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 <asm/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 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 int sit_net_id __read_mostly;
  80 struct sit_net {
  81         struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
  82         struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
  83         struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
  84         struct ip_tunnel __rcu *tunnels_wc[1];
  85         struct ip_tunnel __rcu **tunnels[4];
  86
  87         struct net_device *fb_tunnel_dev;
  88 };
  89
  90 /*
  91  * Must be invoked with rcu_read_lock
  92  */
  93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  94                 struct net_device *dev, __be32 remote, __be32 local)
  95 {
  96         unsigned int h0 = HASH(remote);
  97         unsigned int h1 = HASH(local);
  98         struct ip_tunnel *t;
  99         struct sit_net *sitn = net_generic(net, sit_net_id);
 100
 101         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 102                 if (local == t->parms.iph.saddr &&
 103                     remote == t->parms.iph.daddr &&
 104                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 105                     (t->dev->flags & IFF_UP))
 106                         return t;
 107         }
 108         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 109                 if (remote == t->parms.iph.daddr &&
 110                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 111                     (t->dev->flags & IFF_UP))
 112                         return t;
 113         }
 114         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 115                 if (local == t->parms.iph.saddr &&
 116                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 117                     (t->dev->flags & IFF_UP))
 118                         return t;
 119         }
 120         t = rcu_dereference(sitn->tunnels_wc[0]);
 121         if ((t != NULL) && (t->dev->flags & IFF_UP))
 122                 return t;
 123         return NULL;
 124 }
 125
 126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 127                 struct ip_tunnel_parm *parms)
 128 {
 129         __be32 remote = parms->iph.daddr;
 130         __be32 local = parms->iph.saddr;
 131         unsigned int h = 0;
 132         int prio = 0;
 133
 134         if (remote) {
 135                 prio |= 2;
 136                 h ^= HASH(remote);
 137         }
 138         if (local) {
 139                 prio |= 1;
 140                 h ^= HASH(local);
 141         }
 142         return &sitn->tunnels[prio][h];
 143 }
 144
 145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 146                 struct ip_tunnel *t)
 147 {
 148         return __ipip6_bucket(sitn, &t->parms);
 149 }
 150
 151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 152 {
 153         struct ip_tunnel __rcu **tp;
 154         struct ip_tunnel *iter;
 155
 156         for (tp = ipip6_bucket(sitn, t);
 157              (iter = rtnl_dereference(*tp)) != NULL;
 158              tp = &iter->next) {
 159                 if (t == iter) {
 160                         rcu_assign_pointer(*tp, t->next);
 161                         break;
 162                 }
 163         }
 164 }
 165
 166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 167 {
 168         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 169
 170         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 171         rcu_assign_pointer(*tp, t);
 172 }
 173
 174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 175 {
 176 #ifdef CONFIG_IPV6_SIT_6RD
 177         struct ip_tunnel *t = netdev_priv(dev);
 178
 179         if (t->dev == sitn->fb_tunnel_dev) {
 180                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 181                 t->ip6rd.relay_prefix = 0;
 182                 t->ip6rd.prefixlen = 16;
 183                 t->ip6rd.relay_prefixlen = 0;
 184         } else {
 185                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 186                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 187         }
 188 #endif
 189 }
 190
 191 static int ipip6_tunnel_create(struct net_device *dev)
 192 {
 193         struct ip_tunnel *t = netdev_priv(dev);
 194         struct net *net = dev_net(dev);
 195         struct sit_net *sitn = net_generic(net, sit_net_id);
 196         int err;
 197
 198         err = ipip6_tunnel_init(dev);
 199         if (err < 0)
 200                 goto out;
 201         ipip6_tunnel_clone_6rd(dev, sitn);
 202
 203         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 204                 dev->priv_flags |= IFF_ISATAP;
 205
 206         err = register_netdevice(dev);
 207         if (err < 0)
 208                 goto out;
 209
 210         strcpy(t->parms.name, dev->name);
 211         dev->rtnl_link_ops = &sit_link_ops;
 212
 213         dev_hold(dev);
 214
 215         ipip6_tunnel_link(sitn, t);
 216         return 0;
 217
 218 out:
 219         return err;
 220 }
 221
 222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 223                 struct ip_tunnel_parm *parms, int create)
 224 {
 225         __be32 remote = parms->iph.daddr;
 226         __be32 local = parms->iph.saddr;
 227         struct ip_tunnel *t, *nt;
 228         struct ip_tunnel __rcu **tp;
 229         struct net_device *dev;
 230         char name[IFNAMSIZ];
 231         struct sit_net *sitn = net_generic(net, sit_net_id);
 232
 233         for (tp = __ipip6_bucket(sitn, parms);
 234             (t = rtnl_dereference(*tp)) != NULL;
 235              tp = &t->next) {
 236                 if (local == t->parms.iph.saddr &&
 237                     remote == t->parms.iph.daddr &&
 238                     parms->link == t->parms.link) {
 239                         if (create)
 240                                 return NULL;
 241                         else
 242                                 return t;
 243                 }
 244         }
 245         if (!create)
 246                 goto failed;
 247
 248         if (parms->name[0])
 249                 strlcpy(name, parms->name, IFNAMSIZ);
 250         else
 251                 strcpy(name, "sit%d");
 252
 253         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 254         if (dev == NULL)
 255                 return NULL;
 256
 257         dev_net_set(dev, net);
 258
 259         nt = netdev_priv(dev);
 260
 261         nt->parms = *parms;
 262         if (ipip6_tunnel_create(dev) < 0)
 263                 goto failed_free;
 264
 265         return nt;
 266
 267 failed_free:
 268         ipip6_dev_free(dev);
 269 failed:
 270         return NULL;
 271 }
 272
 273 #define for_each_prl_rcu(start)                 \
 274         for (prl = rcu_dereference(start);      \
 275              prl;                               \
 276              prl = rcu_dereference(prl->next))
 277
 278 static struct ip_tunnel_prl_entry *
 279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 280 {
 281         struct ip_tunnel_prl_entry *prl;
 282
 283         for_each_prl_rcu(t->prl)
 284                 if (prl->addr == addr)
 285                         break;
 286         return prl;
 287
 288 }
 289
 290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 291                                 struct ip_tunnel_prl __user *a)
 292 {
 293         struct ip_tunnel_prl kprl, *kp;
 294         struct ip_tunnel_prl_entry *prl;
 295         unsigned int cmax, c = 0, ca, len;
 296         int ret = 0;
 297
 298         if (copy_from_user(&kprl, a, sizeof(kprl)))
 299                 return -EFAULT;
 300         cmax = kprl.datalen / sizeof(kprl);
 301         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 302                 cmax = 1;
 303
 304         /* For simple GET or for root users,
 305          * we try harder to allocate.
