net/ipv6/sit.c

   1 /*
   2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
  10  *
  11  *      This program is free software; you can redistribute it and/or
  12  *      modify it under the terms of the GNU General Public License
  13  *      as published by the Free Software Foundation; either version
  14  *      2 of the License, or (at your option) any later version.
  15  *
  16  *      Changes:
  17  * Roger Venning <r.venning@telstra.com>:       6to4 support
  18  * Nate Thompson <nate@thebog.net>:             6to4 support
  19  */
  20
  21 #include <linux/module.h>
  22 #include <linux/capability.h>
  23 #include <linux/errno.h>
  24 #include <linux/types.h>
  25 #include <linux/socket.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/in6.h>
  29 #include <linux/netdevice.h>
  30 #include <linux/if_arp.h>
  31 #include <linux/icmp.h>
  32 #include <asm/uaccess.h>
  33 #include <linux/init.h>
  34 #include <linux/netfilter_ipv4.h>
  35 #include <linux/if_ether.h>
  36
  37 #include <net/sock.h>
  38 #include <net/snmp.h>
  39
  40 #include <net/ipv6.h>
  41 #include <net/protocol.h>
  42 #include <net/transp_v6.h>
  43 #include <net/ip6_fib.h>
  44 #include <net/ip6_route.h>
  45 #include <net/ndisc.h>
  46 #include <net/addrconf.h>
  47 #include <net/ip.h>
  48 #include <net/udp.h>
  49 #include <net/icmp.h>
  50 #include <net/ipip.h>
  51 #include <net/inet_ecn.h>
  52 #include <net/xfrm.h>
  53 #include <net/dsfield.h>
  54
  55 /*
  56    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  57
  58    For comments look at net/ipv4/ip_gre.c --ANK
  59  */
  60
  61 #define HASH_SIZE  16
  62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  63
  64 static int ipip6_fb_tunnel_init(struct net_device *dev);
  65 static int ipip6_tunnel_init(struct net_device *dev);
  66 static void ipip6_tunnel_setup(struct net_device *dev);
  67
  68 static struct net_device *ipip6_fb_tunnel_dev;
  69
  70 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  71 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  72 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  73 static struct ip_tunnel *tunnels_wc[1];
  74 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  75
  76 static DEFINE_RWLOCK(ipip6_lock);
  77
  78 static struct ip_tunnel * ipip6_tunnel_lookup(__be32 remote, __be32 local)
  79 {
  80         unsigned h0 = HASH(remote);
  81         unsigned h1 = HASH(local);
  82         struct ip_tunnel *t;
  83
  84         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  85                 if (local == t->parms.iph.saddr &&
  86                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  87                         return t;
  88         }
  89         for (t = tunnels_r[h0]; t; t = t->next) {
  90                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  91                         return t;
  92         }
  93         for (t = tunnels_l[h1]; t; t = t->next) {
  94                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  95                         return t;
  96         }
  97         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
  98                 return t;
  99         return NULL;
 100 }
 101
 102 static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
 103 {
 104         __be32 remote = t->parms.iph.daddr;
 105         __be32 local = t->parms.iph.saddr;
 106         unsigned h = 0;
 107         int prio = 0;
 108
 109         if (remote) {
 110                 prio |= 2;
 111                 h ^= HASH(remote);
 112         }
 113         if (local) {
 114                 prio |= 1;
 115                 h ^= HASH(local);
 116         }
 117         return &tunnels[prio][h];
 118 }
 119
 120 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 121 {
 122         struct ip_tunnel **tp;
 123
 124         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 125                 if (t == *tp) {
 126                         write_lock_bh(&ipip6_lock);
 127                         *tp = t->next;
 128                         write_unlock_bh(&ipip6_lock);
 129                         break;
 130                 }
 131         }
 132 }
 133
 134 static void ipip6_tunnel_link(struct ip_tunnel *t)
 135 {
 136         struct ip_tunnel **tp = ipip6_bucket(t);
 137
 138         t->next = *tp;
 139         write_lock_bh(&ipip6_lock);
 140         *tp = t;
 141         write_unlock_bh(&ipip6_lock);
 142 }
 143
 144 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 145 {
 146         __be32 remote = parms->iph.daddr;
 147         __be32 local = parms->iph.saddr;
 148         struct ip_tunnel *t, **tp, *nt;
 149         struct net_device *dev;
 150         unsigned h = 0;
