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                 struct net_device *dev, __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 &&
  93                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
  94                     (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 &&
  99                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 100                     (t->dev->flags & IFF_UP))
 101                         return t;
 102         }
 103         for (t = sitn->tunnels_l[h1]; t; t = t->next) {
 104                 if (local == t->parms.iph.saddr &&
 105                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 106                     (t->dev->flags & IFF_UP))
 107                         return t;
 108         }
 109         t = sitn->tunnels_wc[0];
 110         if ((t != NULL) && (t->dev->flags & IFF_UP))
 111                 return t;
 112         return NULL;
 113 }
 114
 115 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
 116                 struct ip_tunnel_parm *parms)
 117 {
 118         __be32 remote = parms->iph.daddr;
 119         __be32 local = parms->iph.saddr;
 120         unsigned h = 0;
 121         int prio = 0;
 122
 123         if (remote) {
 124                 prio |= 2;
 125                 h ^= HASH(remote);
 126         }
 127         if (local) {
 128                 prio |= 1;
 129                 h ^= HASH(local);
 130         }
 131         return &sitn->tunnels[prio][h];
 132 }
 133
 134 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
 135                 struct ip_tunnel *t)
 136 {
 137         return __ipip6_bucket(sitn, &t->parms);
 138 }
 139
 140 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 141 {
 142         struct ip_tunnel **tp;
 143
 144         for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
 145                 if (t == *tp) {
 146                         write_lock_bh(&ipip6_lock);
 147                         *tp = t->next;
 148                         write_unlock_bh(&ipip6_lock);
 149                         break;
 150                 }
 151         }
 152 }
 153
 154 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 155 {
 156         struct ip_tunnel **tp = ipip6_bucket(sitn, t);
 157
 158         t->next = *tp;
 159         write_lock_bh(&ipip6_lock);
 160         *tp = t;
 161         write_unlock_bh(&ipip6_lock);
 162 }
 163
 164 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 165 {
 166 #ifdef CONFIG_IPV6_SIT_6RD
 167         struct ip_tunnel *t = netdev_priv(dev);
 168
 169         if (t->dev == sitn->fb_tunnel_dev) {
 170                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 171                 t->ip6rd.relay_prefix = 0;
 172                 t->ip6rd.prefixlen = 16;
 173                 t->ip6rd.relay_prefixlen = 0;
 174         } else {
 175                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 176                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 177         }
 178 #endif
 179 }
 180
 181 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
 182                 struct ip_tunnel_parm *parms, int create)
 183 {
 184         __be32 remote = parms->iph.daddr;
 185         __be32 local = parms->iph.saddr;
 186         struct ip_tunnel *t, **tp, *nt;
 187         struct net_device *dev;
 188         char name[IFNAMSIZ];
 189         struct sit_net *sitn = net_generic(net, sit_net_id);
 190
 191         for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
 192                 if (local == t->parms.iph.saddr &&
 193                     remote == t->parms.iph.daddr &&
 194                     parms->link == t->parms.link) {
 195                         if (create)
 196                                 return NULL;
 197                         else
 198                                 return t;
 199                 }
 200         }
 201         if (!create)
 202                 goto failed;
 203
 204         if (parms->name[0])
 205                 strlcpy(name, parms->name, IFNAMSIZ);
 206         else
 207                 sprintf(name, "sit%%d");
 208
 209         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 210         if (dev == NULL)
 211                 return NULL;
 212
 213         dev_net_set(dev, net);
 214
 215         if (strchr(name, '%')) {
 216                 if (dev_alloc_name(dev, name) < 0)
 217                         goto failed_free;
 218         }
 219
 220         nt = netdev_priv(dev);
 221
 222         nt->parms = *parms;
 223         ipip6_tunnel_init(dev);
 224         ipip6_tunnel_clone_6rd(dev, sitn);
 225
 226         if (parms->i_flags & SIT_ISATAP)
 227                 dev->priv_flags |= IFF_ISATAP;
 228
 229         if (register_netdevice(dev) < 0)
 230                 goto failed_free;
 231
 232         dev_hold(dev);
 233
 234         ipip6_tunnel_link(sitn, nt);
 235         return nt;
 236
 237 failed_free:
 238         free_netdev(dev);
 239 failed:
 240         return NULL;
 241 }
 242
 243 static struct ip_tunnel_prl_entry *
 244 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 245 {
 246         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 247
 248         for (p = t->prl; p; p = p->next)
 249                 if (p->addr == addr)
 250                         break;
 251         return p;
 252
 253 }
 254
 255 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 256                                 struct ip_tunnel_prl __user *a)
 257 {
 258         struct ip_tunnel_prl kprl, *kp;
 259         struct ip_tunnel_prl_entry *prl;
 260         unsigned int cmax, c = 0, ca, len;
 261         int ret = 0;
 262
 263         if (copy_from_user(&kprl, a, sizeof(kprl)))
 264                 return -EFAULT;
 265         cmax = kprl.datalen / sizeof(kprl);
 266         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 267                 cmax = 1;
 268
 269         /* For simple GET or for root users,
 270          * we try harder to allocate.
