net/ipv6/sit.c

   1 /*
   2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  *
  14  *      Changes:
  15  * Roger Venning <r.venning@telstra.com>:       6to4 support
  16  * Nate Thompson <nate@thebog.net>:             6to4 support
  17  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  18  */
  19
  20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  21
  22 #include <linux/module.h>
  23 #include <linux/capability.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/socket.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/in6.h>
  30 #include <linux/netdevice.h>
  31 #include <linux/if_arp.h>
  32 #include <linux/icmp.h>
  33 #include <linux/slab.h>
  34 #include <asm/uaccess.h>
  35 #include <linux/init.h>
  36 #include <linux/netfilter_ipv4.h>
  37 #include <linux/if_ether.h>
  38
  39 #include <net/sock.h>
  40 #include <net/snmp.h>
  41
  42 #include <net/ipv6.h>
  43 #include <net/protocol.h>
  44 #include <net/transp_v6.h>
  45 #include <net/ip6_fib.h>
  46 #include <net/ip6_route.h>
  47 #include <net/ndisc.h>
  48 #include <net/addrconf.h>
  49 #include <net/ip.h>
  50 #include <net/udp.h>
  51 #include <net/icmp.h>
  52 #include <net/ipip.h>
  53 #include <net/inet_ecn.h>
  54 #include <net/xfrm.h>
  55 #include <net/dsfield.h>
  56 #include <net/net_namespace.h>
  57 #include <net/netns/generic.h>
  58
  59 /*
  60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  61
  62    For comments look at net/ipv4/ip_gre.c --ANK
  63  */
  64
  65 #define HASH_SIZE  16
  66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  67
  68 static int ipip6_tunnel_init(struct net_device *dev);
  69 static void ipip6_tunnel_setup(struct net_device *dev);
  70 static void ipip6_dev_free(struct net_device *dev);
  71 static struct rtnl_link_ops sit_link_ops __read_mostly;
  72
  73 static int sit_net_id __read_mostly;
  74 struct sit_net {
  75         struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
  76         struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
  77         struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
  78         struct ip_tunnel __rcu *tunnels_wc[1];
  79         struct ip_tunnel __rcu **tunnels[4];
  80
  81         struct net_device *fb_tunnel_dev;
  82 };
  83
  84 /*
  85  * Locking : hash tables are protected by RCU and RTNL
  86  */
  87
  88 #define for_each_ip_tunnel_rcu(start) \
  89         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
  90
  91 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
  92                                                    struct rtnl_link_stats64 *tot)
  93 {
  94         int i;
  95
  96         for_each_possible_cpu(i) {
  97                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
  98                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
  99                 unsigned int start;
 100
 101                 do {
 102                         start = u64_stats_fetch_begin_bh(&tstats->syncp);
 103                         rx_packets = tstats->rx_packets;
 104                         tx_packets = tstats->tx_packets;
 105                         rx_bytes = tstats->rx_bytes;
 106                         tx_bytes = tstats->tx_bytes;
 107                 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
 108
 109                 tot->rx_packets += rx_packets;
 110                 tot->tx_packets += tx_packets;
 111                 tot->rx_bytes   += rx_bytes;
 112                 tot->tx_bytes   += tx_bytes;
 113         }
 114
 115         tot->rx_errors = dev->stats.rx_errors;
 116         tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
 117         tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
 118         tot->tx_dropped = dev->stats.tx_dropped;
 119         tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
 120         tot->tx_errors = dev->stats.tx_errors;
 121
 122         return tot;
 123 }
 124
 125 /*
 126  * Must be invoked with rcu_read_lock
 127  */
 128 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
 129                 struct net_device *dev, __be32 remote, __be32 local)
 130 {
 131         unsigned int h0 = HASH(remote);
 132         unsigned int h1 = HASH(local);
 133         struct ip_tunnel *t;
 134         struct sit_net *sitn = net_generic(net, sit_net_id);
 135
 136         for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
 137                 if (local == t->parms.iph.saddr &&
 138                     remote == t->parms.iph.daddr &&
 139                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 140                     (t->dev->flags & IFF_UP))
 141                         return t;
 142         }
 143         for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
 144                 if (remote == t->parms.iph.daddr &&
 145                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 146                     (t->dev->flags & IFF_UP))
 147                         return t;
 148         }
 149         for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
 150                 if (local == t->parms.iph.saddr &&
 151                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 152                     (t->dev->flags & IFF_UP))
 153                         return t;
 154         }
 155         t = rcu_dereference(sitn->tunnels_wc[0]);
 156         if ((t != NULL) && (t->dev->flags & IFF_UP))
 157                 return t;
 158         return NULL;
 159 }
 160
 161 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 162                 struct ip_tunnel_parm *parms)
 163 {
 164         __be32 remote = parms->iph.daddr;
 165         __be32 local = parms->iph.saddr;
