SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/sit.c

   1 /*
   2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
  10  *
  11  *      This program is free software; you can redistribute it and/or
  12  *      modify it under the terms of the GNU General Public License
  13  *      as published by the Free Software Foundation; either version
  14  *      2 of the License, or (at your option) any later version.
  15  *
  16  *      Changes:
  17  * Roger Venning <r.venning@telstra.com>:       6to4 support
  18  * Nate Thompson <nate@thebog.net>:             6to4 support
  19  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  20  */
  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 <asm/uaccess.h>
  34 #include <linux/init.h>
  35 #include <linux/netfilter_ipv4.h>
  36 #include <linux/if_ether.h>
  37
  38 #include <net/sock.h>
  39 #include <net/snmp.h>
  40
  41 #include <net/ipv6.h>
  42 #include <net/protocol.h>
  43 #include <net/transp_v6.h>
  44 #include <net/ip6_fib.h>
  45 #include <net/ip6_route.h>
  46 #include <net/ndisc.h>
  47 #include <net/addrconf.h>
  48 #include <net/ip.h>
  49 #include <net/udp.h>
  50 #include <net/icmp.h>
  51 #include <net/ipip.h>
  52 #include <net/inet_ecn.h>
  53 #include <net/xfrm.h>
  54 #include <net/dsfield.h>
  55
  56 /*
  57    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  58
  59    For comments look at net/ipv4/ip_gre.c --ANK
  60  */
  61
  62 #define HASH_SIZE  16
  63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  64
  65 static int ipip6_fb_tunnel_init(struct net_device *dev);
  66 static int ipip6_tunnel_init(struct net_device *dev);
  67 static void ipip6_tunnel_setup(struct net_device *dev);
  68
  69 static struct net_device *ipip6_fb_tunnel_dev;
  70
  71 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  72 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  73 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  74 static struct ip_tunnel *tunnels_wc[1];
  75 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  76
  77 static DEFINE_RWLOCK(ipip6_lock);
  78
  79 static struct ip_tunnel * ipip6_tunnel_lookup(__be32 remote, __be32 local)
  80 {
  81         unsigned h0 = HASH(remote);
  82         unsigned h1 = HASH(local);
  83         struct ip_tunnel *t;
  84
  85         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  86                 if (local == t->parms.iph.saddr &&
  87                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  88                         return t;
  89         }
  90         for (t = tunnels_r[h0]; t; t = t->next) {
  91                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  92                         return t;
  93         }
  94         for (t = tunnels_l[h1]; t; t = t->next) {
  95                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  96                         return t;
  97         }
  98         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
  99                 return t;
 100         return NULL;
 101 }
 102
 103 static struct ip_tunnel **__ipip6_bucket(struct ip_tunnel_parm *parms)
 104 {
 105         __be32 remote = parms->iph.daddr;
 106         __be32 local = parms->iph.saddr;
 107         unsigned h = 0;
 108         int prio = 0;
 109
 110         if (remote) {
 111                 prio |= 2;
 112                 h ^= HASH(remote);
 113         }
 114         if (local) {
 115                 prio |= 1;
 116                 h ^= HASH(local);
 117         }
 118         return &tunnels[prio][h];
 119 }
 120
 121 static inline struct ip_tunnel **ipip6_bucket(struct ip_tunnel *t)
 122 {
 123         return __ipip6_bucket(&t->parms);
 124 }
 125
 126 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 127 {
 128         struct ip_tunnel **tp;
 129
 130         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 131                 if (t == *tp) {
 132                         write_lock_bh(&ipip6_lock);
 133                         *tp = t->next;
 134                         write_unlock_bh(&ipip6_lock);
 135                         break;
 136                 }
 137         }
 138 }
 139
 140 static void ipip6_tunnel_link(struct ip_tunnel *t)
 141 {
 142         struct ip_tunnel **tp = ipip6_bucket(t);
 143
 144         t->next = *tp;
 145         write_lock_bh(&ipip6_lock);
 146         *tp = t;
 147         write_unlock_bh(&ipip6_lock);
 148 }
 149
