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 <linux/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 IP6_SIT_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 unsigned int sit_net_id __read_mostly;
  80 struct sit_net {
  81         struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  82         struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  83         struct ip_tunnel __rcu *tunnels_l[IP6_SIT_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,
  95                                              __be32 remote, __be32 local,
  96                                              int sifindex)
  97 {
  98         unsigned int h0 = HASH(remote);
  99         unsigned int h1 = HASH(local);
 100         struct ip_tunnel *t;
 101         struct sit_net *sitn = net_generic(net, sit_net_id);
 102         int ifindex = dev ? dev->ifindex : 0;
 103
 104         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 105                 if (local == t->parms.iph.saddr &&
 106                     remote == t->parms.iph.daddr &&
 107                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 108                      sifindex == t->parms.link) &&
 109                     (t->dev->flags & IFF_UP))
 110                         return t;
 111         }
 112         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 113                 if (remote == t->parms.iph.daddr &&
 114                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 115                      sifindex == t->parms.link) &&
 116                     (t->dev->flags & IFF_UP))
 117                         return t;
 118         }
 119         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 120                 if (local == t->parms.iph.saddr &&
 121                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 122                      sifindex == t->parms.link) &&
 123                     (t->dev->flags & IFF_UP))
 124                         return t;
 125         }
 126         t = rcu_dereference(sitn->tunnels_wc[0]);
 127         if (t && (t->dev->flags & IFF_UP))
 128                 return t;
 129         return NULL;
 130 }
 131
 132 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 133                 struct ip_tunnel_parm *parms)
 134 {
 135         __be32 remote = parms->iph.daddr;
 136         __be32 local = parms->iph.saddr;
 137         unsigned int h = 0;
 138         int prio = 0;
 139
 140         if (remote) {
 141                 prio |= 2;
 142                 h ^= HASH(remote);
 143         }
 144         if (local) {
 145                 prio |= 1;
 146                 h ^= HASH(local);
 147         }
 148         return &sitn->tunnels[prio][h];
 149 }
 150
 151 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 152                 struct ip_tunnel *t)
 153 {
 154         return __ipip6_bucket(sitn, &t->parms);
 155 }
 156
 157 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 158 {
 159         struct ip_tunnel __rcu **tp;
 160         struct ip_tunnel *iter;
 161
 162         for (tp = ipip6_bucket(sitn, t);
 163              (iter = rtnl_dereference(*tp)) != NULL;
 164              tp = &iter->next) {
 165                 if (t == iter) {
 166                         rcu_assign_pointer(*tp, t->next);
 167                         break;
 168                 }
 169         }
 170 }
 171
 172 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 173 {
 174         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 175
 176         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 177         rcu_assign_pointer(*tp, t);
 178 }
 179
 180 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 181 {
 182 #ifdef CONFIG_IPV6_SIT_6RD
 183         struct ip_tunnel *t = netdev_priv(dev);
 184
 185         if (dev == sitn->fb_tunnel_dev) {
 186                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 187                 t->ip6rd.relay_prefix = 0;
 188                 t->ip6rd.prefixlen = 16;
 189                 t->ip6rd.relay_prefixlen = 0;
 190         } else {
 191                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 192                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 193         }
 194 #endif
 195 }
 196
 197 static int ipip6_tunnel_create(struct net_device *dev)
 198 {
 199         struct ip_tunnel *t = netdev_priv(dev);
 200         struct net *net = dev_net(dev);
 201         struct sit_net *sitn = net_generic(net, sit_net_id);
 202         int err;
 203
 204         memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 205         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 206
 207         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 208                 dev->priv_flags |= IFF_ISATAP;
 209
 210         dev->rtnl_link_ops = &sit_link_ops;
 211
 212         err = register_netdevice(dev);
 213         if (err < 0)
 214                 goto out;
 215
 216         ipip6_tunnel_clone_6rd(dev, sitn);
 217
 218         dev_hold(dev);
 219
 220         ipip6_tunnel_link(sitn, t);
 221         return 0;
 222
 223 out:
 224         return err;
 225 }
 226
 227 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 228                 struct ip_tunnel_parm *parms, int create)
 229 {
 230         __be32 remote = parms->iph.daddr;
 231         __be32 local = parms->iph.saddr;
 232         struct ip_tunnel *t, *nt;
 233         struct ip_tunnel __rcu **tp;
 234         struct net_device *dev;
 235         char name[IFNAMSIZ];
 236         struct sit_net *sitn = net_generic(net, sit_net_id);
 237
 238         for (tp = __ipip6_bucket(sitn, parms);
 239             (t = rtnl_dereference(*tp)) != NULL;
 240              tp = &t->next) {
 241                 if (local == t->parms.iph.saddr &&
 242                     remote == t->parms.iph.daddr &&
 243                     parms->link == t->parms.link) {
 244                         if (create)
 245                                 return NULL;
 246                         else
 247                                 return t;
 248                 }
 249         }
 250         if (!create)
 251                 goto failed;
 252
 253         if (parms->name[0])
 254                 strlcpy(name, parms->name, IFNAMSIZ);
 255         else
 256                 strcpy(name, "sit%d");
 257
 258         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 259                            ipip6_tunnel_setup);
 260         if (!dev)
 261                 return NULL;
 262
 263         dev_net_set(dev, net);
 264
 265         nt = netdev_priv(dev);
 266
 267         nt->parms = *parms;
 268         if (ipip6_tunnel_create(dev) < 0)
 269                 goto failed_free;
 270
 271         return nt;
 272
 273 failed_free:
 274         free_netdev(dev);
 275 failed:
 276         return NULL;
 277 }
 278
 279 #define for_each_prl_rcu(start)                 \
 280         for (prl = rcu_dereference(start);      \
 281              prl;                               \
 282              prl = rcu_dereference(prl->next))
 283
 284 static struct ip_tunnel_prl_entry *
 285 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 286 {
 287         struct ip_tunnel_prl_entry *prl;
 288
 289         for_each_prl_rcu(t->prl)
 290                 if (prl->addr == addr)
 291                         break;
 292         return prl;
 293
 294 }
 295
 296 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 297                                 struct ip_tunnel_prl __user *a)
 298 {
 299         struct ip_tunnel_prl kprl, *kp;
 300         struct ip_tunnel_prl_entry *prl;
 301         unsigned int cmax, c = 0, ca, len;
 302         int ret = 0;
 303
 304         if (copy_from_user(&kprl, a, sizeof(kprl)))
 305                 return -EFAULT;
 306         cmax = kprl.datalen / sizeof(kprl);
 307         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 308                 cmax = 1;
 309
 310         /* For simple GET or for root users,
 311          * we try harder to allocate.
