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             <[email protected]>
   7  *      Alexey Kuznetsov        <[email protected]>
   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 <[email protected]>:       6to4 support
  16  * Nate Thompson <[email protected]>:             6to4 support
  17  * Fred Templin <[email protected]>:    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/ipip.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 int ipip6_tunnel_init(struct net_device *dev);
  69 static void ipip6_tunnel_setup(struct net_device *dev);
  70 static void ipip6_dev_free(struct net_device *dev);
  71 static struct rtnl_link_ops sit_link_ops __read_mostly;
  72
  73 static int sit_net_id __read_mostly;
  74 struct sit_net {
  75         struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
  76         struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
  77         struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
  78         struct ip_tunnel __rcu *tunnels_wc[1];
  79         struct ip_tunnel __rcu **tunnels[4];
  80
  81         struct net_device *fb_tunnel_dev;
  82 };
  83
  84 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
  85                                                    struct rtnl_link_stats64 *tot)
  86 {
  87         int i;
  88
  89         for_each_possible_cpu(i) {
  90                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
  91                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
  92                 unsigned int start;
  93
  94                 do {
  95                         start = u64_stats_fetch_begin_bh(&tstats->syncp);
  96                         rx_packets = tstats->rx_packets;
  97                         tx_packets = tstats->tx_packets;
  98                         rx_bytes = tstats->rx_bytes;
  99                         tx_bytes = tstats->tx_bytes;
 100                 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
 101
 102                 tot->rx_packets += rx_packets;
 103                 tot->tx_packets += tx_packets;
 104                 tot->rx_bytes   += rx_bytes;
 105                 tot->tx_bytes   += tx_bytes;
 106         }
 107
 108         tot->rx_errors = dev->stats.rx_errors;
 109         tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
 110         tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
 111         tot->tx_dropped = dev->stats.tx_dropped;
 112         tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
 113         tot->tx_errors = dev->stats.tx_errors;
 114
 115         return tot;
 116 }
 117
 118 /*
 119  * Must be invoked with rcu_read_lock
 120  */
 121 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
 122                 struct net_device *dev, __be32 remote, __be32 local)
 123 {
 124         unsigned int h0 = HASH(remote);
 125         unsigned int h1 = HASH(local);
 126         struct ip_tunnel *t;
 127         struct sit_net *sitn = net_generic(net, sit_net_id);
 128
 129         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 130                 if (local == t->parms.iph.saddr &&
 131                     remote == t->parms.iph.daddr &&
 132                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 133                     (t->dev->flags & IFF_UP))
 134                         return t;
 135         }
 136         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 137                 if (remote == t->parms.iph.daddr &&
 138                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 139                     (t->dev->flags & IFF_UP))
 140                         return t;
 141         }
 142         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 143                 if (local == t->parms.iph.saddr &&
 144                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 145                     (t->dev->flags & IFF_UP))
 146                         return t;
 147         }
 148         t = rcu_dereference(sitn->tunnels_wc[0]);
 149         if ((t != NULL) && (t->dev->flags & IFF_UP))
 150                 return t;
 151         return NULL;
 152 }
 153
 154 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 155                 struct ip_tunnel_parm *parms)
 156 {
 157         __be32 remote = parms->iph.daddr;
 158         __be32 local = parms->iph.saddr;
 159         unsigned int h = 0;
 160         int prio = 0;
 161
 162         if (remote) {
 163                 prio |= 2;
 164                 h ^= HASH(remote);
 165         }
 166         if (local) {
 167                 prio |= 1;
 168                 h ^= HASH(local);
 169         }
 170         return &sitn->tunnels[prio][h];
 171 }
 172
 173 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 174                 struct ip_tunnel *t)
 175 {
 176         return __ipip6_bucket(sitn, &t->parms);
 177 }
 178
 179 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 180 {
 181         struct ip_tunnel __rcu **tp;
 182         struct ip_tunnel *iter;
 183
 184         for (tp = ipip6_bucket(sitn, t);
 185              (iter = rtnl_dereference(*tp)) != NULL;
 186              tp = &iter->next) {
 187                 if (t == iter) {
 188                         rcu_assign_pointer(*tp, t->next);
 189                         break;
 190                 }
 191         }
 192 }
 193
 194 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 195 {
 196         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 197
 198         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 199         rcu_assign_pointer(*tp, t);
 200 }
 201
 202 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 203 {
 204 #ifdef CONFIG_IPV6_SIT_6RD
 205         struct ip_tunnel *t = netdev_priv(dev);
 206
 207         if (t->dev == sitn->fb_tunnel_dev) {
 208                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 209                 t->ip6rd.relay_prefix = 0;
 210                 t->ip6rd.prefixlen = 16;
 211                 t->ip6rd.relay_prefixlen = 0;
 212         } else {
 213                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 214                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 215         }
 216 #endif
 217 }
 218
 219 static int ipip6_tunnel_create(struct net_device *dev)
 220 {
 221         struct ip_tunnel *t = netdev_priv(dev);
 222         struct net *net = dev_net(dev);
 223         struct sit_net *sitn = net_generic(net, sit_net_id);
 224         int err;
 225
 226         err = ipip6_tunnel_init(dev);
 227         if (err < 0)
 228                 goto out;
 229         ipip6_tunnel_clone_6rd(dev, sitn);
 230
 231         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 232                 dev->priv_flags |= IFF_ISATAP;
 233
 234         err = register_netdevice(dev);
 235         if (err < 0)
