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 ip_tunnel *tunnels_r_l[HASH_SIZE];
  74         struct ip_tunnel *tunnels_r[HASH_SIZE];
  75         struct ip_tunnel *tunnels_l[HASH_SIZE];
  76         struct ip_tunnel *tunnels_wc[1];
  77         struct ip_tunnel **tunnels[4];
  78
  79         struct net_device *fb_tunnel_dev;
  80 };
  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         struct sit_net *sitn = net_generic(net, sit_net_id);
  91
  92         for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
  93                 if (local == t->parms.iph.saddr &&
  94                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  95                         return t;
  96         }
  97         for (t = sitn->tunnels_r[h0]; t; t = t->next) {
  98                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  99                         return t;
 100         }
 101         for (t = sitn->tunnels_l[h1]; t; t = t->next) {
 102                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
 103                         return t;
 104         }
 105         if ((t = sitn->tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
 106                 return t;
 107         return NULL;
 108 }
 109
 110 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
 111                 struct ip_tunnel_parm *parms)
 112 {
 113         __be32 remote = parms->iph.daddr;
 114         __be32 local = parms->iph.saddr;
 115         unsigned h = 0;
 116         int prio = 0;
 117
 118         if (remote) {
 119                 prio |= 2;
 120                 h ^= HASH(remote);
 121         }
 122         if (local) {
 123                 prio |= 1;
 124                 h ^= HASH(local);
 125         }
 126         return &sitn->tunnels[prio][h];
 127 }
 128
 129 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
 130                 struct ip_tunnel *t)
 131 {
 132         return __ipip6_bucket(sitn, &t->parms);
 133 }
 134
 135 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 136 {
 137         struct ip_tunnel **tp;
 138
 139         for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
 140                 if (t == *tp) {
 141                         write_lock_bh(&ipip6_lock);
 142                         *tp = t->next;
 143                         write_unlock_bh(&ipip6_lock);
 144                         break;
 145                 }
 146         }
 147 }
 148
 149 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 150 {
 151         struct ip_tunnel **tp = ipip6_bucket(sitn, t);
 152
 153         t->next = *tp;
 154         write_lock_bh(&ipip6_lock);
 155         *tp = t;
 156         write_unlock_bh(&ipip6_lock);
 157 }
 158
 159 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
 160                 struct ip_tunnel_parm *parms, int create)
 161 {
 162         __be32 remote = parms->iph.daddr;
 163         __be32 local = parms->iph.saddr;
 164         struct ip_tunnel *t, **tp, *nt;
 165         struct net_device *dev;
 166         char name[IFNAMSIZ];
 167         struct sit_net *sitn = net_generic(net, sit_net_id);
 168
 169         for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
 170                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 171                         return t;
 172         }
 173         if (!create)
 174                 goto failed;
 175
 176         if (parms->name[0])
 177                 strlcpy(name, parms->name, IFNAMSIZ);
 178         else
 179                 sprintf(name, "sit%%d");
 180
 181         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 182         if (dev == NULL)
 183                 return NULL;
 184
 185         if (strchr(name, '%')) {
 186                 if (dev_alloc_name(dev, name) < 0)
 187                         goto failed_free;
 188         }
 189
 190         nt = netdev_priv(dev);
 191         dev->init = ipip6_tunnel_init;
 192         nt->parms = *parms;
 193
 194         if (parms->i_flags & SIT_ISATAP)
 195                 dev->priv_flags |= IFF_ISATAP;
 196
 197         if (register_netdevice(dev) < 0)
 198                 goto failed_free;
 199
 200         dev_hold(dev);
 201
 202         ipip6_tunnel_link(sitn, nt);
 203         return nt;
 204
 205 failed_free:
 206         free_netdev(dev);
 207 failed:
 208         return NULL;
 209 }
 210
 211 static struct ip_tunnel_prl_entry *
 212 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 213 {
 214         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 215
 216         for (p = t->prl; p; p = p->next)
 217                 if (p->addr == addr)
 218                         break;
 219         return p;
 220
 221 }
 222
 223 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 224 {
 225         struct ip_tunnel_prl *kp;
 226         struct ip_tunnel_prl_entry *prl;
 227         unsigned int cmax, c = 0, ca, len;
 228         int ret = 0;
 229
 230         cmax = a->datalen / sizeof(*a);
 231         if (cmax > 1 && a->addr != htonl(INADDR_ANY))
 232                 cmax = 1;
 233
 234         /* For simple GET or for root users,
 235          * we try harder to allocate.
