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/ip_tunnels.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 bool log_ecn_error = true;
  69 module_param(log_ecn_error, bool, 0644);
  70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  71
  72 static int ipip6_tunnel_init(struct net_device *dev);
  73 static void ipip6_tunnel_setup(struct net_device *dev);
  74 static void ipip6_dev_free(struct net_device *dev);
  75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  76                       __be32 *v4dst);
  77 static struct rtnl_link_ops sit_link_ops __read_mostly;
  78
  79 static int sit_net_id __read_mostly;
  80 struct sit_net {
  81         struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
  82         struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
  83         struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
  84         struct ip_tunnel __rcu *tunnels_wc[1];
  85         struct ip_tunnel __rcu **tunnels[4];
  86
  87         struct net_device *fb_tunnel_dev;
  88 };
  89
  90 /*
  91  * Must be invoked with rcu_read_lock
  92  */
  93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  94                 struct net_device *dev, __be32 remote, __be32 local)
  95 {
  96         unsigned int h0 = HASH(remote);
  97         unsigned int h1 = HASH(local);
  98         struct ip_tunnel *t;
  99         struct sit_net *sitn = net_generic(net, sit_net_id);
 100
 101         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 102                 if (local == t->parms.iph.saddr &&
 103                     remote == t->parms.iph.daddr &&
 104                     (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
 105                     (t->dev->flags & IFF_UP))
 106                         return t;
 107         }
 108         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 109                 if (remote == t->parms.iph.daddr &&
 110                     (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
 111                     (t->dev->flags & IFF_UP))
 112                         return t;
 113         }
 114         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 115                 if (local == t->parms.iph.saddr &&
 116                     (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
 117                     (t->dev->flags & IFF_UP))
 118                         return t;
 119         }
 120         t = rcu_dereference(sitn->tunnels_wc[0]);
 121         if (t && (t->dev->flags & IFF_UP))
 122                 return t;
 123         return NULL;
 124 }
 125
 126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 127                 struct ip_tunnel_parm *parms)
 128 {
 129         __be32 remote = parms->iph.daddr;
 130         __be32 local = parms->iph.saddr;
 131         unsigned int h = 0;
 132         int prio = 0;
 133
 134         if (remote) {
 135                 prio |= 2;
 136                 h ^= HASH(remote);
 137         }
 138         if (local) {
 139                 prio |= 1;
 140                 h ^= HASH(local);
 141         }
 142         return &sitn->tunnels[prio][h];
 143 }
 144
 145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 146                 struct ip_tunnel *t)
 147 {
 148         return __ipip6_bucket(sitn, &t->parms);
 149 }
 150
 151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 152 {
 153         struct ip_tunnel __rcu **tp;
 154         struct ip_tunnel *iter;
 155
 156         for (tp = ipip6_bucket(sitn, t);
 157              (iter = rtnl_dereference(*tp)) != NULL;
 158              tp = &iter->next) {
 159                 if (t == iter) {
 160                         rcu_assign_pointer(*tp, t->next);
 161                         break;
 162                 }
 163         }
 164 }
 165
 166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 167 {
 168         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 169
 170         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 171         rcu_assign_pointer(*tp, t);
 172 }
 173
 174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 175 {
 176 #ifdef CONFIG_IPV6_SIT_6RD
 177         struct ip_tunnel *t = netdev_priv(dev);
 178
 179         if (dev == sitn->fb_tunnel_dev) {
 180                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 181                 t->ip6rd.relay_prefix = 0;
 182                 t->ip6rd.prefixlen = 16;
 183                 t->ip6rd.relay_prefixlen = 0;
 184         } else {
 185                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 186                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 187         }
 188 #endif
 189 }
 190
 191 static int ipip6_tunnel_create(struct net_device *dev)
 192 {
 193         struct ip_tunnel *t = netdev_priv(dev);
 194         struct net *net = dev_net(dev);
 195         struct sit_net *sitn = net_generic(net, sit_net_id);
 196         int err;
 197
 198         memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 199         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 200
 201         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 202                 dev->priv_flags |= IFF_ISATAP;
 203
 204         err = register_netdevice(dev);
 205         if (err < 0)
 206                 goto out;
 207
 208         ipip6_tunnel_clone_6rd(dev, sitn);
 209
 210         dev->rtnl_link_ops = &sit_link_ops;
 211
 212         ipip6_tunnel_link(sitn, t);
 213         return 0;
 214
 215 out:
 216         return err;
 217 }
 218
 219 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 220                 struct ip_tunnel_parm *parms, int create)
 221 {
 222         __be32 remote = parms->iph.daddr;
 223         __be32 local = parms->iph.saddr;
 224         struct ip_tunnel *t, *nt;
 225         struct ip_tunnel __rcu **tp;
 226         struct net_device *dev;
 227         char name[IFNAMSIZ];
 228         struct sit_net *sitn = net_generic(net, sit_net_id);
 229
 230         for (tp = __ipip6_bucket(sitn, parms);
 231             (t = rtnl_dereference(*tp)) != NULL;
 232              tp = &t->next) {
 233                 if (local == t->parms.iph.saddr &&
 234                     remote == t->parms.iph.daddr &&
 235                     parms->link == t->parms.link) {
 236                         if (create)
 237                                 return NULL;
 238                         else
 239                                 return t;
 240                 }
 241         }
 242         if (!create)
 243                 goto failed;
 244
 245         if (parms->name[0]) {
 246                 if (!dev_valid_name(parms->name))
 247                         goto failed;
 248                 strlcpy(name, parms->name, IFNAMSIZ);
 249         } else {
 250                 strcpy(name, "sit%d");
 251         }
 252         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 253                            ipip6_tunnel_setup);
 254         if (!dev)
 255                 return NULL;
 256
 257         dev_net_set(dev, net);
 258
 259         nt = netdev_priv(dev);
 260
 261         nt->parms = *parms;
 262         if (ipip6_tunnel_create(dev) < 0)
 263                 goto failed_free;
 264
 265         return nt;
 266
 267 failed_free:
 268         ipip6_dev_free(dev);
 269 failed:
 270         return NULL;
 271 }
 272
 273 #define for_each_prl_rcu(start)                 \
 274         for (prl = rcu_dereference(start);      \
 275              prl;                               \
 276              prl = rcu_dereference(prl->next))
 277
 278 static struct ip_tunnel_prl_entry *
 279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 280 {
 281         struct ip_tunnel_prl_entry *prl;
 282
 283         for_each_prl_rcu(t->prl)
 284                 if (prl->addr == addr)
 285                         break;
 286         return prl;
 287
 288 }
 289
 290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 291                                 struct ip_tunnel_prl __user *a)
 292 {
 293         struct ip_tunnel_prl kprl, *kp;
 294         struct ip_tunnel_prl_entry *prl;
 295         unsigned int cmax, c = 0, ca, len;
 296         int ret = 0;
 297
 298         if (copy_from_user(&kprl, a, sizeof(kprl)))
 299                 return -EFAULT;
 300         cmax = kprl.datalen / sizeof(kprl);
 301         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 302                 cmax = 1;
 303
 304         /* For simple GET or for root users,
 305          * we try harder to allocate.
