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         memcpy(dev->dev_addr, &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                 strlcpy(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         return nt;
 275
 276 failed_free:
 277         free_netdev(dev);
 278 failed:
 279         return NULL;
 280 }
 281
 282 #define for_each_prl_rcu(start)                 \
 283         for (prl = rcu_dereference(start);      \
 284              prl;                               \
 285              prl = rcu_dereference(prl->next))
 286
 287 static struct ip_tunnel_prl_entry *
 288 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 289 {
 290         struct ip_tunnel_prl_entry *prl;
 291
 292         for_each_prl_rcu(t->prl)
 293                 if (prl->addr == addr)
 294                         break;
 295         return prl;
 296
 297 }
 298
 299 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ifreq *ifr)
 300 {
 301         struct ip_tunnel_prl __user *a = ifr->ifr_ifru.ifru_data;
 302         struct ip_tunnel *t = netdev_priv(dev);
 303         struct ip_tunnel_prl kprl, *kp;
 304         struct ip_tunnel_prl_entry *prl;
 305         unsigned int cmax, c = 0, ca, len;
 306         int ret = 0;
 307
 308         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 309                 return -EINVAL;
 310
 311         if (copy_from_user(&kprl, a, sizeof(kprl)))
 312                 return -EFAULT;
 313         cmax = kprl.datalen / sizeof(kprl);
 314         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 315                 cmax = 1;
 316
 317         /* For simple GET or for root users,
 318          * we try harder to allocate.
 319          */
 320         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 321                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 322                 NULL;
 323
 324         ca = min(t->prl_count, cmax);
 325
 326         if (!kp) {
 327                 /* We don't try hard to allocate much memory for
 328                  * non-root users.
 329                  * For root users, retry allocating enough memory for
 330                  * the answer.
 331                  */
 332                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 333                 if (!kp) {
 334                         ret = -ENOMEM;
 335                         goto out;
 336                 }
 337         }
 338
 339         rcu_read_lock();
 340         for_each_prl_rcu(t->prl) {
 341                 if (c >= cmax)
 342                         break;
 343                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 344                         continue;
 345                 kp[c].addr = prl->addr;
 346                 kp[c].flags = prl->flags;
 347                 c++;
 348                 if (kprl.addr != htonl(INADDR_ANY))
 349                         break;
 350         }
 351
 352         rcu_read_unlock();
 353
 354         len = sizeof(*kp) * c;
 355         ret = 0;
 356         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 357                 ret = -EFAULT;
 358
 359         kfree(kp);
 360 out:
 361         return ret;
 362 }
 363
 364 static int
 365 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 366 {
 367         struct ip_tunnel_prl_entry *p;
 368         int err = 0;
 369
 370         if (a->addr == htonl(INADDR_ANY))
 371                 return -EINVAL;
 372
 373         ASSERT_RTNL();
 374
 375         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 376                 if (p->addr == a->addr) {
 377                         if (chg) {
 378                                 p->flags = a->flags;
 379                                 goto out;
 380                         }
 381                         err = -EEXIST;
 382                         goto out;
 383                 }
 384         }
 385
 386         if (chg) {
 387                 err = -ENXIO;
 388                 goto out;
 389         }
 390
 391         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 392         if (!p) {
 393                 err = -ENOBUFS;
 394                 goto out;
 395         }
 396
 397         p->next = t->prl;
 398         p->addr = a->addr;
 399         p->flags = a->flags;
 400         t->prl_count++;
 401         rcu_assign_pointer(t->prl, p);
 402 out:
 403         return err;
 404 }
 405
 406 static void prl_list_destroy_rcu(struct rcu_head *head)
 407 {
 408         struct ip_tunnel_prl_entry *p, *n;
 409
 410         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 411         do {
 412                 n = rcu_dereference_protected(p->next, 1);
 413                 kfree(p);
 414                 p = n;
 415         } while (p);
 416 }
 417
 418 static int
 419 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 420 {
 421         struct ip_tunnel_prl_entry *x;
 422         struct ip_tunnel_prl_entry __rcu **p;
 423         int err = 0;
 424
 425         ASSERT_RTNL();
 426
 427         if (a && a->addr != htonl(INADDR_ANY)) {
 428                 for (p = &t->prl;
 429                      (x = rtnl_dereference(*p)) != NULL;
 430                      p = &x->next) {
 431                         if (x->addr == a->addr) {
 432                                 *p = x->next;
 433                                 kfree_rcu(x, rcu_head);
 434                                 t->prl_count--;
 435                                 goto out;
 436                         }
 437                 }
 438                 err = -ENXIO;
 439         } else {
 440                 x = rtnl_dereference(t->prl);
 441                 if (x) {
 442                         t->prl_count = 0;
 443                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 444                         t->prl = NULL;
 445                 }
 446         }
 447 out:
 448         return err;
 449 }
 450
 451 static int ipip6_tunnel_prl_ctl(struct net_device *dev, struct ifreq *ifr,
 452                 int cmd)
 453 {
 454         struct ip_tunnel *t = netdev_priv(dev);
 455         struct ip_tunnel_prl prl;
 456         int err;
 457
 458         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
 459                 return -EPERM;
 460         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 461                 return -EINVAL;
 462
 463         if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 464                 return -EFAULT;
 465
 466         switch (cmd) {
 467         case SIOCDELPRL:
 468                 err = ipip6_tunnel_del_prl(t, &prl);
 469                 break;
 470         case SIOCADDPRL:
 471         case SIOCCHGPRL:
 472                 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 473                 break;
 474         }
 475         dst_cache_reset(&t->dst_cache);
 476         netdev_state_change(dev);
 477         return err;
 478 }
 479
 480 static int
 481 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 482 {
 483         struct ip_tunnel_prl_entry *p;
 484         int ok = 1;
 485
 486         rcu_read_lock();
 487         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 488         if (p) {
 489                 if (p->flags & PRL_DEFAULT)
 490                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 491                 else
 492                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 493         } else {
 494                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 495
 496                 if (ipv6_addr_is_isatap(addr6) &&
 497                     (addr6->s6_addr32[3] == iph->saddr) &&
 498                     ipv6_chk_prefix(addr6, t->dev))
 499                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 500                 else
 501                         ok = 0;
 502         }
 503         rcu_read_unlock();
 504         return ok;
 505 }
 506
 507 static void ipip6_tunnel_uninit(struct net_device *dev)
 508 {
 509         struct ip_tunnel *tunnel = netdev_priv(dev);
 510         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 511
 512         if (dev == sitn->fb_tunnel_dev) {
 513                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 514         } else {
 515                 ipip6_tunnel_unlink(sitn, tunnel);
 516                 ipip6_tunnel_del_prl(tunnel, NULL);
 517         }
 518         dst_cache_reset(&tunnel->dst_cache);
 519         dev_put(dev);
 520 }
 521
 522 static int ipip6_err(struct sk_buff *skb, u32 info)
 523 {
 524         const struct iphdr *iph = (const struct iphdr *)skb->data;
 525         const int type = icmp_hdr(skb)->type;
 526         const int code = icmp_hdr(skb)->code;
 527         unsigned int data_len = 0;
 528         struct ip_tunnel *t;
 529         int sifindex;
 530         int err;
 531
 532         switch (type) {
 533         default:
 534         case ICMP_PARAMETERPROB:
 535                 return 0;
 536
 537         case ICMP_DEST_UNREACH:
 538                 switch (code) {
 539                 case ICMP_SR_FAILED:
 540                         /* Impossible event. */
 541                         return 0;
 542                 default:
 543                         /* All others are translated to HOST_UNREACH.
