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 /* copied directly from anycast.c */
 348 static int
 349 ipip6_onlink(struct in6_addr *addr, struct net_device *dev)
 350 {
 351         struct inet6_dev        *idev;
 352         struct inet6_ifaddr     *ifa;
 353         int     onlink;
 354
 355         onlink = 0;
 356         rcu_read_lock();
 357         idev = __in6_dev_get(dev);
 358         if (idev) {
 359                 read_lock_bh(&idev->lock);
 360                 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 361                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
 362                                                    ifa->prefix_len);
 363                         if (onlink)
 364                                 break;
 365                 }
 366                 read_unlock_bh(&idev->lock);
 367         }
 368         rcu_read_unlock();
 369         return onlink;
 370 }
 371
 372 static int
 373 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 374 {
 375         struct ip_tunnel_prl_entry *p;
 376         int ok = 1;
 377
 378         read_lock(&ipip6_lock);
 379         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 380         if (p) {
 381                 if (p->flags & PRL_DEFAULT)
 382                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 383                 else
 384                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 385         } else {
 386                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 387                 if (ipv6_addr_is_isatap(addr6) &&
 388                     (addr6->s6_addr32[3] == iph->saddr) &&
 389                     ipip6_onlink(addr6, t->dev))
 390                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 391                 else
 392                         ok = 0;
 393         }
 394         read_unlock(&ipip6_lock);
 395         return ok;
 396 }
 397
 398 static void ipip6_tunnel_uninit(struct net_device *dev)
 399 {
 400         if (dev == ipip6_fb_tunnel_dev) {
 401                 write_lock_bh(&ipip6_lock);
 402                 tunnels_wc[0] = NULL;
 403                 write_unlock_bh(&ipip6_lock);
 404                 dev_put(dev);
 405         } else {
 406                 ipip6_tunnel_unlink(netdev_priv(dev));
 407                 ipip6_tunnel_del_prl(netdev_priv(dev), 0);
 408                 dev_put(dev);
 409         }
 410 }
 411
 412
 413 static int ipip6_err(struct sk_buff *skb, u32 info)
 414 {
 415 #ifndef I_WISH_WORLD_WERE_PERFECT
 416
 417 /* It is not :-( All the routers (except for Linux) return only
 418    8 bytes of packet payload. It means, that precise relaying of
 419    ICMP in the real Internet is absolutely infeasible.
 420  */
 421         struct iphdr *iph = (struct iphdr*)skb->data;
 422         const int type = icmp_hdr(skb)->type;
 423         const int code = icmp_hdr(skb)->code;
 424         struct ip_tunnel *t;
 425         int err;
 426
 427         switch (type) {
 428         default:
 429         case ICMP_PARAMETERPROB:
 430                 return 0;
 431
 432         case ICMP_DEST_UNREACH:
 433                 switch (code) {
 434                 case ICMP_SR_FAILED:
 435                 case ICMP_PORT_UNREACH:
 436                         /* Impossible event. */
 437                         return 0;
 438                 case ICMP_FRAG_NEEDED:
 439                         /* Soft state for pmtu is maintained by IP core. */
 440                         return 0;
 441                 default:
 442                         /* All others are translated to HOST_UNREACH.
 443                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 444                            I believe they are just ether pollution. --ANK
 445                          */
 446                         break;
 447                 }
 448                 break;
 449         case ICMP_TIME_EXCEEDED:
 450                 if (code != ICMP_EXC_TTL)
 451                         return 0;
 452                 break;
 453         }
 454
 455         err = -ENOENT;
 456
 457         read_lock(&ipip6_lock);
 458         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 459         if (t == NULL || t->parms.iph.daddr == 0)
 460                 goto out;
 461
 462         err = 0;
 463         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 464                 goto out;
 465
 466         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 467                 t->err_count++;
 468         else
 469                 t->err_count = 1;
 470         t->err_time = jiffies;
 471 out:
 472         read_unlock(&ipip6_lock);
 473         return err;
 474 #else
 475         struct iphdr *iph = (struct iphdr*)dp;
 476         int hlen = iph->ihl<<2;
 477         struct ipv6hdr *iph6;
 478         const int type = icmp_hdr(skb)->type;
 479         const int code = icmp_hdr(skb)->code;
 480         int rel_type = 0;
 481         int rel_code = 0;
 482         int rel_info = 0;
 483         struct sk_buff *skb2;
 484         struct rt6_info *rt6i;
 485
 486         if (len < hlen + sizeof(struct ipv6hdr))
 487                 return;
 488         iph6 = (struct ipv6hdr*)(dp + hlen);
 489
 490         switch (type) {
 491         default:
 492                 return;
 493         case ICMP_PARAMETERPROB:
 494                 if (icmp_hdr(skb)->un.gateway < hlen)
 495                         return;
 496
 497                 /* So... This guy found something strange INSIDE encapsulated
 498                    packet. Well, he is fool, but what can we do ?
