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, __be32 remote, __be32 local)
  95 {
  96         unsigned int h0 = HASH(remote);
  97         unsigned int h1 = HASH(local);
  98         struct ip_tunnel *t;
  99         struct sit_net *sitn = net_generic(net, sit_net_id);
 100
 101         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 102                 if (local == t->parms.iph.saddr &&
 103                     remote == t->parms.iph.daddr &&
 104                     (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
 105                     (t->dev->flags & IFF_UP))
 106                         return t;
 107         }
 108         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 109                 if (remote == t->parms.iph.daddr &&
 110                     (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
 111                     (t->dev->flags & IFF_UP))
 112                         return t;
 113         }
 114         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 115                 if (local == t->parms.iph.saddr &&
 116                     (!dev || !t->parms.link || dev->ifindex == t->parms.link) &&
 117                     (t->dev->flags & IFF_UP))
 118                         return t;
 119         }
 120         t = rcu_dereference(sitn->tunnels_wc[0]);
 121         if (t && (t->dev->flags & IFF_UP))
 122                 return t;
 123         return NULL;
 124 }
 125
 126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 127                 struct ip_tunnel_parm *parms)
 128 {
 129         __be32 remote = parms->iph.daddr;
 130         __be32 local = parms->iph.saddr;
 131         unsigned int h = 0;
 132         int prio = 0;
 133
 134         if (remote) {
 135                 prio |= 2;
 136                 h ^= HASH(remote);
 137         }
 138         if (local) {
 139                 prio |= 1;
 140                 h ^= HASH(local);
 141         }
 142         return &sitn->tunnels[prio][h];
 143 }
 144
 145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 146                 struct ip_tunnel *t)
 147 {
 148         return __ipip6_bucket(sitn, &t->parms);
 149 }
 150
 151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 152 {
 153         struct ip_tunnel __rcu **tp;
 154         struct ip_tunnel *iter;
 155
 156         for (tp = ipip6_bucket(sitn, t);
 157              (iter = rtnl_dereference(*tp)) != NULL;
 158              tp = &iter->next) {
 159                 if (t == iter) {
 160                         rcu_assign_pointer(*tp, t->next);
 161                         break;
 162                 }
 163         }
 164 }
 165
 166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 167 {
 168         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 169
 170         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 171         rcu_assign_pointer(*tp, t);
 172 }
 173
 174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 175 {
 176 #ifdef CONFIG_IPV6_SIT_6RD
 177         struct ip_tunnel *t = netdev_priv(dev);
 178
 179         if (dev == sitn->fb_tunnel_dev) {
 180                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 181                 t->ip6rd.relay_prefix = 0;
 182                 t->ip6rd.prefixlen = 16;
 183                 t->ip6rd.relay_prefixlen = 0;
 184         } else {
 185                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 186                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 187         }
 188 #endif
 189 }
 190
 191 static int ipip6_tunnel_create(struct net_device *dev)
 192 {
 193         struct ip_tunnel *t = netdev_priv(dev);
 194         struct net *net = dev_net(dev);
 195         struct sit_net *sitn = net_generic(net, sit_net_id);
 196         int err;
 197
 198         memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 199         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 200
 201         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 202                 dev->priv_flags |= IFF_ISATAP;
 203
 204         dev->rtnl_link_ops = &sit_link_ops;
 205
 206         err = register_netdevice(dev);
 207         if (err < 0)
 208                 goto out;
 209
 210         ipip6_tunnel_clone_6rd(dev, sitn);
 211
 212         ipip6_tunnel_link(sitn, t);
 213         return 0;
 214
 215 out:
 216         return err;
 217 }
 218
 219 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 220                 struct ip_tunnel_parm *parms, int create)
 221 {
 222         __be32 remote = parms->iph.daddr;
 223         __be32 local = parms->iph.saddr;
 224         struct ip_tunnel *t, *nt;
 225         struct ip_tunnel __rcu **tp;
 226         struct net_device *dev;
 227         char name[IFNAMSIZ];
 228         struct sit_net *sitn = net_generic(net, sit_net_id);
 229
 230         for (tp = __ipip6_bucket(sitn, parms);
 231             (t = rtnl_dereference(*tp)) != NULL;
 232              tp = &t->next) {
 233                 if (local == t->parms.iph.saddr &&
 234                     remote == t->parms.iph.daddr &&
 235                     parms->link == t->parms.link) {
 236                         if (create)
 237                                 return NULL;
 238                         else
 239                                 return t;
 240                 }
 241         }
 242         if (!create)
 243                 goto failed;
 244
 245         if (parms->name[0]) {
 246                 if (!dev_valid_name(parms->name))
 247                         goto failed;
 248                 strlcpy(name, parms->name, IFNAMSIZ);
 249         } else {
 250                 strcpy(name, "sit%d");
 251         }
 252         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 253                            ipip6_tunnel_setup);
 254         if (!dev)
 255                 return NULL;
 256
 257         dev_net_set(dev, net);
 258
 259         nt = netdev_priv(dev);
 260
 261         nt->parms = *parms;
 262         if (ipip6_tunnel_create(dev) < 0)
 263                 goto failed_free;
 264
 265         return nt;
 266
 267 failed_free:
 268         free_netdev(dev);
 269 failed:
 270         return NULL;
 271 }
 272
 273 #define for_each_prl_rcu(start)                 \
 274         for (prl = rcu_dereference(start);      \
 275              prl;                               \
 276              prl = rcu_dereference(prl->next))
 277
 278 static struct ip_tunnel_prl_entry *
 279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 280 {
 281         struct ip_tunnel_prl_entry *prl;
 282
 283         for_each_prl_rcu(t->prl)
 284                 if (prl->addr == addr)
 285                         break;
 286         return prl;
 287
 288 }
 289
 290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 291                                 struct ip_tunnel_prl __user *a)
 292 {
 293         struct ip_tunnel_prl kprl, *kp;
 294         struct ip_tunnel_prl_entry *prl;
 295         unsigned int cmax, c = 0, ca, len;
 296         int ret = 0;
 297
 298         if (copy_from_user(&kprl, a, sizeof(kprl)))
 299                 return -EFAULT;
 300         cmax = kprl.datalen / sizeof(kprl);
 301         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 302                 cmax = 1;
 303
 304         /* For simple GET or for root users,
 305          * we try harder to allocate.
 306          */
 307         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 308                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 309                 NULL;
 310
 311         ca = min(t->prl_count, cmax);
 312
 313         if (!kp) {
 314                 /* We don't try hard to allocate much memory for
 315                  * non-root users.
 316                  * For root users, retry allocating enough memory for
 317                  * the answer.
