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  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  *
  14  *      Changes:
  15  * Roger Venning <r.venning@telstra.com>:       6to4 support
  16  * Nate Thompson <nate@thebog.net>:             6to4 support
  17  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  18  */
  19
  20 #include <linux/module.h>
  21 #include <linux/capability.h>
  22 #include <linux/errno.h>
  23 #include <linux/types.h>
  24 #include <linux/socket.h>
  25 #include <linux/sockios.h>
  26 #include <linux/net.h>
  27 #include <linux/in6.h>
  28 #include <linux/netdevice.h>
  29 #include <linux/if_arp.h>
  30 #include <linux/icmp.h>
  31 #include <asm/uaccess.h>
  32 #include <linux/init.h>
  33 #include <linux/netfilter_ipv4.h>
  34 #include <linux/if_ether.h>
  35
  36 #include <net/sock.h>
  37 #include <net/snmp.h>
  38
  39 #include <net/ipv6.h>
  40 #include <net/protocol.h>
  41 #include <net/transp_v6.h>
  42 #include <net/ip6_fib.h>
  43 #include <net/ip6_route.h>
  44 #include <net/ndisc.h>
  45 #include <net/addrconf.h>
  46 #include <net/ip.h>
  47 #include <net/udp.h>
  48 #include <net/icmp.h>
  49 #include <net/ipip.h>
  50 #include <net/inet_ecn.h>
  51 #include <net/xfrm.h>
  52 #include <net/dsfield.h>
  53 #include <net/net_namespace.h>
  54 #include <net/netns/generic.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 void ipip6_fb_tunnel_init(struct net_device *dev);
  66 static void ipip6_tunnel_init(struct net_device *dev);
  67 static void ipip6_tunnel_setup(struct net_device *dev);
  68
  69 static int sit_net_id;
  70 struct sit_net {
  71         struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  72         struct ip_tunnel *tunnels_r[HASH_SIZE];
  73         struct ip_tunnel *tunnels_l[HASH_SIZE];
  74         struct ip_tunnel *tunnels_wc[1];
  75         struct ip_tunnel **tunnels[4];
  76
  77         struct net_device *fb_tunnel_dev;
  78 };
  79
  80 static DEFINE_RWLOCK(ipip6_lock);
  81
  82 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
  83                 __be32 remote, __be32 local)
  84 {
  85         unsigned h0 = HASH(remote);
  86         unsigned h1 = HASH(local);
  87         struct ip_tunnel *t;
  88         struct sit_net *sitn = net_generic(net, sit_net_id);
  89
  90         for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
  91                 if (local == t->parms.iph.saddr &&
  92                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  93                         return t;
  94         }
  95         for (t = sitn->tunnels_r[h0]; t; t = t->next) {
  96                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  97                         return t;
  98         }
  99         for (t = sitn->tunnels_l[h1]; t; t = t->next) {
 100                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
 101                         return t;
 102         }
 103         if ((t = sitn->tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
 104                 return t;
 105         return NULL;
 106 }
 107
 108 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
 109                 struct ip_tunnel_parm *parms)
 110 {
 111         __be32 remote = parms->iph.daddr;
 112         __be32 local = parms->iph.saddr;
 113         unsigned h = 0;
 114         int prio = 0;
 115
 116         if (remote) {
 117                 prio |= 2;
 118                 h ^= HASH(remote);
 119         }
 120         if (local) {
 121                 prio |= 1;
 122                 h ^= HASH(local);
 123         }
 124         return &sitn->tunnels[prio][h];
 125 }
 126
 127 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
 128                 struct ip_tunnel *t)
 129 {
 130         return __ipip6_bucket(sitn, &t->parms);
 131 }
 132
 133 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 134 {
 135         struct ip_tunnel **tp;
 136
 137         for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
