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/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 #include <linux/if_ether.h>
  37
  38 #include <net/sock.h>
  39 #include <net/snmp.h>
  40
  41 #include <net/ipv6.h>
  42 #include <net/protocol.h>
  43 #include <net/transp_v6.h>
  44 #include <net/ip6_fib.h>
  45 #include <net/ip6_route.h>
  46 #include <net/ndisc.h>
  47 #include <net/addrconf.h>
  48 #include <net/ip.h>
  49 #include <net/udp.h>
  50 #include <net/icmp.h>
  51 #include <net/ipip.h>
  52 #include <net/inet_ecn.h>
  53 #include <net/xfrm.h>
  54 #include <net/dsfield.h>
  55
  56 /*
  57    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  58
  59    For comments look at net/ipv4/ip_gre.c --ANK
  60  */
  61
  62 #define HASH_SIZE  16
  63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  64
  65 static int ipip6_fb_tunnel_init(struct net_device *dev);
  66 static int ipip6_tunnel_init(struct net_device *dev);
  67 static void ipip6_tunnel_setup(struct net_device *dev);
  68
  69 static struct net_device *ipip6_fb_tunnel_dev;
  70
  71 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  72 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  73 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  74 static struct ip_tunnel *tunnels_wc[1];
  75 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  76
  77 static DEFINE_RWLOCK(ipip6_lock);
  78
  79 static struct ip_tunnel * ipip6_tunnel_lookup(__be32 remote, __be32 local)
  80 {
  81         unsigned h0 = HASH(remote);
  82         unsigned h1 = HASH(local);
  83         struct ip_tunnel *t;
  84
  85         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  86                 if (local == t->parms.iph.saddr &&
  87                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  88                         return t;
  89         }
  90         for (t = tunnels_r[h0]; t; t = t->next) {
  91                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  92                         return t;
  93         }
  94         for (t = tunnels_l[h1]; t; t = t->next) {
  95                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  96                         return t;
  97         }
  98         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
  99                 return t;
 100         return NULL;
 101 }
 102
 103 static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
 104 {
 105         __be32 remote = t->parms.iph.daddr;
 106         __be32 local = t->parms.iph.saddr;
 107         unsigned h = 0;
 108         int prio = 0;
 109
 110         if (remote) {
 111                 prio |= 2;
 112                 h ^= HASH(remote);
 113         }
 114         if (local) {
 115                 prio |= 1;
 116                 h ^= HASH(local);
 117         }
 118         return &tunnels[prio][h];
 119 }
 120
 121 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 122 {
 123         struct ip_tunnel **tp;
 124
 125         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 126                 if (t == *tp) {
 127                         write_lock_bh(&ipip6_lock);
 128                         *tp = t->next;
 129                         write_unlock_bh(&ipip6_lock);
 130                         break;
 131                 }
 132         }
 133 }
 134
 135 static void ipip6_tunnel_link(struct ip_tunnel *t)
 136 {
 137         struct ip_tunnel **tp = ipip6_bucket(t);
 138
 139         t->next = *tp;
 140         write_lock_bh(&ipip6_lock);
 141         *tp = t;
 142         write_unlock_bh(&ipip6_lock);
 143 }
 144
 145 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 146 {
 147         __be32 remote = parms->iph.daddr;
 148         __be32 local = parms->iph.saddr;
 149         struct ip_tunnel *t, **tp, *nt;
 150         struct net_device *dev;
 151         unsigned h = 0;
 152         int prio = 0;
 153         char name[IFNAMSIZ];
 154
 155         if (remote) {
 156                 prio |= 2;
 157                 h ^= HASH(remote);
 158         }
 159         if (local) {
 160                 prio |= 1;
 161                 h ^= HASH(local);
 162         }
 163         for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
 164                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 165                         return t;
 166         }
 167         if (!create)
 168                 goto failed;
 169
 170         if (parms->name[0])
 171                 strlcpy(name, parms->name, IFNAMSIZ);