 236                 goto out;
 237
 238         strcpy(t->parms.name, dev->name);
 239         dev->rtnl_link_ops = &sit_link_ops;
 240
 241         dev_hold(dev);
 242
 243         ipip6_tunnel_link(sitn, t);
 244         return 0;
 245
 246 out:
 247         return err;
 248 }
 249
 250 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 251                 struct ip_tunnel_parm *parms, int create)
 252 {
 253         __be32 remote = parms->iph.daddr;
 254         __be32 local = parms->iph.saddr;
 255         struct ip_tunnel *t, *nt;
 256         struct ip_tunnel __rcu **tp;
 257         struct net_device *dev;
 258         char name[IFNAMSIZ];
 259         struct sit_net *sitn = net_generic(net, sit_net_id);
 260
 261         for (tp = __ipip6_bucket(sitn, parms);
 262             (t = rtnl_dereference(*tp)) != NULL;
 263              tp = &t->next) {
 264                 if (local == t->parms.iph.saddr &&
 265                     remote == t->parms.iph.daddr &&
 266                     parms->link == t->parms.link) {
 267                         if (create)
 268                                 return NULL;
 269                         else
 270                                 return t;
 271                 }
 272         }
 273         if (!create)
 274                 goto failed;
 275
 276         if (parms->name[0])
 277                 strlcpy(name, parms->name, IFNAMSIZ);
 278         else
 279                 strcpy(name, "sit%d");
 280
 281         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 282         if (dev == NULL)
 283                 return NULL;
 284
 285         dev_net_set(dev, net);
 286
 287         nt = netdev_priv(dev);
 288
 289         nt->parms = *parms;
 290         if (ipip6_tunnel_create(dev) < 0)
 291                 goto failed_free;
 292
 293         return nt;
 294
 295 failed_free:
 296         ipip6_dev_free(dev);
 297 failed:
 298         return NULL;
 299 }
 300
 301 #define for_each_prl_rcu(start)                 \
 302         for (prl = rcu_dereference(start);      \
 303              prl;                               \
 304              prl = rcu_dereference(prl->next))
 305
 306 static struct ip_tunnel_prl_entry *
 307 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 308 {
 309         struct ip_tunnel_prl_entry *prl;
 310
 311         for_each_prl_rcu(t->prl)
 312                 if (prl->addr == addr)
 313                         break;
 314         return prl;
 315
 316 }
 317
 318 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 319                                 struct ip_tunnel_prl __user *a)
 320 {
 321         struct ip_tunnel_prl kprl, *kp;
 322         struct ip_tunnel_prl_entry *prl;
 323         unsigned int cmax, c = 0, ca, len;
 324         int ret = 0;
 325
 326         if (copy_from_user(&kprl, a, sizeof(kprl)))
 327                 return -EFAULT;
 328         cmax = kprl.datalen / sizeof(kprl);
 329         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 330                 cmax = 1;
 331
 332         /* For simple GET or for root users,
 333          * we try harder to allocate.
 334          */
 335         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 336                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 337                 NULL;
 338
 339         rcu_read_lock();
 340
 341         ca = t->prl_count < cmax ? t->prl_count : cmax;
 342
 343         if (!kp) {
 344                 /* We don't try hard to allocate much memory for
 345                  * non-root users.
 346                  * For root users, retry allocating enough memory for
 347                  * the answer.
 348                  */
 349                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 350                 if (!kp) {
 351                         ret = -ENOMEM;
 352                         goto out;
 353                 }
 354         }
 355
 356         c = 0;
 357         for_each_prl_rcu(t->prl) {
 358                 if (c >= cmax)
 359                         break;
 360                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 361                         continue;
 362                 kp[c].addr = prl->addr;
 363                 kp[c].flags = prl->flags;
 364                 c++;
 365                 if (kprl.addr != htonl(INADDR_ANY))
 366                         break;
 367         }
 368 out:
 369         rcu_read_unlock();
 370
 371         len = sizeof(*kp) * c;
 372         ret = 0;
 373         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 374                 ret = -EFAULT;
 375
 376         kfree(kp);
 377
 378         return ret;
 379 }
 380
 381 static int
 382 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 383 {
 384         struct ip_tunnel_prl_entry *p;
 385         int err = 0;
 386
 387         if (a->addr == htonl(INADDR_ANY))
 388                 return -EINVAL;
 389
 390         ASSERT_RTNL();
 391
 392         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 393                 if (p->addr == a->addr) {
 394                         if (chg) {
 395                                 p->flags = a->flags;
 396                                 goto out;
 397                         }
 398                         err = -EEXIST;
 399                         goto out;
 400                 }
 401         }
 402
 403         if (chg) {
 404                 err = -ENXIO;
 405                 goto out;
 406         }
 407
 408         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 409         if (!p) {
 410                 err = -ENOBUFS;
 411                 goto out;
 412         }
 413
 414         p->next = t->prl;
 415         p->addr = a->addr;
 416         p->flags = a->flags;
 417         t->prl_count++;
 418         rcu_assign_pointer(t->prl, p);
 419 out:
 420         return err;
 421 }
 422
 423 static void prl_list_destroy_rcu(struct rcu_head *head)
 424 {
 425         struct ip_tunnel_prl_entry *p, *n;
 426
 427         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 428         do {
 429                 n = rcu_dereference_protected(p->next, 1);
 430                 kfree(p);
 431                 p = n;
 432         } while (p);
 433 }
 434
 435 static int
 436 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 437 {
 438         struct ip_tunnel_prl_entry *x;
 439         struct ip_tunnel_prl_entry __rcu **p;
 440         int err = 0;
 441
 442         ASSERT_RTNL();
 443
 444         if (a && a->addr != htonl(INADDR_ANY)) {
 445                 for (p = &t->prl;
 446                      (x = rtnl_dereference(*p)) != NULL;
 447                      p = &x->next) {
 448                         if (x->addr == a->addr) {
 449                                 *p = x->next;
 450                                 kfree_rcu(x, rcu_head);
