net/ipv6/sit.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  *
  10  *      Changes:
  11  * Roger Venning <r.venning@telstra.com>:       6to4 support
  12  * Nate Thompson <nate@thebog.net>:             6to4 support
  13  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  14  */
  15
  16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18 #include <linux/module.h>
  19 #include <linux/capability.h>
  20 #include <linux/errno.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/in6.h>
  26 #include <linux/netdevice.h>
  27 #include <linux/if_arp.h>
  28 #include <linux/icmp.h>
  29 #include <linux/slab.h>
  30 #include <linux/uaccess.h>
  31 #include <linux/init.h>
  32 #include <linux/netfilter_ipv4.h>
  33 #include <linux/if_ether.h>
  34
  35 #include <net/sock.h>
  36 #include <net/snmp.h>
  37
  38 #include <net/ipv6.h>
  39 #include <net/protocol.h>
  40 #include <net/transp_v6.h>
  41 #include <net/ip6_fib.h>
  42 #include <net/ip6_route.h>
  43 #include <net/ndisc.h>
  44 #include <net/addrconf.h>
  45 #include <net/ip.h>
  46 #include <net/udp.h>
  47 #include <net/icmp.h>
  48 #include <net/ip_tunnels.h>
  49 #include <net/inet_ecn.h>
  50 #include <net/xfrm.h>
  51 #include <net/dsfield.h>
  52 #include <net/net_namespace.h>
  53 #include <net/netns/generic.h>
  54
  55 /*
  56    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  57
  58    For comments look at net/ipv4/ip_gre.c --ANK
  59  */
  60
  61 #define IP6_SIT_HASH_SIZE  16
  62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  63
  64 static bool log_ecn_error = true;
  65 module_param(log_ecn_error, bool, 0644);
  66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  67
  68 static int ipip6_tunnel_init(struct net_device *dev);
  69 static void ipip6_tunnel_setup(struct net_device *dev);
  70 static void ipip6_dev_free(struct net_device *dev);
  71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  72                       __be32 *v4dst);
  73 static struct rtnl_link_ops sit_link_ops __read_mostly;
  74
  75 static unsigned int sit_net_id __read_mostly;
  76 struct sit_net {
  77         struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  78         struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  79         struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
  80         struct ip_tunnel __rcu *tunnels_wc[1];
  81         struct ip_tunnel __rcu **tunnels[4];
  82
  83         struct net_device *fb_tunnel_dev;
  84 };
  85
  86 static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
  87 {
  88         struct ip_tunnel *t = netdev_priv(dev);
  89
  90         return net_generic(t->net, sit_net_id);
  91 }
  92
  93 /*
  94  * Must be invoked with rcu_read_lock
  95  */
  96 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  97                                              struct net_device *dev,
  98                                              __be32 remote, __be32 local,
  99                                              int sifindex)
 100 {
 101         unsigned int h0 = HASH(remote);
 102         unsigned int h1 = HASH(local);
 103         struct ip_tunnel *t;
 104         struct sit_net *sitn = net_generic(net, sit_net_id);
 105         int ifindex = dev ? dev->ifindex : 0;
 106
 107         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 108                 if (local == t->parms.iph.saddr &&
 109                     remote == t->parms.iph.daddr &&
 110                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 111                      sifindex == t->parms.link) &&
 112                     (t->dev->flags & IFF_UP))
 113                         return t;
 114         }
 115         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 116                 if (remote == t->parms.iph.daddr &&
 117                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 118                      sifindex == t->parms.link) &&
 119                     (t->dev->flags & IFF_UP))
 120                         return t;
 121         }
 122         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 123                 if (local == t->parms.iph.saddr &&
 124                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 125                      sifindex == t->parms.link) &&
 126                     (t->dev->flags & IFF_UP))
 127                         return t;
 128         }
 129         t = rcu_dereference(sitn->tunnels_wc[0]);
 130         if (t && (t->dev->flags & IFF_UP))
 131                 return t;
 132         return NULL;
 133 }
 134
 135 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 136                 struct ip_tunnel_parm *parms)
 137 {
 138         __be32 remote = parms->iph.daddr;
 139         __be32 local = parms->iph.saddr;
 140         unsigned int h = 0;
 141         int prio = 0;
 142
 143         if (remote) {
 144                 prio |= 2;
 145                 h ^= HASH(remote);
 146         }
 147         if (local) {
 148                 prio |= 1;
 149                 h ^= HASH(local);
 150         }
 151         return &sitn->tunnels[prio][h];
 152 }
 153
 154 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 155                 struct ip_tunnel *t)
 156 {
 157         return __ipip6_bucket(sitn, &t->parms);
 158 }
 159
 160 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 161 {
 162         struct ip_tunnel __rcu **tp;
 163         struct ip_tunnel *iter;
 164
 165         for (tp = ipip6_bucket(sitn, t);
 166              (iter = rtnl_dereference(*tp)) != NULL;
 167              tp = &iter->next) {
 168                 if (t == iter) {
 169                         rcu_assign_pointer(*tp, t->next);
 170                         break;
 171                 }
 172         }
 173 }
 174
 175 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 176 {
 177         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 178
 179         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 180         rcu_assign_pointer(*tp, t);
 181 }
 182
 183 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 184 {
 185 #ifdef CONFIG_IPV6_SIT_6RD
 186         struct ip_tunnel *t = netdev_priv(dev);
 187
 188         if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 189                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 190                 t->ip6rd.relay_prefix = 0;
 191                 t->ip6rd.prefixlen = 16;
 192                 t->ip6rd.relay_prefixlen = 0;
 193         } else {
 194                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 195                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 196         }
 197 #endif
 198 }
 199
 200 static int ipip6_tunnel_create(struct net_device *dev)
 201 {
 202         struct ip_tunnel *t = netdev_priv(dev);
 203         struct net *net = dev_net(dev);
 204         struct sit_net *sitn = net_generic(net, sit_net_id);
 205         int err;
 206
 207         __dev_addr_set(dev, &t->parms.iph.saddr, 4);
 208         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 209
 210         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 211                 dev->priv_flags |= IFF_ISATAP;
 212
 213         dev->rtnl_link_ops = &sit_link_ops;
 214
 215         err = register_netdevice(dev);
 216         if (err < 0)
 217                 goto out;
 218
 219         ipip6_tunnel_clone_6rd(dev, sitn);
 220
 221         ipip6_tunnel_link(sitn, t);
 222         return 0;
 223
 224 out:
 225         return err;
 226 }
 227
 228 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 229                 struct ip_tunnel_parm *parms, int create)
 230 {
 231         __be32 remote = parms->iph.daddr;
 232         __be32 local = parms->iph.saddr;
 233         struct ip_tunnel *t, *nt;
 234         struct ip_tunnel __rcu **tp;
 235         struct net_device *dev;
 236         char name[IFNAMSIZ];
 237         struct sit_net *sitn = net_generic(net, sit_net_id);
 238
 239         for (tp = __ipip6_bucket(sitn, parms);
 240             (t = rtnl_dereference(*tp)) != NULL;
 241              tp = &t->next) {
 242                 if (local == t->parms.iph.saddr &&
 243                     remote == t->parms.iph.daddr &&
 244                     parms->link == t->parms.link) {
 245                         if (create)
 246                                 return NULL;
 247                         else
 248                                 return t;
 249                 }
 250         }
 251         if (!create)
 252                 goto failed;
 253
 254         if (parms->name[0]) {
 255                 if (!dev_valid_name(parms->name))
 256                         goto failed;
 257                 strscpy(name, parms->name, IFNAMSIZ);
 258         } else {
 259                 strcpy(name, "sit%d");
 260         }
 261         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 262                            ipip6_tunnel_setup);
 263         if (!dev)
 264                 return NULL;
 265
 266         dev_net_set(dev, net);
 267
 268         nt = netdev_priv(dev);
 269
 270         nt->parms = *parms;
 271         if (ipip6_tunnel_create(dev) < 0)
 272                 goto failed_free;
 273
 274         if (!parms->name[0])
 275                 strcpy(parms->name, dev->name);
 276
 277         return nt;
 278
 279 failed_free:
 280         free_netdev(dev);
 281 failed:
 282         return NULL;
 283 }
 284
 285 #define for_each_prl_rcu(start)                 \
 286         for (prl = rcu_dereference(start);      \
 287              prl;                               \
 288              prl = rcu_dereference(prl->next))
 289
 290 static struct ip_tunnel_prl_entry *
 291 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 292 {
 293         struct ip_tunnel_prl_entry *prl;
 294
 295         for_each_prl_rcu(t->prl)
 296                 if (prl->addr == addr)
 297                         break;
 298         return prl;
 299
 300 }
 301
 302 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
 303 {
 304         struct ip_tunnel *t = netdev_priv(dev);
 305         struct ip_tunnel_prl kprl, *kp;
 306         struct ip_tunnel_prl_entry *prl;
 307         unsigned int cmax, c = 0, ca, len;
 308         int ret = 0;
 309
 310         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 311                 return -EINVAL;
 312
 313         if (copy_from_user(&kprl, a, sizeof(kprl)))
 314                 return -EFAULT;
 315         cmax = kprl.datalen / sizeof(kprl);
 316         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 317                 cmax = 1;
 318
 319         /* For simple GET or for root users,
 320          * we try harder to allocate.