 151         int prio = 0;
 152         char name[IFNAMSIZ];
 153
 154         if (remote) {
 155                 prio |= 2;
 156                 h ^= HASH(remote);
 157         }
 158         if (local) {
 159                 prio |= 1;
 160                 h ^= HASH(local);
 161         }
 162         for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
 163                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 164                         return t;
 165         }
 166         if (!create)
 167                 goto failed;
 168
 169         if (parms->name[0])
 170                 strlcpy(name, parms->name, IFNAMSIZ);
 171         else {
 172                 int i;
 173                 for (i=1; i<100; i++) {
 174                         sprintf(name, "sit%d", i);
 175                         if (__dev_get_by_name(name) == NULL)
 176                                 break;
 177                 }
 178                 if (i==100)
 179                         goto failed;
 180         }
 181
 182         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 183         if (dev == NULL)
 184                 return NULL;
 185
 186         nt = netdev_priv(dev);
 187         dev->init = ipip6_tunnel_init;
 188         nt->parms = *parms;
 189
 190         if (register_netdevice(dev) < 0) {
 191                 free_netdev(dev);
 192                 goto failed;
 193         }
 194
 195         dev_hold(dev);
 196
 197         ipip6_tunnel_link(nt);
 198         return nt;
 199
 200 failed:
 201         return NULL;
 202 }
 203
 204 static void ipip6_tunnel_uninit(struct net_device *dev)
 205 {
 206         if (dev == ipip6_fb_tunnel_dev) {
 207                 write_lock_bh(&ipip6_lock);
 208                 tunnels_wc[0] = NULL;
 209                 write_unlock_bh(&ipip6_lock);
 210                 dev_put(dev);
 211         } else {
 212                 ipip6_tunnel_unlink(netdev_priv(dev));
 213                 dev_put(dev);
 214         }
 215 }
 216
 217
 218 static int 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         int err;
 231
 232         switch (type) {
 233         default:
 234         case ICMP_PARAMETERPROB:
 235                 return 0;
 236
 237         case ICMP_DEST_UNREACH:
 238                 switch (code) {
 239                 case ICMP_SR_FAILED:
 240                 case ICMP_PORT_UNREACH:
 241                         /* Impossible event. */
 242                         return 0;
 243                 case ICMP_FRAG_NEEDED:
 244                         /* Soft state for pmtu is maintained by IP core. */
 245                         return 0;
 246                 default:
 247                         /* All others are translated to HOST_UNREACH.
 248                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 249                            I believe they are just ether pollution. --ANK
 250                          */
 251                         break;
 252                 }
 253                 break;
 254         case ICMP_TIME_EXCEEDED:
 255                 if (code != ICMP_EXC_TTL)
 256                         return 0;
 257                 break;
 258         }
 259
 260         err = -ENOENT;
 261
 262         read_lock(&ipip6_lock);
 263         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 264         if (t == NULL || t->parms.iph.daddr == 0)
 265                 goto out;
 266
 267         err = 0;
 268         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 269                 goto out;
 270
 271         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 272                 t->err_count++;
 273         else
 274                 t->err_count = 1;
 275         t->err_time = jiffies;
 276 out:
 277         read_unlock(&ipip6_lock);
 278         return err;
 279 #else
 280         struct iphdr *iph = (struct iphdr*)dp;
 281         int hlen = iph->ihl<<2;
 282         struct ipv6hdr *iph6;
 283         int type = skb->h.icmph->type;
 284         int code = skb->h.icmph->code;
 285         int rel_type = 0;
 286         int rel_code = 0;
 287         int rel_info = 0;
 288         struct sk_buff *skb2;
 289         struct rt6_info *rt6i;
 290
 291         if (len < hlen + sizeof(struct ipv6hdr))
 292                 return;
 293         iph6 = (struct ipv6hdr*)(dp + hlen);
 294
 295         switch (type) {
 296         default:
 297                 return;
 298         case ICMP_PARAMETERPROB:
 299                 if (skb->h.icmph->un.gateway < hlen)
 300                         return;
 301
 302                 /* So... This guy found something strange INSIDE encapsulated
 303                    packet. Well, he is fool, but what can we do ?