 138                 if (t == *tp) {
 139                         write_lock_bh(&ipip6_lock);
 140                         *tp = t->next;
 141                         write_unlock_bh(&ipip6_lock);
 142                         break;
 143                 }
 144         }
 145 }
 146
 147 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 148 {
 149         struct ip_tunnel **tp = ipip6_bucket(sitn, t);
 150
 151         t->next = *tp;
 152         write_lock_bh(&ipip6_lock);
 153         *tp = t;
 154         write_unlock_bh(&ipip6_lock);
 155 }
 156
 157 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
 158                 struct ip_tunnel_parm *parms, int create)
 159 {
 160         __be32 remote = parms->iph.daddr;
 161         __be32 local = parms->iph.saddr;
 162         struct ip_tunnel *t, **tp, *nt;
 163         struct net_device *dev;
 164         char name[IFNAMSIZ];
 165         struct sit_net *sitn = net_generic(net, sit_net_id);
 166
 167         for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
 168                 if (local == t->parms.iph.saddr &&
 169                     remote == t->parms.iph.daddr) {
 170                         if (create)
 171                                 return NULL;
 172                         else
 173                                 return t;
 174                 }
 175         }
 176         if (!create)
 177                 goto failed;
 178
 179         if (parms->name[0])
 180                 strlcpy(name, parms->name, IFNAMSIZ);
 181         else
 182                 sprintf(name, "sit%%d");
 183
 184         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 185         if (dev == NULL)
 186                 return NULL;
 187
 188         dev_net_set(dev, net);
 189
 190         if (strchr(name, '%')) {
 191                 if (dev_alloc_name(dev, name) < 0)
 192                         goto failed_free;
 193         }
 194
 195         nt = netdev_priv(dev);
 196
 197         nt->parms = *parms;
 198         ipip6_tunnel_init(dev);
 199
 200         if (parms->i_flags & SIT_ISATAP)
 201                 dev->priv_flags |= IFF_ISATAP;
 202
 203         if (register_netdevice(dev) < 0)
 204                 goto failed_free;
 205
 206         dev_hold(dev);
 207
 208         ipip6_tunnel_link(sitn, nt);
 209         return nt;
 210
 211 failed_free:
 212         free_netdev(dev);
 213 failed:
 214         return NULL;
 215 }
 216
 217 static struct ip_tunnel_prl_entry *
 218 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 219 {
 220         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 221
 222         for (p = t->prl; p; p = p->next)
 223                 if (p->addr == addr)
 224                         break;
 225         return p;
 226
 227 }
 228
 229 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 230                                 struct ip_tunnel_prl __user *a)
 231 {
 232         struct ip_tunnel_prl kprl, *kp;
 233         struct ip_tunnel_prl_entry *prl;
 234         unsigned int cmax, c = 0, ca, len;
 235         int ret = 0;
 236
 237         if (copy_from_user(&kprl, a, sizeof(kprl)))
 238                 return -EFAULT;
 239         cmax = kprl.datalen / sizeof(kprl);
 240         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 241                 cmax = 1;
 242
 243         /* For simple GET or for root users,
 244          * we try harder to allocate.
 245          */
 246         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 247                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 248                 NULL;
 249
 250         read_lock(&ipip6_lock);
 251
 252         ca = t->prl_count < cmax ? t->prl_count : cmax;
 253
 254         if (!kp) {
 255                 /* We don't try hard to allocate much memory for
 256                  * non-root users.
 257                  * For root users, retry allocating enough memory for
 258                  * the answer.
