net/ipv6/sit.c

   1 /*
   2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  *
  14  *      Changes:
  15  * Roger Venning <r.venning@telstra.com>:       6to4 support
  16  * Nate Thompson <nate@thebog.net>:             6to4 support
  17  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  18  */
  19
  20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  21
  22 #include <linux/module.h>
  23 #include <linux/capability.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/socket.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/in6.h>
  30 #include <linux/netdevice.h>
  31 #include <linux/if_arp.h>
  32 #include <linux/icmp.h>
  33 #include <linux/slab.h>
  34 #include <linux/uaccess.h>
  35 #include <linux/init.h>
  36 #include <linux/netfilter_ipv4.h>
  37 #include <linux/if_ether.h>
  38
  39 #include <net/sock.h>
  40 #include <net/snmp.h>
  41
  42 #include <net/ipv6.h>
  43 #include <net/protocol.h>
  44 #include <net/transp_v6.h>
  45 #include <net/ip6_fib.h>
  46 #include <net/ip6_route.h>
  47 #include <net/ndisc.h>
  48 #include <net/addrconf.h>
  49 #include <net/ip.h>
  50 #include <net/udp.h>
  51 #include <net/icmp.h>
  52 #include <net/ip_tunnels.h>
  53 #include <net/inet_ecn.h>
  54 #include <net/xfrm.h>
  55 #include <net/dsfield.h>
  56 #include <net/net_namespace.h>
  57 #include <net/netns/generic.h>
  58
  59 /*
  60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  61
  62    For comments look at net/ipv4/ip_gre.c --ANK
  63  */
  64
  65 #define IP6_SIT_HASH_SIZE  16
  66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  67
  68 static bool log_ecn_error = true;
  69 module_param(log_ecn_error, bool, 0644);
  70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  71
  72 static int ipip6_tunnel_init(struct net_device *dev);
  73 static void ipip6_tunnel_setup(struct net_device *dev);
  74 static void ipip6_dev_free(struct net_device *dev);
  75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  76                       __be32 *v4dst);
  77 static struct rtnl_link_ops sit_link_ops __read_mostly;
  78
  79 static unsigned int sit_net_id __read_mostly;
  80 struct sit_net {
  81         struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  82         struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  83         struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
  84         struct ip_tunnel __rcu *tunnels_wc[1];
  85         struct ip_tunnel __rcu **tunnels[4];
  86
  87         struct net_device *fb_tunnel_dev;
  88 };
  89
  90 /*
  91  * Must be invoked with rcu_read_lock
  92  */
  93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  94                                              struct net_device *dev,
  95                                              __be32 remote, __be32 local,
  96                                              int sifindex)
  97 {
  98         unsigned int h0 = HASH(remote);
  99         unsigned int h1 = HASH(local);
 100         struct ip_tunnel *t;
 101         struct sit_net *sitn = net_generic(net, sit_net_id);
 102         int ifindex = dev ? dev->ifindex : 0;
 103
 104         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 105                 if (local == t->parms.iph.saddr &&
 106                     remote == t->parms.iph.daddr &&
 107                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 108                      sifindex == t->parms.link) &&
 109                     (t->dev->flags & IFF_UP))
 110                         return t;
 111         }
 112         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 113                 if (remote == t->parms.iph.daddr &&
 114                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 115                      sifindex == t->parms.link) &&
 116                     (t->dev->flags & IFF_UP))
 117                         return t;
 118         }
 119         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 120                 if (local == t->parms.iph.saddr &&
 121                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 122                      sifindex == t->parms.link) &&
 123                     (t->dev->flags & IFF_UP))
 124                         return t;
 125         }
 126         t = rcu_dereference(sitn->tunnels_wc[0]);
 127         if (t && (t->dev->flags & IFF_UP))
 128                 return t;
 129         return NULL;
 130 }
 131
 132 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 133                 struct ip_tunnel_parm *parms)
 134 {
 135         __be32 remote = parms->iph.daddr;
 136         __be32 local = parms->iph.saddr;
 137         unsigned int h = 0;
 138         int prio = 0;
 139
 140         if (remote) {
 141                 prio |= 2;
 142                 h ^= HASH(remote);
 143         }
 144         if (local) {
 145                 prio |= 1;
 146                 h ^= HASH(local);
 147         }
 148         return &sitn->tunnels[prio][h];
 149 }
 150
 151 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 152                 struct ip_tunnel *t)
 153 {
 154         return __ipip6_bucket(sitn, &t->parms);
 155 }
 156
 157 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 158 {
 159         struct ip_tunnel __rcu **tp;
 160         struct ip_tunnel *iter;
 161
 162         for (tp = ipip6_bucket(sitn, t);
 163              (iter = rtnl_dereference(*tp)) != NULL;
 164              tp = &iter->next) {
 165                 if (t == iter) {
 166                         rcu_assign_pointer(*tp, t->next);
 167                         break;
 168                 }
 169         }
 170 }
 171
 172 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 173 {
 174         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 175
 176         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 177         rcu_assign_pointer(*tp, t);
 178 }
 179
 180 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 181 {
 182 #ifdef CONFIG_IPV6_SIT_6RD
 183         struct ip_tunnel *t = netdev_priv(dev);
 184
 185         if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 186                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 187                 t->ip6rd.relay_prefix = 0;
 188                 t->ip6rd.prefixlen = 16;
 189                 t->ip6rd.relay_prefixlen = 0;
 190         } else {
 191                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 192                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 193         }
 194 #endif
 195 }
 196
 197 static int ipip6_tunnel_create(struct net_device *dev)
 198 {
 199         struct ip_tunnel *t = netdev_priv(dev);
 200         struct net *net = dev_net(dev);
 201         struct sit_net *sitn = net_generic(net, sit_net_id);
 202         int err;
 203
 204         memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 205         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 206
 207         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 208                 dev->priv_flags |= IFF_ISATAP;
 209
 210         dev->rtnl_link_ops = &sit_link_ops;
 211
 212         err = register_netdevice(dev);
 213         if (err < 0)
 214                 goto out;
 215
 216         ipip6_tunnel_clone_6rd(dev, sitn);
 217
 218         ipip6_tunnel_link(sitn, t);
 219         return 0;
 220
 221 out:
 222         return err;
 223 }
 224
 225 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 226                 struct ip_tunnel_parm *parms, int create)
 227 {
 228         __be32 remote = parms->iph.daddr;
 229         __be32 local = parms->iph.saddr;
 230         struct ip_tunnel *t, *nt;
 231         struct ip_tunnel __rcu **tp;
 232         struct net_device *dev;
 233         char name[IFNAMSIZ];
 234         struct sit_net *sitn = net_generic(net, sit_net_id);
 235
 236         for (tp = __ipip6_bucket(sitn, parms);
 237             (t = rtnl_dereference(*tp)) != NULL;
 238              tp = &t->next) {
 239                 if (local == t->parms.iph.saddr &&
 240                     remote == t->parms.iph.daddr &&
 241                     parms->link == t->parms.link) {
 242                         if (create)
 243                                 return NULL;
 244                         else
 245                                 return t;
 246                 }
 247         }
 248         if (!create)
 249                 goto failed;
 250
 251         if (parms->name[0]) {
 252                 if (!dev_valid_name(parms->name))
 253                         goto failed;
 254                 strlcpy(name, parms->name, IFNAMSIZ);
 255         } else {
 256                 strcpy(name, "sit%d");
 257         }
 258         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 259                            ipip6_tunnel_setup);
 260         if (!dev)
 261                 return NULL;
 262
 263         dev_net_set(dev, net);
 264
 265         nt = netdev_priv(dev);
 266
 267         nt->parms = *parms;
 268         if (ipip6_tunnel_create(dev) < 0)
 269                 goto failed_free;
 270
 271         return nt;
 272
 273 failed_free:
 274         free_netdev(dev);
 275 failed:
 276         return NULL;
 277 }
 278
 279 #define for_each_prl_rcu(start)                 \
 280         for (prl = rcu_dereference(start);      \
 281              prl;                               \
 282              prl = rcu_dereference(prl->next))
 283
 284 static struct ip_tunnel_prl_entry *
 285 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 286 {
 287         struct ip_tunnel_prl_entry *prl;
 288
 289         for_each_prl_rcu(t->prl)
 290                 if (prl->addr == addr)
 291                         break;
 292         return prl;
 293
 294 }
 295
 296 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 297                                 struct ip_tunnel_prl __user *a)
 298 {
 299         struct ip_tunnel_prl kprl, *kp;
 300         struct ip_tunnel_prl_entry *prl;
 301         unsigned int cmax, c = 0, ca, len;
 302         int ret = 0;
 303
 304         if (copy_from_user(&kprl, a, sizeof(kprl)))
 305                 return -EFAULT;
 306         cmax = kprl.datalen / sizeof(kprl);
 307         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 308                 cmax = 1;
 309
 310         /* For simple GET or for root users,
 311          * we try harder to allocate.
