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             <pedro_m@yahoo.com>
   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
  53 /*
  54    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  55
  56    For comments look at net/ipv4/ip_gre.c --ANK
  57  */
  58
  59 #define HASH_SIZE  16
  60 #define HASH(addr) ((addr^(addr>>4))&0xF)
  61
  62 static int ipip6_fb_tunnel_init(struct net_device *dev);
  63 static int ipip6_tunnel_init(struct net_device *dev);
  64
  65 static struct net_device ipip6_fb_tunnel_dev = {
  66         name:           "sit0",
  67         init:           ipip6_fb_tunnel_init,
  68 };
  69
  70 static struct ip_tunnel ipip6_fb_tunnel = {
  71         NULL, &ipip6_fb_tunnel_dev, {0, }, 0, 0, 0, 0, 0, 0, 0, {"sit0", }
  72 };
  73
  74 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  75 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  76 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  77 static struct ip_tunnel *tunnels_wc[1];
  78 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  79
  80 static rwlock_t ipip6_lock = RW_LOCK_UNLOCKED;
  81
  82 static struct ip_tunnel * ipip6_tunnel_lookup(u32 remote, u32 local)
  83 {
  84         unsigned h0 = HASH(remote);
  85         unsigned h1 = HASH(local);
  86         struct ip_tunnel *t;
  87
  88         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  89                 if (local == t->parms.iph.saddr &&
  90                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  91                         return t;
  92         }
  93         for (t = tunnels_r[h0]; t; t = t->next) {
  94                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  95                         return t;
  96         }
  97         for (t = tunnels_l[h1]; t; t = t->next) {
  98                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  99                         return t;
 100         }
 101         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
 102                 return t;
 103         return NULL;
 104 }
 105
 106 static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
 107 {
 108         u32 remote = t->parms.iph.daddr;
 109         u32 local = t->parms.iph.saddr;
 110         unsigned h = 0;
 111         int prio = 0;
 112
 113         if (remote) {
 114                 prio |= 2;
 115                 h ^= HASH(remote);
 116         }
 117         if (local) {
 118                 prio |= 1;
 119                 h ^= HASH(local);
 120         }
 121         return &tunnels[prio][h];
 122 }
 123
 124 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 125 {
 126         struct ip_tunnel **tp;
 127
 128         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 129                 if (t == *tp) {
 130                         write_lock_bh(&ipip6_lock);
 131                         *tp = t->next;
 132                         write_unlock_bh(&ipip6_lock);
 133                         break;
 134                 }
 135         }
 136 }
 137
 138 static void ipip6_tunnel_link(struct ip_tunnel *t)
 139 {
 140         struct ip_tunnel **tp = ipip6_bucket(t);
 141
 142         t->next = *tp;
 143         write_lock_bh(&ipip6_lock);
 144         *tp = t;
 145         write_unlock_bh(&ipip6_lock);
 146 }
 147
 148 struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 149 {
 150         u32 remote = parms->iph.daddr;
 151         u32 local = parms->iph.saddr;
 152         struct ip_tunnel *t, **tp, *nt;
 153         struct net_device *dev;
 154         unsigned h = 0;
 155         int prio = 0;
 156
 157         if (remote) {
 158                 prio |= 2;
 159                 h ^= HASH(remote);
 160         }
 161         if (local) {
 162                 prio |= 1;
 163                 h ^= HASH(local);
 164         }
 165         for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
 166                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 167                         return t;
 168         }
 169         if (!create)
 170                 return NULL;
 171
 172         MOD_INC_USE_COUNT;
 173         dev = kmalloc(sizeof(*dev) + sizeof(*t), GFP_KERNEL);
 174         if (dev == NULL) {
 175                 MOD_DEC_USE_COUNT;
 176                 return NULL;
 177         }
 178         memset(dev, 0, sizeof(*dev) + sizeof(*t));
 179         dev->priv = (void*)(dev+1);
 180         nt = (struct ip_tunnel*)dev->priv;
 181         nt->dev = dev;
 182         dev->init = ipip6_tunnel_init;
 183         dev->features |= NETIF_F_DYNALLOC;
 184         memcpy(&nt->parms, parms, sizeof(*parms));