 166         unsigned int h = 0;
 167         int prio = 0;
 168
 169         if (remote) {
 170                 prio |= 2;
 171                 h ^= HASH(remote);
 172         }
 173         if (local) {
 174                 prio |= 1;
 175                 h ^= HASH(local);
 176         }
 177         return &sitn->tunnels[prio][h];
 178 }
 179
 180 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 181                 struct ip_tunnel *t)
 182 {
 183         return __ipip6_bucket(sitn, &t->parms);
 184 }
 185
 186 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 187 {
 188         struct ip_tunnel __rcu **tp;
 189         struct ip_tunnel *iter;
 190
 191         for (tp = ipip6_bucket(sitn, t);
 192              (iter = rtnl_dereference(*tp)) != NULL;
 193              tp = &iter->next) {
 194                 if (t == iter) {
 195                         rcu_assign_pointer(*tp, t->next);
 196                         break;
 197                 }
 198         }
 199 }
 200
 201 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 202 {
 203         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 204
 205         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 206         rcu_assign_pointer(*tp, t);
 207 }
 208
 209 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 210 {
 211 #ifdef CONFIG_IPV6_SIT_6RD
 212         struct ip_tunnel *t = netdev_priv(dev);
 213
 214         if (t->dev == sitn->fb_tunnel_dev) {
 215                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 216                 t->ip6rd.relay_prefix = 0;
 217                 t->ip6rd.prefixlen = 16;
 218                 t->ip6rd.relay_prefixlen = 0;
 219         } else {
 220                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 221                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 222         }
 223 #endif
 224 }
 225
 226 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 227                 struct ip_tunnel_parm *parms, int create)
 228 {
 229         __be32 remote = parms->iph.daddr;
 230         __be32 local = parms->iph.saddr;
 231         struct ip_tunnel *t, *nt;
 232         struct ip_tunnel __rcu **tp;
 233         struct net_device *dev;
 234         char name[IFNAMSIZ];
 235         struct sit_net *sitn = net_generic(net, sit_net_id);
 236
 237         for (tp = __ipip6_bucket(sitn, parms);
 238             (t = rtnl_dereference(*tp)) != NULL;
 239              tp = &t->next) {
 240                 if (local == t->parms.iph.saddr &&
 241                     remote == t->parms.iph.daddr &&
 242                     parms->link == t->parms.link) {
 243                         if (create)
 244                                 return NULL;
 245                         else
 246                                 return t;
 247                 }
 248         }
 249         if (!create)
 250                 goto failed;
 251
 252         if (parms->name[0])
 253                 strlcpy(name, parms->name, IFNAMSIZ);
 254         else
 255                 strcpy(name, "sit%d");
 256
 257         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 258         if (dev == NULL)
 259                 return NULL;
 260
 261         dev_net_set(dev, net);
 262
 263         nt = netdev_priv(dev);
 264
 265         nt->parms = *parms;
 266         if (ipip6_tunnel_init(dev) < 0)
 267                 goto failed_free;
 268         ipip6_tunnel_clone_6rd(dev, sitn);
 269
 270         if (parms->i_flags & SIT_ISATAP)
 271                 dev->priv_flags |= IFF_ISATAP;
 272
 273         if (register_netdevice(dev) < 0)
 274                 goto failed_free;
 275
 276         strcpy(nt->parms.name, dev->name);
 277         dev->rtnl_link_ops = &sit_link_ops;
 278
 279         dev_hold(dev);
 280
 281         ipip6_tunnel_link(sitn, nt);
 282         return nt;
 283
 284 failed_free:
 285         ipip6_dev_free(dev);
 286 failed:
 287         return NULL;
 288 }
 289
 290 #define for_each_prl_rcu(start)                 \
 291         for (prl = rcu_dereference(start);      \
 292              prl;                               \
 293              prl = rcu_dereference(prl->next))
 294
 295 static struct ip_tunnel_prl_entry *
 296 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 297 {
 298         struct ip_tunnel_prl_entry *prl;
 299
 300         for_each_prl_rcu(t->prl)
 301                 if (prl->addr == addr)
 302                         break;
 303         return prl;
 304
 305 }
 306
 307 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 308                                 struct ip_tunnel_prl __user *a)
 309 {
 310         struct ip_tunnel_prl kprl, *kp;
 311         struct ip_tunnel_prl_entry *prl;
 312         unsigned int cmax, c = 0, ca, len;
 313         int ret = 0;
 314
 315         if (copy_from_user(&kprl, a, sizeof(kprl)))
 316                 return -EFAULT;
 317         cmax = kprl.datalen / sizeof(kprl);
 318         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 319                 cmax = 1;
 320
 321         /* For simple GET or for root users,
 322          * we try harder to allocate.
 323          */
 324         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 325                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 326                 NULL;
 327
 328         rcu_read_lock();
 329
 330         ca = t->prl_count < cmax ? t->prl_count : cmax;
 331
 332         if (!kp) {
 333                 /* We don't try hard to allocate much memory for
 334                  * non-root users.
 335                  * For root users, retry allocating enough memory for
 336                  * the answer.