 312          */
 313         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 314                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 315                 NULL;
 316
 317         ca = min(t->prl_count, cmax);
 318
 319         if (!kp) {
 320                 /* We don't try hard to allocate much memory for
 321                  * non-root users.
 322                  * For root users, retry allocating enough memory for
 323                  * the answer.
 324                  */
 325                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 326                 if (!kp) {
 327                         ret = -ENOMEM;
 328                         goto out;
 329                 }
 330         }
 331
 332         rcu_read_lock();
 333         for_each_prl_rcu(t->prl) {
 334                 if (c >= cmax)
 335                         break;
 336                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 337                         continue;
 338                 kp[c].addr = prl->addr;
 339                 kp[c].flags = prl->flags;
 340                 c++;
 341                 if (kprl.addr != htonl(INADDR_ANY))
 342                         break;
 343         }
 344
 345         rcu_read_unlock();
 346
 347         len = sizeof(*kp) * c;
 348         ret = 0;
 349         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 350                 ret = -EFAULT;
 351
 352         kfree(kp);
 353 out:
 354         return ret;
 355 }
 356
 357 static int
 358 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 359 {
 360         struct ip_tunnel_prl_entry *p;
 361         int err = 0;
 362
 363         if (a->addr == htonl(INADDR_ANY))
 364                 return -EINVAL;
 365
 366         ASSERT_RTNL();
 367
 368         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 369                 if (p->addr == a->addr) {
 370                         if (chg) {
 371                                 p->flags = a->flags;
 372                                 goto out;
 373                         }
 374                         err = -EEXIST;
 375                         goto out;
 376                 }
 377         }
 378
 379         if (chg) {
 380                 err = -ENXIO;
 381                 goto out;
 382         }
 383
 384         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 385         if (!p) {
 386                 err = -ENOBUFS;
 387                 goto out;
 388         }
 389
 390         p->next = t->prl;
 391         p->addr = a->addr;
 392         p->flags = a->flags;
 393         t->prl_count++;
 394         rcu_assign_pointer(t->prl, p);
 395 out:
 396         return err;
 397 }
 398
 399 static void prl_list_destroy_rcu(struct rcu_head *head)
 400 {
 401         struct ip_tunnel_prl_entry *p, *n;
 402
 403         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 404         do {
 405                 n = rcu_dereference_protected(p->next, 1);
 406                 kfree(p);
 407                 p = n;
 408         } while (p);
 409 }
 410
 411 static int
 412 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 413 {
 414         struct ip_tunnel_prl_entry *x;
 415         struct ip_tunnel_prl_entry __rcu **p;
 416         int err = 0;
 417
 418         ASSERT_RTNL();
 419
 420         if (a && a->addr != htonl(INADDR_ANY)) {
 421                 for (p = &t->prl;
 422                      (x = rtnl_dereference(*p)) != NULL;
 423                      p = &x->next) {
 424                         if (x->addr == a->addr) {
 425                                 *p = x->next;
 426                                 kfree_rcu(x, rcu_head);
 427                                 t->prl_count--;
 428                                 goto out;
 429                         }
 430                 }
 431                 err = -ENXIO;
 432         } else {
 433                 x = rtnl_dereference(t->prl);
 434                 if (x) {
 435                         t->prl_count = 0;
 436                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 437                         t->prl = NULL;
 438                 }
 439         }
 440 out:
 441         return err;
 442 }
 443
 444 static int
 445 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 446 {
 447         struct ip_tunnel_prl_entry *p;
 448         int ok = 1;
 449
 450         rcu_read_lock();
 451         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 452         if (p) {
 453                 if (p->flags & PRL_DEFAULT)
 454                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 455                 else
 456                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 457         } else {
 458                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 459
 460                 if (ipv6_addr_is_isatap(addr6) &&
 461                     (addr6->s6_addr32[3] == iph->saddr) &&
 462                     ipv6_chk_prefix(addr6, t->dev))
 463                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 464                 else
 465                         ok = 0;
 466         }
 467         rcu_read_unlock();
 468         return ok;
 469 }
 470
 471 static void ipip6_tunnel_uninit(struct net_device *dev)
 472 {
 473         struct ip_tunnel *tunnel = netdev_priv(dev);
 474         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 475
 476         if (dev == sitn->fb_tunnel_dev) {
 477                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 478         } else {
 479                 ipip6_tunnel_unlink(sitn, tunnel);
 480                 ipip6_tunnel_del_prl(tunnel, NULL);
 481         }
 482         dst_cache_reset(&tunnel->dst_cache);
 483         dev_put(dev);
 484 }
 485
 486 static int ipip6_err(struct sk_buff *skb, u32 info)
 487 {
 488         const struct iphdr *iph = (const struct iphdr *)skb->data;
 489         const int type = icmp_hdr(skb)->type;
 490         const int code = icmp_hdr(skb)->code;
 491         unsigned int data_len = 0;
 492         struct ip_tunnel *t;
 493         int sifindex;
 494         int err;
 495
 496         switch (type) {
 497         default:
 498         case ICMP_PARAMETERPROB:
 499                 return 0;
 500
 501         case ICMP_DEST_UNREACH:
 502                 switch (code) {
 503                 case ICMP_SR_FAILED:
 504                         /* Impossible event. */
 505                         return 0;
 506                 default:
 507                         /* All others are translated to HOST_UNREACH.