 321          */
 322         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 323                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
 324                 NULL;
 325
 326         ca = min(t->prl_count, cmax);
 327
 328         if (!kp) {
 329                 /* We don't try hard to allocate much memory for
 330                  * non-root users.
 331                  * For root users, retry allocating enough memory for
 332                  * the answer.
 333                  */
 334                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
 335                                               __GFP_NOWARN);
 336                 if (!kp) {
 337                         ret = -ENOMEM;
 338                         goto out;
 339                 }
 340         }
 341
 342         rcu_read_lock();
 343         for_each_prl_rcu(t->prl) {
 344                 if (c >= cmax)
 345                         break;
 346                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 347                         continue;
 348                 kp[c].addr = prl->addr;
 349                 kp[c].flags = prl->flags;
 350                 c++;
 351                 if (kprl.addr != htonl(INADDR_ANY))
 352                         break;
 353         }
 354
 355         rcu_read_unlock();
 356
 357         len = sizeof(*kp) * c;
 358         ret = 0;
 359         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 360                 ret = -EFAULT;
 361
 362         kfree(kp);
 363 out:
 364         return ret;
 365 }
 366
 367 static int
 368 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 369 {
 370         struct ip_tunnel_prl_entry *p;
 371         int err = 0;
 372
 373         if (a->addr == htonl(INADDR_ANY))
 374                 return -EINVAL;
 375
 376         ASSERT_RTNL();
 377
 378         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 379                 if (p->addr == a->addr) {
 380                         if (chg) {
 381                                 p->flags = a->flags;
 382                                 goto out;
 383                         }
 384                         err = -EEXIST;
 385                         goto out;
 386                 }
 387         }
 388
 389         if (chg) {
 390                 err = -ENXIO;
 391                 goto out;
 392         }
 393
 394         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 395         if (!p) {
 396                 err = -ENOBUFS;
 397                 goto out;
 398         }
 399
 400         p->next = t->prl;
 401         p->addr = a->addr;
 402         p->flags = a->flags;
 403         t->prl_count++;
 404         rcu_assign_pointer(t->prl, p);
 405 out:
 406         return err;
 407 }
 408
 409 static void prl_list_destroy_rcu(struct rcu_head *head)
 410 {
 411         struct ip_tunnel_prl_entry *p, *n;
 412
 413         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 414         do {
 415                 n = rcu_dereference_protected(p->next, 1);
 416                 kfree(p);
 417                 p = n;
 418         } while (p);
 419 }
 420
 421 static int
 422 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 423 {
 424         struct ip_tunnel_prl_entry *x;
 425         struct ip_tunnel_prl_entry __rcu **p;
 426         int err = 0;
 427
 428         ASSERT_RTNL();
 429
 430         if (a && a->addr != htonl(INADDR_ANY)) {
 431                 for (p = &t->prl;
 432                      (x = rtnl_dereference(*p)) != NULL;
 433                      p = &x->next) {
 434                         if (x->addr == a->addr) {
 435                                 *p = x->next;
 436                                 kfree_rcu(x, rcu_head);
 437                                 t->prl_count--;
 438                                 goto out;
 439                         }
 440                 }
 441                 err = -ENXIO;
 442         } else {
 443                 x = rtnl_dereference(t->prl);
 444                 if (x) {
 445                         t->prl_count = 0;
 446                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 447                         t->prl = NULL;
 448                 }
 449         }
 450 out:
 451         return err;
 452 }
 453
 454 static int ipip6_tunnel_prl_ctl(struct net_device *dev,
 455                                 struct ip_tunnel_prl __user *data, int cmd)
 456 {
 457         struct ip_tunnel *t = netdev_priv(dev);
 458         struct ip_tunnel_prl prl;
 459         int err;
 460
 461         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
 462                 return -EPERM;
 463         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 464                 return -EINVAL;
 465
 466         if (copy_from_user(&prl, data, sizeof(prl)))
 467                 return -EFAULT;
 468
 469         switch (cmd) {
 470         case SIOCDELPRL:
 471                 err = ipip6_tunnel_del_prl(t, &prl);
 472                 break;
 473         case SIOCADDPRL:
 474         case SIOCCHGPRL:
 475                 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 476                 break;
 477         }
 478         dst_cache_reset(&t->dst_cache);
 479         netdev_state_change(dev);
 480         return err;
 481 }
 482
 483 static int
 484 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 485 {
 486         struct ip_tunnel_prl_entry *p;
 487         int ok = 1;
 488
 489         rcu_read_lock();
 490         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 491         if (p) {
 492                 if (p->flags & PRL_DEFAULT)
 493                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 494                 else
 495                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 496         } else {
 497                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 498
 499                 if (ipv6_addr_is_isatap(addr6) &&
 500                     (addr6->s6_addr32[3] == iph->saddr) &&
 501                     ipv6_chk_prefix(addr6, t->dev))
 502                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 503                 else
 504                         ok = 0;
 505         }
 506         rcu_read_unlock();
 507         return ok;
 508 }
 509
 510 static void ipip6_tunnel_uninit(struct net_device *dev)
 511 {
 512         struct ip_tunnel *tunnel = netdev_priv(dev);
 513         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 514
 515         if (dev == sitn->fb_tunnel_dev) {
 516                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 517         } else {
 518                 ipip6_tunnel_unlink(sitn, tunnel);
 519                 ipip6_tunnel_del_prl(tunnel, NULL);
 520         }
 521         dst_cache_reset(&tunnel->dst_cache);
 522         netdev_put(dev, &tunnel->dev_tracker);
 523 }
 524
 525 static int ipip6_err(struct sk_buff *skb, u32 info)
 526 {
 527         const struct iphdr *iph = (const struct iphdr *)skb->data;
 528         const int type = icmp_hdr(skb)->type;
 529         const int code = icmp_hdr(skb)->code;
 530         unsigned int data_len = 0;
 531         struct ip_tunnel *t;
 532         int sifindex;
 533         int err;
 534
 535         switch (type) {
 536         default:
 537         case ICMP_PARAMETERPROB:
 538                 return 0;
 539
 540         case ICMP_DEST_UNREACH:
 541                 switch (code) {
 542                 case ICMP_SR_FAILED:
 543                         /* Impossible event. */
 544                         return 0;
 545                 default:
 546                         /* All others are translated to HOST_UNREACH.
