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
  56 /*
  57    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  58
  59    For comments look at net/ipv4/ip_gre.c --ANK
  60  */
  61
  62 #define HASH_SIZE  16
  63 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  64
  65 static int ipip6_fb_tunnel_init(struct net_device *dev);
  66 static int ipip6_tunnel_init(struct net_device *dev);
  67 static void ipip6_tunnel_setup(struct net_device *dev);
  68
  69 static struct net_device *ipip6_fb_tunnel_dev;
  70
  71 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  72 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  73 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  74 static struct ip_tunnel *tunnels_wc[1];
  75 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  76
  77 static DEFINE_RWLOCK(ipip6_lock);
  78
  79 static struct ip_tunnel * ipip6_tunnel_lookup(__be32 remote, __be32 local)
  80 {
  81         unsigned h0 = HASH(remote);
  82         unsigned h1 = HASH(local);
  83         struct ip_tunnel *t;
  84
  85         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  86                 if (local == t->parms.iph.saddr &&
  87                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  88                         return t;
  89         }
  90         for (t = tunnels_r[h0]; t; t = t->next) {
  91                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  92                         return t;
  93         }
  94         for (t = tunnels_l[h1]; t; t = t->next) {
  95                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
  96                         return t;
  97         }
  98         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
  99                 return t;
 100         return NULL;
 101 }
 102
 103 static struct ip_tunnel **__ipip6_bucket(struct ip_tunnel_parm *parms)
 104 {
 105         __be32 remote = parms->iph.daddr;
 106         __be32 local = parms->iph.saddr;
 107         unsigned h = 0;
 108         int prio = 0;
 109
 110         if (remote) {
 111                 prio |= 2;
 112                 h ^= HASH(remote);
 113         }
 114         if (local) {
 115                 prio |= 1;
 116                 h ^= HASH(local);
 117         }
 118         return &tunnels[prio][h];
 119 }
 120
 121 static inline struct ip_tunnel **ipip6_bucket(struct ip_tunnel *t)
 122 {
 123         return __ipip6_bucket(&t->parms);
 124 }
 125
 126 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 127 {
 128         struct ip_tunnel **tp;
 129
 130         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 131                 if (t == *tp) {
 132                         write_lock_bh(&ipip6_lock);
 133                         *tp = t->next;
 134                         write_unlock_bh(&ipip6_lock);
 135                         break;
 136                 }
 137         }
 138 }
 139
 140 static void ipip6_tunnel_link(struct ip_tunnel *t)
 141 {
 142         struct ip_tunnel **tp = ipip6_bucket(t);
 143
 144         t->next = *tp;
 145         write_lock_bh(&ipip6_lock);
 146         *tp = t;
 147         write_unlock_bh(&ipip6_lock);
 148 }
 149
 150 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 151 {
 152         __be32 remote = parms->iph.daddr;
 153         __be32 local = parms->iph.saddr;
 154         struct ip_tunnel *t, **tp, *nt;
 155         struct net_device *dev;
 156         char name[IFNAMSIZ];
 157
 158         for (tp = __ipip6_bucket(parms); (t = *tp) != NULL; tp = &t->next) {
 159                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 160                         return t;
 161         }
 162         if (!create)
 163                 goto failed;
 164
 165         if (parms->name[0])
 166                 strlcpy(name, parms->name, IFNAMSIZ);
 167         else
 168                 sprintf(name, "sit%%d");
 169
 170         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 171         if (dev == NULL)
 172                 return NULL;
 173
 174         if (strchr(name, '%')) {
 175                 if (dev_alloc_name(dev, name) < 0)
 176                         goto failed_free;
 177         }
 178
 179         nt = netdev_priv(dev);
 180         dev->init = ipip6_tunnel_init;
 181         nt->parms = *parms;
 182
 183         if (parms->i_flags & SIT_ISATAP)
 184                 dev->priv_flags |= IFF_ISATAP;
 185
 186         if (register_netdevice(dev) < 0)
 187                 goto failed_free;
 188
 189         dev_hold(dev);
 190
 191         ipip6_tunnel_link(nt);
 192         return nt;
 193
 194 failed_free:
 195         free_netdev(dev);
 196 failed:
 197         return NULL;
 198 }
 199
 200 static struct ip_tunnel_prl_entry *
 201 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 202 {
 203         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 204
 205         for (p = t->prl; p; p = p->next)
 206                 if (p->addr == addr)
 207                         break;
 208         return p;
 209
 210 }
 211
 212 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 213 {
 214         struct ip_tunnel_prl *kp;
 215         struct ip_tunnel_prl_entry *prl;
 216         unsigned int cmax, c = 0, ca, len;
 217         int ret = 0;
 218
 219         cmax = a->datalen / sizeof(*a);
 220         if (cmax > 1 && a->addr != htonl(INADDR_ANY))
 221                 cmax = 1;
 222
 223         /* For simple GET or for root users,
 224          * we try harder to allocate.