 306          */
 307         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 308                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 309                 NULL;
 310
 311         rcu_read_lock();
 312
 313         ca = t->prl_count < cmax ? t->prl_count : cmax;
 314
 315         if (!kp) {
 316                 /* We don't try hard to allocate much memory for
 317                  * non-root users.
 318                  * For root users, retry allocating enough memory for
 319                  * the answer.
 320                  */
 321                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 322                 if (!kp) {
 323                         ret = -ENOMEM;
 324                         goto out;
 325                 }
 326         }
 327
 328         c = 0;
 329         for_each_prl_rcu(t->prl) {
 330                 if (c >= cmax)
 331                         break;
 332                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 333                         continue;
 334                 kp[c].addr = prl->addr;
 335                 kp[c].flags = prl->flags;
 336                 c++;
 337                 if (kprl.addr != htonl(INADDR_ANY))
 338                         break;
 339         }
 340 out:
 341         rcu_read_unlock();
 342
 343         len = sizeof(*kp) * c;
 344         ret = 0;
 345         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 346                 ret = -EFAULT;
 347
 348         kfree(kp);
 349
 350         return ret;
 351 }
 352
 353 static int
 354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 355 {
 356         struct ip_tunnel_prl_entry *p;
 357         int err = 0;
 358
 359         if (a->addr == htonl(INADDR_ANY))
 360                 return -EINVAL;
 361
 362         ASSERT_RTNL();
 363
 364         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 365                 if (p->addr == a->addr) {
 366                         if (chg) {
 367                                 p->flags = a->flags;
 368                                 goto out;
 369                         }
 370                         err = -EEXIST;
 371                         goto out;
 372                 }
 373         }
 374
 375         if (chg) {
 376                 err = -ENXIO;
 377                 goto out;
 378         }
 379
 380         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 381         if (!p) {
 382                 err = -ENOBUFS;
 383                 goto out;
 384         }
 385
 386         p->next = t->prl;
 387         p->addr = a->addr;
 388         p->flags = a->flags;
 389         t->prl_count++;
 390         rcu_assign_pointer(t->prl, p);
 391 out:
 392         return err;
 393 }
 394
 395 static void prl_list_destroy_rcu(struct rcu_head *head)
 396 {
 397         struct ip_tunnel_prl_entry *p, *n;
 398
 399         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 400         do {
 401                 n = rcu_dereference_protected(p->next, 1);
 402                 kfree(p);
 403                 p = n;
 404         } while (p);
 405 }
 406
 407 static int
 408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 409 {
 410         struct ip_tunnel_prl_entry *x;
 411         struct ip_tunnel_prl_entry __rcu **p;
 412         int err = 0;
 413
 414         ASSERT_RTNL();
 415
 416         if (a && a->addr != htonl(INADDR_ANY)) {
 417                 for (p = &t->prl;
 418                      (x = rtnl_dereference(*p)) != NULL;
 419                      p = &x->next) {
 420                         if (x->addr == a->addr) {
 421                                 *p = x->next;
 422                                 kfree_rcu(x, rcu_head);
 423                                 t->prl_count--;
 424                                 goto out;
 425                         }
 426                 }
 427                 err = -ENXIO;
 428         } else {
 429                 x = rtnl_dereference(t->prl);
 430                 if (x) {
 431                         t->prl_count = 0;
 432                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 433                         t->prl = NULL;
 434                 }
 435         }
 436 out:
 437         return err;
 438 }
 439
 440 static int
 441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 442 {
 443         struct ip_tunnel_prl_entry *p;
 444         int ok = 1;
 445
 446         rcu_read_lock();
 447         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 448         if (p) {
 449                 if (p->flags & PRL_DEFAULT)
 450                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 451                 else
 452                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 453         } else {
 454                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 455
 456                 if (ipv6_addr_is_isatap(addr6) &&
 457                     (addr6->s6_addr32[3] == iph->saddr) &&
 458                     ipv6_chk_prefix(addr6, t->dev))
 459                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 460                 else
 461                         ok = 0;
 462         }
 463         rcu_read_unlock();
 464         return ok;
 465 }
 466
 467 static void ipip6_tunnel_uninit(struct net_device *dev)
 468 {
 469         struct net *net = dev_net(dev);
 470         struct sit_net *sitn = net_generic(net, sit_net_id);
 471
 472         if (dev == sitn->fb_tunnel_dev) {
 473                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 474         } else {
 475                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 476                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 477         }
 478         dev_put(dev);
 479 }
 480
 481
 482 static int ipip6_err(struct sk_buff *skb, u32 info)