 499                  */
 500                 rel_type = ICMPV6_PARAMPROB;
 501                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 502                 break;
 503
 504         case ICMP_DEST_UNREACH:
 505                 switch (code) {
 506                 case ICMP_SR_FAILED:
 507                 case ICMP_PORT_UNREACH:
 508                         /* Impossible event. */
 509                         return;
 510                 case ICMP_FRAG_NEEDED:
 511                         /* Too complicated case ... */
 512                         return;
 513                 default:
 514                         /* All others are translated to HOST_UNREACH.
 515                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 516                            I believe, it is just ether pollution. --ANK
 517                          */
 518                         rel_type = ICMPV6_DEST_UNREACH;
 519                         rel_code = ICMPV6_ADDR_UNREACH;
 520                         break;
 521                 }
 522                 break;
 523         case ICMP_TIME_EXCEEDED:
 524                 if (code != ICMP_EXC_TTL)
 525                         return;
 526                 rel_type = ICMPV6_TIME_EXCEED;
 527                 rel_code = ICMPV6_EXC_HOPLIMIT;
 528                 break;
 529         }
 530
 531         /* Prepare fake skb to feed it to icmpv6_send */
 532         skb2 = skb_clone(skb, GFP_ATOMIC);
 533         if (skb2 == NULL)
 534                 return 0;
 535         dst_release(skb2->dst);
 536         skb2->dst = NULL;
 537         skb_pull(skb2, skb->data - (u8*)iph6);
 538         skb_reset_network_header(skb2);
 539
 540         /* Try to guess incoming interface */
 541         rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
 542         if (rt6i && rt6i->rt6i_dev) {
 543                 skb2->dev = rt6i->rt6i_dev;
 544
 545                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 546
 547                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 548                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 549                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 550                                 rel_type = ICMPV6_DEST_UNREACH;
 551                                 rel_code = ICMPV6_ADDR_UNREACH;
 552                         }
 553                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 554                 }
 555         }
 556         kfree_skb(skb2);
 557         return 0;
 558 #endif
 559 }
 560
 561 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 562 {
 563         if (INET_ECN_is_ce(iph->tos))
 564                 IP6_ECN_set_ce(ipv6_hdr(skb));
 565 }
 566
 567 static int ipip6_rcv(struct sk_buff *skb)
 568 {
 569         struct iphdr *iph;
 570         struct ip_tunnel *tunnel;
 571
 572         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 573                 goto out;
 574
 575         iph = ip_hdr(skb);
 576
 577         read_lock(&ipip6_lock);
 578         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 579                 secpath_reset(skb);
 580                 skb->mac_header = skb->network_header;
 581                 skb_reset_network_header(skb);
 582                 IPCB(skb)->flags = 0;
 583                 skb->protocol = htons(ETH_P_IPV6);
 584                 skb->pkt_type = PACKET_HOST;
 585
 586                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 587                     !isatap_chksrc(skb, iph, tunnel)) {
 588                         tunnel->stat.rx_errors++;
 589                         read_unlock(&ipip6_lock);
 590                         kfree_skb(skb);
 591                         return 0;
 592                 }
 593                 tunnel->stat.rx_packets++;
 594                 tunnel->stat.rx_bytes += skb->len;
 595                 skb->dev = tunnel->dev;
 596                 dst_release(skb->dst);
 597                 skb->dst = NULL;
 598                 nf_reset(skb);
 599                 ipip6_ecn_decapsulate(iph, skb);
 600                 netif_rx(skb);
 601                 read_unlock(&ipip6_lock);
 602                 return 0;
 603         }
 604
 605         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 606         kfree_skb(skb);
 607         read_unlock(&ipip6_lock);
 608 out:
 609         return 0;
 610 }
 611
 612 /* Returns the embedded IPv4 address if the IPv6 address
 613    comes from 6to4 (RFC 3056) addr space */
 614
 615 static inline __be32 try_6to4(struct in6_addr *v6dst)
 616 {
 617         __be32 dst = 0;
 618
 619         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 620                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 621                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 622         }
 623         return dst;
 624 }
 625
 626 /*
 627  *      This function assumes it is being called from dev_queue_xmit()
 628  *      and that skb is filled properly by that function.
