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
  54 /*
  55    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  56
  57    For comments look at net/ipv4/ip_gre.c --ANK
  58  */
  59
  60 #define HASH_SIZE  16
  61 #define HASH(addr) ((addr^(addr>>4))&0xF)
  62
  63 static int ipip6_fb_tunnel_init(struct net_device *dev);
  64 static int ipip6_tunnel_init(struct net_device *dev);
  65 static void ipip6_tunnel_setup(struct net_device *dev);
  66
  67 static struct net_device *ipip6_fb_tunnel_dev;
  68
  69 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  70 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  71 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  72 static struct ip_tunnel *tunnels_wc[1];
  73 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  74
  75 static rwlock_t ipip6_lock = RW_LOCK_UNLOCKED;
  76
  77 static struct ip_tunnel * ipip6_tunnel_lookup(u32 remote, u32 local)
  78 {
  79         unsigned h0 = HASH(remote);
  80         unsigned h1 = HASH(local);
  81         struct ip_tunnel *t;
  82
  83         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  84                 if (local == t->parms.iph.saddr &&
  85                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  86                         return t;
  87         }
  88         for (t = tunnels_r[h0]; t; t = t->next) {
  89                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  90                         return t;
  91         }
  92         for (t = tunnels_l[h1]; t; t = t->next) {
  93                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  94                         return t;
  95         }
  96         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
  97                 return t;
  98         return NULL;
  99 }
 100
 101 static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
 102 {
 103         u32 remote = t->parms.iph.daddr;
 104         u32 local = t->parms.iph.saddr;
 105         unsigned h = 0;
 106         int prio = 0;
 107
 108         if (remote) {
 109                 prio |= 2;
 110                 h ^= HASH(remote);
 111         }
 112         if (local) {
 113                 prio |= 1;
 114                 h ^= HASH(local);
 115         }
 116         return &tunnels[prio][h];
 117 }
 118
 119 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 120 {
 121         struct ip_tunnel **tp;
 122
 123         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 124                 if (t == *tp) {
 125                         write_lock_bh(&ipip6_lock);
 126                         *tp = t->next;
 127                         write_unlock_bh(&ipip6_lock);
 128                         break;
 129                 }
 130         }
 131 }
 132
 133 static void ipip6_tunnel_link(struct ip_tunnel *t)
 134 {
 135         struct ip_tunnel **tp = ipip6_bucket(t);
 136
 137         write_lock_bh(&ipip6_lock);
 138         t->next = *tp;
 139         write_unlock_bh(&ipip6_lock);
 140         *tp = t;
 141 }
 142
 143 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 144 {
 145         u32 remote = parms->iph.daddr;
 146         u32 local = parms->iph.saddr;
 147         struct ip_tunnel *t, **tp, *nt;
 148         struct net_device *dev;
 149         unsigned h = 0;
 150         int prio = 0;
 151         char name[IFNAMSIZ];
 152
 153         if (remote) {
 154                 prio |= 2;
 155                 h ^= HASH(remote);
 156         }
 157         if (local) {
 158                 prio |= 1;
 159                 h ^= HASH(local);
 160         }
 161         for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
 162                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 163                         return t;
 164         }
 165         if (!create)
 166                 goto failed;
 167
 168         if (parms->name[0])
 169                 strlcpy(name, parms->name, IFNAMSIZ);
 170         else {
 171                 int i;
 172                 for (i=1; i<100; i++) {
 173                         sprintf(name, "sit%d", i);
 174                         if (__dev_get_by_name(name) == NULL)