 236          */
 237         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 238                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 239                 NULL;
 240
 241         read_lock(&ipip6_lock);
 242
 243         ca = t->prl_count < cmax ? t->prl_count : cmax;
 244
 245         if (!kp) {
 246                 /* We don't try hard to allocate much memory for
 247                  * non-root users.
 248                  * For root users, retry allocating enough memory for
 249                  * the answer.
 250                  */
 251                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 252                 if (!kp) {
 253                         ret = -ENOMEM;
 254                         goto out;
 255                 }
 256         }
 257
 258         c = 0;
 259         for (prl = t->prl; prl; prl = prl->next) {
 260                 if (c > cmax)
 261                         break;
 262                 if (a->addr != htonl(INADDR_ANY) && prl->addr != a->addr)
 263                         continue;
 264                 kp[c].addr = prl->addr;
 265                 kp[c].flags = prl->flags;
 266                 c++;
 267                 if (a->addr != htonl(INADDR_ANY))
 268                         break;
 269         }
 270 out:
 271         read_unlock(&ipip6_lock);
 272
 273         len = sizeof(*kp) * c;
 274         ret = len ? copy_to_user(a->data, kp, len) : 0;
 275
 276         kfree(kp);
 277         if (ret)
 278                 return -EFAULT;
 279
 280         a->datalen = len;
 281         return 0;
 282 }
 283
 284 static int
 285 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 286 {
 287         struct ip_tunnel_prl_entry *p;
 288         int err = 0;
 289
 290         if (a->addr == htonl(INADDR_ANY))
 291                 return -EINVAL;
 292
 293         write_lock(&ipip6_lock);
 294
 295         for (p = t->prl; p; p = p->next) {
 296                 if (p->addr == a->addr) {
 297                         if (chg)
 298                                 goto update;
 299                         err = -EEXIST;
 300                         goto out;
 301                 }
 302         }
 303
 304         if (chg) {
 305                 err = -ENXIO;
 306                 goto out;
 307         }
 308
 309         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 310         if (!p) {
 311                 err = -ENOBUFS;
 312                 goto out;
 313         }
 314
 315         p->next = t->prl;
 316         t->prl = p;
 317         t->prl_count++;
 318 update:
 319         p->addr = a->addr;
 320         p->flags = a->flags;
 321 out:
 322         write_unlock(&ipip6_lock);
 323         return err;
 324 }
 325
 326 static int
 327 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 328 {
 329         struct ip_tunnel_prl_entry *x, **p;
 330         int err = 0;
 331
 332         write_lock(&ipip6_lock);
 333
 334         if (a && a->addr != htonl(INADDR_ANY)) {
 335                 for (p = &t->prl; *p; p = &(*p)->next) {
 336                         if ((*p)->addr == a->addr) {
 337                                 x = *p;
 338                                 *p = x->next;
 339                                 kfree(x);
 340                                 t->prl_count--;
 341                                 goto out;
 342                         }
 343                 }
 344                 err = -ENXIO;
 345         } else {
 346                 while (t->prl) {
 347                         x = t->prl;
 348                         t->prl = t->prl->next;
 349                         kfree(x);
 350                         t->prl_count--;
 351                 }
 352         }
 353 out:
 354         write_unlock(&ipip6_lock);
 355         return 0;
 356 }
 357
 358 static int
 359 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 360 {
 361         struct ip_tunnel_prl_entry *p;
 362         int ok = 1;
 363
 364         read_lock(&ipip6_lock);
 365         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 366         if (p) {
 367                 if (p->flags & PRL_DEFAULT)
 368                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 369                 else
 370                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 371         } else {
 372                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 373                 if (ipv6_addr_is_isatap(addr6) &&
 374                     (addr6->s6_addr32[3] == iph->saddr) &&
 375                     ipv6_chk_prefix(addr6, t->dev))
 376                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 377                 else
 378                         ok = 0;
 379         }
 380         read_unlock(&ipip6_lock);
 381         return ok;
 382 }
 383
 384 static void ipip6_tunnel_uninit(struct net_device *dev)
 385 {
 386         struct net *net = dev_net(dev);
 387         struct sit_net *sitn = net_generic(net, sit_net_id);
 388
 389         if (dev == sitn->fb_tunnel_dev) {
 390                 write_lock_bh(&ipip6_lock);
 391                 sitn->tunnels_wc[0] = NULL;
 392                 write_unlock_bh(&ipip6_lock);
 393                 dev_put(dev);
 394         } else {
 395                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 396                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 397                 dev_put(dev);
 398         }
 399 }
 400
 401
 402 static int ipip6_err(struct sk_buff *skb, u32 info)