 312          */
 313         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 314                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 315                 NULL;
 316
 317         rcu_read_lock();
 318
 319         ca = t->prl_count < cmax ? t->prl_count : cmax;
 320
 321         if (!kp) {
 322                 /* We don't try hard to allocate much memory for
 323                  * non-root users.
 324                  * For root users, retry allocating enough memory for
 325                  * the answer.
 326                  */
 327                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 328                 if (!kp) {
 329                         ret = -ENOMEM;
 330                         goto out;
 331                 }
 332         }
 333
 334         c = 0;
 335         for_each_prl_rcu(t->prl) {
 336                 if (c >= cmax)
 337                         break;
 338                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 339                         continue;
 340                 kp[c].addr = prl->addr;
 341                 kp[c].flags = prl->flags;
 342                 c++;
 343                 if (kprl.addr != htonl(INADDR_ANY))
 344                         break;
 345         }
 346 out:
 347         rcu_read_unlock();
 348
 349         len = sizeof(*kp) * c;
 350         ret = 0;
 351         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 352                 ret = -EFAULT;
 353
 354         kfree(kp);
 355
 356         return ret;
 357 }
 358
 359 static int
 360 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 361 {
 362         struct ip_tunnel_prl_entry *p;
 363         int err = 0;
 364
 365         if (a->addr == htonl(INADDR_ANY))
 366                 return -EINVAL;
 367
 368         ASSERT_RTNL();
 369
 370         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 371                 if (p->addr == a->addr) {
 372                         if (chg) {
 373                                 p->flags = a->flags;
 374                                 goto out;
 375                         }
 376                         err = -EEXIST;
 377                         goto out;
 378                 }
 379         }
 380
 381         if (chg) {
 382                 err = -ENXIO;
 383                 goto out;
 384         }
 385
 386         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 387         if (!p) {
 388                 err = -ENOBUFS;
 389                 goto out;
 390         }
 391
 392         p->next = t->prl;
 393         p->addr = a->addr;
 394         p->flags = a->flags;
 395         t->prl_count++;
 396         rcu_assign_pointer(t->prl, p);
 397 out:
 398         return err;
 399 }
 400
 401 static void prl_list_destroy_rcu(struct rcu_head *head)
 402 {
 403         struct ip_tunnel_prl_entry *p, *n;
 404
 405         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 406         do {
 407                 n = rcu_dereference_protected(p->next, 1);
 408                 kfree(p);
 409                 p = n;
 410         } while (p);
 411 }
 412
 413 static int
 414 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 415 {
 416         struct ip_tunnel_prl_entry *x;
 417         struct ip_tunnel_prl_entry __rcu **p;
 418         int err = 0;
 419
 420         ASSERT_RTNL();
 421
 422         if (a && a->addr != htonl(INADDR_ANY)) {
 423                 for (p = &t->prl;
 424                      (x = rtnl_dereference(*p)) != NULL;
 425                      p = &x->next) {
 426                         if (x->addr == a->addr) {
 427                                 *p = x->next;
 428                                 kfree_rcu(x, rcu_head);
 429                                 t->prl_count--;
 430                                 goto out;
 431                         }
 432                 }
 433                 err = -ENXIO;
 434         } else {
 435                 x = rtnl_dereference(t->prl);
 436                 if (x) {
 437                         t->prl_count = 0;
 438                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 439                         t->prl = NULL;
 440                 }
 441         }
 442 out:
 443         return err;
 444 }
 445
 446 static int
 447 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 448 {
 449         struct ip_tunnel_prl_entry *p;
 450         int ok = 1;
 451
 452         rcu_read_lock();
 453         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 454         if (p) {
 455                 if (p->flags & PRL_DEFAULT)
 456                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 457                 else
 458                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 459         } else {
 460                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 461
 462                 if (ipv6_addr_is_isatap(addr6) &&
 463                     (addr6->s6_addr32[3] == iph->saddr) &&
 464                     ipv6_chk_prefix(addr6, t->dev))
 465                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 466                 else
 467                         ok = 0;
 468         }
 469         rcu_read_unlock();
 470         return ok;
 471 }
 472
 473 static void ipip6_tunnel_uninit(struct net_device *dev)
 474 {
 475         struct ip_tunnel *tunnel = netdev_priv(dev);
 476         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 477
 478         if (dev == sitn->fb_tunnel_dev) {
 479                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 480         } else {
 481                 ipip6_tunnel_unlink(sitn, tunnel);
 482                 ipip6_tunnel_del_prl(tunnel, NULL);
 483         }
 484         dst_cache_reset(&tunnel->dst_cache);
 485         dev_put(dev);
 486 }
 487
 488 static int ipip6_err(struct sk_buff *skb, u32 info)
 489 {
 490         const struct iphdr *iph = (const struct iphdr *)skb->data;
 491         const int type = icmp_hdr(skb)->type;
 492         const int code = icmp_hdr(skb)->code;
 493         unsigned int data_len = 0;
 494         struct ip_tunnel *t;
 495         int sifindex;
 496         int err;
 497
 498         switch (type) {
 499         default:
 500         case ICMP_PARAMETERPROB:
 501                 return 0;
 502
 503         case ICMP_DEST_UNREACH:
 504                 switch (code) {
 505                 case ICMP_SR_FAILED:
 506                         /* Impossible event. */
 507                         return 0;
 508                 default:
 509                         /* All others are translated to HOST_UNREACH.
 510                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 511                            I believe they are just ether pollution. --ANK
 512                          */
 513                         break;
 514                 }
 515                 break;
 516         case ICMP_TIME_EXCEEDED:
 517                 if (code != ICMP_EXC_TTL)
 518                         return 0;
 519                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 520                 break;
 521         case ICMP_REDIRECT:
 522                 break;
 523         }
 524
 525         err = -ENOENT;
 526
 527         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 528         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 529                                 iph->daddr, iph->saddr, sifindex);
 530         if (!t)
 531                 goto out;
 532
 533         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 534                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 535                                  t->parms.link, 0, iph->protocol, 0);
 536                 err = 0;
 537                 goto out;
 538         }
 539         if (type == ICMP_REDIRECT) {
 540                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
 541                               iph->protocol, 0);
 542                 err = 0;
 543                 goto out;
 544         }
 545
 546         err = 0;
 547         if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 548                 goto out;
 549
 550         if (t->parms.iph.daddr == 0)
 551                 goto out;
 552
 553         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 554                 goto out;
 555
 556         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 557                 t->err_count++;
 558         else
 559                 t->err_count = 1;
 560         t->err_time = jiffies;
 561 out:
 562         return err;
 563 }
 564
 565 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 566                                   const struct in6_addr *v6addr)
 567 {
 568         __be32 v4embed = 0;
 569         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 570                 return true;
 571         return false;
 572 }
 573
 574 /* Checks if an address matches an address on the tunnel interface.
