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(&init_net, &iph6->saddr, NULL, NULL, 0);
 531         if (rt6i && rt6i->rt6i_dev) {
 532                 skb2->dev = rt6i->rt6i_dev;
 533
 534                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 535
 536                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 537                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 538                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 539                                 rel_type = ICMPV6_DEST_UNREACH;
 540                                 rel_code = ICMPV6_ADDR_UNREACH;
 541                         }
 542                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 543                 }
 544         }
 545         kfree_skb(skb2);
 546         return 0;
 547 #endif
 548 }
 549
 550 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 551 {
 552         if (INET_ECN_is_ce(iph->tos))
 553                 IP6_ECN_set_ce(ipv6_hdr(skb));
 554 }
 555
 556 static int ipip6_rcv(struct sk_buff *skb)
 557 {
 558         struct iphdr *iph;
 559         struct ip_tunnel *tunnel;
 560
 561         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 562                 goto out;
 563
 564         iph = ip_hdr(skb);
 565
 566         read_lock(&ipip6_lock);
 567         if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
 568                                         iph->saddr, iph->daddr)) != NULL) {
 569                 secpath_reset(skb);
 570                 skb->mac_header = skb->network_header;
 571                 skb_reset_network_header(skb);
 572                 IPCB(skb)->flags = 0;
 573                 skb->protocol = htons(ETH_P_IPV6);
 574                 skb->pkt_type = PACKET_HOST;
 575
 576                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 577                     !isatap_chksrc(skb, iph, tunnel)) {
 578                         tunnel->stat.rx_errors++;
 579                         read_unlock(&ipip6_lock);
 580                         kfree_skb(skb);
 581                         return 0;
 582                 }
 583                 tunnel->stat.rx_packets++;
 584                 tunnel->stat.rx_bytes += skb->len;
 585                 skb->dev = tunnel->dev;
 586                 dst_release(skb->dst);
 587                 skb->dst = NULL;
 588                 nf_reset(skb);
 589                 ipip6_ecn_decapsulate(iph, skb);
 590                 netif_rx(skb);
 591                 read_unlock(&ipip6_lock);
 592                 return 0;
 593         }
 594
 595         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 596         kfree_skb(skb);
 597         read_unlock(&ipip6_lock);
 598 out:
 599         return 0;
 600 }
 601
 602 /* Returns the embedded IPv4 address if the IPv6 address
 603    comes from 6to4 (RFC 3056) addr space */
 604
 605 static inline __be32 try_6to4(struct in6_addr *v6dst)
 606 {
 607         __be32 dst = 0;
 608
 609         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 610                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 611                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 612         }
 613         return dst;
 614 }
 615
 616 /*
 617  *      This function assumes it is being called from dev_queue_xmit()
 618  *      and that skb is filled properly by that function.
 619  */
 620
 621 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 622 {
 623         struct ip_tunnel *tunnel = netdev_priv(dev);
 624         struct net_device_stats *stats = &tunnel->stat;
 625         struct iphdr  *tiph = &tunnel->parms.iph;
 626         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 627         u8     tos = tunnel->parms.iph.tos;
 628         struct rtable *rt;                      /* Route to the other host */
 629         struct net_device *tdev;                        /* Device to other host */
 630         struct iphdr  *iph;                     /* Our new IP header */
 631         unsigned int max_headroom;              /* The extra header space needed */
 632         __be32 dst = tiph->daddr;
 633         int    mtu;
 634         struct in6_addr *addr6;
 635         int addr_type;
 636
 637         if (tunnel->recursion++) {
 638                 tunnel->stat.collisions++;
 639                 goto tx_error;
 640         }
 641
 642         if (skb->protocol != htons(ETH_P_IPV6))
 643                 goto tx_error;
 644
 645         /* ISATAP (RFC4214) - must come before 6to4 */
 646         if (dev->priv_flags & IFF_ISATAP) {
 647                 struct neighbour *neigh = NULL;
 648
 649                 if (skb->dst)
 650                         neigh = skb->dst->neighbour;
 651
 652                 if (neigh == NULL) {
 653                         if (net_ratelimit())
 654                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 655                         goto tx_error;