 508                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 509                            I believe they are just ether pollution. --ANK
 510                          */
 511                         break;
 512                 }
 513                 break;
 514         case ICMP_TIME_EXCEEDED:
 515                 if (code != ICMP_EXC_TTL)
 516                         return 0;
 517                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 518                 break;
 519         case ICMP_REDIRECT:
 520                 break;
 521         }
 522
 523         err = -ENOENT;
 524
 525         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 526         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 527                                 iph->daddr, iph->saddr, sifindex);
 528         if (!t)
 529                 goto out;
 530
 531         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 532                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 533                                  t->parms.link, 0, iph->protocol, 0);
 534                 err = 0;
 535                 goto out;
 536         }
 537         if (type == ICMP_REDIRECT) {
 538                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0,
 539                               iph->protocol, 0);
 540                 err = 0;
 541                 goto out;
 542         }
 543
 544         err = 0;
 545         if (__in6_dev_get(skb->dev) &&
 546             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 547                 goto out;
 548
 549         if (t->parms.iph.daddr == 0)
 550                 goto out;
 551
 552         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 553                 goto out;
 554
 555         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 556                 t->err_count++;
 557         else
 558                 t->err_count = 1;
 559         t->err_time = jiffies;
 560 out:
 561         return err;
 562 }
 563
 564 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 565                                   const struct in6_addr *v6addr)
 566 {
 567         __be32 v4embed = 0;
 568         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 569                 return true;
 570         return false;
 571 }
 572
 573 /* Checks if an address matches an address on the tunnel interface.
 574  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 575  * Long story:
 576  * This function is called after we considered the packet as spoofed
 577  * in is_spoofed_6rd.
 578  * We may have a router that is doing NAT for proto 41 packets
 579  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 580  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 581  * function will return true, dropping the packet.
 582  * But, we can still check if is spoofed against the IP
 583  * addresses associated with the interface.
 584  */
 585 static bool only_dnatted(const struct ip_tunnel *tunnel,
 586         const struct in6_addr *v6dst)
 587 {
 588         int prefix_len;
 589
 590 #ifdef CONFIG_IPV6_SIT_6RD
 591         prefix_len = tunnel->ip6rd.prefixlen + 32
 592                 - tunnel->ip6rd.relay_prefixlen;
 593 #else
 594         prefix_len = 48;
 595 #endif
 596         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 597 }
 598
 599 /* Returns true if a packet is spoofed */
 600 static bool packet_is_spoofed(struct sk_buff *skb,
 601                               const struct iphdr *iph,
 602                               struct ip_tunnel *tunnel)
 603 {
 604         const struct ipv6hdr *ipv6h;
 605
 606         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 607                 if (!isatap_chksrc(skb, iph, tunnel))
 608                         return true;
 609
 610                 return false;
 611         }
 612
 613         if (tunnel->dev->flags & IFF_POINTOPOINT)
 614                 return false;
 615
 616         ipv6h = ipv6_hdr(skb);
 617
 618         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 619                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 620                                      &iph->saddr, &ipv6h->saddr,
 621                                      &iph->daddr, &ipv6h->daddr);
 622                 return true;
 623         }
 624
 625         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 626                 return false;
 627
 628         if (only_dnatted(tunnel, &ipv6h->daddr))
 629                 return false;
 630
 631         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 632                              &iph->saddr, &ipv6h->saddr,
 633                              &iph->daddr, &ipv6h->daddr);
 634         return true;
 635 }
 636
 637 static int ipip6_rcv(struct sk_buff *skb)
 638 {
 639         const struct iphdr *iph = ip_hdr(skb);
 640         struct ip_tunnel *tunnel;
 641         int sifindex;
 642         int err;
 643
 644         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 645         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 646                                      iph->saddr, iph->daddr, sifindex);
 647         if (tunnel) {
 648                 struct pcpu_sw_netstats *tstats;
 649
 650                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 651                     tunnel->parms.iph.protocol != 0)
 652                         goto out;
 653
 654                 skb->mac_header = skb->network_header;
 655                 skb_reset_network_header(skb);
 656                 IPCB(skb)->flags = 0;
 657                 skb->dev = tunnel->dev;
 658
 659                 if (packet_is_spoofed(skb, iph, tunnel)) {
 660                         tunnel->dev->stats.rx_errors++;
 661                         goto out;
 662                 }
 663
 664                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 665                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 666                         goto out;
 667
 668                 /* skb can be uncloned in iptunnel_pull_header, so
 669                  * old iph is no longer valid
 670                  */
 671                 iph = (const struct iphdr *)skb_mac_header(skb);
 672                 err = IP_ECN_decapsulate(iph, skb);
 673                 if (unlikely(err)) {
 674                         if (log_ecn_error)
 675                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 676                                                      &iph->saddr, iph->tos);
 677                         if (err > 1) {
 678                                 ++tunnel->dev->stats.rx_frame_errors;
 679                                 ++tunnel->dev->stats.rx_errors;
 680                                 goto out;
 681                         }
 682                 }
 683
 684                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 685                 u64_stats_update_begin(&tstats->syncp);
 686                 tstats->rx_packets++;
 687                 tstats->rx_bytes += skb->len;
 688                 u64_stats_update_end(&tstats->syncp);
 689
 690                 netif_rx(skb);
 691
 692                 return 0;
 693         }
 694
 695         /* no tunnel matched,  let upstream know, ipsec may handle it */
 696         return 1;
 697 out:
 698         kfree_skb(skb);
 699         return 0;
 700 }
 701
 702 static const struct tnl_ptk_info ipip_tpi = {
 703         /* no tunnel info required for ipip. */
 704         .proto = htons(ETH_P_IP),
 705 };
 706
 707 #if IS_ENABLED(CONFIG_MPLS)
 708 static const struct tnl_ptk_info mplsip_tpi = {
 709         /* no tunnel info required for mplsip. */
 710         .proto = htons(ETH_P_MPLS_UC),
 711 };
 712 #endif
 713
 714 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 715 {
 716         const struct iphdr *iph;
 717         struct ip_tunnel *tunnel;
 718         int sifindex;
 719
 720         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 721
 722         iph = ip_hdr(skb);
 723         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 724                                      iph->saddr, iph->daddr, sifindex);
 725         if (tunnel) {
 726                 const struct tnl_ptk_info *tpi;
 727
 728                 if (tunnel->parms.iph.protocol != ipproto &&
 729                     tunnel->parms.iph.protocol != 0)
 730                         goto drop;
 731
 732                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 733                         goto drop;
 734 #if IS_ENABLED(CONFIG_MPLS)
 735                 if (ipproto == IPPROTO_MPLS)
 736                         tpi = &mplsip_tpi;
 737                 else
 738 #endif
 739                         tpi = &ipip_tpi;
 740                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 741                         goto drop;
 742                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 743         }
 744
 745         return 1;
 746
 747 drop:
 748         kfree_skb(skb);
 749         return 0;
 750 }
 751
 752 static int ipip_rcv(struct sk_buff *skb)
 753 {
 754         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 755 }
 756
 757 #if IS_ENABLED(CONFIG_MPLS)
 758 static int mplsip_rcv(struct sk_buff *skb)
 759 {
 760         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 761 }
 762 #endif
 763
 764 /*
 765  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 766  * stores the embedded IPv4 address in v4dst and returns true.
