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 Templin <fred.l.templin@boeing.com>:    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 #include <net/net_namespace.h>
  56 #include <net/netns/generic.h>
  57
  58 /*
  59    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  60
  61    For comments look at net/ipv4/ip_gre.c --ANK
  62  */
  63
  64 #define HASH_SIZE  16
  65 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  66
  67 static int ipip6_fb_tunnel_init(struct net_device *dev);
  68 static int ipip6_tunnel_init(struct net_device *dev);
  69 static void ipip6_tunnel_setup(struct net_device *dev);
  70
  71 static int sit_net_id;
  72 struct sit_net {
  73         struct net_device *fb_tunnel_dev;
  74 };
  75
  76 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  77 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  78 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  79 static struct ip_tunnel *tunnels_wc[1];
  80 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  81
  82 static DEFINE_RWLOCK(ipip6_lock);
  83
  84 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
  85                 __be32 remote, __be32 local)
  86 {
  87         unsigned h0 = HASH(remote);
  88         unsigned h1 = HASH(local);
  89         struct ip_tunnel *t;
  90
  91         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  92                 if (local == t->parms.iph.saddr &&
  93                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  94                         return t;
  95         }
  96         for (t = tunnels_r[h0]; t; t = t->next) {
  97                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  98                         return t;
  99         }
 100         for (t = tunnels_l[h1]; t; t = t->next) {
 101                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
 102                         return t;
 103         }
 104         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
 105                 return t;
 106         return NULL;
 107 }
 108
 109 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
 110                 struct ip_tunnel_parm *parms)
 111 {
 112         __be32 remote = parms->iph.daddr;
 113         __be32 local = parms->iph.saddr;
 114         unsigned h = 0;
 115         int prio = 0;
 116
 117         if (remote) {
 118                 prio |= 2;
 119                 h ^= HASH(remote);
 120         }
 121         if (local) {
 122                 prio |= 1;
 123                 h ^= HASH(local);
 124         }
 125         return &tunnels[prio][h];
 126 }
 127
 128 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
 129                 struct ip_tunnel *t)
 130 {
 131         return __ipip6_bucket(sitn, &t->parms);
 132 }
 133
 134 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 135 {
 136         struct ip_tunnel **tp;
 137
 138         for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
 139                 if (t == *tp) {
 140                         write_lock_bh(&ipip6_lock);
 141                         *tp = t->next;
 142                         write_unlock_bh(&ipip6_lock);
 143                         break;
 144                 }
 145         }
 146 }
 147
 148 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 149 {
 150         struct ip_tunnel **tp = ipip6_bucket(sitn, t);
 151
 152         t->next = *tp;
 153         write_lock_bh(&ipip6_lock);
 154         *tp = t;
 155         write_unlock_bh(&ipip6_lock);
 156 }
 157
 158 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
 159                 struct ip_tunnel_parm *parms, int create)
 160 {
 161         __be32 remote = parms->iph.daddr;
 162         __be32 local = parms->iph.saddr;
 163         struct ip_tunnel *t, **tp, *nt;
 164         struct net_device *dev;
 165         char name[IFNAMSIZ];
 166         struct sit_net *sitn = net_generic(net, sit_net_id);
 167
 168         for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
 169                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 170                         return t;
 171         }
 172         if (!create)
 173                 goto failed;
 174
 175         if (parms->name[0])
 176                 strlcpy(name, parms->name, IFNAMSIZ);
 177         else
 178                 sprintf(name, "sit%%d");
 179
 180         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 181         if (dev == NULL)
 182                 return NULL;
 183
 184         if (strchr(name, '%')) {
 185                 if (dev_alloc_name(dev, name) < 0)
 186                         goto failed_free;
 187         }
 188
 189         nt = netdev_priv(dev);
 190         dev->init = ipip6_tunnel_init;
 191         nt->parms = *parms;
 192
 193         if (parms->i_flags & SIT_ISATAP)
 194                 dev->priv_flags |= IFF_ISATAP;
 195
 196         if (register_netdevice(dev) < 0)
 197                 goto failed_free;
 198
 199         dev_hold(dev);
 200
 201         ipip6_tunnel_link(sitn, nt);
 202         return nt;
 203
 204 failed_free:
 205         free_netdev(dev);
 206 failed:
 207         return NULL;
 208 }
 209
 210 static struct ip_tunnel_prl_entry *
 211 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 212 {
 213         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 214
 215         for (p = t->prl; p; p = p->next)
 216                 if (p->addr == addr)
 217                         break;
 218         return p;
 219
 220 }
 221
 222 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 223 {
 224         struct ip_tunnel_prl *kp;
 225         struct ip_tunnel_prl_entry *prl;
 226         unsigned int cmax, c = 0, ca, len;
 227         int ret = 0;
 228
 229         cmax = a->datalen / sizeof(*a);
 230         if (cmax > 1 && a->addr != htonl(INADDR_ANY))
 231                 cmax = 1;
 232
 233         /* For simple GET or for root users,
 234          * we try harder to allocate.