 656                 }
 657
 658                 addr6 = (struct in6_addr*)&neigh->primary_key;
 659                 addr_type = ipv6_addr_type(addr6);
 660
 661                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 662                      ipv6_addr_is_isatap(addr6))
 663                         dst = addr6->s6_addr32[3];
 664                 else
 665                         goto tx_error;
 666         }
 667
 668         if (!dst)
 669                 dst = try_6to4(&iph6->daddr);
 670
 671         if (!dst) {
 672                 struct neighbour *neigh = NULL;
 673
 674                 if (skb->dst)
 675                         neigh = skb->dst->neighbour;
 676
 677                 if (neigh == NULL) {
 678                         if (net_ratelimit())
 679                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 680                         goto tx_error;
 681                 }
 682
 683                 addr6 = (struct in6_addr*)&neigh->primary_key;
 684                 addr_type = ipv6_addr_type(addr6);
 685
 686                 if (addr_type == IPV6_ADDR_ANY) {
 687                         addr6 = &ipv6_hdr(skb)->daddr;
 688                         addr_type = ipv6_addr_type(addr6);
 689                 }
 690
 691                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 692                         goto tx_error_icmp;
 693
 694                 dst = addr6->s6_addr32[3];
 695         }
 696
 697         {
 698                 struct flowi fl = { .nl_u = { .ip4_u =
 699                                               { .daddr = dst,
 700                                                 .saddr = tiph->saddr,
 701                                                 .tos = RT_TOS(tos) } },
 702                                     .oif = tunnel->parms.link,
 703                                     .proto = IPPROTO_IPV6 };
 704                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 705                         tunnel->stat.tx_carrier_errors++;
 706                         goto tx_error_icmp;
 707                 }
 708         }
 709         if (rt->rt_type != RTN_UNICAST) {
 710                 ip_rt_put(rt);
 711                 tunnel->stat.tx_carrier_errors++;
 712                 goto tx_error_icmp;
 713         }
 714         tdev = rt->u.dst.dev;
 715
 716         if (tdev == dev) {
 717                 ip_rt_put(rt);
 718                 tunnel->stat.collisions++;
 719                 goto tx_error;
 720         }
 721
 722         if (tiph->frag_off)
 723                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 724         else
 725                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 726
 727         if (mtu < 68) {
 728                 tunnel->stat.collisions++;
 729                 ip_rt_put(rt);
 730                 goto tx_error;
 731         }
 732         if (mtu < IPV6_MIN_MTU)
 733                 mtu = IPV6_MIN_MTU;
 734         if (tunnel->parms.iph.daddr && skb->dst)
 735                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 736
 737         if (skb->len > mtu) {
 738                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 739                 ip_rt_put(rt);
 740                 goto tx_error;
 741         }
 742
 743         if (tunnel->err_count > 0) {
 744                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 745                         tunnel->err_count--;
 746                         dst_link_failure(skb);
 747                 } else
 748                         tunnel->err_count = 0;
 749         }
 750
 751         /*
 752          * Okay, now see if we can stuff it in the buffer as-is.
 753          */
 754         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 755
 756         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 757             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 758                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 759                 if (!new_skb) {
 760                         ip_rt_put(rt);
 761                         stats->tx_dropped++;
 762                         dev_kfree_skb(skb);
 763                         tunnel->recursion--;
 764                         return 0;
 765                 }
 766                 if (skb->sk)
 767                         skb_set_owner_w(new_skb, skb->sk);
 768                 dev_kfree_skb(skb);
 769                 skb = new_skb;
 770                 iph6 = ipv6_hdr(skb);
 771         }
 772
 773         skb->transport_header = skb->network_header;
 774         skb_push(skb, sizeof(struct iphdr));
 775         skb_reset_network_header(skb);
 776         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 777         IPCB(skb)->flags = 0;
 778         dst_release(skb->dst);
 779         skb->dst = &rt->u.dst;
 780
 781         /*
 782          *      Push down and install the IPIP header.