 403 {
 404 #ifndef I_WISH_WORLD_WERE_PERFECT
 405
 406 /* It is not :-( All the routers (except for Linux) return only
 407    8 bytes of packet payload. It means, that precise relaying of
 408    ICMP in the real Internet is absolutely infeasible.
 409  */
 410         struct iphdr *iph = (struct iphdr*)skb->data;
 411         const int type = icmp_hdr(skb)->type;
 412         const int code = icmp_hdr(skb)->code;
 413         struct ip_tunnel *t;
 414         int err;
 415
 416         switch (type) {
 417         default:
 418         case ICMP_PARAMETERPROB:
 419                 return 0;
 420
 421         case ICMP_DEST_UNREACH:
 422                 switch (code) {
 423                 case ICMP_SR_FAILED:
 424                 case ICMP_PORT_UNREACH:
 425                         /* Impossible event. */
 426                         return 0;
 427                 case ICMP_FRAG_NEEDED:
 428                         /* Soft state for pmtu is maintained by IP core. */
 429                         return 0;
 430                 default:
 431                         /* All others are translated to HOST_UNREACH.
 432                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 433                            I believe they are just ether pollution. --ANK
 434                          */
 435                         break;
 436                 }
 437                 break;
 438         case ICMP_TIME_EXCEEDED:
 439                 if (code != ICMP_EXC_TTL)
 440                         return 0;
 441                 break;
 442         }
 443
 444         err = -ENOENT;
 445
 446         read_lock(&ipip6_lock);
 447         t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
 448         if (t == NULL || t->parms.iph.daddr == 0)
 449                 goto out;
 450
 451         err = 0;
 452         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 453                 goto out;
 454
 455         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 456                 t->err_count++;
 457         else
 458                 t->err_count = 1;
 459         t->err_time = jiffies;
 460 out:
 461         read_unlock(&ipip6_lock);
 462         return err;
 463 #else
 464         struct iphdr *iph = (struct iphdr*)dp;
 465         int hlen = iph->ihl<<2;
 466         struct ipv6hdr *iph6;
 467         const int type = icmp_hdr(skb)->type;
 468         const int code = icmp_hdr(skb)->code;
 469         int rel_type = 0;
 470         int rel_code = 0;
 471         int rel_info = 0;
 472         struct sk_buff *skb2;
 473         struct rt6_info *rt6i;
 474
 475         if (len < hlen + sizeof(struct ipv6hdr))
 476                 return;
 477         iph6 = (struct ipv6hdr*)(dp + hlen);
 478
 479         switch (type) {
 480         default:
 481                 return;
 482         case ICMP_PARAMETERPROB:
 483                 if (icmp_hdr(skb)->un.gateway < hlen)
 484                         return;
 485
 486                 /* So... This guy found something strange INSIDE encapsulated
 487                    packet. Well, he is fool, but what can we do ?
 488                  */
 489                 rel_type = ICMPV6_PARAMPROB;
 490                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 491                 break;
 492
 493         case ICMP_DEST_UNREACH:
 494                 switch (code) {
 495                 case ICMP_SR_FAILED:
 496                 case ICMP_PORT_UNREACH:
 497                         /* Impossible event. */
 498                         return;
 499                 case ICMP_FRAG_NEEDED:
 500                         /* Too complicated case ... */
 501                         return;
 502                 default:
 503                         /* All others are translated to HOST_UNREACH.