 547                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 548                            I believe they are just ether pollution. --ANK
 549                          */
 550                         break;
 551                 }
 552                 break;
 553         case ICMP_TIME_EXCEEDED:
 554                 if (code != ICMP_EXC_TTL)
 555                         return 0;
 556                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 557                 break;
 558         case ICMP_REDIRECT:
 559                 break;
 560         }
 561
 562         err = -ENOENT;
 563
 564         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 565         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 566                                 iph->daddr, iph->saddr, sifindex);
 567         if (!t)
 568                 goto out;
 569
 570         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 571                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 572                                  t->parms.link, iph->protocol);
 573                 err = 0;
 574                 goto out;
 575         }
 576         if (type == ICMP_REDIRECT) {
 577                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 578                               iph->protocol);
 579                 err = 0;
 580                 goto out;
 581         }
 582
 583         err = 0;
 584         if (__in6_dev_get(skb->dev) &&
 585             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 586                 goto out;
 587
 588         if (t->parms.iph.daddr == 0)
 589                 goto out;
 590
 591         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 592                 goto out;
 593
 594         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 595                 t->err_count++;
 596         else
 597                 t->err_count = 1;
 598         t->err_time = jiffies;
 599 out:
 600         return err;
 601 }
 602
 603 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 604                                   const struct in6_addr *v6addr)
 605 {
 606         __be32 v4embed = 0;
 607         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 608                 return true;
 609         return false;
 610 }
 611
 612 /* Checks if an address matches an address on the tunnel interface.
 613  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 614  * Long story:
 615  * This function is called after we considered the packet as spoofed
 616  * in is_spoofed_6rd.
 617  * We may have a router that is doing NAT for proto 41 packets
 618  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 619  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 620  * function will return true, dropping the packet.
 621  * But, we can still check if is spoofed against the IP
 622  * addresses associated with the interface.
 623  */
 624 static bool only_dnatted(const struct ip_tunnel *tunnel,
 625         const struct in6_addr *v6dst)
 626 {
 627         int prefix_len;
 628
 629 #ifdef CONFIG_IPV6_SIT_6RD
 630         prefix_len = tunnel->ip6rd.prefixlen + 32
 631                 - tunnel->ip6rd.relay_prefixlen;
 632 #else
 633         prefix_len = 48;
 634 #endif
 635         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 636 }
 637
 638 /* Returns true if a packet is spoofed */
 639 static bool packet_is_spoofed(struct sk_buff *skb,
 640                               const struct iphdr *iph,
 641                               struct ip_tunnel *tunnel)
 642 {
 643         const struct ipv6hdr *ipv6h;
 644
 645         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 646                 if (!isatap_chksrc(skb, iph, tunnel))
 647                         return true;
 648
 649                 return false;
 650         }
 651
 652         if (tunnel->dev->flags & IFF_POINTOPOINT)
 653                 return false;
 654
 655         ipv6h = ipv6_hdr(skb);
 656
 657         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 658                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 659                                      &iph->saddr, &ipv6h->saddr,
 660                                      &iph->daddr, &ipv6h->daddr);
 661                 return true;
 662         }
 663
 664         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 665                 return false;
 666
 667         if (only_dnatted(tunnel, &ipv6h->daddr))
 668                 return false;
 669
 670         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 671                              &iph->saddr, &ipv6h->saddr,
 672                              &iph->daddr, &ipv6h->daddr);
 673         return true;
 674 }
 675
 676 static int ipip6_rcv(struct sk_buff *skb)
 677 {
 678         const struct iphdr *iph = ip_hdr(skb);
 679         struct ip_tunnel *tunnel;
 680         int sifindex;
 681         int err;
 682
 683         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 684         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 685                                      iph->saddr, iph->daddr, sifindex);
 686         if (tunnel) {
 687                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 688                     tunnel->parms.iph.protocol != 0)
 689                         goto out;
 690
 691                 skb->mac_header = skb->network_header;
 692                 skb_reset_network_header(skb);
 693                 IPCB(skb)->flags = 0;
 694                 skb->dev = tunnel->dev;
 695
 696                 if (packet_is_spoofed(skb, iph, tunnel)) {
 697                         DEV_STATS_INC(tunnel->dev, rx_errors);
 698                         goto out;
 699                 }
 700
 701                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 702                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 703                         goto out;
 704
 705                 /* skb can be uncloned in iptunnel_pull_header, so
 706                  * old iph is no longer valid
 707                  */
 708                 iph = (const struct iphdr *)skb_mac_header(skb);
 709                 skb_reset_mac_header(skb);
 710
 711                 err = IP_ECN_decapsulate(iph, skb);
 712                 if (unlikely(err)) {
 713                         if (log_ecn_error)
 714                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 715                                                      &iph->saddr, iph->tos);
 716                         if (err > 1) {
 717                                 DEV_STATS_INC(tunnel->dev, rx_frame_errors);
 718                                 DEV_STATS_INC(tunnel->dev, rx_errors);
 719                                 goto out;
 720                         }
 721                 }
 722
 723                 dev_sw_netstats_rx_add(tunnel->dev, skb->len);
 724
 725                 netif_rx(skb);
 726
 727                 return 0;
 728         }
 729
 730         /* no tunnel matched,  let upstream know, ipsec may handle it */
 731         return 1;
 732 out:
 733         kfree_skb(skb);
 734         return 0;
 735 }
 736
 737 static const struct tnl_ptk_info ipip_tpi = {
 738         /* no tunnel info required for ipip. */
 739         .proto = htons(ETH_P_IP),
 740 };
 741
 742 #if IS_ENABLED(CONFIG_MPLS)
 743 static const struct tnl_ptk_info mplsip_tpi = {
 744         /* no tunnel info required for mplsip. */
 745         .proto = htons(ETH_P_MPLS_UC),
 746 };
 747 #endif
 748
 749 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 750 {
 751         const struct iphdr *iph;
 752         struct ip_tunnel *tunnel;
 753         int sifindex;
 754
 755         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 756
 757         iph = ip_hdr(skb);
 758         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 759                                      iph->saddr, iph->daddr, sifindex);
 760         if (tunnel) {
 761                 const struct tnl_ptk_info *tpi;
 762
 763                 if (tunnel->parms.iph.protocol != ipproto &&
 764                     tunnel->parms.iph.protocol != 0)
 765                         goto drop;
 766
 767                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 768                         goto drop;
 769 #if IS_ENABLED(CONFIG_MPLS)
 770                 if (ipproto == IPPROTO_MPLS)
 771                         tpi = &mplsip_tpi;
 772                 else
 773 #endif
 774                         tpi = &ipip_tpi;
 775                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 776                         goto drop;
 777                 skb_reset_mac_header(skb);
 778
 779                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 780         }
 781
 782         return 1;
 783
 784 drop:
 785         kfree_skb(skb);
 786         return 0;
 787 }
 788
 789 static int ipip_rcv(struct sk_buff *skb)
 790 {
 791         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 792 }
 793
 794 #if IS_ENABLED(CONFIG_MPLS)
 795 static int mplsip_rcv(struct sk_buff *skb)
 796 {
 797         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 798 }
 799 #endif
 800
 801 /*
 802  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 803  * stores the embedded IPv4 address in v4dst and returns true.