 271          */
 272         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 273                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 274                 NULL;
 275
 276         read_lock(&ipip6_lock);
 277
 278         ca = t->prl_count < cmax ? t->prl_count : cmax;
 279
 280         if (!kp) {
 281                 /* We don't try hard to allocate much memory for
 282                  * non-root users.
 283                  * For root users, retry allocating enough memory for
 284                  * the answer.
 285                  */
 286                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 287                 if (!kp) {
 288                         ret = -ENOMEM;
 289                         goto out;
 290                 }
 291         }
 292
 293         c = 0;
 294         for (prl = t->prl; prl; prl = prl->next) {
 295                 if (c >= cmax)
 296                         break;
 297                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 298                         continue;
 299                 kp[c].addr = prl->addr;
 300                 kp[c].flags = prl->flags;
 301                 c++;
 302                 if (kprl.addr != htonl(INADDR_ANY))
 303                         break;
 304         }
 305 out:
 306         read_unlock(&ipip6_lock);
 307
 308         len = sizeof(*kp) * c;
 309         ret = 0;
 310         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 311                 ret = -EFAULT;
 312
 313         kfree(kp);
 314
 315         return ret;
 316 }
 317
 318 static int
 319 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 320 {
 321         struct ip_tunnel_prl_entry *p;
 322         int err = 0;
 323
 324         if (a->addr == htonl(INADDR_ANY))
 325                 return -EINVAL;
 326
 327         write_lock(&ipip6_lock);
 328
 329         for (p = t->prl; p; p = p->next) {
 330                 if (p->addr == a->addr) {
 331                         if (chg)
 332                                 goto update;
 333                         err = -EEXIST;
 334                         goto out;
 335                 }
 336         }
 337
 338         if (chg) {
 339                 err = -ENXIO;
 340                 goto out;
 341         }
 342
 343         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 344         if (!p) {
 345                 err = -ENOBUFS;
 346                 goto out;
 347         }
 348
 349         p->next = t->prl;
 350         t->prl = p;
 351         t->prl_count++;
 352 update:
 353         p->addr = a->addr;
 354         p->flags = a->flags;
 355 out:
 356         write_unlock(&ipip6_lock);
 357         return err;
 358 }
 359
 360 static int
 361 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 362 {
 363         struct ip_tunnel_prl_entry *x, **p;
 364         int err = 0;
 365
 366         write_lock(&ipip6_lock);
 367
 368         if (a && a->addr != htonl(INADDR_ANY)) {
 369                 for (p = &t->prl; *p; p = &(*p)->next) {
 370                         if ((*p)->addr == a->addr) {
 371                                 x = *p;
 372                                 *p = x->next;
 373                                 kfree(x);
 374                                 t->prl_count--;
 375                                 goto out;
 376                         }
 377                 }
 378                 err = -ENXIO;
 379         } else {
 380                 while (t->prl) {
 381                         x = t->prl;
 382                         t->prl = t->prl->next;
 383                         kfree(x);
 384                         t->prl_count--;
 385                 }
 386         }
 387 out:
 388         write_unlock(&ipip6_lock);
 389         return err;
 390 }
 391
 392 static int
 393 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 394 {
 395         struct ip_tunnel_prl_entry *p;
 396         int ok = 1;
 397
 398         read_lock(&ipip6_lock);
 399         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 400         if (p) {
 401                 if (p->flags & PRL_DEFAULT)
 402                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 403                 else
 404                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 405         } else {
 406                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 407                 if (ipv6_addr_is_isatap(addr6) &&
 408                     (addr6->s6_addr32[3] == iph->saddr) &&
 409                     ipv6_chk_prefix(addr6, t->dev))
 410                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 411                 else
 412                         ok = 0;
 413         }
 414         read_unlock(&ipip6_lock);
 415         return ok;
 416 }
 417
 418 static void ipip6_tunnel_uninit(struct net_device *dev)
 419 {
 420         struct net *net = dev_net(dev);
