SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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  * Fred L. Templin <fltemplin@acm.org>:         isatap support
  20  */
  21
  22 #include <linux/module.h>
  23 #include <linux/capability.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/socket.h>
  27 #include <linux/sockios.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_parm *parms)
 104 {
 105         __be32 remote = parms->iph.daddr;
 106         __be32 local = 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 inline struct ip_tunnel **ipip6_bucket(struct ip_tunnel *t)
 122 {
 123         return __ipip6_bucket(&t->parms);
 124 }
 125
 126 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 127 {
 128         struct ip_tunnel **tp;
 129
 130         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 131                 if (t == *tp) {
 132                         write_lock_bh(&ipip6_lock);
 133                         *tp = t->next;
 134                         write_unlock_bh(&ipip6_lock);
 135                         break;
 136                 }
 137         }
 138 }
 139
 140 static void ipip6_tunnel_link(struct ip_tunnel *t)
 141 {
 142         struct ip_tunnel **tp = ipip6_bucket(t);
 143
 144         t->next = *tp;
 145         write_lock_bh(&ipip6_lock);
 146         *tp = t;
 147         write_unlock_bh(&ipip6_lock);
 148 }
 149
 150 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 151 {
 152         __be32 remote = parms->iph.daddr;
 153         __be32 local = parms->iph.saddr;
 154         struct ip_tunnel *t, **tp, *nt;
 155         struct net_device *dev;
 156         char name[IFNAMSIZ];
 157
 158         for (tp = __ipip6_bucket(parms); (t = *tp) != NULL; tp = &t->next) {
 159                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 160                         return t;
 161         }
 162         if (!create)
 163                 goto failed;
 164
 165         if (parms->name[0])
 166                 strlcpy(name, parms->name, IFNAMSIZ);
 167         else
 168                 sprintf(name, "sit%%d");
 169
 170         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 171         if (dev == NULL)
 172                 return NULL;
 173
 174         nt = netdev_priv(dev);
 175         dev->init = ipip6_tunnel_init;
 176         nt->parms = *parms;
 177
 178         if (parms->i_flags & SIT_ISATAP)
 179                 dev->priv_flags |= IFF_ISATAP;
 180
 181         if (register_netdevice(dev) < 0) {
 182                 free_netdev(dev);
 183                 goto failed;
 184         }
 185
 186         dev_hold(dev);
 187
 188         ipip6_tunnel_link(nt);
 189         return nt;
 190
 191 failed:
 192         return NULL;
 193 }
 194
 195 static void ipip6_tunnel_uninit(struct net_device *dev)
 196 {
 197         if (dev == ipip6_fb_tunnel_dev) {
 198                 write_lock_bh(&ipip6_lock);
 199                 tunnels_wc[0] = NULL;
 200                 write_unlock_bh(&ipip6_lock);
 201                 dev_put(dev);
 202         } else {
 203                 ipip6_tunnel_unlink(netdev_priv(dev));
 204                 dev_put(dev);
 205         }
 206 }
 207
 208
 209 static int ipip6_err(struct sk_buff *skb, u32 info)
 210 {
 211 #ifndef I_WISH_WORLD_WERE_PERFECT
 212
 213 /* It is not :-( All the routers (except for Linux) return only
 214    8 bytes of packet payload. It means, that precise relaying of
 215    ICMP in the real Internet is absolutely infeasible.
 216  */
 217         struct iphdr *iph = (struct iphdr*)skb->data;
 218         const int type = icmp_hdr(skb)->type;
 219         const int code = icmp_hdr(skb)->code;
 220         struct ip_tunnel *t;
 221         int err;
 222
 223         switch (type) {
 224         default:
 225         case ICMP_PARAMETERPROB:
 226                 return 0;
 227
 228         case ICMP_DEST_UNREACH:
 229                 switch (code) {
 230                 case ICMP_SR_FAILED:
 231                 case ICMP_PORT_UNREACH:
 232                         /* Impossible event. */
 233                         return 0;
 234                 case ICMP_FRAG_NEEDED:
 235                         /* Soft state for pmtu is maintained by IP core. */
 236                         return 0;
 237                 default:
 238                         /* All others are translated to HOST_UNREACH.
