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             <[email protected]>
   7  *      Alexey Kuznetsov        <[email protected]>
   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 <[email protected]>:       6to4 support
  16  * Nate Thompson <[email protected]>:             6to4 support
  17  * Fred Templin <[email protected]>:    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 || !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                 if (!dev_valid_name(parms->name))
 255                         goto failed;
 256                 strlcpy(name, parms->name, IFNAMSIZ);
 257         } else {
 258                 strcpy(name, "sit%d");
 259         }
 260         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 261                            ipip6_tunnel_setup);
 262         if (!dev)
 263                 return NULL;
 264
 265         dev_net_set(dev, net);
 266
 267         nt = netdev_priv(dev);
 268
 269         nt->parms = *parms;
 270         if (ipip6_tunnel_create(dev) < 0)
 271                 goto failed_free;
 272
 273         return nt;
 274
 275 failed_free:
 276         free_netdev(dev);
 277 failed:
 278         return NULL;
 279 }
 280
 281 #define for_each_prl_rcu(start)                 \
 282         for (prl = rcu_dereference(start);      \
 283              prl;                               \
 284              prl = rcu_dereference(prl->next))
 285
 286 static struct ip_tunnel_prl_entry *
 287 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 288 {
 289         struct ip_tunnel_prl_entry *prl;
 290
 291         for_each_prl_rcu(t->prl)
 292                 if (prl->addr == addr)
 293                         break;
 294         return prl;
 295
 296 }
 297
 298 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 299                                 struct ip_tunnel_prl __user *a)
 300 {
 301         struct ip_tunnel_prl kprl, *kp;
 302         struct ip_tunnel_prl_entry *prl;
 303         unsigned int cmax, c = 0, ca, len;
 304         int ret = 0;
 305
 306         if (copy_from_user(&kprl, a, sizeof(kprl)))
 307                 return -EFAULT;
 308         cmax = kprl.datalen / sizeof(kprl);
 309         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 310                 cmax = 1;
 311
 312         /* For simple GET or for root users,
 313          * we try harder to allocate.
 314          */
 315         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 316                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 317                 NULL;
 318
 319         rcu_read_lock();
 320
 321         ca = t->prl_count < cmax ? t->prl_count : cmax;
 322
 323         if (!kp) {
 324                 /* We don't try hard to allocate much memory for
 325                  * non-root users.
 326                  * For root users, retry allocating enough memory for
 327                  * the answer.
 328                  */
 329                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 330                 if (!kp) {
 331                         ret = -ENOMEM;
 332                         goto out;
 333                 }
 334         }
 335
 336         c = 0;
 337         for_each_prl_rcu(t->prl) {
 338                 if (c >= cmax)
 339                         break;
 340                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 341                         continue;
 342                 kp[c].addr = prl->addr;
 343                 kp[c].flags = prl->flags;
 344                 c++;
 345                 if (kprl.addr != htonl(INADDR_ANY))
 346                         break;
 347         }
 348 out:
 349         rcu_read_unlock();
 350
 351         len = sizeof(*kp) * c;
 352         ret = 0;
 353         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 354                 ret = -EFAULT;
 355
 356         kfree(kp);
 357
 358         return ret;
 359 }
 360
 361 static int
 362 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 363 {
 364         struct ip_tunnel_prl_entry *p;
 365         int err = 0;
 366
 367         if (a->addr == htonl(INADDR_ANY))
 368                 return -EINVAL;
 369
 370         ASSERT_RTNL();
 371
 372         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 373                 if (p->addr == a->addr) {
 374                         if (chg) {
 375                                 p->flags = a->flags;
 376                                 goto out;
 377                         }
 378                         err = -EEXIST;
 379                         goto out;
 380                 }
 381         }
 382
 383         if (chg) {
 384                 err = -ENXIO;
 385                 goto out;
 386         }
 387
 388         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 389         if (!p) {
 390                 err = -ENOBUFS;
 391                 goto out;
 392         }
 393
 394         p->next = t->prl;
 395         p->addr = a->addr;
 396         p->flags = a->flags;
 397         t->prl_count++;
 398         rcu_assign_pointer(t->prl, p);
 399 out:
 400         return err;
 401 }
 402
 403 static void prl_list_destroy_rcu(struct rcu_head *head)
 404 {
 405         struct ip_tunnel_prl_entry *p, *n;
 406
 407         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 408         do {
 409                 n = rcu_dereference_protected(p->next, 1);
 410                 kfree(p);
 411                 p = n;
 412         } while (p);
 413 }
 414
 415 static int
 416 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 417 {
 418         struct ip_tunnel_prl_entry *x;
 419         struct ip_tunnel_prl_entry __rcu **p;
 420         int err = 0;
 421
 422         ASSERT_RTNL();
 423
 424         if (a && a->addr != htonl(INADDR_ANY)) {
 425                 for (p = &t->prl;
 426                      (x = rtnl_dereference(*p)) != NULL;
 427                      p = &x->next) {
 428                         if (x->addr == a->addr) {
 429                                 *p = x->next;
 430                                 kfree_rcu(x, rcu_head);
 431                                 t->prl_count--;
 432                                 goto out;
 433                         }
 434                 }
 435                 err = -ENXIO;
 436         } else {
 437                 x = rtnl_dereference(t->prl);
 438                 if (x) {
 439                         t->prl_count = 0;
 440                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 441                         t->prl = NULL;
 442                 }
 443         }
 444 out:
 445         return err;
 446 }
 447
 448 static int
 449 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 450 {
 451         struct ip_tunnel_prl_entry *p;
 452         int ok = 1;
 453
 454         rcu_read_lock();
 455         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 456         if (p) {
 457                 if (p->flags & PRL_DEFAULT)
 458                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 459                 else
 460                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 461         } else {
 462                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 463
 464                 if (ipv6_addr_is_isatap(addr6) &&
 465                     (addr6->s6_addr32[3] == iph->saddr) &&
 466                     ipv6_chk_prefix(addr6, t->dev))
 467                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 468                 else
 469                         ok = 0;
 470         }
 471         rcu_read_unlock();
 472         return ok;
 473 }
 474
 475 static void ipip6_tunnel_uninit(struct net_device *dev)
 476 {
 477         struct ip_tunnel *tunnel = netdev_priv(dev);
 478         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 479
 480         if (dev == sitn->fb_tunnel_dev) {
 481                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 482         } else {
 483                 ipip6_tunnel_unlink(sitn, tunnel);
 484                 ipip6_tunnel_del_prl(tunnel, NULL);
 485         }
 486         dst_cache_reset(&tunnel->dst_cache);
 487         dev_put(dev);
 488 }
 489
 490 static int ipip6_err(struct sk_buff *skb, u32 info)
 491 {
 492         const struct iphdr *iph = (const struct iphdr *)skb->data;
 493         const int type = icmp_hdr(skb)->type;
 494         const int code = icmp_hdr(skb)->code;
 495         unsigned int data_len = 0;
 496         struct ip_tunnel *t;
 497         int sifindex;
 498         int err;
 499
 500         switch (type) {
 501         default:
 502         case ICMP_PARAMETERPROB:
 503                 return 0;
 504
 505         case ICMP_DEST_UNREACH:
 506                 switch (code) {
 507                 case ICMP_SR_FAILED:
 508                         /* Impossible event. */
 509                         return 0;
 510                 default:
 511                         /* All others are translated to HOST_UNREACH.