 304                  */
 305                 rel_type = ICMPV6_PARAMPROB;
 306                 rel_info = skb->h.icmph->un.gateway - hlen;
 307                 break;
 308
 309         case ICMP_DEST_UNREACH:
 310                 switch (code) {
 311                 case ICMP_SR_FAILED:
 312                 case ICMP_PORT_UNREACH:
 313                         /* Impossible event. */
 314                         return;
 315                 case ICMP_FRAG_NEEDED:
 316                         /* Too complicated case ... */
 317                         return;
 318                 default:
 319                         /* All others are translated to HOST_UNREACH.
 320                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 321                            I believe, it is just ether pollution. --ANK
 322                          */
 323                         rel_type = ICMPV6_DEST_UNREACH;
 324                         rel_code = ICMPV6_ADDR_UNREACH;
 325                         break;
 326                 }
 327                 break;
 328         case ICMP_TIME_EXCEEDED:
 329                 if (code != ICMP_EXC_TTL)
 330                         return;
 331                 rel_type = ICMPV6_TIME_EXCEED;
 332                 rel_code = ICMPV6_EXC_HOPLIMIT;
 333                 break;
 334         }
 335
 336         /* Prepare fake skb to feed it to icmpv6_send */
 337         skb2 = skb_clone(skb, GFP_ATOMIC);
 338         if (skb2 == NULL)
 339                 return 0;
 340         dst_release(skb2->dst);
 341         skb2->dst = NULL;
 342         skb_pull(skb2, skb->data - (u8*)iph6);
 343         skb2->nh.raw = skb2->data;
 344
 345         /* Try to guess incoming interface */
 346         rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
 347         if (rt6i && rt6i->rt6i_dev) {
 348                 skb2->dev = rt6i->rt6i_dev;
 349
 350                 rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
 351
 352                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 353                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 354                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 355                                 rel_type = ICMPV6_DEST_UNREACH;
 356                                 rel_code = ICMPV6_ADDR_UNREACH;
 357                         }
 358                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 359                 }
 360         }
 361         kfree_skb(skb2);
 362         return 0;
 363 #endif
 364 }
 365
 366 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 367 {
 368         if (INET_ECN_is_ce(iph->tos))
 369                 IP6_ECN_set_ce(skb->nh.ipv6h);
 370 }
 371
 372 static int ipip6_rcv(struct sk_buff *skb)
 373 {
 374         struct iphdr *iph;
 375         struct ip_tunnel *tunnel;
 376
 377         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 378                 goto out;
 379
 380         iph = skb->nh.iph;
 381
 382         read_lock(&ipip6_lock);
 383         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 384                 secpath_reset(skb);
 385                 skb->mac.raw = skb->nh.raw;
 386                 skb->nh.raw = skb->data;
 387                 IPCB(skb)->flags = 0;
 388                 skb->protocol = htons(ETH_P_IPV6);
 389                 skb->pkt_type = PACKET_HOST;
 390                 tunnel->stat.rx_packets++;
 391                 tunnel->stat.rx_bytes += skb->len;
 392                 skb->dev = tunnel->dev;
 393                 dst_release(skb->dst);
 394                 skb->dst = NULL;
 395                 nf_reset(skb);
 396                 ipip6_ecn_decapsulate(iph, skb);
 397                 netif_rx(skb);
 398                 read_unlock(&ipip6_lock);
 399                 return 0;
 400         }
 401
 402         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 403         kfree_skb(skb);
 404         read_unlock(&ipip6_lock);
 405 out:
 406         return 0;
 407 }
 408
 409 /* Returns the embedded IPv4 address if the IPv6 address
 410    comes from 6to4 (RFC 3056) addr space */
 411
 412 static inline __be32 try_6to4(struct in6_addr *v6dst)
 413 {
 414         __be32 dst = 0;
 415
 416         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 417                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 418                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 419         }
 420         return dst;
 421 }
 422
 423 /*
 424  *      This function assumes it is being called from dev_queue_xmit()
 425  *      and that skb is filled properly by that function.