 225          */
 226         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 227                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 228                 NULL;
 229
 230         read_lock(&ipip6_lock);
 231
 232         ca = t->prl_count < cmax ? t->prl_count : cmax;
 233
 234         if (!kp) {
 235                 /* We don't try hard to allocate much memory for
 236                  * non-root users.
 237                  * For root users, retry allocating enough memory for
 238                  * the answer.
 239                  */
 240                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 241                 if (!kp) {
 242                         ret = -ENOMEM;
 243                         goto out;
 244                 }
 245         }
 246
 247         c = 0;
 248         for (prl = t->prl; prl; prl = prl->next) {
 249                 if (c > cmax)
 250                         break;
 251                 if (a->addr != htonl(INADDR_ANY) && prl->addr != a->addr)
 252                         continue;
 253                 kp[c].addr = prl->addr;
 254                 kp[c].flags = prl->flags;
 255                 c++;
 256                 if (a->addr != htonl(INADDR_ANY))
 257                         break;
 258         }
 259 out:
 260         read_unlock(&ipip6_lock);
 261
 262         len = sizeof(*kp) * c;
 263         ret = len ? copy_to_user(a->data, kp, len) : 0;
 264
 265         kfree(kp);
 266         if (ret)
 267                 return -EFAULT;
 268
 269         a->datalen = len;
 270         return 0;
 271 }
 272
 273 static int
 274 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 275 {
 276         struct ip_tunnel_prl_entry *p;
 277         int err = 0;
 278
 279         if (a->addr == htonl(INADDR_ANY))
 280                 return -EINVAL;
 281
 282         write_lock(&ipip6_lock);
 283
 284         for (p = t->prl; p; p = p->next) {
 285                 if (p->addr == a->addr) {
 286                         if (chg)
 287                                 goto update;
 288                         err = -EEXIST;
 289                         goto out;
 290                 }
 291         }
 292
 293         if (chg) {
 294                 err = -ENXIO;
 295                 goto out;
 296         }
 297
 298         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 299         if (!p) {
 300                 err = -ENOBUFS;
 301                 goto out;
 302         }
 303
 304         p->next = t->prl;
 305         t->prl = p;
 306         t->prl_count++;
 307 update:
 308         p->addr = a->addr;
 309         p->flags = a->flags;
 310 out:
 311         write_unlock(&ipip6_lock);
 312         return err;
 313 }
 314
 315 static int
 316 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 317 {
 318         struct ip_tunnel_prl_entry *x, **p;
 319         int err = 0;
 320
 321         write_lock(&ipip6_lock);
 322
 323         if (a && a->addr != htonl(INADDR_ANY)) {
 324                 for (p = &t->prl; *p; p = &(*p)->next) {
 325                         if ((*p)->addr == a->addr) {
 326                                 x = *p;
 327                                 *p = x->next;
 328                                 kfree(x);
 329                                 t->prl_count--;
 330                                 goto out;
 331                         }
 332                 }
 333                 err = -ENXIO;
 334         } else {
 335                 while (t->prl) {
 336                         x = t->prl;
 337                         t->prl = t->prl->next;
 338                         kfree(x);
 339                         t->prl_count--;
 340                 }
 341         }
 342 out:
 343         write_unlock(&ipip6_lock);
 344         return 0;
 345 }
 346
 347 static int
 348 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 349 {
 350         struct ip_tunnel_prl_entry *p;
 351         int ok = 1;
 352
 353         read_lock(&ipip6_lock);
 354         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 355         if (p) {
 356                 if (p->flags & PRL_DEFAULT)
 357                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 358                 else