 259                  */
 260                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 261                 if (!kp) {
 262                         ret = -ENOMEM;
 263                         goto out;
 264                 }
 265         }
 266
 267         c = 0;
 268         for (prl = t->prl; prl; prl = prl->next) {
 269                 if (c > cmax)
 270                         break;
 271                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 272                         continue;
 273                 kp[c].addr = prl->addr;
 274                 kp[c].flags = prl->flags;
 275                 c++;
 276                 if (kprl.addr != htonl(INADDR_ANY))
 277                         break;
 278         }
 279 out:
 280         read_unlock(&ipip6_lock);
 281
 282         len = sizeof(*kp) * c;
 283         ret = 0;
 284         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 285                 ret = -EFAULT;
 286
 287         kfree(kp);
 288
 289         return ret;
 290 }
 291
 292 static int
 293 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 294 {
 295         struct ip_tunnel_prl_entry *p;
 296         int err = 0;
 297
 298         if (a->addr == htonl(INADDR_ANY))
 299                 return -EINVAL;
 300
 301         write_lock(&ipip6_lock);
 302
 303         for (p = t->prl; p; p = p->next) {
 304                 if (p->addr == a->addr) {
 305                         if (chg)
 306                                 goto update;
 307                         err = -EEXIST;
 308                         goto out;
 309                 }
 310         }
 311
 312         if (chg) {
 313                 err = -ENXIO;
 314                 goto out;
 315         }
 316
 317         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 318         if (!p) {
 319                 err = -ENOBUFS;
 320                 goto out;
 321         }
 322
 323         p->next = t->prl;
 324         t->prl = p;
 325         t->prl_count++;
 326 update:
 327         p->addr = a->addr;
 328         p->flags = a->flags;
 329 out:
 330         write_unlock(&ipip6_lock);
 331         return err;
 332 }
 333
 334 static int
 335 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 336 {
 337         struct ip_tunnel_prl_entry *x, **p;
 338         int err = 0;
 339
 340         write_lock(&ipip6_lock);
 341
 342         if (a && a->addr != htonl(INADDR_ANY)) {
 343                 for (p = &t->prl; *p; p = &(*p)->next) {
 344                         if ((*p)->addr == a->addr) {
 345                                 x = *p;
 346                                 *p = x->next;
 347                                 kfree(x);
 348                                 t->prl_count--;
 349                                 goto out;
 350                         }
 351                 }
 352                 err = -ENXIO;
 353         } else {
 354                 while (t->prl) {
 355                         x = t->prl;
 356                         t->prl = t->prl->next;
 357                         kfree(x);
 358                         t->prl_count--;
 359                 }
 360         }
 361 out:
 362         write_unlock(&ipip6_lock);
 363         return 0;
 364 }
 365
 366 static int
 367 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 368 {
 369         struct ip_tunnel_prl_entry *p;
 370         int ok = 1;
 371
 372         read_lock(&ipip6_lock);
 373         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 374         if (p) {
 375                 if (p->flags & PRL_DEFAULT)
 376                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 377                 else
 378                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 379         } else {
 380                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 381                 if (ipv6_addr_is_isatap(addr6) &&
 382                     (addr6->s6_addr32[3] == iph->saddr) &&
 383                     ipv6_chk_prefix(addr6, t->dev))
 384                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 385                 else
 386                         ok = 0;
 387         }
 388         read_unlock(&ipip6_lock);
 389         return ok;
 390 }
 391
 392 static void ipip6_tunnel_uninit(struct net_device *dev)
 393 {
 394         struct net *net = dev_net(dev);
 395         struct sit_net *sitn = net_generic(net, sit_net_id);
 396
 397         if (dev == sitn->fb_tunnel_dev) {
 398                 write_lock_bh(&ipip6_lock);
 399                 sitn->tunnels_wc[0] = NULL;
 400                 write_unlock_bh(&ipip6_lock);
 401                 dev_put(dev);
 402         } else {
 403                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 404                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 405                 dev_put(dev);
 406         }
 407 }
 408
 409
 410 static int ipip6_err(struct sk_buff *skb, u32 info)
 411 {
 412
 413 /* All the routers (except for Linux) return only
 414    8 bytes of packet payload. It means, that precise relaying of
 415    ICMP in the real Internet is absolutely infeasible.
 416  */
 417         struct iphdr *iph = (struct iphdr*)skb->data;
 418         const int type = icmp_hdr(skb)->type;
 419         const int code = icmp_hdr(skb)->code;
 420         struct ip_tunnel *t;
 421         int err;
 422
 423         switch (type) {
 424         default:
 425         case ICMP_PARAMETERPROB:
 426                 return 0;
 427
 428         case ICMP_DEST_UNREACH:
 429                 switch (code) {
 430                 case ICMP_SR_FAILED:
 431                 case ICMP_PORT_UNREACH:
 432                         /* Impossible event. */
 433                         return 0;
 434                 case ICMP_FRAG_NEEDED:
 435                         /* Soft state for pmtu is maintained by IP core. */
 436                         return 0;
 437                 default:
 438                         /* All others are translated to HOST_UNREACH.