 421         struct sit_net *sitn = net_generic(net, sit_net_id);
 422
 423         if (dev == sitn->fb_tunnel_dev) {
 424                 write_lock_bh(&ipip6_lock);
 425                 sitn->tunnels_wc[0] = NULL;
 426                 write_unlock_bh(&ipip6_lock);
 427                 dev_put(dev);
 428         } else {
 429                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 430                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 431                 dev_put(dev);
 432         }
 433 }
 434
 435
 436 static int ipip6_err(struct sk_buff *skb, u32 info)
 437 {
 438
 439 /* All the routers (except for Linux) return only
 440    8 bytes of packet payload. It means, that precise relaying of
 441    ICMP in the real Internet is absolutely infeasible.
 442  */
 443         struct iphdr *iph = (struct iphdr*)skb->data;
 444         const int type = icmp_hdr(skb)->type;
 445         const int code = icmp_hdr(skb)->code;
 446         struct ip_tunnel *t;
 447         int err;
 448
 449         switch (type) {
 450         default:
 451         case ICMP_PARAMETERPROB:
 452                 return 0;
 453
 454         case ICMP_DEST_UNREACH:
 455                 switch (code) {
 456                 case ICMP_SR_FAILED:
 457                 case ICMP_PORT_UNREACH:
 458                         /* Impossible event. */
 459                         return 0;
 460                 case ICMP_FRAG_NEEDED:
 461                         /* Soft state for pmtu is maintained by IP core. */
 462                         return 0;
 463                 default:
 464                         /* All others are translated to HOST_UNREACH.
 465                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 466                            I believe they are just ether pollution. --ANK
 467                          */
 468                         break;
 469                 }
 470                 break;
 471         case ICMP_TIME_EXCEEDED:
 472                 if (code != ICMP_EXC_TTL)
 473                         return 0;
 474                 break;
 475         }
 476
 477         err = -ENOENT;
 478
 479         read_lock(&ipip6_lock);
 480         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 481                                 skb->dev,
 482                                 iph->daddr,
 483                                 iph->saddr);
 484         if (t == NULL || t->parms.iph.daddr == 0)
 485                 goto out;
 486
 487         err = 0;
 488         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 489                 goto out;
 490
 491         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 492                 t->err_count++;
 493         else
 494                 t->err_count = 1;
 495         t->err_time = jiffies;
 496 out:
 497         read_unlock(&ipip6_lock);
 498         return err;
 499 }
 500
 501 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 502 {
 503         if (INET_ECN_is_ce(iph->tos))
 504                 IP6_ECN_set_ce(ipv6_hdr(skb));
 505 }
 506
 507 static int ipip6_rcv(struct sk_buff *skb)
 508 {
 509         struct iphdr *iph;
 510         struct ip_tunnel *tunnel;
 511
 512         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 513                 goto out;
 514
 515         iph = ip_hdr(skb);
 516
 517         read_lock(&ipip6_lock);
 518         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 519                                      iph->saddr, iph->daddr);
 520         if (tunnel != NULL) {
 521                 secpath_reset(skb);
 522                 skb->mac_header = skb->network_header;
 523                 skb_reset_network_header(skb);
 524                 IPCB(skb)->flags = 0;
 525                 skb->protocol = htons(ETH_P_IPV6);
 526                 skb->pkt_type = PACKET_HOST;
 527
 528                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 529                     !isatap_chksrc(skb, iph, tunnel)) {
 530                         tunnel->dev->stats.rx_errors++;
 531                         read_unlock(&ipip6_lock);
 532                         kfree_skb(skb);
 533                         return 0;
 534                 }
 535                 tunnel->dev->stats.rx_packets++;
 536                 tunnel->dev->stats.rx_bytes += skb->len;
 537                 skb->dev = tunnel->dev;
 538                 skb_dst_drop(skb);
 539                 nf_reset(skb);
 540                 ipip6_ecn_decapsulate(iph, skb);
 541                 netif_rx(skb);
 542                 read_unlock(&ipip6_lock);
 543                 return 0;
 544         }
 545
 546         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 547         read_unlock(&ipip6_lock);
 548 out:
 549         kfree_skb(skb);
 550         return 0;
 551 }
 552
 553 /*
 554  * Returns the embedded IPv4 address if the IPv6 address
 555  * comes from 6rd / 6to4 (RFC 3056) addr space.