 451                                 t->prl_count--;
 452                                 goto out;
 453                         }
 454                 }
 455                 err = -ENXIO;
 456         } else {
 457                 x = rtnl_dereference(t->prl);
 458                 if (x) {
 459                         t->prl_count = 0;
 460                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 461                         t->prl = NULL;
 462                 }
 463         }
 464 out:
 465         return err;
 466 }
 467
 468 static int
 469 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 470 {
 471         struct ip_tunnel_prl_entry *p;
 472         int ok = 1;
 473
 474         rcu_read_lock();
 475         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 476         if (p) {
 477                 if (p->flags & PRL_DEFAULT)
 478                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 479                 else
 480                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 481         } else {
 482                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 483
 484                 if (ipv6_addr_is_isatap(addr6) &&
 485                     (addr6->s6_addr32[3] == iph->saddr) &&
 486                     ipv6_chk_prefix(addr6, t->dev))
 487                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 488                 else
 489                         ok = 0;
 490         }
 491         rcu_read_unlock();
 492         return ok;
 493 }
 494
 495 static void ipip6_tunnel_uninit(struct net_device *dev)
 496 {
 497         struct net *net = dev_net(dev);
 498         struct sit_net *sitn = net_generic(net, sit_net_id);
 499
 500         if (dev == sitn->fb_tunnel_dev) {
 501                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 502         } else {
 503                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 504                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 505         }
 506         dev_put(dev);
 507 }
 508
 509
 510 static int ipip6_err(struct sk_buff *skb, u32 info)
 511 {
 512
 513 /* All the routers (except for Linux) return only
 514    8 bytes of packet payload. It means, that precise relaying of
 515    ICMP in the real Internet is absolutely infeasible.
 516  */
 517         const struct iphdr *iph = (const struct iphdr *)skb->data;
 518         const int type = icmp_hdr(skb)->type;
 519         const int code = icmp_hdr(skb)->code;
 520         struct ip_tunnel *t;
 521         int err;
 522
 523         switch (type) {
 524         default:
 525         case ICMP_PARAMETERPROB:
 526                 return 0;
 527
 528         case ICMP_DEST_UNREACH:
 529                 switch (code) {
 530                 case ICMP_SR_FAILED:
 531                 case ICMP_PORT_UNREACH:
 532                         /* Impossible event. */
 533                         return 0;
 534                 default:
 535                         /* All others are translated to HOST_UNREACH.
 536                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 537                            I believe they are just ether pollution. --ANK
 538                          */
 539                         break;
 540                 }
 541                 break;
 542         case ICMP_TIME_EXCEEDED:
 543                 if (code != ICMP_EXC_TTL)
 544                         return 0;
 545                 break;
 546         case ICMP_REDIRECT:
 547                 break;
 548         }
 549
 550         err = -ENOENT;
 551
 552         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 553                                 skb->dev,
 554                                 iph->daddr,
 555                                 iph->saddr);
 556         if (t == NULL)
 557                 goto out;
 558
 559         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 560                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 561                                  t->dev->ifindex, 0, IPPROTO_IPV6, 0);
 562                 err = 0;
 563                 goto out;
 564         }
 565         if (type == ICMP_REDIRECT) {
 566                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 567                               IPPROTO_IPV6, 0);
 568                 err = 0;
 569                 goto out;
 570         }
 571
 572         if (t->parms.iph.daddr == 0)
 573                 goto out;
 574
 575         err = 0;
 576         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 577                 goto out;
 578
 579         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 580                 t->err_count++;
 581         else
 582                 t->err_count = 1;
 583         t->err_time = jiffies;
 584 out:
 585         return err;
 586 }
 587
 588 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
 589 {
 590         if (INET_ECN_is_ce(iph->tos))
 591                 IP6_ECN_set_ce(ipv6_hdr(skb));
 592 }
 593
 594 static int ipip6_rcv(struct sk_buff *skb)
 595 {
 596         const struct iphdr *iph;
 597         struct ip_tunnel *tunnel;
 598
 599         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 600                 goto out;
 601
 602         iph = ip_hdr(skb);
 603
 604         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 605                                      iph->saddr, iph->daddr);
 606         if (tunnel != NULL) {
 607                 struct pcpu_tstats *tstats;
 608
 609                 secpath_reset(skb);
 610                 skb->mac_header = skb->network_header;
 611                 skb_reset_network_header(skb);
 612                 IPCB(skb)->flags = 0;
 613                 skb->protocol = htons(ETH_P_IPV6);
 614                 skb->pkt_type = PACKET_HOST;
 615
 616                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 617                     !isatap_chksrc(skb, iph, tunnel)) {
 618                         tunnel->dev->stats.rx_errors++;
 619                         kfree_skb(skb);
 620                         return 0;
 621                 }
 622
 623                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 624                 tstats->rx_packets++;
 625                 tstats->rx_bytes += skb->len;
 626
 627                 __skb_tunnel_rx(skb, tunnel->dev);
 628
 629                 ipip6_ecn_decapsulate(iph, skb);
 630
 631                 netif_rx(skb);
 632
 633                 return 0;
 634         }
 635
 636         /* no tunnel matched,  let upstream know, ipsec may handle it */
 637         return 1;
 638 out:
 639         kfree_skb(skb);
 640         return 0;
 641 }
 642
 643 /*
 644  * Returns the embedded IPv4 address if the IPv6 address
 645  * comes from 6rd / 6to4 (RFC 3056) addr space.