 439                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 440                            I believe they are just ether pollution. --ANK
 441                          */
 442                         break;
 443                 }
 444                 break;
 445         case ICMP_TIME_EXCEEDED:
 446                 if (code != ICMP_EXC_TTL)
 447                         return 0;
 448                 break;
 449         }
 450
 451         err = -ENOENT;
 452
 453         read_lock(&ipip6_lock);
 454         t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
 455         if (t == NULL || t->parms.iph.daddr == 0)
 456                 goto out;
 457
 458         err = 0;
 459         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 460                 goto out;
 461
 462         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 463                 t->err_count++;
 464         else
 465                 t->err_count = 1;
 466         t->err_time = jiffies;
 467 out:
 468         read_unlock(&ipip6_lock);
 469         return err;
 470 }
 471
 472 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 473 {
 474         if (INET_ECN_is_ce(iph->tos))
 475                 IP6_ECN_set_ce(ipv6_hdr(skb));
 476 }
 477
 478 static int ipip6_rcv(struct sk_buff *skb)
 479 {
 480         struct iphdr *iph;
 481         struct ip_tunnel *tunnel;
 482
 483         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 484                 goto out;
 485
 486         iph = ip_hdr(skb);
 487
 488         read_lock(&ipip6_lock);
 489         if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
 490                                         iph->saddr, iph->daddr)) != NULL) {
 491                 secpath_reset(skb);
 492                 skb->mac_header = skb->network_header;
 493                 skb_reset_network_header(skb);
 494                 IPCB(skb)->flags = 0;
 495                 skb->protocol = htons(ETH_P_IPV6);
 496                 skb->pkt_type = PACKET_HOST;
 497
 498                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 499                     !isatap_chksrc(skb, iph, tunnel)) {
 500                         tunnel->dev->stats.rx_errors++;
 501                         read_unlock(&ipip6_lock);
 502                         kfree_skb(skb);
 503                         return 0;
 504                 }
 505                 tunnel->dev->stats.rx_packets++;
 506                 tunnel->dev->stats.rx_bytes += skb->len;
 507                 skb->dev = tunnel->dev;
 508                 dst_release(skb->dst);
 509                 skb->dst = NULL;
 510                 nf_reset(skb);
 511                 ipip6_ecn_decapsulate(iph, skb);
 512                 netif_rx(skb);
 513                 read_unlock(&ipip6_lock);
 514                 return 0;
 515         }
 516
 517         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 518         read_unlock(&ipip6_lock);
 519 out:
 520         kfree_skb(skb);
 521         return 0;
 522 }
 523
 524 /* Returns the embedded IPv4 address if the IPv6 address
 525    comes from 6to4 (RFC 3056) addr space */
 526
 527 static inline __be32 try_6to4(struct in6_addr *v6dst)
 528 {
 529         __be32 dst = 0;
 530
 531         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 532                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 533                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 534         }
 535         return dst;
 536 }
 537
 538 /*
 539  *      This function assumes it is being called from dev_queue_xmit()
 540  *      and that skb is filled properly by that function.
