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->entry.addr == addr)
 207                         break;
 208         return p;
 209
 210 }
 211
 212 static int
 213 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 214 {
 215         struct ip_tunnel_prl_entry *p;
 216         int err = 0;
 217
 218         if (a->addr == htonl(INADDR_ANY))
 219                 return -EINVAL;
 220
 221         write_lock(&ipip6_lock);
 222
 223         for (p = t->prl; p; p = p->next) {
 224                 if (p->entry.addr == a->addr) {
 225                         if (chg)
 226                                 goto update;
 227                         err = -EEXIST;
 228                         goto out;
 229                 }
 230         }
 231
 232         if (chg) {
 233                 err = -ENXIO;
 234                 goto out;
 235         }
 236
 237         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 238         if (!p) {
 239                 err = -ENOBUFS;
 240                 goto out;
 241         }
 242
 243         p->next = t->prl;
 244         t->prl = p;
 245 update:
 246         p->entry = *a;
 247 out:
 248         write_unlock(&ipip6_lock);
 249         return err;
 250 }
 251
 252 static int
 253 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 254 {
 255         struct ip_tunnel_prl_entry *x, **p;
 256         int err = 0;
 257
 258         write_lock(&ipip6_lock);
 259
 260         if (a && a->addr != htonl(INADDR_ANY)) {
 261                 for (p = &t->prl; *p; p = &(*p)->next) {
 262                         if ((*p)->entry.addr == a->addr) {
 263                                 x = *p;
 264                                 *p = x->next;
 265                                 kfree(x);
 266                                 goto out;
 267                         }
 268                 }
 269                 err = -ENXIO;
 270         } else {
 271                 while (t->prl) {
 272                         x = t->prl;
 273                         t->prl = t->prl->next;
 274                         kfree(x);
 275                 }
 276         }
 277 out:
 278         write_unlock(&ipip6_lock);
 279         return 0;
 280 }
 281
 282 /* copied directly from anycast.c */
 283 static int
 284 ipip6_onlink(struct in6_addr *addr, struct net_device *dev)
 285 {
 286         struct inet6_dev        *idev;
 287         struct inet6_ifaddr     *ifa;
 288         int     onlink;
 289
 290         onlink = 0;
 291         rcu_read_lock();
 292         idev = __in6_dev_get(dev);
 293         if (idev) {
 294                 read_lock_bh(&idev->lock);
 295                 for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 296                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
 297                                                    ifa->prefix_len);
 298                         if (onlink)
 299                                 break;
 300                 }
 301                 read_unlock_bh(&idev->lock);
 302         }
 303         rcu_read_unlock();
 304         return onlink;
 305 }
 306
 307 static int
 308 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 309 {
 310         struct ip_tunnel_prl_entry *p;
 311         int ok = 1;
 312
 313         read_lock(&ipip6_lock);
 314         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 315         if (p) {
 316                 if (p->entry.flags & PRL_DEFAULT)
 317                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 318                 else
 319                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 320         } else {
 321                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 322                 if (ipv6_addr_is_isatap(addr6) &&
 323                     (addr6->s6_addr32[3] == iph->saddr) &&
 324                     ipip6_onlink(addr6, t->dev))
 325                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 326                 else
 327                         ok = 0;
 328         }
 329         read_unlock(&ipip6_lock);
 330         return ok;
 331 }
 332
 333 static void ipip6_tunnel_uninit(struct net_device *dev)
 334 {
 335         if (dev == ipip6_fb_tunnel_dev) {
 336                 write_lock_bh(&ipip6_lock);
 337                 tunnels_wc[0] = NULL;
 338                 write_unlock_bh(&ipip6_lock);
 339                 dev_put(dev);
 340         } else {
 341                 ipip6_tunnel_unlink(netdev_priv(dev));
 342                 ipip6_tunnel_del_prl(netdev_priv(dev), 0);
 343                 dev_put(dev);
 344         }
 345 }
 346
 347
 348 static int ipip6_err(struct sk_buff *skb, u32 info)