 235          */
 236         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 237                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 238                 NULL;
 239
 240         read_lock(&ipip6_lock);
 241
 242         ca = t->prl_count < cmax ? t->prl_count : cmax;
 243
 244         if (!kp) {
 245                 /* We don't try hard to allocate much memory for
 246                  * non-root users.
 247                  * For root users, retry allocating enough memory for
 248                  * the answer.
 249                  */
 250                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 251                 if (!kp) {
 252                         ret = -ENOMEM;
 253                         goto out;
 254                 }
 255         }
 256
 257         c = 0;
 258         for (prl = t->prl; prl; prl = prl->next) {
 259                 if (c > cmax)
 260                         break;
 261                 if (a->addr != htonl(INADDR_ANY) && prl->addr != a->addr)
 262                         continue;
 263                 kp[c].addr = prl->addr;
 264                 kp[c].flags = prl->flags;
 265                 c++;
 266                 if (a->addr != htonl(INADDR_ANY))
 267                         break;
 268         }
 269 out:
 270         read_unlock(&ipip6_lock);
 271
 272         len = sizeof(*kp) * c;
 273         ret = len ? copy_to_user(a->data, kp, len) : 0;
 274
 275         kfree(kp);
 276         if (ret)
 277                 return -EFAULT;
 278
 279         a->datalen = len;
 280         return 0;
 281 }
 282
 283 static int
 284 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 285 {
 286         struct ip_tunnel_prl_entry *p;
 287         int err = 0;
 288
 289         if (a->addr == htonl(INADDR_ANY))
 290                 return -EINVAL;
 291
 292         write_lock(&ipip6_lock);
 293
 294         for (p = t->prl; p; p = p->next) {
 295                 if (p->addr == a->addr) {
 296                         if (chg)
 297                                 goto update;
 298                         err = -EEXIST;
 299                         goto out;
 300                 }
 301         }
 302
 303         if (chg) {
 304                 err = -ENXIO;
 305                 goto out;
 306         }
 307
 308         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 309         if (!p) {
 310                 err = -ENOBUFS;
 311                 goto out;
 312         }
 313
 314         p->next = t->prl;
 315         t->prl = p;
 316         t->prl_count++;
 317 update:
 318         p->addr = a->addr;
 319         p->flags = a->flags;
 320 out:
 321         write_unlock(&ipip6_lock);
 322         return err;
 323 }
 324
 325 static int
 326 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 327 {
 328         struct ip_tunnel_prl_entry *x, **p;
 329         int err = 0;
 330
 331         write_lock(&ipip6_lock);
 332
 333         if (a && a->addr != htonl(INADDR_ANY)) {
 334                 for (p = &t->prl; *p; p = &(*p)->next) {
 335                         if ((*p)->addr == a->addr) {
 336                                 x = *p;
 337                                 *p = x->next;
 338                                 kfree(x);
 339                                 t->prl_count--;
 340                                 goto out;
 341                         }
 342                 }
 343                 err = -ENXIO;
 344         } else {
 345                 while (t->prl) {
 346                         x = t->prl;
 347                         t->prl = t->prl->next;
 348                         kfree(x);
 349                         t->prl_count--;
 350                 }
 351         }
 352 out:
 353         write_unlock(&ipip6_lock);
 354         return 0;
 355 }
 356
 357 static int
 358 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 359 {
 360         struct ip_tunnel_prl_entry *p;
 361         int ok = 1;
 362
 363         read_lock(&ipip6_lock);
 364         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 365         if (p) {
 366                 if (p->flags & PRL_DEFAULT)
 367                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 368                 else
 369                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 370         } else {