 150 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 151 {
 152         __be32 remote = parms->iph.daddr;
 153         __be32 local = parms->iph.saddr;
 154         struct ip_tunnel *t, **tp, *nt;
 155         struct net_device *dev;
 156         char name[IFNAMSIZ];
 157
 158         for (tp = __ipip6_bucket(parms); (t = *tp) != NULL; tp = &t->next) {
 159                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 160                         return t;
 161         }
 162         if (!create)
 163                 goto failed;
 164
 165         if (parms->name[0])
 166                 strlcpy(name, parms->name, IFNAMSIZ);
 167         else
 168                 sprintf(name, "sit%%d");
 169
 170         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 171         if (dev == NULL)
 172                 return NULL;
 173
 174         if (strchr(name, '%')) {
 175                 if (dev_alloc_name(dev, name) < 0)
 176                         goto failed_free;
 177         }
 178
 179         nt = netdev_priv(dev);
 180         dev->init = ipip6_tunnel_init;
 181         nt->parms = *parms;
 182
 183         if (parms->i_flags & SIT_ISATAP)
 184                 dev->priv_flags |= IFF_ISATAP;
 185
 186         if (register_netdevice(dev) < 0)
 187                 goto failed_free;
 188
 189         dev_hold(dev);
 190
 191         ipip6_tunnel_link(nt);
 192         return nt;
 193
 194 failed_free:
 195         free_netdev(dev);
 196 failed:
 197         return NULL;
 198 }
 199
 200 static struct ip_tunnel_prl_entry *
 201 ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 202 {
 203         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 204
 205         for (p = t->prl; p; p = p->next)
 206                 if (p->entry.addr == addr)
 207                         break;
 208         return p;
 209
 210 }
 211
 212 static int
 213 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 214 {
 215         struct ip_tunnel_prl_entry *p;
 216
 217         for (p = t->prl; p; p = p->next) {
 218                 if (p->entry.addr == a->addr) {
 219                         if (chg) {
 220                                 p->entry = *a;
 221                                 return 0;
 222                         }
 223                         return -EEXIST;
 224                 }
 225         }
 226
 227         if (chg)
 228                 return -ENXIO;
 229
 230         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 231         if (!p)
 232                 return -ENOBUFS;
 233
 234         p->entry = *a;
 235         p->next = t->prl;
 236         t->prl = p;
 237         return 0;
 238 }
 239
 240 static int
 241 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 242 {
 243         struct ip_tunnel_prl_entry *x, **p;
 244
 245         if (a) {
 246                 for (p = &t->prl; *p; p = &(*p)->next) {
 247                         if ((*p)->entry.addr == a->addr) {
 248                                 x = *p;
 249                                 *p = x->next;
 250                                 kfree(x);
 251                                 return 0;
 252                         }
 253                 }
 254                 return -ENXIO;
 255         } else {
 256                 while (t->prl) {
 257                         x = t->prl;
 258                         t->prl = t->prl->next;
 259                         kfree(x);
 260                 }
 261         }
 262         return 0;
 263 }
 264
 265 /* copied directly from anycast.c */
 266 static int
 267 ipip6_onlink(struct in6_addr *addr, struct net_device *dev)
 268 {
 269         struct inet6_dev        *idev;
 270         struct inet6_ifaddr     *ifa;
 271         int     onlink;
 272
 273         onlink = 0;
 274         rcu_read_lock();
 275         idev = __in6_dev_get(dev);
 276         if (idev) {
 277                 read_lock_bh(&idev->lock);
 278                 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 279                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
 280                                                    ifa->prefix_len);
 281                         if (onlink)
 282                                 break;
 283                 }
 284                 read_unlock_bh(&idev->lock);
 285         }
 286         rcu_read_unlock();
 287         return onlink;
 288 }
 289
 290 static int
 291 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 292 {