 426  */
 427
 428 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 429 {
 430         struct ip_tunnel *tunnel = netdev_priv(dev);
 431         struct net_device_stats *stats = &tunnel->stat;
 432         struct iphdr  *tiph = &tunnel->parms.iph;
 433         struct ipv6hdr *iph6 = skb->nh.ipv6h;
 434         u8     tos = tunnel->parms.iph.tos;
 435         struct rtable *rt;                      /* Route to the other host */
 436         struct net_device *tdev;                        /* Device to other host */
 437         struct iphdr  *iph;                     /* Our new IP header */
 438         int    max_headroom;                    /* The extra header space needed */
 439         __be32 dst = tiph->daddr;
 440         int    mtu;
 441         struct in6_addr *addr6;
 442         int addr_type;
 443
 444         if (tunnel->recursion++) {
 445                 tunnel->stat.collisions++;
 446                 goto tx_error;
 447         }
 448
 449         if (skb->protocol != htons(ETH_P_IPV6))
 450                 goto tx_error;
 451
 452         if (!dst)
 453                 dst = try_6to4(&iph6->daddr);
 454
 455         if (!dst) {
 456                 struct neighbour *neigh = NULL;
 457
 458                 if (skb->dst)
 459                         neigh = skb->dst->neighbour;
 460
 461                 if (neigh == NULL) {
 462                         if (net_ratelimit())
 463                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 464                         goto tx_error;
 465                 }
 466
 467                 addr6 = (struct in6_addr*)&neigh->primary_key;
 468                 addr_type = ipv6_addr_type(addr6);
 469
 470                 if (addr_type == IPV6_ADDR_ANY) {
 471                         addr6 = &skb->nh.ipv6h->daddr;
 472                         addr_type = ipv6_addr_type(addr6);
 473                 }
 474
 475                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 476                         goto tx_error_icmp;
 477
 478                 dst = addr6->s6_addr32[3];
 479         }
 480
 481         {
 482                 struct flowi fl = { .nl_u = { .ip4_u =
 483                                               { .daddr = dst,
 484                                                 .saddr = tiph->saddr,
 485                                                 .tos = RT_TOS(tos) } },
 486                                     .oif = tunnel->parms.link,
 487                                     .proto = IPPROTO_IPV6 };
 488                 if (ip_route_output_key(&rt, &fl)) {
 489                         tunnel->stat.tx_carrier_errors++;
 490                         goto tx_error_icmp;
 491                 }
 492         }
 493         if (rt->rt_type != RTN_UNICAST) {
 494                 ip_rt_put(rt);
 495                 tunnel->stat.tx_carrier_errors++;
 496                 goto tx_error_icmp;
 497         }
 498         tdev = rt->u.dst.dev;
 499
 500         if (tdev == dev) {
 501                 ip_rt_put(rt);
 502                 tunnel->stat.collisions++;
 503                 goto tx_error;
 504         }
 505
 506         if (tiph->frag_off)
 507                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 508         else
 509                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 510
 511         if (mtu < 68) {
 512                 tunnel->stat.collisions++;
 513                 ip_rt_put(rt);
 514                 goto tx_error;
 515         }
 516         if (mtu < IPV6_MIN_MTU)
 517                 mtu = IPV6_MIN_MTU;
 518         if (tunnel->parms.iph.daddr && skb->dst)
 519                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 520
 521         if (skb->len > mtu) {
 522                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 523                 ip_rt_put(rt);
 524                 goto tx_error;
 525         }
 526
 527         if (tunnel->err_count > 0) {
 528                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 529                         tunnel->err_count--;
 530                         dst_link_failure(skb);
 531                 } else
 532                         tunnel->err_count = 0;
 533         }
 534
 535         /*
 536          * Okay, now see if we can stuff it in the buffer as-is.