 371                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 372                 if (ipv6_addr_is_isatap(addr6) &&
 373                     (addr6->s6_addr32[3] == iph->saddr) &&
 374                     ipv6_chk_prefix(addr6, t->dev))
 375                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 376                 else
 377                         ok = 0;
 378         }
 379         read_unlock(&ipip6_lock);
 380         return ok;
 381 }
 382
 383 static void ipip6_tunnel_uninit(struct net_device *dev)
 384 {
 385         struct net *net = dev_net(dev);
 386         struct sit_net *sitn = net_generic(net, sit_net_id);
 387
 388         if (dev == sitn->fb_tunnel_dev) {
 389                 write_lock_bh(&ipip6_lock);
 390                 tunnels_wc[0] = NULL;
 391                 write_unlock_bh(&ipip6_lock);
 392                 dev_put(dev);
 393         } else {
 394                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 395                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 396                 dev_put(dev);
 397         }
 398 }
 399
 400
 401 static int ipip6_err(struct sk_buff *skb, u32 info)
 402 {
 403 #ifndef I_WISH_WORLD_WERE_PERFECT
 404
 405 /* It is not :-( All the routers (except for Linux) return only
 406    8 bytes of packet payload. It means, that precise relaying of
 407    ICMP in the real Internet is absolutely infeasible.
 408  */
 409         struct iphdr *iph = (struct iphdr*)skb->data;
 410         const int type = icmp_hdr(skb)->type;
 411         const int code = icmp_hdr(skb)->code;
 412         struct ip_tunnel *t;
 413         int err;
 414
 415         switch (type) {
 416         default:
 417         case ICMP_PARAMETERPROB:
 418                 return 0;
 419
 420         case ICMP_DEST_UNREACH:
 421                 switch (code) {
 422                 case ICMP_SR_FAILED:
 423                 case ICMP_PORT_UNREACH:
 424                         /* Impossible event. */
 425                         return 0;
 426                 case ICMP_FRAG_NEEDED:
 427                         /* Soft state for pmtu is maintained by IP core. */
 428                         return 0;
 429                 default:
 430                         /* All others are translated to HOST_UNREACH.
 431                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 432                            I believe they are just ether pollution. --ANK
 433                          */
 434                         break;
 435                 }
 436                 break;
 437         case ICMP_TIME_EXCEEDED:
 438                 if (code != ICMP_EXC_TTL)
 439                         return 0;
 440                 break;
 441         }
 442
 443         err = -ENOENT;
 444
 445         read_lock(&ipip6_lock);
 446         t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
 447         if (t == NULL || t->parms.iph.daddr == 0)
 448                 goto out;
 449
 450         err = 0;
 451         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 452                 goto out;
 453
 454         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 455                 t->err_count++;
 456         else
 457                 t->err_count = 1;
 458         t->err_time = jiffies;
 459 out:
 460         read_unlock(&ipip6_lock);
 461         return err;
 462 #else
 463         struct iphdr *iph = (struct iphdr*)dp;
 464         int hlen = iph->ihl<<2;
 465         struct ipv6hdr *iph6;
 466         const int type = icmp_hdr(skb)->type;
 467         const int code = icmp_hdr(skb)->code;
 468         int rel_type = 0;
 469         int rel_code = 0;
 470         int rel_info = 0;
 471         struct sk_buff *skb2;
 472         struct rt6_info *rt6i;
 473
 474         if (len < hlen + sizeof(struct ipv6hdr))
 475                 return;
 476         iph6 = (struct ipv6hdr*)(dp + hlen);
 477
 478         switch (type) {
 479         default:
 480                 return;
 481         case ICMP_PARAMETERPROB:
 482                 if (icmp_hdr(skb)->un.gateway < hlen)
 483                         return;
 484
 485                 /* So... This guy found something strange INSIDE encapsulated
 486                    packet. Well, he is fool, but what can we do ?