 804  */
 805 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 806                       __be32 *v4dst)
 807 {
 808 #ifdef CONFIG_IPV6_SIT_6RD
 809         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 810                               tunnel->ip6rd.prefixlen)) {
 811                 unsigned int pbw0, pbi0;
 812                 int pbi1;
 813                 u32 d;
 814
 815                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 816                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 817
 818                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 819                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 820                     tunnel->ip6rd.relay_prefixlen : 0;
 821
 822                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 823                 if (pbi1 > 0)
 824                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 825                              (32 - pbi1);
 826
 827                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 828                 return true;
 829         }
 830 #else
 831         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 832                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 833                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 834                 return true;
 835         }
 836 #endif
 837         return false;
 838 }
 839
 840 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 841                              const struct in6_addr *v6dst)
 842 {
 843         __be32 dst = 0;
 844         check_6rd(tunnel, v6dst, &dst);
 845         return dst;
 846 }
 847
 848 /*
 849  *      This function assumes it is being called from dev_queue_xmit()
 850  *      and that skb is filled properly by that function.
 851  */
 852
 853 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 854                                      struct net_device *dev)
 855 {
 856         struct ip_tunnel *tunnel = netdev_priv(dev);
 857         const struct iphdr  *tiph = &tunnel->parms.iph;
 858         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 859         u8     tos = tunnel->parms.iph.tos;
 860         __be16 df = tiph->frag_off;
 861         struct rtable *rt;              /* Route to the other host */
 862         struct net_device *tdev;        /* Device to other host */
 863         unsigned int max_headroom;      /* The extra header space needed */
 864         __be32 dst = tiph->daddr;
 865         struct flowi4 fl4;
 866         int    mtu;
 867         const struct in6_addr *addr6;
 868         int addr_type;
 869         u8 ttl;
 870         u8 protocol = IPPROTO_IPV6;
 871         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 872
 873         if (tos == 1)
 874                 tos = ipv6_get_dsfield(iph6);
 875
 876         /* ISATAP (RFC4214) - must come before 6to4 */
 877         if (dev->priv_flags & IFF_ISATAP) {
 878                 struct neighbour *neigh = NULL;
 879                 bool do_tx_error = false;
 880
 881                 if (skb_dst(skb))
 882                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 883
 884                 if (!neigh) {
 885                         net_dbg_ratelimited("nexthop == NULL\n");
 886                         goto tx_error;
 887                 }
 888
 889                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 890                 addr_type = ipv6_addr_type(addr6);
 891
 892                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 893                      ipv6_addr_is_isatap(addr6))
 894                         dst = addr6->s6_addr32[3];
 895                 else
 896                         do_tx_error = true;
 897
 898                 neigh_release(neigh);
 899                 if (do_tx_error)
 900                         goto tx_error;
 901         }
 902
 903         if (!dst)
 904                 dst = try_6rd(tunnel, &iph6->daddr);
 905
 906         if (!dst) {
 907                 struct neighbour *neigh = NULL;
 908                 bool do_tx_error = false;
 909
 910                 if (skb_dst(skb))
 911                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 912
 913                 if (!neigh) {
 914                         net_dbg_ratelimited("nexthop == NULL\n");
 915                         goto tx_error;
 916                 }
 917
 918                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 919                 addr_type = ipv6_addr_type(addr6);
 920
 921                 if (addr_type == IPV6_ADDR_ANY) {
 922                         addr6 = &ipv6_hdr(skb)->daddr;
 923                         addr_type = ipv6_addr_type(addr6);
 924                 }
 925
 926                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 927                         dst = addr6->s6_addr32[3];
 928                 else
 929                         do_tx_error = true;
 930
 931                 neigh_release(neigh);
 932                 if (do_tx_error)
 933                         goto tx_error;
 934         }
 935
 936         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 937                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 938                            0, dst, tiph->saddr, 0, 0,
 939                            sock_net_uid(tunnel->net, NULL));
 940
 941         rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 942         if (!rt) {
 943                 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 944                 if (IS_ERR(rt)) {
 945                         DEV_STATS_INC(dev, tx_carrier_errors);
 946                         goto tx_error_icmp;
 947                 }
 948                 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 949         }
 950
 951         if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 952                 ip_rt_put(rt);
 953                 DEV_STATS_INC(dev, tx_carrier_errors);
 954                 goto tx_error_icmp;
 955         }
 956         tdev = rt->dst.dev;
 957
 958         if (tdev == dev) {
 959                 ip_rt_put(rt);
 960                 DEV_STATS_INC(dev, collisions);
 961                 goto tx_error;
 962         }
 963
 964         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 965                 ip_rt_put(rt);
 966                 goto tx_error;
 967         }
 968
 969         if (df) {
 970                 mtu = dst_mtu(&rt->dst) - t_hlen;
 971
 972                 if (mtu < IPV4_MIN_MTU) {
 973                         DEV_STATS_INC(dev, collisions);
 974                         ip_rt_put(rt);
 975                         goto tx_error;
 976                 }
 977
 978                 if (mtu < IPV6_MIN_MTU) {
 979                         mtu = IPV6_MIN_MTU;
 980                         df = 0;
 981                 }
 982
 983                 if (tunnel->parms.iph.daddr)
 984                         skb_dst_update_pmtu_no_confirm(skb, mtu);
 985
 986                 if (skb->len > mtu && !skb_is_gso(skb)) {
 987                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 988                         ip_rt_put(rt);
 989                         goto tx_error;
 990                 }
 991         }
 992
 993         if (tunnel->err_count > 0) {
 994                 if (time_before(jiffies,
 995                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 996                         tunnel->err_count--;
 997                         dst_link_failure(skb);
 998                 } else
 999                         tunnel->err_count = 0;
1000         }
1001
1002         /*
1003          * Okay, now see if we can stuff it in the buffer as-is.