 544                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 545                            I believe they are just ether pollution. --ANK
 546                          */
 547                         break;
 548                 }
 549                 break;
 550         case ICMP_TIME_EXCEEDED:
 551                 if (code != ICMP_EXC_TTL)
 552                         return 0;
 553                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 554                 break;
 555         case ICMP_REDIRECT:
 556                 break;
 557         }
 558
 559         err = -ENOENT;
 560
 561         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 562         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 563                                 iph->daddr, iph->saddr, sifindex);
 564         if (!t)
 565                 goto out;
 566
 567         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 568                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 569                                  t->parms.link, iph->protocol);
 570                 err = 0;
 571                 goto out;
 572         }
 573         if (type == ICMP_REDIRECT) {
 574                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 575                               iph->protocol);
 576                 err = 0;
 577                 goto out;
 578         }
 579
 580         err = 0;
 581         if (__in6_dev_get(skb->dev) &&
 582             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 583                 goto out;
 584
 585         if (t->parms.iph.daddr == 0)
 586                 goto out;
 587
 588         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 589                 goto out;
 590
 591         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 592                 t->err_count++;
 593         else
 594                 t->err_count = 1;
 595         t->err_time = jiffies;
 596 out:
 597         return err;
 598 }
 599
 600 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 601                                   const struct in6_addr *v6addr)
 602 {
 603         __be32 v4embed = 0;
 604         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 605                 return true;
 606         return false;
 607 }
 608
 609 /* Checks if an address matches an address on the tunnel interface.
 610  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 611  * Long story:
 612  * This function is called after we considered the packet as spoofed
 613  * in is_spoofed_6rd.
 614  * We may have a router that is doing NAT for proto 41 packets
 615  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 616  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 617  * function will return true, dropping the packet.
 618  * But, we can still check if is spoofed against the IP
 619  * addresses associated with the interface.
 620  */
 621 static bool only_dnatted(const struct ip_tunnel *tunnel,
 622         const struct in6_addr *v6dst)
 623 {
 624         int prefix_len;
 625
 626 #ifdef CONFIG_IPV6_SIT_6RD
 627         prefix_len = tunnel->ip6rd.prefixlen + 32
 628                 - tunnel->ip6rd.relay_prefixlen;
 629 #else
 630         prefix_len = 48;
 631 #endif
 632         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 633 }
 634
 635 /* Returns true if a packet is spoofed */
 636 static bool packet_is_spoofed(struct sk_buff *skb,
 637                               const struct iphdr *iph,
 638                               struct ip_tunnel *tunnel)
 639 {
 640         const struct ipv6hdr *ipv6h;
 641
 642         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 643                 if (!isatap_chksrc(skb, iph, tunnel))
 644                         return true;
 645
 646                 return false;
 647         }
 648
 649         if (tunnel->dev->flags & IFF_POINTOPOINT)
 650                 return false;
 651
 652         ipv6h = ipv6_hdr(skb);
 653
 654         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 655                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 656                                      &iph->saddr, &ipv6h->saddr,
 657                                      &iph->daddr, &ipv6h->daddr);
 658                 return true;
 659         }
 660
 661         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 662                 return false;
 663
 664         if (only_dnatted(tunnel, &ipv6h->daddr))
 665                 return false;
 666
 667         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 668                              &iph->saddr, &ipv6h->saddr,
 669                              &iph->daddr, &ipv6h->daddr);
 670         return true;
 671 }
 672
 673 static int ipip6_rcv(struct sk_buff *skb)
 674 {
 675         const struct iphdr *iph = ip_hdr(skb);
 676         struct ip_tunnel *tunnel;
 677         int sifindex;
 678         int err;
 679
 680         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 681         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 682                                      iph->saddr, iph->daddr, sifindex);
 683         if (tunnel) {
 684                 struct pcpu_sw_netstats *tstats;
 685
 686                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 687                     tunnel->parms.iph.protocol != 0)
 688                         goto out;
 689
 690                 skb->mac_header = skb->network_header;
 691                 skb_reset_network_header(skb);
 692                 IPCB(skb)->flags = 0;
 693                 skb->dev = tunnel->dev;
 694
 695                 if (packet_is_spoofed(skb, iph, tunnel)) {
 696                         tunnel->dev->stats.rx_errors++;
 697                         goto out;
 698                 }
 699
 700                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 701                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 702                         goto out;
 703
 704                 /* skb can be uncloned in iptunnel_pull_header, so
 705                  * old iph is no longer valid
 706                  */
 707                 iph = (const struct iphdr *)skb_mac_header(skb);
 708                 err = IP_ECN_decapsulate(iph, skb);
 709                 if (unlikely(err)) {
 710                         if (log_ecn_error)
 711                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 712                                                      &iph->saddr, iph->tos);
 713                         if (err > 1) {
 714                                 ++tunnel->dev->stats.rx_frame_errors;
 715                                 ++tunnel->dev->stats.rx_errors;
 716                                 goto out;
 717                         }
 718                 }
 719
 720                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 721                 u64_stats_update_begin(&tstats->syncp);
 722                 tstats->rx_packets++;
 723                 tstats->rx_bytes += skb->len;
 724                 u64_stats_update_end(&tstats->syncp);
 725
 726                 netif_rx(skb);
 727
 728                 return 0;
 729         }
 730
 731         /* no tunnel matched,  let upstream know, ipsec may handle it */
 732         return 1;
 733 out:
 734         kfree_skb(skb);
 735         return 0;
 736 }
 737
 738 static const struct tnl_ptk_info ipip_tpi = {
 739         /* no tunnel info required for ipip. */
 740         .proto = htons(ETH_P_IP),
 741 };
 742
 743 #if IS_ENABLED(CONFIG_MPLS)
 744 static const struct tnl_ptk_info mplsip_tpi = {
 745         /* no tunnel info required for mplsip. */
 746         .proto = htons(ETH_P_MPLS_UC),
 747 };
 748 #endif
 749
 750 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 751 {
 752         const struct iphdr *iph;
 753         struct ip_tunnel *tunnel;
 754         int sifindex;
 755
 756         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 757
 758         iph = ip_hdr(skb);
 759         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 760                                      iph->saddr, iph->daddr, sifindex);
 761         if (tunnel) {
 762                 const struct tnl_ptk_info *tpi;
 763
 764                 if (tunnel->parms.iph.protocol != ipproto &&
 765                     tunnel->parms.iph.protocol != 0)
 766                         goto drop;
 767
 768                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 769                         goto drop;
 770 #if IS_ENABLED(CONFIG_MPLS)
 771                 if (ipproto == IPPROTO_MPLS)
 772                         tpi = &mplsip_tpi;
 773                 else
 774 #endif
 775                         tpi = &ipip_tpi;
 776                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 777                         goto drop;
 778                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 779         }
 780
 781         return 1;
 782
 783 drop:
 784         kfree_skb(skb);
 785         return 0;
 786 }
 787
 788 static int ipip_rcv(struct sk_buff *skb)
 789 {
 790         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 791 }
 792
 793 #if IS_ENABLED(CONFIG_MPLS)
 794 static int mplsip_rcv(struct sk_buff *skb)
 795 {
 796         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 797 }
 798 #endif
 799
 800 /*
 801  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 802  * stores the embedded IPv4 address in v4dst and returns true.
 803  */
 804 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 805                       __be32 *v4dst)
 806 {
 807 #ifdef CONFIG_IPV6_SIT_6RD
 808         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 809                               tunnel->ip6rd.prefixlen)) {
 810                 unsigned int pbw0, pbi0;
 811                 int pbi1;
 812                 u32 d;
 813
 814                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 815                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 816
 817                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 818                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 819                     tunnel->ip6rd.relay_prefixlen : 0;
 820
 821                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 822                 if (pbi1 > 0)
 823                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 824                              (32 - pbi1);
 825
 826                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 827                 return true;
 828         }
 829 #else
 830         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 831                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 832                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 833                 return true;
 834         }
 835 #endif
 836         return false;
 837 }
 838
 839 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 840                              const struct in6_addr *v6dst)
 841 {
 842         __be32 dst = 0;
 843         check_6rd(tunnel, v6dst, &dst);
 844         return dst;
 845 }
 846
 847 /*
 848  *      This function assumes it is being called from dev_queue_xmit()
 849  *      and that skb is filled properly by that function.