 175                                 break;
 176                 }
 177                 if (i==100)
 178                         goto failed;
 179         }
 180
 181         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 182         if (dev == NULL)
 183                 return NULL;
 184
 185         nt = dev->priv;
 186         dev->init = ipip6_tunnel_init;
 187         nt->parms = *parms;
 188
 189         if (register_netdevice(dev) < 0) {
 190                 free_netdev(dev);
 191                 goto failed;
 192         }
 193
 194         dev_hold(dev);
 195
 196         ipip6_tunnel_link(nt);
 197         /* Do not decrement MOD_USE_COUNT here. */
 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             INET_ECN_is_not_ce(ip6_get_dsfield(skb->nh.ipv6h)))
 365                 IP6_ECN_set_ce(skb->nh.ipv6h);
 366 }
 367
 368 static int ipip6_rcv(struct sk_buff *skb)
 369 {
 370         struct iphdr *iph;
 371         struct ip_tunnel *tunnel;
 372
 373         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 374                 goto out;
 375
 376         iph = skb->nh.iph;
 377
 378         read_lock(&ipip6_lock);
 379         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 380                 secpath_reset(skb);
 381                 skb->mac.raw = skb->nh.raw;
 382                 skb->nh.raw = skb->data;
 383                 memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
 384                 skb->protocol = htons(ETH_P_IPV6);
 385                 skb->pkt_type = PACKET_HOST;
 386                 tunnel->stat.rx_packets++;
 387                 tunnel->stat.rx_bytes += skb->len;
 388                 skb->dev = tunnel->dev;
 389                 dst_release(skb->dst);
 390                 skb->dst = NULL;
 391 #ifdef CONFIG_NETFILTER
 392                 nf_conntrack_put(skb->nfct);
 393                 skb->nfct = NULL;
 394 #ifdef CONFIG_NETFILTER_DEBUG
 395                 skb->nf_debug = 0;
 396 #endif
 397 #endif
 398                 ipip6_ecn_decapsulate(iph, skb);
 399                 netif_rx(skb);
 400                 read_unlock(&ipip6_lock);
 401                 return 0;
 402         }
 403
 404         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0);
 405         kfree_skb(skb);
 406         read_unlock(&ipip6_lock);
 407 out:
 408         return 0;
 409 }
 410
 411 /* Returns the embedded IPv4 address if the IPv6 address
 412    comes from 6to4 (RFC 3056) addr space */
 413
 414 static inline u32 try_6to4(struct in6_addr *v6dst)
 415 {
 416         u32 dst = 0;
 417
 418         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 419                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 420                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 421         }
 422         return dst;
 423 }
 424
 425 /*
 426  *      This function assumes it is being called from dev_queue_xmit()
 427  *      and that skb is filled properly by that function.
 428  */
 429
 430 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 431 {
 432         struct ip_tunnel *tunnel = (struct ip_tunnel*)dev->priv;
 433         struct net_device_stats *stats = &tunnel->stat;
 434         struct iphdr  *tiph = &tunnel->parms.iph;
 435         struct ipv6hdr *iph6 = skb->nh.ipv6h;
 436         u8     tos = tunnel->parms.iph.tos;
 437         struct rtable *rt;                      /* Route to the other host */
 438         struct net_device *tdev;                        /* Device to other host */
 439         struct iphdr  *iph;                     /* Our new IP header */
 440         int    max_headroom;                    /* The extra header space needed */
 441         u32    dst = tiph->daddr;
 442         int    mtu;
 443         struct in6_addr *addr6;
 444         int addr_type;
 445
 446         if (tunnel->recursion++) {
 447                 tunnel->stat.collisions++;
 448                 goto tx_error;
 449         }
 450
 451         if (skb->protocol != htons(ETH_P_IPV6))
 452                 goto tx_error;
 453
 454         if (!dst)
 455                 dst = try_6to4(&iph6->daddr);
 456
 457         if (!dst) {