 239                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 240                            I believe they are just ether pollution. --ANK
 241                          */
 242                         break;
 243                 }
 244                 break;
 245         case ICMP_TIME_EXCEEDED:
 246                 if (code != ICMP_EXC_TTL)
 247                         return 0;
 248                 break;
 249         }
 250
 251         err = -ENOENT;
 252
 253         read_lock(&ipip6_lock);
 254         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 255         if (t == NULL || t->parms.iph.daddr == 0)
 256                 goto out;
 257
 258         err = 0;
 259         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 260                 goto out;
 261
 262         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 263                 t->err_count++;
 264         else
 265                 t->err_count = 1;
 266         t->err_time = jiffies;
 267 out:
 268         read_unlock(&ipip6_lock);
 269         return err;
 270 #else
 271         struct iphdr *iph = (struct iphdr*)dp;
 272         int hlen = iph->ihl<<2;
 273         struct ipv6hdr *iph6;
 274         const int type = icmp_hdr(skb)->type;
 275         const int code = icmp_hdr(skb)->code;
 276         int rel_type = 0;
 277         int rel_code = 0;
 278         int rel_info = 0;
 279         struct sk_buff *skb2;
 280         struct rt6_info *rt6i;
 281
 282         if (len < hlen + sizeof(struct ipv6hdr))
 283                 return;
 284         iph6 = (struct ipv6hdr*)(dp + hlen);
 285
 286         switch (type) {
 287         default:
 288                 return;
 289         case ICMP_PARAMETERPROB:
 290                 if (icmp_hdr(skb)->un.gateway < hlen)
 291                         return;
 292
 293                 /* So... This guy found something strange INSIDE encapsulated
 294                    packet. Well, he is fool, but what can we do ?
 295                  */
 296                 rel_type = ICMPV6_PARAMPROB;
 297                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 298                 break;
 299
 300         case ICMP_DEST_UNREACH:
 301                 switch (code) {
 302                 case ICMP_SR_FAILED:
 303                 case ICMP_PORT_UNREACH:
 304                         /* Impossible event. */
 305                         return;
 306                 case ICMP_FRAG_NEEDED:
 307                         /* Too complicated case ... */
 308                         return;
 309                 default:
 310                         /* All others are translated to HOST_UNREACH.
 311                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 312                            I believe, it is just ether pollution. --ANK
 313                          */
 314                         rel_type = ICMPV6_DEST_UNREACH;
 315                         rel_code = ICMPV6_ADDR_UNREACH;
 316                         break;
 317                 }
 318                 break;
 319         case ICMP_TIME_EXCEEDED:
 320                 if (code != ICMP_EXC_TTL)
 321                         return;
 322                 rel_type = ICMPV6_TIME_EXCEED;
 323                 rel_code = ICMPV6_EXC_HOPLIMIT;
 324                 break;
 325         }
 326
 327         /* Prepare fake skb to feed it to icmpv6_send */
 328         skb2 = skb_clone(skb, GFP_ATOMIC);
 329         if (skb2 == NULL)
 330                 return 0;
 331         dst_release(skb2->dst);
 332         skb2->dst = NULL;
 333         skb_pull(skb2, skb->data - (u8*)iph6);
 334         skb_reset_network_header(skb2);
 335
 336         /* Try to guess incoming interface */
 337         rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
 338         if (rt6i && rt6i->rt6i_dev) {
 339                 skb2->dev = rt6i->rt6i_dev;
 340
 341                 rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
 342
 343                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 344                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 345                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 346                                 rel_type = ICMPV6_DEST_UNREACH;