 337                  */
 338                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 339                 if (!kp) {
 340                         ret = -ENOMEM;
 341                         goto out;
 342                 }
 343         }
 344
 345         c = 0;
 346         for_each_prl_rcu(t->prl) {
 347                 if (c >= cmax)
 348                         break;
 349                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 350                         continue;
 351                 kp[c].addr = prl->addr;
 352                 kp[c].flags = prl->flags;
 353                 c++;
 354                 if (kprl.addr != htonl(INADDR_ANY))
 355                         break;
 356         }
 357 out:
 358         rcu_read_unlock();
 359
 360         len = sizeof(*kp) * c;
 361         ret = 0;
 362         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 363                 ret = -EFAULT;
 364
 365         kfree(kp);
 366
 367         return ret;
 368 }
 369
 370 static int
 371 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 372 {
 373         struct ip_tunnel_prl_entry *p;
 374         int err = 0;
 375
 376         if (a->addr == htonl(INADDR_ANY))
 377                 return -EINVAL;
 378
 379         ASSERT_RTNL();
 380
 381         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 382                 if (p->addr == a->addr) {
 383                         if (chg) {
 384                                 p->flags = a->flags;
 385                                 goto out;
 386                         }
 387                         err = -EEXIST;
 388                         goto out;
 389                 }
 390         }
 391
 392         if (chg) {
 393                 err = -ENXIO;
 394                 goto out;
 395         }
 396
 397         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 398         if (!p) {
 399                 err = -ENOBUFS;
 400                 goto out;
 401         }
 402
 403         p->next = t->prl;
 404         p->addr = a->addr;
 405         p->flags = a->flags;
 406         t->prl_count++;
 407         rcu_assign_pointer(t->prl, p);
 408 out:
 409         return err;
 410 }
 411
 412 static void prl_list_destroy_rcu(struct rcu_head *head)
 413 {
 414         struct ip_tunnel_prl_entry *p, *n;
 415
 416         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 417         do {
 418                 n = rcu_dereference_protected(p->next, 1);
 419                 kfree(p);
 420                 p = n;
 421         } while (p);
 422 }
 423
 424 static int
 425 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 426 {
 427         struct ip_tunnel_prl_entry *x;
 428         struct ip_tunnel_prl_entry __rcu **p;
 429         int err = 0;
 430
 431         ASSERT_RTNL();
 432
 433         if (a && a->addr != htonl(INADDR_ANY)) {
 434                 for (p = &t->prl;
 435                      (x = rtnl_dereference(*p)) != NULL;
 436                      p = &x->next) {
 437                         if (x->addr == a->addr) {
 438                                 *p = x->next;
 439                                 kfree_rcu(x, rcu_head);
 440                                 t->prl_count--;
 441                                 goto out;
 442                         }
 443                 }
 444                 err = -ENXIO;
 445         } else {
 446                 x = rtnl_dereference(t->prl);
 447                 if (x) {
 448                         t->prl_count = 0;
 449                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 450                         t->prl = NULL;
 451                 }
 452         }
 453 out:
 454         return err;
 455 }
 456
 457 static int
 458 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 459 {
 460         struct ip_tunnel_prl_entry *p;
 461         int ok = 1;
 462
 463         rcu_read_lock();
 464         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 465         if (p) {
 466                 if (p->flags & PRL_DEFAULT)
 467                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 468                 else
 469                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 470         } else {
 471                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 472
 473                 if (ipv6_addr_is_isatap(addr6) &&
 474                     (addr6->s6_addr32[3] == iph->saddr) &&
 475                     ipv6_chk_prefix(addr6, t->dev))
 476                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 477                 else
 478                         ok = 0;
 479         }
 480         rcu_read_unlock();
 481         return ok;
 482 }
 483
 484 static void ipip6_tunnel_uninit(struct net_device *dev)
 485 {
 486         struct net *net = dev_net(dev);
 487         struct sit_net *sitn = net_generic(net, sit_net_id);
 488
 489         if (dev == sitn->fb_tunnel_dev) {
 490                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 491         } else {
 492                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 493                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 494         }
 495         dev_put(dev);
 496 }
 497
 498
 499 static int ipip6_err(struct sk_buff *skb, u32 info)
 500 {
 501
 502 /* All the routers (except for Linux) return only
 503    8 bytes of packet payload. It means, that precise relaying of
 504    ICMP in the real Internet is absolutely infeasible.
 505  */
 506         const struct iphdr *iph = (const struct iphdr *)skb->data;
 507         const int type = icmp_hdr(skb)->type;
 508         const int code = icmp_hdr(skb)->code;
 509         struct ip_tunnel *t;
 510         int err;
 511
 512         switch (type) {
 513         default:
 514         case ICMP_PARAMETERPROB:
 515                 return 0;
 516
 517         case ICMP_DEST_UNREACH:
 518                 switch (code) {
 519                 case ICMP_SR_FAILED:
 520                 case ICMP_PORT_UNREACH:
 521                         /* Impossible event. */
 522                         return 0;
 523                 default:
 524                         /* All others are translated to HOST_UNREACH.