 359                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 360         } else {
 361                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 362                 if (ipv6_addr_is_isatap(addr6) &&
 363                     (addr6->s6_addr32[3] == iph->saddr) &&
 364                     ipv6_chk_prefix(addr6, t->dev))
 365                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 366                 else
 367                         ok = 0;
 368         }
 369         read_unlock(&ipip6_lock);
 370         return ok;
 371 }
 372
 373 static void ipip6_tunnel_uninit(struct net_device *dev)
 374 {
 375         if (dev == ipip6_fb_tunnel_dev) {
 376                 write_lock_bh(&ipip6_lock);
 377                 tunnels_wc[0] = NULL;
 378                 write_unlock_bh(&ipip6_lock);
 379                 dev_put(dev);
 380         } else {
 381                 ipip6_tunnel_unlink(netdev_priv(dev));
 382                 ipip6_tunnel_del_prl(netdev_priv(dev), 0);
 383                 dev_put(dev);
 384         }
 385 }
 386
 387
 388 static int ipip6_err(struct sk_buff *skb, u32 info)
 389 {
 390 #ifndef I_WISH_WORLD_WERE_PERFECT
 391
 392 /* It is not :-( All the routers (except for Linux) return only
 393    8 bytes of packet payload. It means, that precise relaying of
 394    ICMP in the real Internet is absolutely infeasible.
 395  */
 396         struct iphdr *iph = (struct iphdr*)skb->data;
 397         const int type = icmp_hdr(skb)->type;
 398         const int code = icmp_hdr(skb)->code;
 399         struct ip_tunnel *t;
 400         int err;
 401
 402         switch (type) {
 403         default:
 404         case ICMP_PARAMETERPROB:
 405                 return 0;
 406
 407         case ICMP_DEST_UNREACH:
 408                 switch (code) {
 409                 case ICMP_SR_FAILED:
 410                 case ICMP_PORT_UNREACH:
 411                         /* Impossible event. */
 412                         return 0;
 413                 case ICMP_FRAG_NEEDED:
 414                         /* Soft state for pmtu is maintained by IP core. */
 415                         return 0;
 416                 default:
 417                         /* All others are translated to HOST_UNREACH.
 418                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 419                            I believe they are just ether pollution. --ANK
 420                          */
 421                         break;
 422                 }
 423                 break;
 424         case ICMP_TIME_EXCEEDED:
 425                 if (code != ICMP_EXC_TTL)
 426                         return 0;
 427                 break;
 428         }
 429
 430         err = -ENOENT;
 431
 432         read_lock(&ipip6_lock);
 433         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 434         if (t == NULL || t->parms.iph.daddr == 0)
 435                 goto out;
 436
 437         err = 0;
 438         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 439                 goto out;
 440
 441         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 442                 t->err_count++;
 443         else
 444                 t->err_count = 1;
 445         t->err_time = jiffies;
 446 out:
 447         read_unlock(&ipip6_lock);
 448         return err;
 449 #else
 450         struct iphdr *iph = (struct iphdr*)dp;
 451         int hlen = iph->ihl<<2;
 452         struct ipv6hdr *iph6;
 453         const int type = icmp_hdr(skb)->type;
 454         const int code = icmp_hdr(skb)->code;
 455         int rel_type = 0;
 456         int rel_code = 0;
 457         int rel_info = 0;
 458         struct sk_buff *skb2;
 459         struct rt6_info *rt6i;
 460
 461         if (len < hlen + sizeof(struct ipv6hdr))
 462                 return;
 463         iph6 = (struct ipv6hdr*)(dp + hlen);
 464
 465         switch (type) {
 466         default:
 467                 return;
 468         case ICMP_PARAMETERPROB:
 469                 if (icmp_hdr(skb)->un.gateway < hlen)
 470                         return;
 471
 472                 /* So... This guy found something strange INSIDE encapsulated
 473                    packet. Well, he is fool, but what can we do ?