 629  */
 630
 631 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 632 {
 633         struct ip_tunnel *tunnel = netdev_priv(dev);
 634         struct net_device_stats *stats = &tunnel->stat;
 635         struct iphdr  *tiph = &tunnel->parms.iph;
 636         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 637         u8     tos = tunnel->parms.iph.tos;
 638         struct rtable *rt;                      /* Route to the other host */
 639         struct net_device *tdev;                        /* Device to other host */
 640         struct iphdr  *iph;                     /* Our new IP header */
 641         unsigned int max_headroom;              /* The extra header space needed */
 642         __be32 dst = tiph->daddr;
 643         int    mtu;
 644         struct in6_addr *addr6;
 645         int addr_type;
 646
 647         if (tunnel->recursion++) {
 648                 tunnel->stat.collisions++;
 649                 goto tx_error;
 650         }
 651
 652         if (skb->protocol != htons(ETH_P_IPV6))
 653                 goto tx_error;
 654
 655         /* ISATAP (RFC4214) - must come before 6to4 */
 656         if (dev->priv_flags & IFF_ISATAP) {
 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_UNICAST) &&
 672                      ipv6_addr_is_isatap(addr6))
 673                         dst = addr6->s6_addr32[3];
 674                 else
 675                         goto tx_error;
 676         }
 677
 678         if (!dst)
 679                 dst = try_6to4(&iph6->daddr);
 680
 681         if (!dst) {
 682                 struct neighbour *neigh = NULL;
 683
 684                 if (skb->dst)
 685                         neigh = skb->dst->neighbour;
 686
 687                 if (neigh == NULL) {
 688                         if (net_ratelimit())
 689                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 690                         goto tx_error;
 691                 }
 692
 693                 addr6 = (struct in6_addr*)&neigh->primary_key;
 694                 addr_type = ipv6_addr_type(addr6);
 695
 696                 if (addr_type == IPV6_ADDR_ANY) {
 697                         addr6 = &ipv6_hdr(skb)->daddr;
 698                         addr_type = ipv6_addr_type(addr6);
 699                 }
 700
 701                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 702                         goto tx_error_icmp;
 703
 704                 dst = addr6->s6_addr32[3];
 705         }
 706
 707         {
 708                 struct flowi fl = { .nl_u = { .ip4_u =
 709                                               { .daddr = dst,
 710                                                 .saddr = tiph->saddr,
 711                                                 .tos = RT_TOS(tos) } },
 712                                     .oif = tunnel->parms.link,
 713                                     .proto = IPPROTO_IPV6 };
 714                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 715                         tunnel->stat.tx_carrier_errors++;
 716                         goto tx_error_icmp;
 717                 }
 718         }
 719         if (rt->rt_type != RTN_UNICAST) {
 720                 ip_rt_put(rt);
 721                 tunnel->stat.tx_carrier_errors++;
 722                 goto tx_error_icmp;
 723         }
 724         tdev = rt->u.dst.dev;
 725
 726         if (tdev == dev) {
 727                 ip_rt_put(rt);
 728                 tunnel->stat.collisions++;
 729                 goto tx_error;
 730         }
 731
 732         if (tiph->frag_off)
 733                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 734         else
 735                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 736
 737         if (mtu < 68) {
 738                 tunnel->stat.collisions++;
 739                 ip_rt_put(rt);
 740                 goto tx_error;
 741         }
 742         if (mtu < IPV6_MIN_MTU)
 743                 mtu = IPV6_MIN_MTU;
 744         if (tunnel->parms.iph.daddr && skb->dst)
 745                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 746
 747         if (skb->len > mtu) {
 748                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 749                 ip_rt_put(rt);
 750                 goto tx_error;
 751         }
 752
 753         if (tunnel->err_count > 0) {
 754                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 755                         tunnel->err_count--;
 756                         dst_link_failure(skb);
 757                 } else
 758                         tunnel->err_count = 0;
 759         }
 760
 761         /*
 762          * Okay, now see if we can stuff it in the buffer as-is.