 306          */
 307         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 308                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 309                 NULL;
 310
 311         rcu_read_lock();
 312
 313         ca = t->prl_count < cmax ? t->prl_count : cmax;
 314
 315         if (!kp) {
 316                 /* We don't try hard to allocate much memory for
 317                  * non-root users.
 318                  * For root users, retry allocating enough memory for
 319                  * the answer.
 320                  */
 321                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 322                 if (!kp) {
 323                         ret = -ENOMEM;
 324                         goto out;
 325                 }
 326         }
 327
 328         c = 0;
 329         for_each_prl_rcu(t->prl) {
 330                 if (c >= cmax)
 331                         break;
 332                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 333                         continue;
 334                 kp[c].addr = prl->addr;
 335                 kp[c].flags = prl->flags;
 336                 c++;
 337                 if (kprl.addr != htonl(INADDR_ANY))
 338                         break;
 339         }
 340 out:
 341         rcu_read_unlock();
 342
 343         len = sizeof(*kp) * c;
 344         ret = 0;
 345         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 346                 ret = -EFAULT;
 347
 348         kfree(kp);
 349
 350         return ret;
 351 }
 352
 353 static int
 354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 355 {
 356         struct ip_tunnel_prl_entry *p;
 357         int err = 0;
 358
 359         if (a->addr == htonl(INADDR_ANY))
 360                 return -EINVAL;
 361
 362         ASSERT_RTNL();
 363
 364         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 365                 if (p->addr == a->addr) {
 366                         if (chg) {
 367                                 p->flags = a->flags;
 368                                 goto out;
 369                         }
 370                         err = -EEXIST;
 371                         goto out;
 372                 }
 373         }
 374
 375         if (chg) {
 376                 err = -ENXIO;
 377                 goto out;
 378         }
 379
 380         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 381         if (!p) {
 382                 err = -ENOBUFS;
 383                 goto out;
 384         }
 385
 386         p->next = t->prl;
 387         p->addr = a->addr;
 388         p->flags = a->flags;
 389         t->prl_count++;
 390         rcu_assign_pointer(t->prl, p);
 391 out:
 392         return err;
 393 }
 394
 395 static void prl_list_destroy_rcu(struct rcu_head *head)
 396 {
 397         struct ip_tunnel_prl_entry *p, *n;
 398
 399         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 400         do {
 401                 n = rcu_dereference_protected(p->next, 1);
 402                 kfree(p);
 403                 p = n;
 404         } while (p);
 405 }
 406
 407 static int
 408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 409 {
 410         struct ip_tunnel_prl_entry *x;
 411         struct ip_tunnel_prl_entry __rcu **p;
 412         int err = 0;
 413
 414         ASSERT_RTNL();
 415
 416         if (a && a->addr != htonl(INADDR_ANY)) {
 417                 for (p = &t->prl;
 418                      (x = rtnl_dereference(*p)) != NULL;
 419                      p = &x->next) {
 420                         if (x->addr == a->addr) {
 421                                 *p = x->next;
 422                                 kfree_rcu(x, rcu_head);
 423                                 t->prl_count--;
 424                                 goto out;
 425                         }
 426                 }
 427                 err = -ENXIO;
 428         } else {
 429                 x = rtnl_dereference(t->prl);
 430                 if (x) {
 431                         t->prl_count = 0;
 432                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 433                         t->prl = NULL;
 434                 }
 435         }
 436 out:
 437         return err;
 438 }
 439
 440 static int
 441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 442 {
 443         struct ip_tunnel_prl_entry *p;
 444         int ok = 1;
 445
 446         rcu_read_lock();
 447         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 448         if (p) {
 449                 if (p->flags & PRL_DEFAULT)
 450                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 451                 else
 452                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 453         } else {
 454                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 455
 456                 if (ipv6_addr_is_isatap(addr6) &&
 457                     (addr6->s6_addr32[3] == iph->saddr) &&
 458                     ipv6_chk_prefix(addr6, t->dev))
 459                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 460                 else
 461                         ok = 0;
 462         }
 463         rcu_read_unlock();
 464         return ok;
 465 }
 466
 467 static void ipip6_tunnel_uninit(struct net_device *dev)
 468 {
 469         struct ip_tunnel *tunnel = netdev_priv(dev);
 470         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 471
 472         if (dev == sitn->fb_tunnel_dev) {
 473                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 474         } else {
 475                 ipip6_tunnel_unlink(sitn, tunnel);
 476                 ipip6_tunnel_del_prl(tunnel, NULL);
 477         }
 478         dst_cache_reset(&tunnel->dst_cache);
 479         dev_put(dev);
 480 }
 481
 482 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
 483  * if sufficient data bytes are available
 484  */
 485 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
 486 {
 487         int ihl = ((const struct iphdr *)skb->data)->ihl*4;
 488         struct rt6_info *rt;
 489         struct sk_buff *skb2;
 490
 491         if (!pskb_may_pull(skb, ihl + sizeof(struct ipv6hdr) + 8))
 492                 return 1;
 493
 494         skb2 = skb_clone(skb, GFP_ATOMIC);
 495
 496         if (!skb2)
 497                 return 1;
 498
 499         skb_dst_drop(skb2);
 500         skb_pull(skb2, ihl);
 501         skb_reset_network_header(skb2);
 502
 503         rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
 504
 505         if (rt && rt->dst.dev)
 506                 skb2->dev = rt->dst.dev;
 507
 508         icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
 509
 510         if (rt)
 511                 ip6_rt_put(rt);
 512
 513         kfree_skb(skb2);
 514
 515         return 0;
 516 }
 517
 518 static int ipip6_err(struct sk_buff *skb, u32 info)
 519 {
 520         const struct iphdr *iph = (const struct iphdr *)skb->data;
 521         const int type = icmp_hdr(skb)->type;
 522         const int code = icmp_hdr(skb)->code;
 523         struct ip_tunnel *t;
 524         int err;
 525
 526         switch (type) {
 527         default:
 528         case ICMP_PARAMETERPROB:
 529                 return 0;
 530
 531         case ICMP_DEST_UNREACH:
 532                 switch (code) {
 533                 case ICMP_SR_FAILED:
 534                         /* Impossible event. */
 535                         return 0;
 536                 default:
 537                         /* All others are translated to HOST_UNREACH.