 487                  */
 488                 rel_type = ICMPV6_PARAMPROB;
 489                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 490                 break;
 491
 492         case ICMP_DEST_UNREACH:
 493                 switch (code) {
 494                 case ICMP_SR_FAILED:
 495                 case ICMP_PORT_UNREACH:
 496                         /* Impossible event. */
 497                         return;
 498                 case ICMP_FRAG_NEEDED:
 499                         /* Too complicated case ... */
 500                         return;
 501                 default:
 502                         /* All others are translated to HOST_UNREACH.
 503                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 504                            I believe, it is just ether pollution. --ANK
 505                          */
 506                         rel_type = ICMPV6_DEST_UNREACH;
 507                         rel_code = ICMPV6_ADDR_UNREACH;
 508                         break;
 509                 }
 510                 break;
 511         case ICMP_TIME_EXCEEDED:
 512                 if (code != ICMP_EXC_TTL)
 513                         return;
 514                 rel_type = ICMPV6_TIME_EXCEED;
 515                 rel_code = ICMPV6_EXC_HOPLIMIT;
 516                 break;
 517         }
 518
 519         /* Prepare fake skb to feed it to icmpv6_send */
 520         skb2 = skb_clone(skb, GFP_ATOMIC);
 521         if (skb2 == NULL)
 522                 return 0;
 523         dst_release(skb2->dst);
 524         skb2->dst = NULL;
 525         skb_pull(skb2, skb->data - (u8*)iph6);
 526         skb_reset_network_header(skb2);
 527
 528         /* Try to guess incoming interface */
 529         rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
 530         if (rt6i && rt6i->rt6i_dev) {
 531                 skb2->dev = rt6i->rt6i_dev;
 532
 533                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 534
 535                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 536                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 537                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 538                                 rel_type = ICMPV6_DEST_UNREACH;
 539                                 rel_code = ICMPV6_ADDR_UNREACH;
 540                         }
 541                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 542                 }
 543         }
 544         kfree_skb(skb2);
 545         return 0;
 546 #endif
 547 }
 548
 549 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 550 {
 551         if (INET_ECN_is_ce(iph->tos))
 552                 IP6_ECN_set_ce(ipv6_hdr(skb));
 553 }
 554
 555 static int ipip6_rcv(struct sk_buff *skb)
 556 {
 557         struct iphdr *iph;
 558         struct ip_tunnel *tunnel;
 559
 560         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 561                 goto out;
 562
 563         iph = ip_hdr(skb);
 564
 565         read_lock(&ipip6_lock);
 566         if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
 567                                         iph->saddr, iph->daddr)) != NULL) {
 568                 secpath_reset(skb);
 569                 skb->mac_header = skb->network_header;
 570                 skb_reset_network_header(skb);
 571                 IPCB(skb)->flags = 0;
 572                 skb->protocol = htons(ETH_P_IPV6);
 573                 skb->pkt_type = PACKET_HOST;
 574
 575                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 576                     !isatap_chksrc(skb, iph, tunnel)) {
 577                         tunnel->stat.rx_errors++;
 578                         read_unlock(&ipip6_lock);
 579                         kfree_skb(skb);
 580                         return 0;
 581                 }
 582                 tunnel->stat.rx_packets++;
 583                 tunnel->stat.rx_bytes += skb->len;
 584                 skb->dev = tunnel->dev;
 585                 dst_release(skb->dst);
 586                 skb->dst = NULL;
 587                 nf_reset(skb);
 588                 ipip6_ecn_decapsulate(iph, skb);
 589                 netif_rx(skb);
 590                 read_unlock(&ipip6_lock);
 591                 return 0;
 592         }
 593
 594         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 595         kfree_skb(skb);
 596         read_unlock(&ipip6_lock);
 597 out:
 598         return 0;
 599 }
 600
 601 /* Returns the embedded IPv4 address if the IPv6 address
 602    comes from 6to4 (RFC 3056) addr space */
 603
 604 static inline __be32 try_6to4(struct in6_addr *v6dst)
 605 {
 606         __be32 dst = 0;
 607
 608         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 609                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 610                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 611         }
 612         return dst;
 613 }
 614
 615 /*
 616  *      This function assumes it is being called from dev_queue_xmit()
 617  *      and that skb is filled properly by that function.