 783          */
 784
 785         iph                     =       ip_hdr(skb);
 786         iph->version            =       4;
 787         iph->ihl                =       sizeof(struct iphdr)>>2;
 788         if (mtu > IPV6_MIN_MTU)
 789                 iph->frag_off   =       htons(IP_DF);
 790         else
 791                 iph->frag_off   =       0;
 792
 793         iph->protocol           =       IPPROTO_IPV6;
 794         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 795         iph->daddr              =       rt->rt_dst;
 796         iph->saddr              =       rt->rt_src;
 797
 798         if ((iph->ttl = tiph->ttl) == 0)
 799                 iph->ttl        =       iph6->hop_limit;
 800
 801         nf_reset(skb);
 802
 803         IPTUNNEL_XMIT();
 804         tunnel->recursion--;
 805         return 0;
 806
 807 tx_error_icmp:
 808         dst_link_failure(skb);
 809 tx_error:
 810         stats->tx_errors++;
 811         dev_kfree_skb(skb);
 812         tunnel->recursion--;
 813         return 0;
 814 }
 815
 816 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 817 {
 818         struct net_device *tdev = NULL;
 819         struct ip_tunnel *tunnel;
 820         struct iphdr *iph;
 821
 822         tunnel = netdev_priv(dev);
 823         iph = &tunnel->parms.iph;
 824
 825         if (iph->daddr) {
 826                 struct flowi fl = { .nl_u = { .ip4_u =
 827                                               { .daddr = iph->daddr,
 828                                                 .saddr = iph->saddr,
 829                                                 .tos = RT_TOS(iph->tos) } },
 830                                     .oif = tunnel->parms.link,
 831                                     .proto = IPPROTO_IPV6 };
 832                 struct rtable *rt;
 833                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 834                         tdev = rt->u.dst.dev;
 835                         ip_rt_put(rt);
 836                 }
 837                 dev->flags |= IFF_POINTOPOINT;
 838         }
 839
 840         if (!tdev && tunnel->parms.link)
 841                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 842
 843         if (tdev) {
 844                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 845                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 846                 if (dev->mtu < IPV6_MIN_MTU)
 847                         dev->mtu = IPV6_MIN_MTU;
 848         }
 849         dev->iflink = tunnel->parms.link;
 850 }
 851
 852 static int
 853 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 854 {
 855         int err = 0;
 856         struct ip_tunnel_parm p;
 857         struct ip_tunnel_prl prl;
 858         struct ip_tunnel *t;
 859         struct net *net = dev_net(dev);
 860         struct sit_net *sitn = net_generic(net, sit_net_id);
 861
 862         switch (cmd) {
 863         case SIOCGETTUNNEL:
 864                 t = NULL;
 865                 if (dev == sitn->fb_tunnel_dev) {
 866                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 867                                 err = -EFAULT;
 868                                 break;
 869                         }
 870                         t = ipip6_tunnel_locate(net, &p, 0);
 871                 }
 872                 if (t == NULL)
 873                         t = netdev_priv(dev);
 874                 memcpy(&p, &t->parms, sizeof(p));
 875                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 876                         err = -EFAULT;
 877                 break;
 878
 879         case SIOCADDTUNNEL:
 880         case SIOCCHGTUNNEL:
 881                 err = -EPERM;
 882                 if (!capable(CAP_NET_ADMIN))
 883                         goto done;
 884
 885                 err = -EFAULT;
 886                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 887                         goto done;
 888
 889                 err = -EINVAL;
 890                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 891                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 892                         goto done;
 893                 if (p.iph.ttl)
 894                         p.iph.frag_off |= htons(IP_DF);
 895
 896                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 897
 898                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 899                         if (t != NULL) {
 900                                 if (t->dev != dev) {
 901                                         err = -EEXIST;
 902                                         break;
 903                                 }
 904                         } else {
 905                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 906                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 907                                         err = -EINVAL;