 646  */
 647 static inline
 648 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 649 {
 650         __be32 dst = 0;
 651
 652 #ifdef CONFIG_IPV6_SIT_6RD
 653         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 654                               tunnel->ip6rd.prefixlen)) {
 655                 unsigned int pbw0, pbi0;
 656                 int pbi1;
 657                 u32 d;
 658
 659                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 660                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 661
 662                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 663                     tunnel->ip6rd.relay_prefixlen;
 664
 665                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 666                 if (pbi1 > 0)
 667                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 668                              (32 - pbi1);
 669
 670                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
 671         }
 672 #else
 673         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 674                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 675                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 676         }
 677 #endif
 678         return dst;
 679 }
 680
 681 /*
 682  *      This function assumes it is being called from dev_queue_xmit()
 683  *      and that skb is filled properly by that function.
 684  */
 685
 686 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 687                                      struct net_device *dev)
 688 {
 689         struct ip_tunnel *tunnel = netdev_priv(dev);
 690         const struct iphdr  *tiph = &tunnel->parms.iph;
 691         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 692         u8     tos = tunnel->parms.iph.tos;
 693         __be16 df = tiph->frag_off;
 694         struct rtable *rt;                      /* Route to the other host */
 695         struct net_device *tdev;                /* Device to other host */
 696         struct iphdr  *iph;                     /* Our new IP header */
 697         unsigned int max_headroom;              /* The extra header space needed */
 698         __be32 dst = tiph->daddr;
 699         struct flowi4 fl4;
 700         int    mtu;
 701         const struct in6_addr *addr6;
 702         int addr_type;
 703
 704         if (skb->protocol != htons(ETH_P_IPV6))
 705                 goto tx_error;
 706
 707         if (tos == 1)
 708                 tos = ipv6_get_dsfield(iph6);
 709
 710         /* ISATAP (RFC4214) - must come before 6to4 */
 711         if (dev->priv_flags & IFF_ISATAP) {
 712                 struct neighbour *neigh = NULL;
 713                 bool do_tx_error = false;
 714
 715                 if (skb_dst(skb))
 716                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 717
 718                 if (neigh == NULL) {
 719                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 720                         goto tx_error;
 721                 }
 722
 723                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 724                 addr_type = ipv6_addr_type(addr6);
 725
 726                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 727                      ipv6_addr_is_isatap(addr6))
 728                         dst = addr6->s6_addr32[3];
 729                 else
 730                         do_tx_error = true;
 731
 732                 neigh_release(neigh);
 733                 if (do_tx_error)
 734                         goto tx_error;
 735         }
 736
 737         if (!dst)
 738                 dst = try_6rd(&iph6->daddr, tunnel);
 739
 740         if (!dst) {
 741                 struct neighbour *neigh = NULL;
 742                 bool do_tx_error = false;
 743
 744                 if (skb_dst(skb))
 745                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 746
 747                 if (neigh == NULL) {
 748                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 749                         goto tx_error;
 750                 }
 751
 752                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 753                 addr_type = ipv6_addr_type(addr6);
 754
 755                 if (addr_type == IPV6_ADDR_ANY) {
 756                         addr6 = &ipv6_hdr(skb)->daddr;
 757                         addr_type = ipv6_addr_type(addr6);
 758                 }
 759
 760                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 761                         dst = addr6->s6_addr32[3];
 762                 else
 763                         do_tx_error = true;
 764
 765                 neigh_release(neigh);
 766                 if (do_tx_error)
 767                         goto tx_error;
 768         }
 769
 770         rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 771                                    dst, tiph->saddr,
 772                                    0, 0,
 773                                    IPPROTO_IPV6, RT_TOS(tos),
 774                                    tunnel->parms.link);
 775         if (IS_ERR(rt)) {
 776                 dev->stats.tx_carrier_errors++;
 777                 goto tx_error_icmp;
 778         }
 779         if (rt->rt_type != RTN_UNICAST) {
 780                 ip_rt_put(rt);
 781                 dev->stats.tx_carrier_errors++;
 782                 goto tx_error_icmp;
 783         }
 784         tdev = rt->dst.dev;
 785
 786         if (tdev == dev) {
 787                 ip_rt_put(rt);
 788                 dev->stats.collisions++;
 789                 goto tx_error;
 790         }
 791
 792         if (df) {
 793                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 794
 795                 if (mtu < 68) {
 796                         dev->stats.collisions++;
 797                         ip_rt_put(rt);
 798                         goto tx_error;
 799                 }
 800
 801                 if (mtu < IPV6_MIN_MTU) {
 802                         mtu = IPV6_MIN_MTU;
 803                         df = 0;
 804                 }
 805
 806                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 807                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 808
 809                 if (skb->len > mtu) {
 810                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 811                         ip_rt_put(rt);
 812                         goto tx_error;
 813                 }
 814         }
 815
 816         if (tunnel->err_count > 0) {
 817                 if (time_before(jiffies,
 818                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 819                         tunnel->err_count--;
 820                         dst_link_failure(skb);
 821                 } else
 822                         tunnel->err_count = 0;
 823         }
 824
 825         /*
 826          * Okay, now see if we can stuff it in the buffer as-is.