 538                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 539                            I believe they are just ether pollution. --ANK
 540                          */
 541                         break;
 542                 }
 543                 break;
 544         case ICMP_TIME_EXCEEDED:
 545                 if (code != ICMP_EXC_TTL)
 546                         return 0;
 547                 break;
 548         case ICMP_REDIRECT:
 549                 break;
 550         }
 551
 552         err = -ENOENT;
 553
 554         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 555                                 skb->dev,
 556                                 iph->daddr,
 557                                 iph->saddr);
 558         if (!t)
 559                 goto out;
 560
 561         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 562                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 563                                  t->parms.link, 0, iph->protocol, 0);
 564                 err = 0;
 565                 goto out;
 566         }
 567         if (type == ICMP_REDIRECT) {
 568                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
 569                               iph->protocol, 0);
 570                 err = 0;
 571                 goto out;
 572         }
 573
 574         if (t->parms.iph.daddr == 0)
 575                 goto out;
 576
 577         err = 0;
 578         if (__in6_dev_get(skb->dev) && !ipip6_err_gen_icmpv6_unreach(skb))
 579                 goto out;
 580
 581         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 582                 goto out;
 583
 584         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 585                 t->err_count++;
 586         else
 587                 t->err_count = 1;
 588         t->err_time = jiffies;
 589 out:
 590         return err;
 591 }
 592
 593 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 594                                   const struct in6_addr *v6addr)
 595 {
 596         __be32 v4embed = 0;
 597         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 598                 return true;
 599         return false;
 600 }
 601
 602 /* Checks if an address matches an address on the tunnel interface.
 603  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 604  * Long story:
 605  * This function is called after we considered the packet as spoofed
 606  * in is_spoofed_6rd.
 607  * We may have a router that is doing NAT for proto 41 packets
 608  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 609  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 610  * function will return true, dropping the packet.
 611  * But, we can still check if is spoofed against the IP
 612  * addresses associated with the interface.
 613  */
 614 static bool only_dnatted(const struct ip_tunnel *tunnel,
 615         const struct in6_addr *v6dst)
 616 {
 617         int prefix_len;
 618
 619 #ifdef CONFIG_IPV6_SIT_6RD
 620         prefix_len = tunnel->ip6rd.prefixlen + 32
 621                 - tunnel->ip6rd.relay_prefixlen;
 622 #else
 623         prefix_len = 48;
 624 #endif
 625         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 626 }
 627
 628 /* Returns true if a packet is spoofed */
 629 static bool packet_is_spoofed(struct sk_buff *skb,
 630                               const struct iphdr *iph,
 631                               struct ip_tunnel *tunnel)
 632 {
 633         const struct ipv6hdr *ipv6h;
 634
 635         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 636                 if (!isatap_chksrc(skb, iph, tunnel))
 637                         return true;
 638
 639                 return false;
 640         }
 641
 642         if (tunnel->dev->flags & IFF_POINTOPOINT)
 643                 return false;
 644
 645         ipv6h = ipv6_hdr(skb);
 646
 647         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 648                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 649                                      &iph->saddr, &ipv6h->saddr,
 650                                      &iph->daddr, &ipv6h->daddr);
 651                 return true;
 652         }
 653
 654         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 655                 return false;
 656
 657         if (only_dnatted(tunnel, &ipv6h->daddr))
 658                 return false;
 659
 660         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 661                              &iph->saddr, &ipv6h->saddr,
 662                              &iph->daddr, &ipv6h->daddr);
 663         return true;
 664 }
 665
 666 static int ipip6_rcv(struct sk_buff *skb)
 667 {
 668         const struct iphdr *iph = ip_hdr(skb);
 669         struct ip_tunnel *tunnel;
 670         int err;
 671
 672         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 673                                      iph->saddr, iph->daddr);
 674         if (tunnel) {
 675                 struct pcpu_sw_netstats *tstats;
 676
 677                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 678                     tunnel->parms.iph.protocol != 0)
 679                         goto out;
 680
 681                 skb->mac_header = skb->network_header;
 682                 skb_reset_network_header(skb);
 683                 IPCB(skb)->flags = 0;
 684                 skb->dev = tunnel->dev;
 685
 686                 if (packet_is_spoofed(skb, iph, tunnel)) {
 687                         tunnel->dev->stats.rx_errors++;
 688                         goto out;
 689                 }
 690
 691                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6)))
 692                         goto out;
 693
 694                 /* skb can be uncloned in iptunnel_pull_header, so
 695                  * old iph is no longer valid
 696                  */
 697                 iph = (const struct iphdr *)skb_mac_header(skb);
 698                 err = IP_ECN_decapsulate(iph, skb);
 699                 if (unlikely(err)) {
 700                         if (log_ecn_error)
 701                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 702                                                      &iph->saddr, iph->tos);
 703                         if (err > 1) {
 704                                 ++tunnel->dev->stats.rx_frame_errors;
 705                                 ++tunnel->dev->stats.rx_errors;
 706                                 goto out;
 707                         }
 708                 }
 709
 710                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 711                 u64_stats_update_begin(&tstats->syncp);
 712                 tstats->rx_packets++;
 713                 tstats->rx_bytes += skb->len;
 714                 u64_stats_update_end(&tstats->syncp);
 715
 716                 netif_rx(skb);
 717
 718                 return 0;
 719         }
 720
 721         /* no tunnel matched,  let upstream know, ipsec may handle it */
 722         return 1;
 723 out:
 724         kfree_skb(skb);
 725         return 0;
 726 }
 727
 728 static const struct tnl_ptk_info tpi = {
 729         /* no tunnel info required for ipip. */
 730         .proto = htons(ETH_P_IP),
 731 };
 732
 733 static int ipip_rcv(struct sk_buff *skb)
 734 {
 735         const struct iphdr *iph;
 736         struct ip_tunnel *tunnel;
 737
 738         iph = ip_hdr(skb);
 739         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 740                                      iph->saddr, iph->daddr);
 741         if (tunnel) {
 742                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 743                     tunnel->parms.iph.protocol != 0)
 744                         goto drop;
 745
 746                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 747                         goto drop;
 748                 if (iptunnel_pull_header(skb, 0, tpi.proto))
 749                         goto drop;
 750                 return ip_tunnel_rcv(tunnel, skb, &tpi, NULL, log_ecn_error);
 751         }
 752
 753         return 1;
 754
 755 drop:
 756         kfree_skb(skb);
 757         return 0;
 758 }
 759
 760 /*
 761  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 762  * stores the embedded IPv4 address in v4dst and returns true.