 318                  */
 319                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 320                 if (!kp) {
 321                         ret = -ENOMEM;
 322                         goto out;
 323                 }
 324         }
 325
 326         rcu_read_lock();
 327         for_each_prl_rcu(t->prl) {
 328                 if (c >= cmax)
 329                         break;
 330                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 331                         continue;
 332                 kp[c].addr = prl->addr;
 333                 kp[c].flags = prl->flags;
 334                 c++;
 335                 if (kprl.addr != htonl(INADDR_ANY))
 336                         break;
 337         }
 338
 339         rcu_read_unlock();
 340
 341         len = sizeof(*kp) * c;
 342         ret = 0;
 343         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 344                 ret = -EFAULT;
 345
 346         kfree(kp);
 347 out:
 348         return ret;
 349 }
 350
 351 static int
 352 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 353 {
 354         struct ip_tunnel_prl_entry *p;
 355         int err = 0;
 356
 357         if (a->addr == htonl(INADDR_ANY))
 358                 return -EINVAL;
 359
 360         ASSERT_RTNL();
 361
 362         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 363                 if (p->addr == a->addr) {
 364                         if (chg) {
 365                                 p->flags = a->flags;
 366                                 goto out;
 367                         }
 368                         err = -EEXIST;
 369                         goto out;
 370                 }
 371         }
 372
 373         if (chg) {
 374                 err = -ENXIO;
 375                 goto out;
 376         }
 377
 378         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 379         if (!p) {
 380                 err = -ENOBUFS;
 381                 goto out;
 382         }
 383
 384         p->next = t->prl;
 385         p->addr = a->addr;
 386         p->flags = a->flags;
 387         t->prl_count++;
 388         rcu_assign_pointer(t->prl, p);
 389 out:
 390         return err;
 391 }
 392
 393 static void prl_list_destroy_rcu(struct rcu_head *head)
 394 {
 395         struct ip_tunnel_prl_entry *p, *n;
 396
 397         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 398         do {
 399                 n = rcu_dereference_protected(p->next, 1);
 400                 kfree(p);
 401                 p = n;
 402         } while (p);
 403 }
 404
 405 static int
 406 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 407 {
 408         struct ip_tunnel_prl_entry *x;
 409         struct ip_tunnel_prl_entry __rcu **p;
 410         int err = 0;
 411
 412         ASSERT_RTNL();
 413
 414         if (a && a->addr != htonl(INADDR_ANY)) {
 415                 for (p = &t->prl;
 416                      (x = rtnl_dereference(*p)) != NULL;
 417                      p = &x->next) {
 418                         if (x->addr == a->addr) {
 419                                 *p = x->next;
 420                                 kfree_rcu(x, rcu_head);
 421                                 t->prl_count--;
 422                                 goto out;
 423                         }
 424                 }
 425                 err = -ENXIO;
 426         } else {
 427                 x = rtnl_dereference(t->prl);
 428                 if (x) {
 429                         t->prl_count = 0;
 430                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 431                         t->prl = NULL;
 432                 }
 433         }
 434 out:
 435         return err;
 436 }
 437
 438 static int
 439 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 440 {
 441         struct ip_tunnel_prl_entry *p;
 442         int ok = 1;
 443
 444         rcu_read_lock();
 445         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 446         if (p) {
 447                 if (p->flags & PRL_DEFAULT)
 448                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 449                 else
 450                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 451         } else {
 452                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 453
 454                 if (ipv6_addr_is_isatap(addr6) &&
 455                     (addr6->s6_addr32[3] == iph->saddr) &&
 456                     ipv6_chk_prefix(addr6, t->dev))
 457                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 458                 else
 459                         ok = 0;
 460         }
 461         rcu_read_unlock();
 462         return ok;
 463 }
 464
 465 static void ipip6_tunnel_uninit(struct net_device *dev)
 466 {
 467         struct ip_tunnel *tunnel = netdev_priv(dev);
 468         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 469
 470         if (dev == sitn->fb_tunnel_dev) {
 471                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 472         } else {
 473                 ipip6_tunnel_unlink(sitn, tunnel);
 474                 ipip6_tunnel_del_prl(tunnel, NULL);
 475         }
 476         dst_cache_reset(&tunnel->dst_cache);
 477         dev_put(dev);
 478 }
 479
 480 static int ipip6_err(struct sk_buff *skb, u32 info)
 481 {
 482         const struct iphdr *iph = (const struct iphdr *)skb->data;
 483         const int type = icmp_hdr(skb)->type;
 484         const int code = icmp_hdr(skb)->code;
 485         unsigned int data_len = 0;
 486         struct ip_tunnel *t;
 487         int err;
 488
 489         switch (type) {
 490         default:
 491         case ICMP_PARAMETERPROB:
 492                 return 0;
 493
 494         case ICMP_DEST_UNREACH:
 495                 switch (code) {
 496                 case ICMP_SR_FAILED:
 497                         /* Impossible event. */
 498                         return 0;
 499                 default:
 500                         /* All others are translated to HOST_UNREACH.
 501                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 502                            I believe they are just ether pollution. --ANK
 503                          */
 504                         break;
 505                 }
 506                 break;
 507         case ICMP_TIME_EXCEEDED:
 508                 if (code != ICMP_EXC_TTL)
 509                         return 0;
 510                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 511                 break;
 512         case ICMP_REDIRECT:
 513                 break;
 514         }
 515
 516         err = -ENOENT;
 517
 518         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 519                                 skb->dev,
 520                                 iph->daddr,
 521                                 iph->saddr);
 522         if (!t)
 523                 goto out;
 524
 525         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 526                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 527                                  t->parms.link, 0, iph->protocol, 0);
 528                 err = 0;
 529                 goto out;
 530         }
 531         if (type == ICMP_REDIRECT) {
 532                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
 533                               iph->protocol, 0);
 534                 err = 0;
 535                 goto out;
 536         }
 537
 538         err = 0;
 539         if (__in6_dev_get(skb->dev) &&
 540             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 541                 goto out;
 542
 543         if (t->parms.iph.daddr == 0)
 544                 goto out;
 545
 546         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 547                 goto out;
 548
 549         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 550                 t->err_count++;
 551         else
 552                 t->err_count = 1;
 553         t->err_time = jiffies;
 554 out:
 555         return err;
 556 }
 557
 558 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 559                                   const struct in6_addr *v6addr)
 560 {
 561         __be32 v4embed = 0;
 562         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 563                 return true;
 564         return false;
 565 }
 566
 567 /* Checks if an address matches an address on the tunnel interface.
 568  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 569  * Long story:
 570  * This function is called after we considered the packet as spoofed
 571  * in is_spoofed_6rd.
 572  * We may have a router that is doing NAT for proto 41 packets
 573  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 574  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 575  * function will return true, dropping the packet.
 576  * But, we can still check if is spoofed against the IP
 577  * addresses associated with the interface.