 850  */
 851
 852 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 853                                      struct net_device *dev)
 854 {
 855         struct ip_tunnel *tunnel = netdev_priv(dev);
 856         const struct iphdr  *tiph = &tunnel->parms.iph;
 857         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 858         u8     tos = tunnel->parms.iph.tos;
 859         __be16 df = tiph->frag_off;
 860         struct rtable *rt;              /* Route to the other host */
 861         struct net_device *tdev;        /* Device to other host */
 862         unsigned int max_headroom;      /* The extra header space needed */
 863         __be32 dst = tiph->daddr;
 864         struct flowi4 fl4;
 865         int    mtu;
 866         const struct in6_addr *addr6;
 867         int addr_type;
 868         u8 ttl;
 869         u8 protocol = IPPROTO_IPV6;
 870         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 871
 872         if (tos == 1)
 873                 tos = ipv6_get_dsfield(iph6);
 874
 875         /* ISATAP (RFC4214) - must come before 6to4 */
 876         if (dev->priv_flags & IFF_ISATAP) {
 877                 struct neighbour *neigh = NULL;
 878                 bool do_tx_error = false;
 879
 880                 if (skb_dst(skb))
 881                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 882
 883                 if (!neigh) {
 884                         net_dbg_ratelimited("nexthop == NULL\n");
 885                         goto tx_error;
 886                 }
 887
 888                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 889                 addr_type = ipv6_addr_type(addr6);
 890
 891                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 892                      ipv6_addr_is_isatap(addr6))
 893                         dst = addr6->s6_addr32[3];
 894                 else
 895                         do_tx_error = true;
 896
 897                 neigh_release(neigh);
 898                 if (do_tx_error)
 899                         goto tx_error;
 900         }
 901
 902         if (!dst)
 903                 dst = try_6rd(tunnel, &iph6->daddr);
 904
 905         if (!dst) {
 906                 struct neighbour *neigh = NULL;
 907                 bool do_tx_error = false;
 908
 909                 if (skb_dst(skb))
 910                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 911
 912                 if (!neigh) {
 913                         net_dbg_ratelimited("nexthop == NULL\n");
 914                         goto tx_error;
 915                 }
 916
 917                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 918                 addr_type = ipv6_addr_type(addr6);
 919
 920                 if (addr_type == IPV6_ADDR_ANY) {
 921                         addr6 = &ipv6_hdr(skb)->daddr;
 922                         addr_type = ipv6_addr_type(addr6);
 923                 }
 924
 925                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 926                         dst = addr6->s6_addr32[3];
 927                 else
 928                         do_tx_error = true;
 929
 930                 neigh_release(neigh);
 931                 if (do_tx_error)
 932                         goto tx_error;
 933         }
 934
 935         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 936                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 937                            0, dst, tiph->saddr, 0, 0,
 938                            sock_net_uid(tunnel->net, NULL));
 939
 940         rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 941         if (!rt) {
 942                 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 943                 if (IS_ERR(rt)) {
 944                         dev->stats.tx_carrier_errors++;
 945                         goto tx_error_icmp;
 946                 }
 947                 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 948         }
 949
 950         if (rt->rt_type != RTN_UNICAST) {
 951                 ip_rt_put(rt);
 952                 dev->stats.tx_carrier_errors++;
 953                 goto tx_error_icmp;
 954         }
 955         tdev = rt->dst.dev;
 956
 957         if (tdev == dev) {
 958                 ip_rt_put(rt);
 959                 dev->stats.collisions++;
 960                 goto tx_error;
 961         }
 962
 963         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 964                 ip_rt_put(rt);
 965                 goto tx_error;
 966         }
 967
 968         if (df) {
 969                 mtu = dst_mtu(&rt->dst) - t_hlen;
 970
 971                 if (mtu < 68) {
 972                         dev->stats.collisions++;
 973                         ip_rt_put(rt);
 974                         goto tx_error;
 975                 }
 976
 977                 if (mtu < IPV6_MIN_MTU) {
 978                         mtu = IPV6_MIN_MTU;
 979                         df = 0;
 980                 }
 981
 982                 if (tunnel->parms.iph.daddr)
 983                         skb_dst_update_pmtu_no_confirm(skb, mtu);
 984
 985                 if (skb->len > mtu && !skb_is_gso(skb)) {
 986                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 987                         ip_rt_put(rt);
 988                         goto tx_error;
 989                 }
 990         }
 991
 992         if (tunnel->err_count > 0) {
 993                 if (time_before(jiffies,
 994                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 995                         tunnel->err_count--;
 996                         dst_link_failure(skb);
 997                 } else
 998                         tunnel->err_count = 0;
 999         }
1000
1001         /*
1002          * Okay, now see if we can stuff it in the buffer as-is.
1003          */
1004         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1005
1006         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1007             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1008                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1009                 if (!new_skb) {
1010                         ip_rt_put(rt);
1011                         dev->stats.tx_dropped++;
1012                         kfree_skb(skb);
1013                         return NETDEV_TX_OK;
1014                 }
1015                 if (skb->sk)
1016                         skb_set_owner_w(new_skb, skb->sk);
1017                 dev_kfree_skb(skb);
1018                 skb = new_skb;