 185         nt->parms.name[IFNAMSIZ-1] = '\0';
 186         strcpy(dev->name, nt->parms.name);
 187         if (dev->name[0] == 0) {
 188                 int i;
 189                 for (i=1; i<100; i++) {
 190                         sprintf(dev->name, "sit%d", i);
 191                         if (__dev_get_by_name(dev->name) == NULL)
 192                                 break;
 193                 }
 194                 if (i==100)
 195                         goto failed;
 196                 memcpy(nt->parms.name, dev->name, IFNAMSIZ);
 197         }
 198         if (register_netdevice(dev) < 0)
 199                 goto failed;
 200
 201         dev_hold(dev);
 202         ipip6_tunnel_link(nt);
 203         /* Do not decrement MOD_USE_COUNT here. */
 204         return nt;
 205
 206 failed:
 207         kfree(dev);
 208         MOD_DEC_USE_COUNT;
 209         return NULL;
 210 }
 211
 212 static void ipip6_tunnel_destructor(struct net_device *dev)
 213 {
 214         if (dev != &ipip6_fb_tunnel_dev) {
 215                 MOD_DEC_USE_COUNT;
 216         }
 217 }
 218
 219 static void ipip6_tunnel_uninit(struct net_device *dev)
 220 {
 221         if (dev == &ipip6_fb_tunnel_dev) {
 222                 write_lock_bh(&ipip6_lock);
 223                 tunnels_wc[0] = NULL;
 224                 write_unlock_bh(&ipip6_lock);
 225                 dev_put(dev);
 226         } else {
 227                 ipip6_tunnel_unlink((struct ip_tunnel*)dev->priv);
 228                 dev_put(dev);
 229         }
 230 }
 231
 232
 233 void ipip6_err(struct sk_buff *skb, u32 info)
 234 {
 235 #ifndef I_WISH_WORLD_WERE_PERFECT
 236
 237 /* It is not :-( All the routers (except for Linux) return only
 238    8 bytes of packet payload. It means, that precise relaying of
 239    ICMP in the real Internet is absolutely infeasible.
 240  */
 241         struct iphdr *iph = (struct iphdr*)skb->data;
 242         int type = skb->h.icmph->type;
 243         int code = skb->h.icmph->code;
 244         struct ip_tunnel *t;
 245
 246         switch (type) {
 247         default:
 248         case ICMP_PARAMETERPROB:
 249                 return;
 250
 251         case ICMP_DEST_UNREACH:
 252                 switch (code) {
 253                 case ICMP_SR_FAILED:
 254                 case ICMP_PORT_UNREACH:
 255                         /* Impossible event. */
 256                         return;
 257                 case ICMP_FRAG_NEEDED:
 258                         /* Soft state for pmtu is maintained by IP core. */
 259                         return;
 260                 default:
 261                         /* All others are translated to HOST_UNREACH.
 262                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 263                            I believe they are just ether pollution. --ANK
 264                          */
 265                         break;
 266                 }
 267                 break;
 268         case ICMP_TIME_EXCEEDED:
 269                 if (code != ICMP_EXC_TTL)
 270                         return;
 271                 break;
 272         }
 273
 274         read_lock(&ipip6_lock);
 275         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 276         if (t == NULL || t->parms.iph.daddr == 0)
 277                 goto out;
 278         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 279                 goto out;
 280
 281         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 282                 t->err_count++;
 283         else
 284                 t->err_count = 1;
 285         t->err_time = jiffies;
 286 out:
 287         read_unlock(&ipip6_lock);
 288         return;
 289 #else
 290         struct iphdr *iph = (struct iphdr*)dp;
 291         int hlen = iph->ihl<<2;
 292         struct ipv6hdr *iph6;
 293         int type = skb->h.icmph->type;
 294         int code = skb->h.icmph->code;
 295         int rel_type = 0;
 296         int rel_code = 0;
 297         int rel_info = 0;
 298         struct sk_buff *skb2;
 299         struct rt6_info *rt6i;
 300
 301         if (len < hlen + sizeof(struct ipv6hdr))
 302                 return;
 303         iph6 = (struct ipv6hdr*)(dp + hlen);
 304
 305         switch (type) {
 306         default:
 307                 return;
 308         case ICMP_PARAMETERPROB:
 309                 if (skb->h.icmph->un.gateway < hlen)
 310                         return;
 311
 312                 /* So... This guy found something strange INSIDE encapsulated
 313                    packet. Well, he is fool, but what can we do ?