 349 {
 350 #ifndef I_WISH_WORLD_WERE_PERFECT
 351
 352 /* It is not :-( All the routers (except for Linux) return only
 353    8 bytes of packet payload. It means, that precise relaying of
 354    ICMP in the real Internet is absolutely infeasible.
 355  */
 356         struct iphdr *iph = (struct iphdr*)skb->data;
 357         const int type = icmp_hdr(skb)->type;
 358         const int code = icmp_hdr(skb)->code;
 359         struct ip_tunnel *t;
 360         int err;
 361
 362         switch (type) {
 363         default:
 364         case ICMP_PARAMETERPROB:
 365                 return 0;
 366
 367         case ICMP_DEST_UNREACH:
 368                 switch (code) {
 369                 case ICMP_SR_FAILED:
 370                 case ICMP_PORT_UNREACH:
 371                         /* Impossible event. */
 372                         return 0;
 373                 case ICMP_FRAG_NEEDED:
 374                         /* Soft state for pmtu is maintained by IP core. */
 375                         return 0;
 376                 default:
 377                         /* All others are translated to HOST_UNREACH.
 378                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 379                            I believe they are just ether pollution. --ANK
 380                          */
 381                         break;
 382                 }
 383                 break;
 384         case ICMP_TIME_EXCEEDED:
 385                 if (code != ICMP_EXC_TTL)
 386                         return 0;
 387                 break;
 388         }
 389
 390         err = -ENOENT;
 391
 392         read_lock(&ipip6_lock);
 393         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 394         if (t == NULL || t->parms.iph.daddr == 0)
 395                 goto out;
 396
 397         err = 0;
 398         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 399                 goto out;
 400
 401         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 402                 t->err_count++;
 403         else
 404                 t->err_count = 1;
 405         t->err_time = jiffies;
 406 out:
 407         read_unlock(&ipip6_lock);
 408         return err;
 409 #else
 410         struct iphdr *iph = (struct iphdr*)dp;
 411         int hlen = iph->ihl<<2;
 412         struct ipv6hdr *iph6;
 413         const int type = icmp_hdr(skb)->type;
 414         const int code = icmp_hdr(skb)->code;
 415         int rel_type = 0;
 416         int rel_code = 0;
 417         int rel_info = 0;
 418         struct sk_buff *skb2;
 419         struct rt6_info *rt6i;
 420
 421         if (len < hlen + sizeof(struct ipv6hdr))
 422                 return;
 423         iph6 = (struct ipv6hdr*)(dp + hlen);
 424
 425         switch (type) {
 426         default:
 427                 return;
 428         case ICMP_PARAMETERPROB:
 429                 if (icmp_hdr(skb)->un.gateway < hlen)
 430                         return;
 431
 432                 /* So... This guy found something strange INSIDE encapsulated
 433                    packet. Well, he is fool, but what can we do ?
 434                  */
 435                 rel_type = ICMPV6_PARAMPROB;
 436                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 437                 break;
 438
 439         case ICMP_DEST_UNREACH:
 440                 switch (code) {
 441                 case ICMP_SR_FAILED:
 442                 case ICMP_PORT_UNREACH:
 443                         /* Impossible event. */
 444                         return;
 445                 case ICMP_FRAG_NEEDED:
 446                         /* Too complicated case ... */
 447                         return;
 448                 default:
 449                         /* All others are translated to HOST_UNREACH.
 450                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 451                            I believe, it is just ether pollution. --ANK
 452                          */
 453                         rel_type = ICMPV6_DEST_UNREACH;
 454                         rel_code = ICMPV6_ADDR_UNREACH;
 455                         break;
 456                 }
 457                 break;
 458         case ICMP_TIME_EXCEEDED:
 459                 if (code != ICMP_EXC_TTL)
 460                         return;
 461                 rel_type = ICMPV6_TIME_EXCEED;
 462                 rel_code = ICMPV6_EXC_HOPLIMIT;
 463                 break;
 464         }
 465
 466         /* Prepare fake skb to feed it to icmpv6_send */
 467         skb2 = skb_clone(skb, GFP_ATOMIC);
 468         if (skb2 == NULL)
 469                 return 0;
 470         dst_release(skb2->dst);
 471         skb2->dst = NULL;
 472         skb_pull(skb2, skb->data - (u8*)iph6);
 473         skb_reset_network_header(skb2);
 474
 475         /* Try to guess incoming interface */