 537          */
 538         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 539
 540         if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
 541                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 542                 if (!new_skb) {
 543                         ip_rt_put(rt);
 544                         stats->tx_dropped++;
 545                         dev_kfree_skb(skb);
 546                         tunnel->recursion--;
 547                         return 0;
 548                 }
 549                 if (skb->sk)
 550                         skb_set_owner_w(new_skb, skb->sk);
 551                 dev_kfree_skb(skb);
 552                 skb = new_skb;
 553                 iph6 = skb->nh.ipv6h;
 554         }
 555
 556         skb->h.raw = skb->nh.raw;
 557         skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
 558         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 559         IPCB(skb)->flags = 0;
 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, ipv6_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         nf_reset(skb);
 584
 585         IPTUNNEL_XMIT();
 586         tunnel->recursion--;
 587         return 0;
 588
 589 tx_error_icmp:
 590         dst_link_failure(skb);
 591 tx_error:
 592         stats->tx_errors++;
 593         dev_kfree_skb(skb);
 594         tunnel->recursion--;
 595         return 0;
 596 }
 597
 598 static int
 599 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 600 {
 601         int err = 0;
 602         struct ip_tunnel_parm p;
 603         struct ip_tunnel *t;
 604
 605         switch (cmd) {
 606         case SIOCGETTUNNEL:
 607                 t = NULL;
 608                 if (dev == ipip6_fb_tunnel_dev) {
 609                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 610                                 err = -EFAULT;
 611                                 break;
 612                         }
 613                         t = ipip6_tunnel_locate(&p, 0);
 614                 }
 615                 if (t == NULL)
 616                         t = netdev_priv(dev);
 617                 memcpy(&p, &t->parms, sizeof(p));
 618                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 619                         err = -EFAULT;
 620                 break;
 621
 622         case SIOCADDTUNNEL:
 623         case SIOCCHGTUNNEL:
 624                 err = -EPERM;
 625                 if (!capable(CAP_NET_ADMIN))
 626                         goto done;
 627
 628                 err = -EFAULT;
 629                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 630                         goto done;
 631
 632                 err = -EINVAL;
 633                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 634                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 635                         goto done;
 636                 if (p.iph.ttl)
 637                         p.iph.frag_off |= htons(IP_DF);
 638
 639                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 640
 641                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 642                         if (t != NULL) {
 643                                 if (t->dev != dev) {
 644                                         err = -EEXIST;
 645                                         break;
 646                                 }
 647                         } else {
 648                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 649                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 650                                         err = -EINVAL;
 651                                         break;
 652                                 }
 653                                 t = netdev_priv(dev);
 654                                 ipip6_tunnel_unlink(t);
 655                                 t->parms.iph.saddr = p.iph.saddr;
 656                                 t->parms.iph.daddr = p.iph.daddr;
 657                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 658                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 659                                 ipip6_tunnel_link(t);
 660                                 netdev_state_change(dev);
 661                         }
 662                 }
 663
 664                 if (t) {
 665                         err = 0;
 666                         if (cmd == SIOCCHGTUNNEL) {
 667                                 t->parms.iph.ttl = p.iph.ttl;
 668                                 t->parms.iph.tos = p.iph.tos;
 669                         }
 670                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 671                                 err = -EFAULT;
 672                 } else
 673                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 674                 break;
 675
 676         case SIOCDELTUNNEL:
 677                 err = -EPERM;
 678                 if (!capable(CAP_NET_ADMIN))
 679                         goto done;
 680
 681                 if (dev == ipip6_fb_tunnel_dev) {
 682                         err = -EFAULT;
 683                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 684                                 goto done;
 685                         err = -ENOENT;
 686                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 687                                 goto done;
 688                         err = -EPERM;
 689                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 690                                 goto done;
 691                         dev = t->dev;
 692                 }
 693                 unregister_netdevice(dev);