 575  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 576  * Long story:
 577  * This function is called after we considered the packet as spoofed
 578  * in is_spoofed_6rd.
 579  * We may have a router that is doing NAT for proto 41 packets
 580  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 581  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 582  * function will return true, dropping the packet.
 583  * But, we can still check if is spoofed against the IP
 584  * addresses associated with the interface.
 585  */
 586 static bool only_dnatted(const struct ip_tunnel *tunnel,
 587         const struct in6_addr *v6dst)
 588 {
 589         int prefix_len;
 590
 591 #ifdef CONFIG_IPV6_SIT_6RD
 592         prefix_len = tunnel->ip6rd.prefixlen + 32
 593                 - tunnel->ip6rd.relay_prefixlen;
 594 #else
 595         prefix_len = 48;
 596 #endif
 597         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 598 }
 599
 600 /* Returns true if a packet is spoofed */
 601 static bool packet_is_spoofed(struct sk_buff *skb,
 602                               const struct iphdr *iph,
 603                               struct ip_tunnel *tunnel)
 604 {
 605         const struct ipv6hdr *ipv6h;
 606
 607         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 608                 if (!isatap_chksrc(skb, iph, tunnel))
 609                         return true;
 610
 611                 return false;
 612         }
 613
 614         if (tunnel->dev->flags & IFF_POINTOPOINT)
 615                 return false;
 616
 617         ipv6h = ipv6_hdr(skb);
 618
 619         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 620                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 621                                      &iph->saddr, &ipv6h->saddr,
 622                                      &iph->daddr, &ipv6h->daddr);
 623                 return true;
 624         }
 625
 626         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 627                 return false;
 628
 629         if (only_dnatted(tunnel, &ipv6h->daddr))
 630                 return false;
 631
 632         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 633                              &iph->saddr, &ipv6h->saddr,
 634                              &iph->daddr, &ipv6h->daddr);
 635         return true;
 636 }
 637
 638 static int ipip6_rcv(struct sk_buff *skb)
 639 {
 640         const struct iphdr *iph = ip_hdr(skb);
 641         struct ip_tunnel *tunnel;
 642         int sifindex;
 643         int err;
 644
 645         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 646         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 647                                      iph->saddr, iph->daddr, sifindex);
 648         if (tunnel) {
 649                 struct pcpu_sw_netstats *tstats;
 650
 651                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 652                     tunnel->parms.iph.protocol != 0)
 653                         goto out;
 654
 655                 skb->mac_header = skb->network_header;
 656                 skb_reset_network_header(skb);
 657                 IPCB(skb)->flags = 0;
 658                 skb->dev = tunnel->dev;
 659
 660                 if (packet_is_spoofed(skb, iph, tunnel)) {
 661                         tunnel->dev->stats.rx_errors++;
 662                         goto out;
 663                 }
 664
 665                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 666                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 667                         goto out;
 668
 669                 err = IP_ECN_decapsulate(iph, skb);
 670                 if (unlikely(err)) {
 671                         if (log_ecn_error)
 672                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 673                                                      &iph->saddr, iph->tos);
 674                         if (err > 1) {
 675                                 ++tunnel->dev->stats.rx_frame_errors;
 676                                 ++tunnel->dev->stats.rx_errors;
 677                                 goto out;
 678                         }
 679                 }
 680
 681                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 682                 u64_stats_update_begin(&tstats->syncp);
 683                 tstats->rx_packets++;
 684                 tstats->rx_bytes += skb->len;
 685                 u64_stats_update_end(&tstats->syncp);
 686
 687                 netif_rx(skb);
 688
 689                 return 0;
 690         }
 691
 692         /* no tunnel matched,  let upstream know, ipsec may handle it */
 693         return 1;
 694 out:
 695         kfree_skb(skb);
 696         return 0;
 697 }
 698
 699 static const struct tnl_ptk_info ipip_tpi = {
 700         /* no tunnel info required for ipip. */
 701         .proto = htons(ETH_P_IP),
 702 };
 703
 704 #if IS_ENABLED(CONFIG_MPLS)
 705 static const struct tnl_ptk_info mplsip_tpi = {
 706         /* no tunnel info required for mplsip. */
 707         .proto = htons(ETH_P_MPLS_UC),
 708 };
 709 #endif
 710
 711 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 712 {
 713         const struct iphdr *iph;
 714         struct ip_tunnel *tunnel;
 715         int sifindex;
 716
 717         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 718
 719         iph = ip_hdr(skb);
 720         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 721                                      iph->saddr, iph->daddr, sifindex);
 722         if (tunnel) {
 723                 const struct tnl_ptk_info *tpi;
 724
 725                 if (tunnel->parms.iph.protocol != ipproto &&
 726                     tunnel->parms.iph.protocol != 0)
 727                         goto drop;
 728
 729                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 730                         goto drop;
 731 #if IS_ENABLED(CONFIG_MPLS)
 732                 if (ipproto == IPPROTO_MPLS)
 733                         tpi = &mplsip_tpi;
 734                 else
 735 #endif
 736                         tpi = &ipip_tpi;
 737                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 738                         goto drop;
 739                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 740         }
 741
 742         return 1;
 743
 744 drop:
 745         kfree_skb(skb);
 746         return 0;
 747 }
 748
 749 static int ipip_rcv(struct sk_buff *skb)
 750 {
 751         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 752 }
 753
 754 #if IS_ENABLED(CONFIG_MPLS)
 755 static int mplsip_rcv(struct sk_buff *skb)
 756 {
 757         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 758 }
 759 #endif
 760
 761 /*
 762  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 763  * stores the embedded IPv4 address in v4dst and returns true.