 476         rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
 477         if (rt6i && rt6i->rt6i_dev) {
 478                 skb2->dev = rt6i->rt6i_dev;
 479
 480                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 481
 482                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 483                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 484                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 485                                 rel_type = ICMPV6_DEST_UNREACH;
 486                                 rel_code = ICMPV6_ADDR_UNREACH;
 487                         }
 488                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 489                 }
 490         }
 491         kfree_skb(skb2);
 492         return 0;
 493 #endif
 494 }
 495
 496 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 497 {
 498         if (INET_ECN_is_ce(iph->tos))
 499                 IP6_ECN_set_ce(ipv6_hdr(skb));
 500 }
 501
 502 static int ipip6_rcv(struct sk_buff *skb)
 503 {
 504         struct iphdr *iph;
 505         struct ip_tunnel *tunnel;
 506
 507         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 508                 goto out;
 509
 510         iph = ip_hdr(skb);
 511
 512         read_lock(&ipip6_lock);
 513         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 514                 secpath_reset(skb);
 515                 skb->mac_header = skb->network_header;
 516                 skb_reset_network_header(skb);
 517                 IPCB(skb)->flags = 0;
 518                 skb->protocol = htons(ETH_P_IPV6);
 519                 skb->pkt_type = PACKET_HOST;
 520
 521                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 522                     !isatap_chksrc(skb, iph, tunnel)) {
 523                         tunnel->stat.rx_errors++;
 524                         read_unlock(&ipip6_lock);
 525                         kfree_skb(skb);
 526                         return 0;
 527                 }
 528                 tunnel->stat.rx_packets++;
 529                 tunnel->stat.rx_bytes += skb->len;
 530                 skb->dev = tunnel->dev;
 531                 dst_release(skb->dst);
 532                 skb->dst = NULL;
 533                 nf_reset(skb);
 534                 ipip6_ecn_decapsulate(iph, skb);
 535                 netif_rx(skb);
 536                 read_unlock(&ipip6_lock);
 537                 return 0;
 538         }
 539
 540         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 541         kfree_skb(skb);
 542         read_unlock(&ipip6_lock);
 543 out:
 544         return 0;
 545 }
 546
 547 /* Returns the embedded IPv4 address if the IPv6 address
 548    comes from 6to4 (RFC 3056) addr space */
 549
 550 static inline __be32 try_6to4(struct in6_addr *v6dst)
 551 {
 552         __be32 dst = 0;
 553
 554         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 555                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 556                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 557         }
 558         return dst;
 559 }
 560
 561 /*
 562  *      This function assumes it is being called from dev_queue_xmit()
 563  *      and that skb is filled properly by that function.
 564  */
 565
 566 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 567 {
 568         struct ip_tunnel *tunnel = netdev_priv(dev);
 569         struct net_device_stats *stats = &tunnel->stat;
 570         struct iphdr  *tiph = &tunnel->parms.iph;
 571         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 572         u8     tos = tunnel->parms.iph.tos;
 573         struct rtable *rt;                      /* Route to the other host */
 574         struct net_device *tdev;                        /* Device to other host */
 575         struct iphdr  *iph;                     /* Our new IP header */
 576         unsigned int max_headroom;              /* The extra header space needed */
 577         __be32 dst = tiph->daddr;
 578         int    mtu;
 579         struct in6_addr *addr6;
 580         int addr_type;
 581
 582         if (tunnel->recursion++) {
 583                 tunnel->stat.collisions++;
 584                 goto tx_error;
 585         }
 586
 587         if (skb->protocol != htons(ETH_P_IPV6))
 588                 goto tx_error;
 589
 590         /* ISATAP (RFC4214) - must come before 6to4 */
 591         if (dev->priv_flags & IFF_ISATAP) {
 592                 struct neighbour *neigh = NULL;
 593
 594                 if (skb->dst)
 595                         neigh = skb->dst->neighbour;
 596
 597                 if (neigh == NULL) {
 598                         if (net_ratelimit())
 599                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 600                         goto tx_error;
 601                 }
 602
