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  */
  20
  21 #include <linux/module.h>
  22 #include <linux/capability.h>
  23 #include <linux/errno.h>
  24 #include <linux/types.h>
  25 #include <linux/socket.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/in6.h>
  29 #include <linux/netdevice.h>
  30 #include <linux/if_arp.h>
  31 #include <linux/icmp.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
  55 /*
  56    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  57
  58    For comments look at net/ipv4/ip_gre.c --ANK
  59  */
  60
  61 #define HASH_SIZE  16
  62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  63
  64 static int ipip6_fb_tunnel_init(struct net_device *dev);
  65 static int ipip6_tunnel_init(struct net_device *dev);
  66 static void ipip6_tunnel_setup(struct net_device *dev);
  67
  68 static struct net_device *ipip6_fb_tunnel_dev;
  69
  70 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  71 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  72 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  73 static struct ip_tunnel *tunnels_wc[1];
  74 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  75
  76 static DEFINE_RWLOCK(ipip6_lock);
  77
  78 static struct ip_tunnel * ipip6_tunnel_lookup(__be32 remote, __be32 local)
  79 {
  80         unsigned h0 = HASH(remote);
  81         unsigned h1 = HASH(local);
  82         struct ip_tunnel *t;
  83
  84         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  85                 if (local == t->parms.iph.saddr &&
  86                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  87                         return t;
  88         }
  89         for (t = tunnels_r[h0]; t; t = t->next) {
  90                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  91                         return t;
  92         }
  93         for (t = tunnels_l[h1]; t; t = t->next) {
  94                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  95                         return t;
  96         }
  97         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
  98                 return t;
  99         return NULL;
 100 }
 101
 102 static struct ip_tunnel **__ipip6_bucket(struct ip_tunnel_parm *parms)
 103 {
 104         __be32 remote = parms->iph.daddr;
 105         __be32 local = parms->iph.saddr;
 106         unsigned h = 0;
 107         int prio = 0;
 108
 109         if (remote) {
 110                 prio |= 2;
 111                 h ^= HASH(remote);
 112         }
 113         if (local) {
 114                 prio |= 1;
 115                 h ^= HASH(local);
 116         }
 117         return &tunnels[prio][h];
 118 }
 119
 120 static inline struct ip_tunnel **ipip6_bucket(struct ip_tunnel *t)
 121 {
 122         return __ipip6_bucket(&t->parms);
 123 }
 124
 125 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 126 {
 127         struct ip_tunnel **tp;
 128
 129         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 130                 if (t == *tp) {
 131                         write_lock_bh(&ipip6_lock);
 132                         *tp = t->next;
 133                         write_unlock_bh(&ipip6_lock);
 134                         break;
 135                 }
 136         }
 137 }
 138
 139 static void ipip6_tunnel_link(struct ip_tunnel *t)
 140 {
 141         struct ip_tunnel **tp = ipip6_bucket(t);
 142
 143         t->next = *tp;
 144         write_lock_bh(&ipip6_lock);
 145         *tp = t;
 146         write_unlock_bh(&ipip6_lock);
 147 }
 148
 149 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 150 {
 151         __be32 remote = parms->iph.daddr;
 152         __be32 local = parms->iph.saddr;
 153         struct ip_tunnel *t, **tp, *nt;
 154         struct net_device *dev;
 155         char name[IFNAMSIZ];
 156
 157         for (tp = __ipip6_bucket(parms); (t = *tp) != NULL; tp = &t->next) {
 158                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 159                         return t;
 160         }
 161         if (!create)
 162                 goto failed;
 163
 164         if (parms->name[0])
 165                 strlcpy(name, parms->name, IFNAMSIZ);
 166         else {
 167                 int i;
 168                 for (i=1; i<100; i++) {
 169                         sprintf(name, "sit%d", i);
 170                         if (__dev_get_by_name(&init_net, name) == NULL)
 171                                 break;
 172                 }
 173                 if (i==100)
 174                         goto failed;
 175         }
 176
 177         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 178         if (dev == NULL)
 179                 return NULL;
 180
 181         nt = netdev_priv(dev);
 182         dev->init = ipip6_tunnel_init;
 183         nt->parms = *parms;
 184
 185         if (register_netdevice(dev) < 0) {
 186                 free_netdev(dev);
 187                 goto failed;
 188         }
 189
 190         dev_hold(dev);
 191
 192         ipip6_tunnel_link(nt);
 193         return nt;
 194
 195 failed:
 196         return NULL;
 197 }
 198
 199 static void ipip6_tunnel_uninit(struct net_device *dev)
 200 {
 201         if (dev == ipip6_fb_tunnel_dev) {
 202                 write_lock_bh(&ipip6_lock);
 203                 tunnels_wc[0] = NULL;
 204                 write_unlock_bh(&ipip6_lock);
 205                 dev_put(dev);
 206         } else {
 207                 ipip6_tunnel_unlink(netdev_priv(dev));
 208                 dev_put(dev);
 209         }
 210 }
 211
 212
 213 static int ipip6_err(struct sk_buff *skb, u32 info)
