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