1004          */
1005         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1006
1007         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1008             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1009                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1010                 if (!new_skb) {
1011                         ip_rt_put(rt);
1012                         DEV_STATS_INC(dev, tx_dropped);
1013                         kfree_skb(skb);
1014                         return NETDEV_TX_OK;
1015                 }
1016                 if (skb->sk)
1017                         skb_set_owner_w(new_skb, skb->sk);
1018                 dev_kfree_skb(skb);
1019                 skb = new_skb;
1020                 iph6 = ipv6_hdr(skb);
1021         }
1022         ttl = tiph->ttl;
1023         if (ttl == 0)
1024                 ttl = iph6->hop_limit;
1025         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1026
1027         if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
1028                 ip_rt_put(rt);
1029                 goto tx_error;
1030         }
1031
1032         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1033
1034         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1035                       df, !net_eq(tunnel->net, dev_net(dev)));
1036         return NETDEV_TX_OK;
1037
1038 tx_error_icmp:
1039         dst_link_failure(skb);
1040 tx_error:
1041         kfree_skb(skb);
1042         DEV_STATS_INC(dev, tx_errors);
1043         return NETDEV_TX_OK;
1044 }
1045
1046 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1047                                      struct net_device *dev, u8 ipproto)
1048 {
1049         struct ip_tunnel *tunnel = netdev_priv(dev);
1050         const struct iphdr  *tiph = &tunnel->parms.iph;
1051
1052         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1053                 goto tx_error;
1054
1055         skb_set_inner_ipproto(skb, ipproto);
1056
1057         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1058         return NETDEV_TX_OK;
1059 tx_error:
1060         kfree_skb(skb);
1061         DEV_STATS_INC(dev, tx_errors);
1062         return NETDEV_TX_OK;
1063 }
1064
1065 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1066                                    struct net_device *dev)
1067 {
1068         if (!pskb_inet_may_pull(skb))
1069                 goto tx_err;
1070
1071         switch (skb->protocol) {
1072         case htons(ETH_P_IP):
1073                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1074                 break;
1075         case htons(ETH_P_IPV6):
1076                 ipip6_tunnel_xmit(skb, dev);
1077                 break;
1078 #if IS_ENABLED(CONFIG_MPLS)
1079         case htons(ETH_P_MPLS_UC):
1080                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1081                 break;
1082 #endif
1083         default:
1084                 goto tx_err;
1085         }
1086
1087         return NETDEV_TX_OK;
1088
1089 tx_err:
1090         DEV_STATS_INC(dev, tx_errors);
1091         kfree_skb(skb);
1092         return NETDEV_TX_OK;
1093
1094 }
1095
1096 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1097 {
1098         struct net_device *tdev = NULL;
1099         struct ip_tunnel *tunnel;
1100         const struct iphdr *iph;
1101         struct flowi4 fl4;
1102
1103         tunnel = netdev_priv(dev);
1104         iph = &tunnel->parms.iph;
1105
1106         if (iph->daddr) {
1107                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1108                                                           NULL,
1109                                                           iph->daddr, iph->saddr,
1110                                                           0, 0,
1111                                                           IPPROTO_IPV6,
1112                                                           RT_TOS(iph->tos),
1113                                                           tunnel->parms.link);
1114
1115                 if (!IS_ERR(rt)) {
1116                         tdev = rt->dst.dev;
1117                         ip_rt_put(rt);
1118                 }
1119                 dev->flags |= IFF_POINTOPOINT;
1120         }
1121
1122         if (!tdev && tunnel->parms.link)
1123                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1124
1125         if (tdev && !netif_is_l3_master(tdev)) {
1126                 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1127                 int mtu;
1128
1129                 mtu = tdev->mtu - t_hlen;
1130                 if (mtu < IPV6_MIN_MTU)
1131                         mtu = IPV6_MIN_MTU;
1132                 WRITE_ONCE(dev->mtu, mtu);
1133         }
1134 }
1135
1136 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1137                                 __u32 fwmark)
1138 {
1139         struct net *net = t->net;
1140         struct sit_net *sitn = net_generic(net, sit_net_id);
1141
1142         ipip6_tunnel_unlink(sitn, t);
1143         synchronize_net();
1144         t->parms.iph.saddr = p->iph.saddr;
1145         t->parms.iph.daddr = p->iph.daddr;
1146         __dev_addr_set(t->dev, &p->iph.saddr, 4);
1147         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1148         ipip6_tunnel_link(sitn, t);
1149         t->parms.iph.ttl = p->iph.ttl;
1150         t->parms.iph.tos = p->iph.tos;
1151         t->parms.iph.frag_off = p->iph.frag_off;
1152         if (t->parms.link != p->link || t->fwmark != fwmark) {
1153                 t->parms.link = p->link;
1154                 t->fwmark = fwmark;
1155                 ipip6_tunnel_bind_dev(t->dev);
1156         }
1157         dst_cache_reset(&t->dst_cache);
1158         netdev_state_change(t->dev);
1159 }
1160
1161 #ifdef CONFIG_IPV6_SIT_6RD
1162 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1163                                    struct ip_tunnel_6rd *ip6rd)
1164 {
1165         struct in6_addr prefix;
1166         __be32 relay_prefix;
1167
1168         if (ip6rd->relay_prefixlen > 32 ||
1169             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1170                 return -EINVAL;
1171
1172         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1173         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1174                 return -EINVAL;
1175         if (ip6rd->relay_prefixlen)
1176                 relay_prefix = ip6rd->relay_prefix &
1177                                htonl(0xffffffffUL <<
1178                                      (32 - ip6rd->relay_prefixlen));
1179         else
1180                 relay_prefix = 0;
1181         if (relay_prefix != ip6rd->relay_prefix)
1182                 return -EINVAL;
1183
1184         t->ip6rd.prefix = prefix;
1185         t->ip6rd.relay_prefix = relay_prefix;
1186         t->ip6rd.prefixlen = ip6rd->prefixlen;
1187         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1188         dst_cache_reset(&t->dst_cache);
1189         netdev_state_change(t->dev);
1190         return 0;
1191 }
1192
1193 static int
1194 ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1195 {
1196         struct ip_tunnel *t = netdev_priv(dev);
1197         struct ip_tunnel_6rd ip6rd;
1198         struct ip_tunnel_parm p;
1199
1200         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1201                 if (copy_from_user(&p, data, sizeof(p)))
1202                         return -EFAULT;
1203                 t = ipip6_tunnel_locate(t->net, &p, 0);
1204         }
1205         if (!t)
1206                 t = netdev_priv(dev);
1207
1208         ip6rd.prefix = t->ip6rd.prefix;
1209         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1210         ip6rd.prefixlen = t->ip6rd.prefixlen;
1211         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1212         if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1213                 return -EFAULT;
1214         return 0;
1215 }
1216
1217 static int
1218 ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1219                     int cmd)
1220 {
1221         struct ip_tunnel *t = netdev_priv(dev);
1222         struct ip_tunnel_6rd ip6rd;
1223         int err;
1224
1225         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1226                 return -EPERM;
1227         if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1228                 return -EFAULT;
1229
1230         if (cmd != SIOCDEL6RD) {
1231                 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1232                 if (err < 0)
1233                         return err;
1234         } else
1235                 ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1236         return 0;
1237 }
1238
1239 #endif /* CONFIG_IPV6_SIT_6RD */
1240
1241 static bool ipip6_valid_ip_proto(u8 ipproto)