 618  */
 619
 620 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 621 {
 622         struct ip_tunnel *tunnel = netdev_priv(dev);
 623         struct net_device_stats *stats = &tunnel->stat;
 624         struct iphdr  *tiph = &tunnel->parms.iph;
 625         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 626         u8     tos = tunnel->parms.iph.tos;
 627         struct rtable *rt;                      /* Route to the other host */
 628         struct net_device *tdev;                        /* Device to other host */
 629         struct iphdr  *iph;                     /* Our new IP header */
 630         unsigned int max_headroom;              /* The extra header space needed */
 631         __be32 dst = tiph->daddr;
 632         int    mtu;
 633         struct in6_addr *addr6;
 634         int addr_type;
 635
 636         if (tunnel->recursion++) {
 637                 tunnel->stat.collisions++;
 638                 goto tx_error;
 639         }
 640
 641         if (skb->protocol != htons(ETH_P_IPV6))
 642                 goto tx_error;
 643
 644         /* ISATAP (RFC4214) - must come before 6to4 */
 645         if (dev->priv_flags & IFF_ISATAP) {
 646                 struct neighbour *neigh = NULL;
 647
 648                 if (skb->dst)
 649                         neigh = skb->dst->neighbour;
 650
 651                 if (neigh == NULL) {
 652                         if (net_ratelimit())
 653                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 654                         goto tx_error;
 655                 }
 656
 657                 addr6 = (struct in6_addr*)&neigh->primary_key;
 658                 addr_type = ipv6_addr_type(addr6);
 659
 660                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 661                      ipv6_addr_is_isatap(addr6))
 662                         dst = addr6->s6_addr32[3];
 663                 else
 664                         goto tx_error;
 665         }
 666
 667         if (!dst)
 668                 dst = try_6to4(&iph6->daddr);
 669
 670         if (!dst) {
 671                 struct neighbour *neigh = NULL;
 672
 673                 if (skb->dst)
 674                         neigh = skb->dst->neighbour;
 675
 676                 if (neigh == NULL) {
 677                         if (net_ratelimit())
 678                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 679                         goto tx_error;
 680                 }
 681
 682                 addr6 = (struct in6_addr*)&neigh->primary_key;
 683                 addr_type = ipv6_addr_type(addr6);
 684
 685                 if (addr_type == IPV6_ADDR_ANY) {
 686                         addr6 = &ipv6_hdr(skb)->daddr;
 687                         addr_type = ipv6_addr_type(addr6);
 688                 }
 689
 690                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 691                         goto tx_error_icmp;
 692
 693                 dst = addr6->s6_addr32[3];
 694         }
 695
 696         {
 697                 struct flowi fl = { .nl_u = { .ip4_u =
 698                                               { .daddr = dst,
 699                                                 .saddr = tiph->saddr,
 700                                                 .tos = RT_TOS(tos) } },
 701                                     .oif = tunnel->parms.link,
 702                                     .proto = IPPROTO_IPV6 };
 703                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 704                         tunnel->stat.tx_carrier_errors++;
 705                         goto tx_error_icmp;
 706                 }
 707         }
 708         if (rt->rt_type != RTN_UNICAST) {
 709                 ip_rt_put(rt);
 710                 tunnel->stat.tx_carrier_errors++;
 711                 goto tx_error_icmp;
 712         }
 713         tdev = rt->u.dst.dev;
 714
 715         if (tdev == dev) {
 716                 ip_rt_put(rt);
 717                 tunnel->stat.collisions++;
 718                 goto tx_error;
 719         }
 720
 721         if (tiph->frag_off)
 722                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 723         else
 724                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 725
 726         if (mtu < 68) {
 727                 tunnel->stat.collisions++;
 728                 ip_rt_put(rt);
 729                 goto tx_error;
 730         }
 731         if (mtu < IPV6_MIN_MTU)
 732                 mtu = IPV6_MIN_MTU;
 733         if (tunnel->parms.iph.daddr && skb->dst)
 734                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 735
 736         if (skb->len > mtu) {
 737                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 738                 ip_rt_put(rt);
 739                 goto tx_error;
 740         }
 741
 742         if (tunnel->err_count > 0) {
 743                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 744                         tunnel->err_count--;
 745                         dst_link_failure(skb);
 746                 } else
 747                         tunnel->err_count = 0;
 748         }
 749
 750         /*
 751          * Okay, now see if we can stuff it in the buffer as-is.