 525                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 526                            I believe they are just ether pollution. --ANK
 527                          */
 528                         break;
 529                 }
 530                 break;
 531         case ICMP_TIME_EXCEEDED:
 532                 if (code != ICMP_EXC_TTL)
 533                         return 0;
 534                 break;
 535         case ICMP_REDIRECT:
 536                 break;
 537         }
 538
 539         err = -ENOENT;
 540
 541         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 542                                 skb->dev,
 543                                 iph->daddr,
 544                                 iph->saddr);
 545         if (t == NULL)
 546                 goto out;
 547
 548         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 549                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 550                                  t->dev->ifindex, 0, IPPROTO_IPV6, 0);
 551                 err = 0;
 552                 goto out;
 553         }
 554         if (type == ICMP_REDIRECT) {
 555                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 556                               IPPROTO_IPV6, 0);
 557                 err = 0;
 558                 goto out;
 559         }
 560
 561         if (t->parms.iph.daddr == 0)
 562                 goto out;
 563
 564         err = 0;
 565         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 566                 goto out;
 567
 568         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 569                 t->err_count++;
 570         else
 571                 t->err_count = 1;
 572         t->err_time = jiffies;
 573 out:
 574         return err;
 575 }
 576
 577 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
 578 {
 579         if (INET_ECN_is_ce(iph->tos))
 580                 IP6_ECN_set_ce(ipv6_hdr(skb));
 581 }
 582
 583 static int ipip6_rcv(struct sk_buff *skb)
 584 {
 585         const struct iphdr *iph;
 586         struct ip_tunnel *tunnel;
 587
 588         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 589                 goto out;
 590
 591         iph = ip_hdr(skb);
 592
 593         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 594                                      iph->saddr, iph->daddr);
 595         if (tunnel != NULL) {
 596                 struct pcpu_tstats *tstats;
 597
 598                 secpath_reset(skb);
 599                 skb->mac_header = skb->network_header;
 600                 skb_reset_network_header(skb);
 601                 IPCB(skb)->flags = 0;
 602                 skb->protocol = htons(ETH_P_IPV6);
 603                 skb->pkt_type = PACKET_HOST;
 604
 605                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 606                     !isatap_chksrc(skb, iph, tunnel)) {
 607                         tunnel->dev->stats.rx_errors++;
 608                         kfree_skb(skb);
 609                         return 0;
 610                 }
 611
 612                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 613                 tstats->rx_packets++;
 614                 tstats->rx_bytes += skb->len;
 615
 616                 __skb_tunnel_rx(skb, tunnel->dev);
 617
 618                 ipip6_ecn_decapsulate(iph, skb);
 619
 620                 netif_rx(skb);
 621
 622                 return 0;
 623         }
 624
 625         /* no tunnel matched,  let upstream know, ipsec may handle it */
 626         return 1;
 627 out:
 628         kfree_skb(skb);
 629         return 0;
 630 }
 631
 632 /*
 633  * Returns the embedded IPv4 address if the IPv6 address
 634  * comes from 6rd / 6to4 (RFC 3056) addr space.
 635  */
 636 static inline
 637 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 638 {
 639         __be32 dst = 0;
 640
 641 #ifdef CONFIG_IPV6_SIT_6RD
 642         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 643                               tunnel->ip6rd.prefixlen)) {
 644                 unsigned int pbw0, pbi0;
 645                 int pbi1;
 646                 u32 d;
 647
 648                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 649                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 650
 651                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 652                     tunnel->ip6rd.relay_prefixlen;
 653
 654                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 655                 if (pbi1 > 0)
 656                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 657                              (32 - pbi1);
 658
 659                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
 660         }
 661 #else
 662         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 663                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 664                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 665         }
 666 #endif
 667         return dst;
 668 }
 669
 670 /*
 671  *      This function assumes it is being called from dev_queue_xmit()
 672  *      and that skb is filled properly by that function.
