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