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 #include <linux/module.h>
  21 #include <linux/capability.h>
  22 #include <linux/errno.h>
  23 #include <linux/types.h>
  24 #include <linux/socket.h>
  25 #include <linux/sockios.h>
  26 #include <linux/net.h>
  27 #include <linux/in6.h>
  28 #include <linux/netdevice.h>
  29 #include <linux/if_arp.h>
  30 #include <linux/icmp.h>
  31 #include <linux/slab.h>
  32 #include <asm/uaccess.h>
  33 #include <linux/init.h>
  34 #include <linux/netfilter_ipv4.h>
  35 #include <linux/if_ether.h>
  36
  37 #include <net/sock.h>
  38 #include <net/snmp.h>
  39
  40 #include <net/ipv6.h>
  41 #include <net/protocol.h>
  42 #include <net/transp_v6.h>
  43 #include <net/ip6_fib.h>
  44 #include <net/ip6_route.h>
  45 #include <net/ndisc.h>
  46 #include <net/addrconf.h>
  47 #include <net/ip.h>
  48 #include <net/udp.h>
  49 #include <net/icmp.h>
  50 #include <net/ipip.h>
  51 #include <net/inet_ecn.h>
  52 #include <net/xfrm.h>
  53 #include <net/dsfield.h>
  54 #include <net/net_namespace.h>
  55 #include <net/netns/generic.h>
  56
  57 /*
  58    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  59
  60    For comments look at net/ipv4/ip_gre.c --ANK
  61  */
  62
  63 #define HASH_SIZE  16
  64 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  65
  66 static void ipip6_tunnel_init(struct net_device *dev);
  67 static void ipip6_tunnel_setup(struct net_device *dev);
  68
  69 static int sit_net_id __read_mostly;
  70 struct sit_net {
  71         struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  72         struct ip_tunnel *tunnels_r[HASH_SIZE];
  73         struct ip_tunnel *tunnels_l[HASH_SIZE];
  74         struct ip_tunnel *tunnels_wc[1];
  75         struct ip_tunnel **tunnels[4];
  76
  77         struct net_device *fb_tunnel_dev;
  78 };
  79
  80 /*
  81  * Locking : hash tables are protected by RCU and a spinlock
  82  */
  83 static DEFINE_SPINLOCK(ipip6_lock);
  84
  85 #define for_each_ip_tunnel_rcu(start) \
  86         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
  87
  88 /*
  89  * Must be invoked with rcu_read_lock
  90  */
  91 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
  92                 struct net_device *dev, __be32 remote, __be32 local)
  93 {
  94         unsigned h0 = HASH(remote);
  95         unsigned h1 = HASH(local);
  96         struct ip_tunnel *t;
  97         struct sit_net *sitn = net_generic(net, sit_net_id);
  98
  99         for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
 100                 if (local == t->parms.iph.saddr &&
 101                     remote == t->parms.iph.daddr &&
 102                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 103                     (t->dev->flags & IFF_UP))
 104                         return t;
 105         }
 106         for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
 107                 if (remote == t->parms.iph.daddr &&
 108                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 109                     (t->dev->flags & IFF_UP))
 110                         return t;
 111         }
 112         for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
 113                 if (local == t->parms.iph.saddr &&
 114                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 115                     (t->dev->flags & IFF_UP))
 116                         return t;
 117         }
 118         t = rcu_dereference(sitn->tunnels_wc[0]);
 119         if ((t != NULL) && (t->dev->flags & IFF_UP))
 120                 return t;
 121         return NULL;
 122 }
 123
 124 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
 125                 struct ip_tunnel_parm *parms)
 126 {
 127         __be32 remote = parms->iph.daddr;
 128         __be32 local = parms->iph.saddr;
 129         unsigned h = 0;
 130         int prio = 0;
 131
 132         if (remote) {
 133                 prio |= 2;
 134                 h ^= HASH(remote);
 135         }
 136         if (local) {
 137                 prio |= 1;
 138                 h ^= HASH(local);
 139         }
 140         return &sitn->tunnels[prio][h];
 141 }
 142
 143 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
 144                 struct ip_tunnel *t)
 145 {
 146         return __ipip6_bucket(sitn, &t->parms);
 147 }
 148
 149 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 150 {
 151         struct ip_tunnel **tp;
 152
 153         for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
 154                 if (t == *tp) {
 155                         spin_lock_bh(&ipip6_lock);
 156                         *tp = t->next;
 157                         spin_unlock_bh(&ipip6_lock);