 752          */
 753         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 754
 755         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 756             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 757                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 758                 if (!new_skb) {
 759                         ip_rt_put(rt);
 760                         stats->tx_dropped++;
 761                         dev_kfree_skb(skb);
 762                         tunnel->recursion--;
 763                         return 0;
 764                 }
 765                 if (skb->sk)
 766                         skb_set_owner_w(new_skb, skb->sk);
 767                 dev_kfree_skb(skb);
 768                 skb = new_skb;
 769                 iph6 = ipv6_hdr(skb);
 770         }
 771
 772         skb->transport_header = skb->network_header;
 773         skb_push(skb, sizeof(struct iphdr));
 774         skb_reset_network_header(skb);
 775         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 776         IPCB(skb)->flags = 0;
 777         dst_release(skb->dst);
 778         skb->dst = &rt->u.dst;
 779
 780         /*
 781          *      Push down and install the IPIP header.
 782          */
 783
 784         iph                     =       ip_hdr(skb);
 785         iph->version            =       4;
 786         iph->ihl                =       sizeof(struct iphdr)>>2;
 787         if (mtu > IPV6_MIN_MTU)
 788                 iph->frag_off   =       htons(IP_DF);
 789         else
 790                 iph->frag_off   =       0;
 791
 792         iph->protocol           =       IPPROTO_IPV6;
 793         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 794         iph->daddr              =       rt->rt_dst;
 795         iph->saddr              =       rt->rt_src;
 796
 797         if ((iph->ttl = tiph->ttl) == 0)
 798                 iph->ttl        =       iph6->hop_limit;
 799
 800         nf_reset(skb);
 801
 802         IPTUNNEL_XMIT();
 803         tunnel->recursion--;
 804         return 0;
 805
 806 tx_error_icmp:
 807         dst_link_failure(skb);
 808 tx_error:
 809         stats->tx_errors++;
 810         dev_kfree_skb(skb);
 811         tunnel->recursion--;
 812         return 0;
 813 }
 814
 815 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 816 {
 817         struct net_device *tdev = NULL;
 818         struct ip_tunnel *tunnel;
 819         struct iphdr *iph;
 820
 821         tunnel = netdev_priv(dev);
 822         iph = &tunnel->parms.iph;
 823
 824         if (iph->daddr) {
 825                 struct flowi fl = { .nl_u = { .ip4_u =
 826                                               { .daddr = iph->daddr,
 827                                                 .saddr = iph->saddr,
 828                                                 .tos = RT_TOS(iph->tos) } },
 829                                     .oif = tunnel->parms.link,
 830                                     .proto = IPPROTO_IPV6 };
 831                 struct rtable *rt;
 832                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 833                         tdev = rt->u.dst.dev;
 834                         ip_rt_put(rt);
 835                 }
 836                 dev->flags |= IFF_POINTOPOINT;
 837         }
 838
 839         if (!tdev && tunnel->parms.link)
 840                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 841
 842         if (tdev) {
 843                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 844                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 845                 if (dev->mtu < IPV6_MIN_MTU)
 846                         dev->mtu = IPV6_MIN_MTU;
 847         }
 848         dev->iflink = tunnel->parms.link;
 849 }
 850
 851 static int
 852 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 853 {
 854         int err = 0;
 855         struct ip_tunnel_parm p;
 856         struct ip_tunnel_prl prl;
 857         struct ip_tunnel *t;
 858         struct net *net = dev_net(dev);
 859         struct sit_net *sitn = net_generic(net, sit_net_id);
 860
 861         switch (cmd) {
 862         case SIOCGETTUNNEL:
 863                 t = NULL;
 864                 if (dev == sitn->fb_tunnel_dev) {
 865                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 866                                 err = -EFAULT;
 867                                 break;
 868                         }
 869                         t = ipip6_tunnel_locate(net, &p, 0);
 870                 }
 871                 if (t == NULL)
 872                         t = netdev_priv(dev);
 873                 memcpy(&p, &t->parms, sizeof(p));
 874                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 875                         err = -EFAULT;
 876                 break;
 877
 878         case SIOCADDTUNNEL:
 879         case SIOCCHGTUNNEL:
 880                 err = -EPERM;
 881                 if (!capable(CAP_NET_ADMIN))
 882                         goto done;
 883
 884                 err = -EFAULT;
 885                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 886                         goto done;
 887
 888                 err = -EINVAL;