 603                 addr6 = (struct in6_addr*)&neigh->primary_key;
 604                 addr_type = ipv6_addr_type(addr6);
 605
 606                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 607                      ipv6_addr_is_isatap(addr6))
 608                         dst = addr6->s6_addr32[3];
 609                 else
 610                         goto tx_error;
 611         }
 612
 613         if (!dst)
 614                 dst = try_6to4(&iph6->daddr);
 615
 616         if (!dst) {
 617                 struct neighbour *neigh = NULL;
 618
 619                 if (skb->dst)
 620                         neigh = skb->dst->neighbour;
 621
 622                 if (neigh == NULL) {
 623                         if (net_ratelimit())
 624                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 625                         goto tx_error;
 626                 }
 627
 628                 addr6 = (struct in6_addr*)&neigh->primary_key;
 629                 addr_type = ipv6_addr_type(addr6);
 630
 631                 if (addr_type == IPV6_ADDR_ANY) {
 632                         addr6 = &ipv6_hdr(skb)->daddr;
 633                         addr_type = ipv6_addr_type(addr6);
 634                 }
 635
 636                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 637                         goto tx_error_icmp;
 638
 639                 dst = addr6->s6_addr32[3];
 640         }
 641
 642         {
 643                 struct flowi fl = { .nl_u = { .ip4_u =
 644                                               { .daddr = dst,
 645                                                 .saddr = tiph->saddr,
 646                                                 .tos = RT_TOS(tos) } },
 647                                     .oif = tunnel->parms.link,
 648                                     .proto = IPPROTO_IPV6 };
 649                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 650                         tunnel->stat.tx_carrier_errors++;
 651                         goto tx_error_icmp;
 652                 }
 653         }
 654         if (rt->rt_type != RTN_UNICAST) {
 655                 ip_rt_put(rt);
 656                 tunnel->stat.tx_carrier_errors++;
 657                 goto tx_error_icmp;
 658         }
 659         tdev = rt->u.dst.dev;
 660
 661         if (tdev == dev) {
 662                 ip_rt_put(rt);
 663                 tunnel->stat.collisions++;
 664                 goto tx_error;
 665         }
 666
 667         if (tiph->frag_off)
 668                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 669         else
 670                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 671
 672         if (mtu < 68) {
 673                 tunnel->stat.collisions++;
 674                 ip_rt_put(rt);
 675                 goto tx_error;
 676         }
 677         if (mtu < IPV6_MIN_MTU)
 678                 mtu = IPV6_MIN_MTU;
 679         if (tunnel->parms.iph.daddr && skb->dst)
 680                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 681
 682         if (skb->len > mtu) {
 683                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 684                 ip_rt_put(rt);
 685                 goto tx_error;
 686         }
 687
 688         if (tunnel->err_count > 0) {
 689                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 690                         tunnel->err_count--;
 691                         dst_link_failure(skb);
 692                 } else
 693                         tunnel->err_count = 0;
 694         }
 695
 696         /*
 697          * Okay, now see if we can stuff it in the buffer as-is.
 698          */
 699         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 700
 701         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 702             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 703                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 704                 if (!new_skb) {
 705                         ip_rt_put(rt);
 706                         stats->tx_dropped++;
 707                         dev_kfree_skb(skb);
 708                         tunnel->recursion--;
 709                         return 0;
 710                 }
 711                 if (skb->sk)
 712                         skb_set_owner_w(new_skb, skb->sk);
 713                 dev_kfree_skb(skb);
 714                 skb = new_skb;
 715                 iph6 = ipv6_hdr(skb);
 716         }
 717
 718         skb->transport_header = skb->network_header;
 719         skb_push(skb, sizeof(struct iphdr));
 720         skb_reset_network_header(skb);
 721         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 722         IPCB(skb)->flags = 0;
 723         dst_release(skb->dst);
 724         skb->dst = &rt->u.dst;
 725
 726         /*
 727          *      Push down and install the IPIP header.
