register(&ip6_link_ops);
1888         if (err < 0)
1889                 goto rtnl_link_failed;
1890
1891         return 0;
1892
1893 rtnl_link_failed:
1894         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1895 out_ip6ip6:
1896         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1897 out_ip4ip6:
1898         unregister_pernet_device(&ip6_tnl_net_ops);
1899 out_pernet:
1900         return err;
1901 }
1902
1903 /**
1904  * ip6_tunnel_cleanup - free resources and unregister protocol
1905  **/
1906
1907 static void __exit ip6_tunnel_cleanup(void)
1908 {
1909         rtnl_link_unregister(&ip6_link_ops);
1910         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1911                 pr_info("%s: can't deregister ip4ip6\n", __func__);
1912
1913         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1914                 pr_info("%s: can't deregister ip6ip6\n", __func__);
1915
1916         unregister_pernet_device(&ip6_tnl_net_ops);
1917 }
1918
1919 module_init(ip6_tunnel_init);
1920 module_exit(ip6_tunnel_cleanup);