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/config.h>
  22 #include <linux/module.h>
  23 #include <linux/errno.h>
  24 #include <linux/types.h>
  25 #include <linux/socket.h>
  26 #include <linux/sockios.h>
  27 #include <linux/sched.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
  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) ((addr^(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(u32 remote, u32 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 *t)
 103 {
 104         u32 remote = t->parms.iph.daddr;
 105         u32 local = t->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 void ipip6_tunnel_unlink(struct ip_tunnel *t)
 121 {
 122         struct ip_tunnel **tp;
 123
 124         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 125                 if (t == *tp) {
 126                         write_lock_bh(&ipip6_lock);
 127                         *tp = t->next;
 128                         write_unlock_bh(&ipip6_lock);
 129                         break;
 130                 }
 131         }
 132 }
 133
 134 static void ipip6_tunnel_link(struct ip_tunnel *t)
 135 {
 136         struct ip_tunnel **tp = ipip6_bucket(t);
 137
 138         t->next = *tp;
 139         write_lock_bh(&ipip6_lock);
 140         *tp = t;
 141         write_unlock_bh(&ipip6_lock);
 142 }
 143
 144 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 145 {
 146         u32 remote = parms->iph.daddr;
 147         u32 local = parms->iph.saddr;
 148         struct ip_tunnel *t, **tp, *nt;
 149         struct net_device *dev;
 150         unsigned h = 0;
 151         int prio = 0;
 152         char name[IFNAMSIZ];
 153
 154         if (remote) {
 155                 prio |= 2;
 156                 h ^= HASH(remote);
 157         }
 158         if (local) {
 159                 prio |= 1;
 160                 h ^= HASH(local);
 161         }
 162         for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
 163                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 164                         return t;
 165         }
 166         if (!create)
 167                 goto failed;
 168
 169         if (parms->name[0])
 170                 strlcpy(name, parms->name, IFNAMSIZ);
 171         else {
 172                 int i;
 173                 for (i=1; i<100; i++) {
 174                         sprintf(name, "sit%d", i);
 175                         if (__dev_get_by_name(name) == NULL)
 176                                 break;
 177                 }
 178                 if (i==100)
 179                         goto failed;
 180         }
 181
 182         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 183         if (dev == NULL)
 184                 return NULL;
 185
 186         nt = dev->priv;
 187         dev->init = ipip6_tunnel_init;
 188         nt->parms = *parms;
 189
 190         if (register_netdevice(dev) < 0) {
 191                 free_netdev(dev);
 192                 goto failed;
 193         }
 194
 195         dev_hold(dev);
 196
 197         ipip6_tunnel_link(nt);
 198         return nt;
 199
 200 failed:
 201         return NULL;
 202 }
 203
 204 static void ipip6_tunnel_uninit(struct net_device *dev)
 205 {
 206         if (dev == ipip6_fb_tunnel_dev) {
 207                 write_lock_bh(&ipip6_lock);
 208                 tunnels_wc[0] = NULL;
 209                 write_unlock_bh(&ipip6_lock);
 210                 dev_put(dev);
 211         } else {
 212                 ipip6_tunnel_unlink((struct ip_tunnel*)dev->priv);
 213                 dev_put(dev);
 214         }
 215 }
 216
 217
 218 static void ipip6_err(struct sk_buff *skb, u32 info)
 219 {
 220 #ifndef I_WISH_WORLD_WERE_PERFECT
 221
 222 /* It is not :-( All the routers (except for Linux) return only
 223    8 bytes of packet payload. It means, that precise relaying of
 224    ICMP in the real Internet is absolutely infeasible.
 225  */
 226         struct iphdr *iph = (struct iphdr*)skb->data;
 227         int type = skb->h.icmph->type;
 228         int code = skb->h.icmph->code;
 229         struct ip_tunnel *t;
 230
 231         switch (type) {
 232         default:
 233         case ICMP_PARAMETERPROB:
 234                 return;
 235
 236         case ICMP_DEST_UNREACH:
 237                 switch (code) {
 238                 case ICMP_SR_FAILED:
 239                 case ICMP_PORT_UNREACH:
 240                         /* Impossible event. */
 241                         return;
 242                 case ICMP_FRAG_NEEDED:
 243                         /* Soft state for pmtu is maintained by IP core. */
 244                         return;
 245                 default:
 246                         /* All others are translated to HOST_UNREACH.