 483 {
 484
 485 /* All the routers (except for Linux) return only
 486    8 bytes of packet payload. It means, that precise relaying of
 487    ICMP in the real Internet is absolutely infeasible.
 488  */
 489         const struct iphdr *iph = (const struct iphdr *)skb->data;
 490         const int type = icmp_hdr(skb)->type;
 491         const int code = icmp_hdr(skb)->code;
 492         struct ip_tunnel *t;
 493         int err;
 494
 495         switch (type) {
 496         default:
 497         case ICMP_PARAMETERPROB:
 498                 return 0;
 499
 500         case ICMP_DEST_UNREACH:
 501                 switch (code) {
 502                 case ICMP_SR_FAILED:
 503                 case ICMP_PORT_UNREACH:
 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                 break;
 518         case ICMP_REDIRECT:
 519                 break;
 520         }
 521
 522         err = -ENOENT;
 523
 524         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 525                                 skb->dev,
 526                                 iph->daddr,
 527                                 iph->saddr);
 528         if (t == NULL)
 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->dev->ifindex, 0, IPPROTO_IPV6, 0);
 534                 err = 0;
 535                 goto out;
 536         }
 537         if (type == ICMP_REDIRECT) {
 538                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 539                               IPPROTO_IPV6, 0);
 540                 err = 0;
 541                 goto out;
 542         }
 543
 544         if (t->parms.iph.daddr == 0)
 545                 goto out;
 546
 547         err = 0;
 548         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 549                 goto out;
 550
 551         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 552                 t->err_count++;
 553         else
 554                 t->err_count = 1;
 555         t->err_time = jiffies;
 556 out:
 557         return err;
 558 }
 559
 560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 561                                   const struct in6_addr *v6addr)
 562 {
 563         __be32 v4embed = 0;
 564         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 565                 return true;
 566         return false;
 567 }
 568
 569 /* Checks if an address matches an address on the tunnel interface.
 570  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 571  * Long story:
 572  * This function is called after we considered the packet as spoofed
 573  * in is_spoofed_6rd.
 574  * We may have a router that is doing NAT for proto 41 packets
 575  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 576  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 577  * function will return true, dropping the packet.
 578  * But, we can still check if is spoofed against the IP
 579  * addresses associated with the interface.
 580  */
 581 static bool only_dnatted(const struct ip_tunnel *tunnel,
 582         const struct in6_addr *v6dst)
 583 {
 584         int prefix_len;
 585
 586 #ifdef CONFIG_IPV6_SIT_6RD
 587         prefix_len = tunnel->ip6rd.prefixlen + 32
 588                 - tunnel->ip6rd.relay_prefixlen;
 589 #else
 590         prefix_len = 48;
 591 #endif
 592         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 593 }
 594
 595 /* Returns true if a packet is spoofed */
 596 static bool packet_is_spoofed(struct sk_buff *skb,
 597                               const struct iphdr *iph,
 598                               struct ip_tunnel *tunnel)
 599 {
 600         const struct ipv6hdr *ipv6h;
 601
 602         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 603                 if (!isatap_chksrc(skb, iph, tunnel))
 604                         return true;
 605
 606                 return false;
 607         }
 608
 609         if (tunnel->dev->flags & IFF_POINTOPOINT)
 610                 return false;
 611
 612         ipv6h = ipv6_hdr(skb);
 613
 614         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 615                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 616                                      &iph->saddr, &ipv6h->saddr,
 617                                      &iph->daddr, &ipv6h->daddr);
 618                 return true;
 619         }
 620
 621         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 622                 return false;
 623
 624         if (only_dnatted(tunnel, &ipv6h->daddr))
 625                 return false;
 626
 627         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 628                              &iph->saddr, &ipv6h->saddr,
 629                              &iph->daddr, &ipv6h->daddr);
 630         return true;
 631 }
 632
 633 static int ipip6_rcv(struct sk_buff *skb)
 634 {
 635         const struct iphdr *iph = ip_hdr(skb);
 636         struct ip_tunnel *tunnel;
 637         int err;
 638
 639         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 640                                      iph->saddr, iph->daddr);
 641         if (tunnel != NULL) {
 642                 struct pcpu_tstats *tstats;
 643
 644                 secpath_reset(skb);
 645                 skb->mac_header = skb->network_header;
 646                 skb_reset_network_header(skb);
 647                 IPCB(skb)->flags = 0;
 648                 skb->protocol = htons(ETH_P_IPV6);
 649                 skb->pkt_type = PACKET_HOST;
 650
 651                 if (packet_is_spoofed(skb, iph, tunnel)) {
 652                         tunnel->dev->stats.rx_errors++;
 653                         goto out;
 654                 }
 655
 656                 __skb_tunnel_rx(skb, tunnel->dev);
 657
 658                 err = IP_ECN_decapsulate(iph, skb);
 659                 if (unlikely(err)) {
 660                         if (log_ecn_error)
 661                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 662                                                      &iph->saddr, iph->tos);
 663                         if (err > 1) {
 664                                 ++tunnel->dev->stats.rx_frame_errors;
 665                                 ++tunnel->dev->stats.rx_errors;
 666                                 goto out;
 667                         }
 668                 }
 669
 670                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 671                 tstats->rx_packets++;
 672                 tstats->rx_bytes += skb->len;
 673
 674                 netif_rx(skb);
 675
 676                 return 0;
 677         }
 678
 679         /* no tunnel matched,  let upstream know, ipsec may handle it */
 680         return 1;
 681 out:
 682         kfree_skb(skb);
 683         return 0;
 684 }
 685
 686 /*
 687  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 688  * stores the embedded IPv4 address in v4dst and returns true.