 214 {
 215 #ifndef I_WISH_WORLD_WERE_PERFECT
 216
 217 /* It is not :-( All the routers (except for Linux) return only
 218    8 bytes of packet payload. It means, that precise relaying of
 219    ICMP in the real Internet is absolutely infeasible.
 220  */
 221         struct iphdr *iph = (struct iphdr*)skb->data;
 222         const int type = icmp_hdr(skb)->type;
 223         const int code = icmp_hdr(skb)->code;
 224         struct ip_tunnel *t;
 225         int err;
 226
 227         switch (type) {
 228         default:
 229         case ICMP_PARAMETERPROB:
 230                 return 0;
 231
 232         case ICMP_DEST_UNREACH:
 233                 switch (code) {
 234                 case ICMP_SR_FAILED:
 235                 case ICMP_PORT_UNREACH:
 236                         /* Impossible event. */
 237                         return 0;
 238                 case ICMP_FRAG_NEEDED:
 239                         /* Soft state for pmtu is maintained by IP core. */
 240                         return 0;
 241                 default:
 242                         /* All others are translated to HOST_UNREACH.
 243                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 244                            I believe they are just ether pollution. --ANK
 245                          */
 246                         break;
 247                 }
 248                 break;
 249         case ICMP_TIME_EXCEEDED:
 250                 if (code != ICMP_EXC_TTL)
 251                         return 0;
 252                 break;
 253         }
 254
 255         err = -ENOENT;
 256
 257         read_lock(&ipip6_lock);
 258         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 259         if (t == NULL || t->parms.iph.daddr == 0)
 260                 goto out;
 261
 262         err = 0;
 263         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 264                 goto out;
 265
 266         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 267                 t->err_count++;
 268         else
 269                 t->err_count = 1;
 270         t->err_time = jiffies;
 271 out:
 272         read_unlock(&ipip6_lock);
 273         return err;
 274 #else
 275         struct iphdr *iph = (struct iphdr*)dp;
 276         int hlen = iph->ihl<<2;
 277         struct ipv6hdr *iph6;
 278         const int type = icmp_hdr(skb)->type;
 279         const int code = icmp_hdr(skb)->code;
 280         int rel_type = 0;
 281         int rel_code = 0;
 282         int rel_info = 0;
 283         struct sk_buff *skb2;
 284         struct rt6_info *rt6i;
 285
 286         if (len < hlen + sizeof(struct ipv6hdr))
 287                 return;
 288         iph6 = (struct ipv6hdr*)(dp + hlen);
 289
 290         switch (type) {
 291         default:
 292                 return;
 293         case ICMP_PARAMETERPROB:
 294                 if (icmp_hdr(skb)->un.gateway < hlen)
 295                         return;
 296
 297                 /* So... This guy found something strange INSIDE encapsulated
 298                    packet. Well, he is fool, but what can we do ?
 299                  */
 300                 rel_type = ICMPV6_PARAMPROB;
 301                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 302                 break;
 303
 304         case ICMP_DEST_UNREACH:
 305                 switch (code) {
 306                 case ICMP_SR_FAILED:
 307                 case ICMP_PORT_UNREACH:
 308                         /* Impossible event. */
 309                         return;
 310                 case ICMP_FRAG_NEEDED:
 311                         /* Too complicated case ... */
 312                         return;
 313                 default:
 314                         /* All others are translated to HOST_UNREACH.