 578  */
 579 static bool only_dnatted(const struct ip_tunnel *tunnel,
 580         const struct in6_addr *v6dst)
 581 {
 582         int prefix_len;
 583
 584 #ifdef CONFIG_IPV6_SIT_6RD
 585         prefix_len = tunnel->ip6rd.prefixlen + 32
 586                 - tunnel->ip6rd.relay_prefixlen;
 587 #else
 588         prefix_len = 48;
 589 #endif
 590         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 591 }
 592
 593 /* Returns true if a packet is spoofed */
 594 static bool packet_is_spoofed(struct sk_buff *skb,
 595                               const struct iphdr *iph,
 596                               struct ip_tunnel *tunnel)
 597 {
 598         const struct ipv6hdr *ipv6h;
 599
 600         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 601                 if (!isatap_chksrc(skb, iph, tunnel))
 602                         return true;
 603
 604                 return false;
 605         }
 606
 607         if (tunnel->dev->flags & IFF_POINTOPOINT)
 608                 return false;
 609
 610         ipv6h = ipv6_hdr(skb);
 611
 612         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 613                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 614                                      &iph->saddr, &ipv6h->saddr,
 615                                      &iph->daddr, &ipv6h->daddr);
 616                 return true;
 617         }
 618
 619         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 620                 return false;
 621
 622         if (only_dnatted(tunnel, &ipv6h->daddr))
 623                 return false;
 624
 625         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 626                              &iph->saddr, &ipv6h->saddr,
 627                              &iph->daddr, &ipv6h->daddr);
 628         return true;
 629 }
 630
 631 static int ipip6_rcv(struct sk_buff *skb)
 632 {
 633         const struct iphdr *iph = ip_hdr(skb);
 634         struct ip_tunnel *tunnel;
 635         int err;
 636
 637         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 638                                      iph->saddr, iph->daddr);
 639         if (tunnel) {
 640                 struct pcpu_sw_netstats *tstats;
 641
 642                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 643                     tunnel->parms.iph.protocol != 0)
 644                         goto out;
 645
 646                 skb->mac_header = skb->network_header;
 647                 skb_reset_network_header(skb);
 648                 IPCB(skb)->flags = 0;
 649                 skb->dev = tunnel->dev;
 650
 651                 if (packet_is_spoofed(skb, iph, tunnel)) {
 652                         tunnel->dev->stats.rx_errors++;
 653                         goto out;
 654                 }
 655
 656                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 657                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 658                         goto out;
 659
 660                 /* skb can be uncloned in iptunnel_pull_header, so
 661                  * old iph is no longer valid
 662                  */
 663                 iph = (const struct iphdr *)skb_mac_header(skb);
 664                 err = IP_ECN_decapsulate(iph, skb);
 665                 if (unlikely(err)) {
 666                         if (log_ecn_error)
 667                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 668                                                      &iph->saddr, iph->tos);
 669                         if (err > 1) {
 670                                 ++tunnel->dev->stats.rx_frame_errors;
 671                                 ++tunnel->dev->stats.rx_errors;
 672                                 goto out;
 673                         }
 674                 }
 675
 676                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 677                 u64_stats_update_begin(&tstats->syncp);
 678                 tstats->rx_packets++;
 679                 tstats->rx_bytes += skb->len;
 680                 u64_stats_update_end(&tstats->syncp);
 681
 682                 netif_rx(skb);
 683
 684                 return 0;
 685         }
 686
 687         /* no tunnel matched,  let upstream know, ipsec may handle it */
 688         return 1;
 689 out:
 690         kfree_skb(skb);
 691         return 0;
 692 }
 693
 694 static const struct tnl_ptk_info ipip_tpi = {
 695         /* no tunnel info required for ipip. */
 696         .proto = htons(ETH_P_IP),
 697 };
 698
 699 #if IS_ENABLED(CONFIG_MPLS)
 700 static const struct tnl_ptk_info mplsip_tpi = {
 701         /* no tunnel info required for mplsip. */
 702         .proto = htons(ETH_P_MPLS_UC),
 703 };
 704 #endif
 705
 706 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 707 {
 708         const struct iphdr *iph;
 709         struct ip_tunnel *tunnel;
 710
 711         iph = ip_hdr(skb);
 712         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 713                                      iph->saddr, iph->daddr);
 714         if (tunnel) {
 715                 const struct tnl_ptk_info *tpi;
 716
 717                 if (tunnel->parms.iph.protocol != ipproto &&
 718                     tunnel->parms.iph.protocol != 0)
 719                         goto drop;
 720
 721                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 722                         goto drop;
 723 #if IS_ENABLED(CONFIG_MPLS)
 724                 if (ipproto == IPPROTO_MPLS)
 725                         tpi = &mplsip_tpi;
 726                 else
 727 #endif
 728                         tpi = &ipip_tpi;
 729                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 730                         goto drop;
 731                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 732         }
 733
 734         return 1;
 735
 736 drop:
 737         kfree_skb(skb);
 738         return 0;
 739 }
 740
 741 static int ipip_rcv(struct sk_buff *skb)
 742 {
 743         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 744 }
 745
 746 #if IS_ENABLED(CONFIG_MPLS)
 747 static int mplsip_rcv(struct sk_buff *skb)
 748 {
 749         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 750 }
 751 #endif
 752
 753 /*
 754  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 755  * stores the embedded IPv4 address in v4dst and returns true.
 756  */
 757 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 758                       __be32 *v4dst)
 759 {
 760 #ifdef CONFIG_IPV6_SIT_6RD
 761         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 762                               tunnel->ip6rd.prefixlen)) {
 763                 unsigned int pbw0, pbi0;
 764                 int pbi1;
 765                 u32 d;
 766
 767                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 768                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 769
 770                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 771                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 772                     tunnel->ip6rd.relay_prefixlen : 0;
 773
 774                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 775                 if (pbi1 > 0)
 776                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 777                              (32 - pbi1);
 778
 779                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 780                 return true;
 781         }
 782 #else
 783         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 784                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 785                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 786                 return true;
 787         }
 788 #endif
 789         return false;
 790 }
 791
 792 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 793                              const struct in6_addr *v6dst)
 794 {
 795         __be32 dst = 0;
 796         check_6rd(tunnel, v6dst, &dst);
 797         return dst;
 798 }
 799
 800 /*
 801  *      This function assumes it is being called from dev_queue_xmit()
 802  *      and that skb is filled properly by that function.