 158                         break;
 159                 }
 160         }
 161 }
 162
 163 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 164 {
 165         struct ip_tunnel **tp = ipip6_bucket(sitn, t);
 166
 167         spin_lock_bh(&ipip6_lock);
 168         t->next = *tp;
 169         rcu_assign_pointer(*tp, t);
 170         spin_unlock_bh(&ipip6_lock);
 171 }
 172
 173 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 174 {
 175 #ifdef CONFIG_IPV6_SIT_6RD
 176         struct ip_tunnel *t = netdev_priv(dev);
 177
 178         if (t->dev == sitn->fb_tunnel_dev) {
 179                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 180                 t->ip6rd.relay_prefix = 0;
 181                 t->ip6rd.prefixlen = 16;
 182                 t->ip6rd.relay_prefixlen = 0;
 183         } else {
 184                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 185                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 186         }
 187 #endif
 188 }
 189
 190 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
 191                 struct ip_tunnel_parm *parms, int create)
 192 {
 193         __be32 remote = parms->iph.daddr;
 194         __be32 local = parms->iph.saddr;
 195         struct ip_tunnel *t, **tp, *nt;
 196         struct net_device *dev;
 197         char name[IFNAMSIZ];
 198         struct sit_net *sitn = net_generic(net, sit_net_id);
 199
 200         for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
 201                 if (local == t->parms.iph.saddr &&
 202                     remote == t->parms.iph.daddr &&
 203                     parms->link == t->parms.link) {
 204                         if (create)
 205                                 return NULL;
 206                         else
 207                                 return t;
 208                 }
 209         }
 210         if (!create)
 211                 goto failed;
 212
 213         if (parms->name[0])
 214                 strlcpy(name, parms->name, IFNAMSIZ);
 215         else
 216                 sprintf(name, "sit%%d");
 217
 218         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 219         if (dev == NULL)
 220                 return NULL;
 221
 222         dev_net_set(dev, net);
 223
 224         if (strchr(name, '%')) {
 225                 if (dev_alloc_name(dev, name) < 0)
 226                         goto failed_free;
 227         }
 228
 229         nt = netdev_priv(dev);
 230
 231         nt->parms = *parms;
 232         ipip6_tunnel_init(dev);
 233         ipip6_tunnel_clone_6rd(dev, sitn);
 234
 235         if (parms->i_flags & SIT_ISATAP)
 236                 dev->priv_flags |= IFF_ISATAP;
 237
 238         if (register_netdevice(dev) < 0)
 239                 goto failed_free;
 240
 241         dev_hold(dev);
 242
 243         ipip6_tunnel_link(sitn, nt);
 244         return nt;
 245
 246 failed_free:
 247         free_netdev(dev);
 248 failed:
 249         return NULL;
 250 }
 251
 252 static DEFINE_SPINLOCK(ipip6_prl_lock);
 253
 254 #define for_each_prl_rcu(start)                 \
 255         for (prl = rcu_dereference(start);      \
 256              prl;                               \
 257              prl = rcu_dereference(prl->next))
 258
 259 static struct ip_tunnel_prl_entry *
 260 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 261 {
 262         struct ip_tunnel_prl_entry *prl;
 263
 264         for_each_prl_rcu(t->prl)
 265                 if (prl->addr == addr)
 266                         break;
 267         return prl;
 268
 269 }
 270
 271 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 272                                 struct ip_tunnel_prl __user *a)
 273 {
 274         struct ip_tunnel_prl kprl, *kp;
 275         struct ip_tunnel_prl_entry *prl;
 276         unsigned int cmax, c = 0, ca, len;
 277         int ret = 0;
 278
 279         if (copy_from_user(&kprl, a, sizeof(kprl)))
 280                 return -EFAULT;
 281         cmax = kprl.datalen / sizeof(kprl);
 282         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 283                 cmax = 1;
 284
 285         /* For simple GET or for root users,
 286          * we try harder to allocate.
 287          */
 288         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 289                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 290                 NULL;
 291
 292         rcu_read_lock();
 293
 294         ca = t->prl_count < cmax ? t->prl_count : cmax;
 295
 296         if (!kp) {
 297                 /* We don't try hard to allocate much memory for
 298                  * non-root users.
 299                  * For root users, retry allocating enough memory for
 300                  * the answer.