 315                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 316                            I believe, it is just ether pollution. --ANK
 317                          */
 318                         rel_type = ICMPV6_DEST_UNREACH;
 319                         rel_code = ICMPV6_ADDR_UNREACH;
 320                         break;
 321                 }
 322                 break;
 323         case ICMP_TIME_EXCEEDED:
 324                 if (code != ICMP_EXC_TTL)
 325                         return;
 326                 rel_type = ICMPV6_TIME_EXCEED;
 327                 rel_code = ICMPV6_EXC_HOPLIMIT;
 328                 break;
 329         }
 330
 331         /* Prepare fake skb to feed it to icmpv6_send */
 332         skb2 = skb_clone(skb, GFP_ATOMIC);
 333         if (skb2 == NULL)
 334                 return 0;
 335         dst_release(skb2->dst);
 336         skb2->dst = NULL;
 337         skb_pull(skb2, skb->data - (u8*)iph6);
 338         skb_reset_network_header(skb2);
 339
 340         /* Try to guess incoming interface */
 341         rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
 342         if (rt6i && rt6i->rt6i_dev) {
 343                 skb2->dev = rt6i->rt6i_dev;
 344
 345                 rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
 346
 347                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 348                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 349                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 350                                 rel_type = ICMPV6_DEST_UNREACH;
 351                                 rel_code = ICMPV6_ADDR_UNREACH;
 352                         }
 353                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 354                 }
 355         }
 356         kfree_skb(skb2);
 357         return 0;
 358 #endif
 359 }
 360
 361 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 362 {
 363         if (INET_ECN_is_ce(iph->tos))
 364                 IP6_ECN_set_ce(ipv6_hdr(skb));
 365 }
 366
 367 static int ipip6_rcv(struct sk_buff *skb)
 368 {
 369         struct iphdr *iph;
 370         struct ip_tunnel *tunnel;
 371
 372         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 373                 goto out;
 374
 375         iph = ip_hdr(skb);
 376
 377         read_lock(&ipip6_lock);
 378         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 379                 secpath_reset(skb);
 380                 skb->mac_header = skb->network_header;
 381                 skb_reset_network_header(skb);
 382                 IPCB(skb)->flags = 0;
 383                 skb->protocol = htons(ETH_P_IPV6);
 384                 skb->pkt_type = PACKET_HOST;
 385                 tunnel->stat.rx_packets++;
 386                 tunnel->stat.rx_bytes += skb->len;
 387                 skb->dev = tunnel->dev;
 388                 dst_release(skb->dst);
 389                 skb->dst = NULL;
 390                 nf_reset(skb);
 391                 ipip6_ecn_decapsulate(iph, skb);
 392                 netif_rx(skb);
 393                 read_unlock(&ipip6_lock);
 394                 return 0;
 395         }
 396
 397         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 398         kfree_skb(skb);
 399         read_unlock(&ipip6_lock);
 400 out:
 401         return 0;
 402 }
 403
 404 /* Returns the embedded IPv4 address if the IPv6 address
 405    comes from 6to4 (RFC 3056) addr space */
 406
 407 static inline __be32 try_6to4(struct in6_addr *v6dst)
 408 {
 409         __be32 dst = 0;
 410
 411         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 412                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 413                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 414         }
 415         return dst;
 416 }
 417
 418 /*
 419  *      This function assumes it is being called from dev_queue_xmit()
 420  *      and that skb is filled properly by that function.