 689  */
 690 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 691                       __be32 *v4dst)
 692 {
 693 #ifdef CONFIG_IPV6_SIT_6RD
 694         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 695                               tunnel->ip6rd.prefixlen)) {
 696                 unsigned int pbw0, pbi0;
 697                 int pbi1;
 698                 u32 d;
 699
 700                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 701                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 702
 703                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 704                     tunnel->ip6rd.relay_prefixlen;
 705
 706                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 707                 if (pbi1 > 0)
 708                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 709                              (32 - pbi1);
 710
 711                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 712                 return true;
 713         }
 714 #else
 715         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 716                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 717                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 718                 return true;
 719         }
 720 #endif
 721         return false;
 722 }
 723
 724 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 725                              const struct in6_addr *v6dst)
 726 {
 727         __be32 dst = 0;
 728         check_6rd(tunnel, v6dst, &dst);
 729         return dst;
 730 }
 731
 732 /*
 733  *      This function assumes it is being called from dev_queue_xmit()
 734  *      and that skb is filled properly by that function.
 735  */
 736
 737 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 738                                      struct net_device *dev)
 739 {
 740         struct ip_tunnel *tunnel = netdev_priv(dev);
 741         const struct iphdr  *tiph = &tunnel->parms.iph;
 742         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 743         u8     tos = tunnel->parms.iph.tos;
 744         __be16 df = tiph->frag_off;
 745         struct rtable *rt;                      /* Route to the other host */
 746         struct net_device *tdev;                /* Device to other host */
 747         struct iphdr  *iph;                     /* Our new IP header */
 748         unsigned int max_headroom;              /* The extra header space needed */
 749         __be32 dst = tiph->daddr;
 750         struct flowi4 fl4;
 751         int    mtu;
 752         const struct in6_addr *addr6;
 753         int addr_type;
 754
 755         if (skb->protocol != htons(ETH_P_IPV6))
 756                 goto tx_error;
 757
 758         if (tos == 1)
 759                 tos = ipv6_get_dsfield(iph6);
 760
 761         /* ISATAP (RFC4214) - must come before 6to4 */
 762         if (dev->priv_flags & IFF_ISATAP) {
 763                 struct neighbour *neigh = NULL;
 764                 bool do_tx_error = false;
 765
 766                 if (skb_dst(skb))
 767                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 768
 769                 if (neigh == NULL) {
 770                         net_dbg_ratelimited("nexthop == NULL\n");
 771                         goto tx_error;
 772                 }
 773
 774                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 775                 addr_type = ipv6_addr_type(addr6);
 776
 777                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 778                      ipv6_addr_is_isatap(addr6))
 779                         dst = addr6->s6_addr32[3];
 780                 else
 781                         do_tx_error = true;
 782
 783                 neigh_release(neigh);
 784                 if (do_tx_error)
 785                         goto tx_error;
 786         }
 787
 788         if (!dst)
 789                 dst = try_6rd(tunnel, &iph6->daddr);
 790
 791         if (!dst) {
 792                 struct neighbour *neigh = NULL;
 793                 bool do_tx_error = false;
 794
 795                 if (skb_dst(skb))
 796                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 797
 798                 if (neigh == NULL) {
 799                         net_dbg_ratelimited("nexthop == NULL\n");
 800                         goto tx_error;
 801                 }
 802
 803                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 804                 addr_type = ipv6_addr_type(addr6);
 805
 806                 if (addr_type == IPV6_ADDR_ANY) {
 807                         addr6 = &ipv6_hdr(skb)->daddr;
 808                         addr_type = ipv6_addr_type(addr6);
 809                 }
 810
 811                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 812                         dst = addr6->s6_addr32[3];
 813                 else
 814                         do_tx_error = true;
 815
 816                 neigh_release(neigh);
 817                 if (do_tx_error)
 818                         goto tx_error;
 819         }
 820
 821         rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 822                                    dst, tiph->saddr,
 823                                    0, 0,
 824                                    IPPROTO_IPV6, RT_TOS(tos),
 825                                    tunnel->parms.link);
 826         if (IS_ERR(rt)) {
 827                 dev->stats.tx_carrier_errors++;
 828                 goto tx_error_icmp;
 829         }
 830         if (rt->rt_type != RTN_UNICAST) {
 831                 ip_rt_put(rt);
 832                 dev->stats.tx_carrier_errors++;
 833                 goto tx_error_icmp;
 834         }
 835         tdev = rt->dst.dev;
 836
 837         if (tdev == dev) {
 838                 ip_rt_put(rt);
 839                 dev->stats.collisions++;
 840                 goto tx_error;
 841         }
 842
 843         if (df) {
 844                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 845
 846                 if (mtu < 68) {
 847                         dev->stats.collisions++;
 848                         ip_rt_put(rt);
 849                         goto tx_error;
 850                 }
 851
 852                 if (mtu < IPV6_MIN_MTU) {
 853                         mtu = IPV6_MIN_MTU;
 854                         df = 0;
 855                 }
 856
 857                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 858                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 859
 860                 if (skb->len > mtu) {
 861                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 862                         ip_rt_put(rt);
 863                         goto tx_error;
 864                 }
 865         }
 866
 867         if (tunnel->err_count > 0) {
 868                 if (time_before(jiffies,
 869                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 870                         tunnel->err_count--;
 871                         dst_link_failure(skb);
 872                 } else
 873                         tunnel->err_count = 0;
 874         }
 875
 876         /*
 877          * Okay, now see if we can stuff it in the buffer as-is.