 764  */
 765 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 766                       __be32 *v4dst)
 767 {
 768 #ifdef CONFIG_IPV6_SIT_6RD
 769         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 770                               tunnel->ip6rd.prefixlen)) {
 771                 unsigned int pbw0, pbi0;
 772                 int pbi1;
 773                 u32 d;
 774
 775                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 776                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 777
 778                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 779                     tunnel->ip6rd.relay_prefixlen;
 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         u8 protocol = IPPROTO_IPV6;
 830         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 831
 832         if (tos == 1)
 833                 tos = ipv6_get_dsfield(iph6);
 834
 835         /* ISATAP (RFC4214) - must come before 6to4 */
 836         if (dev->priv_flags & IFF_ISATAP) {
 837                 struct neighbour *neigh = NULL;
 838                 bool do_tx_error = false;
 839
 840                 if (skb_dst(skb))
 841                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 842
 843                 if (!neigh) {
 844                         net_dbg_ratelimited("nexthop == NULL\n");
 845                         goto tx_error;
 846                 }
 847
 848                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 849                 addr_type = ipv6_addr_type(addr6);
 850
 851                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 852                      ipv6_addr_is_isatap(addr6))
 853                         dst = addr6->s6_addr32[3];
 854                 else
 855                         do_tx_error = true;
 856
 857                 neigh_release(neigh);
 858                 if (do_tx_error)
 859                         goto tx_error;
 860         }
 861
 862         if (!dst)
 863                 dst = try_6rd(tunnel, &iph6->daddr);
 864
 865         if (!dst) {
 866                 struct neighbour *neigh = NULL;
 867                 bool do_tx_error = false;
 868
 869                 if (skb_dst(skb))
 870                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 871
 872                 if (!neigh) {
 873                         net_dbg_ratelimited("nexthop == NULL\n");
 874                         goto tx_error;
 875                 }
 876
 877                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 878                 addr_type = ipv6_addr_type(addr6);
 879
 880                 if (addr_type == IPV6_ADDR_ANY) {
 881                         addr6 = &ipv6_hdr(skb)->daddr;
 882                         addr_type = ipv6_addr_type(addr6);
 883                 }
 884
 885                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 886                         dst = addr6->s6_addr32[3];
 887                 else
 888                         do_tx_error = true;
 889
 890                 neigh_release(neigh);
 891                 if (do_tx_error)
 892                         goto tx_error;
 893         }
 894
 895         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 896                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 897                            0, dst, tiph->saddr, 0, 0,
 898                            sock_net_uid(tunnel->net, NULL));
 899         rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 900
 901         if (IS_ERR(rt)) {
 902                 dev->stats.tx_carrier_errors++;
 903                 goto tx_error_icmp;
 904         }
 905         if (rt->rt_type != RTN_UNICAST) {
 906                 ip_rt_put(rt);
 907                 dev->stats.tx_carrier_errors++;
 908                 goto tx_error_icmp;
 909         }
 910         tdev = rt->dst.dev;
 911
 912         if (tdev == dev) {
 913                 ip_rt_put(rt);
 914                 dev->stats.collisions++;
 915                 goto tx_error;
 916         }
 917
 918         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 919                 ip_rt_put(rt);
 920                 goto tx_error;
 921         }
 922
 923         if (df) {
 924                 mtu = dst_mtu(&rt->dst) - t_hlen;
 925
 926                 if (mtu < 68) {
 927                         dev->stats.collisions++;
 928                         ip_rt_put(rt);
 929                         goto tx_error;
 930                 }
 931
 932                 if (mtu < IPV6_MIN_MTU) {
 933                         mtu = IPV6_MIN_MTU;
 934                         df = 0;
 935                 }
 936
 937                 if (tunnel->parms.iph.daddr)
 938                         skb_dst_update_pmtu(skb, mtu);
 939
 940                 if (skb->len > mtu && !skb_is_gso(skb)) {
 941                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 942                         ip_rt_put(rt);
 943                         goto tx_error;
 944                 }
 945         }
 946
 947         if (tunnel->err_count > 0) {
 948                 if (time_before(jiffies,
 949                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 950                         tunnel->err_count--;
 951                         dst_link_failure(skb);
 952                 } else
 953                         tunnel->err_count = 0;
 954         }
 955
 956         /*
 957          * Okay, now see if we can stuff it in the buffer as-is.
 958          */
 959         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 960
 961         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 962             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 963                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 964                 if (!new_skb) {
 965                         ip_rt_put(rt);
 966                         dev->stats.tx_dropped++;
 967                         kfree_skb(skb);
 968                         return NETDEV_TX_OK;
 969                 }
 970                 if (skb->sk)
 971                         skb_set_owner_w(new_skb, skb->sk);
 972                 dev_kfree_skb(skb);
 973                 skb = new_skb;
 974                 iph6 = ipv6_hdr(skb);
 975         }
 976         ttl = tiph->ttl;
 977         if (ttl == 0)
 978                 ttl = iph6->hop_limit;
 979         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 980
 981         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 982                 ip_rt_put(rt);
 983                 goto tx_error;
 984         }
 985
 986         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 987
 988         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 989                       df, !net_eq(tunnel->net, dev_net(dev)));
 990         return NETDEV_TX_OK;
 991
 992 tx_error_icmp:
 993         dst_link_failure(skb);
 994 tx_error:
 995         kfree_skb(skb);
 996         dev->stats.tx_errors++;
 997         return NETDEV_TX_OK;
 998 }
 999
1000 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1001                                      struct net_device *dev, u8 ipproto)
1002 {
1003         struct ip_tunnel *tunnel = netdev_priv(dev);
1004         const struct iphdr  *tiph = &tunnel->parms.iph;
1005
1006         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1007                 goto tx_error;
1008
1009         skb_set_inner_ipproto(skb, ipproto);
1010
1011         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1012         return NETDEV_TX_OK;
1013 tx_error:
1014         kfree_skb(skb);
1015         dev->stats.tx_errors++;
1016         return NETDEV_TX_OK;
1017 }
1018
1019 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1020                                    struct net_device *dev)
1021 {
1022         switch (skb->protocol) {
1023         case htons(ETH_P_IP):
1024                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1025                 break;
1026         case htons(ETH_P_IPV6):
1027                 ipip6_tunnel_xmit(skb, dev);
1028                 break;
1029 #if IS_ENABLED(CONFIG_MPLS)
1030         case htons(ETH_P_MPLS_UC):
1031                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1032                 break;
1033 #endif
1034         default:
1035                 goto tx_err;
1036         }
1037
1038         return NETDEV_TX_OK;
1039
1040 tx_err:
1041         dev->stats.tx_errors++;
1042         kfree_skb(skb);
1043         return NETDEV_TX_OK;
1044
1045 }
1046
1047 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1048 {
1049         struct net_device *tdev = NULL;