 556  */
 557 static inline
 558 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 559 {
 560         __be32 dst = 0;
 561
 562 #ifdef CONFIG_IPV6_SIT_6RD
 563         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 564                               tunnel->ip6rd.prefixlen)) {
 565                 unsigned pbw0, pbi0;
 566                 int pbi1;
 567                 u32 d;
 568
 569                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 570                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 571
 572                 d = (ntohl(tunnel->ip6rd.prefix.s6_addr32[pbw0]) << pbi0) >>
 573                     tunnel->ip6rd.relay_prefixlen;
 574
 575                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 576                 if (pbi1 > 0)
 577                         d |= ntohl(tunnel->ip6rd.prefix.s6_addr32[pbw0 + 1]) >>
 578                              (32 - pbi1);
 579
 580                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
 581         }
 582 #else
 583         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 584                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 585                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 586         }
 587 #endif
 588         return dst;
 589 }
 590
 591 /*
 592  *      This function assumes it is being called from dev_queue_xmit()
 593  *      and that skb is filled properly by that function.
 594  */
 595
 596 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 597                                      struct net_device *dev)
 598 {
 599         struct ip_tunnel *tunnel = netdev_priv(dev);
 600         struct net_device_stats *stats = &dev->stats;
 601         struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
 602         struct iphdr  *tiph = &tunnel->parms.iph;
 603         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 604         u8     tos = tunnel->parms.iph.tos;
 605         struct rtable *rt;                      /* Route to the other host */
 606         struct net_device *tdev;                        /* Device to other host */
 607         struct iphdr  *iph;                     /* Our new IP header */
 608         unsigned int max_headroom;              /* The extra header space needed */
 609         __be32 dst = tiph->daddr;
 610         int    mtu;
 611         struct in6_addr *addr6;
 612         int addr_type;
 613
 614         if (skb->protocol != htons(ETH_P_IPV6))
 615                 goto tx_error;
 616
 617         /* ISATAP (RFC4214) - must come before 6to4 */
 618         if (dev->priv_flags & IFF_ISATAP) {
 619                 struct neighbour *neigh = NULL;
 620
 621                 if (skb_dst(skb))
 622                         neigh = skb_dst(skb)->neighbour;
 623
 624                 if (neigh == NULL) {
 625                         if (net_ratelimit())
 626                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 627                         goto tx_error;
 628                 }
 629
 630                 addr6 = (struct in6_addr*)&neigh->primary_key;
 631                 addr_type = ipv6_addr_type(addr6);
 632
 633                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 634                      ipv6_addr_is_isatap(addr6))
 635                         dst = addr6->s6_addr32[3];
 636                 else
 637                         goto tx_error;
 638         }
 639
 640         if (!dst)
 641                 dst = try_6rd(&iph6->daddr, tunnel);
 642
 643         if (!dst) {
 644                 struct neighbour *neigh = NULL;
 645
 646                 if (skb_dst(skb))
 647                         neigh = skb_dst(skb)->neighbour;
 648
 649                 if (neigh == NULL) {
 650                         if (net_ratelimit())
 651                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 652                         goto tx_error;
 653                 }
 654