1242 {
1243         return ipproto == IPPROTO_IPV6 ||
1244                 ipproto == IPPROTO_IPIP ||
1245 #if IS_ENABLED(CONFIG_MPLS)
1246                 ipproto == IPPROTO_MPLS ||
1247 #endif
1248                 ipproto == 0;
1249 }
1250
1251 static int
1252 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1253 {
1254         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1255                 return -EPERM;
1256
1257         if (!ipip6_valid_ip_proto(p->iph.protocol))
1258                 return -EINVAL;
1259         if (p->iph.version != 4 ||
1260             p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1261                 return -EINVAL;
1262
1263         if (p->iph.ttl)
1264                 p->iph.frag_off |= htons(IP_DF);
1265         return 0;
1266 }
1267
1268 static int
1269 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1270 {
1271         struct ip_tunnel *t = netdev_priv(dev);
1272
1273         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1274                 t = ipip6_tunnel_locate(t->net, p, 0);
1275         if (!t)
1276                 t = netdev_priv(dev);
1277         memcpy(p, &t->parms, sizeof(*p));
1278         return 0;
1279 }
1280
1281 static int
1282 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1283 {
1284         struct ip_tunnel *t = netdev_priv(dev);
1285         int err;
1286
1287         err = __ipip6_tunnel_ioctl_validate(t->net, p);
1288         if (err)
1289                 return err;
1290
1291         t = ipip6_tunnel_locate(t->net, p, 1);
1292         if (!t)
1293                 return -ENOBUFS;
1294         return 0;
1295 }
1296
1297 static int
1298 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1299 {
1300         struct ip_tunnel *t = netdev_priv(dev);
1301         int err;
1302
1303         err = __ipip6_tunnel_ioctl_validate(t->net, p);
1304         if (err)
1305                 return err;
1306
1307         t = ipip6_tunnel_locate(t->net, p, 0);
1308         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1309                 if (!t)
1310                         return -ENOENT;
1311         } else {
1312                 if (t) {
1313                         if (t->dev != dev)
1314                                 return -EEXIST;
1315                 } else {
1316                         if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1317                             (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1318                                 return -EINVAL;
1319                         t = netdev_priv(dev);
1320                 }
1321
1322                 ipip6_tunnel_update(t, p, t->fwmark);
1323         }
1324
1325         return 0;
1326 }
1327
1328 static int
1329 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1330 {
1331         struct ip_tunnel *t = netdev_priv(dev);
1332
1333         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1334                 return -EPERM;
1335
1336         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1337                 t = ipip6_tunnel_locate(t->net, p, 0);
1338                 if (!t)
1339                         return -ENOENT;
1340                 if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1341                         return -EPERM;
1342                 dev = t->dev;
1343         }
1344         unregister_netdevice(dev);
1345         return 0;
1346 }
1347
1348 static int
1349 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1350 {
1351         switch (cmd) {
1352         case SIOCGETTUNNEL:
1353                 return ipip6_tunnel_get(dev, p);
1354         case SIOCADDTUNNEL:
1355                 return ipip6_tunnel_add(dev, p);
1356         case SIOCCHGTUNNEL:
1357                 return ipip6_tunnel_change(dev, p);
1358         case SIOCDELTUNNEL:
1359                 return ipip6_tunnel_del(dev, p);
1360         default:
1361                 return -EINVAL;
1362         }
1363 }
1364
1365 static int
1366 ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1367                             void __user *data, int cmd)
1368 {
1369         switch (cmd) {
1370         case SIOCGETTUNNEL:
1371         case SIOCADDTUNNEL:
1372         case SIOCCHGTUNNEL:
1373         case SIOCDELTUNNEL:
1374                 return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1375         case SIOCGETPRL:
1376                 return ipip6_tunnel_get_prl(dev, data);
1377         case SIOCADDPRL:
1378         case SIOCDELPRL:
1379         case SIOCCHGPRL:
1380                 return ipip6_tunnel_prl_ctl(dev, data, cmd);
1381 #ifdef CONFIG_IPV6_SIT_6RD
1382         case SIOCGET6RD:
1383                 return ipip6_tunnel_get6rd(dev, data);
1384         case SIOCADD6RD:
1385         case SIOCCHG6RD:
1386         case SIOCDEL6RD:
1387                 return ipip6_tunnel_6rdctl(dev, data, cmd);
1388 #endif
1389         default:
1390                 return -EINVAL;
1391         }
1392 }
1393
1394 static const struct net_device_ops ipip6_netdev_ops = {
1395         .ndo_init       = ipip6_tunnel_init,
1396         .ndo_uninit     = ipip6_tunnel_uninit,
1397         .ndo_start_xmit = sit_tunnel_xmit,
1398         .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1399         .ndo_get_stats64 = dev_get_tstats64,
1400         .ndo_get_iflink = ip_tunnel_get_iflink,
1401         .ndo_tunnel_ctl = ipip6_tunnel_ctl,
1402 };
1403
1404 static void ipip6_dev_free(struct net_device *dev)
1405 {
1406         struct ip_tunnel *tunnel = netdev_priv(dev);
1407
1408         dst_cache_destroy(&tunnel->dst_cache);
1409         free_percpu(dev->tstats);
1410 }
1411
1412 #define SIT_FEATURES (NETIF_F_SG           | \
1413                       NETIF_F_FRAGLIST     | \
1414                       NETIF_F_HIGHDMA      | \
1415                       NETIF_F_GSO_SOFTWARE | \
1416                       NETIF_F_HW_CSUM)
1417
1418 static void ipip6_tunnel_setup(struct net_device *dev)
1419 {
1420         struct ip_tunnel *tunnel = netdev_priv(dev);
1421         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1422
1423         dev->netdev_ops         = &ipip6_netdev_ops;
1424         dev->header_ops         = &ip_tunnel_header_ops;
1425         dev->needs_free_netdev  = true;
1426         dev->priv_destructor    = ipip6_dev_free;
1427
1428         dev->type               = ARPHRD_SIT;
1429         dev->mtu                = ETH_DATA_LEN - t_hlen;
1430         dev->min_mtu            = IPV6_MIN_MTU;
1431         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1432         dev->flags              = IFF_NOARP;
1433         netif_keep_dst(dev);
1434         dev->addr_len           = 4;
1435         dev->features           |= NETIF_F_LLTX;
1436         dev->features           |= SIT_FEATURES;
1437         dev->hw_features        |= SIT_FEATURES;
1438 }
1439
1440 static int ipip6_tunnel_init(struct net_device *dev)
1441 {
1442         struct ip_tunnel *tunnel = netdev_priv(dev);
1443         int err;
1444
1445         tunnel->dev = dev;
1446         tunnel->net = dev_net(dev);
1447         strcpy(tunnel->parms.name, dev->name);
1448
1449         ipip6_tunnel_bind_dev(dev);
1450         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1451         if (!dev->tstats)
1452                 return -ENOMEM;
1453
1454         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1455         if (err) {
1456                 free_percpu(dev->tstats);
1457                 dev->tstats = NULL;
1458                 return err;
1459         }
1460         netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1461         return 0;
1462 }
1463
1464 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1465 {
1466         struct ip_tunnel *tunnel = netdev_priv(dev);
1467         struct iphdr *iph = &tunnel->parms.iph;
1468         struct net *net = dev_net(dev);
1469         struct sit_net *sitn = net_generic(net, sit_net_id);
1470
1471         iph->version            = 4;
1472         iph->protocol           = IPPROTO_IPV6;
1473         iph->ihl                = 5;
1474         iph->ttl                = 64;
1475
1476         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1477 }
1478
1479 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1480                           struct netlink_ext_ack *extack)
1481 {
1482         u8 proto;
1483
1484         if (!data || !data[IFLA_IPTUN_PROTO])
1485                 return 0;
1486
1487         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1488         if (!ipip6_valid_ip_proto(proto))
1489                 return -EINVAL;
1490
1491         return 0;
1492 }
1493
1494 static void ipip6_netlink_parms(struct nlattr *data[],
1495                                 struct ip_tunnel_parm *parms,
1496                                 __u32 *fwmark)
1497 {
1498         memset(parms, 0, sizeof(*parms));