 458                 struct neighbour *neigh = NULL;
 459
 460                 if (skb->dst)
 461                         neigh = skb->dst->neighbour;
 462
 463                 if (neigh == NULL) {
 464                         if (net_ratelimit())
 465                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 466                         goto tx_error;
 467                 }
 468
 469                 addr6 = (struct in6_addr*)&neigh->primary_key;
 470                 addr_type = ipv6_addr_type(addr6);
 471
 472                 if (addr_type == IPV6_ADDR_ANY) {
 473                         addr6 = &skb->nh.ipv6h->daddr;
 474                         addr_type = ipv6_addr_type(addr6);
 475                 }
 476
 477                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 478                         goto tx_error_icmp;
 479
 480                 dst = addr6->s6_addr32[3];
 481         }
 482
 483         {
 484                 struct flowi fl = { .nl_u = { .ip4_u =
 485                                               { .daddr = dst,
 486                                                 .saddr = tiph->saddr,
 487                                                 .tos = RT_TOS(tos) } },
 488                                     .oif = tunnel->parms.link,
 489                                     .proto = IPPROTO_IPV6 };
 490                 if (ip_route_output_key(&rt, &fl)) {
 491                         tunnel->stat.tx_carrier_errors++;
 492                         goto tx_error_icmp;
 493                 }
 494         }
 495         if (rt->rt_type != RTN_UNICAST) {
 496                 tunnel->stat.tx_carrier_errors++;
 497                 goto tx_error_icmp;
 498         }
 499         tdev = rt->u.dst.dev;
 500
 501         if (tdev == dev) {
 502                 ip_rt_put(rt);
 503                 tunnel->stat.collisions++;
 504                 goto tx_error;
 505         }
 506
 507         if (tiph->frag_off)
 508                 mtu = dst_pmtu(&rt->u.dst) - sizeof(struct iphdr);
 509         else
 510                 mtu = skb->dst ? dst_pmtu(skb->dst) : dev->mtu;
 511
 512         if (mtu < 68) {
 513                 tunnel->stat.collisions++;
 514                 ip_rt_put(rt);
 515                 goto tx_error;
 516         }
 517         if (mtu < IPV6_MIN_MTU)
 518                 mtu = IPV6_MIN_MTU;
 519         if (tunnel->parms.iph.daddr && skb->dst)
 520                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 521
 522         if (skb->len > mtu) {
 523                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 524                 ip_rt_put(rt);
 525                 goto tx_error;
 526         }
 527
 528         if (tunnel->err_count > 0) {
 529                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 530                         tunnel->err_count--;
 531                         dst_link_failure(skb);
 532                 } else
 533                         tunnel->err_count = 0;
 534         }
 535
 536         /*
 537          * Okay, now see if we can stuff it in the buffer as-is.
 538          */
 539         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 540
 541         if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
 542                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 543                 if (!new_skb) {
 544                         ip_rt_put(rt);
 545                         stats->tx_dropped++;
 546                         dev_kfree_skb(skb);
 547                         tunnel->recursion--;
 548                         return 0;
 549                 }
 550                 if (skb->sk)
 551                         skb_set_owner_w(new_skb, skb->sk);
 552                 dev_kfree_skb(skb);
 553                 skb = new_skb;
 554                 iph6 = skb->nh.ipv6h;
 555         }
 556
 557         skb->h.raw = skb->nh.raw;
 558         skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
 559         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 560         dst_release(skb->dst);
 561         skb->dst = &rt->u.dst;
 562
 563         /*
 564          *      Push down and install the IPIP header.