 541  */
 542
 543 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 544 {
 545         struct ip_tunnel *tunnel = netdev_priv(dev);
 546         struct net_device_stats *stats = &tunnel->dev->stats;
 547         struct iphdr  *tiph = &tunnel->parms.iph;
 548         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 549         u8     tos = tunnel->parms.iph.tos;
 550         struct rtable *rt;                      /* Route to the other host */
 551         struct net_device *tdev;                        /* Device to other host */
 552         struct iphdr  *iph;                     /* Our new IP header */
 553         unsigned int max_headroom;              /* The extra header space needed */
 554         __be32 dst = tiph->daddr;
 555         int    mtu;
 556         struct in6_addr *addr6;
 557         int addr_type;
 558
 559         if (tunnel->recursion++) {
 560                 stats->collisions++;
 561                 goto tx_error;
 562         }
 563
 564         if (skb->protocol != htons(ETH_P_IPV6))
 565                 goto tx_error;
 566
 567         /* ISATAP (RFC4214) - must come before 6to4 */
 568         if (dev->priv_flags & IFF_ISATAP) {
 569                 struct neighbour *neigh = NULL;
 570
 571                 if (skb->dst)
 572                         neigh = skb->dst->neighbour;
 573
 574                 if (neigh == NULL) {
 575                         if (net_ratelimit())
 576                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 577                         goto tx_error;
 578                 }
 579
 580                 addr6 = (struct in6_addr*)&neigh->primary_key;
 581                 addr_type = ipv6_addr_type(addr6);
 582
 583                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 584                      ipv6_addr_is_isatap(addr6))
 585                         dst = addr6->s6_addr32[3];
 586                 else
 587                         goto tx_error;
 588         }
 589
 590         if (!dst)
 591                 dst = try_6to4(&iph6->daddr);
 592
 593         if (!dst) {
 594                 struct neighbour *neigh = NULL;
 595
 596                 if (skb->dst)
 597                         neigh = skb->dst->neighbour;
 598
 599                 if (neigh == NULL) {
 600                         if (net_ratelimit())
 601                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 602                         goto tx_error;
 603                 }
 604
 605                 addr6 = (struct in6_addr*)&neigh->primary_key;
 606                 addr_type = ipv6_addr_type(addr6);
 607
 608                 if (addr_type == IPV6_ADDR_ANY) {
 609                         addr6 = &ipv6_hdr(skb)->daddr;
 610                         addr_type = ipv6_addr_type(addr6);
 611                 }
 612
 613                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 614                         goto tx_error_icmp;
 615
 616                 dst = addr6->s6_addr32[3];
 617         }
 618
 619         {
 620                 struct flowi fl = { .nl_u = { .ip4_u =
 621                                               { .daddr = dst,
 622                                                 .saddr = tiph->saddr,
 623                                                 .tos = RT_TOS(tos) } },
 624                                     .oif = tunnel->parms.link,
 625                                     .proto = IPPROTO_IPV6 };
 626                 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
 627                         stats->tx_carrier_errors++;
 628                         goto tx_error_icmp;
 629                 }
 630         }
 631         if (rt->rt_type != RTN_UNICAST) {
 632                 ip_rt_put(rt);
 633                 stats->tx_carrier_errors++;
 634                 goto tx_error_icmp;
 635         }
 636         tdev = rt->u.dst.dev;
 637
 638         if (tdev == dev) {
 639                 ip_rt_put(rt);
 640                 stats->collisions++;
 641                 goto tx_error;
 642         }
 643
 644         if (tiph->frag_off)
 645                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 646         else
 647                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 648
 649         if (mtu < 68) {
 650                 stats->collisions++;
 651                 ip_rt_put(rt);
 652                 goto tx_error;
 653         }
 654         if (mtu < IPV6_MIN_MTU)
 655                 mtu = IPV6_MIN_MTU;
 656         if (tunnel->parms.iph.daddr && skb->dst)
 657                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 658
 659         if (skb->len > mtu) {
 660                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 661                 ip_rt_put(rt);
 662                 goto tx_error;
 663         }
 664
 665         if (tunnel->err_count > 0) {
 666                 if (time_before(jiffies,
 667                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 668                         tunnel->err_count--;
 669                         dst_link_failure(skb);
 670                 } else
 671                         tunnel->err_count = 0;
 672         }
 673
 674         /*
 675          * Okay, now see if we can stuff it in the buffer as-is.
 676          */
 677         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 678
 679         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 680             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 681                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 682                 if (!new_skb) {
 683                         ip_rt_put(rt);
 684                         stats->tx_dropped++;
 685                         dev_kfree_skb(skb);
 686                         tunnel->recursion--;
 687                         return 0;
 688                 }
 689                 if (skb->sk)
 690                         skb_set_owner_w(new_skb, skb->sk);
 691                 dev_kfree_skb(skb);
 692                 skb = new_skb;
 693                 iph6 = ipv6_hdr(skb);
 694         }
 695
 696         skb->transport_header = skb->network_header;
 697         skb_push(skb, sizeof(struct iphdr));
 698         skb_reset_network_header(skb);
 699         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 700         IPCB(skb)->flags = 0;
 701         dst_release(skb->dst);
 702         skb->dst = &rt->u.dst;
 703
 704         /*
 705          *      Push down and install the IPIP header.