 512                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 513                            I believe they are just ether pollution. --ANK
 514                          */
 515                         break;
 516                 }
 517                 break;
 518         case ICMP_TIME_EXCEEDED:
 519                 if (code != ICMP_EXC_TTL)
 520                         return 0;
 521                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 522                 break;
 523         case ICMP_REDIRECT:
 524                 break;
 525         }
 526
 527         err = -ENOENT;
 528
 529         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 530         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 531                                 iph->daddr, iph->saddr, sifindex);
 532         if (!t)
 533                 goto out;
 534
 535         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 536                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 537                                  t->parms.link, iph->protocol);
 538                 err = 0;
 539                 goto out;
 540         }
 541         if (type == ICMP_REDIRECT) {
 542                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 543                               iph->protocol);
 544                 err = 0;
 545                 goto out;
 546         }
 547
 548         err = 0;
 549         if (__in6_dev_get(skb->dev) &&
 550             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 551                 goto out;
 552
 553         if (t->parms.iph.daddr == 0)
 554                 goto out;
 555
 556         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 557                 goto out;
 558
 559         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 560                 t->err_count++;
 561         else
 562                 t->err_count = 1;
 563         t->err_time = jiffies;
 564 out:
 565         return err;
 566 }
 567
 568 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 569                                   const struct in6_addr *v6addr)
 570 {
 571         __be32 v4embed = 0;
 572         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 573                 return true;
 574         return false;
 575 }
 576
 577 /* Checks if an address matches an address on the tunnel interface.
 578  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 579  * Long story:
 580  * This function is called after we considered the packet as spoofed
 581  * in is_spoofed_6rd.
 582  * We may have a router that is doing NAT for proto 41 packets
 583  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 584  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 585  * function will return true, dropping the packet.
 586  * But, we can still check if is spoofed against the IP
 587  * addresses associated with the interface.
 588  */
 589 static bool only_dnatted(const struct ip_tunnel *tunnel,
 590         const struct in6_addr *v6dst)
 591 {
 592         int prefix_len;
 593
 594 #ifdef CONFIG_IPV6_SIT_6RD
 595         prefix_len = tunnel->ip6rd.prefixlen + 32
 596                 - tunnel->ip6rd.relay_prefixlen;
 597 #else
 598         prefix_len = 48;
 599 #endif
 600         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 601 }
 602
 603 /* Returns true if a packet is spoofed */
 604 static bool packet_is_spoofed(struct sk_buff *skb,
 605                               const struct iphdr *iph,
 606                               struct ip_tunnel *tunnel)
 607 {
 608         const struct ipv6hdr *ipv6h;
 609
 610         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 611                 if (!isatap_chksrc(skb, iph, tunnel))
 612                         return true;
 613
 614                 return false;
 615         }
 616
 617         if (tunnel->dev->flags & IFF_POINTOPOINT)
 618                 return false;
 619
 620         ipv6h = ipv6_hdr(skb);
 621
 622         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 623                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 624                                      &iph->saddr, &ipv6h->saddr,
 625                                      &iph->daddr, &ipv6h->daddr);
 626                 return true;
 627         }
 628
 629         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 630                 return false;
 631
 632         if (only_dnatted(tunnel, &ipv6h->daddr))
 633                 return false;
 634
 635         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 636                              &iph->saddr, &ipv6h->saddr,
 637                              &iph->daddr, &ipv6h->daddr);
 638         return true;
 639 }
 640
 641 static int ipip6_rcv(struct sk_buff *skb)
 642 {
 643         const struct iphdr *iph = ip_hdr(skb);
 644         struct ip_tunnel *tunnel;
 645         int sifindex;
 646         int err;
 647
 648         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 649         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 650                                      iph->saddr, iph->daddr, sifindex);
 651         if (tunnel) {
 652                 struct pcpu_sw_netstats *tstats;
 653
 654                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 655                     tunnel->parms.iph.protocol != 0)
 656                         goto out;
 657
 658                 skb->mac_header = skb->network_header;
 659                 skb_reset_network_header(skb);
 660                 IPCB(skb)->flags = 0;
 661                 skb->dev = tunnel->dev;
 662
 663                 if (packet_is_spoofed(skb, iph, tunnel)) {
 664                         tunnel->dev->stats.rx_errors++;
 665                         goto out;
 666                 }
 667
 668                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 669                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 670                         goto out;
 671
 672                 err = IP_ECN_decapsulate(iph, skb);
 673                 if (unlikely(err)) {
 674                         if (log_ecn_error)
 675                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 676                                                      &iph->saddr, iph->tos);
 677                         if (err > 1) {
 678                                 ++tunnel->dev->stats.rx_frame_errors;
 679                                 ++tunnel->dev->stats.rx_errors;
 680                                 goto out;
 681                         }
 682                 }
 683
 684                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 685                 u64_stats_update_begin(&tstats->syncp);
 686                 tstats->rx_packets++;
 687                 tstats->rx_bytes += skb->len;
 688                 u64_stats_update_end(&tstats->syncp);
 689
 690                 netif_rx(skb);
 691
 692                 return 0;
 693         }
 694
 695         /* no tunnel matched,  let upstream know, ipsec may handle it */
 696         return 1;
 697 out:
 698         kfree_skb(skb);
 699         return 0;
 700 }
 701
 702 static const struct tnl_ptk_info ipip_tpi = {
 703         /* no tunnel info required for ipip. */
 704         .proto = htons(ETH_P_IP),
 705 };
 706
 707 #if IS_ENABLED(CONFIG_MPLS)
 708 static const struct tnl_ptk_info mplsip_tpi = {
 709         /* no tunnel info required for mplsip. */
 710         .proto = htons(ETH_P_MPLS_UC),
 711 };
 712 #endif
 713
 714 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 715 {
 716         const struct iphdr *iph;
 717         struct ip_tunnel *tunnel;
 718         int sifindex;
 719
 720         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 721
 722         iph = ip_hdr(skb);
 723         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 724                                      iph->saddr, iph->daddr, sifindex);
 725         if (tunnel) {
 726                 const struct tnl_ptk_info *tpi;
 727
 728                 if (tunnel->parms.iph.protocol != ipproto &&