 827          */
 828         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 829
 830         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 831             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 832                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 833                 if (!new_skb) {
 834                         ip_rt_put(rt);
 835                         dev->stats.tx_dropped++;
 836                         dev_kfree_skb(skb);
 837                         return NETDEV_TX_OK;
 838                 }
 839                 if (skb->sk)
 840                         skb_set_owner_w(new_skb, skb->sk);
 841                 dev_kfree_skb(skb);
 842                 skb = new_skb;
 843                 iph6 = ipv6_hdr(skb);
 844         }
 845
 846         skb->transport_header = skb->network_header;
 847         skb_push(skb, sizeof(struct iphdr));
 848         skb_reset_network_header(skb);
 849         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 850         IPCB(skb)->flags = 0;
 851         skb_dst_drop(skb);
 852         skb_dst_set(skb, &rt->dst);
 853
 854         /*
 855          *      Push down and install the IPIP header.
 856          */
 857
 858         iph                     =       ip_hdr(skb);
 859         iph->version            =       4;
 860         iph->ihl                =       sizeof(struct iphdr)>>2;
 861         iph->frag_off           =       df;
 862         iph->protocol           =       IPPROTO_IPV6;
 863         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 864         iph->daddr              =       fl4.daddr;
 865         iph->saddr              =       fl4.saddr;
 866
 867         if ((iph->ttl = tiph->ttl) == 0)
 868                 iph->ttl        =       iph6->hop_limit;
 869
 870         iptunnel_xmit(skb, dev);
 871         return NETDEV_TX_OK;
 872
 873 tx_error_icmp:
 874         dst_link_failure(skb);
 875 tx_error:
 876         dev->stats.tx_errors++;
 877         dev_kfree_skb(skb);
 878         return NETDEV_TX_OK;
 879 }
 880
 881 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 882 {
 883         struct net_device *tdev = NULL;
 884         struct ip_tunnel *tunnel;
 885         const struct iphdr *iph;
 886         struct flowi4 fl4;
 887
 888         tunnel = netdev_priv(dev);
 889         iph = &tunnel->parms.iph;
 890
 891         if (iph->daddr) {
 892                 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 893                                                           iph->daddr, iph->saddr,
 894                                                           0, 0,
 895                                                           IPPROTO_IPV6,
 896                                                           RT_TOS(iph->tos),
 897                                                           tunnel->parms.link);
 898
 899                 if (!IS_ERR(rt)) {
 900                         tdev = rt->dst.dev;
 901                         ip_rt_put(rt);
 902                 }
 903                 dev->flags |= IFF_POINTOPOINT;
 904         }
 905
 906         if (!tdev && tunnel->parms.link)
 907                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 908
 909         if (tdev) {
 910                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 911                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 912                 if (dev->mtu < IPV6_MIN_MTU)
 913                         dev->mtu = IPV6_MIN_MTU;
 914         }
 915         dev->iflink = tunnel->parms.link;
 916 }
 917
 918 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 919 {
 920         struct net *net = dev_net(t->dev);
 921         struct sit_net *sitn = net_generic(net, sit_net_id);
 922
 923         ipip6_tunnel_unlink(sitn, t);
 924         synchronize_net();
 925         t->parms.iph.saddr = p->iph.saddr;
 926         t->parms.iph.daddr = p->iph.daddr;
 927         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 928         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 929         ipip6_tunnel_link(sitn, t);
 930         t->parms.iph.ttl = p->iph.ttl;
 931         t->parms.iph.tos = p->iph.tos;
 932         if (t->parms.link != p->link) {
 933                 t->parms.link = p->link;
 934                 ipip6_tunnel_bind_dev(t->dev);
 935         }
 936         netdev_state_change(t->dev);
 937 }
 938
 939 static int
 940 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 941 {
 942         int err = 0;
 943         struct ip_tunnel_parm p;
 944         struct ip_tunnel_prl prl;
 945         struct ip_tunnel *t;
 946         struct net *net = dev_net(dev);
 947         struct sit_net *sitn = net_generic(net, sit_net_id);
 948 #ifdef CONFIG_IPV6_SIT_6RD
 949         struct ip_tunnel_6rd ip6rd;
 950 #endif
 951
 952         switch (cmd) {
 953         case SIOCGETTUNNEL:
 954 #ifdef CONFIG_IPV6_SIT_6RD
 955         case SIOCGET6RD:
 956 #endif
 957                 t = NULL;
 958                 if (dev == sitn->fb_tunnel_dev) {
 959                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 960                                 err = -EFAULT;
 961                                 break;
 962                         }
 963                         t = ipip6_tunnel_locate(net, &p, 0);
 964                 }
 965                 if (t == NULL)
 966                         t = netdev_priv(dev);
 967
 968                 err = -EFAULT;
 969                 if (cmd == SIOCGETTUNNEL) {
 970                         memcpy(&p, &t->parms, sizeof(p));
 971                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
 972                                          sizeof(p)))
 973                                 goto done;
 974 #ifdef CONFIG_IPV6_SIT_6RD
 975                 } else {
 976                         ip6rd.prefix = t->ip6rd.prefix;
 977                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
 978                         ip6rd.prefixlen = t->ip6rd.prefixlen;
 979                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
 980                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
 981                                          sizeof(ip6rd)))
 982                                 goto done;
 983 #endif
 984                 }