 706          */
 707
 708         iph                     =       ip_hdr(skb);
 709         iph->version            =       4;
 710         iph->ihl                =       sizeof(struct iphdr)>>2;
 711         if (mtu > IPV6_MIN_MTU)
 712                 iph->frag_off   =       htons(IP_DF);
 713         else
 714                 iph->frag_off   =       0;
 715
 716         iph->protocol           =       IPPROTO_IPV6;
 717         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 718         iph->daddr              =       rt->rt_dst;
 719         iph->saddr              =       rt->rt_src;
 720
 721         if ((iph->ttl = tiph->ttl) == 0)
 722                 iph->ttl        =       iph6->hop_limit;
 723
 724         nf_reset(skb);
 725
 726         IPTUNNEL_XMIT();
 727         tunnel->recursion--;
 728         return 0;
 729
 730 tx_error_icmp:
 731         dst_link_failure(skb);
 732 tx_error:
 733         stats->tx_errors++;
 734         dev_kfree_skb(skb);
 735         tunnel->recursion--;
 736         return 0;
 737 }
 738
 739 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 740 {
 741         struct net_device *tdev = NULL;
 742         struct ip_tunnel *tunnel;
 743         struct iphdr *iph;
 744
 745         tunnel = netdev_priv(dev);
 746         iph = &tunnel->parms.iph;
 747
 748         if (iph->daddr) {
 749                 struct flowi fl = { .nl_u = { .ip4_u =
 750                                               { .daddr = iph->daddr,
 751                                                 .saddr = iph->saddr,
 752                                                 .tos = RT_TOS(iph->tos) } },
 753                                     .oif = tunnel->parms.link,
 754                                     .proto = IPPROTO_IPV6 };
 755                 struct rtable *rt;
 756                 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
 757                         tdev = rt->u.dst.dev;
 758                         ip_rt_put(rt);
 759                 }
 760                 dev->flags |= IFF_POINTOPOINT;
 761         }
 762
 763         if (!tdev && tunnel->parms.link)
 764                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 765
 766         if (tdev) {
 767                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 768                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 769                 if (dev->mtu < IPV6_MIN_MTU)
 770                         dev->mtu = IPV6_MIN_MTU;
 771         }
 772         dev->iflink = tunnel->parms.link;
 773 }
 774
 775 static int
 776 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 777 {
 778         int err = 0;
 779         struct ip_tunnel_parm p;
 780         struct ip_tunnel_prl prl;
 781         struct ip_tunnel *t;
 782         struct net *net = dev_net(dev);
 783         struct sit_net *sitn = net_generic(net, sit_net_id);
 784
 785         switch (cmd) {
 786         case SIOCGETTUNNEL:
 787                 t = NULL;
 788                 if (dev == sitn->fb_tunnel_dev) {
 789                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 790                                 err = -EFAULT;
 791                                 break;
 792                         }
 793                         t = ipip6_tunnel_locate(net, &p, 0);
 794                 }
 795                 if (t == NULL)
 796                         t = netdev_priv(dev);
 797                 memcpy(&p, &t->parms, sizeof(p));
 798                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 799                         err = -EFAULT;
 800                 break;
 801
 802         case SIOCADDTUNNEL:
 803         case SIOCCHGTUNNEL:
 804                 err = -EPERM;
 805                 if (!capable(CAP_NET_ADMIN))
 806                         goto done;
 807
 808                 err = -EFAULT;
 809                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 810                         goto done;
 811
 812                 err = -EINVAL;
 813                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 814                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 815                         goto done;
 816                 if (p.iph.ttl)
 817                         p.iph.frag_off |= htons(IP_DF);
 818
 819                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 820
 821                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 822                         if (t != NULL) {
 823                                 if (t->dev != dev) {
 824                                         err = -EEXIST;
 825                                         break;
 826                                 }
 827                         } else {
 828                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 829                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 830                                         err = -EINVAL;