 767  */
 768 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 769                       __be32 *v4dst)
 770 {
 771 #ifdef CONFIG_IPV6_SIT_6RD
 772         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 773                               tunnel->ip6rd.prefixlen)) {
 774                 unsigned int pbw0, pbi0;
 775                 int pbi1;
 776                 u32 d;
 777
 778                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 779                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 780
 781                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 782                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 783                     tunnel->ip6rd.relay_prefixlen : 0;
 784
 785                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 786                 if (pbi1 > 0)
 787                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 788                              (32 - pbi1);
 789
 790                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 791                 return true;
 792         }
 793 #else
 794         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 795                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 796                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 797                 return true;
 798         }
 799 #endif
 800         return false;
 801 }
 802
 803 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 804                              const struct in6_addr *v6dst)
 805 {
 806         __be32 dst = 0;
 807         check_6rd(tunnel, v6dst, &dst);
 808         return dst;
 809 }
 810
 811 /*
 812  *      This function assumes it is being called from dev_queue_xmit()
 813  *      and that skb is filled properly by that function.
 814  */
 815
 816 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 817                                      struct net_device *dev)
 818 {
 819         struct ip_tunnel *tunnel = netdev_priv(dev);
 820         const struct iphdr  *tiph = &tunnel->parms.iph;
 821         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 822         u8     tos = tunnel->parms.iph.tos;
 823         __be16 df = tiph->frag_off;
 824         struct rtable *rt;              /* Route to the other host */
 825         struct net_device *tdev;        /* Device to other host */
 826         unsigned int max_headroom;      /* The extra header space needed */
 827         __be32 dst = tiph->daddr;
 828         struct flowi4 fl4;
 829         int    mtu;
 830         const struct in6_addr *addr6;
 831         int addr_type;
 832         u8 ttl;
 833         u8 protocol = IPPROTO_IPV6;
 834         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 835
 836         if (tos == 1)
 837                 tos = ipv6_get_dsfield(iph6);
 838
 839         /* ISATAP (RFC4214) - must come before 6to4 */
 840         if (dev->priv_flags & IFF_ISATAP) {
 841                 struct neighbour *neigh = NULL;
 842                 bool do_tx_error = false;
 843
 844                 if (skb_dst(skb))
 845                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 846
 847                 if (!neigh) {
 848                         net_dbg_ratelimited("nexthop == NULL\n");
 849                         goto tx_error;
 850                 }
 851
 852                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 853                 addr_type = ipv6_addr_type(addr6);
 854
 855                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 856                      ipv6_addr_is_isatap(addr6))
 857                         dst = addr6->s6_addr32[3];
 858                 else
 859                         do_tx_error = true;
 860
 861                 neigh_release(neigh);
 862                 if (do_tx_error)
 863                         goto tx_error;
 864         }
 865
 866         if (!dst)
 867                 dst = try_6rd(tunnel, &iph6->daddr);
 868
 869         if (!dst) {
 870                 struct neighbour *neigh = NULL;
 871                 bool do_tx_error = false;
 872
 873                 if (skb_dst(skb))
 874                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 875
 876                 if (!neigh) {
 877                         net_dbg_ratelimited("nexthop == NULL\n");
 878                         goto tx_error;
 879                 }
 880
 881                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 882                 addr_type = ipv6_addr_type(addr6);
 883
 884                 if (addr_type == IPV6_ADDR_ANY) {
 885                         addr6 = &ipv6_hdr(skb)->daddr;
 886                         addr_type = ipv6_addr_type(addr6);
 887                 }
 888
 889                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 890                         dst = addr6->s6_addr32[3];
 891                 else
 892                         do_tx_error = true;
 893
 894                 neigh_release(neigh);
 895                 if (do_tx_error)
 896                         goto tx_error;
 897         }
 898
 899         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 900                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 901                            0, dst, tiph->saddr, 0, 0,
 902                            sock_net_uid(tunnel->net, NULL));
 903         rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 904
 905         if (IS_ERR(rt)) {
 906                 dev->stats.tx_carrier_errors++;
 907                 goto tx_error_icmp;
 908         }
 909         if (rt->rt_type != RTN_UNICAST) {
 910                 ip_rt_put(rt);
 911                 dev->stats.tx_carrier_errors++;
 912                 goto tx_error_icmp;
 913         }
 914         tdev = rt->dst.dev;
 915
 916         if (tdev == dev) {
 917                 ip_rt_put(rt);
 918                 dev->stats.collisions++;
 919                 goto tx_error;
 920         }
 921
 922         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 923                 ip_rt_put(rt);
 924                 goto tx_error;
 925         }
 926
 927         if (df) {
 928                 mtu = dst_mtu(&rt->dst) - t_hlen;
 929
 930                 if (mtu < 68) {
 931                         dev->stats.collisions++;
 932                         ip_rt_put(rt);
 933                         goto tx_error;
 934                 }
 935
 936                 if (mtu < IPV6_MIN_MTU) {
 937                         mtu = IPV6_MIN_MTU;
 938                         df = 0;
 939                 }
 940
 941                 if (tunnel->parms.iph.daddr)
 942                         skb_dst_update_pmtu_no_confirm(skb, mtu);
 943
 944                 if (skb->len > mtu && !skb_is_gso(skb)) {
 945                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 946                         ip_rt_put(rt);
 947                         goto tx_error;
 948                 }
 949         }
 950
 951         if (tunnel->err_count > 0) {
 952                 if (time_before(jiffies,
 953                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 954                         tunnel->err_count--;
 955                         dst_link_failure(skb);
 956                 } else
 957                         tunnel->err_count = 0;
 958         }
 959
 960         /*
 961          * Okay, now see if we can stuff it in the buffer as-is.