1050         struct ip_tunnel *tunnel;
1051         const struct iphdr *iph;
1052         struct flowi4 fl4;
1053
1054         tunnel = netdev_priv(dev);
1055         iph = &tunnel->parms.iph;
1056
1057         if (iph->daddr) {
1058                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1059                                                           NULL,
1060                                                           iph->daddr, iph->saddr,
1061                                                           0, 0,
1062                                                           IPPROTO_IPV6,
1063                                                           RT_TOS(iph->tos),
1064                                                           tunnel->parms.link);
1065
1066                 if (!IS_ERR(rt)) {
1067                         tdev = rt->dst.dev;
1068                         ip_rt_put(rt);
1069                 }
1070                 dev->flags |= IFF_POINTOPOINT;
1071         }
1072
1073         if (!tdev && tunnel->parms.link)
1074                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1075
1076         if (tdev) {
1077                 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1078
1079                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
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                                 __u32 fwmark)
1088 {
1089         struct net *net = t->net;
1090         struct sit_net *sitn = net_generic(net, sit_net_id);
1091
1092         ipip6_tunnel_unlink(sitn, t);
1093         synchronize_net();
1094         t->parms.iph.saddr = p->iph.saddr;
1095         t->parms.iph.daddr = p->iph.daddr;
1096         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1097         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1098         ipip6_tunnel_link(sitn, t);
1099         t->parms.iph.ttl = p->iph.ttl;
1100         t->parms.iph.tos = p->iph.tos;
1101         t->parms.iph.frag_off = p->iph.frag_off;
1102         if (t->parms.link != p->link || t->fwmark != fwmark) {
1103                 t->parms.link = p->link;
1104                 t->fwmark = fwmark;
1105                 ipip6_tunnel_bind_dev(t->dev);
1106         }
1107         dst_cache_reset(&t->dst_cache);
1108         netdev_state_change(t->dev);
1109 }
1110
1111 #ifdef CONFIG_IPV6_SIT_6RD
1112 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1113                                    struct ip_tunnel_6rd *ip6rd)
1114 {
1115         struct in6_addr prefix;
1116         __be32 relay_prefix;
1117
1118         if (ip6rd->relay_prefixlen > 32 ||
1119             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1120                 return -EINVAL;
1121
1122         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1123         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1124                 return -EINVAL;
1125         if (ip6rd->relay_prefixlen)
1126                 relay_prefix = ip6rd->relay_prefix &
1127                                htonl(0xffffffffUL <<
1128                                      (32 - ip6rd->relay_prefixlen));
1129         else
1130                 relay_prefix = 0;
1131         if (relay_prefix != ip6rd->relay_prefix)
1132                 return -EINVAL;
1133
1134         t->ip6rd.prefix = prefix;
1135         t->ip6rd.relay_prefix = relay_prefix;
1136         t->ip6rd.prefixlen = ip6rd->prefixlen;
1137         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1138         dst_cache_reset(&t->dst_cache);
1139         netdev_state_change(t->dev);
1140         return 0;
1141 }
1142 #endif
1143
1144 static bool ipip6_valid_ip_proto(u8 ipproto)
1145 {
1146         return ipproto == IPPROTO_IPV6 ||
1147                 ipproto == IPPROTO_IPIP ||
1148 #if IS_ENABLED(CONFIG_MPLS)
1149                 ipproto == IPPROTO_MPLS ||
1150 #endif
1151                 ipproto == 0;
1152 }
1153
1154 static int
1155 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1156 {
1157         int err = 0;
1158         struct ip_tunnel_parm p;
1159         struct ip_tunnel_prl prl;
1160         struct ip_tunnel *t = netdev_priv(dev);
1161         struct net *net = t->net;
1162         struct sit_net *sitn = net_generic(net, sit_net_id);
1163 #ifdef CONFIG_IPV6_SIT_6RD
1164         struct ip_tunnel_6rd ip6rd;
1165 #endif
1166
1167         switch (cmd) {
1168         case SIOCGETTUNNEL:
1169 #ifdef CONFIG_IPV6_SIT_6RD
1170         case SIOCGET6RD:
1171 #endif
1172                 if (dev == sitn->fb_tunnel_dev) {
1173                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1174                                 err = -EFAULT;
1175                                 break;
1176                         }
1177                         t = ipip6_tunnel_locate(net, &p, 0);
1178                         if (!t)
1179                                 t = netdev_priv(dev);
1180                 }
1181
1182                 err = -EFAULT;
1183                 if (cmd == SIOCGETTUNNEL) {
1184                         memcpy(&p, &t->parms, sizeof(p));
1185                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1186                                          sizeof(p)))
1187                                 goto done;
1188 #ifdef CONFIG_IPV6_SIT_6RD
1189                 } else {
1190                         ip6rd.prefix = t->ip6rd.prefix;
1191                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1192                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1193                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1194                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1195                                          sizeof(ip6rd)))
1196                                 goto done;
1197 #endif
1198                 }
1199                 err = 0;
1200                 break;
1201
1202         case SIOCADDTUNNEL:
1203         case SIOCCHGTUNNEL:
1204                 err = -EPERM;
1205                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1206                         goto done;
1207
1208                 err = -EFAULT;
1209                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1210                         goto done;
1211
1212                 err = -EINVAL;
1213                 if (!ipip6_valid_ip_proto(p.iph.protocol))
1214                         goto done;
1215                 if (p.iph.version != 4 ||
1216                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1217                         goto done;
1218                 if (p.iph.ttl)
1219                         p.iph.frag_off |= htons(IP_DF);
1220
1221                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1222
1223                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1224                         if (t) {
1225                                 if (t->dev != dev) {
1226                                         err = -EEXIST;
1227                                         break;
1228                                 }
1229                         } else {
1230                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1231                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1232                                         err = -EINVAL;
1233                                         break;
1234                                 }
1235                                 t = netdev_priv(dev);
1236                         }
1237
1238                         ipip6_tunnel_update(t, &p, t->fwmark);
1239                 }
1240
1241                 if (t) {
1242                         err = 0;
1243                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1244                                 err = -EFAULT;
1245                 } else
1246                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1247                 break;
1248
1249         case SIOCDELTUNNEL:
1250                 err = -EPERM;
1251                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1252                         goto done;
1253
1254                 if (dev == sitn->fb_tunnel_dev) {
1255                         err = -EFAULT;
1256                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1257                                 goto done;
1258                         err = -ENOENT;
1259                         t = ipip6_tunnel_locate(net, &p, 0);
1260                         if (!t)
1261                                 goto done;
1262                         err = -EPERM;
1263                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1264                                 goto done;