 763  */
 764 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 765                       __be32 *v4dst)
 766 {
 767 #ifdef CONFIG_IPV6_SIT_6RD
 768         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 769                               tunnel->ip6rd.prefixlen)) {
 770                 unsigned int pbw0, pbi0;
 771                 int pbi1;
 772                 u32 d;
 773
 774                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 775                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 776
 777                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 778                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 779                     tunnel->ip6rd.relay_prefixlen : 0;
 780
 781                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 782                 if (pbi1 > 0)
 783                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 784                              (32 - pbi1);
 785
 786                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 787                 return true;
 788         }
 789 #else
 790         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 791                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 792                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 793                 return true;
 794         }
 795 #endif
 796         return false;
 797 }
 798
 799 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 800                              const struct in6_addr *v6dst)
 801 {
 802         __be32 dst = 0;
 803         check_6rd(tunnel, v6dst, &dst);
 804         return dst;
 805 }
 806
 807 /*
 808  *      This function assumes it is being called from dev_queue_xmit()
 809  *      and that skb is filled properly by that function.
 810  */
 811
 812 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 813                                      struct net_device *dev)
 814 {
 815         struct ip_tunnel *tunnel = netdev_priv(dev);
 816         const struct iphdr  *tiph = &tunnel->parms.iph;
 817         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 818         u8     tos = tunnel->parms.iph.tos;
 819         __be16 df = tiph->frag_off;
 820         struct rtable *rt;              /* Route to the other host */
 821         struct net_device *tdev;        /* Device to other host */
 822         unsigned int max_headroom;      /* The extra header space needed */
 823         __be32 dst = tiph->daddr;
 824         struct flowi4 fl4;
 825         int    mtu;
 826         const struct in6_addr *addr6;
 827         int addr_type;
 828         u8 ttl;
 829         int err;
 830         u8 protocol = IPPROTO_IPV6;
 831         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 832
 833         if (skb->protocol != htons(ETH_P_IPV6))
 834                 goto tx_error;
 835
 836         if (tos == 1)
 837                 tos = ipv6_get_dsfield(iph6);
 838
 839         /* ISATAP (RFC4214) - must come before 6to4 */
 840         if (dev->priv_flags & IFF_ISATAP) {
 841                 struct neighbour *neigh = NULL;
 842                 bool do_tx_error = false;
 843
 844                 if (skb_dst(skb))
 845                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 846
 847                 if (!neigh) {
 848                         net_dbg_ratelimited("nexthop == NULL\n");
 849                         goto tx_error;
 850                 }
 851
 852                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 853                 addr_type = ipv6_addr_type(addr6);
 854
 855                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 856                      ipv6_addr_is_isatap(addr6))
 857                         dst = addr6->s6_addr32[3];
 858                 else
 859                         do_tx_error = true;
 860
 861                 neigh_release(neigh);
 862                 if (do_tx_error)
 863                         goto tx_error;
 864         }
 865
 866         if (!dst)
 867                 dst = try_6rd(tunnel, &iph6->daddr);
 868
 869         if (!dst) {
 870                 struct neighbour *neigh = NULL;
 871                 bool do_tx_error = false;
 872
 873                 if (skb_dst(skb))
 874                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 875
 876                 if (!neigh) {
 877                         net_dbg_ratelimited("nexthop == NULL\n");
 878                         goto tx_error;
 879                 }
 880
 881                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 882                 addr_type = ipv6_addr_type(addr6);
 883
 884                 if (addr_type == IPV6_ADDR_ANY) {
 885                         addr6 = &ipv6_hdr(skb)->daddr;
 886                         addr_type = ipv6_addr_type(addr6);
 887                 }
 888
 889                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 890                         dst = addr6->s6_addr32[3];
 891                 else
 892                         do_tx_error = true;
 893
 894                 neigh_release(neigh);
 895                 if (do_tx_error)
 896                         goto tx_error;
 897         }
 898
 899         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 900                                    dst, tiph->saddr,
 901                                    0, 0,
 902                                    IPPROTO_IPV6, RT_TOS(tos),
 903                                    tunnel->parms.link);
 904         if (IS_ERR(rt)) {
 905                 dev->stats.tx_carrier_errors++;
 906                 goto tx_error_icmp;
 907         }
 908         if (rt->rt_type != RTN_UNICAST) {
 909                 ip_rt_put(rt);
 910                 dev->stats.tx_carrier_errors++;
 911                 goto tx_error_icmp;
 912         }
 913         tdev = rt->dst.dev;
 914
 915         if (tdev == dev) {
 916                 ip_rt_put(rt);
 917                 dev->stats.collisions++;
 918                 goto tx_error;
 919         }
 920
 921         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
 922         if (IS_ERR(skb)) {
 923                 ip_rt_put(rt);
 924                 goto out;
 925         }
 926
 927         if (df) {
 928                 mtu = dst_mtu(&rt->dst) - t_hlen;
 929
 930                 if (mtu < 68) {
 931                         dev->stats.collisions++;
 932                         ip_rt_put(rt);
 933                         goto tx_error;
 934                 }
 935
 936                 if (mtu < IPV6_MIN_MTU) {
 937                         mtu = IPV6_MIN_MTU;
 938                         df = 0;
 939                 }
 940
 941                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 942                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 943
 944                 if (skb->len > mtu && !skb_is_gso(skb)) {
 945                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 946                         ip_rt_put(rt);
 947                         goto tx_error;
 948                 }
 949         }
 950
 951         if (tunnel->err_count > 0) {
 952                 if (time_before(jiffies,
 953                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 954                         tunnel->err_count--;
 955                         dst_link_failure(skb);
 956                 } else
 957                         tunnel->err_count = 0;
 958         }
 959
 960         /*
 961          * Okay, now see if we can stuff it in the buffer as-is.