 293         struct ip_tunnel_prl_entry *p = ipip6_tunnel_locate_prl(t, iph->saddr);
 294         int ok = 1;
 295
 296         if (p) {
 297                 if (p->entry.flags & PRL_DEFAULT)
 298                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 299                 else
 300                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 301         } else {
 302                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 303                 if (ipv6_addr_is_isatap(addr6) &&
 304                     (addr6->s6_addr32[3] == iph->saddr) &&
 305                     ipip6_onlink(addr6, t->dev))
 306                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 307                 else
 308                         ok = 0;
 309         }
 310         return ok;
 311 }
 312
 313 static void ipip6_tunnel_uninit(struct net_device *dev)
 314 {
 315         if (dev == ipip6_fb_tunnel_dev) {
 316                 write_lock_bh(&ipip6_lock);
 317                 tunnels_wc[0] = NULL;
 318                 write_unlock_bh(&ipip6_lock);
 319                 dev_put(dev);
 320         } else {
 321                 ipip6_tunnel_unlink(netdev_priv(dev));
 322                 ipip6_tunnel_del_prl(netdev_priv(dev), 0);
 323                 dev_put(dev);
 324         }
 325 }
 326
 327
 328 static int ipip6_err(struct sk_buff *skb, u32 info)
 329 {
 330 #ifndef I_WISH_WORLD_WERE_PERFECT
 331
 332 /* It is not :-( All the routers (except for Linux) return only
 333    8 bytes of packet payload. It means, that precise relaying of
 334    ICMP in the real Internet is absolutely infeasible.
 335  */
 336         struct iphdr *iph = (struct iphdr*)skb->data;
 337         const int type = icmp_hdr(skb)->type;
 338         const int code = icmp_hdr(skb)->code;
 339         struct ip_tunnel *t;
 340         int err;
 341
 342         switch (type) {
 343         default:
 344         case ICMP_PARAMETERPROB:
 345                 return 0;
 346
 347         case ICMP_DEST_UNREACH:
 348                 switch (code) {
 349                 case ICMP_SR_FAILED:
 350                 case ICMP_PORT_UNREACH:
 351                         /* Impossible event. */
 352                         return 0;
 353                 case ICMP_FRAG_NEEDED:
 354                         /* Soft state for pmtu is maintained by IP core. */
 355                         return 0;
 356                 default:
 357                         /* All others are translated to HOST_UNREACH.
 358                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 359                            I believe they are just ether pollution. --ANK
 360                          */
 361                         break;
 362                 }
 363                 break;
 364         case ICMP_TIME_EXCEEDED:
 365                 if (code != ICMP_EXC_TTL)
 366                         return 0;
 367                 break;
 368         }
 369
 370         err = -ENOENT;
 371
 372         read_lock(&ipip6_lock);
 373         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 374         if (t == NULL || t->parms.iph.daddr == 0)
 375                 goto out;
 376
 377         err = 0;
 378         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 379                 goto out;
 380
 381         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 382                 t->err_count++;
 383         else
 384                 t->err_count = 1;
 385         t->err_time = jiffies;
 386 out:
 387         read_unlock(&ipip6_lock);
 388         return err;
 389 #else
 390         struct iphdr *iph = (struct iphdr*)dp;
 391         int hlen = iph->ihl<<2;
 392         struct ipv6hdr *iph6;
 393         const int type = icmp_hdr(skb)->type;
 394         const int code = icmp_hdr(skb)->code;
 395         int rel_type = 0;
 396         int rel_code = 0;
 397         int rel_info = 0;
 398         struct sk_buff *skb2;
 399         struct rt6_info *rt6i;
 400
 401         if (len < hlen + sizeof(struct ipv6hdr))
 402                 return;
 403         iph6 = (struct ipv6hdr*)(dp + hlen);
 404
 405         switch (type) {
 406         default:
 407                 return;
 408         case ICMP_PARAMETERPROB:
 409                 if (icmp_hdr(skb)->un.gateway < hlen)
 410                         return;
 411
 412                 /* So... This guy found something strange INSIDE encapsulated
 413                    packet. Well, he is fool, but what can we do ?