 673  */
 674
 675 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 676                                      struct net_device *dev)
 677 {
 678         struct ip_tunnel *tunnel = netdev_priv(dev);
 679         const struct iphdr  *tiph = &tunnel->parms.iph;
 680         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 681         u8     tos = tunnel->parms.iph.tos;
 682         __be16 df = tiph->frag_off;
 683         struct rtable *rt;                      /* Route to the other host */
 684         struct net_device *tdev;                /* Device to other host */
 685         struct iphdr  *iph;                     /* Our new IP header */
 686         unsigned int max_headroom;              /* The extra header space needed */
 687         __be32 dst = tiph->daddr;
 688         struct flowi4 fl4;
 689         int    mtu;
 690         const struct in6_addr *addr6;
 691         int addr_type;
 692
 693         if (skb->protocol != htons(ETH_P_IPV6))
 694                 goto tx_error;
 695
 696         if (tos == 1)
 697                 tos = ipv6_get_dsfield(iph6);
 698
 699         /* ISATAP (RFC4214) - must come before 6to4 */
 700         if (dev->priv_flags & IFF_ISATAP) {
 701                 struct neighbour *neigh = NULL;
 702                 bool do_tx_error = false;
 703
 704                 if (skb_dst(skb))
 705                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 706
 707                 if (neigh == NULL) {
 708                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 709                         goto tx_error;
 710                 }
 711
 712                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 713                 addr_type = ipv6_addr_type(addr6);
 714
 715                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 716                      ipv6_addr_is_isatap(addr6))
 717                         dst = addr6->s6_addr32[3];
 718                 else
 719                         do_tx_error = true;
 720
 721                 neigh_release(neigh);
 722                 if (do_tx_error)
 723                         goto tx_error;
 724         }
 725
 726         if (!dst)
 727                 dst = try_6rd(&iph6->daddr, tunnel);
 728
 729         if (!dst) {
 730                 struct neighbour *neigh = NULL;
 731                 bool do_tx_error = false;
 732
 733                 if (skb_dst(skb))
 734                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 735
 736                 if (neigh == NULL) {
 737                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 738                         goto tx_error;
 739                 }
 740
 741                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 742                 addr_type = ipv6_addr_type(addr6);
 743
 744                 if (addr_type == IPV6_ADDR_ANY) {
 745                         addr6 = &ipv6_hdr(skb)->daddr;
 746                         addr_type = ipv6_addr_type(addr6);
 747                 }
 748
 749                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 750                         dst = addr6->s6_addr32[3];
 751                 else
 752                         do_tx_error = true;
 753
 754                 neigh_release(neigh);
 755                 if (do_tx_error)
 756                         goto tx_error;
 757         }
 758
 759         rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 760                                    dst, tiph->saddr,
 761                                    0, 0,
 762                                    IPPROTO_IPV6, RT_TOS(tos),
 763                                    tunnel->parms.link);
 764         if (IS_ERR(rt)) {
 765                 dev->stats.tx_carrier_errors++;
 766                 goto tx_error_icmp;
 767         }
 768         if (rt->rt_type != RTN_UNICAST) {
 769                 ip_rt_put(rt);
 770                 dev->stats.tx_carrier_errors++;
 771                 goto tx_error_icmp;
 772         }
 773         tdev = rt->dst.dev;
 774
 775         if (tdev == dev) {
 776                 ip_rt_put(rt);
 777                 dev->stats.collisions++;
 778                 goto tx_error;
 779         }
 780
 781         if (df) {
 782                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 783
 784                 if (mtu < 68) {
 785                         dev->stats.collisions++;
 786                         ip_rt_put(rt);
 787                         goto tx_error;
 788                 }
 789
 790                 if (mtu < IPV6_MIN_MTU) {
 791                         mtu = IPV6_MIN_MTU;
 792                         df = 0;
 793                 }
 794
 795                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 796                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 797
 798                 if (skb->len > mtu) {
 799                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 800                         ip_rt_put(rt);
 801                         goto tx_error;
 802                 }
 803         }
 804
 805         if (tunnel->err_count > 0) {
 806                 if (time_before(jiffies,
 807                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 808                         tunnel->err_count--;
 809                         dst_link_failure(skb);
 810                 } else
 811                         tunnel->err_count = 0;
 812         }
 813
 814         /*
 815          * Okay, now see if we can stuff it in the buffer as-is.
 816          */
 817         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 818
 819         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 820             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 821                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 822                 if (!new_skb) {
 823                         ip_rt_put(rt);
 824                         dev->stats.tx_dropped++;
 825                         dev_kfree_skb(skb);
 826                         return NETDEV_TX_OK;
 827                 }
 828                 if (skb->sk)
 829                         skb_set_owner_w(new_skb, skb->sk);
 830                 dev_kfree_skb(skb);
 831                 skb = new_skb;
 832                 iph6 = ipv6_hdr(skb);
 833         }
 834
 835         skb->transport_header = skb->network_header;
 836         skb_push(skb, sizeof(struct iphdr));
 837         skb_reset_network_header(skb);
 838         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 839         IPCB(skb)->flags = 0;
 840         skb_dst_drop(skb);
 841         skb_dst_set(skb, &rt->dst);
 842
 843         /*
 844          *      Push down and install the IPIP header.