 504                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 505                            I believe, it is just ether pollution. --ANK
 506                          */
 507                         rel_type = ICMPV6_DEST_UNREACH;
 508                         rel_code = ICMPV6_ADDR_UNREACH;
 509                         break;
 510                 }
 511                 break;
 512         case ICMP_TIME_EXCEEDED:
 513                 if (code != ICMP_EXC_TTL)
 514                         return;
 515                 rel_type = ICMPV6_TIME_EXCEED;
 516                 rel_code = ICMPV6_EXC_HOPLIMIT;
 517                 break;
 518         }
 519
 520         /* Prepare fake skb to feed it to icmpv6_send */
 521         skb2 = skb_clone(skb, GFP_ATOMIC);
 522         if (skb2 == NULL)
 523                 return 0;
 524         dst_release(skb2->dst);
 525         skb2->dst = NULL;
 526         skb_pull(skb2, skb->data - (u8*)iph6);
 527         skb_reset_network_header(skb2);
 528
 529         /* Try to guess incoming interface */
 530         rt6i = rt6_lookup(dev_net(skb->dev), &iph6->saddr, NULL, NULL, 0);
 531         if (rt6i && rt6i->rt6i_dev) {
 532                 skb2->dev = rt6i->rt6i_dev;
 533
 534                 rt6i = rt6_lookup(dev_net(skb->dev),
 535                                 &iph6->daddr, &iph6->saddr, NULL, 0);
 536
 537                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 538                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 539                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 540                                 rel_type = ICMPV6_DEST_UNREACH;
 541                                 rel_code = ICMPV6_ADDR_UNREACH;
 542                         }
 543                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 544                 }
 545         }
 546         kfree_skb(skb2);
 547         return 0;
 548 #endif
 549 }
 550
 551 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 552 {
 553         if (INET_ECN_is_ce(iph->tos))
 554                 IP6_ECN_set_ce(ipv6_hdr(skb));
 555 }
 556
 557 static int ipip6_rcv(struct sk_buff *skb)
 558 {
 559         struct iphdr *iph;
 560         struct ip_tunnel *tunnel;
 561
 562         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 563                 goto out;
 564
 565         iph = ip_hdr(skb);
 566
 567         read_lock(&ipip6_lock);
 568         if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
 569                                         iph->saddr, iph->daddr)) != NULL) {
 570                 secpath_reset(skb);
 571                 skb->mac_header = skb->network_header;
 572                 skb_reset_network_header(skb);
 573                 IPCB(skb)->flags = 0;
 574                 skb->protocol = htons(ETH_P_IPV6);
 575                 skb->pkt_type = PACKET_HOST;
 576
 577                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 578                     !isatap_chksrc(skb, iph, tunnel)) {
 579                         tunnel->stat.rx_errors++;
 580                         read_unlock(&ipip6_lock);
 581                         kfree_skb(skb);
 582                         return 0;
 583                 }
 584                 tunnel->stat.rx_packets++;
 585                 tunnel->stat.rx_bytes += skb->len;
 586                 skb->dev = tunnel->dev;
 587                 dst_release(skb->dst);
 588                 skb->dst = NULL;
 589                 nf_reset(skb);
 590                 ipip6_ecn_decapsulate(iph, skb);
 591                 netif_rx(skb);
 592                 read_unlock(&ipip6_lock);
 593                 return 0;
 594         }
 595
 596         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 597         kfree_skb(skb);
 598         read_unlock(&ipip6_lock);
 599 out:
 600         return 0;
 601 }
 602
 603 /* Returns the embedded IPv4 address if the IPv6 address
 604    comes from 6to4 (RFC 3056) addr space */
 605
 606 static inline __be32 try_6to4(struct in6_addr *v6dst)
 607 {
 608         __be32 dst = 0;
 609
 610         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 611                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 612                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 613         }
 614         return dst;
 615 }
 616
 617 /*
 618  *      This function assumes it is being called from dev_queue_xmit()
 619  *      and that skb is filled properly by that function.