 414                  */
 415                 rel_type = ICMPV6_PARAMPROB;
 416                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 417                 break;
 418
 419         case ICMP_DEST_UNREACH:
 420                 switch (code) {
 421                 case ICMP_SR_FAILED:
 422                 case ICMP_PORT_UNREACH:
 423                         /* Impossible event. */
 424                         return;
 425                 case ICMP_FRAG_NEEDED:
 426                         /* Too complicated case ... */
 427                         return;
 428                 default:
 429                         /* All others are translated to HOST_UNREACH.
 430                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 431                            I believe, it is just ether pollution. --ANK
 432                          */
 433                         rel_type = ICMPV6_DEST_UNREACH;
 434                         rel_code = ICMPV6_ADDR_UNREACH;
 435                         break;
 436                 }
 437                 break;
 438         case ICMP_TIME_EXCEEDED:
 439                 if (code != ICMP_EXC_TTL)
 440                         return;
 441                 rel_type = ICMPV6_TIME_EXCEED;
 442                 rel_code = ICMPV6_EXC_HOPLIMIT;
 443                 break;
 444         }
 445
 446         /* Prepare fake skb to feed it to icmpv6_send */
 447         skb2 = skb_clone(skb, GFP_ATOMIC);
 448         if (skb2 == NULL)
 449                 return 0;
 450         dst_release(skb2->dst);
 451         skb2->dst = NULL;
 452         skb_pull(skb2, skb->data - (u8*)iph6);
 453         skb_reset_network_header(skb2);
 454
 455         /* Try to guess incoming interface */
 456         rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
 457         if (rt6i && rt6i->rt6i_dev) {
 458                 skb2->dev = rt6i->rt6i_dev;
 459
 460                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 461
 462                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 463                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 464                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 465                                 rel_type = ICMPV6_DEST_UNREACH;
 466                                 rel_code = ICMPV6_ADDR_UNREACH;
 467                         }
 468                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 469                 }
 470         }
 471         kfree_skb(skb2);
 472         return 0;
 473 #endif
 474 }
 475
 476 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 477 {
 478         if (INET_ECN_is_ce(iph->tos))
 479                 IP6_ECN_set_ce(ipv6_hdr(skb));
 480 }
 481
 482 static int ipip6_rcv(struct sk_buff *skb)
 483 {
 484         struct iphdr *iph;
 485         struct ip_tunnel *tunnel;
 486
 487         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 488                 goto out;
 489
 490         iph = ip_hdr(skb);
 491
 492         read_lock(&ipip6_lock);
 493         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 494                 secpath_reset(skb);
 495                 skb->mac_header = skb->network_header;
 496                 skb_reset_network_header(skb);
 497                 IPCB(skb)->flags = 0;
 498                 skb->protocol = htons(ETH_P_IPV6);
 499                 skb->pkt_type = PACKET_HOST;
 500
 501                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 502                     !isatap_chksrc(skb, iph, tunnel)) {
 503                         tunnel->stat.rx_errors++;
 504                         read_unlock(&ipip6_lock);
 505                         kfree_skb(skb);
 506                         return 0;
 507                 }
 508                 tunnel->stat.rx_packets++;
 509                 tunnel->stat.rx_bytes += skb->len;
 510                 skb->dev = tunnel->dev;
 511                 dst_release(skb->dst);
 512                 skb->dst = NULL;
 513                 nf_reset(skb);
 514                 ipip6_ecn_decapsulate(iph, skb);
 515                 netif_rx(skb);
 516                 read_unlock(&ipip6_lock);
 517                 return 0;
 518         }
 519
 520         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 521         kfree_skb(skb);
 522         read_unlock(&ipip6_lock);
 523 out:
 524         return 0;
 525 }
 526
 527 /* Returns the embedded IPv4 address if the IPv6 address
 528    comes from 6to4 (RFC 3056) addr space */
 529
 530 static inline __be32 try_6to4(struct in6_addr *v6dst)
 531 {
 532         __be32 dst = 0;
 533
 534         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 535                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 536                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 537         }
 538         return dst;
 539 }
 540
 541 /*
 542  *      This function assumes it is being called from dev_queue_xmit()
 543  *      and that skb is filled properly by that function.