 421  */
 422
 423 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 424 {
 425         struct ip_tunnel *tunnel = netdev_priv(dev);
 426         struct net_device_stats *stats = &tunnel->stat;
 427         struct iphdr  *tiph = &tunnel->parms.iph;
 428         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 429         u8     tos = tunnel->parms.iph.tos;
 430         struct rtable *rt;                      /* Route to the other host */
 431         struct net_device *tdev;                        /* Device to other host */
 432         struct iphdr  *iph;                     /* Our new IP header */
 433         unsigned int max_headroom;              /* The extra header space needed */
 434         __be32 dst = tiph->daddr;
 435         int    mtu;
 436         struct in6_addr *addr6;
 437         int addr_type;
 438
 439         if (tunnel->recursion++) {
 440                 tunnel->stat.collisions++;
 441                 goto tx_error;
 442         }
 443
 444         if (skb->protocol != htons(ETH_P_IPV6))
 445                 goto tx_error;
 446
 447         if (!dst)
 448                 dst = try_6to4(&iph6->daddr);
 449
 450         if (!dst) {
 451                 struct neighbour *neigh = NULL;
 452
 453                 if (skb->dst)
 454                         neigh = skb->dst->neighbour;
 455
 456                 if (neigh == NULL) {
 457                         if (net_ratelimit())
 458                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 459                         goto tx_error;
 460                 }
 461
 462                 addr6 = (struct in6_addr*)&neigh->primary_key;
 463                 addr_type = ipv6_addr_type(addr6);
 464
 465                 if (addr_type == IPV6_ADDR_ANY) {
 466                         addr6 = &ipv6_hdr(skb)->daddr;
 467                         addr_type = ipv6_addr_type(addr6);
 468                 }
 469
 470                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 471                         goto tx_error_icmp;
 472
 473                 dst = addr6->s6_addr32[3];
 474         }
 475
 476         {
 477                 struct flowi fl = { .nl_u = { .ip4_u =
 478                                               { .daddr = dst,
 479                                                 .saddr = tiph->saddr,
 480                                                 .tos = RT_TOS(tos) } },
 481                                     .oif = tunnel->parms.link,
 482                                     .proto = IPPROTO_IPV6 };
 483                 if (ip_route_output_key(&rt, &fl)) {
 484                         tunnel->stat.tx_carrier_errors++;
 485                         goto tx_error_icmp;
 486                 }
 487         }
 488         if (rt->rt_type != RTN_UNICAST) {
 489                 ip_rt_put(rt);
 490                 tunnel->stat.tx_carrier_errors++;
 491                 goto tx_error_icmp;
 492         }
 493         tdev = rt->u.dst.dev;
 494
 495         if (tdev == dev) {
 496                 ip_rt_put(rt);
 497                 tunnel->stat.collisions++;
 498                 goto tx_error;
 499         }
 500
 501         if (tiph->frag_off)
 502                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 503         else
 504                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 505
 506         if (mtu < 68) {
 507                 tunnel->stat.collisions++;
 508                 ip_rt_put(rt);
 509                 goto tx_error;
 510         }
 511         if (mtu < IPV6_MIN_MTU)
 512                 mtu = IPV6_MIN_MTU;
 513         if (tunnel->parms.iph.daddr && skb->dst)
 514                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 515
 516         if (skb->len > mtu) {
 517                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 518                 ip_rt_put(rt);
 519                 goto tx_error;
 520         }
 521
 522         if (tunnel->err_count > 0) {
 523                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 524                         tunnel->err_count--;
 525                         dst_link_failure(skb);
 526                 } else
 527                         tunnel->err_count = 0;
 528         }
 529
 530         /*
 531          * Okay, now see if we can stuff it in the buffer as-is.
 532          */
 533         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 534
 535         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 536             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 537                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 538                 if (!new_skb) {
 539                         ip_rt_put(rt);
 540                         stats->tx_dropped++;
 541                         dev_kfree_skb(skb);
 542                         tunnel->recursion--;
 543                         return 0;
 544                 }
 545                 if (skb->sk)
 546                         skb_set_owner_w(new_skb, skb->sk);
 547                 dev_kfree_skb(skb);
 548                 skb = new_skb;
 549                 iph6 = ipv6_hdr(skb);
 550         }
 551
 552         skb->transport_header = skb->network_header;
 553         skb_push(skb, sizeof(struct iphdr));
 554         skb_reset_network_header(skb);
 555         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 556         IPCB(skb)->flags = 0;
 557         dst_release(skb->dst);
 558         skb->dst = &rt->u.dst;
 559
 560         /*
 561          *      Push down and install the IPIP header.