 878          */
 879         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 880
 881         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 882             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 883                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 884                 if (!new_skb) {
 885                         ip_rt_put(rt);
 886                         dev->stats.tx_dropped++;
 887                         dev_kfree_skb(skb);
 888                         return NETDEV_TX_OK;
 889                 }
 890                 if (skb->sk)
 891                         skb_set_owner_w(new_skb, skb->sk);
 892                 dev_kfree_skb(skb);
 893                 skb = new_skb;
 894                 iph6 = ipv6_hdr(skb);
 895         }
 896
 897         skb->transport_header = skb->network_header;
 898         skb_push(skb, sizeof(struct iphdr));
 899         skb_reset_network_header(skb);
 900         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 901         IPCB(skb)->flags = 0;
 902         skb_dst_drop(skb);
 903         skb_dst_set(skb, &rt->dst);
 904
 905         /*
 906          *      Push down and install the IPIP header.
 907          */
 908
 909         iph                     =       ip_hdr(skb);
 910         iph->version            =       4;
 911         iph->ihl                =       sizeof(struct iphdr)>>2;
 912         iph->frag_off           =       df;
 913         iph->protocol           =       IPPROTO_IPV6;
 914         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 915         iph->daddr              =       fl4.daddr;
 916         iph->saddr              =       fl4.saddr;
 917
 918         if ((iph->ttl = tiph->ttl) == 0)
 919                 iph->ttl        =       iph6->hop_limit;
 920
 921         skb->ip_summed = CHECKSUM_NONE;
 922         ip_select_ident(skb, skb_dst(skb), NULL);
 923         iptunnel_xmit(skb, dev);
 924         return NETDEV_TX_OK;
 925
 926 tx_error_icmp:
 927         dst_link_failure(skb);
 928 tx_error:
 929         dev->stats.tx_errors++;
 930         dev_kfree_skb(skb);
 931         return NETDEV_TX_OK;
 932 }
 933
 934 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 935 {
 936         struct net_device *tdev = NULL;
 937         struct ip_tunnel *tunnel;
 938         const struct iphdr *iph;
 939         struct flowi4 fl4;
 940
 941         tunnel = netdev_priv(dev);
 942         iph = &tunnel->parms.iph;
 943
 944         if (iph->daddr) {
 945                 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 946                                                           iph->daddr, iph->saddr,
 947                                                           0, 0,
 948                                                           IPPROTO_IPV6,
 949                                                           RT_TOS(iph->tos),
 950                                                           tunnel->parms.link);
 951
 952                 if (!IS_ERR(rt)) {
 953                         tdev = rt->dst.dev;
 954                         ip_rt_put(rt);
 955                 }
 956                 dev->flags |= IFF_POINTOPOINT;
 957         }
 958
 959         if (!tdev && tunnel->parms.link)
 960                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 961
 962         if (tdev) {
 963                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 964                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 965                 if (dev->mtu < IPV6_MIN_MTU)
 966                         dev->mtu = IPV6_MIN_MTU;
 967         }
 968         dev->iflink = tunnel->parms.link;
 969 }
 970
 971 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 972 {
 973         struct net *net = dev_net(t->dev);
 974         struct sit_net *sitn = net_generic(net, sit_net_id);
 975
 976         ipip6_tunnel_unlink(sitn, t);
 977         synchronize_net();
 978         t->parms.iph.saddr = p->iph.saddr;
 979         t->parms.iph.daddr = p->iph.daddr;
 980         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 981         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 982         ipip6_tunnel_link(sitn, t);
 983         t->parms.iph.ttl = p->iph.ttl;
 984         t->parms.iph.tos = p->iph.tos;
 985         if (t->parms.link != p->link) {
 986                 t->parms.link = p->link;
 987                 ipip6_tunnel_bind_dev(t->dev);
 988         }
 989         netdev_state_change(t->dev);
 990 }
 991
 992 #ifdef CONFIG_IPV6_SIT_6RD
 993 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
 994                                    struct ip_tunnel_6rd *ip6rd)
 995 {
 996         struct in6_addr prefix;
 997         __be32 relay_prefix;
 998
 999         if (ip6rd->relay_prefixlen > 32 ||
1000             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1001                 return -EINVAL;
1002
1003         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1004         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1005                 return -EINVAL;
1006         if (ip6rd->relay_prefixlen)
1007                 relay_prefix = ip6rd->relay_prefix &
1008                                htonl(0xffffffffUL <<
1009                                      (32 - ip6rd->relay_prefixlen));
1010         else
1011                 relay_prefix = 0;
1012         if (relay_prefix != ip6rd->relay_prefix)
1013                 return -EINVAL;
1014
1015         t->ip6rd.prefix = prefix;
1016         t->ip6rd.relay_prefix = relay_prefix;
1017         t->ip6rd.prefixlen = ip6rd->prefixlen;
1018         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1019         netdev_state_change(t->dev);
1020         return 0;
1021 }
1022 #endif
1023
1024 static int
1025 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1026 {
1027         int err = 0;
1028         struct ip_tunnel_parm p;
1029         struct ip_tunnel_prl prl;
1030         struct ip_tunnel *t;
1031         struct net *net = dev_net(dev);
1032         struct sit_net *sitn = net_generic(net, sit_net_id);
1033 #ifdef CONFIG_IPV6_SIT_6RD
1034         struct ip_tunnel_6rd ip6rd;
1035 #endif
1036
1037         switch (cmd) {
1038         case SIOCGETTUNNEL:
1039 #ifdef CONFIG_IPV6_SIT_6RD
1040         case SIOCGET6RD:
1041 #endif
1042                 t = NULL;
1043                 if (dev == sitn->fb_tunnel_dev) {
1044                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1045                                 err = -EFAULT;
1046                                 break;
1047                         }
1048                         t = ipip6_tunnel_locate(net, &p, 0);
1049                 }
1050                 if (t == NULL)
1051                         t = netdev_priv(dev);
1052
1053                 err = -EFAULT;
1054                 if (cmd == SIOCGETTUNNEL) {
1055                         memcpy(&p, &t->parms, sizeof(p));
1056                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1057                                          sizeof(p)))
1058                                 goto done;
1059 #ifdef CONFIG_IPV6_SIT_6RD
1060                 } else {
1061                         ip6rd.prefix = t->ip6rd.prefix;