1019                 iph6 = ipv6_hdr(skb);
1020         }
1021         ttl = tiph->ttl;
1022         if (ttl == 0)
1023                 ttl = iph6->hop_limit;
1024         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1025
1026         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1027                 ip_rt_put(rt);
1028                 goto tx_error;
1029         }
1030
1031         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1032
1033         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1034                       df, !net_eq(tunnel->net, dev_net(dev)));
1035         return NETDEV_TX_OK;
1036
1037 tx_error_icmp:
1038         dst_link_failure(skb);
1039 tx_error:
1040         kfree_skb(skb);
1041         dev->stats.tx_errors++;
1042         return NETDEV_TX_OK;
1043 }
1044
1045 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1046                                      struct net_device *dev, u8 ipproto)
1047 {
1048         struct ip_tunnel *tunnel = netdev_priv(dev);
1049         const struct iphdr  *tiph = &tunnel->parms.iph;
1050
1051         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1052                 goto tx_error;
1053
1054         skb_set_inner_ipproto(skb, ipproto);
1055
1056         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1057         return NETDEV_TX_OK;
1058 tx_error:
1059         kfree_skb(skb);
1060         dev->stats.tx_errors++;
1061         return NETDEV_TX_OK;
1062 }
1063
1064 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1065                                    struct net_device *dev)
1066 {
1067         if (!pskb_inet_may_pull(skb))
1068                 goto tx_err;
1069
1070         switch (skb->protocol) {
1071         case htons(ETH_P_IP):
1072                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1073                 break;
1074         case htons(ETH_P_IPV6):
1075                 ipip6_tunnel_xmit(skb, dev);
1076                 break;
1077 #if IS_ENABLED(CONFIG_MPLS)
1078         case htons(ETH_P_MPLS_UC):
1079                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1080                 break;
1081 #endif
1082         default:
1083                 goto tx_err;
1084         }
1085
1086         return NETDEV_TX_OK;
1087
1088 tx_err:
1089         dev->stats.tx_errors++;
1090         kfree_skb(skb);
1091         return NETDEV_TX_OK;
1092
1093 }
1094
1095 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1096 {
1097         struct ip_tunnel *tunnel = netdev_priv(dev);
1098         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1099         struct net_device *tdev = NULL;
1100         int hlen = LL_MAX_HEADER;
1101         const struct iphdr *iph;
1102         struct flowi4 fl4;
1103
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 mtu;
1127
1128                 mtu = tdev->mtu - t_hlen;
1129                 if (mtu < IPV6_MIN_MTU)
1130                         mtu = IPV6_MIN_MTU;
1131                 WRITE_ONCE(dev->mtu, mtu);
1132                 hlen = tdev->hard_header_len + tdev->needed_headroom;
1133         }
1134         dev->needed_headroom = t_hlen + hlen;
1135 }
1136
1137 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1138                                 __u32 fwmark)
1139 {
1140         struct net *net = t->net;
1141         struct sit_net *sitn = net_generic(net, sit_net_id);
1142
1143         ipip6_tunnel_unlink(sitn, t);
1144         synchronize_net();
1145         t->parms.iph.saddr = p->iph.saddr;
1146         t->parms.iph.daddr = p->iph.daddr;
1147         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1148         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1149         ipip6_tunnel_link(sitn, t);
1150         t->parms.iph.ttl = p->iph.ttl;
1151         t->parms.iph.tos = p->iph.tos;
1152         t->parms.iph.frag_off = p->iph.frag_off;
1153         if (t->parms.link != p->link || t->fwmark != fwmark) {
1154                 t->parms.link = p->link;
1155                 t->fwmark = fwmark;
1156                 ipip6_tunnel_bind_dev(t->dev);
1157         }
1158         dst_cache_reset(&t->dst_cache);
1159         netdev_state_change(t->dev);
1160 }
1161
1162 #ifdef CONFIG_IPV6_SIT_6RD
1163 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1164                                    struct ip_tunnel_6rd *ip6rd)
1165 {
1166         struct in6_addr prefix;
1167         __be32 relay_prefix;
1168
1169         if (ip6rd->relay_prefixlen > 32 ||
1170             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1171                 return -EINVAL;
1172
1173         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1174         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1175                 return -EINVAL;
1176         if (ip6rd->relay_prefixlen)
1177                 relay_prefix = ip6rd->relay_prefix &
1178                                htonl(0xffffffffUL <<
1179                                      (32 - ip6rd->relay_prefixlen));
1180         else
1181                 relay_prefix = 0;
1182         if (relay_prefix != ip6rd->relay_prefix)
1183                 return -EINVAL;
1184
1185         t->ip6rd.prefix = prefix;
1186         t->ip6rd.relay_prefix = relay_prefix;
1187         t->ip6rd.prefixlen = ip6rd->prefixlen;
1188         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1189         dst_cache_reset(&t->dst_cache);
1190         netdev_state_change(t->dev);
1191         return 0;
1192 }
1193
1194 static int
1195 ipip6_tunnel_get6rd(struct net_device *dev, struct ifreq *ifr)
1196 {
1197         struct ip_tunnel *t = netdev_priv(dev);
1198         struct ip_tunnel_6rd ip6rd;
1199         struct ip_tunnel_parm p;
1200
1201         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1202                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1203                         return -EFAULT;
1204                 t = ipip6_tunnel_locate(t->net, &p, 0);
1205         }
1206         if (!t)
1207                 t = netdev_priv(dev);
1208
1209         ip6rd.prefix = t->ip6rd.prefix;
1210         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1211         ip6rd.prefixlen = t->ip6rd.prefixlen;
1212         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1213         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, sizeof(ip6rd)))
1214                 return -EFAULT;
1215         return 0;
1216 }
1217
1218 static int
1219 ipip6_tunnel_6rdctl(struct net_device *dev, struct ifreq *ifr, 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, ifr->ifr_ifru.ifru_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_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1367 {