 301                  */
 302                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 303                 if (!kp) {
 304                         ret = -ENOMEM;
 305                         goto out;
 306                 }
 307         }
 308
 309         c = 0;
 310         for_each_prl_rcu(t->prl) {
 311                 if (c >= cmax)
 312                         break;
 313                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 314                         continue;
 315                 kp[c].addr = prl->addr;
 316                 kp[c].flags = prl->flags;
 317                 c++;
 318                 if (kprl.addr != htonl(INADDR_ANY))
 319                         break;
 320         }
 321 out:
 322         rcu_read_unlock();
 323
 324         len = sizeof(*kp) * c;
 325         ret = 0;
 326         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 327                 ret = -EFAULT;
 328
 329         kfree(kp);
 330
 331         return ret;
 332 }
 333
 334 static int
 335 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 336 {
 337         struct ip_tunnel_prl_entry *p;
 338         int err = 0;
 339
 340         if (a->addr == htonl(INADDR_ANY))
 341                 return -EINVAL;
 342
 343         spin_lock(&ipip6_prl_lock);
 344
 345         for (p = t->prl; p; p = p->next) {
 346                 if (p->addr == a->addr) {
 347                         if (chg) {
 348                                 p->flags = a->flags;
 349                                 goto out;
 350                         }
 351                         err = -EEXIST;
 352                         goto out;
 353                 }
 354         }
 355
 356         if (chg) {
 357                 err = -ENXIO;
 358                 goto out;
 359         }
 360
 361         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 362         if (!p) {
 363                 err = -ENOBUFS;
 364                 goto out;
 365         }
 366
 367         p->next = t->prl;
 368         p->addr = a->addr;
 369         p->flags = a->flags;
 370         t->prl_count++;
 371         rcu_assign_pointer(t->prl, p);
 372 out:
 373         spin_unlock(&ipip6_prl_lock);
 374         return err;
 375 }
 376
 377 static void prl_entry_destroy_rcu(struct rcu_head *head)
 378 {
 379         kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
 380 }
 381
 382 static void prl_list_destroy_rcu(struct rcu_head *head)
 383 {
 384         struct ip_tunnel_prl_entry *p, *n;
 385
 386         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 387         do {
 388                 n = p->next;
 389                 kfree(p);
 390                 p = n;
 391         } while (p);
 392 }
 393
 394 static int
 395 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 396 {
 397         struct ip_tunnel_prl_entry *x, **p;
 398         int err = 0;
 399
 400         spin_lock(&ipip6_prl_lock);
 401
 402         if (a && a->addr != htonl(INADDR_ANY)) {
 403                 for (p = &t->prl; *p; p = &(*p)->next) {
 404                         if ((*p)->addr == a->addr) {
 405                                 x = *p;
 406                                 *p = x->next;
 407                                 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
 408                                 t->prl_count--;
 409                                 goto out;
 410                         }
 411                 }
 412                 err = -ENXIO;
 413         } else {
 414                 if (t->prl) {
 415                         t->prl_count = 0;
 416                         x = t->prl;
 417                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 418                         t->prl = NULL;
 419                 }
 420         }
 421 out:
 422         spin_unlock(&ipip6_prl_lock);
 423         return err;
 424 }
 425
 426 static int
 427 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 428 {
 429         struct ip_tunnel_prl_entry *p;
 430         int ok = 1;
 431
 432         rcu_read_lock();
 433         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 434         if (p) {
 435                 if (p->flags & PRL_DEFAULT)
 436                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 437                 else
 438                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 439         } else {
 440                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 441                 if (ipv6_addr_is_isatap(addr6) &&
 442                     (addr6->s6_addr32[3] == iph->saddr) &&
 443                     ipv6_chk_prefix(addr6, t->dev))
 444                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 445                 else
 446                         ok = 0;
 447         }
 448         rcu_read_unlock();
 449         return ok;
 450 }
 451
 452 static void ipip6_tunnel_uninit(struct net_device *dev)
 453 {
 454         struct net *net = dev_net(dev);
 455         struct sit_net *sitn = net_generic(net, sit_net_id);
 456
 457         if (dev == sitn->fb_tunnel_dev) {
 458                 spin_lock_bh(&ipip6_lock);
 459                 sitn->tunnels_wc[0] = NULL;
 460                 spin_unlock_bh(&ipip6_lock);
 461                 dev_put(dev);
 462         } else {
 463                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 464                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 465                 dev_put(dev);
 466         }
 467 }
 468
 469
 470 static int ipip6_err(struct sk_buff *skb, u32 info)
 471 {
 472
 473 /* All the routers (except for Linux) return only
 474    8 bytes of packet payload. It means, that precise relaying of
 475    ICMP in the real Internet is absolutely infeasible.
 476  */
 477         struct iphdr *iph = (struct iphdr*)skb->data;
 478         const int type = icmp_hdr(skb)->type;
 479         const int code = icmp_hdr(skb)->code;
 480         struct ip_tunnel *t;
 481         int err;
 482
 483         switch (type) {
 484         default:
 485         case ICMP_PARAMETERPROB:
 486                 return 0;
 487
 488         case ICMP_DEST_UNREACH:
 489                 switch (code) {
 490                 case ICMP_SR_FAILED:
 491                 case ICMP_PORT_UNREACH:
 492                         /* Impossible event. */
 493                         return 0;
 494                 case ICMP_FRAG_NEEDED:
 495                         /* Soft state for pmtu is maintained by IP core. */
 496                         return 0;
 497                 default:
 498                         /* All others are translated to HOST_UNREACH.