1062                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1063                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1064                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1065                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1066                                          sizeof(ip6rd)))
1067                                 goto done;
1068 #endif
1069                 }
1070                 err = 0;
1071                 break;
1072
1073         case SIOCADDTUNNEL:
1074         case SIOCCHGTUNNEL:
1075                 err = -EPERM;
1076                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1077                         goto done;
1078
1079                 err = -EFAULT;
1080                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1081                         goto done;
1082
1083                 err = -EINVAL;
1084                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1085                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1086                         goto done;
1087                 if (p.iph.ttl)
1088                         p.iph.frag_off |= htons(IP_DF);
1089
1090                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1091
1092                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1093                         if (t != NULL) {
1094                                 if (t->dev != dev) {
1095                                         err = -EEXIST;
1096                                         break;
1097                                 }
1098                         } else {
1099                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1100                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1101                                         err = -EINVAL;
1102                                         break;
1103                                 }
1104                                 t = netdev_priv(dev);
1105                         }
1106
1107                         ipip6_tunnel_update(t, &p);
1108                 }
1109
1110                 if (t) {
1111                         err = 0;
1112                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1113                                 err = -EFAULT;
1114                 } else
1115                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1116                 break;
1117
1118         case SIOCDELTUNNEL:
1119                 err = -EPERM;
1120                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1121                         goto done;
1122
1123                 if (dev == sitn->fb_tunnel_dev) {
1124                         err = -EFAULT;
1125                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1126                                 goto done;
1127                         err = -ENOENT;
1128                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1129                                 goto done;
1130                         err = -EPERM;
1131                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1132                                 goto done;
1133                         dev = t->dev;
1134                 }
1135                 unregister_netdevice(dev);
1136                 err = 0;
1137                 break;
1138
1139         case SIOCGETPRL:
1140                 err = -EINVAL;
1141                 if (dev == sitn->fb_tunnel_dev)
1142                         goto done;
1143                 err = -ENOENT;
1144                 if (!(t = netdev_priv(dev)))
1145                         goto done;
1146                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1147                 break;
1148
1149         case SIOCADDPRL:
1150         case SIOCDELPRL:
1151         case SIOCCHGPRL:
1152                 err = -EPERM;
1153                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1154                         goto done;
1155                 err = -EINVAL;
1156                 if (dev == sitn->fb_tunnel_dev)
1157                         goto done;
1158                 err = -EFAULT;
1159                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1160                         goto done;
1161                 err = -ENOENT;
1162                 if (!(t = netdev_priv(dev)))
1163                         goto done;
1164
1165                 switch (cmd) {
1166                 case SIOCDELPRL:
1167                         err = ipip6_tunnel_del_prl(t, &prl);
1168                         break;
1169                 case SIOCADDPRL:
1170                 case SIOCCHGPRL:
1171                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1172                         break;
1173                 }
1174                 netdev_state_change(dev);
1175                 break;
1176
1177 #ifdef CONFIG_IPV6_SIT_6RD
1178         case SIOCADD6RD:
1179         case SIOCCHG6RD:
1180         case SIOCDEL6RD:
1181                 err = -EPERM;
1182                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1183                         goto done;
1184
1185                 err = -EFAULT;
1186                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1187                                    sizeof(ip6rd)))
1188                         goto done;
1189
1190                 t = netdev_priv(dev);
1191
1192                 if (cmd != SIOCDEL6RD) {
1193                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1194                         if (err < 0)
1195                                 goto done;
1196                 } else
1197                         ipip6_tunnel_clone_6rd(dev, sitn);
1198
1199                 err = 0;
1200                 break;
1201 #endif
1202
1203         default:
1204                 err = -EINVAL;
1205         }
1206
1207 done:
1208         return err;
1209 }
1210
1211 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1212 {
1213         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1214                 return -EINVAL;
1215         dev->mtu = new_mtu;
1216         return 0;
1217 }
1218
1219 static const struct net_device_ops ipip6_netdev_ops = {
1220         .ndo_uninit     = ipip6_tunnel_uninit,
1221         .ndo_start_xmit = ipip6_tunnel_xmit,
1222         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1223         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1224         .ndo_get_stats64 = ip_tunnel_get_stats64,
1225 };
1226
1227 static void ipip6_dev_free(struct net_device *dev)
1228 {
1229         free_percpu(dev->tstats);
1230         free_netdev(dev);
1231 }
1232
1233 static void ipip6_tunnel_setup(struct net_device *dev)
1234 {
1235         dev->netdev_ops         = &ipip6_netdev_ops;
1236         dev->destructor         = ipip6_dev_free;
1237
1238         dev->type               = ARPHRD_SIT;
1239         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1240         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1241         dev->flags              = IFF_NOARP;
1242         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1243         dev->iflink             = 0;
1244         dev->addr_len           = 4;
1245         dev->features           |= NETIF_F_NETNS_LOCAL;
1246         dev->features           |= NETIF_F_LLTX;
1247 }
1248
1249 static int ipip6_tunnel_init(struct net_device *dev)