1265                         dev = t->dev;
1266                 }
1267                 unregister_netdevice(dev);
1268                 err = 0;
1269                 break;
1270
1271         case SIOCGETPRL:
1272                 err = -EINVAL;
1273                 if (dev == sitn->fb_tunnel_dev)
1274                         goto done;
1275                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1276                 break;
1277
1278         case SIOCADDPRL:
1279         case SIOCDELPRL:
1280         case SIOCCHGPRL:
1281                 err = -EPERM;
1282                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1283                         goto done;
1284                 err = -EINVAL;
1285                 if (dev == sitn->fb_tunnel_dev)
1286                         goto done;
1287                 err = -EFAULT;
1288                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1289                         goto done;
1290
1291                 switch (cmd) {
1292                 case SIOCDELPRL:
1293                         err = ipip6_tunnel_del_prl(t, &prl);
1294                         break;
1295                 case SIOCADDPRL:
1296                 case SIOCCHGPRL:
1297                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1298                         break;
1299                 }
1300                 dst_cache_reset(&t->dst_cache);
1301                 netdev_state_change(dev);
1302                 break;
1303
1304 #ifdef CONFIG_IPV6_SIT_6RD
1305         case SIOCADD6RD:
1306         case SIOCCHG6RD:
1307         case SIOCDEL6RD:
1308                 err = -EPERM;
1309                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1310                         goto done;
1311
1312                 err = -EFAULT;
1313                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1314                                    sizeof(ip6rd)))
1315                         goto done;
1316
1317                 if (cmd != SIOCDEL6RD) {
1318                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1319                         if (err < 0)
1320                                 goto done;
1321                 } else
1322                         ipip6_tunnel_clone_6rd(dev, sitn);
1323
1324                 err = 0;
1325                 break;
1326 #endif
1327
1328         default:
1329                 err = -EINVAL;
1330         }
1331
1332 done:
1333         return err;
1334 }
1335
1336 static const struct net_device_ops ipip6_netdev_ops = {
1337         .ndo_init       = ipip6_tunnel_init,
1338         .ndo_uninit     = ipip6_tunnel_uninit,
1339         .ndo_start_xmit = sit_tunnel_xmit,
1340         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1341         .ndo_get_stats64 = ip_tunnel_get_stats64,
1342         .ndo_get_iflink = ip_tunnel_get_iflink,
1343 };
1344
1345 static void ipip6_dev_free(struct net_device *dev)
1346 {
1347         struct ip_tunnel *tunnel = netdev_priv(dev);
1348
1349         dst_cache_destroy(&tunnel->dst_cache);
1350         free_percpu(dev->tstats);
1351 }
1352
1353 #define SIT_FEATURES (NETIF_F_SG           | \
1354                       NETIF_F_FRAGLIST     | \
1355                       NETIF_F_HIGHDMA      | \
1356                       NETIF_F_GSO_SOFTWARE | \
1357                       NETIF_F_HW_CSUM)
1358
1359 static void ipip6_tunnel_setup(struct net_device *dev)
1360 {
1361         struct ip_tunnel *tunnel = netdev_priv(dev);
1362         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1363
1364         dev->netdev_ops         = &ipip6_netdev_ops;
1365         dev->needs_free_netdev  = true;
1366         dev->priv_destructor    = ipip6_dev_free;
1367
1368         dev->type               = ARPHRD_SIT;
1369         dev->hard_header_len    = LL_MAX_HEADER + t_hlen;
1370         dev->mtu                = ETH_DATA_LEN - t_hlen;
1371         dev->min_mtu            = IPV6_MIN_MTU;
1372         dev->max_mtu            = 0xFFF8 - t_hlen;
1373         dev->flags              = IFF_NOARP;
1374         netif_keep_dst(dev);
1375         dev->addr_len           = 4;
1376         dev->features           |= NETIF_F_LLTX;
1377         dev->features           |= SIT_FEATURES;
1378         dev->hw_features        |= SIT_FEATURES;
1379 }
1380
1381 static int ipip6_tunnel_init(struct net_device *dev)
1382 {
1383         struct ip_tunnel *tunnel = netdev_priv(dev);
1384         int err;
1385
1386         tunnel->dev = dev;
1387         tunnel->net = dev_net(dev);
1388         strcpy(tunnel->parms.name, dev->name);
1389
1390         ipip6_tunnel_bind_dev(dev);
1391         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1392         if (!dev->tstats)
1393                 return -ENOMEM;
1394
1395         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1396         if (err) {
1397                 free_percpu(dev->tstats);
1398                 dev->tstats = NULL;
1399                 return err;
1400         }
1401
1402         return 0;
1403 }
1404
1405 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1406 {
1407         struct ip_tunnel *tunnel = netdev_priv(dev);
1408         struct iphdr *iph = &tunnel->parms.iph;
1409         struct net *net = dev_net(dev);
1410         struct sit_net *sitn = net_generic(net, sit_net_id);
1411
1412         iph->version            = 4;
1413         iph->protocol           = IPPROTO_IPV6;
1414         iph->ihl                = 5;
1415         iph->ttl                = 64;
1416
1417         dev_hold(dev);
1418         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1419 }
1420
1421 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1422                           struct netlink_ext_ack *extack)
1423 {
1424         u8 proto;
1425
1426         if (!data || !data[IFLA_IPTUN_PROTO])
1427                 return 0;
1428
1429         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1430         if (!ipip6_valid_ip_proto(proto))
1431                 return -EINVAL;
1432
1433         return 0;
1434 }
1435
1436 static void ipip6_netlink_parms(struct nlattr *data[],
1437                                 struct ip_tunnel_parm *parms,
1438                                 __u32 *fwmark)
1439 {
1440         memset(parms, 0, sizeof(*parms));
1441
1442         parms->iph.version = 4;
1443         parms->iph.protocol = IPPROTO_IPV6;
1444         parms->iph.ihl = 5;
1445         parms->iph.ttl = 64;
1446
1447         if (!data)
1448                 return;
1449
1450         if (data[IFLA_IPTUN_LINK])
1451                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1452
1453         if (data[IFLA_IPTUN_LOCAL])
1454                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1455
1456         if (data[IFLA_IPTUN_REMOTE])
1457                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1458
1459         if (data[IFLA_IPTUN_TTL]) {
1460                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1461                 if (parms->iph.ttl)
1462                         parms->iph.frag_off = htons(IP_DF);
1463         }
1464
1465         if (data[IFLA_IPTUN_TOS])
1466                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1467
1468         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1469                 parms->iph.frag_off = htons(IP_DF);
1470
1471         if (data[IFLA_IPTUN_FLAGS])
1472                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1473
1474         if (data[IFLA_IPTUN_PROTO])
1475                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1476
1477         if (data[IFLA_IPTUN_FWMARK])
1478                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1479 }
1480
1481 /* This function returns true when ENCAP attributes are present in the nl msg */
1482 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1483                                       struct ip_tunnel_encap *ipencap)
1484 {
1485         bool ret = false;
1486
1487         memset(ipencap, 0, sizeof(*ipencap));
1488
1489         if (!data)
1490                 return ret;
1491
1492         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1493                 ret = true;
1494                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1495         }
1496
1497         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1498                 ret = true;
1499                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1500         }
1501
1502         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1503                 ret = true;