 314                  */
 315                 rel_type = ICMPV6_PARAMPROB;
 316                 rel_info = skb->h.icmph->un.gateway - hlen;
 317                 break;
 318
 319         case ICMP_DEST_UNREACH:
 320                 switch (code) {
 321                 case ICMP_SR_FAILED:
 322                 case ICMP_PORT_UNREACH:
 323                         /* Impossible event. */
 324                         return;
 325                 case ICMP_FRAG_NEEDED:
 326                         /* Too complicated case ... */
 327                         return;
 328                 default:
 329                         /* All others are translated to HOST_UNREACH.
 330                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 331                            I believe, it is just ether pollution. --ANK
 332                          */
 333                         rel_type = ICMPV6_DEST_UNREACH;
 334                         rel_code = ICMPV6_ADDR_UNREACH;
 335                         break;
 336                 }
 337                 break;
 338         case ICMP_TIME_EXCEEDED:
 339                 if (code != ICMP_EXC_TTL)
 340                         return;
 341                 rel_type = ICMPV6_TIME_EXCEED;
 342                 rel_code = ICMPV6_EXC_HOPLIMIT;
 343                 break;
 344         }
 345
 346         /* Prepare fake skb to feed it to icmpv6_send */
 347         skb2 = skb_clone(skb, GFP_ATOMIC);
 348         if (skb2 == NULL)
 349                 return;
 350         dst_release(skb2->dst);
 351         skb2->dst = NULL;
 352         skb_pull(skb2, skb->data - (u8*)iph6);
 353         skb2->nh.raw = skb2->data;
 354
 355         /* Try to guess incoming interface */
 356         rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
 357         if (rt6i && rt6i->rt6i_dev) {
 358                 skb2->dev = rt6i->rt6i_dev;
 359
 360                 rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
 361
 362                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 363                         struct ip_tunnel * t = (struct ip_tunnel*)rt6i->rt6i_dev->priv;
 364                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 365                                 rel_type = ICMPV6_DEST_UNREACH;
 366                                 rel_code = ICMPV6_ADDR_UNREACH;
 367                         }
 368                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 369                 }
 370         }
 371         kfree_skb(skb2);
 372         return;
 373 #endif
 374 }
 375
 376 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 377 {
 378         if (INET_ECN_is_ce(iph->tos) &&
 379             INET_ECN_is_not_ce(ip6_get_dsfield(skb->nh.ipv6h)))
 380                 IP6_ECN_set_ce(skb->nh.ipv6h);
 381 }
 382
 383 int ipip6_rcv(struct sk_buff *skb)
 384 {
 385         struct iphdr *iph;
 386         struct ip_tunnel *tunnel;
 387
 388         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 389                 goto out;
 390
 391         iph = skb->nh.iph;
 392
 393         read_lock(&ipip6_lock);
 394         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 395                 skb->mac.raw = skb->nh.raw;
 396                 skb->nh.raw = skb->data;
 397                 memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
 398                 skb->protocol = htons(ETH_P_IPV6);
 399                 skb->pkt_type = PACKET_HOST;
 400                 tunnel->stat.rx_packets++;
 401                 tunnel->stat.rx_bytes += skb->len;
 402                 skb->dev = tunnel->dev;
 403                 dst_release(skb->dst);
 404                 skb->dst = NULL;
 405                 nf_reset(skb);
 406                 ipip6_ecn_decapsulate(iph, skb);
 407                 netif_rx(skb);
 408                 read_unlock(&ipip6_lock);
 409                 return 0;
 410         }
 411
 412         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0);
 413         read_unlock(&ipip6_lock);
 414 out:
 415         kfree_skb(skb);
 416         return 0;
 417 }
 418
 419 /* Need this wrapper because NF_HOOK takes the function address */
 420 static inline int do_ip_send(struct sk_buff *skb)
 421 {
 422         return ip_send(skb);
 423 }
 424
 425
 426 /* Returns the embedded IPv4 address if the IPv6 address
 427    comes from 6to4 (draft-ietf-ngtrans-6to4-04) addr space */
 428
 429 static inline u32 try_6to4(struct in6_addr *v6dst)
 430 {
 431         u32 dst = 0;
 432
 433         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 434                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 435                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 436         }
 437         return dst;
 438 }
 439
 440 /*
 441  *      This function assumes it is being called from dev_queue_xmit()
 442  *      and that skb is filled properly by that function.