 803  */
 804
 805 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 806                                      struct net_device *dev)
 807 {
 808         struct ip_tunnel *tunnel = netdev_priv(dev);
 809         const struct iphdr  *tiph = &tunnel->parms.iph;
 810         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 811         u8     tos = tunnel->parms.iph.tos;
 812         __be16 df = tiph->frag_off;
 813         struct rtable *rt;              /* Route to the other host */
 814         struct net_device *tdev;        /* Device to other host */
 815         unsigned int max_headroom;      /* The extra header space needed */
 816         __be32 dst = tiph->daddr;
 817         struct flowi4 fl4;
 818         int    mtu;
 819         const struct in6_addr *addr6;
 820         int addr_type;
 821         u8 ttl;
 822         u8 protocol = IPPROTO_IPV6;
 823         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 824
 825         if (tos == 1)
 826                 tos = ipv6_get_dsfield(iph6);
 827
 828         /* ISATAP (RFC4214) - must come before 6to4 */
 829         if (dev->priv_flags & IFF_ISATAP) {
 830                 struct neighbour *neigh = NULL;
 831                 bool do_tx_error = false;
 832
 833                 if (skb_dst(skb))
 834                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 835
 836                 if (!neigh) {
 837                         net_dbg_ratelimited("nexthop == NULL\n");
 838                         goto tx_error;
 839                 }
 840
 841                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 842                 addr_type = ipv6_addr_type(addr6);
 843
 844                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 845                      ipv6_addr_is_isatap(addr6))
 846                         dst = addr6->s6_addr32[3];
 847                 else
 848                         do_tx_error = true;
 849
 850                 neigh_release(neigh);
 851                 if (do_tx_error)
 852                         goto tx_error;
 853         }
 854
 855         if (!dst)
 856                 dst = try_6rd(tunnel, &iph6->daddr);
 857
 858         if (!dst) {
 859                 struct neighbour *neigh = NULL;
 860                 bool do_tx_error = false;
 861
 862                 if (skb_dst(skb))
 863                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 864
 865                 if (!neigh) {
 866                         net_dbg_ratelimited("nexthop == NULL\n");
 867                         goto tx_error;
 868                 }
 869
 870                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 871                 addr_type = ipv6_addr_type(addr6);
 872
 873                 if (addr_type == IPV6_ADDR_ANY) {
 874                         addr6 = &ipv6_hdr(skb)->daddr;
 875                         addr_type = ipv6_addr_type(addr6);
 876                 }
 877
 878                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 879                         dst = addr6->s6_addr32[3];
 880                 else
 881                         do_tx_error = true;
 882
 883                 neigh_release(neigh);
 884                 if (do_tx_error)
 885                         goto tx_error;
 886         }
 887
 888         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 889                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 890                            0, dst, tiph->saddr, 0, 0,
 891                            sock_net_uid(tunnel->net, NULL));
 892         rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 893
 894         if (IS_ERR(rt)) {
 895                 dev->stats.tx_carrier_errors++;
 896                 goto tx_error_icmp;
 897         }
 898         if (rt->rt_type != RTN_UNICAST) {
 899                 ip_rt_put(rt);
 900                 dev->stats.tx_carrier_errors++;
 901                 goto tx_error_icmp;
 902         }
 903         tdev = rt->dst.dev;
 904
 905         if (tdev == dev) {
 906                 ip_rt_put(rt);
 907                 dev->stats.collisions++;
 908                 goto tx_error;
 909         }
 910
 911         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 912                 ip_rt_put(rt);
 913                 goto tx_error;
 914         }
 915
 916         if (df) {
 917                 mtu = dst_mtu(&rt->dst) - t_hlen;
 918
 919                 if (mtu < 68) {
 920                         dev->stats.collisions++;
 921                         ip_rt_put(rt);
 922                         goto tx_error;
 923                 }
 924
 925                 if (mtu < IPV6_MIN_MTU) {
 926                         mtu = IPV6_MIN_MTU;
 927                         df = 0;
 928                 }
 929
 930                 if (tunnel->parms.iph.daddr)
 931                         skb_dst_update_pmtu_no_confirm(skb, mtu);
 932
 933                 if (skb->len > mtu && !skb_is_gso(skb)) {
 934                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 935                         ip_rt_put(rt);
 936                         goto tx_error;
 937                 }
 938         }
 939
 940         if (tunnel->err_count > 0) {
 941                 if (time_before(jiffies,
 942                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 943                         tunnel->err_count--;
 944                         dst_link_failure(skb);
 945                 } else
 946                         tunnel->err_count = 0;
 947         }
 948
 949         /*
 950          * Okay, now see if we can stuff it in the buffer as-is.
 951          */
 952         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 953
 954         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 955             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 956                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 957                 if (!new_skb) {
 958                         ip_rt_put(rt);
 959                         dev->stats.tx_dropped++;
 960                         kfree_skb(skb);
 961                         return NETDEV_TX_OK;
 962                 }
 963                 if (skb->sk)
 964                         skb_set_owner_w(new_skb, skb->sk);
 965                 dev_kfree_skb(skb);
 966                 skb = new_skb;
 967                 iph6 = ipv6_hdr(skb);
 968         }
 969         ttl = tiph->ttl;
 970         if (ttl == 0)
 971                 ttl = iph6->hop_limit;
 972         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 973
 974         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 975                 ip_rt_put(rt);
 976                 goto tx_error;
 977         }
 978
 979         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 980
 981         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 982                       df, !net_eq(tunnel->net, dev_net(dev)));
 983         return NETDEV_TX_OK;
 984
 985 tx_error_icmp:
 986         dst_link_failure(skb);
 987 tx_error:
 988         kfree_skb(skb);
 989         dev->stats.tx_errors++;
 990         return NETDEV_TX_OK;
 991 }
 992
 993 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
 994                                      struct net_device *dev, u8 ipproto)
 995 {
 996         struct ip_tunnel *tunnel = netdev_priv(dev);
 997         const struct iphdr  *tiph = &tunnel->parms.iph;
 998
 999         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1000                 goto tx_error;
1001
1002         skb_set_inner_ipproto(skb, ipproto);
1003
1004         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1005         return NETDEV_TX_OK;
1006 tx_error:
1007         kfree_skb(skb);
1008         dev->stats.tx_errors++;
1009         return NETDEV_TX_OK;
1010 }
1011
1012 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1013                                    struct net_device *dev)
1014 {
1015         switch (skb->protocol) {
1016         case htons(ETH_P_IP):
1017                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1018                 break;
1019         case htons(ETH_P_IPV6):
1020                 ipip6_tunnel_xmit(skb, dev);
1021                 break;
1022 #if IS_ENABLED(CONFIG_MPLS)
1023         case htons(ETH_P_MPLS_UC):
1024                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1025                 break;
1026 #endif
1027         default:
1028                 goto tx_err;
1029         }
1030
1031         return NETDEV_TX_OK;
1032
1033 tx_err:
1034         dev->stats.tx_errors++;
1035         kfree_skb(skb);
1036         return NETDEV_TX_OK;
1037
1038 }
1039
1040 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1041 {