 544  */
 545
 546 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 547 {
 548         struct ip_tunnel *tunnel = netdev_priv(dev);
 549         struct net_device_stats *stats = &tunnel->stat;
 550         struct iphdr  *tiph = &tunnel->parms.iph;
 551         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 552         u8     tos = tunnel->parms.iph.tos;
 553         struct rtable *rt;                      /* Route to the other host */
 554         struct net_device *tdev;                        /* Device to other host */
 555         struct iphdr  *iph;                     /* Our new IP header */
 556         unsigned int max_headroom;              /* The extra header space needed */
 557         __be32 dst = tiph->daddr;
 558         int    mtu;
 559         struct in6_addr *addr6;
 560         int addr_type;
 561
 562         if (tunnel->recursion++) {
 563                 tunnel->stat.collisions++;
 564                 goto tx_error;
 565         }
 566
 567         if (skb->protocol != htons(ETH_P_IPV6))
 568                 goto tx_error;
 569
 570         /* ISATAP (RFC4214) - must come before 6to4 */
 571         if (dev->priv_flags & IFF_ISATAP) {
 572                 struct neighbour *neigh = NULL;
 573
 574                 if (skb->dst)
 575                         neigh = skb->dst->neighbour;
 576
 577                 if (neigh == NULL) {
 578                         if (net_ratelimit())
 579                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 580                         goto tx_error;
 581                 }
 582
 583                 addr6 = (struct in6_addr*)&neigh->primary_key;
 584                 addr_type = ipv6_addr_type(addr6);
 585
 586                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 587                      ipv6_addr_is_isatap(addr6))
 588                         dst = addr6->s6_addr32[3];
 589                 else
 590                         goto tx_error;
 591         }
 592
 593         if (!dst)
 594                 dst = try_6to4(&iph6->daddr);
 595
 596         if (!dst) {
 597                 struct neighbour *neigh = NULL;
 598
 599                 if (skb->dst)
 600                         neigh = skb->dst->neighbour;
 601
 602                 if (neigh == NULL) {
 603                         if (net_ratelimit())
 604                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 605                         goto tx_error;
 606                 }
 607
 608                 addr6 = (struct in6_addr*)&neigh->primary_key;
 609                 addr_type = ipv6_addr_type(addr6);
 610
 611                 if (addr_type == IPV6_ADDR_ANY) {
 612                         addr6 = &ipv6_hdr(skb)->daddr;
 613                         addr_type = ipv6_addr_type(addr6);
 614                 }
 615
 616                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 617                         goto tx_error_icmp;
 618
 619                 dst = addr6->s6_addr32[3];
 620         }
 621
 622         {
 623                 struct flowi fl = { .nl_u = { .ip4_u =
 624                                               { .daddr = dst,
 625                                                 .saddr = tiph->saddr,
 626                                                 .tos = RT_TOS(tos) } },
 627                                     .oif = tunnel->parms.link,
 628                                     .proto = IPPROTO_IPV6 };
 629                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 630                         tunnel->stat.tx_carrier_errors++;
 631                         goto tx_error_icmp;
 632                 }
 633         }
 634         if (rt->rt_type != RTN_UNICAST) {
 635                 ip_rt_put(rt);
 636                 tunnel->stat.tx_carrier_errors++;
 637                 goto tx_error_icmp;
 638         }
 639         tdev = rt->u.dst.dev;
 640
 641         if (tdev == dev) {
 642                 ip_rt_put(rt);
 643                 tunnel->stat.collisions++;
 644                 goto tx_error;
 645         }
 646
 647         if (tiph->frag_off)
 648                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 649         else
 650                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 651
 652         if (mtu < 68) {
 653                 tunnel->stat.collisions++;
 654                 ip_rt_put(rt);
 655                 goto tx_error;
 656         }
 657         if (mtu < IPV6_MIN_MTU)
 658                 mtu = IPV6_MIN_MTU;
 659         if (tunnel->parms.iph.daddr && skb->dst)
 660                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 661
 662         if (skb->len > mtu) {
 663                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 664                 ip_rt_put(rt);
 665                 goto tx_error;
 666         }
 667
 668         if (tunnel->err_count > 0) {
 669                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 670                         tunnel->err_count--;
 671                         dst_link_failure(skb);
 672                 } else
 673                         tunnel->err_count = 0;
 674         }
 675
 676         /*
 677          * Okay, now see if we can stuff it in the buffer as-is.