 474                  */
 475                 rel_type = ICMPV6_PARAMPROB;
 476                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 477                 break;
 478
 479         case ICMP_DEST_UNREACH:
 480                 switch (code) {
 481                 case ICMP_SR_FAILED:
 482                 case ICMP_PORT_UNREACH:
 483                         /* Impossible event. */
 484                         return;
 485                 case ICMP_FRAG_NEEDED:
 486                         /* Too complicated case ... */
 487                         return;
 488                 default:
 489                         /* All others are translated to HOST_UNREACH.
 490                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 491                            I believe, it is just ether pollution. --ANK
 492                          */
 493                         rel_type = ICMPV6_DEST_UNREACH;
 494                         rel_code = ICMPV6_ADDR_UNREACH;
 495                         break;
 496                 }
 497                 break;
 498         case ICMP_TIME_EXCEEDED:
 499                 if (code != ICMP_EXC_TTL)
 500                         return;
 501                 rel_type = ICMPV6_TIME_EXCEED;
 502                 rel_code = ICMPV6_EXC_HOPLIMIT;
 503                 break;
 504         }
 505
 506         /* Prepare fake skb to feed it to icmpv6_send */
 507         skb2 = skb_clone(skb, GFP_ATOMIC);
 508         if (skb2 == NULL)
 509                 return 0;
 510         dst_release(skb2->dst);
 511         skb2->dst = NULL;
 512         skb_pull(skb2, skb->data - (u8*)iph6);
 513         skb_reset_network_header(skb2);
 514
 515         /* Try to guess incoming interface */
 516         rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
 517         if (rt6i && rt6i->rt6i_dev) {
 518                 skb2->dev = rt6i->rt6i_dev;
 519
 520                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 521
 522                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 523                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 524                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 525                                 rel_type = ICMPV6_DEST_UNREACH;
 526                                 rel_code = ICMPV6_ADDR_UNREACH;
 527                         }
 528                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 529                 }
 530         }
 531         kfree_skb(skb2);
 532         return 0;
 533 #endif
 534 }
 535
 536 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 537 {
 538         if (INET_ECN_is_ce(iph->tos))
 539                 IP6_ECN_set_ce(ipv6_hdr(skb));
 540 }
 541
 542 static int ipip6_rcv(struct sk_buff *skb)
 543 {
 544         struct iphdr *iph;
 545         struct ip_tunnel *tunnel;
 546
 547         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 548                 goto out;
 549
 550         iph = ip_hdr(skb);
 551
 552         read_lock(&ipip6_lock);
 553         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 554                 secpath_reset(skb);
 555                 skb->mac_header = skb->network_header;
 556                 skb_reset_network_header(skb);
 557                 IPCB(skb)->flags = 0;
 558                 skb->protocol = htons(ETH_P_IPV6);
 559                 skb->pkt_type = PACKET_HOST;
 560
 561                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 562                     !isatap_chksrc(skb, iph, tunnel)) {
 563                         tunnel->stat.rx_errors++;
 564                         read_unlock(&ipip6_lock);
 565                         kfree_skb(skb);
 566                         return 0;
 567                 }
 568                 tunnel->stat.rx_packets++;
 569                 tunnel->stat.rx_bytes += skb->len;
 570                 skb->dev = tunnel->dev;
 571                 dst_release(skb->dst);
 572                 skb->dst = NULL;
 573                 nf_reset(skb);
 574                 ipip6_ecn_decapsulate(iph, skb);
 575                 netif_rx(skb);
 576                 read_unlock(&ipip6_lock);
 577                 return 0;
 578         }
 579
 580         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 581         kfree_skb(skb);
 582         read_unlock(&ipip6_lock);
 583 out:
 584         return 0;
 585 }
 586
 587 /* Returns the embedded IPv4 address if the IPv6 address
 588    comes from 6to4 (RFC 3056) addr space */
 589
 590 static inline __be32 try_6to4(struct in6_addr *v6dst)
 591 {
 592         __be32 dst = 0;
 593
 594         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 595                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 596                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 597         }
 598         return dst;
 599 }
 600
 601 /*
 602  *      This function assumes it is being called from dev_queue_xmit()
 603  *      and that skb is filled properly by that function.