1042         struct ip_tunnel *tunnel = netdev_priv(dev);
1043         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1044         struct net_device *tdev = NULL;
1045         int hlen = LL_MAX_HEADER;
1046         const struct iphdr *iph;
1047         struct flowi4 fl4;
1048
1049         iph = &tunnel->parms.iph;
1050
1051         if (iph->daddr) {
1052                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1053                                                           NULL,
1054                                                           iph->daddr, iph->saddr,
1055                                                           0, 0,
1056                                                           IPPROTO_IPV6,
1057                                                           RT_TOS(iph->tos),
1058                                                           tunnel->parms.link);
1059
1060                 if (!IS_ERR(rt)) {
1061                         tdev = rt->dst.dev;
1062                         ip_rt_put(rt);
1063                 }
1064                 dev->flags |= IFF_POINTOPOINT;
1065         }
1066
1067         if (!tdev && tunnel->parms.link)
1068                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1069
1070         if (tdev && !netif_is_l3_master(tdev)) {
1071                 int mtu;
1072
1073                 mtu = tdev->mtu - t_hlen;
1074                 if (mtu < IPV6_MIN_MTU)
1075                         mtu = IPV6_MIN_MTU;
1076                 WRITE_ONCE(dev->mtu, mtu);
1077                 hlen = tdev->hard_header_len + tdev->needed_headroom;
1078         }
1079         dev->needed_headroom = t_hlen + hlen;
1080 }
1081
1082 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1083                                 __u32 fwmark)
1084 {
1085         struct net *net = t->net;
1086         struct sit_net *sitn = net_generic(net, sit_net_id);
1087
1088         ipip6_tunnel_unlink(sitn, t);
1089         synchronize_net();
1090         t->parms.iph.saddr = p->iph.saddr;
1091         t->parms.iph.daddr = p->iph.daddr;
1092         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1093         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1094         ipip6_tunnel_link(sitn, t);
1095         t->parms.iph.ttl = p->iph.ttl;
1096         t->parms.iph.tos = p->iph.tos;
1097         t->parms.iph.frag_off = p->iph.frag_off;
1098         if (t->parms.link != p->link || t->fwmark != fwmark) {
1099                 t->parms.link = p->link;
1100                 t->fwmark = fwmark;
1101                 ipip6_tunnel_bind_dev(t->dev);
1102         }
1103         dst_cache_reset(&t->dst_cache);
1104         netdev_state_change(t->dev);
1105 }
1106
1107 #ifdef CONFIG_IPV6_SIT_6RD
1108 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1109                                    struct ip_tunnel_6rd *ip6rd)
1110 {
1111         struct in6_addr prefix;
1112         __be32 relay_prefix;
1113
1114         if (ip6rd->relay_prefixlen > 32 ||
1115             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1116                 return -EINVAL;
1117
1118         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1119         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1120                 return -EINVAL;
1121         if (ip6rd->relay_prefixlen)
1122                 relay_prefix = ip6rd->relay_prefix &
1123                                htonl(0xffffffffUL <<
1124                                      (32 - ip6rd->relay_prefixlen));
1125         else
1126                 relay_prefix = 0;
1127         if (relay_prefix != ip6rd->relay_prefix)
1128                 return -EINVAL;
1129
1130         t->ip6rd.prefix = prefix;
1131         t->ip6rd.relay_prefix = relay_prefix;
1132         t->ip6rd.prefixlen = ip6rd->prefixlen;
1133         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1134         dst_cache_reset(&t->dst_cache);
1135         netdev_state_change(t->dev);
1136         return 0;
1137 }
1138 #endif
1139
1140 static bool ipip6_valid_ip_proto(u8 ipproto)
1141 {
1142         return ipproto == IPPROTO_IPV6 ||
1143                 ipproto == IPPROTO_IPIP ||
1144 #if IS_ENABLED(CONFIG_MPLS)
1145                 ipproto == IPPROTO_MPLS ||
1146 #endif
1147                 ipproto == 0;
1148 }
1149
1150 static int
1151 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1152 {
1153         int err = 0;
1154         struct ip_tunnel_parm p;
1155         struct ip_tunnel_prl prl;
1156         struct ip_tunnel *t = netdev_priv(dev);
1157         struct net *net = t->net;
1158         struct sit_net *sitn = net_generic(net, sit_net_id);
1159 #ifdef CONFIG_IPV6_SIT_6RD
1160         struct ip_tunnel_6rd ip6rd;
1161 #endif
1162
1163         switch (cmd) {
1164         case SIOCGETTUNNEL:
1165 #ifdef CONFIG_IPV6_SIT_6RD
1166         case SIOCGET6RD:
1167 #endif
1168                 if (dev == sitn->fb_tunnel_dev) {
1169                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1170                                 err = -EFAULT;
1171                                 break;
1172                         }
1173                         t = ipip6_tunnel_locate(net, &p, 0);
1174                         if (!t)
1175                                 t = netdev_priv(dev);
1176                 }
1177
1178                 err = -EFAULT;
1179                 if (cmd == SIOCGETTUNNEL) {
1180                         memcpy(&p, &t->parms, sizeof(p));
1181                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1182                                          sizeof(p)))
1183                                 goto done;
1184 #ifdef CONFIG_IPV6_SIT_6RD
1185                 } else {
1186                         ip6rd.prefix = t->ip6rd.prefix;
1187                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1188                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1189                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1190                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1191                                          sizeof(ip6rd)))
1192                                 goto done;
1193 #endif
1194                 }
1195                 err = 0;
1196                 break;
1197
1198         case SIOCADDTUNNEL:
1199         case SIOCCHGTUNNEL:
1200                 err = -EPERM;
1201                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1202                         goto done;
1203
1204                 err = -EFAULT;
1205                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1206                         goto done;
1207
1208                 err = -EINVAL;
1209                 if (!ipip6_valid_ip_proto(p.iph.protocol))
1210                         goto done;
1211                 if (p.iph.version != 4 ||
1212                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1213                         goto done;
1214                 if (p.iph.ttl)
1215                         p.iph.frag_off |= htons(IP_DF);
1216
1217                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1218
1219                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1220                         if (t) {
1221                                 if (t->dev != dev) {
1222                                         err = -EEXIST;
1223                                         break;
1224                                 }
1225                         } else {
1226                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1227                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1228                                         err = -EINVAL;
1229                                         break;
1230                                 }
1231                                 t = netdev_priv(dev);
1232                         }
1233
1234                         ipip6_tunnel_update(t, &p, t->fwmark);
1235                 }
1236
1237                 if (t) {
1238                         err = 0;
1239                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1240                                 err = -EFAULT;
1241                 } else
1242                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1243                 break;
1244
1245         case SIOCDELTUNNEL:
1246                 err = -EPERM;
1247                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1248                         goto done;
1249
1250                 if (dev == sitn->fb_tunnel_dev) {
1251                         err = -EFAULT;
1252                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1253                                 goto done;
1254                         err = -ENOENT;