1250 {
1251         struct ip_tunnel *tunnel = netdev_priv(dev);
1252
1253         tunnel->dev = dev;
1254
1255         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1256         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1257
1258         ipip6_tunnel_bind_dev(dev);
1259         dev->tstats = alloc_percpu(struct pcpu_tstats);
1260         if (!dev->tstats)
1261                 return -ENOMEM;
1262
1263         return 0;
1264 }
1265
1266 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1267 {
1268         struct ip_tunnel *tunnel = netdev_priv(dev);
1269         struct iphdr *iph = &tunnel->parms.iph;
1270         struct net *net = dev_net(dev);
1271         struct sit_net *sitn = net_generic(net, sit_net_id);
1272
1273         tunnel->dev = dev;
1274         strcpy(tunnel->parms.name, dev->name);
1275
1276         iph->version            = 4;
1277         iph->protocol           = IPPROTO_IPV6;
1278         iph->ihl                = 5;
1279         iph->ttl                = 64;
1280
1281         dev->tstats = alloc_percpu(struct pcpu_tstats);
1282         if (!dev->tstats)
1283                 return -ENOMEM;
1284         dev_hold(dev);
1285         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1286         return 0;
1287 }
1288
1289 static void ipip6_netlink_parms(struct nlattr *data[],
1290                                 struct ip_tunnel_parm *parms)
1291 {
1292         memset(parms, 0, sizeof(*parms));
1293
1294         parms->iph.version = 4;
1295         parms->iph.protocol = IPPROTO_IPV6;
1296         parms->iph.ihl = 5;
1297         parms->iph.ttl = 64;
1298
1299         if (!data)
1300                 return;
1301
1302         if (data[IFLA_IPTUN_LINK])
1303                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1304
1305         if (data[IFLA_IPTUN_LOCAL])
1306                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1307
1308         if (data[IFLA_IPTUN_REMOTE])
1309                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1310
1311         if (data[IFLA_IPTUN_TTL]) {
1312                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1313                 if (parms->iph.ttl)
1314                         parms->iph.frag_off = htons(IP_DF);
1315         }
1316
1317         if (data[IFLA_IPTUN_TOS])
1318                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1319
1320         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1321                 parms->iph.frag_off = htons(IP_DF);
1322
1323         if (data[IFLA_IPTUN_FLAGS])
1324                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1325 }
1326
1327 #ifdef CONFIG_IPV6_SIT_6RD
1328 /* This function returns true when 6RD attributes are present in the nl msg */
1329 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1330                                     struct ip_tunnel_6rd *ip6rd)
1331 {
1332         bool ret = false;
1333         memset(ip6rd, 0, sizeof(*ip6rd));
1334
1335         if (!data)
1336                 return ret;
1337
1338         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1339                 ret = true;
1340                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1341                            sizeof(struct in6_addr));
1342         }
1343
1344         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1345                 ret = true;
1346                 ip6rd->relay_prefix =
1347                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1348         }
1349
1350         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1351                 ret = true;
1352                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1353         }
1354
1355         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1356                 ret = true;
1357                 ip6rd->relay_prefixlen =
1358                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1359         }
1360
1361         return ret;
1362 }
1363 #endif
1364
1365 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1366                          struct nlattr *tb[], struct nlattr *data[])
1367 {
1368         struct net *net = dev_net(dev);
1369         struct ip_tunnel *nt;
1370 #ifdef CONFIG_IPV6_SIT_6RD
1371         struct ip_tunnel_6rd ip6rd;
1372 #endif
1373         int err;
1374
1375         nt = netdev_priv(dev);
1376         ipip6_netlink_parms(data, &nt->parms);
1377
1378         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1379                 return -EEXIST;
1380
1381         err = ipip6_tunnel_create(dev);
1382         if (err < 0)
1383                 return err;
1384
1385 #ifdef CONFIG_IPV6_SIT_6RD
1386         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1387                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1388 #endif
1389
1390         return err;
1391 }
1392
1393 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1394                           struct nlattr *data[])
1395 {
1396         struct ip_tunnel *t;
1397         struct ip_tunnel_parm p;
1398         struct net *net = dev_net(dev);
1399         struct sit_net *sitn = net_generic(net, sit_net_id);
1400 #ifdef CONFIG_IPV6_SIT_6RD
1401         struct ip_tunnel_6rd ip6rd;
1402 #endif
1403
1404         if (dev == sitn->fb_tunnel_dev)
1405                 return -EINVAL;
1406
1407         ipip6_netlink_parms(data, &p);
1408
1409         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1410             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1411                 return -EINVAL;
1412
1413         t = ipip6_tunnel_locate(net, &p, 0);
1414
1415         if (t) {
1416                 if (t->dev != dev)
1417                         return -EEXIST;
1418         } else
1419                 t = netdev_priv(dev);
1420
1421         ipip6_tunnel_update(t, &p);
1422
1423 #ifdef CONFIG_IPV6_SIT_6RD
1424         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1425                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1426 #endif
1427
1428         return 0;
1429 }
1430
1431 static size_t ipip6_get_size(const struct net_device *dev)
1432 {
1433         return
1434                 /* IFLA_IPTUN_LINK */
1435                 nla_total_size(4) +
1436                 /* IFLA_IPTUN_LOCAL */
1437                 nla_total_size(4) +
1438                 /* IFLA_IPTUN_REMOTE */
1439                 nla_total_size(4) +
1440                 /* IFLA_IPTUN_TTL */
1441                 nla_total_size(1) +
1442                 /* IFLA_IPTUN_TOS */
1443                 nla_total_size(1) +
1444                 /* IFLA_IPTUN_PMTUDISC */
1445                 nla_total_size(1) +
1446                 /* IFLA_IPTUN_FLAGS */
1447                 nla_total_size(2) +
1448 #ifdef CONFIG_IPV6_SIT_6RD
1449                 /* IFLA_IPTUN_6RD_PREFIX */
1450                 nla_total_size(sizeof(struct in6_addr)) +
1451                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1452                 nla_total_size(4) +
1453                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1454                 nla_total_size(2) +
1455                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1456                 nla_total_size(2) +
1457 #endif
1458                 0;
1459 }
1460