 655                 addr6 = (struct in6_addr*)&neigh->primary_key;
 656                 addr_type = ipv6_addr_type(addr6);
 657
 658                 if (addr_type == IPV6_ADDR_ANY) {
 659                         addr6 = &ipv6_hdr(skb)->daddr;
 660                         addr_type = ipv6_addr_type(addr6);
 661                 }
 662
 663                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 664                         goto tx_error_icmp;
 665
 666                 dst = addr6->s6_addr32[3];
 667         }
 668
 669         {
 670                 struct flowi fl = { .nl_u = { .ip4_u =
 671                                               { .daddr = dst,
 672                                                 .saddr = tiph->saddr,
 673                                                 .tos = RT_TOS(tos) } },
 674                                     .oif = tunnel->parms.link,
 675                                     .proto = IPPROTO_IPV6 };
 676                 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
 677                         stats->tx_carrier_errors++;
 678                         goto tx_error_icmp;
 679                 }
 680         }
 681         if (rt->rt_type != RTN_UNICAST) {
 682                 ip_rt_put(rt);
 683                 stats->tx_carrier_errors++;
 684                 goto tx_error_icmp;
 685         }
 686         tdev = rt->u.dst.dev;
 687
 688         if (tdev == dev) {
 689                 ip_rt_put(rt);
 690                 stats->collisions++;
 691                 goto tx_error;
 692         }
 693
 694         if (tiph->frag_off)
 695                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 696         else
 697                 mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 698
 699         if (mtu < 68) {
 700                 stats->collisions++;
 701                 ip_rt_put(rt);
 702                 goto tx_error;
 703         }
 704         if (mtu < IPV6_MIN_MTU)
 705                 mtu = IPV6_MIN_MTU;
 706         if (tunnel->parms.iph.daddr && skb_dst(skb))
 707                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
 708
 709         if (skb->len > mtu) {
 710                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 711                 ip_rt_put(rt);
 712                 goto tx_error;
 713         }
 714
 715         if (tunnel->err_count > 0) {
 716                 if (time_before(jiffies,
 717                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 718                         tunnel->err_count--;
 719                         dst_link_failure(skb);
 720                 } else
 721                         tunnel->err_count = 0;
 722         }
 723
 724         /*
 725          * Okay, now see if we can stuff it in the buffer as-is.
 726          */
 727         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 728
 729         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 730             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 731                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 732                 if (!new_skb) {
 733                         ip_rt_put(rt);
 734                         txq->tx_dropped++;
 735                         dev_kfree_skb(skb);
 736                         return NETDEV_TX_OK;
 737                 }
 738                 if (skb->sk)
 739                         skb_set_owner_w(new_skb, skb->sk);
 740                 dev_kfree_skb(skb);
 741                 skb = new_skb;
 742                 iph6 = ipv6_hdr(skb);
 743         }
 744
 745         skb->transport_header = skb->network_header;
 746         skb_push(skb, sizeof(struct iphdr));
 747         skb_reset_network_header(skb);
 748         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 749         IPCB(skb)->flags = 0;
 750         skb_dst_drop(skb);
 751         skb_dst_set(skb, &rt->u.dst);
 752
 753         /*
 754          *      Push down and install the IPIP header.