 962          */
 963         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 964
 965         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 966             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 967                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 968                 if (!new_skb) {
 969                         ip_rt_put(rt);
 970                         dev->stats.tx_dropped++;
 971                         kfree_skb(skb);
 972                         return NETDEV_TX_OK;
 973                 }
 974                 if (skb->sk)
 975                         skb_set_owner_w(new_skb, skb->sk);
 976                 dev_kfree_skb(skb);
 977                 skb = new_skb;
 978                 iph6 = ipv6_hdr(skb);
 979         }
 980         ttl = tiph->ttl;
 981         if (ttl == 0)
 982                 ttl = iph6->hop_limit;
 983         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 984
 985         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 986                 ip_rt_put(rt);
 987                 goto tx_error;
 988         }
 989
 990         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 991
 992         err = iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr,
 993                             protocol, tos, ttl, df,
 994                             !net_eq(tunnel->net, dev_net(dev)));
 995         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 996         return NETDEV_TX_OK;
 997
 998 tx_error_icmp:
 999         dst_link_failure(skb);
1000 tx_error:
1001         kfree_skb(skb);
1002 out:
1003         dev->stats.tx_errors++;
1004         return NETDEV_TX_OK;
1005 }
1006
1007 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
1008 {
1009         struct ip_tunnel *tunnel = netdev_priv(dev);
1010         const struct iphdr  *tiph = &tunnel->parms.iph;
1011
1012         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
1013         if (IS_ERR(skb))
1014                 goto out;
1015
1016         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1017
1018         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1019         return NETDEV_TX_OK;
1020 out:
1021         dev->stats.tx_errors++;
1022         return NETDEV_TX_OK;
1023 }
1024
1025 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1026                                    struct net_device *dev)
1027 {
1028         switch (skb->protocol) {
1029         case htons(ETH_P_IP):
1030                 ipip_tunnel_xmit(skb, dev);
1031                 break;
1032         case htons(ETH_P_IPV6):
1033                 ipip6_tunnel_xmit(skb, dev);
1034                 break;
1035         default:
1036                 goto tx_err;
1037         }
1038
1039         return NETDEV_TX_OK;
1040
1041 tx_err:
1042         dev->stats.tx_errors++;
1043         kfree_skb(skb);
1044         return NETDEV_TX_OK;
1045
1046 }
1047
1048 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1049 {
1050         struct net_device *tdev = NULL;
1051         struct ip_tunnel *tunnel;
1052         const struct iphdr *iph;
1053         struct flowi4 fl4;
1054
1055         tunnel = netdev_priv(dev);
1056         iph = &tunnel->parms.iph;
1057
1058         if (iph->daddr) {
1059                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1060                                                           NULL,
1061                                                           iph->daddr, iph->saddr,
1062                                                           0, 0,
1063                                                           IPPROTO_IPV6,
1064                                                           RT_TOS(iph->tos),
1065                                                           tunnel->parms.link);
1066
1067                 if (!IS_ERR(rt)) {
1068                         tdev = rt->dst.dev;
1069                         ip_rt_put(rt);
1070                 }
1071                 dev->flags |= IFF_POINTOPOINT;
1072         }
1073
1074         if (!tdev && tunnel->parms.link)
1075                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1076
1077         if (tdev && !netif_is_l3_master(tdev)) {
1078                 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1079
1080                 dev->mtu = tdev->mtu - t_hlen;
1081                 if (dev->mtu < IPV6_MIN_MTU)
1082                         dev->mtu = IPV6_MIN_MTU;
1083         }
1084 }
1085
1086 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1087 {
1088         struct net *net = t->net;
1089         struct sit_net *sitn = net_generic(net, sit_net_id);
1090
1091         ipip6_tunnel_unlink(sitn, t);
1092         synchronize_net();
1093         t->parms.iph.saddr = p->iph.saddr;
1094         t->parms.iph.daddr = p->iph.daddr;
1095         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1096         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1097         ipip6_tunnel_link(sitn, t);
1098         t->parms.iph.ttl = p->iph.ttl;
1099         t->parms.iph.tos = p->iph.tos;
1100         t->parms.iph.frag_off = p->iph.frag_off;
1101         if (t->parms.link != p->link) {
1102                 t->parms.link = p->link;
1103                 ipip6_tunnel_bind_dev(t->dev);
1104         }
1105         dst_cache_reset(&t->dst_cache);
1106         netdev_state_change(t->dev);
1107 }
1108
1109 #ifdef CONFIG_IPV6_SIT_6RD
1110 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1111                                    struct ip_tunnel_6rd *ip6rd)
1112 {
1113         struct in6_addr prefix;
1114         __be32 relay_prefix;
1115
1116         if (ip6rd->relay_prefixlen > 32 ||
1117             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1118                 return -EINVAL;
1119
1120         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1121         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1122                 return -EINVAL;
1123         if (ip6rd->relay_prefixlen)
1124                 relay_prefix = ip6rd->relay_prefix &
1125                                htonl(0xffffffffUL <<
1126                                      (32 - ip6rd->relay_prefixlen));
1127         else
1128                 relay_prefix = 0;
1129         if (relay_prefix != ip6rd->relay_prefix)
1130                 return -EINVAL;
1131
1132         t->ip6rd.prefix = prefix;
1133         t->ip6rd.relay_prefix = relay_prefix;
1134         t->ip6rd.prefixlen = ip6rd->prefixlen;
1135         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1136         dst_cache_reset(&t->dst_cache);
1137         netdev_state_change(t->dev);
1138         return 0;
1139 }
1140 #endif
1141
1142 static int
1143 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1144 {
1145         int err = 0;
1146         struct ip_tunnel_parm p;
1147         struct ip_tunnel_prl prl;
1148         struct ip_tunnel *t = netdev_priv(dev);
1149         struct net *net = t->net;
1150         struct sit_net *sitn = net_generic(net, sit_net_id);
1151 #ifdef CONFIG_IPV6_SIT_6RD
1152         struct ip_tunnel_6rd ip6rd;
1153 #endif
1154
1155         switch (cmd) {
1156         case SIOCGETTUNNEL:
1157 #ifdef CONFIG_IPV6_SIT_6RD
1158         case SIOCGET6RD:
1159 #endif
1160                 if (dev == sitn->fb_tunnel_dev) {
1161                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1162                                 err = -EFAULT;
1163                                 break;
1164                         }
1165                         t = ipip6_tunnel_locate(net, &p, 0);
1166                         if (!t)
1167                                 t = netdev_priv(dev);
1168                 }
1169
1170                 err = -EFAULT;
1171                 if (cmd == SIOCGETTUNNEL) {
1172                         memcpy(&p, &t->parms, sizeof(p));
1173                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1174                                          sizeof(p)))
1175                                 goto done;
1176 #ifdef CONFIG_IPV6_SIT_6RD
1177                 } else {
1178                         ip6rd.prefix = t->ip6rd.prefix;
1179                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1180                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1181                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1182                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1183                                          sizeof(ip6rd)))
1184                                 goto done;
1185 #endif
1186                 }
1187                 err = 0;
1188                 break;
1189
1190         case SIOCADDTUNNEL:
1191         case SIOCCHGTUNNEL:
1192                 err = -EPERM;
1193                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1194                         goto done;
1195
1196                 err = -EFAULT;
1197                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1198                         goto done;
1199
1200                 err = -EINVAL;