 678          */
 679         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 680
 681         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 682             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 683                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 684                 if (!new_skb) {
 685                         ip_rt_put(rt);
 686                         stats->tx_dropped++;
 687                         dev_kfree_skb(skb);
 688                         tunnel->recursion--;
 689                         return 0;
 690                 }
 691                 if (skb->sk)
 692                         skb_set_owner_w(new_skb, skb->sk);
 693                 dev_kfree_skb(skb);
 694                 skb = new_skb;
 695                 iph6 = ipv6_hdr(skb);
 696         }
 697
 698         skb->transport_header = skb->network_header;
 699         skb_push(skb, sizeof(struct iphdr));
 700         skb_reset_network_header(skb);
 701         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 702         IPCB(skb)->flags = 0;
 703         dst_release(skb->dst);
 704         skb->dst = &rt->u.dst;
 705
 706         /*
 707          *      Push down and install the IPIP header.
 708          */
 709
 710         iph                     =       ip_hdr(skb);
 711         iph->version            =       4;
 712         iph->ihl                =       sizeof(struct iphdr)>>2;
 713         if (mtu > IPV6_MIN_MTU)
 714                 iph->frag_off   =       htons(IP_DF);
 715         else
 716                 iph->frag_off   =       0;
 717
 718         iph->protocol           =       IPPROTO_IPV6;
 719         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 720         iph->daddr              =       rt->rt_dst;
 721         iph->saddr              =       rt->rt_src;
 722
 723         if ((iph->ttl = tiph->ttl) == 0)
 724                 iph->ttl        =       iph6->hop_limit;
 725
 726         nf_reset(skb);
 727
 728         IPTUNNEL_XMIT();
 729         tunnel->recursion--;
 730         return 0;
 731
 732 tx_error_icmp:
 733         dst_link_failure(skb);
 734 tx_error:
 735         stats->tx_errors++;
 736         dev_kfree_skb(skb);
 737         tunnel->recursion--;
 738         return 0;
 739 }
 740
 741 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 742 {
 743         struct net_device *tdev = NULL;
 744         struct ip_tunnel *tunnel;
 745         struct iphdr *iph;
 746
 747         tunnel = netdev_priv(dev);
 748         iph = &tunnel->parms.iph;
 749
 750         if (iph->daddr) {
 751                 struct flowi fl = { .nl_u = { .ip4_u =
 752                                               { .daddr = iph->daddr,
 753                                                 .saddr = iph->saddr,
 754                                                 .tos = RT_TOS(iph->tos) } },
 755                                     .oif = tunnel->parms.link,
 756                                     .proto = IPPROTO_IPV6 };
 757                 struct rtable *rt;
 758                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 759                         tdev = rt->u.dst.dev;
 760                         ip_rt_put(rt);
 761                 }
 762                 dev->flags |= IFF_POINTOPOINT;
 763         }
 764
 765         if (!tdev && tunnel->parms.link)
 766                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 767
 768         if (tdev) {
 769                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 770                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 771                 if (dev->mtu < IPV6_MIN_MTU)
 772                         dev->mtu = IPV6_MIN_MTU;
 773         }
 774         dev->iflink = tunnel->parms.link;
 775 }
 776
 777 static int
 778 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 779 {
 780         int err = 0;
 781         struct ip_tunnel_parm p;
 782         struct ip_tunnel_prl prl;
 783         struct ip_tunnel *t;
 784
 785         switch (cmd) {
 786         case SIOCGETTUNNEL:
 787                 t = NULL;
 788                 if (dev == ipip6_fb_tunnel_dev) {
 789                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 790                                 err = -EFAULT;
 791                                 break;
 792                         }
 793                         t = ipip6_tunnel_locate(&p, 0);
 794                 }
 795                 if (t == NULL)
 796                         t = netdev_priv(dev);
 797                 memcpy(&p, &t->parms, sizeof(p));