 729                     tunnel->parms.iph.protocol != 0)
 730                         goto drop;
 731
 732                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 733                         goto drop;
 734 #if IS_ENABLED(CONFIG_MPLS)
 735                 if (ipproto == IPPROTO_MPLS)
 736                         tpi = &mplsip_tpi;
 737                 else
 738 #endif
 739                         tpi = &ipip_tpi;
 740                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 741                         goto drop;
 742                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 743         }
 744
 745         return 1;
 746
 747 drop:
 748         kfree_skb(skb);
 749         return 0;
 750 }
 751
 752 static int ipip_rcv(struct sk_buff *skb)
 753 {
 754         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 755 }
 756
 757 #if IS_ENABLED(CONFIG_MPLS)
 758 static int mplsip_rcv(struct sk_buff *skb)
 759 {
 760         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 761 }
 762 #endif
 763
 764 /*
 765  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 766  * stores the embedded IPv4 address in v4dst and returns true.
 767  */
 768 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 769                       __be32 *v4dst)
 770 {
 771 #ifdef CONFIG_IPV6_SIT_6RD
 772         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 773                               tunnel->ip6rd.prefixlen)) {
 774                 unsigned int pbw0, pbi0;
 775                 int pbi1;
 776                 u32 d;
 777
 778                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 779                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 780
 781                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 782                     tunnel->ip6rd.relay_prefixlen;
 783
 784                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 785                 if (pbi1 > 0)
 786                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 787                              (32 - pbi1);
 788
 789                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 790                 return true;
 791         }
 792 #else
 793         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 794                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 795                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 796                 return true;
 797         }
 798 #endif
 799         return false;
 800 }
 801
 802 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 803                              const struct in6_addr *v6dst)
 804 {
 805         __be32 dst = 0;
 806         check_6rd(tunnel, v6dst, &dst);
 807         return dst;
 808 }
 809
 810 /*
 811  *      This function assumes it is being called from dev_queue_xmit()
 812  *      and that skb is filled properly by that function.
 813  */
 814
 815 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 816                                      struct net_device *dev)
 817 {
 818         struct ip_tunnel *tunnel = netdev_priv(dev);
 819         const struct iphdr  *tiph = &tunnel->parms.iph;
 820         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 821         u8     tos = tunnel->parms.iph.tos;
 822         __be16 df = tiph->frag_off;
 823         struct rtable *rt;              /* Route to the other host */
 824         struct net_device *tdev;        /* Device to other host */
 825         unsigned int max_headroom;      /* The extra header space needed */
 826         __be32 dst = tiph->daddr;
 827         struct flowi4 fl4;
 828         int    mtu;
 829         const struct in6_addr *addr6;
 830         int addr_type;
 831         u8 ttl;
 832         u8 protocol = IPPROTO_IPV6;
 833         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 834
 835         if (tos == 1)
 836                 tos = ipv6_get_dsfield(iph6);
 837
 838         /* ISATAP (RFC4214) - must come before 6to4 */
 839         if (dev->priv_flags & IFF_ISATAP) {
 840                 struct neighbour *neigh = NULL;
 841                 bool do_tx_error = false;
 842
 843                 if (skb_dst(skb))
 844                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 845
 846                 if (!neigh) {
 847                         net_dbg_ratelimited("nexthop == NULL\n");
 848                         goto tx_error;
 849                 }
 850
 851                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 852                 addr_type = ipv6_addr_type(addr6);
 853
 854                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 855                      ipv6_addr_is_isatap(addr6))
 856                         dst = addr6->s6_addr32[3];
 857                 else
 858                         do_tx_error = true;
 859
 860                 neigh_release(neigh);
 861                 if (do_tx_error)
 862                         goto tx_error;
 863         }
 864
 865         if (!dst)
 866                 dst = try_6rd(tunnel, &iph6->daddr);
 867
 868         if (!dst) {
 869                 struct neighbour *neigh = NULL;
 870                 bool do_tx_error = false;
 871
 872                 if (skb_dst(skb))
 873                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 874
 875                 if (!neigh) {
 876                         net_dbg_ratelimited("nexthop == NULL\n");
 877                         goto tx_error;
 878                 }
 879
 880                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 881                 addr_type = ipv6_addr_type(addr6);
 882
 883                 if (addr_type == IPV6_ADDR_ANY) {
 884                         addr6 = &ipv6_hdr(skb)->daddr;
 885                         addr_type = ipv6_addr_type(addr6);
 886                 }
 887
 888                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 889                         dst = addr6->s6_addr32[3];
 890                 else
 891                         do_tx_error = true;
 892
 893                 neigh_release(neigh);
 894                 if (do_tx_error)
 895                         goto tx_error;
 896         }
 897
 898         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 899                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 900                            0, dst, tiph->saddr, 0, 0,
 901                            sock_net_uid(tunnel->net, NULL));
 902         rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 903
 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         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 922                 ip_rt_put(rt);
 923                 goto tx_error;
 924         }
 925
 926         if (df) {
 927                 mtu = dst_mtu(&rt->dst) - t_hlen;
 928
 929                 if (mtu < 68) {
 930                         dev->stats.collisions++;
 931                         ip_rt_put(rt);
 932                         goto tx_error;
 933                 }
 934
 935                 if (mtu < IPV6_MIN_MTU) {
 936                         mtu = IPV6_MIN_MTU;
 937                         df = 0;
 938                 }
 939
 940                 if (tunnel->parms.iph.daddr)
 941                         skb_dst_update_pmtu(skb, mtu);
 942
 943                 if (skb->len > mtu && !skb_is_gso(skb)) {
 944                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 945                         ip_rt_put(rt);
 946                         goto tx_error;
 947                 }
 948         }
 949
 950         if (tunnel->err_count > 0) {
 951                 if (time_before(jiffies,
 952                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 953                         tunnel->err_count--;
 954                         dst_link_failure(skb);
 955                 } else
 956                         tunnel->err_count = 0;
 957         }
 958
 959         /*
 960          * Okay, now see if we can stuff it in the buffer as-is.