 908                                         break;
 909                                 }
 910                                 t = netdev_priv(dev);
 911                                 ipip6_tunnel_unlink(sitn, t);
 912                                 t->parms.iph.saddr = p.iph.saddr;
 913                                 t->parms.iph.daddr = p.iph.daddr;
 914                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 915                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 916                                 ipip6_tunnel_link(sitn, t);
 917                                 netdev_state_change(dev);
 918                         }
 919                 }
 920
 921                 if (t) {
 922                         err = 0;
 923                         if (cmd == SIOCCHGTUNNEL) {
 924                                 t->parms.iph.ttl = p.iph.ttl;
 925                                 t->parms.iph.tos = p.iph.tos;
 926                                 if (t->parms.link != p.link) {
 927                                         t->parms.link = p.link;
 928                                         ipip6_tunnel_bind_dev(dev);
 929                                         netdev_state_change(dev);
 930                                 }
 931                         }
 932                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 933                                 err = -EFAULT;
 934                 } else
 935                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 936                 break;
 937
 938         case SIOCDELTUNNEL:
 939                 err = -EPERM;
 940                 if (!capable(CAP_NET_ADMIN))
 941                         goto done;
 942
 943                 if (dev == sitn->fb_tunnel_dev) {
 944                         err = -EFAULT;
 945                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 946                                 goto done;
 947                         err = -ENOENT;
 948                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
 949                                 goto done;
 950                         err = -EPERM;
 951                         if (t == netdev_priv(sitn->fb_tunnel_dev))
 952                                 goto done;
 953                         dev = t->dev;
 954                 }
 955                 unregister_netdevice(dev);
 956                 err = 0;
 957                 break;
 958
 959         case SIOCGETPRL:
 960         case SIOCADDPRL:
 961         case SIOCDELPRL:
 962         case SIOCCHGPRL:
 963                 err = -EPERM;
 964                 if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
 965                         goto done;
 966                 err = -EINVAL;
 967                 if (dev == sitn->fb_tunnel_dev)
 968                         goto done;
 969                 err = -EFAULT;
 970                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 971                         goto done;
 972                 err = -ENOENT;
 973                 if (!(t = netdev_priv(dev)))
 974                         goto done;
 975
 976                 switch (cmd) {
 977                 case SIOCGETPRL:
 978                         err = ipip6_tunnel_get_prl(t, &prl);
 979                         if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
 980                                                  &prl, sizeof(prl)))
 981                                 err = -EFAULT;
 982                         break;
 983                 case SIOCDELPRL:
 984                         err = ipip6_tunnel_del_prl(t, &prl);
 985                         break;
 986                 case SIOCADDPRL:
 987                 case SIOCCHGPRL:
 988                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 989                         break;
 990                 }
 991                 if (cmd != SIOCGETPRL)
 992                         netdev_state_change(dev);
 993                 break;
 994
 995         default:
 996                 err = -EINVAL;
 997         }
 998
 999 done:
1000         return err;
1001 }
1002
1003 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
1004 {
1005         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
1006 }
1007
1008 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1009 {
1010         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1011                 return -EINVAL;
1012         dev->mtu = new_mtu;
1013         return 0;
1014 }
1015
1016 static void ipip6_tunnel_setup(struct net_device *dev)
1017 {
1018         dev->uninit             = ipip6_tunnel_uninit;
1019         dev->destructor         = free_netdev;
1020         dev->hard_start_xmit    = ipip6_tunnel_xmit;
1021         dev->get_stats          = ipip6_tunnel_get_stats;
1022         dev->do_ioctl           = ipip6_tunnel_ioctl;