 889                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 890                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 891                         goto done;
 892                 if (p.iph.ttl)
 893                         p.iph.frag_off |= htons(IP_DF);
 894
 895                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 896
 897                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 898                         if (t != NULL) {
 899                                 if (t->dev != dev) {
 900                                         err = -EEXIST;
 901                                         break;
 902                                 }
 903                         } else {
 904                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 905                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 906                                         err = -EINVAL;
 907                                         break;
 908                                 }
 909                                 t = netdev_priv(dev);
 910                                 ipip6_tunnel_unlink(sitn, t);
 911                                 t->parms.iph.saddr = p.iph.saddr;
 912                                 t->parms.iph.daddr = p.iph.daddr;
 913                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 914                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 915                                 ipip6_tunnel_link(sitn, t);
 916                                 netdev_state_change(dev);
 917                         }
 918                 }
 919
 920                 if (t) {
 921                         err = 0;
 922                         if (cmd == SIOCCHGTUNNEL) {
 923                                 t->parms.iph.ttl = p.iph.ttl;
 924                                 t->parms.iph.tos = p.iph.tos;
 925                                 if (t->parms.link != p.link) {
 926                                         t->parms.link = p.link;
 927                                         ipip6_tunnel_bind_dev(dev);
 928                                         netdev_state_change(dev);
 929                                 }
 930                         }
 931                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 932                                 err = -EFAULT;
 933                 } else
 934                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 935                 break;
 936
 937         case SIOCDELTUNNEL:
 938                 err = -EPERM;
 939                 if (!capable(CAP_NET_ADMIN))
 940                         goto done;
 941
 942                 if (dev == sitn->fb_tunnel_dev) {
 943                         err = -EFAULT;
 944                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 945                                 goto done;
 946                         err = -ENOENT;
 947                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
 948                                 goto done;
 949                         err = -EPERM;
 950                         if (t == netdev_priv(sitn->fb_tunnel_dev))
 951                                 goto done;
 952                         dev = t->dev;
 953                 }
 954                 unregister_netdevice(dev);
 955                 err = 0;
 956                 break;
 957
 958         case SIOCGETPRL:
 959         case SIOCADDPRL:
 960         case SIOCDELPRL:
 961         case SIOCCHGPRL:
 962                 err = -EPERM;
 963                 if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
 964                         goto done;
 965                 err = -EINVAL;
 966                 if (dev == sitn->fb_tunnel_dev)
 967                         goto done;
 968                 err = -EFAULT;
 969                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 970                         goto done;
 971                 err = -ENOENT;
 972                 if (!(t = netdev_priv(dev)))
 973                         goto done;
 974
 975                 switch (cmd) {
 976                 case SIOCGETPRL:
 977                         err = ipip6_tunnel_get_prl(t, &prl);
 978                         if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
 979                                                  &prl, sizeof(prl)))
 980                                 err = -EFAULT;
 981                         break;
 982                 case SIOCDELPRL:
 983                         err = ipip6_tunnel_del_prl(t, &prl);
 984                         break;
 985                 case SIOCADDPRL:
 986                 case SIOCCHGPRL:
 987                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 988                         break;
 989                 }
 990                 if (cmd != SIOCGETPRL)
 991                         netdev_state_change(dev);
 992                 break;
 993
 994         default:
 995                 err = -EINVAL;
 996         }
 997
 998 done:
 999         return err;
1000 }
1001
1002 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
1003 {
1004         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
1005 }