1499
1500         parms->iph.version = 4;
1501         parms->iph.protocol = IPPROTO_IPV6;
1502         parms->iph.ihl = 5;
1503         parms->iph.ttl = 64;
1504
1505         if (!data)
1506                 return;
1507
1508         ip_tunnel_netlink_parms(data, parms);
1509
1510         if (data[IFLA_IPTUN_FWMARK])
1511                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1512 }
1513
1514 #ifdef CONFIG_IPV6_SIT_6RD
1515 /* This function returns true when 6RD attributes are present in the nl msg */
1516 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1517                                     struct ip_tunnel_6rd *ip6rd)
1518 {
1519         bool ret = false;
1520         memset(ip6rd, 0, sizeof(*ip6rd));
1521
1522         if (!data)
1523                 return ret;
1524
1525         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1526                 ret = true;
1527                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1528         }
1529
1530         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1531                 ret = true;
1532                 ip6rd->relay_prefix =
1533                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1534         }
1535
1536         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1537                 ret = true;
1538                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1539         }
1540
1541         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1542                 ret = true;
1543                 ip6rd->relay_prefixlen =
1544                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1545         }
1546
1547         return ret;
1548 }
1549 #endif
1550
1551 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1552                          struct nlattr *tb[], struct nlattr *data[],
1553                          struct netlink_ext_ack *extack)
1554 {
1555         struct net *net = dev_net(dev);
1556         struct ip_tunnel *nt;
1557         struct ip_tunnel_encap ipencap;
1558 #ifdef CONFIG_IPV6_SIT_6RD
1559         struct ip_tunnel_6rd ip6rd;
1560 #endif
1561         int err;
1562
1563         nt = netdev_priv(dev);
1564
1565         if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1566                 err = ip_tunnel_encap_setup(nt, &ipencap);
1567                 if (err < 0)
1568                         return err;
1569         }
1570
1571         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1572
1573         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1574                 return -EEXIST;
1575
1576         err = ipip6_tunnel_create(dev);
1577         if (err < 0)
1578                 return err;
1579
1580         if (tb[IFLA_MTU]) {
1581                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1582
1583                 if (mtu >= IPV6_MIN_MTU &&
1584                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1585                         dev->mtu = mtu;
1586         }
1587
1588 #ifdef CONFIG_IPV6_SIT_6RD
1589         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1590                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1591                 if (err < 0)
1592                         unregister_netdevice_queue(dev, NULL);
1593         }
1594 #endif
1595
1596         return err;
1597 }
1598
1599 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1600                             struct nlattr *data[],
1601                             struct netlink_ext_ack *extack)
1602 {
1603         struct ip_tunnel *t = netdev_priv(dev);
1604         struct ip_tunnel_parm p;
1605         struct ip_tunnel_encap ipencap;
1606         struct net *net = t->net;
1607         struct sit_net *sitn = net_generic(net, sit_net_id);
1608 #ifdef CONFIG_IPV6_SIT_6RD
1609         struct ip_tunnel_6rd ip6rd;
1610 #endif
1611         __u32 fwmark = t->fwmark;
1612         int err;
1613
1614         if (dev == sitn->fb_tunnel_dev)
1615                 return -EINVAL;
1616
1617         if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1618                 err = ip_tunnel_encap_setup(t, &ipencap);
1619                 if (err < 0)
1620                         return err;
1621         }
1622
1623         ipip6_netlink_parms(data, &p, &fwmark);
1624
1625         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1626             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1627                 return -EINVAL;
1628
1629         t = ipip6_tunnel_locate(net, &p, 0);
1630
1631         if (t) {
1632                 if (t->dev != dev)
1633                         return -EEXIST;
1634         } else
1635                 t = netdev_priv(dev);
1636
1637         ipip6_tunnel_update(t, &p, fwmark);
1638
1639 #ifdef CONFIG_IPV6_SIT_6RD
1640         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1641                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1642 #endif
1643
1644         return 0;
1645 }
1646
1647 static size_t ipip6_get_size(const struct net_device *dev)
1648 {
1649         return
1650                 /* IFLA_IPTUN_LINK */
1651                 nla_total_size(4) +
1652                 /* IFLA_IPTUN_LOCAL */
1653                 nla_total_size(4) +
1654                 /* IFLA_IPTUN_REMOTE */
1655                 nla_total_size(4) +
1656                 /* IFLA_IPTUN_TTL */
1657                 nla_total_size(1) +
1658                 /* IFLA_IPTUN_TOS */
1659                 nla_total_size(1) +
1660                 /* IFLA_IPTUN_PMTUDISC */
1661                 nla_total_size(1) +
1662                 /* IFLA_IPTUN_FLAGS */
1663                 nla_total_size(2) +
1664                 /* IFLA_IPTUN_PROTO */
1665                 nla_total_size(1) +
1666 #ifdef CONFIG_IPV6_SIT_6RD
1667                 /* IFLA_IPTUN_6RD_PREFIX */
1668                 nla_total_size(sizeof(struct in6_addr)) +
1669                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1670                 nla_total_size(4) +
1671                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1672                 nla_total_size(2) +
1673                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1674                 nla_total_size(2) +
1675 #endif
1676                 /* IFLA_IPTUN_ENCAP_TYPE */
1677                 nla_total_size(2) +
1678                 /* IFLA_IPTUN_ENCAP_FLAGS */
1679                 nla_total_size(2) +
1680                 /* IFLA_IPTUN_ENCAP_SPORT */
1681                 nla_total_size(2) +
1682                 /* IFLA_IPTUN_ENCAP_DPORT */
1683                 nla_total_size(2) +
1684                 /* IFLA_IPTUN_FWMARK */
1685                 nla_total_size(4) +
1686                 0;
1687 }
1688
1689 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1690 {
1691         struct ip_tunnel *tunnel = netdev_priv(dev);
1692         struct ip_tunnel_parm *parm = &tunnel->parms;
1693
1694         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1695             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1696             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1697             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1698             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1699             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1700                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1701             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1702             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1703             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1704                 goto nla_put_failure;
1705
1706 #ifdef CONFIG_IPV6_SIT_6RD
1707         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1708                              &tunnel->ip6rd.prefix) ||
1709             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1710                             tunnel->ip6rd.relay_prefix) ||
1711             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1712                         tunnel->ip6rd.prefixlen) ||
1713             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1714                         tunnel->ip6rd.relay_prefixlen))
1715                 goto nla_put_failure;
1716 #endif
1717
1718         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1719                         tunnel->encap.type) ||
1720             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1721                         tunnel->encap.sport) ||
1722             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1723                         tunnel->encap.dport) ||
1724             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1725                         tunnel->encap.flags))
1726                 goto nla_put_failure;
1727