 694                 err = 0;
 695                 break;
 696
 697         default:
 698                 err = -EINVAL;
 699         }
 700
 701 done:
 702         return err;
 703 }
 704
 705 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 706 {
 707         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 708 }
 709
 710 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 711 {
 712         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 713                 return -EINVAL;
 714         dev->mtu = new_mtu;
 715         return 0;
 716 }
 717
 718 static void ipip6_tunnel_setup(struct net_device *dev)
 719 {
 720         SET_MODULE_OWNER(dev);
 721         dev->uninit             = ipip6_tunnel_uninit;
 722         dev->destructor         = free_netdev;
 723         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 724         dev->get_stats          = ipip6_tunnel_get_stats;
 725         dev->do_ioctl           = ipip6_tunnel_ioctl;
 726         dev->change_mtu         = ipip6_tunnel_change_mtu;
 727
 728         dev->type               = ARPHRD_SIT;
 729         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 730         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 731         dev->flags              = IFF_NOARP;
 732         dev->iflink             = 0;
 733         dev->addr_len           = 4;
 734 }
 735
 736 static int ipip6_tunnel_init(struct net_device *dev)
 737 {
 738         struct net_device *tdev = NULL;
 739         struct ip_tunnel *tunnel;
 740         struct iphdr *iph;
 741
 742         tunnel = netdev_priv(dev);
 743         iph = &tunnel->parms.iph;
 744
 745         tunnel->dev = dev;
 746         strcpy(tunnel->parms.name, dev->name);
 747
 748         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 749         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 750
 751         if (iph->daddr) {
 752                 struct flowi fl = { .nl_u = { .ip4_u =
 753                                               { .daddr = iph->daddr,
 754                                                 .saddr = iph->saddr,
 755                                                 .tos = RT_TOS(iph->tos) } },
 756                                     .oif = tunnel->parms.link,
 757                                     .proto = IPPROTO_IPV6 };
 758                 struct rtable *rt;
 759                 if (!ip_route_output_key(&rt, &fl)) {
 760                         tdev = rt->u.dst.dev;
 761                         ip_rt_put(rt);
 762                 }
 763                 dev->flags |= IFF_POINTOPOINT;
 764         }
 765
 766         if (!tdev && tunnel->parms.link)
 767                 tdev = __dev_get_by_index(tunnel->parms.link);
 768
 769         if (tdev) {
 770                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 771                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 772                 if (dev->mtu < IPV6_MIN_MTU)
 773                         dev->mtu = IPV6_MIN_MTU;
 774         }
 775         dev->iflink = tunnel->parms.link;
 776
 777         return 0;
 778 }
 779
 780 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 781 {
 782         struct ip_tunnel *tunnel = netdev_priv(dev);
 783         struct iphdr *iph = &tunnel->parms.iph;
 784
 785         tunnel->dev = dev;
 786         strcpy(tunnel->parms.name, dev->name);
 787
 788         iph->version            = 4;
 789         iph->protocol           = IPPROTO_IPV6;
 790         iph->ihl                = 5;
 791         iph->ttl                = 64;
 792
 793         dev_hold(dev);
 794         tunnels_wc[0]           = tunnel;
 795         return 0;
 796 }
 797
 798 static struct xfrm_tunnel sit_handler = {
 799         .handler        =       ipip6_rcv,
 800         .err_handler    =       ipip6_err,
 801         .priority       =       1,
 802 };
 803
 804 static void __exit sit_destroy_tunnels(void)
 805 {
 806         int prio;
 807
 808         for (prio = 1; prio < 4; prio++) {
 809                 int h;
 810                 for (h = 0; h < HASH_SIZE; h++) {
 811                         struct ip_tunnel *t;
 812                         while ((t = tunnels[prio][h]) != NULL)
 813                                 unregister_netdevice(t->dev);
 814                 }
 815         }
 816 }
 817
 818 static void __exit sit_cleanup(void)
 819 {
 820         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 821
 822         rtnl_lock();
 823         sit_destroy_tunnels();
 824         unregister_netdevice(ipip6_fb_tunnel_dev);
 825         rtnl_unlock();
 826 }
 827
 828 static int __init sit_init(void)
 829 {
 830         int err;
 831
 832         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 833
 834         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
 835                 printk(KERN_INFO "sit init: Can't add protocol\n");
 836                 return -EAGAIN;
 837         }
 838
 839         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
 840                                            ipip6_tunnel_setup);
 841         if (!ipip6_fb_tunnel_dev) {
 842                 err = -ENOMEM;
 843                 goto err1;
 844         }
 845
 846         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
 847
 848         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
 849                 goto err2;
 850
 851  out:
 852         return err;
 853  err2:
 854         free_netdev(ipip6_fb_tunnel_dev);
 855  err1:
 856         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 857         goto out;
 858 }
 859
 860 module_init(sit_init);
 861 module_exit(sit_cleanup);
 862 MODULE_LICENSE("GPL");
 863 MODULE_ALIAS("sit0");