 798                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 799                         err = -EFAULT;
 800                 break;
 801
 802         case SIOCADDTUNNEL:
 803         case SIOCCHGTUNNEL:
 804                 err = -EPERM;
 805                 if (!capable(CAP_NET_ADMIN))
 806                         goto done;
 807
 808                 err = -EFAULT;
 809                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 810                         goto done;
 811
 812                 err = -EINVAL;
 813                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 814                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 815                         goto done;
 816                 if (p.iph.ttl)
 817                         p.iph.frag_off |= htons(IP_DF);
 818
 819                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 820
 821                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 822                         if (t != NULL) {
 823                                 if (t->dev != dev) {
 824                                         err = -EEXIST;
 825                                         break;
 826                                 }
 827                         } else {
 828                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 829                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 830                                         err = -EINVAL;
 831                                         break;
 832                                 }
 833                                 t = netdev_priv(dev);
 834                                 ipip6_tunnel_unlink(t);
 835                                 t->parms.iph.saddr = p.iph.saddr;
 836                                 t->parms.iph.daddr = p.iph.daddr;
 837                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 838                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 839                                 ipip6_tunnel_link(t);
 840                                 netdev_state_change(dev);
 841                         }
 842                 }
 843
 844                 if (t) {
 845                         err = 0;
 846                         if (cmd == SIOCCHGTUNNEL) {
 847                                 t->parms.iph.ttl = p.iph.ttl;
 848                                 t->parms.iph.tos = p.iph.tos;
 849                                 if (t->parms.link != p.link) {
 850                                         t->parms.link = p.link;
 851                                         ipip6_tunnel_bind_dev(dev);
 852                                         netdev_state_change(dev);
 853                                 }
 854                         }
 855                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 856                                 err = -EFAULT;
 857                 } else
 858                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 859                 break;
 860
 861         case SIOCDELTUNNEL:
 862                 err = -EPERM;
 863                 if (!capable(CAP_NET_ADMIN))
 864                         goto done;
 865
 866                 if (dev == ipip6_fb_tunnel_dev) {
 867                         err = -EFAULT;
 868                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 869                                 goto done;
 870                         err = -ENOENT;
 871                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 872                                 goto done;
 873                         err = -EPERM;
 874                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 875                                 goto done;
 876                         dev = t->dev;
 877                 }
 878                 unregister_netdevice(dev);
 879                 err = 0;
 880                 break;
 881
 882         case SIOCADDPRL:
 883         case SIOCDELPRL:
 884         case SIOCCHGPRL:
 885                 err = -EPERM;
 886                 if (!capable(CAP_NET_ADMIN))
 887                         goto done;
 888                 err = -EINVAL;
 889                 if (dev == ipip6_fb_tunnel_dev)
 890                         goto done;
 891                 err = -EFAULT;
 892                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 893                         goto done;
 894                 err = -ENOENT;
 895                 if (!(t = netdev_priv(dev)))
 896                         goto done;
 897
 898                 ipip6_tunnel_unlink(t);
 899                 if (cmd == SIOCDELPRL)