1368         switch (cmd) {
1369         case SIOCGETTUNNEL:
1370         case SIOCADDTUNNEL:
1371         case SIOCCHGTUNNEL:
1372         case SIOCDELTUNNEL:
1373                 return ip_tunnel_ioctl(dev, ifr, cmd);
1374         case SIOCGETPRL:
1375                 return ipip6_tunnel_get_prl(dev, ifr);
1376         case SIOCADDPRL:
1377         case SIOCDELPRL:
1378         case SIOCCHGPRL:
1379                 return ipip6_tunnel_prl_ctl(dev, ifr, cmd);
1380 #ifdef CONFIG_IPV6_SIT_6RD
1381         case SIOCGET6RD:
1382                 return ipip6_tunnel_get6rd(dev, ifr);
1383         case SIOCADD6RD:
1384         case SIOCCHG6RD:
1385         case SIOCDEL6RD:
1386                 return ipip6_tunnel_6rdctl(dev, ifr, cmd);
1387 #endif
1388         default:
1389                 return -EINVAL;
1390         }
1391 }
1392
1393 static const struct net_device_ops ipip6_netdev_ops = {
1394         .ndo_init       = ipip6_tunnel_init,
1395         .ndo_uninit     = ipip6_tunnel_uninit,
1396         .ndo_start_xmit = sit_tunnel_xmit,
1397         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1398         .ndo_get_stats64 = ip_tunnel_get_stats64,
1399         .ndo_get_iflink = ip_tunnel_get_iflink,
1400         .ndo_tunnel_ctl = ipip6_tunnel_ctl,
1401 };
1402
1403 static void ipip6_dev_free(struct net_device *dev)
1404 {
1405         struct ip_tunnel *tunnel = netdev_priv(dev);
1406
1407         dst_cache_destroy(&tunnel->dst_cache);
1408         free_percpu(dev->tstats);
1409 }
1410
1411 #define SIT_FEATURES (NETIF_F_SG           | \
1412                       NETIF_F_FRAGLIST     | \
1413                       NETIF_F_HIGHDMA      | \
1414                       NETIF_F_GSO_SOFTWARE | \
1415                       NETIF_F_HW_CSUM)
1416
1417 static void ipip6_tunnel_setup(struct net_device *dev)
1418 {
1419         struct ip_tunnel *tunnel = netdev_priv(dev);
1420         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1421
1422         dev->netdev_ops         = &ipip6_netdev_ops;
1423         dev->header_ops         = &ip_tunnel_header_ops;
1424         dev->needs_free_netdev  = true;
1425         dev->priv_destructor    = ipip6_dev_free;
1426
1427         dev->type               = ARPHRD_SIT;
1428         dev->mtu                = ETH_DATA_LEN - t_hlen;
1429         dev->min_mtu            = IPV6_MIN_MTU;
1430         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1431         dev->flags              = IFF_NOARP;
1432         netif_keep_dst(dev);
1433         dev->addr_len           = 4;
1434         dev->features           |= NETIF_F_LLTX;
1435         dev->features           |= SIT_FEATURES;
1436         dev->hw_features        |= SIT_FEATURES;
1437 }
1438
1439 static int ipip6_tunnel_init(struct net_device *dev)
1440 {
1441         struct ip_tunnel *tunnel = netdev_priv(dev);
1442         int err;
1443
1444         tunnel->dev = dev;
1445         tunnel->net = dev_net(dev);
1446         strcpy(tunnel->parms.name, dev->name);
1447
1448         ipip6_tunnel_bind_dev(dev);
1449         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1450         if (!dev->tstats)
1451                 return -ENOMEM;
1452
1453         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1454         if (err) {
1455                 free_percpu(dev->tstats);
1456                 dev->tstats = NULL;
1457                 return err;
1458         }
1459         dev_hold(dev);
1460         return 0;
1461 }
1462
1463 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1464 {
1465         struct ip_tunnel *tunnel = netdev_priv(dev);
1466         struct iphdr *iph = &tunnel->parms.iph;
1467         struct net *net = dev_net(dev);
1468         struct sit_net *sitn = net_generic(net, sit_net_id);
1469
1470         iph->version            = 4;
1471         iph->protocol           = IPPROTO_IPV6;
1472         iph->ihl                = 5;
1473         iph->ttl                = 64;
1474
1475         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1476 }
1477
1478 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1479                           struct netlink_ext_ack *extack)
1480 {
1481         u8 proto;
1482
1483         if (!data || !data[IFLA_IPTUN_PROTO])
1484                 return 0;
1485
1486         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1487         if (!ipip6_valid_ip_proto(proto))
1488                 return -EINVAL;
1489
1490         return 0;
1491 }
1492
1493 static void ipip6_netlink_parms(struct nlattr *data[],
1494                                 struct ip_tunnel_parm *parms,
1495                                 __u32 *fwmark)
1496 {
1497         memset(parms, 0, sizeof(*parms));
1498
1499         parms->iph.version = 4;
1500         parms->iph.protocol = IPPROTO_IPV6;
1501         parms->iph.ihl = 5;
1502         parms->iph.ttl = 64;
1503
1504         if (!data)
1505                 return;
1506
1507         if (data[IFLA_IPTUN_LINK])
1508                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1509
1510         if (data[IFLA_IPTUN_LOCAL])
1511                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1512
1513         if (data[IFLA_IPTUN_REMOTE])
1514                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1515
1516         if (data[IFLA_IPTUN_TTL]) {
1517                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1518                 if (parms->iph.ttl)
1519                         parms->iph.frag_off = htons(IP_DF);
1520         }
1521
1522         if (data[IFLA_IPTUN_TOS])
1523                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1524
1525         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1526                 parms->iph.frag_off = htons(IP_DF);
1527
1528         if (data[IFLA_IPTUN_FLAGS])
1529                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1530
1531         if (data[IFLA_IPTUN_PROTO])
1532                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1533
1534         if (data[IFLA_IPTUN_FWMARK])
1535                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1536 }
1537
1538 /* This function returns true when ENCAP attributes are present in the nl msg */
1539 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1540                                       struct ip_tunnel_encap *ipencap)
1541 {
1542         bool ret = false;
1543
1544         memset(ipencap, 0, sizeof(*ipencap));
1545
1546         if (!data)