 620  */
 621
 622 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 623 {
 624         struct ip_tunnel *tunnel = netdev_priv(dev);
 625         struct net_device_stats *stats = &tunnel->stat;
 626         struct iphdr  *tiph = &tunnel->parms.iph;
 627         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 628         u8     tos = tunnel->parms.iph.tos;
 629         struct rtable *rt;                      /* Route to the other host */
 630         struct net_device *tdev;                        /* Device to other host */
 631         struct iphdr  *iph;                     /* Our new IP header */
 632         unsigned int max_headroom;              /* The extra header space needed */
 633         __be32 dst = tiph->daddr;
 634         int    mtu;
 635         struct in6_addr *addr6;
 636         int addr_type;
 637
 638         if (tunnel->recursion++) {
 639                 tunnel->stat.collisions++;
 640                 goto tx_error;
 641         }
 642
 643         if (skb->protocol != htons(ETH_P_IPV6))
 644                 goto tx_error;
 645
 646         /* ISATAP (RFC4214) - must come before 6to4 */
 647         if (dev->priv_flags & IFF_ISATAP) {
 648                 struct neighbour *neigh = NULL;
 649
 650                 if (skb->dst)
 651                         neigh = skb->dst->neighbour;
 652
 653                 if (neigh == NULL) {
 654                         if (net_ratelimit())
 655                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 656                         goto tx_error;
 657                 }
 658
 659                 addr6 = (struct in6_addr*)&neigh->primary_key;
 660                 addr_type = ipv6_addr_type(addr6);
 661
 662                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 663                      ipv6_addr_is_isatap(addr6))
 664                         dst = addr6->s6_addr32[3];
 665                 else
 666                         goto tx_error;
 667         }
 668
 669         if (!dst)
 670                 dst = try_6to4(&iph6->daddr);
 671
 672         if (!dst) {
 673                 struct neighbour *neigh = NULL;
 674
 675                 if (skb->dst)
 676                         neigh = skb->dst->neighbour;
 677
 678                 if (neigh == NULL) {
 679                         if (net_ratelimit())
 680                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 681                         goto tx_error;
 682                 }
 683
 684                 addr6 = (struct in6_addr*)&neigh->primary_key;
 685                 addr_type = ipv6_addr_type(addr6);
 686
 687                 if (addr_type == IPV6_ADDR_ANY) {
 688                         addr6 = &ipv6_hdr(skb)->daddr;
 689                         addr_type = ipv6_addr_type(addr6);
 690                 }
 691
 692                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 693                         goto tx_error_icmp;
 694
 695                 dst = addr6->s6_addr32[3];
 696         }
 697
 698         {
 699                 struct flowi fl = { .nl_u = { .ip4_u =
 700                                               { .daddr = dst,
 701                                                 .saddr = tiph->saddr,
 702                                                 .tos = RT_TOS(tos) } },
 703                                     .oif = tunnel->parms.link,
 704                                     .proto = IPPROTO_IPV6 };
 705                 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
 706                         tunnel->stat.tx_carrier_errors++;
 707                         goto tx_error_icmp;
 708                 }
 709         }
 710         if (rt->rt_type != RTN_UNICAST) {
 711                 ip_rt_put(rt);
 712                 tunnel->stat.tx_carrier_errors++;
 713                 goto tx_error_icmp;
 714         }
 715         tdev = rt->u.dst.dev;
 716
 717         if (tdev == dev) {
 718                 ip_rt_put(rt);
 719                 tunnel->stat.collisions++;
 720                 goto tx_error;
 721         }
 722
 723         if (tiph->frag_off)
 724                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 725         else
 726                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 727
 728         if (mtu < 68) {
 729                 tunnel->stat.collisions++;
 730                 ip_rt_put(rt);
 731                 goto tx_error;
 732         }
 733         if (mtu < IPV6_MIN_MTU)
 734                 mtu = IPV6_MIN_MTU;
 735         if (tunnel->parms.iph.daddr && skb->dst)
 736                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 737
 738         if (skb->len > mtu) {
 739                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 740                 ip_rt_put(rt);
 741                 goto tx_error;
 742         }
 743
 744         if (tunnel->err_count > 0) {
 745                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 746                         tunnel->err_count--;
 747                         dst_link_failure(skb);
 748                 } else
 749                         tunnel->err_count = 0;
 750         }
 751
 752         /*
 753          * Okay, now see if we can stuff it in the buffer as-is.