 604  */
 605
 606 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 607 {
 608         struct ip_tunnel *tunnel = netdev_priv(dev);
 609         struct net_device_stats *stats = &tunnel->stat;
 610         struct iphdr  *tiph = &tunnel->parms.iph;
 611         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 612         u8     tos = tunnel->parms.iph.tos;
 613         struct rtable *rt;                      /* Route to the other host */
 614         struct net_device *tdev;                        /* Device to other host */
 615         struct iphdr  *iph;                     /* Our new IP header */
 616         unsigned int max_headroom;              /* The extra header space needed */
 617         __be32 dst = tiph->daddr;
 618         int    mtu;
 619         struct in6_addr *addr6;
 620         int addr_type;
 621
 622         if (tunnel->recursion++) {
 623                 tunnel->stat.collisions++;
 624                 goto tx_error;
 625         }
 626
 627         if (skb->protocol != htons(ETH_P_IPV6))
 628                 goto tx_error;
 629
 630         /* ISATAP (RFC4214) - must come before 6to4 */
 631         if (dev->priv_flags & IFF_ISATAP) {
 632                 struct neighbour *neigh = NULL;
 633
 634                 if (skb->dst)
 635                         neigh = skb->dst->neighbour;
 636
 637                 if (neigh == NULL) {
 638                         if (net_ratelimit())
 639                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 640                         goto tx_error;
 641                 }
 642
 643                 addr6 = (struct in6_addr*)&neigh->primary_key;
 644                 addr_type = ipv6_addr_type(addr6);
 645
 646                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 647                      ipv6_addr_is_isatap(addr6))
 648                         dst = addr6->s6_addr32[3];
 649                 else
 650                         goto tx_error;
 651         }
 652
 653         if (!dst)
 654                 dst = try_6to4(&iph6->daddr);
 655
 656         if (!dst) {
 657                 struct neighbour *neigh = NULL;
 658
 659                 if (skb->dst)
 660                         neigh = skb->dst->neighbour;
 661
 662                 if (neigh == NULL) {
 663                         if (net_ratelimit())
 664                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 665                         goto tx_error;
 666                 }
 667
 668                 addr6 = (struct in6_addr*)&neigh->primary_key;
 669                 addr_type = ipv6_addr_type(addr6);
 670
 671                 if (addr_type == IPV6_ADDR_ANY) {
 672                         addr6 = &ipv6_hdr(skb)->daddr;
 673                         addr_type = ipv6_addr_type(addr6);
 674                 }
 675
 676                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 677                         goto tx_error_icmp;
 678
 679                 dst = addr6->s6_addr32[3];
 680         }
 681
 682         {
 683                 struct flowi fl = { .nl_u = { .ip4_u =
 684                                               { .daddr = dst,
 685                                                 .saddr = tiph->saddr,
 686                                                 .tos = RT_TOS(tos) } },
 687                                     .oif = tunnel->parms.link,
 688                                     .proto = IPPROTO_IPV6 };
 689                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 690                         tunnel->stat.tx_carrier_errors++;
 691                         goto tx_error_icmp;
 692                 }
 693         }
 694         if (rt->rt_type != RTN_UNICAST) {
 695                 ip_rt_put(rt);
 696                 tunnel->stat.tx_carrier_errors++;
 697                 goto tx_error_icmp;
 698         }
 699         tdev = rt->u.dst.dev;
 700
 701         if (tdev == dev) {
 702                 ip_rt_put(rt);
 703                 tunnel->stat.collisions++;
 704                 goto tx_error;
 705         }
 706
 707         if (tiph->frag_off)
 708                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 709         else
 710                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 711
 712         if (mtu < 68) {
 713                 tunnel->stat.collisions++;
 714                 ip_rt_put(rt);
 715                 goto tx_error;
 716         }
 717         if (mtu < IPV6_MIN_MTU)
 718                 mtu = IPV6_MIN_MTU;
 719         if (tunnel->parms.iph.daddr && skb->dst)
 720                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 721
 722         if (skb->len > mtu) {
 723                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 724                 ip_rt_put(rt);
 725                 goto tx_error;
 726         }
 727
 728         if (tunnel->err_count > 0) {
 729                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 730                         tunnel->err_count--;
 731                         dst_link_failure(skb);
 732                 } else
 733                         tunnel->err_count = 0;
 734         }
 735
 736         /*
 737          * Okay, now see if we can stuff it in the buffer as-is.