 845          */
 846
 847         iph                     =       ip_hdr(skb);
 848         iph->version            =       4;
 849         iph->ihl                =       sizeof(struct iphdr)>>2;
 850         iph->frag_off           =       df;
 851         iph->protocol           =       IPPROTO_IPV6;
 852         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 853         iph->daddr              =       fl4.daddr;
 854         iph->saddr              =       fl4.saddr;
 855
 856         if ((iph->ttl = tiph->ttl) == 0)
 857                 iph->ttl        =       iph6->hop_limit;
 858
 859         iptunnel_xmit(skb, dev);
 860         return NETDEV_TX_OK;
 861
 862 tx_error_icmp:
 863         dst_link_failure(skb);
 864 tx_error:
 865         dev->stats.tx_errors++;
 866         dev_kfree_skb(skb);
 867         return NETDEV_TX_OK;
 868 }
 869
 870 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 871 {
 872         struct net_device *tdev = NULL;
 873         struct ip_tunnel *tunnel;
 874         const struct iphdr *iph;
 875         struct flowi4 fl4;
 876
 877         tunnel = netdev_priv(dev);
 878         iph = &tunnel->parms.iph;
 879
 880         if (iph->daddr) {
 881                 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 882                                                           iph->daddr, iph->saddr,
 883                                                           0, 0,
 884                                                           IPPROTO_IPV6,
 885                                                           RT_TOS(iph->tos),
 886                                                           tunnel->parms.link);
 887
 888                 if (!IS_ERR(rt)) {
 889                         tdev = rt->dst.dev;
 890                         ip_rt_put(rt);
 891                 }
 892                 dev->flags |= IFF_POINTOPOINT;
 893         }
 894
 895         if (!tdev && tunnel->parms.link)
 896                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 897
 898         if (tdev) {
 899                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 900                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 901                 if (dev->mtu < IPV6_MIN_MTU)
 902                         dev->mtu = IPV6_MIN_MTU;
 903         }
 904         dev->iflink = tunnel->parms.link;
 905 }
 906
 907 static int
 908 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 909 {
 910         int err = 0;
 911         struct ip_tunnel_parm p;
 912         struct ip_tunnel_prl prl;
 913         struct ip_tunnel *t;
 914         struct net *net = dev_net(dev);
 915         struct sit_net *sitn = net_generic(net, sit_net_id);
 916 #ifdef CONFIG_IPV6_SIT_6RD
 917         struct ip_tunnel_6rd ip6rd;
 918 #endif
 919
 920         switch (cmd) {
 921         case SIOCGETTUNNEL:
 922 #ifdef CONFIG_IPV6_SIT_6RD
 923         case SIOCGET6RD:
 924 #endif
 925                 t = NULL;
 926                 if (dev == sitn->fb_tunnel_dev) {
 927                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 928                                 err = -EFAULT;
 929                                 break;
 930                         }
 931                         t = ipip6_tunnel_locate(net, &p, 0);
 932                 }
 933                 if (t == NULL)
 934                         t = netdev_priv(dev);
 935
 936                 err = -EFAULT;
 937                 if (cmd == SIOCGETTUNNEL) {
 938                         memcpy(&p, &t->parms, sizeof(p));
 939                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
 940                                          sizeof(p)))
 941                                 goto done;
 942 #ifdef CONFIG_IPV6_SIT_6RD
 943                 } else {
 944                         ip6rd.prefix = t->ip6rd.prefix;
 945                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
 946                         ip6rd.prefixlen = t->ip6rd.prefixlen;
 947                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
 948                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
 949                                          sizeof(ip6rd)))
 950                                 goto done;
 951 #endif
 952                 }
 953                 err = 0;
 954                 break;
 955
 956         case SIOCADDTUNNEL:
 957         case SIOCCHGTUNNEL:
 958                 err = -EPERM;
 959                 if (!capable(CAP_NET_ADMIN))
 960                         goto done;
 961
 962                 err = -EFAULT;
 963                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 964                         goto done;
 965
 966                 err = -EINVAL;
 967                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 968                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 969                         goto done;
 970                 if (p.iph.ttl)
 971                         p.iph.frag_off |= htons(IP_DF);
 972
 973                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 974
 975                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 976                         if (t != NULL) {
 977                                 if (t->dev != dev) {
 978                                         err = -EEXIST;
 979                                         break;
 980                                 }
 981                         } else {
 982                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 983                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 984                                         err = -EINVAL;
 985                                         break;
 986                                 }
 987                                 t = netdev_priv(dev);
 988                                 ipip6_tunnel_unlink(sitn, t);
 989                                 synchronize_net();
 990                                 t->parms.iph.saddr = p.iph.saddr;
 991                                 t->parms.iph.daddr = p.iph.daddr;
 992                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 993                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 994                                 ipip6_tunnel_link(sitn, t);
 995                                 netdev_state_change(dev);
 996                         }
 997                 }
 998
 999                 if (t) {
1000                         err = 0;
1001                         if (cmd == SIOCCHGTUNNEL) {
1002                                 t->parms.iph.ttl = p.iph.ttl;
1003                                 t->parms.iph.tos = p.iph.tos;
1004                                 if (t->parms.link != p.link) {