 247                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 248                            I believe they are just ether pollution. --ANK
 249                          */
 250                         break;
 251                 }
 252                 break;
 253         case ICMP_TIME_EXCEEDED:
 254                 if (code != ICMP_EXC_TTL)
 255                         return;
 256                 break;
 257         }
 258
 259         read_lock(&ipip6_lock);
 260         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 261         if (t == NULL || t->parms.iph.daddr == 0)
 262                 goto out;
 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;
 274 #else
 275         struct iphdr *iph = (struct iphdr*)dp;
 276         int hlen = iph->ihl<<2;
 277         struct ipv6hdr *iph6;
 278         int type = skb->h.icmph->type;
 279         int code = skb->h.icmph->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 (skb->h.icmph->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 = skb->h.icmph->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;
 335         dst_release(skb2->dst);
 336         skb2->dst = NULL;
 337         skb_pull(skb2, skb->data - (u8*)iph6);
 338         skb2->nh.raw = skb2->data;
 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 = (struct ip_tunnel*)rt6i->rt6i_dev->priv;
 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;
 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(skb->nh.ipv6h);
 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 = skb->nh.iph;
 376
 377         read_lock(&ipip6_lock);
 378         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 379                 secpath_reset(skb);
 380                 skb->mac.raw = skb->nh.raw;
 381                 skb->nh.raw = skb->data;
 382                 memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
 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_PROT_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 u32 try_6to4(struct in6_addr *v6dst)
 408 {
 409         u32 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 = (struct ip_tunnel*)dev->priv;
 426         struct net_device_stats *stats = &tunnel->stat;
 427         struct iphdr  *tiph = &tunnel->parms.iph;
 428         struct ipv6hdr *iph6 = skb->nh.ipv6h;
 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         int    max_headroom;                    /* The extra header space needed */
 434         u32    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 = &skb->nh.ipv6h->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_cloned(skb) || skb_shared(skb)) {
 536                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 537                 if (!new_skb) {
 538                         ip_rt_put(rt);
 539                         stats->tx_dropped++;
 540                         dev_kfree_skb(skb);
 541                         tunnel->recursion--;
 542                         return 0;
 543                 }
 544                 if (skb->sk)
 545                         skb_set_owner_w(new_skb, skb->sk);
 546                 dev_kfree_skb(skb);
 547                 skb = new_skb;
 548                 iph6 = skb->nh.ipv6h;
 549         }
 550
 551         skb->h.raw = skb->nh.raw;
 552         skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
 553         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 554         dst_release(skb->dst);
 555         skb->dst = &rt->u.dst;
 556
 557         /*
 558          *      Push down and install the IPIP header.