 347                                 rel_code = ICMPV6_ADDR_UNREACH;
 348                         }
 349                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 350                 }
 351         }
 352         kfree_skb(skb2);
 353         return 0;
 354 #endif
 355 }
 356
 357 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 358 {
 359         if (INET_ECN_is_ce(iph->tos))
 360                 IP6_ECN_set_ce(ipv6_hdr(skb));
 361 }
 362
 363 /* ISATAP (RFC4214) - check source address */
 364 static int
 365 isatap_srcok(struct sk_buff *skb, struct iphdr *iph, struct net_device *dev)
 366 {
 367         struct neighbour *neigh;
 368         struct dst_entry *dst;
 369         struct rt6_info *rt;
 370         struct flowi fl;
 371         struct in6_addr *addr6;
 372         struct in6_addr rtr;
 373         struct ipv6hdr *iph6;
 374         int ok = 0;
 375
 376         /* from onlink default router */
 377         ipv6_addr_set(&rtr,  htonl(0xFE800000), 0, 0, 0);
 378         ipv6_isatap_eui64(rtr.s6_addr + 8, iph->saddr);
 379         if ((rt = rt6_get_dflt_router(&rtr, dev))) {
 380                 dst_release(&rt->u.dst);
 381                 return 1;
 382         }
 383
 384         iph6 = ipv6_hdr(skb);
 385         memset(&fl, 0, sizeof(fl));
 386         fl.proto = iph6->nexthdr;
 387         ipv6_addr_copy(&fl.fl6_dst, &iph6->saddr);
 388         fl.oif = dev->ifindex;
 389         security_skb_classify_flow(skb, &fl);
 390
 391         dst = ip6_route_output(NULL, &fl);
 392         if (!dst->error && (dst->dev == dev) && (neigh = dst->neighbour)) {
 393
 394                 addr6 = (struct in6_addr*)&neigh->primary_key;
 395
 396                 /* from correct previous hop */
 397                 if (ipv6_addr_is_isatap(addr6) &&
 398                     (addr6->s6_addr32[3] == iph->saddr))
 399                         ok = 1;
 400         }
 401         dst_release(dst);
 402         return ok;
 403 }
 404
 405 static int ipip6_rcv(struct sk_buff *skb)
 406 {
 407         struct iphdr *iph;
 408         struct ip_tunnel *tunnel;
 409
 410         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 411                 goto out;
 412
 413         iph = ip_hdr(skb);
 414
 415         read_lock(&ipip6_lock);
 416         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 417                 secpath_reset(skb);
 418                 skb->mac_header = skb->network_header;
 419                 skb_reset_network_header(skb);
 420                 IPCB(skb)->flags = 0;
 421                 skb->protocol = htons(ETH_P_IPV6);
 422                 skb->pkt_type = PACKET_HOST;
 423
 424                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 425                     !isatap_srcok(skb, iph, tunnel->dev)) {
 426                         tunnel->stat.rx_errors++;
 427                         read_unlock(&ipip6_lock);
 428                         kfree_skb(skb);
 429                         return 0;
 430                 }
 431                 tunnel->stat.rx_packets++;
 432                 tunnel->stat.rx_bytes += skb->len;
 433                 skb->dev = tunnel->dev;
 434                 dst_release(skb->dst);
 435                 skb->dst = NULL;
 436                 nf_reset(skb);
 437                 ipip6_ecn_decapsulate(iph, skb);
 438                 netif_rx(skb);
 439                 read_unlock(&ipip6_lock);
 440                 return 0;
 441         }
 442
 443         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 444         kfree_skb(skb);
 445         read_unlock(&ipip6_lock);
 446 out:
 447         return 0;
 448 }
 449
 450 /* Returns the embedded IPv4 address if the IPv6 address
 451    comes from 6to4 (RFC 3056) addr space */
 452
 453 static inline __be32 try_6to4(struct in6_addr *v6dst)
 454 {
 455         __be32 dst = 0;
 456
 457         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 458                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 459                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 460         }
 461         return dst;
 462 }
 463
 464 /*
 465  *      This function assumes it is being called from dev_queue_xmit()
 466  *      and that skb is filled properly by that function.