 962          */
 963         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 964
 965         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 966             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 967                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 968                 if (!new_skb) {
 969                         ip_rt_put(rt);
 970                         dev->stats.tx_dropped++;
 971                         kfree_skb(skb);
 972                         return NETDEV_TX_OK;
 973                 }
 974                 if (skb->sk)
 975                         skb_set_owner_w(new_skb, skb->sk);
 976                 dev_kfree_skb(skb);
 977                 skb = new_skb;
 978                 iph6 = ipv6_hdr(skb);
 979         }
 980         ttl = tiph->ttl;
 981         if (ttl == 0)
 982                 ttl = iph6->hop_limit;
 983         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 984
 985         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 986                 ip_rt_put(rt);
 987                 goto tx_error;
 988         }
 989
 990         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 991
 992         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 993                       df, !net_eq(tunnel->net, dev_net(dev)));
 994         return NETDEV_TX_OK;
 995
 996 tx_error_icmp:
 997         dst_link_failure(skb);
 998 tx_error:
 999         kfree_skb(skb);
1000         dev->stats.tx_errors++;
1001         return NETDEV_TX_OK;
1002 }
1003
1004 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1005                                      struct net_device *dev, u8 ipproto)
1006 {
1007         struct ip_tunnel *tunnel = netdev_priv(dev);
1008         const struct iphdr  *tiph = &tunnel->parms.iph;
1009
1010         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1011                 goto tx_error;
1012
1013         skb_set_inner_ipproto(skb, ipproto);
1014
1015         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1016         return NETDEV_TX_OK;
1017 tx_error:
1018         kfree_skb(skb);
1019         dev->stats.tx_errors++;
1020         return NETDEV_TX_OK;
1021 }
1022
1023 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1024                                    struct net_device *dev)
1025 {
1026         if (!pskb_inet_may_pull(skb))
1027                 goto tx_err;
1028
1029         switch (skb->protocol) {
1030         case htons(ETH_P_IP):
1031                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1032                 break;
1033         case htons(ETH_P_IPV6):
1034                 ipip6_tunnel_xmit(skb, dev);
1035                 break;
1036 #if IS_ENABLED(CONFIG_MPLS)
1037         case htons(ETH_P_MPLS_UC):
1038                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1039                 break;
1040 #endif
1041         default:
1042                 goto tx_err;
1043         }
1044
1045         return NETDEV_TX_OK;
1046
1047 tx_err:
1048         dev->stats.tx_errors++;
1049         kfree_skb(skb);
1050         return NETDEV_TX_OK;
1051
1052 }
1053
1054 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1055 {
1056         struct ip_tunnel *tunnel = netdev_priv(dev);
1057         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1058         struct net_device *tdev = NULL;
1059         int hlen = LL_MAX_HEADER;
1060         const struct iphdr *iph;
1061         struct flowi4 fl4;
1062
1063         iph = &tunnel->parms.iph;
1064
1065         if (iph->daddr) {
1066                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1067                                                           NULL,
1068                                                           iph->daddr, iph->saddr,
1069                                                           0, 0,
1070                                                           IPPROTO_IPV6,
1071                                                           RT_TOS(iph->tos),
1072                                                           tunnel->parms.link);
1073
1074                 if (!IS_ERR(rt)) {
1075                         tdev = rt->dst.dev;
1076                         ip_rt_put(rt);
1077                 }
1078                 dev->flags |= IFF_POINTOPOINT;
1079         }
1080
1081         if (!tdev && tunnel->parms.link)
1082                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1083
1084         if (tdev && !netif_is_l3_master(tdev)) {
1085                 int mtu;
1086
1087                 mtu = tdev->mtu - t_hlen;
1088                 if (mtu < IPV6_MIN_MTU)
1089                         mtu = IPV6_MIN_MTU;
1090                 WRITE_ONCE(dev->mtu, mtu);
1091                 hlen = tdev->hard_header_len + tdev->needed_headroom;
1092         }
1093         dev->needed_headroom = t_hlen + hlen;
1094 }
1095
1096 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1097                                 __u32 fwmark)
1098 {
1099         struct net *net = t->net;
1100         struct sit_net *sitn = net_generic(net, sit_net_id);
1101
1102         ipip6_tunnel_unlink(sitn, t);
1103         synchronize_net();
1104         t->parms.iph.saddr = p->iph.saddr;
1105         t->parms.iph.daddr = p->iph.daddr;
1106         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1107         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1108         ipip6_tunnel_link(sitn, t);
1109         t->parms.iph.ttl = p->iph.ttl;
1110         t->parms.iph.tos = p->iph.tos;
1111         t->parms.iph.frag_off = p->iph.frag_off;
1112         if (t->parms.link != p->link || t->fwmark != fwmark) {
1113                 t->parms.link = p->link;
1114                 t->fwmark = fwmark;
1115                 ipip6_tunnel_bind_dev(t->dev);
1116         }
1117         dst_cache_reset(&t->dst_cache);
1118         netdev_state_change(t->dev);
1119 }
1120
1121 #ifdef CONFIG_IPV6_SIT_6RD
1122 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1123                                    struct ip_tunnel_6rd *ip6rd)
1124 {
1125         struct in6_addr prefix;
1126         __be32 relay_prefix;
1127
1128         if (ip6rd->relay_prefixlen > 32 ||
1129             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1130                 return -EINVAL;
1131
1132         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1133         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1134                 return -EINVAL;
1135         if (ip6rd->relay_prefixlen)
1136                 relay_prefix = ip6rd->relay_prefix &
1137                                htonl(0xffffffffUL <<
1138                                      (32 - ip6rd->relay_prefixlen));
1139         else
1140                 relay_prefix = 0;
1141         if (relay_prefix != ip6rd->relay_prefix)
1142                 return -EINVAL;
1143
1144         t->ip6rd.prefix = prefix;
1145         t->ip6rd.relay_prefix = relay_prefix;
1146         t->ip6rd.prefixlen = ip6rd->prefixlen;
1147         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1148         dst_cache_reset(&t->dst_cache);
1149         netdev_state_change(t->dev);
1150         return 0;
1151 }
1152 #endif
1153
1154 static bool ipip6_valid_ip_proto(u8 ipproto)
1155 {
1156         return ipproto == IPPROTO_IPV6 ||
1157                 ipproto == IPPROTO_IPIP ||
1158 #if IS_ENABLED(CONFIG_MPLS)
1159                 ipproto == IPPROTO_MPLS ||
1160 #endif
1161                 ipproto == 0;
1162 }
1163
1164 static int
1165 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1166 {
1167         int err = 0;
1168         struct ip_tunnel_parm p;
1169         struct ip_tunnel_prl prl;
1170         struct ip_tunnel *t = netdev_priv(dev);
1171         struct net *net = t->net;
1172         struct sit_net *sitn = net_generic(net, sit_net_id);
1173 #ifdef CONFIG_IPV6_SIT_6RD
1174         struct ip_tunnel_6rd ip6rd;
1175 #endif
1176
1177         switch (cmd) {
1178         case SIOCGETTUNNEL:
1179 #ifdef CONFIG_IPV6_SIT_6RD
1180         case SIOCGET6RD:
1181 #endif
1182                 if (dev == sitn->fb_tunnel_dev) {
1183                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1184                                 err = -EFAULT;
1185                                 break;
1186                         }
1187                         t = ipip6_tunnel_locate(net, &p, 0);
1188                         if (!t)
1189                                 t = netdev_priv(dev);
1190                 }
1191
1192                 err = -EFAULT;
1193                 if (cmd == SIOCGETTUNNEL) {
1194                         memcpy(&p, &t->parms, sizeof(p));
1195                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1196                                          sizeof(p)))
1197                                 goto done;
1198 #ifdef CONFIG_IPV6_SIT_6RD
1199                 } else {
1200                         ip6rd.prefix = t->ip6rd.prefix;
1201                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1202                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1203                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1204                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1205                                          sizeof(ip6rd)))
1206                                 goto done;
1207 #endif
1208                 }
1209                 err = 0;
1210                 break;
1211
1212         case SIOCADDTUNNEL:
1213         case SIOCCHGTUNNEL:
1214                 err = -EPERM;