1201                 if (p.iph.protocol != IPPROTO_IPV6 &&
1202                     p.iph.protocol != IPPROTO_IPIP &&
1203                     p.iph.protocol != 0)
1204                         goto done;
1205                 if (p.iph.version != 4 ||
1206                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1207                         goto done;
1208                 if (p.iph.ttl)
1209                         p.iph.frag_off |= htons(IP_DF);
1210
1211                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1212
1213                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1214                         if (t) {
1215                                 if (t->dev != dev) {
1216                                         err = -EEXIST;
1217                                         break;
1218                                 }
1219                         } else {
1220                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1221                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1222                                         err = -EINVAL;
1223                                         break;
1224                                 }
1225                                 t = netdev_priv(dev);
1226                         }
1227
1228                         ipip6_tunnel_update(t, &p);
1229                 }
1230
1231                 if (t) {
1232                         err = 0;
1233                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1234                                 err = -EFAULT;
1235                 } else
1236                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1237                 break;
1238
1239         case SIOCDELTUNNEL:
1240                 err = -EPERM;
1241                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1242                         goto done;
1243
1244                 if (dev == sitn->fb_tunnel_dev) {
1245                         err = -EFAULT;
1246                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1247                                 goto done;
1248                         err = -ENOENT;
1249                         t = ipip6_tunnel_locate(net, &p, 0);
1250                         if (!t)
1251                                 goto done;
1252                         err = -EPERM;
1253                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1254                                 goto done;
1255                         dev = t->dev;
1256                 }
1257                 unregister_netdevice(dev);
1258                 err = 0;
1259                 break;
1260
1261         case SIOCGETPRL:
1262                 err = -EINVAL;
1263                 if (dev == sitn->fb_tunnel_dev)
1264                         goto done;
1265                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1266                 break;
1267
1268         case SIOCADDPRL:
1269         case SIOCDELPRL:
1270         case SIOCCHGPRL:
1271                 err = -EPERM;
1272                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1273                         goto done;
1274                 err = -EINVAL;
1275                 if (dev == sitn->fb_tunnel_dev)
1276                         goto done;
1277                 err = -EFAULT;
1278                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1279                         goto done;
1280
1281                 switch (cmd) {
1282                 case SIOCDELPRL:
1283                         err = ipip6_tunnel_del_prl(t, &prl);
1284                         break;
1285                 case SIOCADDPRL:
1286                 case SIOCCHGPRL:
1287                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1288                         break;
1289                 }
1290                 dst_cache_reset(&t->dst_cache);
1291                 netdev_state_change(dev);
1292                 break;
1293
1294 #ifdef CONFIG_IPV6_SIT_6RD
1295         case SIOCADD6RD:
1296         case SIOCCHG6RD:
1297         case SIOCDEL6RD:
1298                 err = -EPERM;
1299                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1300                         goto done;
1301
1302                 err = -EFAULT;
1303                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1304                                    sizeof(ip6rd)))
1305                         goto done;
1306
1307                 if (cmd != SIOCDEL6RD) {
1308                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1309                         if (err < 0)
1310                                 goto done;
1311                 } else
1312                         ipip6_tunnel_clone_6rd(dev, sitn);
1313
1314                 err = 0;
1315                 break;
1316 #endif
1317
1318         default:
1319                 err = -EINVAL;
1320         }
1321
1322 done:
1323         return err;
1324 }
1325
1326 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1327 {
1328         struct ip_tunnel *tunnel = netdev_priv(dev);
1329         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1330
1331         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - t_hlen)
1332                 return -EINVAL;
1333         dev->mtu = new_mtu;
1334         return 0;
1335 }
1336
1337 static const struct net_device_ops ipip6_netdev_ops = {
1338         .ndo_init       = ipip6_tunnel_init,
1339         .ndo_uninit     = ipip6_tunnel_uninit,
1340         .ndo_start_xmit = sit_tunnel_xmit,
1341         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1342         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1343         .ndo_get_stats64 = ip_tunnel_get_stats64,
1344         .ndo_get_iflink = ip_tunnel_get_iflink,
1345 };
1346
1347 static void ipip6_dev_free(struct net_device *dev)
1348 {
1349         struct ip_tunnel *tunnel = netdev_priv(dev);
1350
1351         dst_cache_destroy(&tunnel->dst_cache);
1352         free_percpu(dev->tstats);
1353         free_netdev(dev);
1354 }
1355
1356 #define SIT_FEATURES (NETIF_F_SG           | \
1357                       NETIF_F_FRAGLIST     | \
1358                       NETIF_F_HIGHDMA      | \
1359                       NETIF_F_GSO_SOFTWARE | \
1360                       NETIF_F_HW_CSUM)
1361
1362 static void ipip6_tunnel_setup(struct net_device *dev)
1363 {
1364         struct ip_tunnel *tunnel = netdev_priv(dev);
1365         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1366
1367         dev->netdev_ops         = &ipip6_netdev_ops;
1368         dev->destructor         = ipip6_dev_free;
1369
1370         dev->type               = ARPHRD_SIT;
1371         dev->mtu                = ETH_DATA_LEN - t_hlen;
1372         dev->flags              = IFF_NOARP;
1373         netif_keep_dst(dev);
1374         dev->addr_len           = 4;
1375         dev->features           |= NETIF_F_LLTX;
1376         dev->features           |= SIT_FEATURES;
1377         dev->hw_features        |= SIT_FEATURES;
1378 }
1379
1380 static int ipip6_tunnel_init(struct net_device *dev)
1381 {
1382         struct ip_tunnel *tunnel = netdev_priv(dev);
1383         int err;
1384
1385         tunnel->dev = dev;
1386         tunnel->net = dev_net(dev);
1387         strcpy(tunnel->parms.name, dev->name);
1388
1389         ipip6_tunnel_bind_dev(dev);
1390         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1391         if (!dev->tstats)
1392                 return -ENOMEM;
1393
1394         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1395         if (err) {
1396                 free_percpu(dev->tstats);
1397                 dev->tstats = NULL;
1398                 return err;
1399         }
1400         dev_hold(dev);
1401         return 0;
1402 }
1403
1404 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1405 {
1406         struct ip_tunnel *tunnel = netdev_priv(dev);
1407         struct iphdr *iph = &tunnel->parms.iph;
1408         struct net *net = dev_net(dev);
1409         struct sit_net *sitn = net_generic(net, sit_net_id);
1410
1411         iph->version            = 4;
1412         iph->protocol           = IPPROTO_IPV6;
1413         iph->ihl                = 5;
1414         iph->ttl                = 64;
1415
1416         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1417 }
1418
1419 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1420 {
1421         u8 proto;
1422
1423         if (!data || !data[IFLA_IPTUN_PROTO])
1424                 return 0;
1425
1426         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1427         if (proto != IPPROTO_IPV6 &&
1428             proto != IPPROTO_IPIP &&
1429             proto != 0)
1430                 return -EINVAL;
1431
1432         return 0;
1433 }
1434
1435 static void ipip6_netlink_parms(struct nlattr *data[],
1436                                 struct ip_tunnel_parm *parms)
1437 {
1438         memset(parms, 0, sizeof(*parms));
1439
1440         parms->iph.version = 4;
1441         parms->iph.protocol = IPPROTO_IPV6;
1442         parms->iph.ihl = 5;