 467  */
 468
 469 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 470 {
 471         struct ip_tunnel *tunnel = netdev_priv(dev);
 472         struct net_device_stats *stats = &tunnel->stat;
 473         struct iphdr  *tiph = &tunnel->parms.iph;
 474         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 475         u8     tos = tunnel->parms.iph.tos;
 476         struct rtable *rt;                      /* Route to the other host */
 477         struct net_device *tdev;                        /* Device to other host */
 478         struct iphdr  *iph;                     /* Our new IP header */
 479         unsigned int max_headroom;              /* The extra header space needed */
 480         __be32 dst = tiph->daddr;
 481         int    mtu;
 482         struct in6_addr *addr6;
 483         int addr_type;
 484
 485         if (tunnel->recursion++) {
 486                 tunnel->stat.collisions++;
 487                 goto tx_error;
 488         }
 489
 490         if (skb->protocol != htons(ETH_P_IPV6))
 491                 goto tx_error;
 492
 493         /* ISATAP (RFC4214) - must come before 6to4 */
 494         if (dev->priv_flags & IFF_ISATAP) {
 495                 struct neighbour *neigh = NULL;
 496
 497                 if (skb->dst)
 498                         neigh = skb->dst->neighbour;
 499
 500                 if (neigh == NULL) {
 501                         if (net_ratelimit())
 502                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 503                         goto tx_error;
 504                 }
 505
 506                 addr6 = (struct in6_addr*)&neigh->primary_key;
 507                 addr_type = ipv6_addr_type(addr6);
 508
 509                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 510                      ipv6_addr_is_isatap(addr6))
 511                         dst = addr6->s6_addr32[3];
 512                 else
 513                         goto tx_error;
 514         }
 515
 516         if (!dst)
 517                 dst = try_6to4(&iph6->daddr);
 518
 519         if (!dst) {
 520                 struct neighbour *neigh = NULL;
 521
 522                 if (skb->dst)
 523                         neigh = skb->dst->neighbour;
 524
 525                 if (neigh == NULL) {
 526                         if (net_ratelimit())
 527                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 528                         goto tx_error;
 529                 }
 530
 531                 addr6 = (struct in6_addr*)&neigh->primary_key;
 532                 addr_type = ipv6_addr_type(addr6);
 533
 534                 if (addr_type == IPV6_ADDR_ANY) {
 535                         addr6 = &ipv6_hdr(skb)->daddr;
 536                         addr_type = ipv6_addr_type(addr6);
 537                 }
 538
 539                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 540                         goto tx_error_icmp;
 541
 542                 dst = addr6->s6_addr32[3];
 543         }
 544
 545         {
 546                 struct flowi fl = { .nl_u = { .ip4_u =
 547                                               { .daddr = dst,
 548                                                 .saddr = tiph->saddr,
 549                                                 .tos = RT_TOS(tos) } },
 550                                     .oif = tunnel->parms.link,
 551                                     .proto = IPPROTO_IPV6 };
 552                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 553                         tunnel->stat.tx_carrier_errors++;
 554                         goto tx_error_icmp;
 555                 }
 556         }
 557         if (rt->rt_type != RTN_UNICAST) {
 558                 ip_rt_put(rt);
 559                 tunnel->stat.tx_carrier_errors++;
 560                 goto tx_error_icmp;
 561         }
 562         tdev = rt->u.dst.dev;
 563
 564         if (tdev == dev) {
 565                 ip_rt_put(rt);
 566                 tunnel->stat.collisions++;
 567                 goto tx_error;
 568         }
 569
 570         if (tiph->frag_off)
 571                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 572         else
 573                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 574
 575         if (mtu < 68) {
 576                 tunnel->stat.collisions++;
 577                 ip_rt_put(rt);
 578                 goto tx_error;
 579         }
 580         if (mtu < IPV6_MIN_MTU)
 581                 mtu = IPV6_MIN_MTU;
 582         if (tunnel->parms.iph.daddr && skb->dst)
 583                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 584
 585         if (skb->len > mtu) {
 586                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 587                 ip_rt_put(rt);
 588                 goto tx_error;
 589         }
 590
 591         if (tunnel->err_count > 0) {
 592                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 593                         tunnel->err_count--;
 594                         dst_link_failure(skb);
 595                 } else
 596                         tunnel->err_count = 0;
 597         }
 598
 599         /*
 600          * Okay, now see if we can stuff it in the buffer as-is.
 601          */
 602         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 603
 604         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 605             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 606                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 607                 if (!new_skb) {
 608                         ip_rt_put(rt);
 609                         stats->tx_dropped++;
 610                         dev_kfree_skb(skb);
 611                         tunnel->recursion--;
 612                         return 0;
 613                 }
 614                 if (skb->sk)
 615                         skb_set_owner_w(new_skb, skb->sk);
 616                 dev_kfree_skb(skb);
 617                 skb = new_skb;
 618                 iph6 = ipv6_hdr(skb);
 619         }
 620
 621         skb->transport_header = skb->network_header;
 622         skb_push(skb, sizeof(struct iphdr));
 623         skb_reset_network_header(skb);
 624         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 625         IPCB(skb)->flags = 0;
 626         dst_release(skb->dst);
 627         skb->dst = &rt->u.dst;
 628
 629         /*
 630          *      Push down and install the IPIP header.