 961          */
 962         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 963
 964         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 965             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 966                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 967                 if (!new_skb) {
 968                         ip_rt_put(rt);
 969                         dev->stats.tx_dropped++;
 970                         kfree_skb(skb);
 971                         return NETDEV_TX_OK;
 972                 }
 973                 if (skb->sk)
 974                         skb_set_owner_w(new_skb, skb->sk);
 975                 dev_kfree_skb(skb);
 976                 skb = new_skb;
 977                 iph6 = ipv6_hdr(skb);
 978         }
 979         ttl = tiph->ttl;
 980         if (ttl == 0)
 981                 ttl = iph6->hop_limit;
 982         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 983
 984         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 985                 ip_rt_put(rt);
 986                 goto tx_error;
 987         }
 988
 989         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 990
 991         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 992                       df, !net_eq(tunnel->net, dev_net(dev)));
 993         return NETDEV_TX_OK;
 994
 995 tx_error_icmp:
 996         dst_link_failure(skb);
 997 tx_error:
 998         kfree_skb(skb);
 999         dev->stats.tx_errors++;
1000         return NETDEV_TX_OK;
1001 }
1002
1003 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1004                                      struct net_device *dev, u8 ipproto)
1005 {
1006         struct ip_tunnel *tunnel = netdev_priv(dev);
1007         const struct iphdr  *tiph = &tunnel->parms.iph;
1008
1009         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1010                 goto tx_error;
1011
1012         skb_set_inner_ipproto(skb, ipproto);
1013
1014         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1015         return NETDEV_TX_OK;
1016 tx_error:
1017         kfree_skb(skb);
1018         dev->stats.tx_errors++;
1019         return NETDEV_TX_OK;
1020 }
1021
1022 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1023                                    struct net_device *dev)
1024 {
1025         if (!pskb_inet_may_pull(skb))
1026                 goto tx_err;
1027
1028         switch (skb->protocol) {
1029         case htons(ETH_P_IP):
1030                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1031                 break;
1032         case htons(ETH_P_IPV6):
1033                 ipip6_tunnel_xmit(skb, dev);
1034                 break;
1035 #if IS_ENABLED(CONFIG_MPLS)
1036         case htons(ETH_P_MPLS_UC):
1037                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1038                 break;
1039 #endif
1040         default:
1041                 goto tx_err;
1042         }
1043
1044         return NETDEV_TX_OK;
1045
1046 tx_err:
1047         dev->stats.tx_errors++;
1048         kfree_skb(skb);
1049         return NETDEV_TX_OK;
1050
1051 }
1052
1053 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1054 {
1055         struct net_device *tdev = NULL;
1056         struct ip_tunnel *tunnel;
1057         const struct iphdr *iph;
1058         struct flowi4 fl4;
1059
1060         tunnel = netdev_priv(dev);
1061         iph = &tunnel->parms.iph;
1062
1063         if (iph->daddr) {
1064                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1065                                                           NULL,
1066                                                           iph->daddr, iph->saddr,
1067                                                           0, 0,
1068                                                           IPPROTO_IPV6,
1069                                                           RT_TOS(iph->tos),
1070                                                           tunnel->parms.link);
1071
1072                 if (!IS_ERR(rt)) {
1073                         tdev = rt->dst.dev;
1074                         ip_rt_put(rt);
1075                 }
1076                 dev->flags |= IFF_POINTOPOINT;
1077         }
1078
1079         if (!tdev && tunnel->parms.link)
1080                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1081
1082         if (tdev) {
1083                 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1084
1085                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1086                 dev->mtu = tdev->mtu - t_hlen;
1087                 if (dev->mtu < IPV6_MIN_MTU)
1088                         dev->mtu = IPV6_MIN_MTU;
1089         }
1090 }
1091
1092 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1093                                 __u32 fwmark)
1094 {
1095         struct net *net = t->net;
1096         struct sit_net *sitn = net_generic(net, sit_net_id);
1097
1098         ipip6_tunnel_unlink(sitn, t);
1099         synchronize_net();
1100         t->parms.iph.saddr = p->iph.saddr;
1101         t->parms.iph.daddr = p->iph.daddr;
1102         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1103         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1104         ipip6_tunnel_link(sitn, t);
1105         t->parms.iph.ttl = p->iph.ttl;
1106         t->parms.iph.tos = p->iph.tos;
1107         t->parms.iph.frag_off = p->iph.frag_off;
1108         if (t->parms.link != p->link || t->fwmark != fwmark) {
1109                 t->parms.link = p->link;
1110                 t->fwmark = fwmark;
1111                 ipip6_tunnel_bind_dev(t->dev);
1112         }
1113         dst_cache_reset(&t->dst_cache);
1114         netdev_state_change(t->dev);
1115 }
1116
1117 #ifdef CONFIG_IPV6_SIT_6RD
1118 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1119                                    struct ip_tunnel_6rd *ip6rd)
1120 {
1121         struct in6_addr prefix;
1122         __be32 relay_prefix;
1123
1124         if (ip6rd->relay_prefixlen > 32 ||
1125             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1126                 return -EINVAL;
1127
1128         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1129         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1130                 return -EINVAL;
1131         if (ip6rd->relay_prefixlen)
1132                 relay_prefix = ip6rd->relay_prefix &
1133                                htonl(0xffffffffUL <<
1134                                      (32 - ip6rd->relay_prefixlen));
1135         else
1136                 relay_prefix = 0;
1137         if (relay_prefix != ip6rd->relay_prefix)
1138                 return -EINVAL;
1139
1140         t->ip6rd.prefix = prefix;
1141         t->ip6rd.relay_prefix = relay_prefix;
1142         t->ip6rd.prefixlen = ip6rd->prefixlen;
1143         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1144         dst_cache_reset(&t->dst_cache);
1145         netdev_state_change(t->dev);
1146         return 0;
1147 }
1148 #endif
1149
1150 static bool ipip6_valid_ip_proto(u8 ipproto)
1151 {
1152         return ipproto == IPPROTO_IPV6 ||
1153                 ipproto == IPPROTO_IPIP ||
1154 #if IS_ENABLED(CONFIG_MPLS)
1155                 ipproto == IPPROTO_MPLS ||
1156 #endif
1157                 ipproto == 0;
1158 }
1159
1160 static int
1161 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1162 {
1163         int err = 0;
1164         struct ip_tunnel_parm p;
1165         struct ip_tunnel_prl prl;
1166         struct ip_tunnel *t = netdev_priv(dev);
1167         struct net *net = t->net;