 499                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 500                            I believe they are just ether pollution. --ANK
 501                          */
 502                         break;
 503                 }
 504                 break;
 505         case ICMP_TIME_EXCEEDED:
 506                 if (code != ICMP_EXC_TTL)
 507                         return 0;
 508                 break;
 509         }
 510
 511         err = -ENOENT;
 512
 513         rcu_read_lock();
 514         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 515                                 skb->dev,
 516                                 iph->daddr,
 517                                 iph->saddr);
 518         if (t == NULL || t->parms.iph.daddr == 0)
 519                 goto out;
 520
 521         err = 0;
 522         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 523                 goto out;
 524
 525         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 526                 t->err_count++;
 527         else
 528                 t->err_count = 1;
 529         t->err_time = jiffies;
 530 out:
 531         rcu_read_unlock();
 532         return err;
 533 }
 534
 535 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 536 {
 537         if (INET_ECN_is_ce(iph->tos))
 538                 IP6_ECN_set_ce(ipv6_hdr(skb));
 539 }
 540
 541 static int ipip6_rcv(struct sk_buff *skb)
 542 {
 543         struct iphdr *iph;
 544         struct ip_tunnel *tunnel;
 545
 546         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 547                 goto out;
 548
 549         iph = ip_hdr(skb);
 550
 551         rcu_read_lock();
 552         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 553                                      iph->saddr, iph->daddr);
 554         if (tunnel != NULL) {
 555                 secpath_reset(skb);
 556                 skb->mac_header = skb->network_header;
 557                 skb_reset_network_header(skb);
 558                 IPCB(skb)->flags = 0;
 559                 skb->protocol = htons(ETH_P_IPV6);
 560                 skb->pkt_type = PACKET_HOST;
 561
 562                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 563                     !isatap_chksrc(skb, iph, tunnel)) {
 564                         tunnel->dev->stats.rx_errors++;
 565                         rcu_read_unlock();
 566                         kfree_skb(skb);
 567                         return 0;
 568                 }
 569
 570                 skb_tunnel_rx(skb, tunnel->dev);
 571
 572                 ipip6_ecn_decapsulate(iph, skb);
 573                 netif_rx(skb);
 574                 rcu_read_unlock();
 575                 return 0;
 576         }
 577
 578         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 579         rcu_read_unlock();
 580 out:
 581         kfree_skb(skb);
 582         return 0;
 583 }
 584
 585 /*
 586  * Returns the embedded IPv4 address if the IPv6 address
 587  * comes from 6rd / 6to4 (RFC 3056) addr space.
 588  */
 589 static inline
 590 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 591 {
 592         __be32 dst = 0;
 593
 594 #ifdef CONFIG_IPV6_SIT_6RD
 595         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 596                               tunnel->ip6rd.prefixlen)) {
 597                 unsigned pbw0, pbi0;
 598                 int pbi1;
 599                 u32 d;
 600
 601                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 602                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 603
 604                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 605                     tunnel->ip6rd.relay_prefixlen;
 606
 607                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 608                 if (pbi1 > 0)
 609                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 610                              (32 - pbi1);
 611
 612                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
 613         }
 614 #else
 615         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 616                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 617                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 618         }
 619 #endif
 620         return dst;
 621 }
 622
 623 /*
 624  *      This function assumes it is being called from dev_queue_xmit()
 625  *      and that skb is filled properly by that function.