 631          */
 632
 633         iph                     =       ip_hdr(skb);
 634         iph->version            =       4;
 635         iph->ihl                =       sizeof(struct iphdr)>>2;
 636         if (mtu > IPV6_MIN_MTU)
 637                 iph->frag_off   =       htons(IP_DF);
 638         else
 639                 iph->frag_off   =       0;
 640
 641         iph->protocol           =       IPPROTO_IPV6;
 642         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 643         iph->daddr              =       rt->rt_dst;
 644         iph->saddr              =       rt->rt_src;
 645
 646         if ((iph->ttl = tiph->ttl) == 0)
 647                 iph->ttl        =       iph6->hop_limit;
 648
 649         nf_reset(skb);
 650
 651         IPTUNNEL_XMIT();
 652         tunnel->recursion--;
 653         return 0;
 654
 655 tx_error_icmp:
 656         dst_link_failure(skb);
 657 tx_error:
 658         stats->tx_errors++;
 659         dev_kfree_skb(skb);
 660         tunnel->recursion--;
 661         return 0;
 662 }
 663
 664 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 665 {
 666         struct net_device *tdev = NULL;
 667         struct ip_tunnel *tunnel;
 668         struct iphdr *iph;
 669
 670         tunnel = netdev_priv(dev);
 671         iph = &tunnel->parms.iph;
 672
 673         if (iph->daddr) {
 674                 struct flowi fl = { .nl_u = { .ip4_u =
 675                                               { .daddr = iph->daddr,
 676                                                 .saddr = iph->saddr,
 677                                                 .tos = RT_TOS(iph->tos) } },
 678                                     .oif = tunnel->parms.link,
 679                                     .proto = IPPROTO_IPV6 };
 680                 struct rtable *rt;
 681                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 682                         tdev = rt->u.dst.dev;
 683                         ip_rt_put(rt);
 684                 }
 685                 dev->flags |= IFF_POINTOPOINT;
 686         }
 687
 688         if (!tdev && tunnel->parms.link)
 689                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 690
 691         if (tdev) {
 692                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 693                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 694                 if (dev->mtu < IPV6_MIN_MTU)
 695                         dev->mtu = IPV6_MIN_MTU;
 696         }
 697         dev->iflink = tunnel->parms.link;
 698 }
 699
 700 static int
 701 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 702 {
 703         int err = 0;
 704         struct ip_tunnel_parm p;
 705         struct ip_tunnel *t;
 706
 707         switch (cmd) {
 708         case SIOCGETTUNNEL:
 709                 t = NULL;
 710                 if (dev == ipip6_fb_tunnel_dev) {
 711                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 712                                 err = -EFAULT;
 713                                 break;
 714                         }
 715                         t = ipip6_tunnel_locate(&p, 0);
 716                 }
 717                 if (t == NULL)
 718                         t = netdev_priv(dev);
 719                 memcpy(&p, &t->parms, sizeof(p));
 720                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 721                         err = -EFAULT;
 722                 break;
 723
 724         case SIOCADDTUNNEL:
 725         case SIOCCHGTUNNEL:
 726                 err = -EPERM;
 727                 if (!capable(CAP_NET_ADMIN))
 728                         goto done;
 729
 730                 err = -EFAULT;
 731                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 732                         goto done;
 733
 734                 err = -EINVAL;
 735                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 736                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 737                         goto done;
 738                 if (p.iph.ttl)
 739                         p.iph.frag_off |= htons(IP_DF);
 740
 741                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 742
 743                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 744                         if (t != NULL) {
 745                                 if (t->dev != dev) {
 746                                         err = -EEXIST;
 747                                         break;
 748                                 }
 749                         } else {
 750                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 751                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 752                                         err = -EINVAL;
 753                                         break;
 754                                 }
 755                                 t = netdev_priv(dev);
 756                                 ipip6_tunnel_unlink(t);
 757                                 t->parms.iph.saddr = p.iph.saddr;
 758                                 t->parms.iph.daddr = p.iph.daddr;
 759                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 760                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 761                                 ipip6_tunnel_link(t);
 762                                 netdev_state_change(dev);
 763                         }
 764                 }
 765
 766                 if (t) {
 767                         err = 0;
 768                         if (cmd == SIOCCHGTUNNEL) {
 769                                 t->parms.iph.ttl = p.iph.ttl;
 770                                 t->parms.iph.tos = p.iph.tos;
 771                                 if (t->parms.link != p.link) {