1168         struct sit_net *sitn = net_generic(net, sit_net_id);
1169 #ifdef CONFIG_IPV6_SIT_6RD
1170         struct ip_tunnel_6rd ip6rd;
1171 #endif
1172
1173         switch (cmd) {
1174         case SIOCGETTUNNEL:
1175 #ifdef CONFIG_IPV6_SIT_6RD
1176         case SIOCGET6RD:
1177 #endif
1178                 if (dev == sitn->fb_tunnel_dev) {
1179                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1180                                 err = -EFAULT;
1181                                 break;
1182                         }
1183                         t = ipip6_tunnel_locate(net, &p, 0);
1184                         if (!t)
1185                                 t = netdev_priv(dev);
1186                 }
1187
1188                 err = -EFAULT;
1189                 if (cmd == SIOCGETTUNNEL) {
1190                         memcpy(&p, &t->parms, sizeof(p));
1191                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1192                                          sizeof(p)))
1193                                 goto done;
1194 #ifdef CONFIG_IPV6_SIT_6RD
1195                 } else {
1196                         ip6rd.prefix = t->ip6rd.prefix;
1197                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1198                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1199                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1200                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1201                                          sizeof(ip6rd)))
1202                                 goto done;
1203 #endif
1204                 }
1205                 err = 0;
1206                 break;
1207
1208         case SIOCADDTUNNEL:
1209         case SIOCCHGTUNNEL:
1210                 err = -EPERM;
1211                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1212                         goto done;
1213
1214                 err = -EFAULT;
1215                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1216                         goto done;
1217
1218                 err = -EINVAL;
1219                 if (!ipip6_valid_ip_proto(p.iph.protocol))
1220                         goto done;
1221                 if (p.iph.version != 4 ||
1222                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1223                         goto done;
1224                 if (p.iph.ttl)
1225                         p.iph.frag_off |= htons(IP_DF);
1226
1227                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1228
1229                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1230                         if (t) {
1231                                 if (t->dev != dev) {
1232                                         err = -EEXIST;
1233                                         break;
1234                                 }
1235                         } else {
1236                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1237                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1238                                         err = -EINVAL;
1239                                         break;
1240                                 }
1241                                 t = netdev_priv(dev);
1242                         }
1243
1244                         ipip6_tunnel_update(t, &p, t->fwmark);
1245                 }
1246
1247                 if (t) {
1248                         err = 0;
1249                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1250                                 err = -EFAULT;
1251                 } else
1252                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1253                 break;
1254
1255         case SIOCDELTUNNEL:
1256                 err = -EPERM;
1257                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1258                         goto done;
1259
1260                 if (dev == sitn->fb_tunnel_dev) {
1261                         err = -EFAULT;
1262                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1263                                 goto done;
1264                         err = -ENOENT;
1265                         t = ipip6_tunnel_locate(net, &p, 0);
1266                         if (!t)
1267                                 goto done;
1268                         err = -EPERM;
1269                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1270                                 goto done;
1271                         dev = t->dev;
1272                 }
1273                 unregister_netdevice(dev);
1274                 err = 0;
1275                 break;
1276
1277         case SIOCGETPRL:
1278                 err = -EINVAL;
1279                 if (dev == sitn->fb_tunnel_dev)
1280                         goto done;
1281                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1282                 break;
1283
1284         case SIOCADDPRL:
1285         case SIOCDELPRL:
1286         case SIOCCHGPRL:
1287                 err = -EPERM;
1288                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1289                         goto done;
1290                 err = -EINVAL;
1291                 if (dev == sitn->fb_tunnel_dev)
1292                         goto done;
1293                 err = -EFAULT;
1294                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1295                         goto done;
1296
1297                 switch (cmd) {
1298                 case SIOCDELPRL:
1299                         err = ipip6_tunnel_del_prl(t, &prl);
1300                         break;
1301                 case SIOCADDPRL:
1302                 case SIOCCHGPRL:
1303                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1304                         break;
1305                 }
1306                 dst_cache_reset(&t->dst_cache);
1307                 netdev_state_change(dev);
1308                 break;
1309
1310 #ifdef CONFIG_IPV6_SIT_6RD
1311         case SIOCADD6RD:
1312         case SIOCCHG6RD:
1313         case SIOCDEL6RD:
1314                 err = -EPERM;
1315                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1316                         goto done;
1317
1318                 err = -EFAULT;
1319                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1320                                    sizeof(ip6rd)))
1321                         goto done;
1322
1323                 if (cmd != SIOCDEL6RD) {
1324                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1325                         if (err < 0)
1326                                 goto done;
1327                 } else
1328                         ipip6_tunnel_clone_6rd(dev, sitn);
1329
1330                 err = 0;
1331                 break;
1332 #endif
1333
1334         default:
1335                 err = -EINVAL;
1336         }
1337
1338 done:
1339         return err;
1340 }
1341
1342 static const struct net_device_ops ipip6_netdev_ops = {
1343         .ndo_init       = ipip6_tunnel_init,
1344         .ndo_uninit     = ipip6_tunnel_uninit,
1345         .ndo_start_xmit = sit_tunnel_xmit,
1346         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1347         .ndo_get_stats64 = ip_tunnel_get_stats64,
1348         .ndo_get_iflink = ip_tunnel_get_iflink,
1349 };
1350
1351 static void ipip6_dev_free(struct net_device *dev)
1352 {
1353         struct ip_tunnel *tunnel = netdev_priv(dev);
1354
1355         dst_cache_destroy(&tunnel->dst_cache);
1356         free_percpu(dev->tstats);
1357 }
1358
1359 #define SIT_FEATURES (NETIF_F_SG           | \