1443         parms->iph.ttl = 64;
1444
1445         if (!data)
1446                 return;
1447
1448         if (data[IFLA_IPTUN_LINK])
1449                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1450
1451         if (data[IFLA_IPTUN_LOCAL])
1452                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1453
1454         if (data[IFLA_IPTUN_REMOTE])
1455                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1456
1457         if (data[IFLA_IPTUN_TTL]) {
1458                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1459                 if (parms->iph.ttl)
1460                         parms->iph.frag_off = htons(IP_DF);
1461         }
1462
1463         if (data[IFLA_IPTUN_TOS])
1464                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1465
1466         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1467                 parms->iph.frag_off = htons(IP_DF);
1468
1469         if (data[IFLA_IPTUN_FLAGS])
1470                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1471
1472         if (data[IFLA_IPTUN_PROTO])
1473                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1474
1475 }
1476
1477 /* This function returns true when ENCAP attributes are present in the nl msg */
1478 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1479                                       struct ip_tunnel_encap *ipencap)
1480 {
1481         bool ret = false;
1482
1483         memset(ipencap, 0, sizeof(*ipencap));
1484
1485         if (!data)
1486                 return ret;
1487
1488         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1489                 ret = true;
1490                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1491         }
1492
1493         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1494                 ret = true;
1495                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1496         }
1497
1498         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1499                 ret = true;
1500                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1501         }
1502
1503         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1504                 ret = true;
1505                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1506         }
1507
1508         return ret;
1509 }
1510
1511 #ifdef CONFIG_IPV6_SIT_6RD
1512 /* This function returns true when 6RD attributes are present in the nl msg */
1513 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1514                                     struct ip_tunnel_6rd *ip6rd)
1515 {
1516         bool ret = false;
1517         memset(ip6rd, 0, sizeof(*ip6rd));
1518
1519         if (!data)
1520                 return ret;
1521
1522         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1523                 ret = true;
1524                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1525         }
1526
1527         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1528                 ret = true;
1529                 ip6rd->relay_prefix =
1530                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1531         }
1532
1533         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1534                 ret = true;
1535                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1536         }
1537
1538         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1539                 ret = true;
1540                 ip6rd->relay_prefixlen =
1541                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1542         }
1543
1544         return ret;
1545 }
1546 #endif
1547
1548 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1549                          struct nlattr *tb[], struct nlattr *data[])
1550 {
1551         struct net *net = dev_net(dev);
1552         struct ip_tunnel *nt;
1553         struct ip_tunnel_encap ipencap;
1554 #ifdef CONFIG_IPV6_SIT_6RD
1555         struct ip_tunnel_6rd ip6rd;
1556 #endif
1557         int err;
1558
1559         nt = netdev_priv(dev);
1560
1561         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1562                 err = ip_tunnel_encap_setup(nt, &ipencap);
1563                 if (err < 0)
1564                         return err;
1565         }
1566
1567         ipip6_netlink_parms(data, &nt->parms);
1568
1569         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1570                 return -EEXIST;
1571
1572         err = ipip6_tunnel_create(dev);
1573         if (err < 0)
1574                 return err;
1575
1576         if (tb[IFLA_MTU]) {
1577                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1578
1579                 if (mtu >= IPV6_MIN_MTU && mtu <= 0xFFF8 - dev->hard_header_len)
1580                         dev->mtu = mtu;
1581         }
1582
1583 #ifdef CONFIG_IPV6_SIT_6RD
1584         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1585                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1586                 if (err < 0)
1587                         unregister_netdevice_queue(dev, NULL);
1588         }
1589 #endif
1590
1591         return err;
1592 }
1593
1594 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1595                           struct nlattr *data[])
1596 {
1597         struct ip_tunnel *t = netdev_priv(dev);
1598         struct ip_tunnel_parm p;
1599         struct ip_tunnel_encap ipencap;
1600         struct net *net = t->net;
1601         struct sit_net *sitn = net_generic(net, sit_net_id);
1602 #ifdef CONFIG_IPV6_SIT_6RD
1603         struct ip_tunnel_6rd ip6rd;
1604 #endif
1605         int err;
1606
1607         if (dev == sitn->fb_tunnel_dev)
1608                 return -EINVAL;
1609
1610         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1611                 err = ip_tunnel_encap_setup(t, &ipencap);
1612                 if (err < 0)
1613                         return err;
1614         }
1615
1616         ipip6_netlink_parms(data, &p);
1617
1618         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1619             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1620                 return -EINVAL;
1621
1622         t = ipip6_tunnel_locate(net, &p, 0);
1623
1624         if (t) {
1625                 if (t->dev != dev)
1626                         return -EEXIST;
1627         } else
1628                 t = netdev_priv(dev);
1629
1630         ipip6_tunnel_update(t, &p);
1631
1632 #ifdef CONFIG_IPV6_SIT_6RD
1633         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1634                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1635 #endif
1636
1637         return 0;
1638 }
1639
1640 static size_t ipip6_get_size(const struct net_device *dev)
1641 {
1642         return
1643                 /* IFLA_IPTUN_LINK */
1644                 nla_total_size(4) +
1645                 /* IFLA_IPTUN_LOCAL */
1646                 nla_total_size(4) +
1647                 /* IFLA_IPTUN_REMOTE */
1648                 nla_total_size(4) +
1649                 /* IFLA_IPTUN_TTL */
1650                 nla_total_size(1) +
1651                 /* IFLA_IPTUN_TOS */
1652                 nla_total_size(1) +
1653                 /* IFLA_IPTUN_PMTUDISC */
1654                 nla_total_size(1) +
1655                 /* IFLA_IPTUN_FLAGS */
1656                 nla_total_size(2) +
1657                 /* IFLA_IPTUN_PROTO */
1658                 nla_total_size(1) +
1659 #ifdef CONFIG_IPV6_SIT_6RD
1660                 /* IFLA_IPTUN_6RD_PREFIX */
1661                 nla_total_size(sizeof(struct in6_addr)) +
1662                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1663                 nla_total_size(4) +
1664                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1665                 nla_total_size(2) +
1666                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1667                 nla_total_size(2) +
1668 #endif
1669                 /* IFLA_IPTUN_ENCAP_TYPE */
1670                 nla_total_size(2) +
1671                 /* IFLA_IPTUN_ENCAP_FLAGS */
1672                 nla_total_size(2) +
1673                 /* IFLA_IPTUN_ENCAP_SPORT */
1674                 nla_total_size(2) +
1675                 /* IFLA_IPTUN_ENCAP_DPORT */
1676                 nla_total_size(2) +
1677                 0;
1678 }
1679
1680 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1681 {
1682         struct ip_tunnel *tunnel = netdev_priv(dev);
1683         struct ip_tunnel_parm *parm = &tunnel->parms;
1684
1685         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1686             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1687             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1688             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1689             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1690             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1691                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1692             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1693             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1694                 goto nla_put_failure;