 626  */
 627
 628 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 629                                      struct net_device *dev)
 630 {
 631         struct ip_tunnel *tunnel = netdev_priv(dev);
 632         struct net_device_stats *stats = &dev->stats;
 633         struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
 634         struct iphdr  *tiph = &tunnel->parms.iph;
 635         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 636         u8     tos = tunnel->parms.iph.tos;
 637         __be16 df = tiph->frag_off;
 638         struct rtable *rt;                      /* Route to the other host */
 639         struct net_device *tdev;                        /* Device to other host */
 640         struct iphdr  *iph;                     /* Our new IP header */
 641         unsigned int max_headroom;              /* The extra header space needed */
 642         __be32 dst = tiph->daddr;
 643         int    mtu;
 644         struct in6_addr *addr6;
 645         int addr_type;
 646
 647         if (skb->protocol != htons(ETH_P_IPV6))
 648                 goto tx_error;
 649
 650         /* ISATAP (RFC4214) - must come before 6to4 */
 651         if (dev->priv_flags & IFF_ISATAP) {
 652                 struct neighbour *neigh = NULL;
 653
 654                 if (skb_dst(skb))
 655                         neigh = skb_dst(skb)->neighbour;
 656
 657                 if (neigh == NULL) {
 658                         if (net_ratelimit())
 659                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 660                         goto tx_error;
 661                 }
 662
 663                 addr6 = (struct in6_addr*)&neigh->primary_key;
 664                 addr_type = ipv6_addr_type(addr6);
 665
 666                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 667                      ipv6_addr_is_isatap(addr6))
 668                         dst = addr6->s6_addr32[3];
 669                 else
 670                         goto tx_error;
 671         }
 672
 673         if (!dst)
 674                 dst = try_6rd(&iph6->daddr, tunnel);
 675
 676         if (!dst) {
 677                 struct neighbour *neigh = NULL;
 678
 679                 if (skb_dst(skb))
 680                         neigh = skb_dst(skb)->neighbour;
 681
 682                 if (neigh == NULL) {
 683                         if (net_ratelimit())
 684                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 685                         goto tx_error;
 686                 }
 687
 688                 addr6 = (struct in6_addr*)&neigh->primary_key;
 689                 addr_type = ipv6_addr_type(addr6);
 690
 691                 if (addr_type == IPV6_ADDR_ANY) {
 692                         addr6 = &ipv6_hdr(skb)->daddr;
 693                         addr_type = ipv6_addr_type(addr6);
 694                 }
 695
 696                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 697                         goto tx_error_icmp;
 698
 699                 dst = addr6->s6_addr32[3];
 700         }
 701
 702         {
 703                 struct flowi fl = { .nl_u = { .ip4_u =
 704                                               { .daddr = dst,
 705                                                 .saddr = tiph->saddr,
 706                                                 .tos = RT_TOS(tos) } },
 707                                     .oif = tunnel->parms.link,
 708                                     .proto = IPPROTO_IPV6 };
 709                 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
 710                         stats->tx_carrier_errors++;
 711                         goto tx_error_icmp;
 712                 }
 713         }
 714         if (rt->rt_type != RTN_UNICAST) {
 715                 ip_rt_put(rt);
 716                 stats->tx_carrier_errors++;
 717                 goto tx_error_icmp;
 718         }
 719         tdev = rt->u.dst.dev;
 720
 721         if (tdev == dev) {
 722                 ip_rt_put(rt);
 723                 stats->collisions++;
 724                 goto tx_error;
 725         }
 726
 727         if (df) {
 728                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 729
 730                 if (mtu < 68) {
 731                         stats->collisions++;
 732                         ip_rt_put(rt);
 733                         goto tx_error;
 734                 }
 735
 736                 if (mtu < IPV6_MIN_MTU) {
 737                         mtu = IPV6_MIN_MTU;
 738                         df = 0;
 739                 }
 740
 741                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 742                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
 743
 744                 if (skb->len > mtu) {
 745                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 746                         ip_rt_put(rt);
 747                         goto tx_error;
 748                 }
 749         }
 750
 751         if (tunnel->err_count > 0) {
 752                 if (time_before(jiffies,
 753                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 754                         tunnel->err_count--;
 755                         dst_link_failure(skb);
 756                 } else
 757                         tunnel->err_count = 0;
 758         }
 759
 760         /*
 761          * Okay, now see if we can stuff it in the buffer as-is.
 762          */
 763         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 764
 765         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 766             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 767                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 768                 if (!new_skb) {
 769                         ip_rt_put(rt);
 770                         txq->tx_dropped++;
 771                         dev_kfree_skb(skb);
 772                         return NETDEV_TX_OK;
 773                 }
 774                 if (skb->sk)
 775                         skb_set_owner_w(new_skb, skb->sk);
 776                 dev_kfree_skb(skb);
 777                 skb = new_skb;
 778                 iph6 = ipv6_hdr(skb);
 779         }
 780
 781         skb->transport_header = skb->network_header;
 782         skb_push(skb, sizeof(struct iphdr));
 783         skb_reset_network_header(skb);
 784         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 785         IPCB(skb)->flags = 0;
 786         skb_dst_drop(skb);
 787         skb_dst_set(skb, &rt->u.dst);
 788
 789         /*
 790          *      Push down and install the IPIP header.