1728         return 0;
1729
1730 nla_put_failure:
1731         return -EMSGSIZE;
1732 }
1733
1734 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1735         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1736         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1737         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1738         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1739         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1740         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1741         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1742         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1743 #ifdef CONFIG_IPV6_SIT_6RD
1744         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1745         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1746         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1747         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1748 #endif
1749         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1750         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1751         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1752         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1753         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1754 };
1755
1756 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1757 {
1758         struct net *net = dev_net(dev);
1759         struct sit_net *sitn = net_generic(net, sit_net_id);
1760
1761         if (dev != sitn->fb_tunnel_dev)
1762                 unregister_netdevice_queue(dev, head);
1763 }
1764
1765 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1766         .kind           = "sit",
1767         .maxtype        = IFLA_IPTUN_MAX,
1768         .policy         = ipip6_policy,
1769         .priv_size      = sizeof(struct ip_tunnel),
1770         .setup          = ipip6_tunnel_setup,
1771         .validate       = ipip6_validate,
1772         .newlink        = ipip6_newlink,
1773         .changelink     = ipip6_changelink,
1774         .get_size       = ipip6_get_size,
1775         .fill_info      = ipip6_fill_info,
1776         .dellink        = ipip6_dellink,
1777         .get_link_net   = ip_tunnel_get_link_net,
1778 };
1779
1780 static struct xfrm_tunnel sit_handler __read_mostly = {
1781         .handler        =       ipip6_rcv,
1782         .err_handler    =       ipip6_err,
1783         .priority       =       1,
1784 };
1785
1786 static struct xfrm_tunnel ipip_handler __read_mostly = {
1787         .handler        =       ipip_rcv,
1788         .err_handler    =       ipip6_err,
1789         .priority       =       2,
1790 };
1791
1792 #if IS_ENABLED(CONFIG_MPLS)
1793 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1794         .handler        =       mplsip_rcv,
1795         .err_handler    =       ipip6_err,
1796         .priority       =       2,
1797 };
1798 #endif
1799
1800 static void __net_exit sit_destroy_tunnels(struct net *net,
1801                                            struct list_head *head)
1802 {
1803         struct sit_net *sitn = net_generic(net, sit_net_id);
1804         struct net_device *dev, *aux;
1805         int prio;
1806
1807         for_each_netdev_safe(net, dev, aux)
1808                 if (dev->rtnl_link_ops == &sit_link_ops)
1809                         unregister_netdevice_queue(dev, head);
1810
1811         for (prio = 0; prio < 4; prio++) {
1812                 int h;
1813                 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1814                         struct ip_tunnel *t;
1815
1816                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1817                         while (t) {
1818                                 /* If dev is in the same netns, it has already
1819                                  * been added to the list by the previous loop.
1820                                  */
1821                                 if (!net_eq(dev_net(t->dev), net))
1822                                         unregister_netdevice_queue(t->dev,
1823                                                                    head);
1824                                 t = rtnl_dereference(t->next);
1825                         }
1826                 }
1827         }
1828 }
1829
1830 static int __net_init sit_init_net(struct net *net)
1831 {
1832         struct sit_net *sitn = net_generic(net, sit_net_id);
1833         struct ip_tunnel *t;
1834         int err;
1835
1836         sitn->tunnels[0] = sitn->tunnels_wc;
1837         sitn->tunnels[1] = sitn->tunnels_l;
1838         sitn->tunnels[2] = sitn->tunnels_r;
1839         sitn->tunnels[3] = sitn->tunnels_r_l;
1840
1841         if (!net_has_fallback_tunnels(net))
1842                 return 0;
1843
1844         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1845                                            NET_NAME_UNKNOWN,
1846                                            ipip6_tunnel_setup);
1847         if (!sitn->fb_tunnel_dev) {
1848                 err = -ENOMEM;
1849                 goto err_alloc_dev;
1850         }
1851         dev_net_set(sitn->fb_tunnel_dev, net);
1852         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1853         /* FB netdevice is special: we have one, and only one per netns.
1854          * Allowing to move it to another netns is clearly unsafe.
1855          */
1856         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1857
1858         err = register_netdev(sitn->fb_tunnel_dev);
1859         if (err)
1860                 goto err_reg_dev;
1861
1862         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1863         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1864
1865         t = netdev_priv(sitn->fb_tunnel_dev);
1866
1867         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1868         return 0;
1869
1870 err_reg_dev:
1871         free_netdev(sitn->fb_tunnel_dev);
1872 err_alloc_dev:
1873         return err;
1874 }
1875
1876 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1877 {
1878         LIST_HEAD(list);
1879         struct net *net;
1880
1881         rtnl_lock();
1882         list_for_each_entry(net, net_list, exit_list)
1883                 sit_destroy_tunnels(net, &list);
1884
1885         unregister_netdevice_many(&list);
1886         rtnl_unlock();
1887 }
1888
1889 static struct pernet_operations sit_net_ops = {
1890         .init = sit_init_net,
1891         .exit_batch = sit_exit_batch_net,
1892         .id   = &sit_net_id,
1893         .size = sizeof(struct sit_net),
1894 };
1895
1896 static void __exit sit_cleanup(void)
1897 {
1898         rtnl_link_unregister(&sit_link_ops);
1899         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1900         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1901 #if IS_ENABLED(CONFIG_MPLS)
1902         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1903 #endif
1904
1905         unregister_pernet_device(&sit_net_ops);
1906         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1907 }
1908
1909 static int __init sit_init(void)
1910 {
1911         int err;
1912
1913         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1914
1915         err = register_pernet_device(&sit_net_ops);
1916         if (err < 0)
1917                 return err;
1918         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1919         if (err < 0) {
1920                 pr_info("%s: can't register ip6ip4\n", __func__);
1921                 goto xfrm_tunnel_failed;
1922         }
1923         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1924         if (err < 0) {
1925                 pr_info("%s: can't register ip4ip4\n", __func__);
1926                 goto xfrm_tunnel4_failed;
1927         }
1928 #if IS_ENABLED(CONFIG_MPLS)
1929         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1930         if (err < 0) {
1931                 pr_info("%s: can't register mplsip\n", __func__);
1932                 goto xfrm_tunnel_mpls_failed;
1933         }
1934 #endif
1935         err = rtnl_link_register(&sit_link_ops);
1936         if (err < 0)
1937                 goto rtnl_link_failed;
1938
1939 out:
1940         return err;
1941
1942 rtnl_link_failed:
1943 #if IS_ENABLED(CONFIG_MPLS)
1944         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1945 xfrm_tunnel_mpls_failed:
1946 #endif
1947         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1948 xfrm_tunnel4_failed:
1949         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1950 xfrm_tunnel_failed:
1951         unregister_pernet_device(&sit_net_ops);
1952         goto out;
1953 }
1954
1955 module_init(sit_init);
1956 module_exit(sit_cleanup);
1957 MODULE_LICENSE("GPL");
1958 MODULE_ALIAS_RTNL_LINK("sit");
1959 MODULE_ALIAS_NETDEV("sit0");