1547                 return ret;
1548
1549         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1550                 ret = true;
1551                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1552         }
1553
1554         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1555                 ret = true;
1556                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1557         }
1558
1559         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1560                 ret = true;
1561                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1562         }
1563
1564         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1565                 ret = true;
1566                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1567         }
1568
1569         return ret;
1570 }
1571
1572 #ifdef CONFIG_IPV6_SIT_6RD
1573 /* This function returns true when 6RD attributes are present in the nl msg */
1574 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1575                                     struct ip_tunnel_6rd *ip6rd)
1576 {
1577         bool ret = false;
1578         memset(ip6rd, 0, sizeof(*ip6rd));
1579
1580         if (!data)
1581                 return ret;
1582
1583         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1584                 ret = true;
1585                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1586         }
1587
1588         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1589                 ret = true;
1590                 ip6rd->relay_prefix =
1591                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1592         }
1593
1594         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1595                 ret = true;
1596                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1597         }
1598
1599         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1600                 ret = true;
1601                 ip6rd->relay_prefixlen =
1602                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1603         }
1604
1605         return ret;
1606 }
1607 #endif
1608
1609 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1610                          struct nlattr *tb[], struct nlattr *data[],
1611                          struct netlink_ext_ack *extack)
1612 {
1613         struct net *net = dev_net(dev);
1614         struct ip_tunnel *nt;
1615         struct ip_tunnel_encap ipencap;
1616 #ifdef CONFIG_IPV6_SIT_6RD
1617         struct ip_tunnel_6rd ip6rd;
1618 #endif
1619         int err;
1620
1621         nt = netdev_priv(dev);
1622
1623         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1624                 err = ip_tunnel_encap_setup(nt, &ipencap);
1625                 if (err < 0)
1626                         return err;
1627         }
1628
1629         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1630
1631         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1632                 return -EEXIST;
1633
1634         err = ipip6_tunnel_create(dev);
1635         if (err < 0)
1636                 return err;
1637
1638         if (tb[IFLA_MTU]) {
1639                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1640
1641                 if (mtu >= IPV6_MIN_MTU &&
1642                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1643                         dev->mtu = mtu;
1644         }
1645
1646 #ifdef CONFIG_IPV6_SIT_6RD
1647         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1648                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1649                 if (err < 0)
1650                         unregister_netdevice_queue(dev, NULL);
1651         }
1652 #endif
1653
1654         return err;
1655 }
1656
1657 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1658                             struct nlattr *data[],
1659                             struct netlink_ext_ack *extack)
1660 {
1661         struct ip_tunnel *t = netdev_priv(dev);
1662         struct ip_tunnel_parm p;
1663         struct ip_tunnel_encap ipencap;
1664         struct net *net = t->net;
1665         struct sit_net *sitn = net_generic(net, sit_net_id);
1666 #ifdef CONFIG_IPV6_SIT_6RD
1667         struct ip_tunnel_6rd ip6rd;
1668 #endif
1669         __u32 fwmark = t->fwmark;
1670         int err;
1671
1672         if (dev == sitn->fb_tunnel_dev)
1673                 return -EINVAL;
1674
1675         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1676                 err = ip_tunnel_encap_setup(t, &ipencap);
1677                 if (err < 0)
1678                         return err;
1679         }
1680
1681         ipip6_netlink_parms(data, &p, &fwmark);
1682
1683         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1684             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1685                 return -EINVAL;
1686
1687         t = ipip6_tunnel_locate(net, &p, 0);
1688
1689         if (t) {
1690                 if (t->dev != dev)
1691                         return -EEXIST;
1692         } else
1693                 t = netdev_priv(dev);
1694
1695         ipip6_tunnel_update(t, &p, fwmark);
1696
1697 #ifdef CONFIG_IPV6_SIT_6RD
1698         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1699                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1700 #endif
1701
1702         return 0;
1703 }
1704
1705 static size_t ipip6_get_size(const struct net_device *dev)
1706 {
1707         return
1708                 /* IFLA_IPTUN_LINK */
1709                 nla_total_size(4) +
1710                 /* IFLA_IPTUN_LOCAL */
1711                 nla_total_size(4) +
1712                 /* IFLA_IPTUN_REMOTE */
1713                 nla_total_size(4) +
1714                 /* IFLA_IPTUN_TTL */
1715                 nla_total_size(1) +
1716                 /* IFLA_IPTUN_TOS */
1717                 nla_total_size(1) +
1718                 /* IFLA_IPTUN_PMTUDISC */
1719                 nla_total_size(1) +
1720                 /* IFLA_IPTUN_FLAGS */
1721                 nla_total_size(2) +
1722                 /* IFLA_IPTUN_PROTO */
1723                 nla_total_size(1) +
1724 #ifdef CONFIG_IPV6_SIT_6RD
1725                 /* IFLA_IPTUN_6RD_PREFIX */
1726                 nla_total_size(sizeof(struct in6_addr)) +
1727                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1728                 nla_total_size(4) +
1729                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1730                 nla_total_size(2) +