1360                       NETIF_F_FRAGLIST     | \
1361                       NETIF_F_HIGHDMA      | \
1362                       NETIF_F_GSO_SOFTWARE | \
1363                       NETIF_F_HW_CSUM)
1364
1365 static void ipip6_tunnel_setup(struct net_device *dev)
1366 {
1367         struct ip_tunnel *tunnel = netdev_priv(dev);
1368         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1369
1370         dev->netdev_ops         = &ipip6_netdev_ops;
1371         dev->needs_free_netdev  = true;
1372         dev->priv_destructor    = ipip6_dev_free;
1373
1374         dev->type               = ARPHRD_SIT;
1375         dev->hard_header_len    = LL_MAX_HEADER + t_hlen;
1376         dev->mtu                = ETH_DATA_LEN - t_hlen;
1377         dev->min_mtu            = IPV6_MIN_MTU;
1378         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1379         dev->flags              = IFF_NOARP;
1380         netif_keep_dst(dev);
1381         dev->addr_len           = 4;
1382         dev->features           |= NETIF_F_LLTX;
1383         dev->features           |= SIT_FEATURES;
1384         dev->hw_features        |= SIT_FEATURES;
1385 }
1386
1387 static int ipip6_tunnel_init(struct net_device *dev)
1388 {
1389         struct ip_tunnel *tunnel = netdev_priv(dev);
1390         int err;
1391
1392         tunnel->dev = dev;
1393         tunnel->net = dev_net(dev);
1394         strcpy(tunnel->parms.name, dev->name);
1395
1396         ipip6_tunnel_bind_dev(dev);
1397         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1398         if (!dev->tstats)
1399                 return -ENOMEM;
1400
1401         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1402         if (err) {
1403                 free_percpu(dev->tstats);
1404                 dev->tstats = NULL;
1405                 return err;
1406         }
1407
1408         return 0;
1409 }
1410
1411 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1412 {
1413         struct ip_tunnel *tunnel = netdev_priv(dev);
1414         struct iphdr *iph = &tunnel->parms.iph;
1415         struct net *net = dev_net(dev);
1416         struct sit_net *sitn = net_generic(net, sit_net_id);
1417
1418         iph->version            = 4;
1419         iph->protocol           = IPPROTO_IPV6;
1420         iph->ihl                = 5;
1421         iph->ttl                = 64;
1422
1423         dev_hold(dev);
1424         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1425 }
1426
1427 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1428                           struct netlink_ext_ack *extack)
1429 {
1430         u8 proto;
1431
1432         if (!data || !data[IFLA_IPTUN_PROTO])
1433                 return 0;
1434
1435         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1436         if (!ipip6_valid_ip_proto(proto))
1437                 return -EINVAL;
1438
1439         return 0;
1440 }
1441
1442 static void ipip6_netlink_parms(struct nlattr *data[],
1443                                 struct ip_tunnel_parm *parms,
1444                                 __u32 *fwmark)
1445 {
1446         memset(parms, 0, sizeof(*parms));
1447
1448         parms->iph.version = 4;
1449         parms->iph.protocol = IPPROTO_IPV6;
1450         parms->iph.ihl = 5;
1451         parms->iph.ttl = 64;
1452
1453         if (!data)
1454                 return;
1455
1456         if (data[IFLA_IPTUN_LINK])
1457                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1458
1459         if (data[IFLA_IPTUN_LOCAL])
1460                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1461
1462         if (data[IFLA_IPTUN_REMOTE])
1463                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1464
1465         if (data[IFLA_IPTUN_TTL]) {
1466                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1467                 if (parms->iph.ttl)
1468                         parms->iph.frag_off = htons(IP_DF);
1469         }
1470
1471         if (data[IFLA_IPTUN_TOS])
1472                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1473
1474         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1475                 parms->iph.frag_off = htons(IP_DF);
1476
1477         if (data[IFLA_IPTUN_FLAGS])
1478                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1479
1480         if (data[IFLA_IPTUN_PROTO])
1481                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1482
1483         if (data[IFLA_IPTUN_FWMARK])
1484                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1485 }
1486
1487 /* This function returns true when ENCAP attributes are present in the nl msg */
1488 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1489                                       struct ip_tunnel_encap *ipencap)
1490 {
1491         bool ret = false;
1492
1493         memset(ipencap, 0, sizeof(*ipencap));
1494
1495         if (!data)
1496                 return ret;
1497
1498         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1499                 ret = true;
1500                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1501         }
1502
1503         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1504                 ret = true;
1505                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1506         }
1507
1508         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1509                 ret = true;
1510                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1511         }
1512
1513         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1514                 ret = true;
1515                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1516         }
1517
1518         return ret;
1519 }
1520
1521 #ifdef CONFIG_IPV6_SIT_6RD
1522 /* This function returns true when 6RD attributes are present in the nl msg */
1523 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1524                                     struct ip_tunnel_6rd *ip6rd)
1525 {
1526         bool ret = false;
1527         memset(ip6rd, 0, sizeof(*ip6rd));
1528
1529         if (!data)
1530                 return ret;
1531
1532         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1533                 ret = true;
1534                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1535         }
1536
1537         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1538                 ret = true;
1539                 ip6rd->relay_prefix =
1540                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1541         }
1542
1543         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1544                 ret = true;
1545                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1546         }
1547
1548         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1549                 ret = true;
1550                 ip6rd->relay_prefixlen =
1551                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1552         }
1553
1554         return ret;
1555 }
1556 #endif
1557
1558 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1559                          struct nlattr *tb[], struct nlattr *data[],
1560                          struct netlink_ext_ack *extack)
1561 {
1562         struct net *net = dev_net(dev);
1563         struct ip_tunnel *nt;
1564         struct ip_tunnel_encap ipencap;
1565 #ifdef CONFIG_IPV6_SIT_6RD
1566         struct ip_tunnel_6rd ip6rd;
1567 #endif
1568         int err;
1569
1570         nt = netdev_priv(dev);
1571
1572         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1573                 err = ip_tunnel_encap_setup(nt, &ipencap);