 985                 err = 0;
 986                 break;
 987
 988         case SIOCADDTUNNEL:
 989         case SIOCCHGTUNNEL:
 990                 err = -EPERM;
 991                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 992                         goto done;
 993
 994                 err = -EFAULT;
 995                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 996                         goto done;
 997
 998                 err = -EINVAL;
 999                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1000                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1001                         goto done;
1002                 if (p.iph.ttl)
1003                         p.iph.frag_off |= htons(IP_DF);
1004
1005                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1006
1007                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1008                         if (t != NULL) {
1009                                 if (t->dev != dev) {
1010                                         err = -EEXIST;
1011                                         break;
1012                                 }
1013                         } else {
1014                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1015                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1016                                         err = -EINVAL;
1017                                         break;
1018                                 }
1019                                 t = netdev_priv(dev);
1020                         }
1021
1022                         ipip6_tunnel_update(t, &p);
1023                 }
1024
1025                 if (t) {
1026                         err = 0;
1027                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1028                                 err = -EFAULT;
1029                 } else
1030                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1031                 break;
1032
1033         case SIOCDELTUNNEL:
1034                 err = -EPERM;
1035                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1036                         goto done;
1037
1038                 if (dev == sitn->fb_tunnel_dev) {
1039                         err = -EFAULT;
1040                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1041                                 goto done;
1042                         err = -ENOENT;
1043                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1044                                 goto done;
1045                         err = -EPERM;
1046                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1047                                 goto done;
1048                         dev = t->dev;
1049                 }
1050                 unregister_netdevice(dev);
1051                 err = 0;
1052                 break;
1053
1054         case SIOCGETPRL:
1055                 err = -EINVAL;
1056                 if (dev == sitn->fb_tunnel_dev)
1057                         goto done;
1058                 err = -ENOENT;
1059                 if (!(t = netdev_priv(dev)))
1060                         goto done;
1061                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1062                 break;
1063
1064         case SIOCADDPRL:
1065         case SIOCDELPRL:
1066         case SIOCCHGPRL:
1067                 err = -EPERM;
1068                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1069                         goto done;
1070                 err = -EINVAL;
1071                 if (dev == sitn->fb_tunnel_dev)
1072                         goto done;
1073                 err = -EFAULT;
1074                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1075                         goto done;
1076                 err = -ENOENT;
1077                 if (!(t = netdev_priv(dev)))
1078                         goto done;
1079
1080                 switch (cmd) {
1081                 case SIOCDELPRL:
1082                         err = ipip6_tunnel_del_prl(t, &prl);
1083                         break;
1084                 case SIOCADDPRL:
1085                 case SIOCCHGPRL:
1086                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1087                         break;
1088                 }
1089                 netdev_state_change(dev);
1090                 break;
1091
1092 #ifdef CONFIG_IPV6_SIT_6RD
1093         case SIOCADD6RD:
1094         case SIOCCHG6RD:
1095         case SIOCDEL6RD:
1096                 err = -EPERM;
1097                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1098                         goto done;
1099
1100                 err = -EFAULT;
1101                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1102                                    sizeof(ip6rd)))
1103                         goto done;
1104
1105                 t = netdev_priv(dev);
1106
1107                 if (cmd != SIOCDEL6RD) {
1108                         struct in6_addr prefix;
1109                         __be32 relay_prefix;
1110
1111                         err = -EINVAL;
1112                         if (ip6rd.relay_prefixlen > 32 ||
1113                             ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1114                                 goto done;
1115
1116                         ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1117                                          ip6rd.prefixlen);
1118                         if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1119                                 goto done;
1120                         if (ip6rd.relay_prefixlen)
1121                                 relay_prefix = ip6rd.relay_prefix &
1122                                                htonl(0xffffffffUL <<
1123                                                      (32 - ip6rd.relay_prefixlen));
1124                         else
1125                                 relay_prefix = 0;
1126                         if (relay_prefix != ip6rd.relay_prefix)
1127                                 goto done;
1128
1129                         t->ip6rd.prefix = prefix;
1130                         t->ip6rd.relay_prefix = relay_prefix;
1131                         t->ip6rd.prefixlen = ip6rd.prefixlen;