 172         else {
 173                 int i;
 174                 for (i=1; i<100; i++) {
 175                         sprintf(name, "sit%d", i);
 176                         if (__dev_get_by_name(name) == NULL)
 177                                 break;
 178                 }
 179                 if (i==100)
 180                         goto failed;
 181         }
 182
 183         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 184         if (dev == NULL)
 185                 return NULL;
 186
 187         nt = netdev_priv(dev);
 188         dev->init = ipip6_tunnel_init;
 189         nt->parms = *parms;
 190
 191         if (register_netdevice(dev) < 0) {
 192                 free_netdev(dev);
 193                 goto failed;
 194         }
 195
 196         dev_hold(dev);
 197
 198         ipip6_tunnel_link(nt);
 199         return nt;
 200
 201 failed:
 202         return NULL;
 203 }
 204
 205 static void ipip6_tunnel_uninit(struct net_device *dev)
 206 {
 207         if (dev == ipip6_fb_tunnel_dev) {
 208                 write_lock_bh(&ipip6_lock);
 209                 tunnels_wc[0] = NULL;
 210                 write_unlock_bh(&ipip6_lock);
 211                 dev_put(dev);
 212         } else {
 213                 ipip6_tunnel_unlink(netdev_priv(dev));
 214                 dev_put(dev);
 215         }
 216 }
 217
 218
 219 static int ipip6_err(struct sk_buff *skb, u32 info)
 220 {
 221 #ifndef I_WISH_WORLD_WERE_PERFECT
 222
 223 /* It is not :-( All the routers (except for Linux) return only
 224    8 bytes of packet payload. It means, that precise relaying of
 225    ICMP in the real Internet is absolutely infeasible.
 226  */
 227         struct iphdr *iph = (struct iphdr*)skb->data;
 228         int type = skb->h.icmph->type;
 229         int code = skb->h.icmph->code;
 230         struct ip_tunnel *t;
 231         int err;
 232
 233         switch (type) {
 234         default:
 235         case ICMP_PARAMETERPROB:
 236                 return 0;
 237
 238         case ICMP_DEST_UNREACH:
 239                 switch (code) {
 240                 case ICMP_SR_FAILED:
 241                 case ICMP_PORT_UNREACH:
 242                         /* Impossible event. */
 243                         return 0;
 244                 case ICMP_FRAG_NEEDED:
 245                         /* Soft state for pmtu is maintained by IP core. */
 246                         return 0;
 247                 default:
 248                         /* All others are translated to HOST_UNREACH.
 249                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 250                            I believe they are just ether pollution. --ANK
 251                          */
 252                         break;
 253                 }
 254                 break;
 255         case ICMP_TIME_EXCEEDED:
 256                 if (code != ICMP_EXC_TTL)
 257                         return 0;
 258                 break;
 259         }
 260
 261         err = -ENOENT;
 262
 263         read_lock(&ipip6_lock);
 264         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 265         if (t == NULL || t->parms.iph.daddr == 0)
 266                 goto out;
 267
 268         err = 0;
 269         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 270                 goto out;
 271
 272         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 273                 t->err_count++;
 274         else
 275                 t->err_count = 1;
 276         t->err_time = jiffies;
 277 out:
 278         read_unlock(&ipip6_lock);
 279         return err;
 280 #else
 281         struct iphdr *iph = (struct iphdr*)dp;
 282         int hlen = iph->ihl<<2;
 283         struct ipv6hdr *iph6;
 284         int type = skb->h.icmph->type;
 285         int code = skb->h.icmph->code;
 286         int rel_type = 0;
 287         int rel_code = 0;
 288         int rel_info = 0;
 289         struct sk_buff *skb2;
 290         struct rt6_info *rt6i;
 291
 292         if (len < hlen + sizeof(struct ipv6hdr))
 293                 return;
 294         iph6 = (struct ipv6hdr*)(dp + hlen);
 295
 296         switch (type) {
 297         default:
 298                 return;
 299         case ICMP_PARAMETERPROB:
 300                 if (skb->h.icmph->un.gateway < hlen)
 301                         return;
 302
 303                 /* So... This guy found something strange INSIDE encapsulated
 304                    packet. Well, he is fool, but what can we do ?