 754          */
 755         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 756
 757         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 758             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 759                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 760                 if (!new_skb) {
 761                         ip_rt_put(rt);
 762                         stats->tx_dropped++;
 763                         dev_kfree_skb(skb);
 764                         tunnel->recursion--;
 765                         return 0;
 766                 }
 767                 if (skb->sk)
 768                         skb_set_owner_w(new_skb, skb->sk);
 769                 dev_kfree_skb(skb);
 770                 skb = new_skb;
 771                 iph6 = ipv6_hdr(skb);
 772         }
 773
 774         skb->transport_header = skb->network_header;
 775         skb_push(skb, sizeof(struct iphdr));
 776         skb_reset_network_header(skb);
 777         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 778         IPCB(skb)->flags = 0;
 779         dst_release(skb->dst);
 780         skb->dst = &rt->u.dst;
 781
 782         /*
 783          *      Push down and install the IPIP header.
 784          */
 785
 786         iph                     =       ip_hdr(skb);
 787         iph->version            =       4;
 788         iph->ihl                =       sizeof(struct iphdr)>>2;
 789         if (mtu > IPV6_MIN_MTU)
 790                 iph->frag_off   =       htons(IP_DF);
 791         else
 792                 iph->frag_off   =       0;
 793
 794         iph->protocol           =       IPPROTO_IPV6;
 795         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 796         iph->daddr              =       rt->rt_dst;
 797         iph->saddr              =       rt->rt_src;
 798
 799         if ((iph->ttl = tiph->ttl) == 0)
 800                 iph->ttl        =       iph6->hop_limit;
 801
 802         nf_reset(skb);
 803
 804         IPTUNNEL_XMIT();
 805         tunnel->recursion--;
 806         return 0;
 807
 808 tx_error_icmp:
 809         dst_link_failure(skb);
 810 tx_error:
 811         stats->tx_errors++;
 812         dev_kfree_skb(skb);
 813         tunnel->recursion--;
 814         return 0;
 815 }
 816
 817 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 818 {
 819         struct net_device *tdev = NULL;
 820         struct ip_tunnel *tunnel;
 821         struct iphdr *iph;
 822
 823         tunnel = netdev_priv(dev);
 824         iph = &tunnel->parms.iph;
 825
 826         if (iph->daddr) {
 827                 struct flowi fl = { .nl_u = { .ip4_u =
 828                                               { .daddr = iph->daddr,
 829                                                 .saddr = iph->saddr,
 830                                                 .tos = RT_TOS(iph->tos) } },
 831                                     .oif = tunnel->parms.link,
 832                                     .proto = IPPROTO_IPV6 };
 833                 struct rtable *rt;
 834                 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
 835                         tdev = rt->u.dst.dev;
 836                         ip_rt_put(rt);
 837                 }
 838                 dev->flags |= IFF_POINTOPOINT;
 839         }
 840
 841         if (!tdev && tunnel->parms.link)
 842                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 843
 844         if (tdev) {
 845                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 846                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 847                 if (dev->mtu < IPV6_MIN_MTU)
 848                         dev->mtu = IPV6_MIN_MTU;
 849         }
 850         dev->iflink = tunnel->parms.link;
 851 }
 852
 853 static int
 854 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 855 {
 856         int err = 0;
 857         struct ip_tunnel_parm p;
 858         struct ip_tunnel_prl prl;
 859         struct ip_tunnel *t;
 860         struct net *net = dev_net(dev);
 861         struct sit_net *sitn = net_generic(net, sit_net_id);
 862
 863         switch (cmd) {
 864         case SIOCGETTUNNEL:
 865                 t = NULL;
 866                 if (dev == sitn->fb_tunnel_dev) {
 867                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 868                                 err = -EFAULT;
 869                                 break;
 870                         }
 871                         t = ipip6_tunnel_locate(net, &p, 0);
 872                 }
 873                 if (t == NULL)
 874                         t = netdev_priv(dev);
 875                 memcpy(&p, &t->parms, sizeof(p));
 876                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 877                         err = -EFAULT;
 878                 break;
 879
 880         case SIOCADDTUNNEL:
 881         case SIOCCHGTUNNEL:
 882                 err = -EPERM;
 883                 if (!capable(CAP_NET_ADMIN))
 884                         goto done;
 885
 886                 err = -EFAULT;
 887                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 888                         goto done;
 889
 890                 err = -EINVAL;