 565          */
 566
 567         iph                     =       skb->nh.iph;
 568         iph->version            =       4;
 569         iph->ihl                =       sizeof(struct iphdr)>>2;
 570         if (mtu > IPV6_MIN_MTU)
 571                 iph->frag_off   =       htons(IP_DF);
 572         else
 573                 iph->frag_off   =       0;
 574
 575         iph->protocol           =       IPPROTO_IPV6;
 576         iph->tos                =       INET_ECN_encapsulate(tos, ip6_get_dsfield(iph6));
 577         iph->daddr              =       rt->rt_dst;
 578         iph->saddr              =       rt->rt_src;
 579
 580         if ((iph->ttl = tiph->ttl) == 0)
 581                 iph->ttl        =       iph6->hop_limit;
 582
 583 #ifdef CONFIG_NETFILTER
 584         nf_conntrack_put(skb->nfct);
 585         skb->nfct = NULL;
 586 #ifdef CONFIG_NETFILTER_DEBUG
 587         skb->nf_debug = 0;
 588 #endif
 589 #endif
 590
 591         IPTUNNEL_XMIT();
 592         tunnel->recursion--;
 593         return 0;
 594
 595 tx_error_icmp:
 596         dst_link_failure(skb);
 597 tx_error:
 598         stats->tx_errors++;
 599         dev_kfree_skb(skb);
 600         tunnel->recursion--;
 601         return 0;
 602 }
 603
 604 static int
 605 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 606 {
 607         int err = 0;
 608         struct ip_tunnel_parm p;
 609         struct ip_tunnel *t;
 610
 611         switch (cmd) {
 612         case SIOCGETTUNNEL:
 613                 t = NULL;
 614                 if (dev == ipip6_fb_tunnel_dev) {
 615                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 616                                 err = -EFAULT;
 617                                 break;
 618                         }
 619                         t = ipip6_tunnel_locate(&p, 0);
 620                 }
 621                 if (t == NULL)
 622                         t = (struct ip_tunnel*)dev->priv;
 623                 memcpy(&p, &t->parms, sizeof(p));
 624                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 625                         err = -EFAULT;
 626                 break;
 627
 628         case SIOCADDTUNNEL:
 629         case SIOCCHGTUNNEL:
 630                 err = -EPERM;
 631                 if (!capable(CAP_NET_ADMIN))
 632                         goto done;
 633
 634                 err = -EFAULT;
 635                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 636                         goto done;
 637
 638                 err = -EINVAL;
 639                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 640                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 641                         goto done;
 642                 if (p.iph.ttl)
 643                         p.iph.frag_off |= htons(IP_DF);
 644
 645                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 646
 647                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 648                         if (t != NULL) {
 649                                 if (t->dev != dev) {
 650                                         err = -EEXIST;
 651                                         break;
 652                                 }
 653                         } else {
 654                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 655                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 656                                         err = -EINVAL;
 657                                         break;
 658                                 }
 659                                 t = (struct ip_tunnel*)dev->priv;
 660                                 ipip6_tunnel_unlink(t);
 661                                 t->parms.iph.saddr = p.iph.saddr;
 662                                 t->parms.iph.daddr = p.iph.daddr;
 663                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 664                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 665                                 ipip6_tunnel_link(t);
 666                                 netdev_state_change(dev);
 667                         }
 668                 }
 669
 670                 if (t) {
 671                         err = 0;
 672                         if (cmd == SIOCCHGTUNNEL) {
 673                                 t->parms.iph.ttl = p.iph.ttl;
 674                                 t->parms.iph.tos = p.iph.tos;
 675                         }
 676                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 677                                 err = -EFAULT;
 678                 } else
 679                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 680                 break;
 681
 682         case SIOCDELTUNNEL:
 683                 err = -EPERM;
 684                 if (!capable(CAP_NET_ADMIN))
 685                         goto done;
 686
 687                 if (dev == ipip6_fb_tunnel_dev) {
 688                         err = -EFAULT;
 689                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 690                                 goto done;
 691                         err = -ENOENT;
 692                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 693                                 goto done;