1215                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1216                         goto done;
1217
1218                 err = -EFAULT;
1219                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1220                         goto done;
1221
1222                 err = -EINVAL;
1223                 if (!ipip6_valid_ip_proto(p.iph.protocol))
1224                         goto done;
1225                 if (p.iph.version != 4 ||
1226                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1227                         goto done;
1228                 if (p.iph.ttl)
1229                         p.iph.frag_off |= htons(IP_DF);
1230
1231                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1232
1233                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1234                         if (t) {
1235                                 if (t->dev != dev) {
1236                                         err = -EEXIST;
1237                                         break;
1238                                 }
1239                         } else {
1240                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1241                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1242                                         err = -EINVAL;
1243                                         break;
1244                                 }
1245                                 t = netdev_priv(dev);
1246                         }
1247
1248                         ipip6_tunnel_update(t, &p, t->fwmark);
1249                 }
1250
1251                 if (t) {
1252                         err = 0;
1253                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1254                                 err = -EFAULT;
1255                 } else
1256                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1257                 break;
1258
1259         case SIOCDELTUNNEL:
1260                 err = -EPERM;
1261                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1262                         goto done;
1263
1264                 if (dev == sitn->fb_tunnel_dev) {
1265                         err = -EFAULT;
1266                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1267                                 goto done;
1268                         err = -ENOENT;
1269                         t = ipip6_tunnel_locate(net, &p, 0);
1270                         if (!t)
1271                                 goto done;
1272                         err = -EPERM;
1273                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1274                                 goto done;
1275                         dev = t->dev;
1276                 }
1277                 unregister_netdevice(dev);
1278                 err = 0;
1279                 break;
1280
1281         case SIOCGETPRL:
1282                 err = -EINVAL;
1283                 if (dev == sitn->fb_tunnel_dev)
1284                         goto done;
1285                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1286                 break;
1287
1288         case SIOCADDPRL:
1289         case SIOCDELPRL:
1290         case SIOCCHGPRL:
1291                 err = -EPERM;
1292                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1293                         goto done;
1294                 err = -EINVAL;
1295                 if (dev == sitn->fb_tunnel_dev)
1296                         goto done;
1297                 err = -EFAULT;
1298                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1299                         goto done;
1300
1301                 switch (cmd) {
1302                 case SIOCDELPRL:
1303                         err = ipip6_tunnel_del_prl(t, &prl);
1304                         break;
1305                 case SIOCADDPRL:
1306                 case SIOCCHGPRL:
1307                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1308                         break;
1309                 }
1310                 dst_cache_reset(&t->dst_cache);
1311                 netdev_state_change(dev);
1312                 break;
1313
1314 #ifdef CONFIG_IPV6_SIT_6RD
1315         case SIOCADD6RD:
1316         case SIOCCHG6RD:
1317         case SIOCDEL6RD:
1318                 err = -EPERM;
1319                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1320                         goto done;
1321
1322                 err = -EFAULT;
1323                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1324                                    sizeof(ip6rd)))
1325                         goto done;
1326
1327                 if (cmd != SIOCDEL6RD) {
1328                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1329                         if (err < 0)
1330                                 goto done;
1331                 } else
1332                         ipip6_tunnel_clone_6rd(dev, sitn);
1333
1334                 err = 0;
1335                 break;
1336 #endif
1337
1338         default:
1339                 err = -EINVAL;
1340         }
1341
1342 done:
1343         return err;
1344 }
1345
1346 static const struct net_device_ops ipip6_netdev_ops = {
1347         .ndo_init       = ipip6_tunnel_init,
1348         .ndo_uninit     = ipip6_tunnel_uninit,
1349         .ndo_start_xmit = sit_tunnel_xmit,
1350         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1351         .ndo_get_stats64 = ip_tunnel_get_stats64,
1352         .ndo_get_iflink = ip_tunnel_get_iflink,
1353 };
1354
1355 static void ipip6_dev_free(struct net_device *dev)
1356 {
1357         struct ip_tunnel *tunnel = netdev_priv(dev);
1358
1359         dst_cache_destroy(&tunnel->dst_cache);
1360         free_percpu(dev->tstats);
1361 }
1362
1363 #define SIT_FEATURES (NETIF_F_SG           | \
1364                       NETIF_F_FRAGLIST     | \
1365                       NETIF_F_HIGHDMA      | \
1366                       NETIF_F_GSO_SOFTWARE | \
1367                       NETIF_F_HW_CSUM)
1368
1369 static void ipip6_tunnel_setup(struct net_device *dev)
1370 {
1371         struct ip_tunnel *tunnel = netdev_priv(dev);
1372         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1373
1374         dev->netdev_ops         = &ipip6_netdev_ops;
1375         dev->needs_free_netdev  = true;
1376         dev->priv_destructor    = ipip6_dev_free;
1377
1378         dev->type               = ARPHRD_SIT;
1379         dev->mtu                = ETH_DATA_LEN - t_hlen;
1380         dev->min_mtu            = IPV6_MIN_MTU;
1381         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1382         dev->flags              = IFF_NOARP;
1383         netif_keep_dst(dev);
1384         dev->addr_len           = 4;
1385         dev->features           |= NETIF_F_LLTX;
1386         dev->features           |= SIT_FEATURES;
1387         dev->hw_features        |= SIT_FEATURES;
1388 }
1389
1390 static int ipip6_tunnel_init(struct net_device *dev)
1391 {
1392         struct ip_tunnel *tunnel = netdev_priv(dev);
1393         int err;
1394
1395         tunnel->dev = dev;
1396         tunnel->net = dev_net(dev);
1397         strcpy(tunnel->parms.name, dev->name);
1398
1399         ipip6_tunnel_bind_dev(dev);
1400         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1401         if (!dev->tstats)
1402                 return -ENOMEM;
1403
1404         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1405         if (err) {
1406                 free_percpu(dev->tstats);
1407                 dev->tstats = NULL;
1408                 return err;
1409         }
1410         dev_hold(dev);
1411         return 0;
1412 }
1413
1414 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1415 {
1416         struct ip_tunnel *tunnel = netdev_priv(dev);
1417         struct iphdr *iph = &tunnel->parms.iph;
1418         struct net *net = dev_net(dev);
1419         struct sit_net *sitn = net_generic(net, sit_net_id);
1420
1421         iph->version            = 4;
1422         iph->protocol           = IPPROTO_IPV6;
1423         iph->ihl                = 5;
1424         iph->ttl                = 64;
1425
1426         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1427 }
1428
1429 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1430                           struct netlink_ext_ack *extack)
1431 {
1432         u8 proto;
1433
1434         if (!data || !data[IFLA_IPTUN_PROTO])
1435                 return 0;
1436
1437         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1438         if (!ipip6_valid_ip_proto(proto))
1439                 return -EINVAL;
1440
1441         return 0;
1442 }
1443
1444 static void ipip6_netlink_parms(struct nlattr *data[],
1445                                 struct ip_tunnel_parm *parms,
1446                                 __u32 *fwmark)
1447 {
1448         memset(parms, 0, sizeof(*parms));
1449
1450         parms->iph.version = 4;
1451         parms->iph.protocol = IPPROTO_IPV6;
1452         parms->iph.ihl = 5;
1453         parms->iph.ttl = 64;
1454
1455         if (!data)
1456                 return;
1457
1458         if (data[IFLA_IPTUN_LINK])
1459                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1460
1461         if (data[IFLA_IPTUN_LOCAL])
1462                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1463
1464         if (data[IFLA_IPTUN_REMOTE])
1465                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1466
1467         if (data[IFLA_IPTUN_TTL]) {
1468                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1469                 if (parms->iph.ttl)