 305                  */
 306                 rel_type = ICMPV6_PARAMPROB;
 307                 rel_info = skb->h.icmph->un.gateway - hlen;
 308                 break;
 309
 310         case ICMP_DEST_UNREACH:
 311                 switch (code) {
 312                 case ICMP_SR_FAILED:
 313                 case ICMP_PORT_UNREACH:
 314                         /* Impossible event. */
 315                         return;
 316                 case ICMP_FRAG_NEEDED:
 317                         /* Too complicated case ... */
 318                         return;
 319                 default:
 320                         /* All others are translated to HOST_UNREACH.
 321                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 322                            I believe, it is just ether pollution. --ANK
 323                          */
 324                         rel_type = ICMPV6_DEST_UNREACH;
 325                         rel_code = ICMPV6_ADDR_UNREACH;
 326                         break;
 327                 }
 328                 break;
 329         case ICMP_TIME_EXCEEDED:
 330                 if (code != ICMP_EXC_TTL)
 331                         return;
 332                 rel_type = ICMPV6_TIME_EXCEED;
 333                 rel_code = ICMPV6_EXC_HOPLIMIT;
 334                 break;
 335         }
 336
 337         /* Prepare fake skb to feed it to icmpv6_send */
 338         skb2 = skb_clone(skb, GFP_ATOMIC);
 339         if (skb2 == NULL)
 340                 return 0;
 341         dst_release(skb2->dst);
 342         skb2->dst = NULL;
 343         skb_pull(skb2, skb->data - (u8*)iph6);
 344         skb2->nh.raw = skb2->data;
 345
 346         /* Try to guess incoming interface */
 347         rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
 348         if (rt6i && rt6i->rt6i_dev) {
 349                 skb2->dev = rt6i->rt6i_dev;
 350
 351                 rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
 352
 353                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 354                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 355                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 356                                 rel_type = ICMPV6_DEST_UNREACH;
 357                                 rel_code = ICMPV6_ADDR_UNREACH;
 358                         }
 359                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 360                 }
 361         }
 362         kfree_skb(skb2);
 363         return 0;
 364 #endif
 365 }
 366
 367 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 368 {
 369         if (INET_ECN_is_ce(iph->tos))
 370                 IP6_ECN_set_ce(skb->nh.ipv6h);
 371 }
 372
 373 static int ipip6_rcv(struct sk_buff *skb)
 374 {
 375         struct iphdr *iph;
 376         struct ip_tunnel *tunnel;
 377
 378         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 379                 goto out;
 380
 381         iph = skb->nh.iph;
 382
 383         read_lock(&ipip6_lock);
 384         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 385                 secpath_reset(skb);
 386                 skb->mac.raw = skb->nh.raw;
 387                 skb->nh.raw = skb->data;
 388                 IPCB(skb)->flags = 0;
 389                 skb->protocol = htons(ETH_P_IPV6);
 390                 skb->pkt_type = PACKET_HOST;
 391                 tunnel->stat.rx_packets++;
 392                 tunnel->stat.rx_bytes += skb->len;
 393                 skb->dev = tunnel->dev;
 394                 dst_release(skb->dst);
 395                 skb->dst = NULL;
 396                 nf_reset(skb);
 397                 ipip6_ecn_decapsulate(iph, skb);
 398                 netif_rx(skb);
 399                 read_unlock(&ipip6_lock);
 400                 return 0;
 401         }
 402
 403         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 404         kfree_skb(skb);
 405         read_unlock(&ipip6_lock);
 406 out:
 407         return 0;
 408 }
 409
 410 /* Returns the embedded IPv4 address if the IPv6 address
 411    comes from 6to4 (RFC 3056) addr space */
 412
 413 static inline __be32 try_6to4(struct in6_addr *v6dst)
 414 {
 415         __be32 dst = 0;
 416
 417         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 418                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 419                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 420         }
 421         return dst;
 422 }
 423
 424 /*
 425  *      This function assumes it is being called from dev_queue_xmit()
 426  *      and that skb is filled properly by that function.