1132                         t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1133                 } else
1134                         ipip6_tunnel_clone_6rd(dev, sitn);
1135
1136                 err = 0;
1137                 break;
1138 #endif
1139
1140         default:
1141                 err = -EINVAL;
1142         }
1143
1144 done:
1145         return err;
1146 }
1147
1148 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1149 {
1150         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1151                 return -EINVAL;
1152         dev->mtu = new_mtu;
1153         return 0;
1154 }
1155
1156 static const struct net_device_ops ipip6_netdev_ops = {
1157         .ndo_uninit     = ipip6_tunnel_uninit,
1158         .ndo_start_xmit = ipip6_tunnel_xmit,
1159         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1160         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1161         .ndo_get_stats64= ipip6_get_stats64,
1162 };
1163
1164 static void ipip6_dev_free(struct net_device *dev)
1165 {
1166         free_percpu(dev->tstats);
1167         free_netdev(dev);
1168 }
1169
1170 static void ipip6_tunnel_setup(struct net_device *dev)
1171 {
1172         dev->netdev_ops         = &ipip6_netdev_ops;
1173         dev->destructor         = ipip6_dev_free;
1174
1175         dev->type               = ARPHRD_SIT;
1176         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1177         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1178         dev->flags              = IFF_NOARP;
1179         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1180         dev->iflink             = 0;
1181         dev->addr_len           = 4;
1182         dev->features           |= NETIF_F_NETNS_LOCAL;
1183         dev->features           |= NETIF_F_LLTX;
1184 }
1185
1186 static int ipip6_tunnel_init(struct net_device *dev)
1187 {
1188         struct ip_tunnel *tunnel = netdev_priv(dev);
1189
1190         tunnel->dev = dev;
1191
1192         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1193         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1194
1195         ipip6_tunnel_bind_dev(dev);
1196         dev->tstats = alloc_percpu(struct pcpu_tstats);
1197         if (!dev->tstats)
1198                 return -ENOMEM;
1199
1200         return 0;
1201 }
1202
1203 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1204 {
1205         struct ip_tunnel *tunnel = netdev_priv(dev);
1206         struct iphdr *iph = &tunnel->parms.iph;
1207         struct net *net = dev_net(dev);
1208         struct sit_net *sitn = net_generic(net, sit_net_id);
1209
1210         tunnel->dev = dev;
1211         strcpy(tunnel->parms.name, dev->name);
1212
1213         iph->version            = 4;
1214         iph->protocol           = IPPROTO_IPV6;
1215         iph->ihl                = 5;
1216         iph->ttl                = 64;
1217
1218         dev->tstats = alloc_percpu(struct pcpu_tstats);
1219         if (!dev->tstats)
1220                 return -ENOMEM;
1221         dev_hold(dev);
1222         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1223         return 0;
1224 }
1225
1226 static void ipip6_netlink_parms(struct nlattr *data[],
1227                                 struct ip_tunnel_parm *parms)
1228 {
1229         memset(parms, 0, sizeof(*parms));
1230
1231         parms->iph.version = 4;
1232         parms->iph.protocol = IPPROTO_IPV6;
1233         parms->iph.ihl = 5;
1234         parms->iph.ttl = 64;
1235
1236         if (!data)
1237                 return;
1238
1239         if (data[IFLA_IPTUN_LINK])
1240                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1241
1242         if (data[IFLA_IPTUN_LOCAL])
1243                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1244
1245         if (data[IFLA_IPTUN_REMOTE])
1246                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1247
1248         if (data[IFLA_IPTUN_TTL]) {
1249                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1250                 if (parms->iph.ttl)
1251                         parms->iph.frag_off = htons(IP_DF);
1252         }
1253
1254         if (data[IFLA_IPTUN_TOS])
1255                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1256
1257         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1258                 parms->iph.frag_off = htons(IP_DF);
1259
1260         if (data[IFLA_IPTUN_FLAGS])
1261                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1262 }
1263
1264 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1265                          struct nlattr *tb[], struct nlattr *data[])
1266 {
1267         struct net *net = dev_net(dev);
1268         struct ip_tunnel *nt;
1269
1270         nt = netdev_priv(dev);
1271         ipip6_netlink_parms(data, &nt->parms);
1272
1273         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1274                 return -EEXIST;
1275
1276         return ipip6_tunnel_create(dev);
1277 }
1278
1279 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1280                           struct nlattr *data[])
1281 {
1282         struct ip_tunnel *t;
1283         struct ip_tunnel_parm p;
1284         struct net *net = dev_net(dev);
1285         struct sit_net *sitn = net_generic(net, sit_net_id);
1286
1287         if (dev == sitn->fb_tunnel_dev)
1288                 return -EINVAL;
1289
1290         ipip6_netlink_parms(data, &p);
1291
1292         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1293             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1294                 return -EINVAL;
1295
1296         t = ipip6_tunnel_locate(net, &p, 0);
1297
1298         if (t) {
1299                 if (t->dev != dev)
1300                         return -EEXIST;
1301         } else
1302                 t = netdev_priv(dev);
1303
1304         ipip6_tunnel_update(t, &p);
1305         return 0;
1306 }
1307
1308 static size_t ipip6_get_size(const struct net_device *dev)
1309 {
1310         return
1311                 /* IFLA_IPTUN_LINK */
1312                 nla_total_size(4) +
1313                 /* IFLA_IPTUN_LOCAL */
1314                 nla_total_size(4) +
1315                 /* IFLA_IPTUN_REMOTE */
1316                 nla_total_size(4) +
1317                 /* IFLA_IPTUN_TTL */