1023         dev->change_mtu         = ipip6_tunnel_change_mtu;
1024
1025         dev->type               = ARPHRD_SIT;
1026         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1027         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1028         dev->flags              = IFF_NOARP;
1029         dev->iflink             = 0;
1030         dev->addr_len           = 4;
1031 }
1032
1033 static int ipip6_tunnel_init(struct net_device *dev)
1034 {
1035         struct ip_tunnel *tunnel;
1036
1037         tunnel = netdev_priv(dev);
1038
1039         tunnel->dev = dev;
1040         strcpy(tunnel->parms.name, dev->name);
1041
1042         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1043         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1044
1045         ipip6_tunnel_bind_dev(dev);
1046
1047         return 0;
1048 }
1049
1050 static int ipip6_fb_tunnel_init(struct net_device *dev)
1051 {
1052         struct ip_tunnel *tunnel = netdev_priv(dev);
1053         struct iphdr *iph = &tunnel->parms.iph;
1054         struct net *net = dev_net(dev);
1055         struct sit_net *sitn = net_generic(net, sit_net_id);
1056
1057         tunnel->dev = dev;
1058         strcpy(tunnel->parms.name, dev->name);
1059
1060         iph->version            = 4;
1061         iph->protocol           = IPPROTO_IPV6;
1062         iph->ihl                = 5;
1063         iph->ttl                = 64;
1064
1065         dev_hold(dev);
1066         sitn->tunnels_wc[0]     = tunnel;
1067         return 0;
1068 }
1069
1070 static struct xfrm_tunnel sit_handler = {
1071         .handler        =       ipip6_rcv,
1072         .err_handler    =       ipip6_err,
1073         .priority       =       1,
1074 };
1075
1076 static void sit_destroy_tunnels(struct sit_net *sitn)
1077 {
1078         int prio;
1079
1080         for (prio = 1; prio < 4; prio++) {
1081                 int h;
1082                 for (h = 0; h < HASH_SIZE; h++) {
1083                         struct ip_tunnel *t;
1084                         while ((t = sitn->tunnels[prio][h]) != NULL)
1085                                 unregister_netdevice(t->dev);
1086                 }
1087         }
1088 }
1089
1090 static int sit_init_net(struct net *net)
1091 {
1092         int err;
1093         struct sit_net *sitn;
1094
1095         err = -ENOMEM;
1096         sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1097         if (sitn == NULL)
1098                 goto err_alloc;
1099
1100         err = net_assign_generic(net, sit_net_id, sitn);
1101         if (err < 0)
1102                 goto err_assign;
1103
1104         sitn->tunnels[0] = sitn->tunnels_wc;
1105         sitn->tunnels[1] = sitn->tunnels_l;
1106         sitn->tunnels[2] = sitn->tunnels_r;
1107         sitn->tunnels[3] = sitn->tunnels_r_l;
1108
1109         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1110                                            ipip6_tunnel_setup);
1111         if (!sitn->fb_tunnel_dev) {
1112                 err = -ENOMEM;
1113                 goto err_alloc_dev;
1114         }
1115
1116         sitn->fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1117         dev_net_set(sitn->fb_tunnel_dev, net);
1118
1119         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1120                 goto err_reg_dev;
1121
1122         return 0;
1123
1124 err_reg_dev:
1125         free_netdev(sitn->fb_tunnel_dev);
1126 err_alloc_dev:
1127         /* nothing */
1128 err_assign:
1129         kfree(sitn);
1130 err_alloc:
1131         return err;
1132 }
1133
1134 static void sit_exit_net(struct net *net)
1135 {
1136         struct sit_net *sitn;
1137
1138         sitn = net_generic(net, sit_net_id);
1139         rtnl_lock();
1140         sit_destroy_tunnels(sitn);
1141         unregister_netdevice(sitn->fb_tunnel_dev);
1142         rtnl_unlock();
1143         kfree(sitn);
1144 }
1145
1146 static struct pernet_operations sit_net_ops = {
1147         .init = sit_init_net,
1148         .exit = sit_exit_net,
1149 };
1150
1151 static void __exit sit_cleanup(void)
1152 {
1153         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1154
1155         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1156 }
1157
1158 static int __init sit_init(void)
1159 {
1160         int err;
1161
1162         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1163
1164         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1165                 printk(KERN_INFO "sit init: Can't add protocol\n");
1166                 return -EAGAIN;
1167         }
1168
1169         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1170         if (err < 0)
1171                 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1172
1173         return err;
1174 }
1175
1176 module_init(sit_init);
1177 module_exit(sit_cleanup);
1178 MODULE_LICENSE("GPL");
1179 MODULE_ALIAS("sit0");