 443  */
 444
 445 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 446 {
 447         struct ip_tunnel *tunnel = (struct ip_tunnel*)dev->priv;
 448         struct net_device_stats *stats = &tunnel->stat;
 449         struct iphdr  *tiph = &tunnel->parms.iph;
 450         struct ipv6hdr *iph6 = skb->nh.ipv6h;
 451         u8     tos = tunnel->parms.iph.tos;
 452         struct rtable *rt;                      /* Route to the other host */
 453         struct net_device *tdev;                        /* Device to other host */
 454         struct iphdr  *iph;                     /* Our new IP header */
 455         int    max_headroom;                    /* The extra header space needed */
 456         u32    dst = tiph->daddr;
 457         int    mtu;
 458         struct in6_addr *addr6;
 459         int addr_type;
 460
 461         if (tunnel->recursion++) {
 462                 tunnel->stat.collisions++;
 463                 goto tx_error;
 464         }
 465
 466         if (skb->protocol != htons(ETH_P_IPV6))
 467                 goto tx_error;
 468
 469         if (!dst)
 470                 dst = try_6to4(&iph6->daddr);
 471
 472         if (!dst) {
 473                 struct neighbour *neigh = NULL;
 474
 475                 if (skb->dst)
 476                         neigh = skb->dst->neighbour;
 477
 478                 if (neigh == NULL) {
 479                         if (net_ratelimit())
 480                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 481                         goto tx_error;
 482                 }
 483
 484                 addr6 = (struct in6_addr*)&neigh->primary_key;
 485                 addr_type = ipv6_addr_type(addr6);
 486
 487                 if (addr_type == IPV6_ADDR_ANY) {
 488                         addr6 = &skb->nh.ipv6h->daddr;
 489                         addr_type = ipv6_addr_type(addr6);
 490                 }
 491
 492                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 493                         goto tx_error_icmp;
 494
 495                 dst = addr6->s6_addr32[3];
 496         }
 497
 498         if (ip_route_output(&rt, dst, tiph->saddr, RT_TOS(tos), tunnel->parms.link)) {
 499                 tunnel->stat.tx_carrier_errors++;
 500                 goto tx_error_icmp;
 501         }
 502         if (rt->rt_type != RTN_UNICAST) {
 503                 ip_rt_put(rt);
 504                 tunnel->stat.tx_carrier_errors++;
 505                 goto tx_error_icmp;
 506         }
 507         tdev = rt->u.dst.dev;
 508
 509         if (tdev == dev) {
 510                 ip_rt_put(rt);
 511                 tunnel->stat.collisions++;
 512                 goto tx_error;
 513         }
 514
 515         if (tiph->frag_off)
 516                 mtu = rt->u.dst.pmtu - sizeof(struct iphdr);
 517         else
 518                 mtu = skb->dst ? skb->dst->pmtu : dev->mtu;
 519
 520         if (mtu < 68) {
 521                 tunnel->stat.collisions++;
 522                 ip_rt_put(rt);
 523                 goto tx_error;
 524         }
 525         if (mtu < IPV6_MIN_MTU)
 526                 mtu = IPV6_MIN_MTU;
 527         if (skb->dst && mtu < skb->dst->pmtu) {
 528                 struct rt6_info *rt6 = (struct rt6_info*)skb->dst;
 529                 if (mtu < rt6->u.dst.pmtu) {
 530                         if (tunnel->parms.iph.daddr || rt6->rt6i_dst.plen == 128) {
 531                                 rt6->rt6i_flags |= RTF_MODIFIED;
 532                                 rt6->u.dst.pmtu = mtu;
 533                         }
 534                 }
 535         }
 536         if (skb->len > mtu) {
 537                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 538                 ip_rt_put(rt);
 539                 goto tx_error;
 540         }
 541
 542         if (tunnel->err_count > 0) {
 543                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 544                         tunnel->err_count--;
 545                         dst_link_failure(skb);
 546                 } else
 547                         tunnel->err_count = 0;
 548         }
 549
 550         /*
 551          * Okay, now see if we can stuff it in the buffer as-is.
 552          */
 553         max_headroom = (((tdev->hard_header_len+15)&~15)+sizeof(struct iphdr));
 554
 555         if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
 556                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 557                 if (!new_skb) {
 558                         ip_rt_put(rt);
 559                         stats->tx_dropped++;
 560                         dev_kfree_skb(skb);
 561                         tunnel->recursion--;
 562                         return 0;
 563                 }
 564                 if (skb->sk)
 565                         skb_set_owner_w(new_skb, skb->sk);
 566                 dev_kfree_skb(skb);
 567                 skb = new_skb;
 568                 iph6 = skb->nh.ipv6h;
 569         }
 570
 571         skb->h.raw = skb->nh.raw;
 572         skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
 573         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 574         dst_release(skb->dst);
 575         skb->dst = &rt->u.dst;
 576
 577         /*
 578          *      Push down and install the IPIP header.