1005                                         t->parms.link = p.link;
1006                                         ipip6_tunnel_bind_dev(dev);
1007                                         netdev_state_change(dev);
1008                                 }
1009                         }
1010                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1011                                 err = -EFAULT;
1012                 } else
1013                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1014                 break;
1015
1016         case SIOCDELTUNNEL:
1017                 err = -EPERM;
1018                 if (!capable(CAP_NET_ADMIN))
1019                         goto done;
1020
1021                 if (dev == sitn->fb_tunnel_dev) {
1022                         err = -EFAULT;
1023                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1024                                 goto done;
1025                         err = -ENOENT;
1026                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1027                                 goto done;
1028                         err = -EPERM;
1029                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1030                                 goto done;
1031                         dev = t->dev;
1032                 }
1033                 unregister_netdevice(dev);
1034                 err = 0;
1035                 break;
1036
1037         case SIOCGETPRL:
1038                 err = -EINVAL;
1039                 if (dev == sitn->fb_tunnel_dev)
1040                         goto done;
1041                 err = -ENOENT;
1042                 if (!(t = netdev_priv(dev)))
1043                         goto done;
1044                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1045                 break;
1046
1047         case SIOCADDPRL:
1048         case SIOCDELPRL:
1049         case SIOCCHGPRL:
1050                 err = -EPERM;
1051                 if (!capable(CAP_NET_ADMIN))
1052                         goto done;
1053                 err = -EINVAL;
1054                 if (dev == sitn->fb_tunnel_dev)
1055                         goto done;
1056                 err = -EFAULT;
1057                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1058                         goto done;
1059                 err = -ENOENT;
1060                 if (!(t = netdev_priv(dev)))
1061                         goto done;
1062
1063                 switch (cmd) {
1064                 case SIOCDELPRL:
1065                         err = ipip6_tunnel_del_prl(t, &prl);
1066                         break;
1067                 case SIOCADDPRL:
1068                 case SIOCCHGPRL:
1069                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1070                         break;
1071                 }
1072                 netdev_state_change(dev);
1073                 break;
1074
1075 #ifdef CONFIG_IPV6_SIT_6RD
1076         case SIOCADD6RD:
1077         case SIOCCHG6RD:
1078         case SIOCDEL6RD:
1079                 err = -EPERM;
1080                 if (!capable(CAP_NET_ADMIN))
1081                         goto done;
1082
1083                 err = -EFAULT;
1084                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1085                                    sizeof(ip6rd)))
1086                         goto done;
1087
1088                 t = netdev_priv(dev);
1089
1090                 if (cmd != SIOCDEL6RD) {
1091                         struct in6_addr prefix;
1092                         __be32 relay_prefix;
1093
1094                         err = -EINVAL;
1095                         if (ip6rd.relay_prefixlen > 32 ||
1096                             ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1097                                 goto done;
1098
1099                         ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1100                                          ip6rd.prefixlen);
1101                         if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1102                                 goto done;
1103                         if (ip6rd.relay_prefixlen)
1104                                 relay_prefix = ip6rd.relay_prefix &
1105                                                htonl(0xffffffffUL <<
1106                                                      (32 - ip6rd.relay_prefixlen));
1107                         else
1108                                 relay_prefix = 0;
1109                         if (relay_prefix != ip6rd.relay_prefix)
1110                                 goto done;
1111
1112                         t->ip6rd.prefix = prefix;
1113                         t->ip6rd.relay_prefix = relay_prefix;
1114                         t->ip6rd.prefixlen = ip6rd.prefixlen;
1115                         t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1116                 } else
1117                         ipip6_tunnel_clone_6rd(dev, sitn);
1118
1119                 err = 0;
1120                 break;
1121 #endif
1122
1123         default:
1124                 err = -EINVAL;
1125         }
1126
1127 done:
1128         return err;
1129 }
1130
1131 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1132 {
1133         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1134                 return -EINVAL;
1135         dev->mtu = new_mtu;
1136         return 0;
1137 }
1138
1139 static const struct net_device_ops ipip6_netdev_ops = {
1140         .ndo_uninit     = ipip6_tunnel_uninit,
1141         .ndo_start_xmit = ipip6_tunnel_xmit,
1142         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1143         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1144         .ndo_get_stats64= ipip6_get_stats64,
1145 };
1146
1147 static void ipip6_dev_free(struct net_device *dev)
1148 {
1149         free_percpu(dev->tstats);
1150         free_netdev(dev);
1151 }
1152
1153 static void ipip6_tunnel_setup(struct net_device *dev)
1154 {
1155         dev->netdev_ops         = &ipip6_netdev_ops;
1156         dev->destructor         = ipip6_dev_free;
1157
1158         dev->type               = ARPHRD_SIT;
1159         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1160         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1161         dev->flags              = IFF_NOARP;
1162         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1163         dev->iflink             = 0;
1164         dev->addr_len           = 4;
1165         dev->features           |= NETIF_F_NETNS_LOCAL;
1166         dev->features           |= NETIF_F_LLTX;
1167 }
1168
1169 static int ipip6_tunnel_init(struct net_device *dev)
1170 {
1171         struct ip_tunnel *tunnel = netdev_priv(dev);
1172
1173         tunnel->dev = dev;
1174