 791          */
 792
 793         iph                     =       ip_hdr(skb);
 794         iph->version            =       4;
 795         iph->ihl                =       sizeof(struct iphdr)>>2;
 796         iph->frag_off           =       df;
 797         iph->protocol           =       IPPROTO_IPV6;
 798         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 799         iph->daddr              =       rt->rt_dst;
 800         iph->saddr              =       rt->rt_src;
 801
 802         if ((iph->ttl = tiph->ttl) == 0)
 803                 iph->ttl        =       iph6->hop_limit;
 804
 805         nf_reset(skb);
 806
 807         IPTUNNEL_XMIT();
 808         return NETDEV_TX_OK;
 809
 810 tx_error_icmp:
 811         dst_link_failure(skb);
 812 tx_error:
 813         stats->tx_errors++;
 814         dev_kfree_skb(skb);
 815         return NETDEV_TX_OK;
 816 }
 817
 818 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 819 {
 820         struct net_device *tdev = NULL;
 821         struct ip_tunnel *tunnel;
 822         struct iphdr *iph;
 823
 824         tunnel = netdev_priv(dev);
 825         iph = &tunnel->parms.iph;
 826
 827         if (iph->daddr) {
 828                 struct flowi fl = { .nl_u = { .ip4_u =
 829                                               { .daddr = iph->daddr,
 830                                                 .saddr = iph->saddr,
 831                                                 .tos = RT_TOS(iph->tos) } },
 832                                     .oif = tunnel->parms.link,
 833                                     .proto = IPPROTO_IPV6 };
 834                 struct rtable *rt;
 835                 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
 836                         tdev = rt->u.dst.dev;
 837                         ip_rt_put(rt);
 838                 }
 839                 dev->flags |= IFF_POINTOPOINT;
 840         }
 841
 842         if (!tdev && tunnel->parms.link)
 843                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 844
 845         if (tdev) {
 846                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 847                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 848                 if (dev->mtu < IPV6_MIN_MTU)
 849                         dev->mtu = IPV6_MIN_MTU;
 850         }
 851         dev->iflink = tunnel->parms.link;
 852 }
 853
 854 static int
 855 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 856 {
 857         int err = 0;
 858         struct ip_tunnel_parm p;
 859         struct ip_tunnel_prl prl;
 860         struct ip_tunnel *t;
 861         struct net *net = dev_net(dev);
 862         struct sit_net *sitn = net_generic(net, sit_net_id);
 863 #ifdef CONFIG_IPV6_SIT_6RD
 864         struct ip_tunnel_6rd ip6rd;
 865 #endif
 866
 867         switch (cmd) {
 868         case SIOCGETTUNNEL:
 869 #ifdef CONFIG_IPV6_SIT_6RD
 870         case SIOCGET6RD:
 871 #endif
 872                 t = NULL;
 873                 if (dev == sitn->fb_tunnel_dev) {
 874                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 875                                 err = -EFAULT;
 876                                 break;
 877                         }
 878                         t = ipip6_tunnel_locate(net, &p, 0);
 879                 }
 880                 if (t == NULL)
 881                         t = netdev_priv(dev);
 882
 883                 err = -EFAULT;
 884                 if (cmd == SIOCGETTUNNEL) {
 885                         memcpy(&p, &t->parms, sizeof(p));
 886                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
 887                                          sizeof(p)))
 888                                 goto done;
 889 #ifdef CONFIG_IPV6_SIT_6RD
 890                 } else {
 891                         ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
 892                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
 893                         ip6rd.prefixlen = t->ip6rd.prefixlen;
 894                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
 895                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
 896                                          sizeof(ip6rd)))
 897                                 goto done;
 898 #endif
 899                 }
 900                 err = 0;
 901                 break;
 902
 903         case SIOCADDTUNNEL:
 904         case SIOCCHGTUNNEL:
 905                 err = -EPERM;
 906                 if (!capable(CAP_NET_ADMIN))
 907                         goto done;
 908
 909                 err = -EFAULT;
 910                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 911                         goto done;
 912
 913                 err = -EINVAL;
 914                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 915                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 916                         goto done;
 917                 if (p.iph.ttl)
 918                         p.iph.frag_off |= htons(IP_DF);
 919
 920                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 921
 922                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 923                         if (t != NULL) {
 924                                 if (t->dev != dev) {
 925                                         err = -EEXIST;
 926                                         break;
 927                                 }
 928                         } else {
 929                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 930                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 931                                         err = -EINVAL;
 932                                         break;
 933                                 }
 934                                 t = netdev_priv(dev);