 427  */
 428
 429 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 430 {
 431         struct ip_tunnel *tunnel = netdev_priv(dev);
 432         struct net_device_stats *stats = &tunnel->stat;
 433         struct iphdr  *tiph = &tunnel->parms.iph;
 434         struct ipv6hdr *iph6 = skb->nh.ipv6h;
 435         u8     tos = tunnel->parms.iph.tos;
 436         struct rtable *rt;                      /* Route to the other host */
 437         struct net_device *tdev;                        /* Device to other host */
 438         struct iphdr  *iph;                     /* Our new IP header */
 439         int    max_headroom;                    /* The extra header space needed */
 440         __be32 dst = tiph->daddr;
 441         int    mtu;
 442         struct in6_addr *addr6;
 443         int addr_type;
 444
 445         if (tunnel->recursion++) {
 446                 tunnel->stat.collisions++;
 447                 goto tx_error;
 448         }
 449
 450         if (skb->protocol != htons(ETH_P_IPV6))
 451                 goto tx_error;
 452
 453         if (!dst)
 454                 dst = try_6to4(&iph6->daddr);
 455
 456         if (!dst) {
 457                 struct neighbour *neigh = NULL;
 458
 459                 if (skb->dst)
 460                         neigh = skb->dst->neighbour;
 461
 462                 if (neigh == NULL) {
 463                         if (net_ratelimit())
 464                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 465                         goto tx_error;
 466                 }
 467
 468                 addr6 = (struct in6_addr*)&neigh->primary_key;
 469                 addr_type = ipv6_addr_type(addr6);
 470
 471                 if (addr_type == IPV6_ADDR_ANY) {
 472                         addr6 = &skb->nh.ipv6h->daddr;
 473                         addr_type = ipv6_addr_type(addr6);
 474                 }
 475
 476                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 477                         goto tx_error_icmp;
 478
 479                 dst = addr6->s6_addr32[3];
 480         }
 481
 482         {
 483                 struct flowi fl = { .nl_u = { .ip4_u =
 484                                               { .daddr = dst,
 485                                                 .saddr = tiph->saddr,
 486                                                 .tos = RT_TOS(tos) } },
 487                                     .oif = tunnel->parms.link,
 488                                     .proto = IPPROTO_IPV6 };
 489                 if (ip_route_output_key(&rt, &fl)) {
 490                         tunnel->stat.tx_carrier_errors++;
 491                         goto tx_error_icmp;
 492                 }
 493         }
 494         if (rt->rt_type != RTN_UNICAST) {
 495                 ip_rt_put(rt);
 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_mtu(&rt->u.dst) - sizeof(struct iphdr);
 509         else
 510                 mtu = skb->dst ? dst_mtu(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         IPCB(skb)->flags = 0;
 561         dst_release(skb->dst);
 562         skb->dst = &rt->u.dst;
 563
 564         /*
 565          *      Push down and install the IPIP header.