1318                 nla_total_size(1) +
1319                 /* IFLA_IPTUN_TOS */
1320                 nla_total_size(1) +
1321                 /* IFLA_IPTUN_PMTUDISC */
1322                 nla_total_size(1) +
1323                 /* IFLA_IPTUN_FLAGS */
1324                 nla_total_size(2) +
1325                 0;
1326 }
1327
1328 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1329 {
1330         struct ip_tunnel *tunnel = netdev_priv(dev);
1331         struct ip_tunnel_parm *parm = &tunnel->parms;
1332
1333         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1334             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1335             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1336             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1337             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1338             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1339                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1340             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1341                 goto nla_put_failure;
1342         return 0;
1343
1344 nla_put_failure:
1345         return -EMSGSIZE;
1346 }
1347
1348 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1349         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1350         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1351         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1352         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1353         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1354         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1355         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1356 };
1357
1358 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1359         .kind           = "sit",
1360         .maxtype        = IFLA_IPTUN_MAX,
1361         .policy         = ipip6_policy,
1362         .priv_size      = sizeof(struct ip_tunnel),
1363         .setup          = ipip6_tunnel_setup,
1364         .newlink        = ipip6_newlink,
1365         .changelink     = ipip6_changelink,
1366         .get_size       = ipip6_get_size,
1367         .fill_info      = ipip6_fill_info,
1368 };
1369
1370 static struct xfrm_tunnel sit_handler __read_mostly = {
1371         .handler        =       ipip6_rcv,
1372         .err_handler    =       ipip6_err,
1373         .priority       =       1,
1374 };
1375
1376 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1377 {
1378         int prio;
1379
1380         for (prio = 1; prio < 4; prio++) {
1381                 int h;
1382                 for (h = 0; h < HASH_SIZE; h++) {
1383                         struct ip_tunnel *t;
1384
1385                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1386                         while (t != NULL) {
1387                                 unregister_netdevice_queue(t->dev, head);
1388                                 t = rtnl_dereference(t->next);
1389                         }
1390                 }
1391         }
1392 }
1393
1394 static int __net_init sit_init_net(struct net *net)
1395 {
1396         struct sit_net *sitn = net_generic(net, sit_net_id);
1397         struct ip_tunnel *t;
1398         int err;
1399
1400         sitn->tunnels[0] = sitn->tunnels_wc;
1401         sitn->tunnels[1] = sitn->tunnels_l;
1402         sitn->tunnels[2] = sitn->tunnels_r;
1403         sitn->tunnels[3] = sitn->tunnels_r_l;
1404
1405         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1406                                            ipip6_tunnel_setup);
1407         if (!sitn->fb_tunnel_dev) {
1408                 err = -ENOMEM;
1409                 goto err_alloc_dev;
1410         }
1411         dev_net_set(sitn->fb_tunnel_dev, net);
1412
1413         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1414         if (err)
1415                 goto err_dev_free;
1416
1417         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1418
1419         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1420                 goto err_reg_dev;
1421
1422         t = netdev_priv(sitn->fb_tunnel_dev);
1423
1424         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1425         return 0;
1426
1427 err_reg_dev:
1428         dev_put(sitn->fb_tunnel_dev);
1429 err_dev_free:
1430         ipip6_dev_free(sitn->fb_tunnel_dev);
1431 err_alloc_dev:
1432         return err;
1433 }
1434
1435 static void __net_exit sit_exit_net(struct net *net)
1436 {
1437         struct sit_net *sitn = net_generic(net, sit_net_id);
1438         LIST_HEAD(list);
1439
1440         rtnl_lock();
1441         sit_destroy_tunnels(sitn, &list);
1442         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1443         unregister_netdevice_many(&list);
1444         rtnl_unlock();
1445 }
1446
1447 static struct pernet_operations sit_net_ops = {
1448         .init = sit_init_net,
1449         .exit = sit_exit_net,
1450         .id   = &sit_net_id,
1451         .size = sizeof(struct sit_net),
1452 };
1453
1454 static void __exit sit_cleanup(void)
1455 {
1456         rtnl_link_unregister(&sit_link_ops);
1457         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1458
1459         unregister_pernet_device(&sit_net_ops);
1460         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1461 }
1462
1463 static int __init sit_init(void)
1464 {
1465         int err;
1466
1467         pr_info("IPv6 over IPv4 tunneling driver\n");
1468
1469         err = register_pernet_device(&sit_net_ops);
1470         if (err < 0)
1471                 return err;
1472         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1473         if (err < 0) {
1474                 pr_info("%s: can't add protocol\n", __func__);
1475                 goto xfrm_tunnel_failed;
1476         }
1477         err = rtnl_link_register(&sit_link_ops);
1478         if (err < 0)
1479                 goto rtnl_link_failed;
1480
1481 out:
1482         return err;
1483
1484 rtnl_link_failed:
1485         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1486 xfrm_tunnel_failed:
1487         unregister_pernet_device(&sit_net_ops);
1488         goto out;
1489 }
1490
1491 module_init(sit_init);
1492 module_exit(sit_cleanup);
1493 MODULE_LICENSE("GPL");
1494 MODULE_ALIAS_NETDEV("sit0");