 900                         err = ipip6_tunnel_del_prl(t, &prl);
 901                 else
 902                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 903                 ipip6_tunnel_link(t);
 904                 netdev_state_change(dev);
 905                 break;
 906
 907         default:
 908                 err = -EINVAL;
 909         }
 910
 911 done:
 912         return err;
 913 }
 914
 915 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 916 {
 917         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 918 }
 919
 920 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 921 {
 922         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 923                 return -EINVAL;
 924         dev->mtu = new_mtu;
 925         return 0;
 926 }
 927
 928 static void ipip6_tunnel_setup(struct net_device *dev)
 929 {
 930         dev->uninit             = ipip6_tunnel_uninit;
 931         dev->destructor         = free_netdev;
 932         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 933         dev->get_stats          = ipip6_tunnel_get_stats;
 934         dev->do_ioctl           = ipip6_tunnel_ioctl;
 935         dev->change_mtu         = ipip6_tunnel_change_mtu;
 936
 937         dev->type               = ARPHRD_SIT;
 938         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 939         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 940         dev->flags              = IFF_NOARP;
 941         dev->iflink             = 0;
 942         dev->addr_len           = 4;
 943 }
 944
 945 static int ipip6_tunnel_init(struct net_device *dev)
 946 {
 947         struct ip_tunnel *tunnel;
 948
 949         tunnel = netdev_priv(dev);
 950
 951         tunnel->dev = dev;
 952         strcpy(tunnel->parms.name, dev->name);
 953
 954         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 955         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 956
 957         ipip6_tunnel_bind_dev(dev);
 958
 959         return 0;
 960 }
 961
 962 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 963 {
 964         struct ip_tunnel *tunnel = netdev_priv(dev);
 965         struct iphdr *iph = &tunnel->parms.iph;
 966
 967         tunnel->dev = dev;
 968         strcpy(tunnel->parms.name, dev->name);
 969
 970         iph->version            = 4;
 971         iph->protocol           = IPPROTO_IPV6;
 972         iph->ihl                = 5;
 973         iph->ttl                = 64;
 974
 975         dev_hold(dev);
 976         tunnels_wc[0]           = tunnel;
 977         return 0;
 978 }
 979
 980 static struct xfrm_tunnel sit_handler = {
 981         .handler        =       ipip6_rcv,
 982         .err_handler    =       ipip6_err,
 983         .priority       =       1,
 984 };
 985
 986 static void __exit sit_destroy_tunnels(void)
 987 {
 988         int prio;
 989
 990         for (prio = 1; prio < 4; prio++) {
 991                 int h;
 992                 for (h = 0; h < HASH_SIZE; h++) {
 993                         struct ip_tunnel *t;
 994                         while ((t = tunnels[prio][h]) != NULL)
 995                                 unregister_netdevice(t->dev);
 996                 }
 997         }
 998 }
 999
1000 static void __exit sit_cleanup(void)
1001 {
1002         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1003
1004         rtnl_lock();
1005         sit_destroy_tunnels();
1006         unregister_netdevice(ipip6_fb_tunnel_dev);
1007         rtnl_unlock();
1008 }
1009
1010 static int __init sit_init(void)
1011 {
1012         int err;
1013
1014         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1015
1016         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1017                 printk(KERN_INFO "sit init: Can't add protocol\n");
1018                 return -EAGAIN;
1019         }
1020
1021         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1022                                            ipip6_tunnel_setup);
1023         if (!ipip6_fb_tunnel_dev) {
1024                 err = -ENOMEM;
1025                 goto err1;
1026         }
1027
1028         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1029
1030         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
1031                 goto err2;
1032
1033  out:
1034         return err;
1035  err2:
1036         free_netdev(ipip6_fb_tunnel_dev);
1037  err1:
1038         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1039         goto out;
1040 }
1041
1042 module_init(sit_init);
1043 module_exit(sit_cleanup);
1044 MODULE_LICENSE("GPL");
1045 MODULE_ALIAS("sit0");