 694                         err = -EPERM;
 695                         if (t == ipip6_fb_tunnel_dev->priv)
 696                                 goto done;
 697                         dev = t->dev;
 698                 }
 699                 err = unregister_netdevice(dev);
 700                 break;
 701
 702         default:
 703                 err = -EINVAL;
 704         }
 705
 706 done:
 707         return err;
 708 }
 709
 710 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 711 {
 712         return &(((struct ip_tunnel*)dev->priv)->stat);
 713 }
 714
 715 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 716 {
 717         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 718                 return -EINVAL;
 719         dev->mtu = new_mtu;
 720         return 0;
 721 }
 722
 723 static void ipip6_tunnel_setup(struct net_device *dev)
 724 {
 725         SET_MODULE_OWNER(dev);
 726         dev->uninit             = ipip6_tunnel_uninit;
 727         dev->destructor         = free_netdev;
 728         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 729         dev->get_stats          = ipip6_tunnel_get_stats;
 730         dev->do_ioctl           = ipip6_tunnel_ioctl;
 731         dev->change_mtu         = ipip6_tunnel_change_mtu;
 732
 733         dev->type               = ARPHRD_SIT;
 734         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 735         dev->mtu                = 1500 - sizeof(struct iphdr);
 736         dev->flags              = IFF_NOARP;
 737         dev->iflink             = 0;
 738         dev->addr_len           = 4;
 739 }
 740
 741 static int ipip6_tunnel_init(struct net_device *dev)
 742 {
 743         struct net_device *tdev = NULL;
 744         struct ip_tunnel *tunnel;
 745         struct iphdr *iph;
 746
 747         tunnel = (struct ip_tunnel*)dev->priv;
 748         iph = &tunnel->parms.iph;
 749
 750         tunnel->dev = dev;
 751         strcpy(tunnel->parms.name, dev->name);
 752
 753         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 754         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 755
 756         if (iph->daddr) {
 757                 struct flowi fl = { .nl_u = { .ip4_u =
 758                                               { .daddr = iph->daddr,
 759                                                 .saddr = iph->saddr,
 760                                                 .tos = RT_TOS(iph->tos) } },
 761                                     .oif = tunnel->parms.link,
 762                                     .proto = IPPROTO_IPV6 };
 763                 struct rtable *rt;
 764                 if (!ip_route_output_key(&rt, &fl)) {
 765                         tdev = rt->u.dst.dev;
 766                         ip_rt_put(rt);
 767                 }
 768                 dev->flags |= IFF_POINTOPOINT;
 769         }
 770
 771         if (!tdev && tunnel->parms.link)
 772                 tdev = __dev_get_by_index(tunnel->parms.link);
 773
 774         if (tdev) {
 775                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 776                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 777                 if (dev->mtu < IPV6_MIN_MTU)
 778                         dev->mtu = IPV6_MIN_MTU;
 779         }
 780         dev->iflink = tunnel->parms.link;
 781
 782         return 0;
 783 }
 784
 785 int __init ipip6_fb_tunnel_init(struct net_device *dev)
 786 {
 787         struct ip_tunnel *tunnel = dev->priv;
 788         struct iphdr *iph = &tunnel->parms.iph;
 789
 790         tunnel->dev = dev;
 791         strcpy(tunnel->parms.name, dev->name);
 792
 793         iph->version            = 4;
 794         iph->protocol           = IPPROTO_IPV6;
 795         iph->ihl                = 5;
 796         iph->ttl                = 64;
 797
 798         dev_hold(dev);
 799         tunnels_wc[0]           = tunnel;
 800         return 0;
 801 }
 802
 803 static struct inet_protocol sit_protocol = {
 804         .handler        =       ipip6_rcv,
 805         .err_handler    =       ipip6_err,
 806 };
 807
 808 void __exit sit_cleanup(void)
 809 {
 810         inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
 811         unregister_netdev(ipip6_fb_tunnel_dev);
 812 }
 813
 814 int __init sit_init(void)
 815 {
 816         int err;
 817
 818         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 819
 820         if (inet_add_protocol(&sit_protocol, IPPROTO_IPV6) < 0) {
 821                 printk(KERN_INFO "sit init: Can't add protocol\n");
 822                 return -EAGAIN;
 823         }
 824
 825         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
 826                                            ipip6_tunnel_setup);
 827         if (!ipip6_fb_tunnel_dev) {
 828                 err = -ENOMEM;
 829                 goto fail;
 830         }
 831
 832         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
 833
 834         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
 835                 goto fail;
 836
 837  out:
 838         return err;
 839  fail:
 840         inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
 841         free_netdev(ipip6_fb_tunnel_dev);
 842         goto out;
 843 }