 831                                         break;
 832                                 }
 833                                 t = netdev_priv(dev);
 834                                 ipip6_tunnel_unlink(sitn, t);
 835                                 t->parms.iph.saddr = p.iph.saddr;
 836                                 t->parms.iph.daddr = p.iph.daddr;
 837                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 838                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 839                                 ipip6_tunnel_link(sitn, t);
 840                                 netdev_state_change(dev);
 841                         }
 842                 }
 843
 844                 if (t) {
 845                         err = 0;
 846                         if (cmd == SIOCCHGTUNNEL) {
 847                                 t->parms.iph.ttl = p.iph.ttl;
 848                                 t->parms.iph.tos = p.iph.tos;
 849                                 if (t->parms.link != p.link) {
 850                                         t->parms.link = p.link;
 851                                         ipip6_tunnel_bind_dev(dev);
 852                                         netdev_state_change(dev);
 853                                 }
 854                         }
 855                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 856                                 err = -EFAULT;
 857                 } else
 858                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 859                 break;
 860
 861         case SIOCDELTUNNEL:
 862                 err = -EPERM;
 863                 if (!capable(CAP_NET_ADMIN))
 864                         goto done;
 865
 866                 if (dev == sitn->fb_tunnel_dev) {
 867                         err = -EFAULT;
 868                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 869                                 goto done;
 870                         err = -ENOENT;
 871                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
 872                                 goto done;
 873                         err = -EPERM;
 874                         if (t == netdev_priv(sitn->fb_tunnel_dev))
 875                                 goto done;
 876                         dev = t->dev;
 877                 }
 878                 unregister_netdevice(dev);
 879                 err = 0;
 880                 break;
 881
 882         case SIOCGETPRL:
 883                 err = -EINVAL;
 884                 if (dev == sitn->fb_tunnel_dev)
 885                         goto done;
 886                 err = -ENOENT;
 887                 if (!(t = netdev_priv(dev)))
 888                         goto done;
 889                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
 890                 break;
 891
 892         case SIOCADDPRL:
 893         case SIOCDELPRL:
 894         case SIOCCHGPRL:
 895                 err = -EPERM;
 896                 if (!capable(CAP_NET_ADMIN))
 897                         goto done;
 898                 err = -EINVAL;
 899                 if (dev == sitn->fb_tunnel_dev)
 900                         goto done;
 901                 err = -EFAULT;
 902                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 903                         goto done;
 904                 err = -ENOENT;
 905                 if (!(t = netdev_priv(dev)))
 906                         goto done;
 907
 908                 switch (cmd) {
 909                 case SIOCDELPRL:
 910                         err = ipip6_tunnel_del_prl(t, &prl);
 911                         break;
 912                 case SIOCADDPRL:
 913                 case SIOCCHGPRL:
 914                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 915                         break;
 916                 }
 917                 netdev_state_change(dev);
 918                 break;
 919
 920         default:
 921                 err = -EINVAL;
 922         }
 923
 924 done:
 925         return err;
 926 }
 927
 928 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 929 {
 930         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
 931                 return -EINVAL;
 932         dev->mtu = new_mtu;
 933         return 0;
 934 }
 935
 936 static const struct net_device_ops ipip6_netdev_ops = {
 937         .ndo_uninit     = ipip6_tunnel_uninit,
 938         .ndo_start_xmit = ipip6_tunnel_xmit,
 939         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
 940         .ndo_change_mtu = ipip6_tunnel_change_mtu,
 941 };
 942
 943 static void ipip6_tunnel_setup(struct net_device *dev)
 944 {
 945         dev->netdev_ops         = &ipip6_netdev_ops;
 946         dev->destructor         = free_netdev;