1504                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1505         }
1506
1507         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1508                 ret = true;
1509                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1510         }
1511
1512         return ret;
1513 }
1514
1515 #ifdef CONFIG_IPV6_SIT_6RD
1516 /* This function returns true when 6RD attributes are present in the nl msg */
1517 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1518                                     struct ip_tunnel_6rd *ip6rd)
1519 {
1520         bool ret = false;
1521         memset(ip6rd, 0, sizeof(*ip6rd));
1522
1523         if (!data)
1524                 return ret;
1525
1526         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1527                 ret = true;
1528                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1529         }
1530
1531         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1532                 ret = true;
1533                 ip6rd->relay_prefix =
1534                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1535         }
1536
1537         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1538                 ret = true;
1539                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1540         }
1541
1542         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1543                 ret = true;
1544                 ip6rd->relay_prefixlen =
1545                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1546         }
1547
1548         return ret;
1549 }
1550 #endif
1551
1552 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1553                          struct nlattr *tb[], struct nlattr *data[],
1554                          struct netlink_ext_ack *extack)
1555 {
1556         struct net *net = dev_net(dev);
1557         struct ip_tunnel *nt;
1558         struct ip_tunnel_encap ipencap;
1559 #ifdef CONFIG_IPV6_SIT_6RD
1560         struct ip_tunnel_6rd ip6rd;
1561 #endif
1562         int err;
1563
1564         nt = netdev_priv(dev);
1565
1566         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1567                 err = ip_tunnel_encap_setup(nt, &ipencap);
1568                 if (err < 0)
1569                         return err;
1570         }
1571
1572         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1573
1574         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1575                 return -EEXIST;
1576
1577         err = ipip6_tunnel_create(dev);
1578         if (err < 0)
1579                 return err;
1580
1581         if (tb[IFLA_MTU]) {
1582                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1583
1584                 if (mtu >= IPV6_MIN_MTU && mtu <= 0xFFF8 - dev->hard_header_len)
1585                         dev->mtu = mtu;
1586         }
1587
1588 #ifdef CONFIG_IPV6_SIT_6RD
1589         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1590                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1591 #endif
1592
1593         return err;
1594 }
1595
1596 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1597                             struct nlattr *data[],
1598                             struct netlink_ext_ack *extack)
1599 {
1600         struct ip_tunnel *t = netdev_priv(dev);
1601         struct ip_tunnel_parm p;
1602         struct ip_tunnel_encap ipencap;
1603         struct net *net = t->net;
1604         struct sit_net *sitn = net_generic(net, sit_net_id);
1605 #ifdef CONFIG_IPV6_SIT_6RD
1606         struct ip_tunnel_6rd ip6rd;
1607 #endif
1608         __u32 fwmark = t->fwmark;
1609         int err;
1610
1611         if (dev == sitn->fb_tunnel_dev)
1612                 return -EINVAL;
1613
1614         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1615                 err = ip_tunnel_encap_setup(t, &ipencap);
1616                 if (err < 0)
1617                         return err;
1618         }
1619
1620         ipip6_netlink_parms(data, &p, &fwmark);
1621
1622         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1623             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1624                 return -EINVAL;
1625
1626         t = ipip6_tunnel_locate(net, &p, 0);
1627
1628         if (t) {
1629                 if (t->dev != dev)
1630                         return -EEXIST;
1631         } else
1632                 t = netdev_priv(dev);
1633
1634         ipip6_tunnel_update(t, &p, fwmark);
1635
1636 #ifdef CONFIG_IPV6_SIT_6RD
1637         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1638                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1639 #endif
1640
1641         return 0;
1642 }
1643
1644 static size_t ipip6_get_size(const struct net_device *dev)
1645 {
1646         return
1647                 /* IFLA_IPTUN_LINK */
1648                 nla_total_size(4) +
1649                 /* IFLA_IPTUN_LOCAL */
1650                 nla_total_size(4) +
1651                 /* IFLA_IPTUN_REMOTE */
1652                 nla_total_size(4) +
1653                 /* IFLA_IPTUN_TTL */
1654                 nla_total_size(1) +
1655                 /* IFLA_IPTUN_TOS */
1656                 nla_total_size(1) +
1657                 /* IFLA_IPTUN_PMTUDISC */
1658                 nla_total_size(1) +
1659                 /* IFLA_IPTUN_FLAGS */
1660                 nla_total_size(2) +
1661                 /* IFLA_IPTUN_PROTO */
1662                 nla_total_size(1) +
1663 #ifdef CONFIG_IPV6_SIT_6RD
1664                 /* IFLA_IPTUN_6RD_PREFIX */
1665                 nla_total_size(sizeof(struct in6_addr)) +
1666                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1667                 nla_total_size(4) +
1668                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1669                 nla_total_size(2) +
1670                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1671                 nla_total_size(2) +
1672 #endif
1673                 /* IFLA_IPTUN_ENCAP_TYPE */
1674                 nla_total_size(2) +
1675                 /* IFLA_IPTUN_ENCAP_FLAGS */
1676                 nla_total_size(2) +
1677                 /* IFLA_IPTUN_ENCAP_SPORT */
1678                 nla_total_size(2) +
1679                 /* IFLA_IPTUN_ENCAP_DPORT */
1680                 nla_total_size(2) +
1681                 /* IFLA_IPTUN_FWMARK */
1682                 nla_total_size(4) +
1683                 0;
1684 }
1685
1686 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1687 {
1688         struct ip_tunnel *tunnel = netdev_priv(dev);
1689         struct ip_tunnel_parm *parm = &tunnel->parms;
1690
1691         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1692             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1693             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1694             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1695             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1696             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1697                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1698             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1699             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1700             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1701                 goto nla_put_failure;
1702
1703 #ifdef CONFIG_IPV6_SIT_6RD
1704         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1705                              &tunnel->ip6rd.prefix) ||
1706             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1707                             tunnel->ip6rd.relay_prefix) ||
1708             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1709                         tunnel->ip6rd.prefixlen) ||
1710             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1711                         tunnel->ip6rd.relay_prefixlen))
1712                 goto nla_put_failure;
1713 #endif
1714
1715         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1716                         tunnel->encap.type) ||
1717             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1718                         tunnel->encap.sport) ||
1719             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1720                         tunnel->encap.dport) ||
1721             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1722                         tunnel->encap.flags))
1723                 goto nla_put_failure;
1724
1725         return 0;
1726
1727 nla_put_failure:
1728         return -EMSGSIZE;
1729 }
1730