 728          */
 729
 730         iph                     =       ip_hdr(skb);
 731         iph->version            =       4;
 732         iph->ihl                =       sizeof(struct iphdr)>>2;
 733         if (mtu > IPV6_MIN_MTU)
 734                 iph->frag_off   =       htons(IP_DF);
 735         else
 736                 iph->frag_off   =       0;
 737
 738         iph->protocol           =       IPPROTO_IPV6;
 739         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 740         iph->daddr              =       rt->rt_dst;
 741         iph->saddr              =       rt->rt_src;
 742
 743         if ((iph->ttl = tiph->ttl) == 0)
 744                 iph->ttl        =       iph6->hop_limit;
 745
 746         nf_reset(skb);
 747
 748         IPTUNNEL_XMIT();
 749         tunnel->recursion--;
 750         return 0;
 751
 752 tx_error_icmp:
 753         dst_link_failure(skb);
 754 tx_error:
 755         stats->tx_errors++;
 756         dev_kfree_skb(skb);
 757         tunnel->recursion--;
 758         return 0;
 759 }
 760
 761 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 762 {
 763         struct net_device *tdev = NULL;
 764         struct ip_tunnel *tunnel;
 765         struct iphdr *iph;
 766
 767         tunnel = netdev_priv(dev);
 768         iph = &tunnel->parms.iph;
 769
 770         if (iph->daddr) {
 771                 struct flowi fl = { .nl_u = { .ip4_u =
 772                                               { .daddr = iph->daddr,
 773                                                 .saddr = iph->saddr,
 774                                                 .tos = RT_TOS(iph->tos) } },
 775                                     .oif = tunnel->parms.link,
 776                                     .proto = IPPROTO_IPV6 };
 777                 struct rtable *rt;
 778                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 779                         tdev = rt->u.dst.dev;
 780                         ip_rt_put(rt);
 781                 }
 782                 dev->flags |= IFF_POINTOPOINT;
 783         }
 784
 785         if (!tdev && tunnel->parms.link)
 786                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 787
 788         if (tdev) {
 789                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 790                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 791                 if (dev->mtu < IPV6_MIN_MTU)
 792                         dev->mtu = IPV6_MIN_MTU;
 793         }
 794         dev->iflink = tunnel->parms.link;
 795 }
 796
 797 static int
 798 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 799 {
 800         int err = 0;
 801         struct ip_tunnel_parm p;
 802         struct ip_tunnel_prl prl;
 803         struct ip_tunnel *t;
 804
 805         switch (cmd) {
 806         case SIOCGETTUNNEL:
 807                 t = NULL;
 808                 if (dev == ipip6_fb_tunnel_dev) {
 809                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 810                                 err = -EFAULT;
 811                                 break;
 812                         }
 813                         t = ipip6_tunnel_locate(&p, 0);
 814                 }
 815                 if (t == NULL)
 816                         t = netdev_priv(dev);
 817                 memcpy(&p, &t->parms, sizeof(p));
 818                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 819                         err = -EFAULT;
 820                 break;
 821
 822         case SIOCADDTUNNEL:
 823         case SIOCCHGTUNNEL:
 824                 err = -EPERM;
 825                 if (!capable(CAP_NET_ADMIN))
 826                         goto done;
 827
 828                 err = -EFAULT;
 829                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 830                         goto done;
 831
 832                 err = -EINVAL;
 833                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 834                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 835                         goto done;
 836                 if (p.iph.ttl)
 837                         p.iph.frag_off |= htons(IP_DF);
 838
 839                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 840
 841                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 842                         if (t != NULL) {
 843                                 if (t->dev != dev) {
 844                                         err = -EEXIST;
 845                                         break;
 846                                 }
 847                         } else {
 848                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 849                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 850                                         err = -EINVAL;
 851                                         break;
 852                                 }
 853                                 t = netdev_priv(dev);
 854                                 ipip6_tunnel_unlink(t);
 855                                 t->parms.iph.saddr = p.iph.saddr;
 856                                 t->parms.iph.daddr = p.iph.daddr;
 857                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 858                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 859                                 ipip6_tunnel_link(t);
 860                                 netdev_state_change(dev);
 861                         }
 862                 }
 863
 864                 if (t) {
 865                         err = 0;
 866                         if (cmd == SIOCCHGTUNNEL) {
 867                                 t->parms.iph.ttl = p.iph.ttl;
 868                                 t->parms.iph.tos = p.iph.tos;
 869                                 if (t->parms.link != p.link) {
 870                                         t->parms.link = p.link;
 871                                         ipip6_tunnel_bind_dev(dev);
 872                                         netdev_state_change(dev);
 873                                 }
 874                         }
 875                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 876                                 err = -EFAULT;