1470                         parms->iph.frag_off = htons(IP_DF);
1471         }
1472
1473         if (data[IFLA_IPTUN_TOS])
1474                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1475
1476         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1477                 parms->iph.frag_off = htons(IP_DF);
1478
1479         if (data[IFLA_IPTUN_FLAGS])
1480                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1481
1482         if (data[IFLA_IPTUN_PROTO])
1483                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1484
1485         if (data[IFLA_IPTUN_FWMARK])
1486                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1487 }
1488
1489 /* This function returns true when ENCAP attributes are present in the nl msg */
1490 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1491                                       struct ip_tunnel_encap *ipencap)
1492 {
1493         bool ret = false;
1494
1495         memset(ipencap, 0, sizeof(*ipencap));
1496
1497         if (!data)
1498                 return ret;
1499
1500         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1501                 ret = true;
1502                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1503         }
1504
1505         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1506                 ret = true;
1507                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1508         }
1509
1510         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1511                 ret = true;
1512                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1513         }
1514
1515         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1516                 ret = true;
1517                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1518         }
1519
1520         return ret;
1521 }
1522
1523 #ifdef CONFIG_IPV6_SIT_6RD
1524 /* This function returns true when 6RD attributes are present in the nl msg */
1525 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1526                                     struct ip_tunnel_6rd *ip6rd)
1527 {
1528         bool ret = false;
1529         memset(ip6rd, 0, sizeof(*ip6rd));
1530
1531         if (!data)
1532                 return ret;
1533
1534         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1535                 ret = true;
1536                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1537         }
1538
1539         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1540                 ret = true;
1541                 ip6rd->relay_prefix =
1542                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1543         }
1544
1545         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1546                 ret = true;
1547                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1548         }
1549
1550         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1551                 ret = true;
1552                 ip6rd->relay_prefixlen =
1553                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1554         }
1555
1556         return ret;
1557 }
1558 #endif
1559
1560 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1561                          struct nlattr *tb[], struct nlattr *data[],
1562                          struct netlink_ext_ack *extack)
1563 {
1564         struct net *net = dev_net(dev);
1565         struct ip_tunnel *nt;
1566         struct ip_tunnel_encap ipencap;
1567 #ifdef CONFIG_IPV6_SIT_6RD
1568         struct ip_tunnel_6rd ip6rd;
1569 #endif
1570         int err;
1571
1572         nt = netdev_priv(dev);
1573
1574         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1575                 err = ip_tunnel_encap_setup(nt, &ipencap);
1576                 if (err < 0)
1577                         return err;
1578         }
1579
1580         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1581
1582         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1583                 return -EEXIST;
1584
1585         err = ipip6_tunnel_create(dev);
1586         if (err < 0)
1587                 return err;
1588
1589         if (tb[IFLA_MTU]) {
1590                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1591
1592                 if (mtu >= IPV6_MIN_MTU &&
1593                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1594                         dev->mtu = mtu;
1595         }
1596
1597 #ifdef CONFIG_IPV6_SIT_6RD
1598         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1599                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1600                 if (err < 0)
1601                         unregister_netdevice_queue(dev, NULL);
1602         }
1603 #endif
1604
1605         return err;
1606 }
1607
1608 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1609                             struct nlattr *data[],
1610                             struct netlink_ext_ack *extack)
1611 {
1612         struct ip_tunnel *t = netdev_priv(dev);
1613         struct ip_tunnel_parm p;
1614         struct ip_tunnel_encap ipencap;
1615         struct net *net = t->net;
1616         struct sit_net *sitn = net_generic(net, sit_net_id);
1617 #ifdef CONFIG_IPV6_SIT_6RD
1618         struct ip_tunnel_6rd ip6rd;
1619 #endif
1620         __u32 fwmark = t->fwmark;
1621         int err;
1622
1623         if (dev == sitn->fb_tunnel_dev)
1624                 return -EINVAL;
1625
1626         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1627                 err = ip_tunnel_encap_setup(t, &ipencap);
1628                 if (err < 0)
1629                         return err;
1630         }
1631
1632         ipip6_netlink_parms(data, &p, &fwmark);
1633
1634         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1635             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1636                 return -EINVAL;
1637
1638         t = ipip6_tunnel_locate(net, &p, 0);
1639
1640         if (t) {
1641                 if (t->dev != dev)
1642                         return -EEXIST;
1643         } else
1644                 t = netdev_priv(dev);
1645
1646         ipip6_tunnel_update(t, &p, fwmark);
1647
1648 #ifdef CONFIG_IPV6_SIT_6RD
1649         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1650                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1651 #endif
1652
1653         return 0;
1654 }
1655
1656 static size_t ipip6_get_size(const struct net_device *dev)
1657 {
1658         return
1659                 /* IFLA_IPTUN_LINK */
1660                 nla_total_size(4) +
1661                 /* IFLA_IPTUN_LOCAL */
1662                 nla_total_size(4) +
1663                 /* IFLA_IPTUN_REMOTE */
1664                 nla_total_size(4) +
1665                 /* IFLA_IPTUN_TTL */
1666                 nla_total_size(1) +
1667                 /* IFLA_IPTUN_TOS */
1668                 nla_total_size(1) +
1669                 /* IFLA_IPTUN_PMTUDISC */
1670                 nla_total_size(1) +
1671                 /* IFLA_IPTUN_FLAGS */
1672                 nla_total_size(2) +
1673                 /* IFLA_IPTUN_PROTO */
1674                 nla_total_size(1) +
1675 #ifdef CONFIG_IPV6_SIT_6RD
1676                 /* IFLA_IPTUN_6RD_PREFIX */
1677                 nla_total_size(sizeof(struct in6_addr)) +
1678                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1679                 nla_total_size(4) +
1680                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1681                 nla_total_size(2) +
1682                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1683                 nla_total_size(2) +
1684 #endif
1685                 /* IFLA_IPTUN_ENCAP_TYPE */
1686                 nla_total_size(2) +
1687                 /* IFLA_IPTUN_ENCAP_FLAGS */
1688                 nla_total_size(2) +
1689                 /* IFLA_IPTUN_ENCAP_SPORT */
1690                 nla_total_size(2) +
1691                 /* IFLA_IPTUN_ENCAP_DPORT */
1692                 nla_total_size(2) +
1693                 /* IFLA_IPTUN_FWMARK */
1694                 nla_total_size(4) +
1695                 0;
1696 }
1697
1698 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1699 {
1700         struct ip_tunnel *tunnel = netdev_priv(dev);
1701         struct ip_tunnel_parm *parm = &tunnel->parms;
1702
1703         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1704             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1705             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1706             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1707             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1708             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1709                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1710             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1711             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1712             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1713                 goto nla_put_failure;
1714
1715 #ifdef CONFIG_IPV6_SIT_6RD
1716         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1717                              &tunnel->ip6rd.prefix) ||
1718             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1719                             tunnel->ip6rd.relay_prefix) ||
1720             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1721                         tunnel->ip6rd.prefixlen) ||