1731                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1732                 nla_total_size(2) +
1733 #endif
1734                 /* IFLA_IPTUN_ENCAP_TYPE */
1735                 nla_total_size(2) +
1736                 /* IFLA_IPTUN_ENCAP_FLAGS */
1737                 nla_total_size(2) +
1738                 /* IFLA_IPTUN_ENCAP_SPORT */
1739                 nla_total_size(2) +
1740                 /* IFLA_IPTUN_ENCAP_DPORT */
1741                 nla_total_size(2) +
1742                 /* IFLA_IPTUN_FWMARK */
1743                 nla_total_size(4) +
1744                 0;
1745 }
1746
1747 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1748 {
1749         struct ip_tunnel *tunnel = netdev_priv(dev);
1750         struct ip_tunnel_parm *parm = &tunnel->parms;
1751
1752         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1753             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1754             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1755             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1756             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1757             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1758                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1759             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1760             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1761             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1762                 goto nla_put_failure;
1763
1764 #ifdef CONFIG_IPV6_SIT_6RD
1765         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1766                              &tunnel->ip6rd.prefix) ||
1767             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1768                             tunnel->ip6rd.relay_prefix) ||
1769             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1770                         tunnel->ip6rd.prefixlen) ||
1771             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1772                         tunnel->ip6rd.relay_prefixlen))
1773                 goto nla_put_failure;
1774 #endif
1775
1776         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1777                         tunnel->encap.type) ||
1778             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1779                         tunnel->encap.sport) ||
1780             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1781                         tunnel->encap.dport) ||
1782             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1783                         tunnel->encap.flags))
1784                 goto nla_put_failure;
1785
1786         return 0;
1787
1788 nla_put_failure:
1789         return -EMSGSIZE;
1790 }
1791
1792 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1793         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1794         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1795         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1796         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1797         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1798         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1799         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1800         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1801 #ifdef CONFIG_IPV6_SIT_6RD
1802         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1803         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1804         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1805         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1806 #endif
1807         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1808         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1809         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1810         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1811         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1812 };
1813
1814 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1815 {
1816         struct net *net = dev_net(dev);
1817         struct sit_net *sitn = net_generic(net, sit_net_id);
1818
1819         if (dev != sitn->fb_tunnel_dev)
1820                 unregister_netdevice_queue(dev, head);
1821 }
1822
1823 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1824         .kind           = "sit",
1825         .maxtype        = IFLA_IPTUN_MAX,
1826         .policy         = ipip6_policy,
1827         .priv_size      = sizeof(struct ip_tunnel),
1828         .setup          = ipip6_tunnel_setup,
1829         .validate       = ipip6_validate,
1830         .newlink        = ipip6_newlink,
1831         .changelink     = ipip6_changelink,
1832         .get_size       = ipip6_get_size,
1833         .fill_info      = ipip6_fill_info,
1834         .dellink        = ipip6_dellink,
1835         .get_link_net   = ip_tunnel_get_link_net,
1836 };
1837
1838 static struct xfrm_tunnel sit_handler __read_mostly = {
1839         .handler        =       ipip6_rcv,
1840         .err_handler    =       ipip6_err,
1841         .priority       =       1,
1842 };
1843
1844 static struct xfrm_tunnel ipip_handler __read_mostly = {
1845         .handler        =       ipip_rcv,
1846         .err_handler    =       ipip6_err,
1847         .priority       =       2,
1848 };
1849
1850 #if IS_ENABLED(CONFIG_MPLS)
1851 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1852         .handler        =       mplsip_rcv,
1853         .err_handler    =       ipip6_err,
1854         .priority       =       2,
1855 };
1856 #endif
1857
1858 static void __net_exit sit_destroy_tunnels(struct net *net,
1859                                            struct list_head *head)
1860 {
1861         struct sit_net *sitn = net_generic(net, sit_net_id);
1862         struct net_device *dev, *aux;
1863         int prio;
1864
1865         for_each_netdev_safe(net, dev, aux)
1866                 if (dev->rtnl_link_ops == &sit_link_ops)
1867                         unregister_netdevice_queue(dev, head);
1868
1869         for (prio = 0; prio < 4; prio++) {
1870                 int h;
1871                 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1872                         struct ip_tunnel *t;
1873
1874                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1875                         while (t) {
1876                                 /* If dev is in the same netns, it has already
1877                                  * been added to the list by the previous loop.
