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