 755          */
 756
 757         iph                     =       ip_hdr(skb);
 758         iph->version            =       4;
 759         iph->ihl                =       sizeof(struct iphdr)>>2;
 760         if (mtu > IPV6_MIN_MTU)
 761                 iph->frag_off   =       tiph->frag_off;
 762         else
 763                 iph->frag_off   =       0;
 764
 765         iph->protocol           =       IPPROTO_IPV6;
 766         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 767         iph->daddr              =       rt->rt_dst;
 768         iph->saddr              =       rt->rt_src;
 769
 770         if ((iph->ttl = tiph->ttl) == 0)
 771                 iph->ttl        =       iph6->hop_limit;
 772
 773         nf_reset(skb);
 774
 775         IPTUNNEL_XMIT();
 776         return NETDEV_TX_OK;
 777
 778 tx_error_icmp:
 779         dst_link_failure(skb);
 780 tx_error:
 781         stats->tx_errors++;
 782         dev_kfree_skb(skb);
 783         return NETDEV_TX_OK;
 784 }
 785
 786 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 787 {
 788         struct net_device *tdev = NULL;
 789         struct ip_tunnel *tunnel;
 790         struct iphdr *iph;
 791
 792         tunnel = netdev_priv(dev);
 793         iph = &tunnel->parms.iph;
 794
 795         if (iph->daddr) {
 796                 struct flowi fl = { .nl_u = { .ip4_u =
 797                                               { .daddr = iph->daddr,
 798                                                 .saddr = iph->saddr,
 799                                                 .tos = RT_TOS(iph->tos) } },
 800                                     .oif = tunnel->parms.link,
 801                                     .proto = IPPROTO_IPV6 };
 802                 struct rtable *rt;
 803                 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
 804                         tdev = rt->u.dst.dev;
 805                         ip_rt_put(rt);
 806                 }
 807                 dev->flags |= IFF_POINTOPOINT;
 808         }
 809
 810         if (!tdev && tunnel->parms.link)
 811                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 812
 813         if (tdev) {
 814                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 815                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 816                 if (dev->mtu < IPV6_MIN_MTU)
 817                         dev->mtu = IPV6_MIN_MTU;
 818         }
 819         dev->iflink = tunnel->parms.link;
 820 }
 821
 822 static int
 823 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 824 {
 825         int err = 0;
 826         struct ip_tunnel_parm p;
 827         struct ip_tunnel_prl prl;
 828         struct ip_tunnel *t;
 829         struct net *net = dev_net(dev);
 830         struct sit_net *sitn = net_generic(net, sit_net_id);
 831 #ifdef CONFIG_IPV6_SIT_6RD
 832         struct ip_tunnel_6rd ip6rd;
 833 #endif
 834
 835         switch (cmd) {
 836         case SIOCGETTUNNEL:
 837 #ifdef CONFIG_IPV6_SIT_6RD
 838         case SIOCGET6RD:
 839 #endif
 840                 t = NULL;
 841                 if (dev == sitn->fb_tunnel_dev) {
 842                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 843                                 err = -EFAULT;
 844                                 break;
 845                         }
 846                         t = ipip6_tunnel_locate(net, &p, 0);
 847                 }
 848                 if (t == NULL)
 849                         t = netdev_priv(dev);
 850
 851                 err = -EFAULT;
 852                 if (cmd == SIOCGETTUNNEL) {
 853                         memcpy(&p, &t->parms, sizeof(p));
 854                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
 855                                          sizeof(p)))
 856                                 goto done;
 857 #ifdef CONFIG_IPV6_SIT_6RD
 858                 } else {
 859                         ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
 860                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
 861                         ip6rd.prefixlen = t->ip6rd.prefixlen;
 862                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
 863                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
 864                                          sizeof(ip6rd)))
 865                                 goto done;
 866 #endif
 867                 }
 868                 err = 0;
 869                 break;
 870
 871         case SIOCADDTUNNEL:
 872         case SIOCCHGTUNNEL:
 873                 err = -EPERM;
 874                 if (!capable(CAP_NET_ADMIN))
 875                         goto done;
 876
 877                 err = -EFAULT;
 878                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 879                         goto done;
 880
 881                 err = -EINVAL;
 882                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 883                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 884                         goto done;
 885                 if (p.iph.ttl)
 886                         p.iph.frag_off |= htons(IP_DF);
 887
 888                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 889
 890                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 891                         if (t != NULL) {
 892                                 if (t->dev != dev) {