 566          */
 567
 568         iph                     =       skb->nh.iph;
 569         iph->version            =       4;
 570         iph->ihl                =       sizeof(struct iphdr)>>2;
 571         if (mtu > IPV6_MIN_MTU)
 572                 iph->frag_off   =       htons(IP_DF);
 573         else
 574                 iph->frag_off   =       0;
 575
 576         iph->protocol           =       IPPROTO_IPV6;
 577         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 578         iph->daddr              =       rt->rt_dst;
 579         iph->saddr              =       rt->rt_src;
 580
 581         if ((iph->ttl = tiph->ttl) == 0)
 582                 iph->ttl        =       iph6->hop_limit;
 583
 584         nf_reset(skb);
 585
 586         IPTUNNEL_XMIT();
 587         tunnel->recursion--;
 588         return 0;
 589
 590 tx_error_icmp:
 591         dst_link_failure(skb);
 592 tx_error:
 593         stats->tx_errors++;
 594         dev_kfree_skb(skb);
 595         tunnel->recursion--;
 596         return 0;
 597 }
 598
 599 static int
 600 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 601 {
 602         int err = 0;
 603         struct ip_tunnel_parm p;
 604         struct ip_tunnel *t;
 605
 606         switch (cmd) {
 607         case SIOCGETTUNNEL:
 608                 t = NULL;
 609                 if (dev == ipip6_fb_tunnel_dev) {
 610                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 611                                 err = -EFAULT;
 612                                 break;
 613                         }
 614                         t = ipip6_tunnel_locate(&p, 0);
 615                 }
 616                 if (t == NULL)
 617                         t = netdev_priv(dev);
 618                 memcpy(&p, &t->parms, sizeof(p));
 619                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 620                         err = -EFAULT;
 621                 break;
 622
 623         case SIOCADDTUNNEL:
 624         case SIOCCHGTUNNEL:
 625                 err = -EPERM;
 626                 if (!capable(CAP_NET_ADMIN))
 627                         goto done;
 628
 629                 err = -EFAULT;
 630                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 631                         goto done;
 632
 633                 err = -EINVAL;
 634                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 635                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 636                         goto done;
 637                 if (p.iph.ttl)
 638                         p.iph.frag_off |= htons(IP_DF);
 639
 640                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 641
 642                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 643                         if (t != NULL) {
 644                                 if (t->dev != dev) {
 645                                         err = -EEXIST;
 646                                         break;
 647                                 }
 648                         } else {
 649                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 650                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 651                                         err = -EINVAL;
 652                                         break;
 653                                 }
 654                                 t = netdev_priv(dev);
 655                                 ipip6_tunnel_unlink(t);
 656                                 t->parms.iph.saddr = p.iph.saddr;
 657                                 t->parms.iph.daddr = p.iph.daddr;
 658                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 659                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 660                                 ipip6_tunnel_link(t);
 661                                 netdev_state_change(dev);
 662                         }
 663                 }
 664
 665                 if (t) {
 666                         err = 0;
 667                         if (cmd == SIOCCHGTUNNEL) {
 668                                 t->parms.iph.ttl = p.iph.ttl;
 669                                 t->parms.iph.tos = p.iph.tos;
 670                         }
 671                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 672                                 err = -EFAULT;
 673                 } else
 674                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 675                 break;
 676
 677         case SIOCDELTUNNEL:
 678                 err = -EPERM;
 679                 if (!capable(CAP_NET_ADMIN))
 680                         goto done;
 681
 682                 if (dev == ipip6_fb_tunnel_dev) {
 683                         err = -EFAULT;
 684                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 685                                 goto done;
 686                         err = -ENOENT;
 687                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 688                                 goto done;
 689                         err = -EPERM;
 690                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 691                                 goto done;
 692                         dev = t->dev;
 693                 }
 694                 unregister_netdevice(dev);
 695                 err = 0;
 696                 break;
 697
 698         default:
 699                 err = -EINVAL;
 700         }
 701
 702 done:
 703         return err;