1574                 if (err < 0)
1575                         return err;
1576         }
1577
1578         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1579
1580         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1581                 return -EEXIST;
1582
1583         err = ipip6_tunnel_create(dev);
1584         if (err < 0)
1585                 return err;
1586
1587         if (tb[IFLA_MTU]) {
1588                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1589
1590                 if (mtu >= IPV6_MIN_MTU &&
1591                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1592                         dev->mtu = mtu;
1593         }
1594
1595 #ifdef CONFIG_IPV6_SIT_6RD
1596         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1597                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1598 #endif
1599
1600         return err;
1601 }
1602
1603 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1604                             struct nlattr *data[],
1605                             struct netlink_ext_ack *extack)
1606 {
1607         struct ip_tunnel *t = netdev_priv(dev);
1608         struct ip_tunnel_parm p;
1609         struct ip_tunnel_encap ipencap;
1610         struct net *net = t->net;
1611         struct sit_net *sitn = net_generic(net, sit_net_id);
1612 #ifdef CONFIG_IPV6_SIT_6RD
1613         struct ip_tunnel_6rd ip6rd;
1614 #endif
1615         __u32 fwmark = t->fwmark;
1616         int err;
1617
1618         if (dev == sitn->fb_tunnel_dev)
1619                 return -EINVAL;
1620
1621         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1622                 err = ip_tunnel_encap_setup(t, &ipencap);
1623                 if (err < 0)
1624                         return err;
1625         }
1626
1627         ipip6_netlink_parms(data, &p, &fwmark);
1628
1629         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1630             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1631                 return -EINVAL;
1632
1633         t = ipip6_tunnel_locate(net, &p, 0);
1634
1635         if (t) {
1636                 if (t->dev != dev)
1637                         return -EEXIST;
1638         } else
1639                 t = netdev_priv(dev);
1640
1641         ipip6_tunnel_update(t, &p, fwmark);
1642
1643 #ifdef CONFIG_IPV6_SIT_6RD
1644         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1645                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1646 #endif
1647
1648         return 0;
1649 }
1650
1651 static size_t ipip6_get_size(const struct net_device *dev)
1652 {
1653         return
1654                 /* IFLA_IPTUN_LINK */
1655                 nla_total_size(4) +
1656                 /* IFLA_IPTUN_LOCAL */
1657                 nla_total_size(4) +
1658                 /* IFLA_IPTUN_REMOTE */
1659                 nla_total_size(4) +
1660                 /* IFLA_IPTUN_TTL */
1661                 nla_total_size(1) +
1662                 /* IFLA_IPTUN_TOS */
1663                 nla_total_size(1) +
1664                 /* IFLA_IPTUN_PMTUDISC */
1665                 nla_total_size(1) +
1666                 /* IFLA_IPTUN_FLAGS */
1667                 nla_total_size(2) +
1668                 /* IFLA_IPTUN_PROTO */
1669                 nla_total_size(1) +
1670 #ifdef CONFIG_IPV6_SIT_6RD
1671                 /* IFLA_IPTUN_6RD_PREFIX */
1672                 nla_total_size(sizeof(struct in6_addr)) +
1673                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1674                 nla_total_size(4) +
1675                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1676                 nla_total_size(2) +
1677                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1678                 nla_total_size(2) +
1679 #endif
1680                 /* IFLA_IPTUN_ENCAP_TYPE */
1681                 nla_total_size(2) +
1682                 /* IFLA_IPTUN_ENCAP_FLAGS */
1683                 nla_total_size(2) +
1684                 /* IFLA_IPTUN_ENCAP_SPORT */
1685                 nla_total_size(2) +
1686                 /* IFLA_IPTUN_ENCAP_DPORT */
1687                 nla_total_size(2) +
1688                 /* IFLA_IPTUN_FWMARK */
1689                 nla_total_size(4) +
1690                 0;
1691 }
1692
1693 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1694 {
1695         struct ip_tunnel *tunnel = netdev_priv(dev);
1696         struct ip_tunnel_parm *parm = &tunnel->parms;
1697
1698         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1699             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1700             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1701             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1702             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1703             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1704                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1705             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1706             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1707             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1708                 goto nla_put_failure;
1709
1710 #ifdef CONFIG_IPV6_SIT_6RD
1711         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1712                              &tunnel->ip6rd.prefix) ||
1713             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1714                             tunnel->ip6rd.relay_prefix) ||
1715             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1716                         tunnel->ip6rd.prefixlen) ||
1717             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1718                         tunnel->ip6rd.relay_prefixlen))
1719                 goto nla_put_failure;
1720 #endif
1721
1722         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1723                         tunnel->encap.type) ||
1724             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1725                         tunnel->encap.sport) ||
1726             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1727                         tunnel->encap.dport) ||
1728             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1729                         tunnel->encap.flags))
1730                 goto nla_put_failure;
1731
1732         return 0;
1733
1734 nla_put_failure:
1735         return -EMSGSIZE;
1736 }
1737
1738 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1739         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1740         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1741         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1742         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1743         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1744         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1745         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1746         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1747 #ifdef CONFIG_IPV6_SIT_6RD
1748         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1749         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1750         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1751         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1752 #endif
1753         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1754         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1755         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1756         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1757         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1758 };