 562          */
 563
 564         iph                     =       ip_hdr(skb);
 565         iph->version            =       4;
 566         iph->ihl                =       sizeof(struct iphdr)>>2;
 567         if (mtu > IPV6_MIN_MTU)
 568                 iph->frag_off   =       htons(IP_DF);
 569         else
 570                 iph->frag_off   =       0;
 571
 572         iph->protocol           =       IPPROTO_IPV6;
 573         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 574         iph->daddr              =       rt->rt_dst;
 575         iph->saddr              =       rt->rt_src;
 576
 577         if ((iph->ttl = tiph->ttl) == 0)
 578                 iph->ttl        =       iph6->hop_limit;
 579
 580         nf_reset(skb);
 581
 582         IPTUNNEL_XMIT();
 583         tunnel->recursion--;
 584         return 0;
 585
 586 tx_error_icmp:
 587         dst_link_failure(skb);
 588 tx_error:
 589         stats->tx_errors++;
 590         dev_kfree_skb(skb);
 591         tunnel->recursion--;
 592         return 0;
 593 }
 594
 595 static int
 596 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 597 {
 598         int err = 0;
 599         struct ip_tunnel_parm p;
 600         struct ip_tunnel *t;
 601
 602         switch (cmd) {
 603         case SIOCGETTUNNEL:
 604                 t = NULL;
 605                 if (dev == ipip6_fb_tunnel_dev) {
 606                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 607                                 err = -EFAULT;
 608                                 break;
 609                         }
 610                         t = ipip6_tunnel_locate(&p, 0);
 611                 }
 612                 if (t == NULL)
 613                         t = netdev_priv(dev);
 614                 memcpy(&p, &t->parms, sizeof(p));
 615                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 616                         err = -EFAULT;
 617                 break;
 618
 619         case SIOCADDTUNNEL:
 620         case SIOCCHGTUNNEL:
 621                 err = -EPERM;
 622                 if (!capable(CAP_NET_ADMIN))
 623                         goto done;
 624
 625                 err = -EFAULT;
 626                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 627                         goto done;
 628
 629                 err = -EINVAL;
 630                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 631                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 632                         goto done;
 633                 if (p.iph.ttl)
 634                         p.iph.frag_off |= htons(IP_DF);
 635
 636                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 637
 638                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 639                         if (t != NULL) {
 640                                 if (t->dev != dev) {
 641                                         err = -EEXIST;
 642                                         break;
 643                                 }
 644                         } else {
 645                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 646                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 647                                         err = -EINVAL;
 648                                         break;
 649                                 }
 650                                 t = netdev_priv(dev);
 651                                 ipip6_tunnel_unlink(t);
 652                                 t->parms.iph.saddr = p.iph.saddr;
 653                                 t->parms.iph.daddr = p.iph.daddr;
 654                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 655                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 656                                 ipip6_tunnel_link(t);
 657                                 netdev_state_change(dev);
 658                         }
 659                 }
 660
 661                 if (t) {
 662                         err = 0;
 663                         if (cmd == SIOCCHGTUNNEL) {
 664                                 t->parms.iph.ttl = p.iph.ttl;
 665                                 t->parms.iph.tos = p.iph.tos;
 666                         }
 667                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 668                                 err = -EFAULT;
 669                 } else
 670                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 671                 break;
 672
 673         case SIOCDELTUNNEL:
 674                 err = -EPERM;
 675                 if (!capable(CAP_NET_ADMIN))
 676                         goto done;
 677
 678                 if (dev == ipip6_fb_tunnel_dev) {
 679                         err = -EFAULT;
 680                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 681                                 goto done;
 682                         err = -ENOENT;
 683                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 684                                 goto done;
 685                         err = -EPERM;
 686                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 687                                 goto done;
 688                         dev = t->dev;
 689                 }
 690                 unregister_netdevice(dev);
 691                 err = 0;
 692                 break;
 693
 694         default:
 695                 err = -EINVAL;
 696         }
 697
 698 done:
 699         return err;
 700 }
 701
 702 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 703 {
 704         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 705 }
 706
 707 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 708 {
 709         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 710                 return -EINVAL;
 711         dev->mtu = new_mtu;
 712         return 0;
 713 }
 714
 715 static void ipip6_tunnel_setup(struct net_device *dev)
 716 {
 717         dev->uninit             = ipip6_tunnel_uninit;
 718         dev->destructor         = free_netdev;
 719         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 720         dev->get_stats          = ipip6_tunnel_get_stats;
 721         dev->do_ioctl           = ipip6_tunnel_ioctl;
 722         dev->change_mtu         = ipip6_tunnel_change_mtu;
 723
 724         dev->type               = ARPHRD_SIT;
 725         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 726         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 727         dev->flags              = IFF_NOARP;
 728         dev->iflink             = 0;
 729         dev->addr_len           = 4;
 730 }
 731
 732 static int ipip6_tunnel_init(struct net_device *dev)
 733 {
 734         struct net_device *tdev = NULL;
 735         struct ip_tunnel *tunnel;
 736         struct iphdr *iph;
 737
 738         tunnel = netdev_priv(dev);
 739         iph = &tunnel->parms.iph;
 740
 741         tunnel->dev = dev;
 742         strcpy(tunnel->parms.name, dev->name);
 743
 744         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 745         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 746
 747         if (iph->daddr) {
 748                 struct flowi fl = { .nl_u = { .ip4_u =
 749                                               { .daddr = iph->daddr,
 750                                                 .saddr = iph->saddr,
 751                                                 .tos = RT_TOS(iph->tos) } },
 752                                     .oif = tunnel->parms.link,
 753                                     .proto = IPPROTO_IPV6 };
 754                 struct rtable *rt;
 755                 if (!ip_route_output_key(&rt, &fl)) {
 756                         tdev = rt->u.dst.dev;
 757                         ip_rt_put(rt);
 758                 }
 759                 dev->flags |= IFF_POINTOPOINT;
 760         }
 761
 762         if (!tdev && tunnel->parms.link)
 763                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 764
 765         if (tdev) {
 766                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 767                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 768                 if (dev->mtu < IPV6_MIN_MTU)
 769                         dev->mtu = IPV6_MIN_MTU;
 770         }
 771         dev->iflink = tunnel->parms.link;
 772
 773         return 0;
 774 }
 775
 776 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 777 {
 778         struct ip_tunnel *tunnel = netdev_priv(dev);
 779         struct iphdr *iph = &tunnel->parms.iph;
 780
 781         tunnel->dev = dev;
 782         strcpy(tunnel->parms.name, dev->name);
 783
 784         iph->version            = 4;
 785         iph->protocol           = IPPROTO_IPV6;
 786         iph->ihl                = 5;
 787         iph->ttl                = 64;
 788
 789         dev_hold(dev);
 790         tunnels_wc[0]           = tunnel;
 791         return 0;
 792 }
 793
 794 static struct xfrm_tunnel sit_handler = {
 795         .handler        =       ipip6_rcv,
 796         .err_handler    =       ipip6_err,
 797         .priority       =       1,
 798 };
 799
 800 static void __exit sit_destroy_tunnels(void)
 801 {
 802         int prio;
 803
 804         for (prio = 1; prio < 4; prio++) {
 805                 int h;
 806                 for (h = 0; h < HASH_SIZE; h++) {
 807                         struct ip_tunnel *t;
 808                         while ((t = tunnels[prio][h]) != NULL)
 809                                 unregister_netdevice(t->dev);
 810                 }
 811         }
 812 }
 813
 814 static void __exit sit_cleanup(void)
 815 {
 816         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 817
 818         rtnl_lock();
 819         sit_destroy_tunnels();
 820         unregister_netdevice(ipip6_fb_tunnel_dev);
 821         rtnl_unlock();
 822 }
 823
 824 static int __init sit_init(void)
 825 {
 826         int err;
 827
 828         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 829
 830         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
 831                 printk(KERN_INFO "sit init: Can't add protocol\n");
 832                 return -EAGAIN;
 833         }
 834
 835         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
 836                                            ipip6_tunnel_setup);
 837         if (!ipip6_fb_tunnel_dev) {
 838                 err = -ENOMEM;
 839                 goto err1;
 840         }
 841
 842         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
 843
 844         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
 845                 goto err2;
 846
 847  out:
 848         return err;
 849  err2:
 850         free_netdev(ipip6_fb_tunnel_dev);
 851  err1:
 852         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 853         goto out;
 854 }
 855
 856 module_init(sit_init);
 857 module_exit(sit_cleanup);
 858 MODULE_LICENSE("GPL");
 859 MODULE_ALIAS("sit0");