 738          */
 739         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 740
 741         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 742             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 743                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 744                 if (!new_skb) {
 745                         ip_rt_put(rt);
 746                         stats->tx_dropped++;
 747                         dev_kfree_skb(skb);
 748                         tunnel->recursion--;
 749                         return 0;
 750                 }
 751                 if (skb->sk)
 752                         skb_set_owner_w(new_skb, skb->sk);
 753                 dev_kfree_skb(skb);
 754                 skb = new_skb;
 755                 iph6 = ipv6_hdr(skb);
 756         }
 757
 758         skb->transport_header = skb->network_header;
 759         skb_push(skb, sizeof(struct iphdr));
 760         skb_reset_network_header(skb);
 761         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 762         IPCB(skb)->flags = 0;
 763         dst_release(skb->dst);
 764         skb->dst = &rt->u.dst;
 765
 766         /*
 767          *      Push down and install the IPIP header.
 768          */
 769
 770         iph                     =       ip_hdr(skb);
 771         iph->version            =       4;
 772         iph->ihl                =       sizeof(struct iphdr)>>2;
 773         if (mtu > IPV6_MIN_MTU)
 774                 iph->frag_off   =       htons(IP_DF);
 775         else
 776                 iph->frag_off   =       0;
 777
 778         iph->protocol           =       IPPROTO_IPV6;
 779         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 780         iph->daddr              =       rt->rt_dst;
 781         iph->saddr              =       rt->rt_src;
 782
 783         if ((iph->ttl = tiph->ttl) == 0)
 784                 iph->ttl        =       iph6->hop_limit;
 785
 786         nf_reset(skb);
 787
 788         IPTUNNEL_XMIT();
 789         tunnel->recursion--;
 790         return 0;
 791
 792 tx_error_icmp:
 793         dst_link_failure(skb);
 794 tx_error:
 795         stats->tx_errors++;
 796         dev_kfree_skb(skb);
 797         tunnel->recursion--;
 798         return 0;
 799 }
 800
 801 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 802 {
 803         struct net_device *tdev = NULL;
 804         struct ip_tunnel *tunnel;
 805         struct iphdr *iph;
 806
 807         tunnel = netdev_priv(dev);
 808         iph = &tunnel->parms.iph;
 809
 810         if (iph->daddr) {
 811                 struct flowi fl = { .nl_u = { .ip4_u =
 812                                               { .daddr = iph->daddr,
 813                                                 .saddr = iph->saddr,
 814                                                 .tos = RT_TOS(iph->tos) } },
 815                                     .oif = tunnel->parms.link,
 816                                     .proto = IPPROTO_IPV6 };
 817                 struct rtable *rt;
 818                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 819                         tdev = rt->u.dst.dev;
 820                         ip_rt_put(rt);
 821                 }
 822                 dev->flags |= IFF_POINTOPOINT;
 823         }
 824
 825         if (!tdev && tunnel->parms.link)
 826                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 827
 828         if (tdev) {
 829                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 830                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 831                 if (dev->mtu < IPV6_MIN_MTU)
 832                         dev->mtu = IPV6_MIN_MTU;
 833         }
 834         dev->iflink = tunnel->parms.link;
 835 }
 836
 837 static int
 838 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 839 {
 840         int err = 0;
 841         struct ip_tunnel_parm p;
 842         struct ip_tunnel_prl prl;
 843         struct ip_tunnel *t;
 844
 845         switch (cmd) {
 846         case SIOCGETTUNNEL:
 847                 t = NULL;
 848                 if (dev == ipip6_fb_tunnel_dev) {
 849                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 850                                 err = -EFAULT;
 851                                 break;
 852                         }
 853                         t = ipip6_tunnel_locate(&p, 0);
 854                 }
 855                 if (t == NULL)
 856                         t = netdev_priv(dev);
 857                 memcpy(&p, &t->parms, sizeof(p));
 858                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 859                         err = -EFAULT;
 860                 break;
 861
 862         case SIOCADDTUNNEL:
 863         case SIOCCHGTUNNEL:
 864                 err = -EPERM;
 865                 if (!capable(CAP_NET_ADMIN))
 866                         goto done;
 867
 868                 err = -EFAULT;
 869                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 870                         goto done;
 871
 872                 err = -EINVAL;
 873                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 874                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 875                         goto done;
 876                 if (p.iph.ttl)
 877                         p.iph.frag_off |= htons(IP_DF);
 878
 879                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 880
 881                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 882                         if (t != NULL) {
 883                                 if (t->dev != dev) {
 884                                         err = -EEXIST;
 885                                         break;
 886                                 }
 887                         } else {
 888                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 889                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 890                                         err = -EINVAL;
 891                                         break;
 892                                 }
 893                                 t = netdev_priv(dev);
 894                                 ipip6_tunnel_unlink(t);
 895                                 t->parms.iph.saddr = p.iph.saddr;
 896                                 t->parms.iph.daddr = p.iph.daddr;
 897                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 898                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 899                                 ipip6_tunnel_link(t);
 900                                 netdev_state_change(dev);
 901                         }
 902                 }
 903
 904                 if (t) {
 905                         err = 0;
 906                         if (cmd == SIOCCHGTUNNEL) {
 907                                 t->parms.iph.ttl = p.iph.ttl;
 908                                 t->parms.iph.tos = p.iph.tos;
 909                                 if (t->parms.link != p.link) {
 910                                         t->parms.link = p.link;
 911                                         ipip6_tunnel_bind_dev(dev);
 912                                         netdev_state_change(dev);
 913                                 }
 914                         }
 915                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 916                                 err = -EFAULT;
 917                 } else
 918                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 919                 break;
 920
 921         case SIOCDELTUNNEL:
 922                 err = -EPERM;
 923                 if (!capable(CAP_NET_ADMIN))
 924                         goto done;
 925
 926                 if (dev == ipip6_fb_tunnel_dev) {
 927                         err = -EFAULT;
 928                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 929                                 goto done;
 930                         err = -ENOENT;
 931                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 932                                 goto done;
 933                         err = -EPERM;
 934                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 935                                 goto done;
 936                         dev = t->dev;
 937                 }
 938                 unregister_netdevice(dev);
 939                 err = 0;
 940                 break;
 941
 942         case SIOCGETPRL:
 943         case SIOCADDPRL:
 944         case SIOCDELPRL:
 945         case SIOCCHGPRL:
 946                 err = -EPERM;
 947                 if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
 948                         goto done;
 949                 err = -EINVAL;
 950                 if (dev == ipip6_fb_tunnel_dev)
 951                         goto done;
 952                 err = -EFAULT;
 953                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 954                         goto done;
 955                 err = -ENOENT;
 956                 if (!(t = netdev_priv(dev)))
 957                         goto done;
 958
 959                 switch (cmd) {
 960                 case SIOCGETPRL:
 961                         err = ipip6_tunnel_get_prl(t, &prl);
 962                         if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
 963                                                  &prl, sizeof(prl)))
 964                                 err = -EFAULT;
 965                         break;