 579          */
 580
 581         iph                     =       skb->nh.iph;
 582         iph->version            =       4;
 583         iph->ihl                =       sizeof(struct iphdr)>>2;
 584         if (mtu > IPV6_MIN_MTU)
 585                 iph->frag_off   =       htons(IP_DF);
 586         else
 587                 iph->frag_off   =       0;
 588
 589         iph->protocol           =       IPPROTO_IPV6;
 590         iph->tos                =       INET_ECN_encapsulate(tos, ip6_get_dsfield(iph6));
 591         iph->daddr              =       rt->rt_dst;
 592         iph->saddr              =       rt->rt_src;
 593
 594         if ((iph->ttl = tiph->ttl) == 0)
 595                 iph->ttl        =       iph6->hop_limit;
 596
 597         nf_reset(skb);
 598
 599         IPTUNNEL_XMIT();
 600         tunnel->recursion--;
 601         return 0;
 602
 603 tx_error_icmp:
 604         dst_link_failure(skb);
 605 tx_error:
 606         stats->tx_errors++;
 607         dev_kfree_skb(skb);
 608         tunnel->recursion--;
 609         return 0;
 610 }
 611
 612 static int
 613 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 614 {
 615         int err = 0;
 616         struct ip_tunnel_parm p;
 617         struct ip_tunnel *t;
 618
 619         MOD_INC_USE_COUNT;
 620
 621         switch (cmd) {
 622         case SIOCGETTUNNEL:
 623                 t = NULL;
 624                 if (dev == &ipip6_fb_tunnel_dev) {
 625                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 626                                 err = -EFAULT;
 627                                 break;
 628                         }
 629                         t = ipip6_tunnel_locate(&p, 0);
 630                 }
 631                 if (t == NULL)
 632                         t = (struct ip_tunnel*)dev->priv;
 633                 memcpy(&p, &t->parms, sizeof(p));
 634                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 635                         err = -EFAULT;
 636                 break;
 637
 638         case SIOCADDTUNNEL:
 639         case SIOCCHGTUNNEL:
 640                 err = -EPERM;
 641                 if (!capable(CAP_NET_ADMIN))
 642                         goto done;
 643
 644                 err = -EFAULT;
 645                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 646                         goto done;
 647
 648                 err = -EINVAL;
 649                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 650                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 651                         goto done;
 652                 if (p.iph.ttl)
 653                         p.iph.frag_off |= htons(IP_DF);
 654
 655                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 656
 657                 if (dev != &ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL &&
 658                     t != &ipip6_fb_tunnel) {
 659                         if (t != NULL) {
 660                                 if (t->dev != dev) {
 661                                         err = -EEXIST;
 662                                         break;
 663                                 }
 664                         } else {
 665                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 666                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 667                                         err = -EINVAL;
 668                                         break;
 669                                 }
 670                                 t = (struct ip_tunnel*)dev->priv;
 671                                 ipip6_tunnel_unlink(t);
 672                                 t->parms.iph.saddr = p.iph.saddr;
 673                                 t->parms.iph.daddr = p.iph.daddr;
 674                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 675                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 676                                 ipip6_tunnel_link(t);
 677                                 netdev_state_change(dev);
 678                         }
 679                 }
 680
 681                 if (t) {
 682                         err = 0;
 683                         if (cmd == SIOCCHGTUNNEL) {
 684                                 t->parms.iph.ttl = p.iph.ttl;
 685                                 t->parms.iph.tos = p.iph.tos;
 686                         }
 687                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 688                                 err = -EFAULT;
 689                 } else
 690                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 691                 break;
 692
 693         case SIOCDELTUNNEL:
 694                 err = -EPERM;
 695                 if (!capable(CAP_NET_ADMIN))
 696                         goto done;
 697
 698                 if (dev == &ipip6_fb_tunnel_dev) {
 699                         err = -EFAULT;