 559          */
 560
 561         iph                     =       skb->nh.iph;
 562         iph->version            =       4;
 563         iph->ihl                =       sizeof(struct iphdr)>>2;
 564         if (mtu > IPV6_MIN_MTU)
 565                 iph->frag_off   =       htons(IP_DF);
 566         else
 567                 iph->frag_off   =       0;
 568
 569         iph->protocol           =       IPPROTO_IPV6;
 570         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 571         iph->daddr              =       rt->rt_dst;
 572         iph->saddr              =       rt->rt_src;
 573
 574         if ((iph->ttl = tiph->ttl) == 0)
 575                 iph->ttl        =       iph6->hop_limit;
 576
 577         nf_reset(skb);
 578
 579         IPTUNNEL_XMIT();
 580         tunnel->recursion--;
 581         return 0;
 582
 583 tx_error_icmp:
 584         dst_link_failure(skb);
 585 tx_error:
 586         stats->tx_errors++;
 587         dev_kfree_skb(skb);
 588         tunnel->recursion--;
 589         return 0;
 590 }
 591
 592 static int
 593 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 594 {
 595         int err = 0;
 596         struct ip_tunnel_parm p;
 597         struct ip_tunnel *t;
 598
 599         switch (cmd) {
 600         case SIOCGETTUNNEL:
 601                 t = NULL;
 602                 if (dev == ipip6_fb_tunnel_dev) {
 603                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 604                                 err = -EFAULT;
 605                                 break;
 606                         }
 607                         t = ipip6_tunnel_locate(&p, 0);
 608                 }
 609                 if (t == NULL)
 610                         t = (struct ip_tunnel*)dev->priv;
 611                 memcpy(&p, &t->parms, sizeof(p));
 612                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 613                         err = -EFAULT;
 614                 break;
 615
 616         case SIOCADDTUNNEL:
 617         case SIOCCHGTUNNEL:
 618                 err = -EPERM;
 619                 if (!capable(CAP_NET_ADMIN))
 620                         goto done;
 621
 622                 err = -EFAULT;
 623                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 624                         goto done;
 625
 626                 err = -EINVAL;
 627                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 628                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 629                         goto done;
 630                 if (p.iph.ttl)
 631                         p.iph.frag_off |= htons(IP_DF);
 632
 633                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 634
 635                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 636                         if (t != NULL) {
 637                                 if (t->dev != dev) {
 638                                         err = -EEXIST;
 639                                         break;
 640                                 }
 641                         } else {
 642                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 643                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 644                                         err = -EINVAL;
 645                                         break;
 646                                 }
 647                                 t = (struct ip_tunnel*)dev->priv;
 648                                 ipip6_tunnel_unlink(t);
 649                                 t->parms.iph.saddr = p.iph.saddr;
 650                                 t->parms.iph.daddr = p.iph.daddr;
 651                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 652                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 653                                 ipip6_tunnel_link(t);
 654                                 netdev_state_change(dev);
 655                         }
 656                 }
 657
 658                 if (t) {
 659                         err = 0;
 660                         if (cmd == SIOCCHGTUNNEL) {
 661                                 t->parms.iph.ttl = p.iph.ttl;
 662                                 t->parms.iph.tos = p.iph.tos;
 663                         }
 664                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 665                                 err = -EFAULT;
 666                 } else
 667                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 668                 break;
 669
 670         case SIOCDELTUNNEL:
 671                 err = -EPERM;
 672                 if (!capable(CAP_NET_ADMIN))
 673                         goto done;
 674
 675                 if (dev == ipip6_fb_tunnel_dev) {
 676                         err = -EFAULT;
 677                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 678                                 goto done;
 679                         err = -ENOENT;
 680                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 681                                 goto done;
 682                         err = -EPERM;
 683                         if (t == ipip6_fb_tunnel_dev->priv)
 684                                 goto done;
 685                         dev = t->dev;
 686                 }
 687                 err = unregister_netdevice(dev);
 688                 break;
 689
 690         default:
 691                 err = -EINVAL;