 947
 948         dev->type               = ARPHRD_SIT;
 949         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 950         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
 951         dev->flags              = IFF_NOARP;
 952         dev->iflink             = 0;
 953         dev->addr_len           = 4;
 954         dev->features           |= NETIF_F_NETNS_LOCAL;
 955 }
 956
 957 static void ipip6_tunnel_init(struct net_device *dev)
 958 {
 959         struct ip_tunnel *tunnel = netdev_priv(dev);
 960
 961         tunnel->dev = dev;
 962         strcpy(tunnel->parms.name, dev->name);
 963
 964         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
 965         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 966
 967         ipip6_tunnel_bind_dev(dev);
 968 }
 969
 970 static void ipip6_fb_tunnel_init(struct net_device *dev)
 971 {
 972         struct ip_tunnel *tunnel = netdev_priv(dev);
 973         struct iphdr *iph = &tunnel->parms.iph;
 974         struct net *net = dev_net(dev);
 975         struct sit_net *sitn = net_generic(net, sit_net_id);
 976
 977         tunnel->dev = dev;
 978         strcpy(tunnel->parms.name, dev->name);
 979
 980         iph->version            = 4;
 981         iph->protocol           = IPPROTO_IPV6;
 982         iph->ihl                = 5;
 983         iph->ttl                = 64;
 984
 985         dev_hold(dev);
 986         sitn->tunnels_wc[0]     = tunnel;
 987 }
 988
 989 static struct xfrm_tunnel sit_handler = {
 990         .handler        =       ipip6_rcv,
 991         .err_handler    =       ipip6_err,
 992         .priority       =       1,
 993 };
 994
 995 static void sit_destroy_tunnels(struct sit_net *sitn)
 996 {
 997         int prio;
 998
 999         for (prio = 1; prio < 4; prio++) {
1000                 int h;
1001                 for (h = 0; h < HASH_SIZE; h++) {
1002                         struct ip_tunnel *t;
1003                         while ((t = sitn->tunnels[prio][h]) != NULL)
1004                                 unregister_netdevice(t->dev);
1005                 }
1006         }
1007 }
1008
1009 static int sit_init_net(struct net *net)
1010 {
1011         int err;
1012         struct sit_net *sitn;
1013
1014         err = -ENOMEM;
1015         sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1016         if (sitn == NULL)
1017                 goto err_alloc;
1018
1019         err = net_assign_generic(net, sit_net_id, sitn);
1020         if (err < 0)
1021                 goto err_assign;
1022
1023         sitn->tunnels[0] = sitn->tunnels_wc;
1024         sitn->tunnels[1] = sitn->tunnels_l;
1025         sitn->tunnels[2] = sitn->tunnels_r;
1026         sitn->tunnels[3] = sitn->tunnels_r_l;
1027
1028         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1029                                            ipip6_tunnel_setup);
1030         if (!sitn->fb_tunnel_dev) {
1031                 err = -ENOMEM;
1032                 goto err_alloc_dev;
1033         }
1034         dev_net_set(sitn->fb_tunnel_dev, net);
1035
1036         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1037
1038         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1039                 goto err_reg_dev;
1040
1041         return 0;
1042
1043 err_reg_dev:
1044         dev_put(sitn->fb_tunnel_dev);
1045         free_netdev(sitn->fb_tunnel_dev);
1046 err_alloc_dev:
1047         /* nothing */
1048 err_assign:
1049         kfree(sitn);
1050 err_alloc:
1051         return err;
1052 }
1053
1054 static void sit_exit_net(struct net *net)
1055 {
1056         struct sit_net *sitn;
1057
1058         sitn = net_generic(net, sit_net_id);
1059         rtnl_lock();
1060         sit_destroy_tunnels(sitn);
1061         unregister_netdevice(sitn->fb_tunnel_dev);
1062         rtnl_unlock();
1063         kfree(sitn);
1064 }
1065
1066 static struct pernet_operations sit_net_ops = {
1067         .init = sit_init_net,
1068         .exit = sit_exit_net,
1069 };
1070
1071 static void __exit sit_cleanup(void)
1072 {
1073         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1074
1075         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1076 }
1077
1078 static int __init sit_init(void)
1079 {
1080         int err;
1081
1082         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1083
1084         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1085                 printk(KERN_INFO "sit init: Can't add protocol\n");
1086                 return -EAGAIN;
1087         }
1088
1089         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1090         if (err < 0)
1091                 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1092
1093         return err;
1094 }
1095
1096 module_init(sit_init);
1097 module_exit(sit_cleanup);
1098 MODULE_LICENSE("GPL");
1099 MODULE_ALIAS("sit0");