1175         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1176         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1177
1178         ipip6_tunnel_bind_dev(dev);
1179         dev->tstats = alloc_percpu(struct pcpu_tstats);
1180         if (!dev->tstats)
1181                 return -ENOMEM;
1182
1183         return 0;
1184 }
1185
1186 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1187 {
1188         struct ip_tunnel *tunnel = netdev_priv(dev);
1189         struct iphdr *iph = &tunnel->parms.iph;
1190         struct net *net = dev_net(dev);
1191         struct sit_net *sitn = net_generic(net, sit_net_id);
1192
1193         tunnel->dev = dev;
1194         strcpy(tunnel->parms.name, dev->name);
1195
1196         iph->version            = 4;
1197         iph->protocol           = IPPROTO_IPV6;
1198         iph->ihl                = 5;
1199         iph->ttl                = 64;
1200
1201         dev->tstats = alloc_percpu(struct pcpu_tstats);
1202         if (!dev->tstats)
1203                 return -ENOMEM;
1204         dev_hold(dev);
1205         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1206         return 0;
1207 }
1208
1209 static size_t sit_get_size(const struct net_device *dev)
1210 {
1211         return
1212                 /* IFLA_IPTUN_LINK */
1213                 nla_total_size(4) +
1214                 /* IFLA_IPTUN_LOCAL */
1215                 nla_total_size(4) +
1216                 /* IFLA_IPTUN_REMOTE */
1217                 nla_total_size(4) +
1218                 /* IFLA_IPTUN_TTL */
1219                 nla_total_size(1) +
1220                 /* IFLA_IPTUN_TOS */
1221                 nla_total_size(1) +
1222                 0;
1223 }
1224
1225 static int sit_fill_info(struct sk_buff *skb, const struct net_device *dev)
1226 {
1227         struct ip_tunnel *tunnel = netdev_priv(dev);
1228         struct ip_tunnel_parm *parm = &tunnel->parms;
1229
1230         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1231             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1232             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1233             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1234             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos))
1235                 goto nla_put_failure;
1236         return 0;
1237
1238 nla_put_failure:
1239         return -EMSGSIZE;
1240 }
1241
1242 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1243         .kind           = "sit",
1244         .maxtype        = IFLA_IPTUN_MAX,
1245         .priv_size      = sizeof(struct ip_tunnel),
1246         .get_size       = sit_get_size,
1247         .fill_info      = sit_fill_info,
1248 };
1249
1250 static struct xfrm_tunnel sit_handler __read_mostly = {
1251         .handler        =       ipip6_rcv,
1252         .err_handler    =       ipip6_err,
1253         .priority       =       1,
1254 };
1255
1256 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1257 {
1258         int prio;
1259
1260         for (prio = 1; prio < 4; prio++) {
1261                 int h;
1262                 for (h = 0; h < HASH_SIZE; h++) {
1263                         struct ip_tunnel *t;
1264
1265                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1266                         while (t != NULL) {
1267                                 unregister_netdevice_queue(t->dev, head);
1268                                 t = rtnl_dereference(t->next);
1269                         }
1270                 }
1271         }
1272 }
1273
1274 static int __net_init sit_init_net(struct net *net)
1275 {
1276         struct sit_net *sitn = net_generic(net, sit_net_id);
1277         struct ip_tunnel *t;
1278         int err;
1279
1280         sitn->tunnels[0] = sitn->tunnels_wc;
1281         sitn->tunnels[1] = sitn->tunnels_l;
1282         sitn->tunnels[2] = sitn->tunnels_r;
1283         sitn->tunnels[3] = sitn->tunnels_r_l;
1284
1285         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1286                                            ipip6_tunnel_setup);
1287         if (!sitn->fb_tunnel_dev) {
1288                 err = -ENOMEM;
1289                 goto err_alloc_dev;
1290         }
1291         dev_net_set(sitn->fb_tunnel_dev, net);
1292
1293         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1294         if (err)
1295                 goto err_dev_free;
1296
1297         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1298
1299         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1300                 goto err_reg_dev;
1301
1302         t = netdev_priv(sitn->fb_tunnel_dev);
1303
1304         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1305         return 0;
1306
1307 err_reg_dev:
1308         dev_put(sitn->fb_tunnel_dev);
1309 err_dev_free:
1310         ipip6_dev_free(sitn->fb_tunnel_dev);
1311 err_alloc_dev:
1312         return err;
1313 }
1314
1315 static void __net_exit sit_exit_net(struct net *net)
1316 {
1317         struct sit_net *sitn = net_generic(net, sit_net_id);
1318         LIST_HEAD(list);
1319
1320         rtnl_lock();
1321         sit_destroy_tunnels(sitn, &list);
1322         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1323         unregister_netdevice_many(&list);
1324         rtnl_unlock();
1325 }
1326
1327 static struct pernet_operations sit_net_ops = {
1328         .init = sit_init_net,
1329         .exit = sit_exit_net,
1330         .id   = &sit_net_id,
1331         .size = sizeof(struct sit_net),
1332 };
1333
1334 static void __exit sit_cleanup(void)
1335 {
1336         rtnl_link_unregister(&sit_link_ops);
1337         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1338
1339         unregister_pernet_device(&sit_net_ops);
1340         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1341 }
1342
1343 static int __init sit_init(void)
1344 {
1345         int err;
1346
1347         pr_info("IPv6 over IPv4 tunneling driver\n");
1348
1349         err = register_pernet_device(&sit_net_ops);
1350         if (err < 0)
1351                 return err;
1352         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1353         if (err < 0) {
1354                 pr_info("%s: can't add protocol\n", __func__);
1355                 goto xfrm_tunnel_failed;
1356         }
1357         err = rtnl_link_register(&sit_link_ops);
1358         if (err < 0)
1359                 goto rtnl_link_failed;
1360
1361 out:
1362         return err;
1363
1364 rtnl_link_failed:
1365         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1366 xfrm_tunnel_failed:
1367         unregister_pernet_device(&sit_net_ops);
1368         goto out;
1369 }
1370
1371 module_init(sit_init);
1372 module_exit(sit_cleanup);
1373 MODULE_LICENSE("GPL");
1374 MODULE_ALIAS_NETDEV("sit0");