 891                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 892                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 893                         goto done;
 894                 if (p.iph.ttl)
 895                         p.iph.frag_off |= htons(IP_DF);
 896
 897                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 898
 899                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 900                         if (t != NULL) {
 901                                 if (t->dev != dev) {
 902                                         err = -EEXIST;
 903                                         break;
 904                                 }
 905                         } else {
 906                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 907                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 908                                         err = -EINVAL;
 909                                         break;
 910                                 }
 911                                 t = netdev_priv(dev);
 912                                 ipip6_tunnel_unlink(sitn, t);
 913                                 t->parms.iph.saddr = p.iph.saddr;
 914                                 t->parms.iph.daddr = p.iph.daddr;
 915                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 916                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 917                                 ipip6_tunnel_link(sitn, t);
 918                                 netdev_state_change(dev);
 919                         }
 920                 }
 921
 922                 if (t) {
 923                         err = 0;
 924                         if (cmd == SIOCCHGTUNNEL) {
 925                                 t->parms.iph.ttl = p.iph.ttl;
 926                                 t->parms.iph.tos = p.iph.tos;
 927                                 if (t->parms.link != p.link) {
 928                                         t->parms.link = p.link;
 929                                         ipip6_tunnel_bind_dev(dev);
 930                                         netdev_state_change(dev);
 931                                 }
 932                         }
 933                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 934                                 err = -EFAULT;
 935                 } else
 936                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 937                 break;
 938
 939         case SIOCDELTUNNEL:
 940                 err = -EPERM;
 941                 if (!capable(CAP_NET_ADMIN))
 942                         goto done;
 943
 944                 if (dev == sitn->fb_tunnel_dev) {
 945                         err = -EFAULT;
 946                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 947                                 goto done;
 948                         err = -ENOENT;
 949                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
 950                                 goto done;
 951                         err = -EPERM;
 952                         if (t == netdev_priv(sitn->fb_tunnel_dev))
 953                                 goto done;
 954                         dev = t->dev;
 955                 }
 956                 unregister_netdevice(dev);
 957                 err = 0;
 958                 break;
 959
 960         case SIOCGETPRL:
 961         case SIOCADDPRL:
 962         case SIOCDELPRL:
 963         case SIOCCHGPRL:
 964                 err = -EPERM;
 965                 if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
 966                         goto done;
 967                 err = -EINVAL;
 968                 if (dev == sitn->fb_tunnel_dev)
 969                         goto done;
 970                 err = -EFAULT;
 971                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 972                         goto done;
 973                 err = -ENOENT;
 974                 if (!(t = netdev_priv(dev)))
 975                         goto done;
 976
 977                 switch (cmd) {
 978                 case SIOCGETPRL:
 979                         err = ipip6_tunnel_get_prl(t, &prl);
 980                         if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
 981                                                  &prl, sizeof(prl)))
 982                                 err = -EFAULT;
 983                         break;
 984                 case SIOCDELPRL:
 985                         err = ipip6_tunnel_del_prl(t, &prl);
 986                         break;
 987                 case SIOCADDPRL:
 988                 case SIOCCHGPRL:
 989                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 990                         break;
 991                 }
 992                 if (cmd != SIOCGETPRL)
 993                         netdev_state_change(dev);
 994                 break;
 995
 996         default:
 997                 err = -EINVAL;
 998         }
 999
1000 done:
1001         return err;
1002 }
1003
1004 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
1005 {
1006         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
1007 }
1008
1009 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1010 {
1011         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1012                 return -EINVAL;