1722             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1723                         tunnel->ip6rd.relay_prefixlen))
1724                 goto nla_put_failure;
1725 #endif
1726
1727         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1728                         tunnel->encap.type) ||
1729             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1730                         tunnel->encap.sport) ||
1731             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1732                         tunnel->encap.dport) ||
1733             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1734                         tunnel->encap.flags))
1735                 goto nla_put_failure;
1736
1737         return 0;
1738
1739 nla_put_failure:
1740         return -EMSGSIZE;
1741 }
1742
1743 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1744         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1745         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1746         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1747         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1748         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1749         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1750         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1751         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1752 #ifdef CONFIG_IPV6_SIT_6RD
1753         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1754         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1755         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1756         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1757 #endif
1758         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1759         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1760         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1761         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1762         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1763 };
1764
1765 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1766 {
1767         struct net *net = dev_net(dev);
1768         struct sit_net *sitn = net_generic(net, sit_net_id);
1769
1770         if (dev != sitn->fb_tunnel_dev)
1771                 unregister_netdevice_queue(dev, head);
1772 }
1773
1774 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1775         .kind           = "sit",
1776         .maxtype        = IFLA_IPTUN_MAX,
1777         .policy         = ipip6_policy,
1778         .priv_size      = sizeof(struct ip_tunnel),
1779         .setup          = ipip6_tunnel_setup,
1780         .validate       = ipip6_validate,
1781         .newlink        = ipip6_newlink,
1782         .changelink     = ipip6_changelink,
1783         .get_size       = ipip6_get_size,
1784         .fill_info      = ipip6_fill_info,
1785         .dellink        = ipip6_dellink,
1786         .get_link_net   = ip_tunnel_get_link_net,
1787 };
1788
1789 static struct xfrm_tunnel sit_handler __read_mostly = {
1790         .handler        =       ipip6_rcv,
1791         .err_handler    =       ipip6_err,
1792         .priority       =       1,
1793 };
1794
1795 static struct xfrm_tunnel ipip_handler __read_mostly = {
1796         .handler        =       ipip_rcv,
1797         .err_handler    =       ipip6_err,
1798         .priority       =       2,
1799 };
1800
1801 #if IS_ENABLED(CONFIG_MPLS)
1802 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1803         .handler        =       mplsip_rcv,
1804         .err_handler    =       ipip6_err,
1805         .priority       =       2,
1806 };
1807 #endif
1808
1809 static void __net_exit sit_destroy_tunnels(struct net *net,
1810                                            struct list_head *head)
1811 {
1812         struct sit_net *sitn = net_generic(net, sit_net_id);
1813         struct net_device *dev, *aux;
1814         int prio;
1815
1816         for_each_netdev_safe(net, dev, aux)
1817                 if (dev->rtnl_link_ops == &sit_link_ops)
1818                         unregister_netdevice_queue(dev, head);
1819
1820         for (prio = 0; prio < 4; prio++) {
1821                 int h;
1822                 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1823                         struct ip_tunnel *t;
1824
1825                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1826                         while (t) {
1827                                 /* If dev is in the same netns, it has already
1828                                  * been added to the list by the previous loop.
1829                                  */
1830                                 if (!net_eq(dev_net(t->dev), net))
1831                                         unregister_netdevice_queue(t->dev,
1832                                                                    head);
1833                                 t = rtnl_dereference(t->next);
1834                         }
1835                 }
1836         }
1837 }
1838
1839 static int __net_init sit_init_net(struct net *net)
1840 {
1841         struct sit_net *sitn = net_generic(net, sit_net_id);
1842         struct ip_tunnel *t;
1843         int err;
1844
1845         sitn->tunnels[0] = sitn->tunnels_wc;
1846         sitn->tunnels[1] = sitn->tunnels_l;
1847         sitn->tunnels[2] = sitn->tunnels_r;
1848         sitn->tunnels[3] = sitn->tunnels_r_l;
1849
1850         if (!net_has_fallback_tunnels(net))
1851                 return 0;
1852
1853         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1854                                            NET_NAME_UNKNOWN,
1855                                            ipip6_tunnel_setup);
1856         if (!sitn->fb_tunnel_dev) {
1857                 err = -ENOMEM;
1858                 goto err_alloc_dev;
1859         }
1860         dev_net_set(sitn->fb_tunnel_dev, net);
1861         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1862         /* FB netdevice is special: we have one, and only one per netns.
1863          * Allowing to move it to another netns is clearly unsafe.
1864          */
1865         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1866
1867         err = register_netdev(sitn->fb_tunnel_dev);
1868         if (err)
1869                 goto err_reg_dev;
1870
1871         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1872         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1873
1874         t = netdev_priv(sitn->fb_tunnel_dev);
1875
1876         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1877         return 0;
1878
1879 err_reg_dev:
1880         free_netdev(sitn->fb_tunnel_dev);
1881 err_alloc_dev:
1882         return err;
1883 }
1884
1885 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1886 {
1887         LIST_HEAD(list);
1888         struct net *net;
1889
1890         rtnl_lock();
1891         list_for_each_entry(net, net_list, exit_list)
1892                 sit_destroy_tunnels(net, &list);
1893
1894         unregister_netdevice_many(&list);
1895         rtnl_unlock();
1896 }
1897
1898 static struct pernet_operations sit_net_ops = {
1899         .init = sit_init_net,
1900         .exit_batch = sit_exit_batch_net,
1901         .id   = &sit_net_id,
1902         .size = sizeof(struct sit_net),
1903 };
1904
1905 static void __exit sit_cleanup(void)
1906 {
1907         rtnl_link_unregister(&sit_link_ops);
1908         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1909         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1910 #if IS_ENABLED(CONFIG_MPLS)
1911         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1912 #endif
1913
1914         unregister_pernet_device(&sit_net_ops);
1915         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1916 }
1917
1918 static int __init sit_init(void)
1919 {
1920         int err;
1921
1922         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1923
1924         err = register_pernet_device(&sit_net_ops);
1925         if (err < 0)
1926                 return err;
1927         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1928         if (err < 0) {
1929                 pr_info("%s: can't register ip6ip4\n", __func__);
1930                 goto xfrm_tunnel_failed;
1931         }
1932         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1933         if (err < 0) {
1934                 pr_info("%s: can't register ip4ip4\n", __func__);
1935                 goto xfrm_tunnel4_failed;
1936         }
1937 #if IS_ENABLED(CONFIG_MPLS)
1938         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1939         if (err < 0) {
1940                 pr_info("%s: can't register mplsip\n", __func__);
1941                 goto xfrm_tunnel_mpls_failed;
1942         }
1943 #endif
1944         err = rtnl_link_register(&sit_link_ops);
1945         if (err < 0)
1946                 goto rtnl_link_failed;
1947
1948 out:
1949         return err;
1950
1951 rtnl_link_failed:
1952 #if IS_ENABLED(CONFIG_MPLS)
1953         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1954 xfrm_tunnel_mpls_failed:
1955 #endif
1956         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1957 xfrm_tunnel4_failed:
1958         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1959 xfrm_tunnel_failed:
1960         unregister_pernet_device(&sit_net_ops);
1961         goto out;
1962 }
1963
1964 module_init(sit_init);
1965 module_exit(sit_cleanup);
1966 MODULE_LICENSE("GPL");
1967 MODULE_ALIAS_RTNL_LINK("sit");
1968 MODULE_ALIAS_NETDEV("sit0");