1759
1760 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1761 {
1762         struct net *net = dev_net(dev);
1763         struct sit_net *sitn = net_generic(net, sit_net_id);
1764
1765         if (dev != sitn->fb_tunnel_dev)
1766                 unregister_netdevice_queue(dev, head);
1767 }
1768
1769 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1770         .kind           = "sit",
1771         .maxtype        = IFLA_IPTUN_MAX,
1772         .policy         = ipip6_policy,
1773         .priv_size      = sizeof(struct ip_tunnel),
1774         .setup          = ipip6_tunnel_setup,
1775         .validate       = ipip6_validate,
1776         .newlink        = ipip6_newlink,
1777         .changelink     = ipip6_changelink,
1778         .get_size       = ipip6_get_size,
1779         .fill_info      = ipip6_fill_info,
1780         .dellink        = ipip6_dellink,
1781         .get_link_net   = ip_tunnel_get_link_net,
1782 };
1783
1784 static struct xfrm_tunnel sit_handler __read_mostly = {
1785         .handler        =       ipip6_rcv,
1786         .err_handler    =       ipip6_err,
1787         .priority       =       1,
1788 };
1789
1790 static struct xfrm_tunnel ipip_handler __read_mostly = {
1791         .handler        =       ipip_rcv,
1792         .err_handler    =       ipip6_err,
1793         .priority       =       2,
1794 };
1795
1796 #if IS_ENABLED(CONFIG_MPLS)
1797 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1798         .handler        =       mplsip_rcv,
1799         .err_handler    =       ipip6_err,
1800         .priority       =       2,
1801 };
1802 #endif
1803
1804 static void __net_exit sit_destroy_tunnels(struct net *net,
1805                                            struct list_head *head)
1806 {
1807         struct sit_net *sitn = net_generic(net, sit_net_id);
1808         struct net_device *dev, *aux;
1809         int prio;
1810
1811         for_each_netdev_safe(net, dev, aux)
1812                 if (dev->rtnl_link_ops == &sit_link_ops)
1813                         unregister_netdevice_queue(dev, head);
1814
1815         for (prio = 1; prio < 4; prio++) {
1816                 int h;
1817                 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1818                         struct ip_tunnel *t;
1819
1820                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1821                         while (t) {
1822                                 /* If dev is in the same netns, it has already
1823                                  * been added to the list by the previous loop.
1824                                  */
1825                                 if (!net_eq(dev_net(t->dev), net))
1826                                         unregister_netdevice_queue(t->dev,
1827                                                                    head);
1828                                 t = rtnl_dereference(t->next);
1829                         }
1830                 }
1831         }
1832 }
1833
1834 static int __net_init sit_init_net(struct net *net)
1835 {
1836         struct sit_net *sitn = net_generic(net, sit_net_id);
1837         struct ip_tunnel *t;
1838         int err;
1839
1840         sitn->tunnels[0] = sitn->tunnels_wc;
1841         sitn->tunnels[1] = sitn->tunnels_l;
1842         sitn->tunnels[2] = sitn->tunnels_r;
1843         sitn->tunnels[3] = sitn->tunnels_r_l;
1844
1845         if (!net_has_fallback_tunnels(net))
1846                 return 0;
1847
1848         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1849                                            NET_NAME_UNKNOWN,
1850                                            ipip6_tunnel_setup);
1851         if (!sitn->fb_tunnel_dev) {
1852                 err = -ENOMEM;
1853                 goto err_alloc_dev;
1854         }
1855         dev_net_set(sitn->fb_tunnel_dev, net);
1856         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1857         /* FB netdevice is special: we have one, and only one per netns.
1858          * Allowing to move it to another netns is clearly unsafe.
1859          */
1860         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1861
1862         err = register_netdev(sitn->fb_tunnel_dev);
1863         if (err)
1864                 goto err_reg_dev;
1865
1866         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1867         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1868
1869         t = netdev_priv(sitn->fb_tunnel_dev);
1870
1871         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1872         return 0;
1873
1874 err_reg_dev:
1875         ipip6_dev_free(sitn->fb_tunnel_dev);
1876 err_alloc_dev:
1877         return err;
1878 }
1879
1880 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1881 {
1882         LIST_HEAD(list);
1883         struct net *net;
1884
1885         rtnl_lock();
1886         list_for_each_entry(net, net_list, exit_list)
1887                 sit_destroy_tunnels(net, &list);
1888
1889         unregister_netdevice_many(&list);
1890         rtnl_unlock();
1891 }
1892
1893 static struct pernet_operations sit_net_ops = {
1894         .init = sit_init_net,
1895         .exit_batch = sit_exit_batch_net,
1896         .id   = &sit_net_id,
1897         .size = sizeof(struct sit_net),
1898 };
1899
1900 static void __exit sit_cleanup(void)
1901 {
1902         rtnl_link_unregister(&sit_link_ops);
1903         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1904         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1905 #if IS_ENABLED(CONFIG_MPLS)
1906         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1907 #endif
1908
1909         unregister_pernet_device(&sit_net_ops);
1910         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1911 }
1912
1913 static int __init sit_init(void)
1914 {
1915         int err;
1916
1917         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1918
1919         err = register_pernet_device(&sit_net_ops);
1920         if (err < 0)
1921                 return err;
1922         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1923         if (err < 0) {
1924                 pr_info("%s: can't register ip6ip4\n", __func__);
1925                 goto xfrm_tunnel_failed;
1926         }
1927         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1928         if (err < 0) {
1929                 pr_info("%s: can't register ip4ip4\n", __func__);
1930                 goto xfrm_tunnel4_failed;
1931         }
1932 #if IS_ENABLED(CONFIG_MPLS)
1933         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1934         if (err < 0) {
1935                 pr_info("%s: can't register mplsip\n", __func__);
1936                 goto xfrm_tunnel_mpls_failed;
1937         }
1938 #endif
1939         err = rtnl_link_register(&sit_link_ops);
1940         if (err < 0)
1941                 goto rtnl_link_failed;
1942
1943 out:
1944         return err;
1945
1946 rtnl_link_failed:
1947 #if IS_ENABLED(CONFIG_MPLS)
1948         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1949 xfrm_tunnel_mpls_failed:
1950 #endif
1951         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1952 xfrm_tunnel4_failed:
1953         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1954 xfrm_tunnel_failed:
1955         unregister_pernet_device(&sit_net_ops);
1956         goto out;
1957 }
1958
1959 module_init(sit_init);
1960 module_exit(sit_cleanup);
1961 MODULE_LICENSE("GPL");
1962 MODULE_ALIAS_RTNL_LINK("sit");
1963 MODULE_ALIAS_NETDEV("sit0");