 763          */
 764         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 765
 766         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 767             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 768                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 769                 if (!new_skb) {
 770                         ip_rt_put(rt);
 771                         stats->tx_dropped++;
 772                         dev_kfree_skb(skb);
 773                         tunnel->recursion--;
 774                         return 0;
 775                 }
 776                 if (skb->sk)
 777                         skb_set_owner_w(new_skb, skb->sk);
 778                 dev_kfree_skb(skb);
 779                 skb = new_skb;
 780                 iph6 = ipv6_hdr(skb);
 781         }
 782
 783         skb->transport_header = skb->network_header;
 784         skb_push(skb, sizeof(struct iphdr));
 785         skb_reset_network_header(skb);
 786         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 787         IPCB(skb)->flags = 0;
 788         dst_release(skb->dst);
 789         skb->dst = &rt->u.dst;
 790
 791         /*
 792          *      Push down and install the IPIP header.
 793          */
 794
 795         iph                     =       ip_hdr(skb);
 796         iph->version            =       4;
 797         iph->ihl                =       sizeof(struct iphdr)>>2;
 798         if (mtu > IPV6_MIN_MTU)
 799                 iph->frag_off   =       htons(IP_DF);
 800         else
 801                 iph->frag_off   =       0;
 802
 803         iph->protocol           =       IPPROTO_IPV6;
 804         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 805         iph->daddr              =       rt->rt_dst;
 806         iph->saddr              =       rt->rt_src;
 807
 808         if ((iph->ttl = tiph->ttl) == 0)
 809                 iph->ttl        =       iph6->hop_limit;
 810
 811         nf_reset(skb);
 812
 813         IPTUNNEL_XMIT();
 814         tunnel->recursion--;
 815         return 0;
 816
 817 tx_error_icmp:
 818         dst_link_failure(skb);
 819 tx_error:
 820         stats->tx_errors++;
 821         dev_kfree_skb(skb);
 822         tunnel->recursion--;
 823         return 0;
 824 }
 825
 826 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 827 {
 828         struct net_device *tdev = NULL;
 829         struct ip_tunnel *tunnel;
 830         struct iphdr *iph;
 831
 832         tunnel = netdev_priv(dev);
 833         iph = &tunnel->parms.iph;
 834
 835         if (iph->daddr) {
 836                 struct flowi fl = { .nl_u = { .ip4_u =
 837                                               { .daddr = iph->daddr,
 838                                                 .saddr = iph->saddr,
 839                                                 .tos = RT_TOS(iph->tos) } },
 840                                     .oif = tunnel->parms.link,
 841                                     .proto = IPPROTO_IPV6 };
 842                 struct rtable *rt;
 843                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 844                         tdev = rt->u.dst.dev;
 845                         ip_rt_put(rt);
 846                 }
 847                 dev->flags |= IFF_POINTOPOINT;
 848         }
 849
 850         if (!tdev && tunnel->parms.link)
 851                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 852
 853         if (tdev) {
 854                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 855                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 856                 if (dev->mtu < IPV6_MIN_MTU)
 857                         dev->mtu = IPV6_MIN_MTU;
 858         }
 859         dev->iflink = tunnel->parms.link;
 860 }
 861
 862 static int
 863 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 864 {
 865         int err = 0;
 866         struct ip_tunnel_parm p;
 867         struct ip_tunnel_prl prl;
 868         struct ip_tunnel *t;
 869
 870         switch (cmd) {
 871         case SIOCGETTUNNEL:
 872                 t = NULL;
 873                 if (dev == ipip6_fb_tunnel_dev) {
 874                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 875                                 err = -EFAULT;
 876                                 break;
 877                         }
 878                         t = ipip6_tunnel_locate(&p, 0);
 879                 }
 880                 if (t == NULL)
 881                         t = netdev_priv(dev);
 882                 memcpy(&p, &t->parms, sizeof(p));
 883                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 884                         err = -EFAULT;
 885                 break;
 886
 887         case SIOCADDTUNNEL:
 888         case SIOCCHGTUNNEL:
 889                 err = -EPERM;
 890                 if (!capable(CAP_NET_ADMIN))
 891                         goto done;
 892
 893                 err = -EFAULT;
 894                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 895                         goto done;
 896
 897                 err = -EINVAL;
 898                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 899                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 900                         goto done;
 901                 if (p.iph.ttl)
 902                         p.iph.frag_off |= htons(IP_DF);
 903
 904                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 905
 906                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 907                         if (t != NULL) {
 908                                 if (t->dev != dev) {
 909                                         err = -EEXIST;
 910                                         break;
 911                                 }
 912                         } else {
 913                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 914                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 915                                         err = -EINVAL;
 916                                         break;
 917                                 }
 918                                 t = netdev_priv(dev);
 919                                 ipip6_tunnel_unlink(t);
 920                                 t->parms.iph.saddr = p.iph.saddr;
 921                                 t->parms.iph.daddr = p.iph.daddr;
 922                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 923                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 924                                 ipip6_tunnel_link(t);
 925                                 netdev_state_change(dev);
 926                         }
 927                 }
 928
 929                 if (t) {
 930                         err = 0;
 931                         if (cmd == SIOCCHGTUNNEL) {
 932                                 t->parms.iph.ttl = p.iph.ttl;
 933                                 t->parms.iph.tos = p.iph.tos;
 934                                 if (t->parms.link != p.link) {
 935                                         t->parms.link = p.link;
 936                                         ipip6_tunnel_bind_dev(dev);
 937                                         netdev_state_change(dev);
 938                                 }
 939                         }
 940                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 941                                 err = -EFAULT;
 942                 } else
 943                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 944                 break;
 945
 946         case SIOCDELTUNNEL:
 947                 err = -EPERM;
 948                 if (!capable(CAP_NET_ADMIN))
 949                         goto done;
 950
 951                 if (dev == ipip6_fb_tunnel_dev) {
 952                         err = -EFAULT;
 953                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 954                                 goto done;
 955                         err = -ENOENT;
 956                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 957                                 goto done;
 958                         err = -EPERM;
 959                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 960                                 goto done;
 961                         dev = t->dev;
 962                 }
 963                 unregister_netdevice(dev);
 964                 err = 0;
 965                 break;
 966
 967         case SIOCGETPRL:
 968         case SIOCADDPRL:
 969         case SIOCDELPRL:
 970         case SIOCCHGPRL:
 971                 err = -EPERM;
 972                 if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
 973                         goto done;
 974                 err = -EINVAL;
 975                 if (dev == ipip6_fb_tunnel_dev)
 976                         goto done;
 977                 err = -EFAULT;
 978                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 979                         goto done;
 980                 err = -ENOENT;
 981                 if (!(t = netdev_priv(dev)))
 982                         goto done;
 983
 984                 switch (cmd) {
 985                 case SIOCGETPRL:
 986                         err = ipip6_tunnel_get_prl(t, &prl);
 987                         if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
 988                                                  &prl, sizeof(prl)))
 989                                 err = -EFAULT;
 990                         break;