1461 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1462 {
1463         struct ip_tunnel *tunnel = netdev_priv(dev);
1464         struct ip_tunnel_parm *parm = &tunnel->parms;
1465
1466         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1467             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1468             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1469             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1470             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1471             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1472                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1473             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1474                 goto nla_put_failure;
1475
1476 #ifdef CONFIG_IPV6_SIT_6RD
1477         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1478                     &tunnel->ip6rd.prefix) ||
1479             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1480                          tunnel->ip6rd.relay_prefix) ||
1481             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1482                         tunnel->ip6rd.prefixlen) ||
1483             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1484                         tunnel->ip6rd.relay_prefixlen))
1485                 goto nla_put_failure;
1486 #endif
1487
1488         return 0;
1489
1490 nla_put_failure:
1491         return -EMSGSIZE;
1492 }
1493
1494 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1495         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1496         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1497         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1498         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1499         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1500         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1501         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1502 #ifdef CONFIG_IPV6_SIT_6RD
1503         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1504         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1505         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1506         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1507 #endif
1508 };
1509
1510 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1511 {
1512         struct net *net = dev_net(dev);
1513         struct sit_net *sitn = net_generic(net, sit_net_id);
1514
1515         if (dev != sitn->fb_tunnel_dev)
1516                 unregister_netdevice_queue(dev, head);
1517 }
1518
1519 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1520         .kind           = "sit",
1521         .maxtype        = IFLA_IPTUN_MAX,
1522         .policy         = ipip6_policy,
1523         .priv_size      = sizeof(struct ip_tunnel),
1524         .setup          = ipip6_tunnel_setup,
1525         .newlink        = ipip6_newlink,
1526         .changelink     = ipip6_changelink,
1527         .get_size       = ipip6_get_size,
1528         .fill_info      = ipip6_fill_info,
1529         .dellink        = ipip6_dellink,
1530 };
1531
1532 static struct xfrm_tunnel sit_handler __read_mostly = {
1533         .handler        =       ipip6_rcv,
1534         .err_handler    =       ipip6_err,
1535         .priority       =       1,
1536 };
1537
1538 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1539 {
1540         int prio;
1541
1542         for (prio = 1; prio < 4; prio++) {
1543                 int h;
1544                 for (h = 0; h < HASH_SIZE; h++) {
1545                         struct ip_tunnel *t;
1546
1547                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1548                         while (t != NULL) {
1549                                 unregister_netdevice_queue(t->dev, head);
1550                                 t = rtnl_dereference(t->next);
1551                         }
1552                 }
1553         }
1554 }
1555
1556 static int __net_init sit_init_net(struct net *net)
1557 {
1558         struct sit_net *sitn = net_generic(net, sit_net_id);
1559         struct ip_tunnel *t;
1560         int err;
1561
1562         sitn->tunnels[0] = sitn->tunnels_wc;
1563         sitn->tunnels[1] = sitn->tunnels_l;
1564         sitn->tunnels[2] = sitn->tunnels_r;
1565         sitn->tunnels[3] = sitn->tunnels_r_l;
1566
1567         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1568                                            ipip6_tunnel_setup);
1569         if (!sitn->fb_tunnel_dev) {
1570                 err = -ENOMEM;
1571                 goto err_alloc_dev;
1572         }
1573         dev_net_set(sitn->fb_tunnel_dev, net);
1574         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1575
1576         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1577         if (err)
1578                 goto err_dev_free;
1579
1580         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1581
1582         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1583                 goto err_reg_dev;
1584
1585         t = netdev_priv(sitn->fb_tunnel_dev);
1586
1587         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1588         return 0;
1589
1590 err_reg_dev:
1591         dev_put(sitn->fb_tunnel_dev);
1592 err_dev_free:
1593         ipip6_dev_free(sitn->fb_tunnel_dev);
1594 err_alloc_dev:
1595         return err;
1596 }
1597
1598 static void __net_exit sit_exit_net(struct net *net)
1599 {
1600         struct sit_net *sitn = net_generic(net, sit_net_id);
1601         LIST_HEAD(list);
1602
1603         rtnl_lock();
1604         sit_destroy_tunnels(sitn, &list);
1605         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1606         unregister_netdevice_many(&list);
1607         rtnl_unlock();
1608 }
1609
1610 static struct pernet_operations sit_net_ops = {
1611         .init = sit_init_net,
1612         .exit = sit_exit_net,
1613         .id   = &sit_net_id,
1614         .size = sizeof(struct sit_net),
1615 };
1616
1617 static void __exit sit_cleanup(void)
1618 {
1619         rtnl_link_unregister(&sit_link_ops);
1620         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1621
1622         unregister_pernet_device(&sit_net_ops);
1623         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1624 }
1625
1626 static int __init sit_init(void)
1627 {
1628         int err;
1629
1630         pr_info("IPv6 over IPv4 tunneling driver\n");
1631
1632         err = register_pernet_device(&sit_net_ops);
1633         if (err < 0)
1634                 return err;
1635         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1636         if (err < 0) {
1637                 pr_info("%s: can't add protocol\n", __func__);
1638                 goto xfrm_tunnel_failed;
1639         }
1640         err = rtnl_link_register(&sit_link_ops);
1641         if (err < 0)
1642                 goto rtnl_link_failed;
1643
1644 out:
1645         return err;
1646
1647 rtnl_link_failed:
1648         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1649 xfrm_tunnel_failed:
1650         unregister_pernet_device(&sit_net_ops);
1651         goto out;
1652 }
1653
1654 module_init(sit_init);
1655 module_exit(sit_cleanup);
1656 MODULE_LICENSE("GPL");
1657 MODULE_ALIAS_NETDEV("sit0");