1255                         t = ipip6_tunnel_locate(net, &p, 0);
1256                         if (!t)
1257                                 goto done;
1258                         err = -EPERM;
1259                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1260                                 goto done;
1261                         dev = t->dev;
1262                 }
1263                 unregister_netdevice(dev);
1264                 err = 0;
1265                 break;
1266
1267         case SIOCGETPRL:
1268                 err = -EINVAL;
1269                 if (dev == sitn->fb_tunnel_dev)
1270                         goto done;
1271                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1272                 break;
1273
1274         case SIOCADDPRL:
1275         case SIOCDELPRL:
1276         case SIOCCHGPRL:
1277                 err = -EPERM;
1278                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1279                         goto done;
1280                 err = -EINVAL;
1281                 if (dev == sitn->fb_tunnel_dev)
1282                         goto done;
1283                 err = -EFAULT;
1284                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1285                         goto done;
1286
1287                 switch (cmd) {
1288                 case SIOCDELPRL:
1289                         err = ipip6_tunnel_del_prl(t, &prl);
1290                         break;
1291                 case SIOCADDPRL:
1292                 case SIOCCHGPRL:
1293                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1294                         break;
1295                 }
1296                 dst_cache_reset(&t->dst_cache);
1297                 netdev_state_change(dev);
1298                 break;
1299
1300 #ifdef CONFIG_IPV6_SIT_6RD
1301         case SIOCADD6RD:
1302         case SIOCCHG6RD:
1303         case SIOCDEL6RD:
1304                 err = -EPERM;
1305                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1306                         goto done;
1307
1308                 err = -EFAULT;
1309                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1310                                    sizeof(ip6rd)))
1311                         goto done;
1312
1313                 if (cmd != SIOCDEL6RD) {
1314                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1315                         if (err < 0)
1316                                 goto done;
1317                 } else
1318                         ipip6_tunnel_clone_6rd(dev, sitn);
1319
1320                 err = 0;
1321                 break;
1322 #endif
1323
1324         default:
1325                 err = -EINVAL;
1326         }
1327
1328 done:
1329         return err;
1330 }
1331
1332 static const struct net_device_ops ipip6_netdev_ops = {
1333         .ndo_init       = ipip6_tunnel_init,
1334         .ndo_uninit     = ipip6_tunnel_uninit,
1335         .ndo_start_xmit = sit_tunnel_xmit,
1336         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1337         .ndo_get_stats64 = ip_tunnel_get_stats64,
1338         .ndo_get_iflink = ip_tunnel_get_iflink,
1339 };
1340
1341 static void ipip6_dev_free(struct net_device *dev)
1342 {
1343         struct ip_tunnel *tunnel = netdev_priv(dev);
1344
1345         dst_cache_destroy(&tunnel->dst_cache);
1346         free_percpu(dev->tstats);
1347 }
1348
1349 #define SIT_FEATURES (NETIF_F_SG           | \
1350                       NETIF_F_FRAGLIST     | \
1351                       NETIF_F_HIGHDMA      | \
1352                       NETIF_F_GSO_SOFTWARE | \
1353                       NETIF_F_HW_CSUM)
1354
1355 static void ipip6_tunnel_setup(struct net_device *dev)
1356 {
1357         struct ip_tunnel *tunnel = netdev_priv(dev);
1358         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1359
1360         dev->netdev_ops         = &ipip6_netdev_ops;
1361         dev->needs_free_netdev  = true;
1362         dev->priv_destructor    = ipip6_dev_free;
1363
1364         dev->type               = ARPHRD_SIT;
1365         dev->mtu                = ETH_DATA_LEN - t_hlen;
1366         dev->min_mtu            = IPV6_MIN_MTU;
1367         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1368         dev->flags              = IFF_NOARP;
1369         netif_keep_dst(dev);
1370         dev->addr_len           = 4;
1371         dev->features           |= NETIF_F_LLTX;
1372         dev->features           |= SIT_FEATURES;
1373         dev->hw_features        |= SIT_FEATURES;
1374 }
1375
1376 static int ipip6_tunnel_init(struct net_device *dev)
1377 {
1378         struct ip_tunnel *tunnel = netdev_priv(dev);
1379         int err;
1380
1381         tunnel->dev = dev;
1382         tunnel->net = dev_net(dev);
1383         strcpy(tunnel->parms.name, dev->name);
1384
1385         ipip6_tunnel_bind_dev(dev);
1386         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1387         if (!dev->tstats)
1388                 return -ENOMEM;
1389
1390         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1391         if (err) {
1392                 free_percpu(dev->tstats);
1393                 dev->tstats = NULL;
1394                 return err;
1395         }
1396         dev_hold(dev);
1397         return 0;
1398 }
1399
1400 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1401 {
1402         struct ip_tunnel *tunnel = netdev_priv(dev);
1403         struct iphdr *iph = &tunnel->parms.iph;
1404         struct net *net = dev_net(dev);
1405         struct sit_net *sitn = net_generic(net, sit_net_id);
1406
1407         iph->version            = 4;
1408         iph->protocol           = IPPROTO_IPV6;
1409         iph->ihl                = 5;
1410         iph->ttl                = 64;
1411
1412         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1413 }
1414
1415 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1416                           struct netlink_ext_ack *extack)
1417 {
1418         u8 proto;
1419
1420         if (!data || !data[IFLA_IPTUN_PROTO])
1421                 return 0;
1422
1423         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1424         if (!ipip6_valid_ip_proto(proto))
1425                 return -EINVAL;
1426
1427         return 0;
1428 }
1429
1430 static void ipip6_netlink_parms(struct nlattr *data[],
1431                                 struct ip_tunnel_parm *parms,
1432                                 __u32 *fwmark)
1433 {
1434         memset(parms, 0, sizeof(*parms));
1435
1436         parms->iph.version = 4;
1437         parms->iph.protocol = IPPROTO_IPV6;
1438         parms->iph.ihl = 5;
1439         parms->iph.ttl = 64;
1440
1441         if (!data)
1442                 return;
1443
1444         if (data[IFLA_IPTUN_LINK])
1445                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1446
1447         if (data[IFLA_IPTUN_LOCAL])
1448                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1449
1450         if (data[IFLA_IPTUN_REMOTE])
1451                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1452
1453         if (data[IFLA_IPTUN_TTL]) {
1454                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1455                 if (parms->iph.ttl)
1456                         parms->iph.frag_off = htons(IP_DF);
1457         }
1458
1459         if (data[IFLA_IPTUN_TOS])
1460                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1461
1462         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1463                 parms->iph.frag_off = htons(IP_DF);
1464
1465         if (data[IFLA_IPTUN_FLAGS])
1466                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1467
1468         if (data[IFLA_IPTUN_PROTO])
1469                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1470
1471         if (data[IFLA_IPTUN_FWMARK])
1472                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1473 }
1474
1475 /* This function returns true when ENCAP attributes are present in the nl msg */
1476 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1477                                       struct ip_tunnel_encap *ipencap)
1478 {
1479         bool ret = false;
1480
1481         memset(ipencap, 0, sizeof(*ipencap));
1482
1483         if (!data)
1484                 return ret;
1485
1486         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1487                 ret = true;
1488                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1489         }
1490
1491         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1492                 ret = true;
1493                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1494         }