 692         }
 693
 694 done:
 695         return err;
 696 }
 697
 698 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 699 {
 700         return &(((struct ip_tunnel*)dev->priv)->stat);
 701 }
 702
 703 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 704 {
 705         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 706                 return -EINVAL;
 707         dev->mtu = new_mtu;
 708         return 0;
 709 }
 710
 711 static void ipip6_tunnel_setup(struct net_device *dev)
 712 {
 713         SET_MODULE_OWNER(dev);
 714         dev->uninit             = ipip6_tunnel_uninit;
 715         dev->destructor         = free_netdev;
 716         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 717         dev->get_stats          = ipip6_tunnel_get_stats;
 718         dev->do_ioctl           = ipip6_tunnel_ioctl;
 719         dev->change_mtu         = ipip6_tunnel_change_mtu;
 720
 721         dev->type               = ARPHRD_SIT;
 722         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 723         dev->mtu                = 1500 - sizeof(struct iphdr);
 724         dev->flags              = IFF_NOARP;
 725         dev->iflink             = 0;
 726         dev->addr_len           = 4;
 727 }
 728
 729 static int ipip6_tunnel_init(struct net_device *dev)
 730 {
 731         struct net_device *tdev = NULL;
 732         struct ip_tunnel *tunnel;
 733         struct iphdr *iph;
 734
 735         tunnel = (struct ip_tunnel*)dev->priv;
 736         iph = &tunnel->parms.iph;
 737
 738         tunnel->dev = dev;
 739         strcpy(tunnel->parms.name, dev->name);
 740
 741         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 742         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 743
 744         if (iph->daddr) {
 745                 struct flowi fl = { .nl_u = { .ip4_u =
 746                                               { .daddr = iph->daddr,
 747                                                 .saddr = iph->saddr,
 748                                                 .tos = RT_TOS(iph->tos) } },
 749                                     .oif = tunnel->parms.link,
 750                                     .proto = IPPROTO_IPV6 };
 751                 struct rtable *rt;
 752                 if (!ip_route_output_key(&rt, &fl)) {
 753                         tdev = rt->u.dst.dev;
 754                         ip_rt_put(rt);
 755                 }
 756                 dev->flags |= IFF_POINTOPOINT;
 757         }
 758
 759         if (!tdev && tunnel->parms.link)
 760                 tdev = __dev_get_by_index(tunnel->parms.link);
 761
 762         if (tdev) {
 763                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 764                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 765                 if (dev->mtu < IPV6_MIN_MTU)
 766                         dev->mtu = IPV6_MIN_MTU;
 767         }
 768         dev->iflink = tunnel->parms.link;
 769
 770         return 0;
 771 }
 772
 773 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 774 {
 775         struct ip_tunnel *tunnel = dev->priv;
 776         struct iphdr *iph = &tunnel->parms.iph;
 777
 778         tunnel->dev = dev;
 779         strcpy(tunnel->parms.name, dev->name);
 780
 781         iph->version            = 4;
 782         iph->protocol           = IPPROTO_IPV6;
 783         iph->ihl                = 5;
 784         iph->ttl                = 64;
 785
 786         dev_hold(dev);
 787         tunnels_wc[0]           = tunnel;
 788         return 0;
 789 }
 790
 791 static struct net_protocol sit_protocol = {
 792         .handler        =       ipip6_rcv,
 793         .err_handler    =       ipip6_err,
 794 };
 795
 796 static void __exit sit_destroy_tunnels(void)
 797 {
 798         int prio;
 799
 800         for (prio = 1; prio < 4; prio++) {
 801                 int h;
 802                 for (h = 0; h < HASH_SIZE; h++) {
 803                         struct ip_tunnel *t;
 804                         while ((t = tunnels[prio][h]) != NULL)
 805                                 unregister_netdevice(t->dev);
 806                 }
 807         }
 808 }
 809
 810 void __exit sit_cleanup(void)
 811 {
 812         inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
 813
 814         rtnl_lock();
 815         sit_destroy_tunnels();
 816         unregister_netdevice(ipip6_fb_tunnel_dev);
 817         rtnl_unlock();
 818 }
 819
 820 int __init sit_init(void)
 821 {
 822         int err;
 823
 824         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 825
 826         if (inet_add_protocol(&sit_protocol, IPPROTO_IPV6) < 0) {
 827                 printk(KERN_INFO "sit init: Can't add protocol\n");
 828                 return -EAGAIN;
 829         }
 830
 831         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
 832                                            ipip6_tunnel_setup);
 833         if (!ipip6_fb_tunnel_dev) {
 834                 err = -ENOMEM;
 835                 goto err1;
 836         }
 837
 838         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
 839
 840         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
 841                 goto err2;
 842
 843  out:
 844         return err;
 845  err2:
 846         free_netdev(ipip6_fb_tunnel_dev);
 847  err1:
 848         inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
 849         goto out;
 850 }