 877                 } else
 878                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 879                 break;
 880
 881         case SIOCDELTUNNEL:
 882                 err = -EPERM;
 883                 if (!capable(CAP_NET_ADMIN))
 884                         goto done;
 885
 886                 if (dev == ipip6_fb_tunnel_dev) {
 887                         err = -EFAULT;
 888                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 889                                 goto done;
 890                         err = -ENOENT;
 891                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 892                                 goto done;
 893                         err = -EPERM;
 894                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 895                                 goto done;
 896                         dev = t->dev;
 897                 }
 898                 unregister_netdevice(dev);
 899                 err = 0;
 900                 break;
 901
 902         case SIOCADDPRL:
 903         case SIOCDELPRL:
 904         case SIOCCHGPRL:
 905                 err = -EPERM;
 906                 if (!capable(CAP_NET_ADMIN))
 907                         goto done;
 908                 err = -EINVAL;
 909                 if (dev == ipip6_fb_tunnel_dev)
 910                         goto done;
 911                 err = -EFAULT;
 912                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 913                         goto done;
 914                 err = -ENOENT;
 915                 if (!(t = netdev_priv(dev)))
 916                         goto done;
 917
 918                 if (cmd == SIOCDELPRL)
 919                         err = ipip6_tunnel_del_prl(t, &prl);
 920                 else
 921                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 922                 netdev_state_change(dev);
 923                 break;
 924
 925         default:
 926                 err = -EINVAL;
 927         }
 928
 929 done:
 930         return err;
 931 }
 932
 933 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 934 {
 935         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 936 }
 937
 938 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 939 {
 940         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 941                 return -EINVAL;
 942         dev->mtu = new_mtu;
 943         return 0;
 944 }
 945
 946 static void ipip6_tunnel_setup(struct net_device *dev)
 947 {
 948         dev->uninit             = ipip6_tunnel_uninit;
 949         dev->destructor         = free_netdev;
 950         dev->hard_start_xmit    = ipip6_tunnel_xmit;
 951         dev->get_stats          = ipip6_tunnel_get_stats;
 952         dev->do_ioctl           = ipip6_tunnel_ioctl;
 953         dev->change_mtu         = ipip6_tunnel_change_mtu;
 954
 955         dev->type               = ARPHRD_SIT;
 956         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 957         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 958         dev->flags              = IFF_NOARP;
 959         dev->iflink             = 0;
 960         dev->addr_len           = 4;
 961 }
 962
 963 static int ipip6_tunnel_init(struct net_device *dev)
 964 {
 965         struct ip_tunnel *tunnel;
 966
 967         tunnel = netdev_priv(dev);
 968
 969         tunnel->dev = dev;
 970         strcpy(tunnel->parms.name, dev->name);
 971
 972         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 973         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 974
 975         ipip6_tunnel_bind_dev(dev);
 976
 977         return 0;
 978 }
 979
 980 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
 981 {
 982         struct ip_tunnel *tunnel = netdev_priv(dev);
 983         struct iphdr *iph = &tunnel->parms.iph;
 984
 985         tunnel->dev = dev;
 986         strcpy(tunnel->parms.name, dev->name);
 987
 988         iph->version            = 4;
 989         iph->protocol           = IPPROTO_IPV6;
 990         iph->ihl                = 5;
 991         iph->ttl                = 64;
 992
 993         dev_hold(dev);
 994         tunnels_wc[0]           = tunnel;
 995         return 0;
 996 }
 997
 998 static struct xfrm_tunnel sit_handler = {
 999         .handler        =       ipip6_rcv,
1000         .err_handler    =       ipip6_err,
1001         .priority       =       1,
1002 };
1003
1004 static void __exit sit_destroy_tunnels(void)
1005 {
1006         int prio;
1007
1008         for (prio = 1; prio < 4; prio++) {
1009                 int h;
1010                 for (h = 0; h < HASH_SIZE; h++) {
1011                         struct ip_tunnel *t;
1012                         while ((t = tunnels[prio][h]) != NULL)
1013                                 unregister_netdevice(t->dev);
1014                 }
1015         }
1016 }
1017
1018 static void __exit sit_cleanup(void)
1019 {
1020         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1021
1022         rtnl_lock();
1023         sit_destroy_tunnels();
1024         unregister_netdevice(ipip6_fb_tunnel_dev);
1025         rtnl_unlock();
1026 }
1027
1028 static int __init sit_init(void)
1029 {
1030         int err;
1031
1032         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1033
1034         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1035                 printk(KERN_INFO "sit init: Can't add protocol\n");
1036                 return -EAGAIN;
1037         }
1038
1039         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1040                                            ipip6_tunnel_setup);
1041         if (!ipip6_fb_tunnel_dev) {
1042                 err = -ENOMEM;
1043                 goto err1;
1044         }
1045
1046         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1047
1048         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
1049                 goto err2;
1050
1051  out:
1052         return err;
1053  err2:
1054         free_netdev(ipip6_fb_tunnel_dev);
1055  err1:
1056         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1057         goto out;
1058 }
1059
1060 module_init(sit_init);
1061 module_exit(sit_cleanup);
1062 MODULE_LICENSE("GPL");
1063 MODULE_ALIAS("sit0");