1495
1496         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1497                 ret = true;
1498                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1499         }
1500
1501         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1502                 ret = true;
1503                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1504         }
1505
1506         return ret;
1507 }
1508
1509 #ifdef CONFIG_IPV6_SIT_6RD
1510 /* This function returns true when 6RD attributes are present in the nl msg */
1511 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1512                                     struct ip_tunnel_6rd *ip6rd)
1513 {
1514         bool ret = false;
1515         memset(ip6rd, 0, sizeof(*ip6rd));
1516
1517         if (!data)
1518                 return ret;
1519
1520         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1521                 ret = true;
1522                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1523         }
1524
1525         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1526                 ret = true;
1527                 ip6rd->relay_prefix =
1528                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1529         }
1530
1531         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1532                 ret = true;
1533                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1534         }
1535
1536         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1537                 ret = true;
1538                 ip6rd->relay_prefixlen =
1539                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1540         }
1541
1542         return ret;
1543 }
1544 #endif
1545
1546 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1547                          struct nlattr *tb[], struct nlattr *data[],
1548                          struct netlink_ext_ack *extack)
1549 {
1550         struct net *net = dev_net(dev);
1551         struct ip_tunnel *nt;
1552         struct ip_tunnel_encap ipencap;
1553 #ifdef CONFIG_IPV6_SIT_6RD
1554         struct ip_tunnel_6rd ip6rd;
1555 #endif
1556         int err;
1557
1558         nt = netdev_priv(dev);
1559
1560         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1561                 err = ip_tunnel_encap_setup(nt, &ipencap);
1562                 if (err < 0)
1563                         return err;
1564         }
1565
1566         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1567
1568         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1569                 return -EEXIST;
1570
1571         err = ipip6_tunnel_create(dev);
1572         if (err < 0)
1573                 return err;
1574
1575         if (tb[IFLA_MTU]) {
1576                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1577
1578                 if (mtu >= IPV6_MIN_MTU &&
1579                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1580                         dev->mtu = mtu;
1581         }
1582
1583 #ifdef CONFIG_IPV6_SIT_6RD
1584         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1585                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1586                 if (err < 0)
1587                         unregister_netdevice_queue(dev, NULL);
1588         }
1589 #endif
1590
1591         return err;
1592 }
1593
1594 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1595                             struct nlattr *data[],
1596                             struct netlink_ext_ack *extack)
1597 {
1598         struct ip_tunnel *t = netdev_priv(dev);
1599         struct ip_tunnel_parm p;
1600         struct ip_tunnel_encap ipencap;
1601         struct net *net = t->net;
1602         struct sit_net *sitn = net_generic(net, sit_net_id);
1603 #ifdef CONFIG_IPV6_SIT_6RD
1604         struct ip_tunnel_6rd ip6rd;
1605 #endif
1606         __u32 fwmark = t->fwmark;
1607         int err;
1608
1609         if (dev == sitn->fb_tunnel_dev)
1610                 return -EINVAL;
1611
1612         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1613                 err = ip_tunnel_encap_setup(t, &ipencap);
1614                 if (err < 0)
1615                         return err;
1616         }
1617
1618         ipip6_netlink_parms(data, &p, &fwmark);
1619
1620         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1621             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1622                 return -EINVAL;
1623
1624         t = ipip6_tunnel_locate(net, &p, 0);
1625
1626         if (t) {
1627                 if (t->dev != dev)
1628                         return -EEXIST;
1629         } else
1630                 t = netdev_priv(dev);
1631
1632         ipip6_tunnel_update(t, &p, fwmark);
1633
1634 #ifdef CONFIG_IPV6_SIT_6RD
1635         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1636                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1637 #endif
1638
1639         return 0;
1640 }
1641
1642 static size_t ipip6_get_size(const struct net_device *dev)
1643 {
1644         return
1645                 /* IFLA_IPTUN_LINK */
1646                 nla_total_size(4) +
1647                 /* IFLA_IPTUN_LOCAL */
1648                 nla_total_size(4) +
1649                 /* IFLA_IPTUN_REMOTE */
1650                 nla_total_size(4) +
1651                 /* IFLA_IPTUN_TTL */
1652                 nla_total_size(1) +
1653                 /* IFLA_IPTUN_TOS */
1654                 nla_total_size(1) +
1655                 /* IFLA_IPTUN_PMTUDISC */
1656                 nla_total_size(1) +
1657                 /* IFLA_IPTUN_FLAGS */
1658                 nla_total_size(2) +
1659                 /* IFLA_IPTUN_PROTO */
1660                 nla_total_size(1) +
1661 #ifdef CONFIG_IPV6_SIT_6RD
1662                 /* IFLA_IPTUN_6RD_PREFIX */
1663                 nla_total_size(sizeof(struct in6_addr)) +
1664                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1665                 nla_total_size(4) +
1666                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1667                 nla_total_size(2) +
1668                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1669                 nla_total_size(2) +
1670 #endif
1671                 /* IFLA_IPTUN_ENCAP_TYPE */
1672                 nla_total_size(2) +
1673                 /* IFLA_IPTUN_ENCAP_FLAGS */
1674                 nla_total_size(2) +
1675                 /* IFLA_IPTUN_ENCAP_SPORT */
1676                 nla_total_size(2) +
1677                 /* IFLA_IPTUN_ENCAP_DPORT */
1678                 nla_total_size(2) +
1679                 /* IFLA_IPTUN_FWMARK */
1680                 nla_total_size(4) +
1681                 0;
1682 }
1683
1684 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1685 {
1686         struct ip_tunnel *tunnel = netdev_priv(dev);
1687         struct ip_tunnel_parm *parm = &tunnel->parms;
1688
1689         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1690             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1691             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1692             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1693             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1694             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1695                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1696             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1697             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1698             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1699                 goto nla_put_failure;
1700
1701 #ifdef CONFIG_IPV6_SIT_6RD
1702         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1703                              &tunnel->ip6rd.prefix) ||
1704             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1705                             tunnel->ip6rd.relay_prefix) ||
1706             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1707                         tunnel->ip6rd.prefixlen) ||
1708             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1709                         tunnel->ip6rd.relay_prefixlen))
1710                 goto nla_put_failure;
1711 #endif
1712
1713         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1714                         tunnel->encap.type) ||
1715             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1716                         tunnel->encap.sport) ||
1717             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1718                         tunnel->encap.dport) ||
1719             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1720                         tunnel->encap.flags))