1006
1007 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1008 {
1009         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1010                 return -EINVAL;
1011         dev->mtu = new_mtu;
1012         return 0;
1013 }
1014
1015 static void ipip6_tunnel_setup(struct net_device *dev)
1016 {
1017         dev->uninit             = ipip6_tunnel_uninit;
1018         dev->destructor         = free_netdev;
1019         dev->hard_start_xmit    = ipip6_tunnel_xmit;
1020         dev->get_stats          = ipip6_tunnel_get_stats;
1021         dev->do_ioctl           = ipip6_tunnel_ioctl;
1022         dev->change_mtu         = ipip6_tunnel_change_mtu;
1023
1024         dev->type               = ARPHRD_SIT;
1025         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1026         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1027         dev->flags              = IFF_NOARP;
1028         dev->iflink             = 0;
1029         dev->addr_len           = 4;
1030 }
1031
1032 static int ipip6_tunnel_init(struct net_device *dev)
1033 {
1034         struct ip_tunnel *tunnel;
1035
1036         tunnel = netdev_priv(dev);
1037
1038         tunnel->dev = dev;
1039         strcpy(tunnel->parms.name, dev->name);
1040
1041         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1042         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1043
1044         ipip6_tunnel_bind_dev(dev);
1045
1046         return 0;
1047 }
1048
1049 static int ipip6_fb_tunnel_init(struct net_device *dev)
1050 {
1051         struct ip_tunnel *tunnel = netdev_priv(dev);
1052         struct iphdr *iph = &tunnel->parms.iph;
1053
1054         tunnel->dev = dev;
1055         strcpy(tunnel->parms.name, dev->name);
1056
1057         iph->version            = 4;
1058         iph->protocol           = IPPROTO_IPV6;
1059         iph->ihl                = 5;
1060         iph->ttl                = 64;
1061
1062         dev_hold(dev);
1063         tunnels_wc[0]           = tunnel;
1064         return 0;
1065 }
1066
1067 static struct xfrm_tunnel sit_handler = {
1068         .handler        =       ipip6_rcv,
1069         .err_handler    =       ipip6_err,
1070         .priority       =       1,
1071 };
1072
1073 static void __exit sit_destroy_tunnels(void)
1074 {
1075         int prio;
1076
1077         for (prio = 1; prio < 4; prio++) {
1078                 int h;
1079                 for (h = 0; h < HASH_SIZE; h++) {
1080                         struct ip_tunnel *t;
1081                         while ((t = tunnels[prio][h]) != NULL)
1082                                 unregister_netdevice(t->dev);
1083                 }
1084         }
1085 }
1086
1087 static int sit_init_net(struct net *net)
1088 {
1089         int err;
1090         struct sit_net *sitn;
1091
1092         err = -ENOMEM;
1093         sitn = kmalloc(sizeof(struct sit_net), GFP_KERNEL);
1094         if (sitn == NULL)
1095                 goto err_alloc;
1096
1097         err = net_assign_generic(net, sit_net_id, sitn);
1098         if (err < 0)
1099                 goto err_assign;
1100
1101         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1102                                            ipip6_tunnel_setup);
1103         if (!sitn->fb_tunnel_dev) {
1104                 err = -ENOMEM;
1105                 goto err_alloc_dev;
1106         }
1107
1108         sitn->fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1109         dev_net_set(sitn->fb_tunnel_dev, net);
1110
1111         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1112                 goto err_reg_dev;
1113
1114         return 0;
1115
1116 err_reg_dev:
1117         free_netdev(sitn->fb_tunnel_dev);
1118 err_alloc_dev:
1119         /* nothing */
1120 err_assign:
1121         kfree(sitn);
1122 err_alloc:
1123         return err;
1124 }
1125
1126 static void sit_exit_net(struct net *net)
1127 {
1128         struct sit_net *sitn;
1129
1130         sitn = net_generic(net, sit_net_id);
1131         rtnl_lock();
1132         unregister_netdevice(sitn->fb_tunnel_dev);
1133         rtnl_unlock();
1134         kfree(sitn);
1135 }
1136
1137 static struct pernet_operations sit_net_ops = {
1138         .init = sit_init_net,
1139         .exit = sit_exit_net,
1140 };
1141
1142 static void __exit sit_cleanup(void)
1143 {
1144         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1145
1146         rtnl_lock();
1147         sit_destroy_tunnels();
1148         rtnl_unlock();
1149
1150         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1151 }
1152
1153 static int __init sit_init(void)
1154 {
1155         int err;
1156
1157         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1158
1159         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1160                 printk(KERN_INFO "sit init: Can't add protocol\n");
1161                 return -EAGAIN;
1162         }
1163
1164         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1165         if (err < 0)
1166                 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1167
1168         return err;
1169 }
1170
1171 module_init(sit_init);
1172 module_exit(sit_cleanup);
1173 MODULE_LICENSE("GPL");
1174 MODULE_ALIAS("sit0");