1013         dev->mtu = new_mtu;
1014         return 0;
1015 }
1016
1017 static void ipip6_tunnel_setup(struct net_device *dev)
1018 {
1019         dev->uninit             = ipip6_tunnel_uninit;
1020         dev->destructor         = free_netdev;
1021         dev->hard_start_xmit    = ipip6_tunnel_xmit;
1022         dev->get_stats          = ipip6_tunnel_get_stats;
1023         dev->do_ioctl           = ipip6_tunnel_ioctl;
1024         dev->change_mtu         = ipip6_tunnel_change_mtu;
1025
1026         dev->type               = ARPHRD_SIT;
1027         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1028         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1029         dev->flags              = IFF_NOARP;
1030         dev->iflink             = 0;
1031         dev->addr_len           = 4;
1032 }
1033
1034 static int ipip6_tunnel_init(struct net_device *dev)
1035 {
1036         struct ip_tunnel *tunnel;
1037
1038         tunnel = netdev_priv(dev);
1039
1040         tunnel->dev = dev;
1041         strcpy(tunnel->parms.name, dev->name);
1042
1043         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1044         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1045
1046         ipip6_tunnel_bind_dev(dev);
1047
1048         return 0;
1049 }
1050
1051 static int ipip6_fb_tunnel_init(struct net_device *dev)
1052 {
1053         struct ip_tunnel *tunnel = netdev_priv(dev);
1054         struct iphdr *iph = &tunnel->parms.iph;
1055         struct net *net = dev_net(dev);
1056         struct sit_net *sitn = net_generic(net, sit_net_id);
1057
1058         tunnel->dev = dev;
1059         strcpy(tunnel->parms.name, dev->name);
1060
1061         iph->version            = 4;
1062         iph->protocol           = IPPROTO_IPV6;
1063         iph->ihl                = 5;
1064         iph->ttl                = 64;
1065
1066         dev_hold(dev);
1067         sitn->tunnels_wc[0]     = tunnel;
1068         return 0;
1069 }
1070
1071 static struct xfrm_tunnel sit_handler = {
1072         .handler        =       ipip6_rcv,
1073         .err_handler    =       ipip6_err,
1074         .priority       =       1,
1075 };
1076
1077 static void sit_destroy_tunnels(struct sit_net *sitn)
1078 {
1079         int prio;
1080
1081         for (prio = 1; prio < 4; prio++) {
1082                 int h;
1083                 for (h = 0; h < HASH_SIZE; h++) {
1084                         struct ip_tunnel *t;
1085                         while ((t = sitn->tunnels[prio][h]) != NULL)
1086                                 unregister_netdevice(t->dev);
1087                 }
1088         }
1089 }
1090
1091 static int sit_init_net(struct net *net)
1092 {
1093         int err;
1094         struct sit_net *sitn;
1095
1096         err = -ENOMEM;
1097         sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1098         if (sitn == NULL)
1099                 goto err_alloc;
1100
1101         err = net_assign_generic(net, sit_net_id, sitn);
1102         if (err < 0)
1103                 goto err_assign;
1104
1105         sitn->tunnels[0] = sitn->tunnels_wc;
1106         sitn->tunnels[1] = sitn->tunnels_l;
1107         sitn->tunnels[2] = sitn->tunnels_r;
1108         sitn->tunnels[3] = sitn->tunnels_r_l;
1109
1110         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1111                                            ipip6_tunnel_setup);
1112         if (!sitn->fb_tunnel_dev) {
1113                 err = -ENOMEM;
1114                 goto err_alloc_dev;
1115         }
1116
1117         sitn->fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1118         dev_net_set(sitn->fb_tunnel_dev, net);
1119
1120         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1121                 goto err_reg_dev;
1122
1123         return 0;
1124
1125 err_reg_dev:
1126         free_netdev(sitn->fb_tunnel_dev);
1127 err_alloc_dev:
1128         /* nothing */
1129 err_assign:
1130         kfree(sitn);
1131 err_alloc:
1132         return err;
1133 }
1134
1135 static void sit_exit_net(struct net *net)
1136 {
1137         struct sit_net *sitn;
1138
1139         sitn = net_generic(net, sit_net_id);
1140         rtnl_lock();
1141         sit_destroy_tunnels(sitn);
1142         unregister_netdevice(sitn->fb_tunnel_dev);
1143         rtnl_unlock();
1144         kfree(sitn);
1145 }
1146
1147 static struct pernet_operations sit_net_ops = {
1148         .init = sit_init_net,
1149         .exit = sit_exit_net,
1150 };
1151
1152 static void __exit sit_cleanup(void)
1153 {
1154         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1155
1156         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1157 }
1158
1159 static int __init sit_init(void)
1160 {
1161         int err;
1162
1163         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1164
1165         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1166                 printk(KERN_INFO "sit init: Can't add protocol\n");
1167                 return -EAGAIN;
1168         }
1169
1170         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1171         if (err < 0)
1172                 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1173
1174         return err;
1175 }
1176
1177 module_init(sit_init);
1178 module_exit(sit_cleanup);
1179 MODULE_LICENSE("GPL");
1180 MODULE_ALIAS("sit0");