 935                                 ipip6_tunnel_unlink(sitn, t);
 936                                 t->parms.iph.saddr = p.iph.saddr;
 937                                 t->parms.iph.daddr = p.iph.daddr;
 938                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 939                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 940                                 ipip6_tunnel_link(sitn, t);
 941                                 netdev_state_change(dev);
 942                         }
 943                 }
 944
 945                 if (t) {
 946                         err = 0;
 947                         if (cmd == SIOCCHGTUNNEL) {
 948                                 t->parms.iph.ttl = p.iph.ttl;
 949                                 t->parms.iph.tos = p.iph.tos;
 950                                 if (t->parms.link != p.link) {
 951                                         t->parms.link = p.link;
 952                                         ipip6_tunnel_bind_dev(dev);
 953                                         netdev_state_change(dev);
 954                                 }
 955                         }
 956                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 957                                 err = -EFAULT;
 958                 } else
 959                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 960                 break;
 961
 962         case SIOCDELTUNNEL:
 963                 err = -EPERM;
 964                 if (!capable(CAP_NET_ADMIN))
 965                         goto done;
 966
 967                 if (dev == sitn->fb_tunnel_dev) {
 968                         err = -EFAULT;
 969                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 970                                 goto done;
 971                         err = -ENOENT;
 972                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
 973                                 goto done;
 974                         err = -EPERM;
 975                         if (t == netdev_priv(sitn->fb_tunnel_dev))
 976                                 goto done;
 977                         dev = t->dev;
 978                 }
 979                 unregister_netdevice(dev);
 980                 err = 0;
 981                 break;
 982
 983         case SIOCGETPRL:
 984                 err = -EINVAL;
 985                 if (dev == sitn->fb_tunnel_dev)
 986                         goto done;
 987                 err = -ENOENT;
 988                 if (!(t = netdev_priv(dev)))
 989                         goto done;
 990                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
 991                 break;
 992
 993         case SIOCADDPRL:
 994         case SIOCDELPRL:
 995         case SIOCCHGPRL:
 996                 err = -EPERM;
 997                 if (!capable(CAP_NET_ADMIN))
 998                         goto done;
 999                 err = -EINVAL;
1000                 if (dev == sitn->fb_tunnel_dev)
1001                         goto done;
1002                 err = -EFAULT;
1003                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1004                         goto done;
1005                 err = -ENOENT;
1006                 if (!(t = netdev_priv(dev)))
1007                         goto done;
1008
1009                 switch (cmd) {
1010                 case SIOCDELPRL:
1011                         err = ipip6_tunnel_del_prl(t, &prl);
1012                         break;
1013                 case SIOCADDPRL:
1014                 case SIOCCHGPRL:
1015                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1016                         break;
1017                 }
1018                 netdev_state_change(dev);
1019                 break;
1020
1021 #ifdef CONFIG_IPV6_SIT_6RD
1022         case SIOCADD6RD:
1023         case SIOCCHG6RD:
1024         case SIOCDEL6RD:
1025                 err = -EPERM;
1026                 if (!capable(CAP_NET_ADMIN))
1027                         goto done;
1028
1029                 err = -EFAULT;
1030                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1031                                    sizeof(ip6rd)))
1032                         goto done;
1033
1034                 t = netdev_priv(dev);
1035
1036                 if (cmd != SIOCDEL6RD) {
1037                         struct in6_addr prefix;
1038                         __be32 relay_prefix;
1039
1040                         err = -EINVAL;
1041                         if (ip6rd.relay_prefixlen > 32 ||
1042                             ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1043                                 goto done;
1044
1045                         ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1046                                          ip6rd.prefixlen);
1047                         if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1048                                 goto done;
1049                         if (ip6rd.relay_prefixlen)
1050                                 relay_prefix = ip6rd.relay_prefix &
1051                                                htonl(0xffffffffUL <<
1052                                                      (32 - ip6rd.relay_prefixlen));
1053                         else
1054                                 relay_prefix = 0;
1055                         if (relay_prefix != ip6rd.relay_prefix)
1056                                 goto done;
1057
1058                         ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1059                         t->ip6rd.relay_prefix = relay_prefix;
1060                         t->ip6rd.prefixlen = ip6rd.prefixlen;
1061                         t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1062                 } else
1063                         ipip6_tunnel_clone_6rd(dev, sitn);