 700                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 701                                 goto done;
 702                         err = -ENOENT;
 703                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 704                                 goto done;
 705                         err = -EPERM;
 706                         if (t == &ipip6_fb_tunnel)
 707                                 goto done;
 708                         dev = t->dev;
 709                 }
 710                 err = unregister_netdevice(dev);
 711                 break;
 712
 713         default:
 714                 err = -EINVAL;
 715         }
 716
 717 done:
 718         MOD_DEC_USE_COUNT;
 719         return err;
 720 }
 721
 722 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 723 {
 724         return &(((struct ip_tunnel*)dev->priv)->stat);
 725 }
 726
 727 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 728 {
 729         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 730                 return -EINVAL;
 731         dev->mtu = new_mtu;
 732         return 0;
 733 }
 734
 735 static void ipip6_tunnel_init_gen(struct net_device *dev)
 736 {
 737         struct ip_tunnel *t = (struct ip_tunnel*)dev->priv;
 738
 739         dev->destructor         = ipip6_tunnel_destructor;
 740         dev->uninit             = ipip6_tunnel_uninit;
 741         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 742         dev->get_stats          = ipip6_tunnel_get_stats;
 743         dev->do_ioctl           = ipip6_tunnel_ioctl;
 744         dev->change_mtu         = ipip6_tunnel_change_mtu;
 745
 746         dev->type               = ARPHRD_SIT;
 747         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 748         dev->mtu                = 1500 - sizeof(struct iphdr);
 749         dev->flags              = IFF_NOARP;
 750         dev->iflink             = 0;
 751         dev->addr_len           = 4;
 752         memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 753         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 754 }
 755
 756 static int ipip6_tunnel_init(struct net_device *dev)
 757 {
 758         struct net_device *tdev = NULL;
 759         struct ip_tunnel *tunnel;
 760         struct iphdr *iph;
 761
 762         tunnel = (struct ip_tunnel*)dev->priv;
 763         iph = &tunnel->parms.iph;
 764
 765         ipip6_tunnel_init_gen(dev);
 766
 767         if (iph->daddr) {
 768                 struct rtable *rt;
 769                 if (!ip_route_output(&rt, iph->daddr, iph->saddr, RT_TOS(iph->tos), tunnel->parms.link)) {
 770                         tdev = rt->u.dst.dev;
 771                         ip_rt_put(rt);
 772                 }
 773                 dev->flags |= IFF_POINTOPOINT;
 774         }
 775
 776         if (!tdev && tunnel->parms.link)
 777                 tdev = __dev_get_by_index(tunnel->parms.link);
 778
 779         if (tdev) {
 780                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 781                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 782                 if (dev->mtu < IPV6_MIN_MTU)
 783                         dev->mtu = IPV6_MIN_MTU;
 784         }
 785         dev->iflink = tunnel->parms.link;
 786
 787         return 0;
 788 }
 789
 790 #ifdef MODULE
 791 static int ipip6_fb_tunnel_open(struct net_device *dev)
 792 {
 793         MOD_INC_USE_COUNT;
 794         return 0;
 795 }
 796
 797 static int ipip6_fb_tunnel_close(struct net_device *dev)
 798 {
 799         MOD_DEC_USE_COUNT;
 800         return 0;
 801 }
 802 #endif
 803
 804 int __init ipip6_fb_tunnel_init(struct net_device *dev)
 805 {
 806         struct iphdr *iph;
 807
 808         ipip6_tunnel_init_gen(dev);
 809 #ifdef MODULE
 810         dev->open               = ipip6_fb_tunnel_open;
 811         dev->stop               = ipip6_fb_tunnel_close;
 812 #endif
 813
 814         iph = &ipip6_fb_tunnel.parms.iph;
 815         iph->version            = 4;
 816         iph->protocol           = IPPROTO_IPV6;
 817         iph->ihl                = 5;
 818         iph->ttl                = 64;
 819
 820         dev_hold(dev);
 821         tunnels_wc[0]           = &ipip6_fb_tunnel;
 822         return 0;
 823 }
 824
 825 static struct inet_protocol sit_protocol = {
 826         ipip6_rcv,
 827         ipip6_err,
 828         0,
 829         IPPROTO_IPV6,
 830         0,
 831         NULL,
 832         "IPv6"
 833 };
 834
 835 #ifdef MODULE
 836 void sit_cleanup(void)
 837 {
 838         inet_del_protocol(&sit_protocol);
 839         unregister_netdev(&ipip6_fb_tunnel_dev);
 840 }
 841 #endif
 842
 843 int __init sit_init(void)
 844 {
 845         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 846
 847         ipip6_fb_tunnel_dev.priv = (void*)&ipip6_fb_tunnel;
 848         strcpy(ipip6_fb_tunnel_dev.name, ipip6_fb_tunnel.parms.name);
 849         register_netdev(&ipip6_fb_tunnel_dev);
 850         inet_add_protocol(&sit_protocol);
 851         return 0;
 852 }