 772                                         t->parms.link = p.link;
 773                                         ipip6_tunnel_bind_dev(dev);
 774                                         netdev_state_change(dev);
 775                                 }
 776                         }
 777                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 778                                 err = -EFAULT;
 779                 } else
 780                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 781                 break;
 782
 783         case SIOCDELTUNNEL:
 784                 err = -EPERM;
 785                 if (!capable(CAP_NET_ADMIN))
 786                         goto done;
 787
 788                 if (dev == ipip6_fb_tunnel_dev) {
 789                         err = -EFAULT;
 790                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 791                                 goto done;
 792                         err = -ENOENT;
 793                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 794                                 goto done;
 795                         err = -EPERM;
 796                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 797                                 goto done;
 798                         dev = t->dev;
 799                 }
 800                 unregister_netdevice(dev);
 801                 err = 0;
 802                 break;
 803
 804         default:
 805                 err = -EINVAL;
 806         }
 807
 808 done:
 809         return err;
 810 }
 811
 812 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 813 {
 814         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 815 }
 816
 817 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 818 {
 819         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 820                 return -EINVAL;
 821         dev->mtu = new_mtu;
 822         return 0;
 823 }
 824
 825 static void ipip6_tunnel_setup(struct net_device *dev)
 826 {
 827         dev->uninit             = ipip6_tunnel_uninit;
 828         dev->destructor         = free_netdev;
 829         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 830         dev->get_stats          = ipip6_tunnel_get_stats;
 831         dev->do_ioctl           = ipip6_tunnel_ioctl;
 832         dev->change_mtu         = ipip6_tunnel_change_mtu;
 833
 834         dev->type               = ARPHRD_SIT;
 835         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 836         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 837         dev->flags              = IFF_NOARP;
 838         dev->iflink             = 0;
 839         dev->addr_len           = 4;
 840 }
 841
 842 static int ipip6_tunnel_init(struct net_device *dev)
 843 {
 844         struct ip_tunnel *tunnel;
 845
 846         tunnel = netdev_priv(dev);
 847
 848         tunnel->dev = dev;
 849         strcpy(tunnel->parms.name, dev->name);
 850
 851         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 852         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 853
 854         ipip6_tunnel_bind_dev(dev);
 855
 856         return 0;
 857 }
 858
 859 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 860 {
 861         struct ip_tunnel *tunnel = netdev_priv(dev);
 862         struct iphdr *iph = &tunnel->parms.iph;
 863
 864         tunnel->dev = dev;
 865         strcpy(tunnel->parms.name, dev->name);
 866
 867         iph->version            = 4;
 868         iph->protocol           = IPPROTO_IPV6;
 869         iph->ihl                = 5;
 870         iph->ttl                = 64;
 871
 872         dev_hold(dev);
 873         tunnels_wc[0]           = tunnel;
 874         return 0;
 875 }
 876
 877 static struct xfrm_tunnel sit_handler = {
 878         .handler        =       ipip6_rcv,
 879         .err_handler    =       ipip6_err,
 880         .priority       =       1,
 881 };
 882
 883 static void __exit sit_destroy_tunnels(void)
 884 {
 885         int prio;
 886
 887         for (prio = 1; prio < 4; prio++) {
 888                 int h;
 889                 for (h = 0; h < HASH_SIZE; h++) {
 890                         struct ip_tunnel *t;
 891                         while ((t = tunnels[prio][h]) != NULL)
 892                                 unregister_netdevice(t->dev);
 893                 }
 894         }
 895 }
 896
 897 static void __exit sit_cleanup(void)
 898 {
 899         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 900
 901         rtnl_lock();
 902         sit_destroy_tunnels();
 903         unregister_netdevice(ipip6_fb_tunnel_dev);
 904         rtnl_unlock();
 905 }
 906
 907 static int __init sit_init(void)
 908 {
 909         int err;
 910
 911         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 912
 913         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
 914                 printk(KERN_INFO "sit init: Can't add protocol\n");
 915                 return -EAGAIN;
 916         }
 917
 918         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
 919                                            ipip6_tunnel_setup);
 920         if (!ipip6_fb_tunnel_dev) {
 921                 err = -ENOMEM;
 922                 goto err1;
 923         }
 924
 925         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
 926
 927         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
 928                 goto err2;
 929
 930  out:
 931         return err;
 932  err2:
 933         free_netdev(ipip6_fb_tunnel_dev);
 934  err1:
 935         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 936         goto out;
 937 }
 938
 939 module_init(sit_init);
 940 module_exit(sit_cleanup);
 941 MODULE_LICENSE("GPL");
 942 MODULE_ALIAS("sit0");