 966                 case SIOCDELPRL:
 967                         err = ipip6_tunnel_del_prl(t, &prl);
 968                         break;
 969                 case SIOCADDPRL:
 970                 case SIOCCHGPRL:
 971                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 972                         break;
 973                 }
 974                 if (cmd != SIOCGETPRL)
 975                         netdev_state_change(dev);
 976                 break;
 977
 978         default:
 979                 err = -EINVAL;
 980         }
 981
 982 done:
 983         return err;
 984 }
 985
 986 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 987 {
 988         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 989 }
 990
 991 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 992 {
 993         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 994                 return -EINVAL;
 995         dev->mtu = new_mtu;
 996         return 0;
 997 }
 998
 999 static void ipip6_tunnel_setup(struct net_device *dev)
1000 {
1001         dev->uninit             = ipip6_tunnel_uninit;
1002         dev->destructor         = free_netdev;
1003         dev->hard_start_xmit    = ipip6_tunnel_xmit;
1004         dev->get_stats          = ipip6_tunnel_get_stats;
1005         dev->do_ioctl           = ipip6_tunnel_ioctl;
1006         dev->change_mtu         = ipip6_tunnel_change_mtu;
1007
1008         dev->type               = ARPHRD_SIT;
1009         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1010         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1011         dev->flags              = IFF_NOARP;
1012         dev->iflink             = 0;
1013         dev->addr_len           = 4;
1014 }
1015
1016 static int ipip6_tunnel_init(struct net_device *dev)
1017 {
1018         struct ip_tunnel *tunnel;
1019
1020         tunnel = netdev_priv(dev);
1021
1022         tunnel->dev = dev;
1023         strcpy(tunnel->parms.name, dev->name);
1024
1025         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1026         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1027
1028         ipip6_tunnel_bind_dev(dev);
1029
1030         return 0;
1031 }
1032
1033 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
1034 {
1035         struct ip_tunnel *tunnel = netdev_priv(dev);
1036         struct iphdr *iph = &tunnel->parms.iph;
1037
1038         tunnel->dev = dev;
1039         strcpy(tunnel->parms.name, dev->name);
1040
1041         iph->version            = 4;
1042         iph->protocol           = IPPROTO_IPV6;
1043         iph->ihl                = 5;
1044         iph->ttl                = 64;
1045
1046         dev_hold(dev);
1047         tunnels_wc[0]           = tunnel;
1048         return 0;
1049 }
1050
1051 static struct xfrm_tunnel sit_handler = {
1052         .handler        =       ipip6_rcv,
1053         .err_handler    =       ipip6_err,
1054         .priority       =       1,
1055 };
1056
1057 static void __exit sit_destroy_tunnels(void)
1058 {
1059         int prio;
1060
1061         for (prio = 1; prio < 4; prio++) {
1062                 int h;
1063                 for (h = 0; h < HASH_SIZE; h++) {
1064                         struct ip_tunnel *t;
1065                         while ((t = tunnels[prio][h]) != NULL)
1066                                 unregister_netdevice(t->dev);
1067                 }
1068         }
1069 }
1070
1071 static void __exit sit_cleanup(void)
1072 {
1073         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1074
1075         rtnl_lock();
1076         sit_destroy_tunnels();
1077         unregister_netdevice(ipip6_fb_tunnel_dev);
1078         rtnl_unlock();
1079 }
1080
1081 static int __init sit_init(void)
1082 {
1083         int err;
1084
1085         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1086
1087         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1088                 printk(KERN_INFO "sit init: Can't add protocol\n");
1089                 return -EAGAIN;
1090         }
1091
1092         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1093                                            ipip6_tunnel_setup);
1094         if (!ipip6_fb_tunnel_dev) {
1095                 err = -ENOMEM;
1096                 goto err1;
1097         }
1098
1099         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1100
1101         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
1102                 goto err2;
1103
1104  out:
1105         return err;
1106  err2:
1107         free_netdev(ipip6_fb_tunnel_dev);
1108  err1:
1109         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1110         goto out;
1111 }
1112
1113 module_init(sit_init);
1114 module_exit(sit_cleanup);
1115 MODULE_LICENSE("GPL");
1116 MODULE_ALIAS("sit0");