1695
1696 #ifdef CONFIG_IPV6_SIT_6RD
1697         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1698                              &tunnel->ip6rd.prefix) ||
1699             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1700                             tunnel->ip6rd.relay_prefix) ||
1701             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1702                         tunnel->ip6rd.prefixlen) ||
1703             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1704                         tunnel->ip6rd.relay_prefixlen))
1705                 goto nla_put_failure;
1706 #endif
1707
1708         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1709                         tunnel->encap.type) ||
1710             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1711                         tunnel->encap.sport) ||
1712             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1713                         tunnel->encap.dport) ||
1714             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1715                         tunnel->encap.flags))
1716                 goto nla_put_failure;
1717
1718         return 0;
1719
1720 nla_put_failure:
1721         return -EMSGSIZE;
1722 }
1723
1724 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1725         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1726         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1727         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1728         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1729         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1730         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1731         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1732         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1733 #ifdef CONFIG_IPV6_SIT_6RD
1734         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1735         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1736         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1737         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1738 #endif
1739         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1740         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1741         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1742         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1743 };
1744
1745 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1746 {
1747         struct net *net = dev_net(dev);
1748         struct sit_net *sitn = net_generic(net, sit_net_id);
1749
1750         if (dev != sitn->fb_tunnel_dev)
1751                 unregister_netdevice_queue(dev, head);
1752 }
1753
1754 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1755         .kind           = "sit",
1756         .maxtype        = IFLA_IPTUN_MAX,
1757         .policy         = ipip6_policy,
1758         .priv_size      = sizeof(struct ip_tunnel),
1759         .setup          = ipip6_tunnel_setup,
1760         .validate       = ipip6_validate,
1761         .newlink        = ipip6_newlink,
1762         .changelink     = ipip6_changelink,
1763         .get_size       = ipip6_get_size,
1764         .fill_info      = ipip6_fill_info,
1765         .dellink        = ipip6_dellink,
1766         .get_link_net   = ip_tunnel_get_link_net,
1767 };
1768
1769 static struct xfrm_tunnel sit_handler __read_mostly = {
1770         .handler        =       ipip6_rcv,
1771         .err_handler    =       ipip6_err,
1772         .priority       =       1,
1773 };
1774
1775 static struct xfrm_tunnel ipip_handler __read_mostly = {
1776         .handler        =       ipip_rcv,
1777         .err_handler    =       ipip6_err,
1778         .priority       =       2,
1779 };
1780
1781 static void __net_exit sit_destroy_tunnels(struct net *net,
1782                                            struct list_head *head)
1783 {
1784         struct sit_net *sitn = net_generic(net, sit_net_id);
1785         struct net_device *dev, *aux;
1786         int prio;
1787
1788         for_each_netdev_safe(net, dev, aux)
1789                 if (dev->rtnl_link_ops == &sit_link_ops)
1790                         unregister_netdevice_queue(dev, head);
1791
1792         for (prio = 1; prio < 4; prio++) {
1793                 int h;
1794                 for (h = 0; h < HASH_SIZE; h++) {
1795                         struct ip_tunnel *t;
1796
1797                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1798                         while (t) {
1799                                 /* If dev is in the same netns, it has already
1800                                  * been added to the list by the previous loop.
1801                                  */
1802                                 if (!net_eq(dev_net(t->dev), net))
1803                                         unregister_netdevice_queue(t->dev,
1804                                                                    head);
1805                                 t = rtnl_dereference(t->next);
1806                         }
1807                 }
1808         }
1809 }
1810
1811 static int __net_init sit_init_net(struct net *net)
1812 {
1813         struct sit_net *sitn = net_generic(net, sit_net_id);
1814         struct ip_tunnel *t;
1815         int err;
1816
1817         sitn->tunnels[0] = sitn->tunnels_wc;
1818         sitn->tunnels[1] = sitn->tunnels_l;
1819         sitn->tunnels[2] = sitn->tunnels_r;
1820         sitn->tunnels[3] = sitn->tunnels_r_l;
1821
1822         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1823                                            NET_NAME_UNKNOWN,
1824                                            ipip6_tunnel_setup);
1825         if (!sitn->fb_tunnel_dev) {
1826                 err = -ENOMEM;
1827                 goto err_alloc_dev;
1828         }
1829         dev_net_set(sitn->fb_tunnel_dev, net);
1830         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1831         /* FB netdevice is special: we have one, and only one per netns.
1832          * Allowing to move it to another netns is clearly unsafe.
1833          */
1834         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1835
1836         err = register_netdev(sitn->fb_tunnel_dev);
1837         if (err)
1838                 goto err_reg_dev;
1839
1840         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1841         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1842
1843         t = netdev_priv(sitn->fb_tunnel_dev);
1844
1845         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1846         return 0;
1847
1848 err_reg_dev:
1849         ipip6_dev_free(sitn->fb_tunnel_dev);
1850 err_alloc_dev:
1851         return err;
1852 }
1853
1854 static void __net_exit sit_exit_net(struct net *net)
1855 {
1856         LIST_HEAD(list);
1857
1858         rtnl_lock();
1859         sit_destroy_tunnels(net, &list);
1860         unregister_netdevice_many(&list);
1861         rtnl_unlock();
1862 }
1863
1864 static struct pernet_operations sit_net_ops = {
1865         .init = sit_init_net,
1866         .exit = sit_exit_net,
1867         .id   = &sit_net_id,
1868         .size = sizeof(struct sit_net),
1869 };
1870
1871 static void __exit sit_cleanup(void)
1872 {
1873         rtnl_link_unregister(&sit_link_ops);
1874         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1875         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1876
1877         unregister_pernet_device(&sit_net_ops);
1878         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1879 }
1880
1881 static int __init sit_init(void)
1882 {
1883         int err;
1884
1885         pr_info("IPv6 over IPv4 tunneling driver\n");
1886
1887         err = register_pernet_device(&sit_net_ops);
1888         if (err < 0)
1889                 return err;
1890         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1891         if (err < 0) {
1892                 pr_info("%s: can't register ip6ip4\n", __func__);
1893                 goto xfrm_tunnel_failed;
1894         }
1895         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1896         if (err < 0) {
1897                 pr_info("%s: can't register ip4ip4\n", __func__);
1898                 goto xfrm_tunnel4_failed;
1899         }
1900         err = rtnl_link_register(&sit_link_ops);
1901         if (err < 0)
1902                 goto rtnl_link_failed;
1903
1904 out:
1905         return err;
1906
1907 rtnl_link_failed:
1908         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1909 xfrm_tunnel4_failed:
1910         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1911 xfrm_tunnel_failed:
1912         unregister_pernet_device(&sit_net_ops);
1913         goto out;
1914 }
1915
1916 module_init(sit_init);
1917 module_exit(sit_cleanup);
1918 MODULE_LICENSE("GPL");
1919 MODULE_ALIAS_RTNL_LINK("sit");
1920 MODULE_ALIAS_NETDEV("sit0");