 893                                         err = -EEXIST;
 894                                         break;
 895                                 }
 896                         } else {
 897                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 898                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 899                                         err = -EINVAL;
 900                                         break;
 901                                 }
 902                                 t = netdev_priv(dev);
 903                                 ipip6_tunnel_unlink(sitn, t);
 904                                 t->parms.iph.saddr = p.iph.saddr;
 905                                 t->parms.iph.daddr = p.iph.daddr;
 906                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 907                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 908                                 ipip6_tunnel_link(sitn, t);
 909                                 netdev_state_change(dev);
 910                         }
 911                 }
 912
 913                 if (t) {
 914                         err = 0;
 915                         if (cmd == SIOCCHGTUNNEL) {
 916                                 t->parms.iph.ttl = p.iph.ttl;
 917                                 t->parms.iph.tos = p.iph.tos;
 918                                 if (t->parms.link != p.link) {
 919                                         t->parms.link = p.link;
 920                                         ipip6_tunnel_bind_dev(dev);
 921                                         netdev_state_change(dev);
 922                                 }
 923                         }
 924                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 925                                 err = -EFAULT;
 926                 } else
 927                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 928                 break;
 929
 930         case SIOCDELTUNNEL:
 931                 err = -EPERM;
 932                 if (!capable(CAP_NET_ADMIN))
 933                         goto done;
 934
 935                 if (dev == sitn->fb_tunnel_dev) {
 936                         err = -EFAULT;
 937                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 938                                 goto done;
 939                         err = -ENOENT;
 940                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
 941                                 goto done;
 942                         err = -EPERM;
 943                         if (t == netdev_priv(sitn->fb_tunnel_dev))
 944                                 goto done;
 945                         dev = t->dev;
 946                 }
 947                 unregister_netdevice(dev);
 948                 err = 0;
 949                 break;
 950
 951         case SIOCGETPRL:
 952                 err = -EINVAL;
 953                 if (dev == sitn->fb_tunnel_dev)
 954                         goto done;
 955                 err = -ENOENT;
 956                 if (!(t = netdev_priv(dev)))
 957                         goto done;
 958                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
 959                 break;
 960
 961         case SIOCADDPRL:
 962         case SIOCDELPRL:
 963         case SIOCCHGPRL:
 964                 err = -EPERM;
 965                 if (!capable(CAP_NET_ADMIN))
 966                         goto done;
 967                 err = -EINVAL;
 968                 if (dev == sitn->fb_tunnel_dev)
 969                         goto done;
 970                 err = -EFAULT;
 971                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 972                         goto done;
 973                 err = -ENOENT;
 974                 if (!(t = netdev_priv(dev)))
 975                         goto done;
 976
 977                 switch (cmd) {
 978                 case SIOCDELPRL:
 979                         err = ipip6_tunnel_del_prl(t, &prl);
 980                         break;
 981                 case SIOCADDPRL:
 982                 case SIOCCHGPRL:
 983                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 984                         break;
 985                 }
 986                 netdev_state_change(dev);
 987                 break;
 988
 989 #ifdef CONFIG_IPV6_SIT_6RD
 990         case SIOCADD6RD:
 991         case SIOCCHG6RD:
 992         case SIOCDEL6RD:
 993                 err = -EPERM;
 994                 if (!capable(CAP_NET_ADMIN))
 995                         goto done;
 996
 997                 err = -EFAULT;
 998                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
 999                                    sizeof(ip6rd)))
1000                         goto done;
1001
1002                 t = netdev_priv(dev);
1003
1004                 if (cmd != SIOCDEL6RD) {
1005                         struct in6_addr prefix;
1006                         __be32 relay_prefix;
1007
1008                         err = -EINVAL;
1009                         if (ip6rd.relay_prefixlen > 32 ||
1010                             ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1011                                 goto done;
1012
1013                         ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1014                                          ip6rd.prefixlen);
1015                         if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1016                                 goto done;