1064
1065                 err = 0;
1066                 break;
1067 #endif
1068
1069         default:
1070                 err = -EINVAL;
1071         }
1072
1073 done:
1074         return err;
1075 }
1076
1077 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1078 {
1079         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1080                 return -EINVAL;
1081         dev->mtu = new_mtu;
1082         return 0;
1083 }
1084
1085 static const struct net_device_ops ipip6_netdev_ops = {
1086         .ndo_uninit     = ipip6_tunnel_uninit,
1087         .ndo_start_xmit = ipip6_tunnel_xmit,
1088         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1089         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1090 };
1091
1092 static void ipip6_tunnel_setup(struct net_device *dev)
1093 {
1094         dev->netdev_ops         = &ipip6_netdev_ops;
1095         dev->destructor         = free_netdev;
1096
1097         dev->type               = ARPHRD_SIT;
1098         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1099         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1100         dev->flags              = IFF_NOARP;
1101         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1102         dev->iflink             = 0;
1103         dev->addr_len           = 4;
1104         dev->features           |= NETIF_F_NETNS_LOCAL;
1105 }
1106
1107 static void ipip6_tunnel_init(struct net_device *dev)
1108 {
1109         struct ip_tunnel *tunnel = netdev_priv(dev);
1110
1111         tunnel->dev = dev;
1112         strcpy(tunnel->parms.name, dev->name);
1113
1114         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1115         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1116
1117         ipip6_tunnel_bind_dev(dev);
1118 }
1119
1120 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1121 {
1122         struct ip_tunnel *tunnel = netdev_priv(dev);
1123         struct iphdr *iph = &tunnel->parms.iph;
1124         struct net *net = dev_net(dev);
1125         struct sit_net *sitn = net_generic(net, sit_net_id);
1126
1127         tunnel->dev = dev;
1128         strcpy(tunnel->parms.name, dev->name);
1129
1130         iph->version            = 4;
1131         iph->protocol           = IPPROTO_IPV6;
1132         iph->ihl                = 5;
1133         iph->ttl                = 64;
1134
1135         dev_hold(dev);
1136         sitn->tunnels_wc[0]     = tunnel;
1137 }
1138
1139 static struct xfrm_tunnel sit_handler = {
1140         .handler        =       ipip6_rcv,
1141         .err_handler    =       ipip6_err,
1142         .priority       =       1,
1143 };
1144
1145 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1146 {
1147         int prio;
1148
1149         for (prio = 1; prio < 4; prio++) {
1150                 int h;
1151                 for (h = 0; h < HASH_SIZE; h++) {
1152                         struct ip_tunnel *t = sitn->tunnels[prio][h];
1153
1154                         while (t != NULL) {
1155                                 unregister_netdevice_queue(t->dev, head);
1156                                 t = t->next;
1157                         }
1158                 }
1159         }
1160 }
1161
1162 static int __net_init sit_init_net(struct net *net)
1163 {
1164         struct sit_net *sitn = net_generic(net, sit_net_id);
1165         int err;
1166
1167         sitn->tunnels[0] = sitn->tunnels_wc;
1168         sitn->tunnels[1] = sitn->tunnels_l;
1169         sitn->tunnels[2] = sitn->tunnels_r;
1170         sitn->tunnels[3] = sitn->tunnels_r_l;
1171
1172         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1173                                            ipip6_tunnel_setup);
1174         if (!sitn->fb_tunnel_dev) {
1175                 err = -ENOMEM;
1176                 goto err_alloc_dev;
1177         }
1178         dev_net_set(sitn->fb_tunnel_dev, net);
1179
1180         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1181         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1182
1183         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1184                 goto err_reg_dev;
1185
1186         return 0;
1187
1188 err_reg_dev:
1189         dev_put(sitn->fb_tunnel_dev);
1190         free_netdev(sitn->fb_tunnel_dev);
1191 err_alloc_dev:
1192         return err;
1193 }
1194
1195 static void __net_exit sit_exit_net(struct net *net)
1196 {
1197         struct sit_net *sitn = net_generic(net, sit_net_id);
1198         LIST_HEAD(list);
1199
1200         rtnl_lock();
1201         sit_destroy_tunnels(sitn, &list);
1202         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1203         unregister_netdevice_many(&list);
1204         rtnl_unlock();
1205 }
1206
1207 static struct pernet_operations sit_net_ops = {
1208         .init = sit_init_net,
1209         .exit = sit_exit_net,
1210         .id   = &sit_net_id,
1211         .size = sizeof(struct sit_net),
1212 };
1213
1214 static void __exit sit_cleanup(void)
1215 {
1216         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1217
1218         unregister_pernet_device(&sit_net_ops);
1219         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1220 }
1221
1222 static int __init sit_init(void)
1223 {
1224         int err;
1225
1226         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1227
1228         err = register_pernet_device(&sit_net_ops);
1229         if (err < 0)
1230                 return err;
1231         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1232         if (err < 0) {
1233                 unregister_pernet_device(&sit_net_ops);
1234                 printk(KERN_INFO "sit init: Can't add protocol\n");
1235         }
1236         return err;
1237 }
1238
1239 module_init(sit_init);
1240 module_exit(sit_cleanup);
1241 MODULE_LICENSE("GPL");
1242 MODULE_ALIAS("sit0");