 704 }
 705
 706 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 707 {
 708         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 709 }
 710
 711 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 712 {
 713         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 714                 return -EINVAL;
 715         dev->mtu = new_mtu;
 716         return 0;
 717 }
 718
 719 static void ipip6_tunnel_setup(struct net_device *dev)
 720 {
 721         SET_MODULE_OWNER(dev);
 722         dev->uninit             = ipip6_tunnel_uninit;
 723         dev->destructor         = free_netdev;
 724         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 725         dev->get_stats          = ipip6_tunnel_get_stats;
 726         dev->do_ioctl           = ipip6_tunnel_ioctl;
 727         dev->change_mtu         = ipip6_tunnel_change_mtu;
 728
 729         dev->type               = ARPHRD_SIT;
 730         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 731         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 732         dev->flags              = IFF_NOARP;
 733         dev->iflink             = 0;
 734         dev->addr_len           = 4;
 735 }
 736
 737 static int ipip6_tunnel_init(struct net_device *dev)
 738 {
 739         struct net_device *tdev = NULL;
 740         struct ip_tunnel *tunnel;
 741         struct iphdr *iph;
 742
 743         tunnel = netdev_priv(dev);
 744         iph = &tunnel->parms.iph;
 745
 746         tunnel->dev = dev;
 747         strcpy(tunnel->parms.name, dev->name);
 748
 749         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 750         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 751
 752         if (iph->daddr) {
 753                 struct flowi fl = { .nl_u = { .ip4_u =
 754                                               { .daddr = iph->daddr,
 755                                                 .saddr = iph->saddr,
 756                                                 .tos = RT_TOS(iph->tos) } },
 757                                     .oif = tunnel->parms.link,
 758                                     .proto = IPPROTO_IPV6 };
 759                 struct rtable *rt;
 760                 if (!ip_route_output_key(&rt, &fl)) {
 761                         tdev = rt->u.dst.dev;
 762                         ip_rt_put(rt);
 763                 }
 764                 dev->flags |= IFF_POINTOPOINT;
 765         }
 766
 767         if (!tdev && tunnel->parms.link)
 768                 tdev = __dev_get_by_index(tunnel->parms.link);
 769
 770         if (tdev) {
 771                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 772                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 773                 if (dev->mtu < IPV6_MIN_MTU)
 774                         dev->mtu = IPV6_MIN_MTU;
 775         }
 776         dev->iflink = tunnel->parms.link;
 777
 778         return 0;
 779 }
 780
 781 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 782 {
 783         struct ip_tunnel *tunnel = netdev_priv(dev);
 784         struct iphdr *iph = &tunnel->parms.iph;
 785
 786         tunnel->dev = dev;
 787         strcpy(tunnel->parms.name, dev->name);
 788
 789         iph->version            = 4;
 790         iph->protocol           = IPPROTO_IPV6;
 791         iph->ihl                = 5;
 792         iph->ttl                = 64;
 793
 794         dev_hold(dev);
 795         tunnels_wc[0]           = tunnel;
 796         return 0;
 797 }
 798
 799 static struct xfrm_tunnel sit_handler = {
 800         .handler        =       ipip6_rcv,
 801         .err_handler    =       ipip6_err,
 802         .priority       =       1,
 803 };
 804
 805 static void __exit sit_destroy_tunnels(void)
 806 {
 807         int prio;
 808
 809         for (prio = 1; prio < 4; prio++) {
 810                 int h;
 811                 for (h = 0; h < HASH_SIZE; h++) {
 812                         struct ip_tunnel *t;
 813                         while ((t = tunnels[prio][h]) != NULL)
 814                                 unregister_netdevice(t->dev);
 815                 }
 816         }
 817 }
 818
 819 static void __exit sit_cleanup(void)
 820 {
 821         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 822
 823         rtnl_lock();
 824         sit_destroy_tunnels();
 825         unregister_netdevice(ipip6_fb_tunnel_dev);
 826         rtnl_unlock();
 827 }
 828
 829 static int __init sit_init(void)
 830 {
 831         int err;
 832
 833         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 834
 835         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
 836                 printk(KERN_INFO "sit init: Can't add protocol\n");
 837                 return -EAGAIN;
 838         }
 839
 840         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
 841                                            ipip6_tunnel_setup);
 842         if (!ipip6_fb_tunnel_dev) {
 843                 err = -ENOMEM;
 844                 goto err1;
 845         }
 846
 847         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
 848
 849         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
 850                 goto err2;
 851
 852  out:
 853         return err;
 854  err2:
 855         free_netdev(ipip6_fb_tunnel_dev);
 856  err1:
 857         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 858         goto out;
 859 }
 860
 861 module_init(sit_init);
 862 module_exit(sit_cleanup);
 863 MODULE_LICENSE("GPL");
 864 MODULE_ALIAS("sit0");