1721                 goto nla_put_failure;
1722
1723         return 0;
1724
1725 nla_put_failure:
1726         return -EMSGSIZE;
1727 }
1728
1729 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1730         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1731         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1732         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1733         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1734         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1735         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1736         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1737         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1738 #ifdef CONFIG_IPV6_SIT_6RD
1739         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1740         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1741         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1742         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1743 #endif
1744         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1745         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1746         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1747         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1748         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1749 };
1750
1751 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1752 {
1753         struct net *net = dev_net(dev);
1754         struct sit_net *sitn = net_generic(net, sit_net_id);
1755
1756         if (dev != sitn->fb_tunnel_dev)
1757                 unregister_netdevice_queue(dev, head);
1758 }
1759
1760 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1761         .kind           = "sit",
1762         .maxtype        = IFLA_IPTUN_MAX,
1763         .policy         = ipip6_policy,
1764         .priv_size      = sizeof(struct ip_tunnel),
1765         .setup          = ipip6_tunnel_setup,
1766         .validate       = ipip6_validate,
1767         .newlink        = ipip6_newlink,
1768         .changelink     = ipip6_changelink,
1769         .get_size       = ipip6_get_size,
1770         .fill_info      = ipip6_fill_info,
1771         .dellink        = ipip6_dellink,
1772         .get_link_net   = ip_tunnel_get_link_net,
1773 };
1774
1775 static struct xfrm_tunnel sit_handler __read_mostly = {
1776         .handler        =       ipip6_rcv,
1777         .err_handler    =       ipip6_err,
1778         .priority       =       1,
1779 };
1780
1781 static struct xfrm_tunnel ipip_handler __read_mostly = {
1782         .handler        =       ipip_rcv,
1783         .err_handler    =       ipip6_err,
1784         .priority       =       2,
1785 };
1786
1787 #if IS_ENABLED(CONFIG_MPLS)
1788 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1789         .handler        =       mplsip_rcv,
1790         .err_handler    =       ipip6_err,
1791         .priority       =       2,
1792 };
1793 #endif
1794
1795 static void __net_exit sit_destroy_tunnels(struct net *net,
1796                                            struct list_head *head)
1797 {
1798         struct sit_net *sitn = net_generic(net, sit_net_id);
1799         struct net_device *dev, *aux;
1800         int prio;
1801
1802         for_each_netdev_safe(net, dev, aux)
1803                 if (dev->rtnl_link_ops == &sit_link_ops)
1804                         unregister_netdevice_queue(dev, head);
1805
1806         for (prio = 0; prio < 4; prio++) {
1807                 int h;
1808                 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1809                         struct ip_tunnel *t;
1810
1811                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1812                         while (t) {
1813                                 /* If dev is in the same netns, it has already
1814                                  * been added to the list by the previous loop.
1815                                  */
1816                                 if (!net_eq(dev_net(t->dev), net))
1817                                         unregister_netdevice_queue(t->dev,
1818                                                                    head);
1819                                 t = rtnl_dereference(t->next);
1820                         }
1821                 }
1822         }
1823 }
1824
1825 static int __net_init sit_init_net(struct net *net)
1826 {
1827         struct sit_net *sitn = net_generic(net, sit_net_id);
1828         struct ip_tunnel *t;
1829         int err;
1830
1831         sitn->tunnels[0] = sitn->tunnels_wc;
1832         sitn->tunnels[1] = sitn->tunnels_l;
1833         sitn->tunnels[2] = sitn->tunnels_r;
1834         sitn->tunnels[3] = sitn->tunnels_r_l;
1835
1836         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1837                                            NET_NAME_UNKNOWN,
1838                                            ipip6_tunnel_setup);
1839         if (!sitn->fb_tunnel_dev) {
1840                 err = -ENOMEM;
1841                 goto err_alloc_dev;
1842         }
1843         dev_net_set(sitn->fb_tunnel_dev, net);
1844         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1845         /* FB netdevice is special: we have one, and only one per netns.
1846          * Allowing to move it to another netns is clearly unsafe.
1847          */
1848         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1849
1850         err = register_netdev(sitn->fb_tunnel_dev);
1851         if (err)
1852                 goto err_reg_dev;
1853
1854         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1855         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1856
1857         t = netdev_priv(sitn->fb_tunnel_dev);
1858
1859         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1860         return 0;
1861
1862 err_reg_dev:
1863         free_netdev(sitn->fb_tunnel_dev);
1864 err_alloc_dev:
1865         return err;
1866 }
1867
1868 static void __net_exit sit_exit_net(struct net *net)
1869 {
1870         LIST_HEAD(list);
1871
1872         rtnl_lock();
1873         sit_destroy_tunnels(net, &list);
1874         unregister_netdevice_many(&list);
1875         rtnl_unlock();
1876 }
1877
1878 static struct pernet_operations sit_net_ops = {
1879         .init = sit_init_net,
1880         .exit = sit_exit_net,
1881         .id   = &sit_net_id,
1882         .size = sizeof(struct sit_net),
1883 };
1884
1885 static void __exit sit_cleanup(void)
1886 {
1887         rtnl_link_unregister(&sit_link_ops);
1888         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1889         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1890 #if IS_ENABLED(CONFIG_MPLS)
1891         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1892 #endif
1893
1894         unregister_pernet_device(&sit_net_ops);
1895         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1896 }
1897
1898 static int __init sit_init(void)
1899 {
1900         int err;
1901
1902         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1903
1904         err = register_pernet_device(&sit_net_ops);
1905         if (err < 0)
1906                 return err;
1907         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1908         if (err < 0) {
1909                 pr_info("%s: can't register ip6ip4\n", __func__);
1910                 goto xfrm_tunnel_failed;
1911         }
1912         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1913         if (err < 0) {
1914                 pr_info("%s: can't register ip4ip4\n", __func__);
1915                 goto xfrm_tunnel4_failed;
1916         }
1917 #if IS_ENABLED(CONFIG_MPLS)
1918         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1919         if (err < 0) {
1920                 pr_info("%s: can't register mplsip\n", __func__);
1921                 goto xfrm_tunnel_mpls_failed;
1922         }
1923 #endif
1924         err = rtnl_link_register(&sit_link_ops);
1925         if (err < 0)
1926                 goto rtnl_link_failed;
1927
1928 out:
1929         return err;
1930
1931 rtnl_link_failed:
1932 #if IS_ENABLED(CONFIG_MPLS)
1933         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1934 xfrm_tunnel_mpls_failed:
1935 #endif
1936         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1937 xfrm_tunnel4_failed:
1938         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1939 xfrm_tunnel_failed:
1940         unregister_pernet_device(&sit_net_ops);
1941         goto out;
1942 }
1943
1944 module_init(sit_init);
1945 module_exit(sit_cleanup);
1946 MODULE_LICENSE("GPL");
1947 MODULE_ALIAS_RTNL_LINK("sit");
1948 MODULE_ALIAS_NETDEV("sit0");