 991                 case SIOCDELPRL:
 992                         err = ipip6_tunnel_del_prl(t, &prl);
 993                         break;
 994                 case SIOCADDPRL:
 995                 case SIOCCHGPRL:
 996                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 997                         break;
 998                 }
 999                 if (cmd != SIOCGETPRL)
1000                         netdev_state_change(dev);
1001                 break;
1002
1003         default:
1004                 err = -EINVAL;
1005         }
1006
1007 done:
1008         return err;
1009 }
1010
1011 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
1012 {
1013         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
1014 }
1015
1016 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1017 {
1018         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1019                 return -EINVAL;
1020         dev->mtu = new_mtu;
1021         return 0;
1022 }
1023
1024 static void ipip6_tunnel_setup(struct net_device *dev)
1025 {
1026         dev->uninit             = ipip6_tunnel_uninit;
1027         dev->destructor         = free_netdev;
1028         dev->hard_start_xmit    = ipip6_tunnel_xmit;
1029         dev->get_stats          = ipip6_tunnel_get_stats;
1030         dev->do_ioctl           = ipip6_tunnel_ioctl;
1031         dev->change_mtu         = ipip6_tunnel_change_mtu;
1032
1033         dev->type               = ARPHRD_SIT;
1034         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1035         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1036         dev->flags              = IFF_NOARP;
1037         dev->iflink             = 0;
1038         dev->addr_len           = 4;
1039 }
1040
1041 static int ipip6_tunnel_init(struct net_device *dev)
1042 {
1043         struct ip_tunnel *tunnel;
1044
1045         tunnel = netdev_priv(dev);
1046
1047         tunnel->dev = dev;
1048         strcpy(tunnel->parms.name, dev->name);
1049
1050         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1051         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1052
1053         ipip6_tunnel_bind_dev(dev);
1054
1055         return 0;
1056 }
1057
1058 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
1059 {
1060         struct ip_tunnel *tunnel = netdev_priv(dev);
1061         struct iphdr *iph = &tunnel->parms.iph;
1062
1063         tunnel->dev = dev;
1064         strcpy(tunnel->parms.name, dev->name);
1065
1066         iph->version            = 4;
1067         iph->protocol           = IPPROTO_IPV6;
1068         iph->ihl                = 5;
1069         iph->ttl                = 64;
1070
1071         dev_hold(dev);
1072         tunnels_wc[0]           = tunnel;
1073         return 0;
1074 }
1075
1076 static struct xfrm_tunnel sit_handler = {
1077         .handler        =       ipip6_rcv,
1078         .err_handler    =       ipip6_err,
1079         .priority       =       1,
1080 };
1081
1082 static void __exit sit_destroy_tunnels(void)
1083 {
1084         int prio;
1085
1086         for (prio = 1; prio < 4; prio++) {
1087                 int h;
1088                 for (h = 0; h < HASH_SIZE; h++) {
1089                         struct ip_tunnel *t;
1090                         while ((t = tunnels[prio][h]) != NULL)
1091                                 unregister_netdevice(t->dev);
1092                 }
1093         }
1094 }
1095
1096 static void __exit sit_cleanup(void)
1097 {
1098         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1099
1100         rtnl_lock();
1101         sit_destroy_tunnels();
1102         unregister_netdevice(ipip6_fb_tunnel_dev);
1103         rtnl_unlock();
1104 }
1105
1106 static int __init sit_init(void)
1107 {
1108         int err;
1109
1110         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1111
1112         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1113                 printk(KERN_INFO "sit init: Can't add protocol\n");
1114                 return -EAGAIN;
1115         }
1116
1117         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1118                                            ipip6_tunnel_setup);
1119         if (!ipip6_fb_tunnel_dev) {
1120                 err = -ENOMEM;
1121                 goto err1;
1122         }
1123
1124         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1125
1126         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
1127                 goto err2;
1128
1129  out:
1130         return err;
1131  err2:
1132         free_netdev(ipip6_fb_tunnel_dev);
1133  err1:
1134         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1135         goto out;
1136 }
1137
1138 module_init(sit_init);
1139 module_exit(sit_cleanup);
1140 MODULE_LICENSE("GPL");
1141 MODULE_ALIAS("sit0");