1878                                  */
1879                                 if (!net_eq(dev_net(t->dev), net))
1880                                         unregister_netdevice_queue(t->dev,
1881                                                                    head);
1882                                 t = rtnl_dereference(t->next);
1883                         }
1884                 }
1885         }
1886 }
1887
1888 static int __net_init sit_init_net(struct net *net)
1889 {
1890         struct sit_net *sitn = net_generic(net, sit_net_id);
1891         struct ip_tunnel *t;
1892         int err;
1893
1894         sitn->tunnels[0] = sitn->tunnels_wc;
1895         sitn->tunnels[1] = sitn->tunnels_l;
1896         sitn->tunnels[2] = sitn->tunnels_r;
1897         sitn->tunnels[3] = sitn->tunnels_r_l;
1898
1899         if (!net_has_fallback_tunnels(net))
1900                 return 0;
1901
1902         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1903                                            NET_NAME_UNKNOWN,
1904                                            ipip6_tunnel_setup);
1905         if (!sitn->fb_tunnel_dev) {
1906                 err = -ENOMEM;
1907                 goto err_alloc_dev;
1908         }
1909         dev_net_set(sitn->fb_tunnel_dev, net);
1910         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1911         /* FB netdevice is special: we have one, and only one per netns.
1912          * Allowing to move it to another netns is clearly unsafe.
1913          */
1914         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1915
1916         err = register_netdev(sitn->fb_tunnel_dev);
1917         if (err)
1918                 goto err_reg_dev;
1919
1920         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1921         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1922
1923         t = netdev_priv(sitn->fb_tunnel_dev);
1924
1925         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1926         return 0;
1927
1928 err_reg_dev:
1929         free_netdev(sitn->fb_tunnel_dev);
1930 err_alloc_dev:
1931         return err;
1932 }
1933
1934 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1935 {
1936         LIST_HEAD(list);
1937         struct net *net;
1938
1939         rtnl_lock();
1940         list_for_each_entry(net, net_list, exit_list)
1941                 sit_destroy_tunnels(net, &list);
1942
1943         unregister_netdevice_many(&list);
1944         rtnl_unlock();
1945 }
1946
1947 static struct pernet_operations sit_net_ops = {
1948         .init = sit_init_net,
1949         .exit_batch = sit_exit_batch_net,
1950         .id   = &sit_net_id,
1951         .size = sizeof(struct sit_net),
1952 };
1953
1954 static void __exit sit_cleanup(void)
1955 {
1956         rtnl_link_unregister(&sit_link_ops);
1957         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1958         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1959 #if IS_ENABLED(CONFIG_MPLS)
1960         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1961 #endif
1962
1963         unregister_pernet_device(&sit_net_ops);
1964         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1965 }
1966
1967 static int __init sit_init(void)
1968 {
1969         int err;
1970
1971         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1972
1973         err = register_pernet_device(&sit_net_ops);
1974         if (err < 0)
1975                 return err;
1976         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1977         if (err < 0) {
1978                 pr_info("%s: can't register ip6ip4\n", __func__);
1979                 goto xfrm_tunnel_failed;
1980         }
1981         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1982         if (err < 0) {
1983                 pr_info("%s: can't register ip4ip4\n", __func__);
1984                 goto xfrm_tunnel4_failed;
1985         }
1986 #if IS_ENABLED(CONFIG_MPLS)
1987         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1988         if (err < 0) {
1989                 pr_info("%s: can't register mplsip\n", __func__);
1990                 goto xfrm_tunnel_mpls_failed;
1991         }
1992 #endif
1993         err = rtnl_link_register(&sit_link_ops);
1994         if (err < 0)
1995                 goto rtnl_link_failed;
1996
1997 out:
1998         return err;
1999
2000 rtnl_link_failed:
2001 #if IS_ENABLED(CONFIG_MPLS)
2002         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2003 xfrm_tunnel_mpls_failed:
2004 #endif
2005         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2006 xfrm_tunnel4_failed:
2007         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2008 xfrm_tunnel_failed:
2009         unregister_pernet_device(&sit_net_ops);
2010         goto out;
2011 }
2012
2013 module_init(sit_init);
2014 module_exit(sit_cleanup);
2015 MODULE_LICENSE("GPL");
2016 MODULE_ALIAS_RTNL_LINK("sit");
2017 MODULE_ALIAS_NETDEV("sit0");