1731 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1732         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1733         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1734         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1735         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1736         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1737         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1738         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1739         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1740 #ifdef CONFIG_IPV6_SIT_6RD
1741         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1742         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1743         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1744         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1745 #endif
1746         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1747         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1748         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1749         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1750         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1751 };
1752
1753 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1754 {
1755         struct net *net = dev_net(dev);
1756         struct sit_net *sitn = net_generic(net, sit_net_id);
1757
1758         if (dev != sitn->fb_tunnel_dev)
1759                 unregister_netdevice_queue(dev, head);
1760 }
1761
1762 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1763         .kind           = "sit",
1764         .maxtype        = IFLA_IPTUN_MAX,
1765         .policy         = ipip6_policy,
1766         .priv_size      = sizeof(struct ip_tunnel),
1767         .setup          = ipip6_tunnel_setup,
1768         .validate       = ipip6_validate,
1769         .newlink        = ipip6_newlink,
1770         .changelink     = ipip6_changelink,
1771         .get_size       = ipip6_get_size,
1772         .fill_info      = ipip6_fill_info,
1773         .dellink        = ipip6_dellink,
1774         .get_link_net   = ip_tunnel_get_link_net,
1775 };
1776
1777 static struct xfrm_tunnel sit_handler __read_mostly = {
1778         .handler        =       ipip6_rcv,
1779         .err_handler    =       ipip6_err,
1780         .priority       =       1,
1781 };
1782
1783 static struct xfrm_tunnel ipip_handler __read_mostly = {
1784         .handler        =       ipip_rcv,
1785         .err_handler    =       ipip6_err,
1786         .priority       =       2,
1787 };
1788
1789 #if IS_ENABLED(CONFIG_MPLS)
1790 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1791         .handler        =       mplsip_rcv,
1792         .err_handler    =       ipip6_err,
1793         .priority       =       2,
1794 };
1795 #endif
1796
1797 static void __net_exit sit_destroy_tunnels(struct net *net,
1798                                            struct list_head *head)
1799 {
1800         struct sit_net *sitn = net_generic(net, sit_net_id);
1801         struct net_device *dev, *aux;
1802         int prio;
1803
1804         for_each_netdev_safe(net, dev, aux)
1805                 if (dev->rtnl_link_ops == &sit_link_ops)
1806                         unregister_netdevice_queue(dev, head);
1807
1808         for (prio = 1; prio < 4; prio++) {
1809                 int h;
1810                 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1811                         struct ip_tunnel *t;
1812
1813                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1814                         while (t) {
1815                                 /* If dev is in the same netns, it has already
1816                                  * been added to the list by the previous loop.
1817                                  */
1818                                 if (!net_eq(dev_net(t->dev), net))
1819                                         unregister_netdevice_queue(t->dev,
1820                                                                    head);
1821                                 t = rtnl_dereference(t->next);
1822                         }
1823                 }
1824         }
1825 }
1826
1827 static int __net_init sit_init_net(struct net *net)
1828 {
1829         struct sit_net *sitn = net_generic(net, sit_net_id);
1830         struct ip_tunnel *t;
1831         int err;
1832
1833         sitn->tunnels[0] = sitn->tunnels_wc;
1834         sitn->tunnels[1] = sitn->tunnels_l;
1835         sitn->tunnels[2] = sitn->tunnels_r;
1836         sitn->tunnels[3] = sitn->tunnels_r_l;
1837
1838         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1839                                            NET_NAME_UNKNOWN,
1840                                            ipip6_tunnel_setup);
1841         if (!sitn->fb_tunnel_dev) {
1842                 err = -ENOMEM;
1843                 goto err_alloc_dev;
1844         }
1845         dev_net_set(sitn->fb_tunnel_dev, net);
1846         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1847         /* FB netdevice is special: we have one, and only one per netns.
1848          * Allowing to move it to another netns is clearly unsafe.
1849          */
1850         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1851
1852         err = register_netdev(sitn->fb_tunnel_dev);
1853         if (err)
1854                 goto err_reg_dev;
1855
1856         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1857         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1858
1859         t = netdev_priv(sitn->fb_tunnel_dev);
1860
1861         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1862         return 0;
1863
1864 err_reg_dev:
1865         ipip6_dev_free(sitn->fb_tunnel_dev);
1866 err_alloc_dev:
1867         return err;
1868 }
1869
1870 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1871 {
1872         LIST_HEAD(list);
1873         struct net *net;
1874
1875         rtnl_lock();
1876         list_for_each_entry(net, net_list, exit_list)
1877                 sit_destroy_tunnels(net, &list);
1878
1879         unregister_netdevice_many(&list);
1880         rtnl_unlock();
1881 }
1882
1883 static struct pernet_operations sit_net_ops = {
1884         .init = sit_init_net,
1885         .exit_batch = sit_exit_batch_net,
1886         .id   = &sit_net_id,
1887         .size = sizeof(struct sit_net),
1888 };
1889
1890 static void __exit sit_cleanup(void)
1891 {
1892         rtnl_link_unregister(&sit_link_ops);
1893         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1894         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1895 #if IS_ENABLED(CONFIG_MPLS)
1896         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1897 #endif
1898
1899         unregister_pernet_device(&sit_net_ops);
1900         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1901 }
1902
1903 static int __init sit_init(void)
1904 {
1905         int err;
1906
1907         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1908
1909         err = register_pernet_device(&sit_net_ops);
1910         if (err < 0)
1911                 return err;
1912         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1913         if (err < 0) {
1914                 pr_info("%s: can't register ip6ip4\n", __func__);
1915                 goto xfrm_tunnel_failed;
1916         }
1917         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1918         if (err < 0) {
1919                 pr_info("%s: can't register ip4ip4\n", __func__);
1920                 goto xfrm_tunnel4_failed;
1921         }
1922 #if IS_ENABLED(CONFIG_MPLS)
1923         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1924         if (err < 0) {
1925                 pr_info("%s: can't register mplsip\n", __func__);
1926                 goto xfrm_tunnel_mpls_failed;
1927         }
1928 #endif
1929         err = rtnl_link_register(&sit_link_ops);
1930         if (err < 0)
1931                 goto rtnl_link_failed;
1932
1933 out:
1934         return err;
1935
1936 rtnl_link_failed:
1937 #if IS_ENABLED(CONFIG_MPLS)
1938         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1939 xfrm_tunnel_mpls_failed:
1940 #endif
1941         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1942 xfrm_tunnel4_failed:
1943         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1944 xfrm_tunnel_failed:
1945         unregister_pernet_device(&sit_net_ops);
1946         goto out;
1947 }
1948
1949 module_init(sit_init);
1950 module_exit(sit_cleanup);
1951 MODULE_LICENSE("GPL");
1952 MODULE_ALIAS_RTNL_LINK("sit");
1953 MODULE_ALIAS_NETDEV("sit0");