1017                         relay_prefix = ip6rd.relay_prefix &
1018                                        htonl(0xffffffffUL <<
1019                                              (32 - ip6rd.relay_prefixlen));
1020                         if (relay_prefix != ip6rd.relay_prefix)
1021                                 goto done;
1022
1023                         ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1024                         t->ip6rd.relay_prefix = relay_prefix;
1025                         t->ip6rd.prefixlen = ip6rd.prefixlen;
1026                         t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1027                 } else
1028                         ipip6_tunnel_clone_6rd(dev, sitn);
1029
1030                 err = 0;
1031                 break;
1032 #endif
1033
1034         default:
1035                 err = -EINVAL;
1036         }
1037
1038 done:
1039         return err;
1040 }
1041
1042 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1043 {
1044         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1045                 return -EINVAL;
1046         dev->mtu = new_mtu;
1047         return 0;
1048 }
1049
1050 static const struct net_device_ops ipip6_netdev_ops = {
1051         .ndo_uninit     = ipip6_tunnel_uninit,
1052         .ndo_start_xmit = ipip6_tunnel_xmit,
1053         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1054         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1055 };
1056
1057 static void ipip6_tunnel_setup(struct net_device *dev)
1058 {
1059         dev->netdev_ops         = &ipip6_netdev_ops;
1060         dev->destructor         = free_netdev;
1061
1062         dev->type               = ARPHRD_SIT;
1063         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1064         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1065         dev->flags              = IFF_NOARP;
1066         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1067         dev->iflink             = 0;
1068         dev->addr_len           = 4;
1069         dev->features           |= NETIF_F_NETNS_LOCAL;
1070 }
1071
1072 static void ipip6_tunnel_init(struct net_device *dev)
1073 {
1074         struct ip_tunnel *tunnel = netdev_priv(dev);
1075
1076         tunnel->dev = dev;
1077         strcpy(tunnel->parms.name, dev->name);
1078
1079         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1080         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1081
1082         ipip6_tunnel_bind_dev(dev);
1083 }
1084
1085 static void ipip6_fb_tunnel_init(struct net_device *dev)
1086 {
1087         struct ip_tunnel *tunnel = netdev_priv(dev);
1088         struct iphdr *iph = &tunnel->parms.iph;
1089         struct net *net = dev_net(dev);
1090         struct sit_net *sitn = net_generic(net, sit_net_id);
1091
1092         tunnel->dev = dev;
1093         strcpy(tunnel->parms.name, dev->name);
1094
1095         iph->version            = 4;
1096         iph->protocol           = IPPROTO_IPV6;
1097         iph->ihl                = 5;
1098         iph->ttl                = 64;
1099
1100         dev_hold(dev);
1101         sitn->tunnels_wc[0]     = tunnel;
1102 }
1103
1104 static struct xfrm_tunnel sit_handler = {
1105         .handler        =       ipip6_rcv,
1106         .err_handler    =       ipip6_err,
1107         .priority       =       1,
1108 };
1109
1110 static void sit_destroy_tunnels(struct sit_net *sitn)
1111 {
1112         int prio;
1113
1114         for (prio = 1; prio < 4; prio++) {
1115                 int h;
1116                 for (h = 0; h < HASH_SIZE; h++) {
1117                         struct ip_tunnel *t;
1118                         while ((t = sitn->tunnels[prio][h]) != NULL)
1119                                 unregister_netdevice(t->dev);
1120                 }
1121         }
1122 }
1123
1124 static int sit_init_net(struct net *net)
1125 {
1126         int err;
1127         struct sit_net *sitn;
1128
1129         err = -ENOMEM;
1130         sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1131         if (sitn == NULL)
1132                 goto err_alloc;
1133
1134         err = net_assign_generic(net, sit_net_id, sitn);
1135         if (err < 0)
1136                 goto err_assign;
1137
1138         sitn->tunnels[0] = sitn->tunnels_wc;
1139         sitn->tunnels[1] = sitn->tunnels_l;
1140         sitn->tunnels[2] = sitn->tunnels_r;
1141         sitn->tunnels[3] = sitn->tunnels_r_l;
1142
1143         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1144                                            ipip6_tunnel_setup);
1145         if (!sitn->fb_tunnel_dev) {
1146                 err = -ENOMEM;
1147                 goto err_alloc_dev;
1148         }
1149         dev_net_set(sitn->fb_tunnel_dev, net);
1150
1151         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1152         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1153
1154         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1155                 goto err_reg_dev;
1156
1157         return 0;
1158
1159 err_reg_dev:
1160         dev_put(sitn->fb_tunnel_dev);
1161         free_netdev(sitn->fb_tunnel_dev);
1162 err_alloc_dev:
1163         /* nothing */
1164 err_assign:
1165         kfree(sitn);
1166 err_alloc:
1167         return err;
1168 }
1169
1170 static void sit_exit_net(struct net *net)
1171 {
1172         struct sit_net *sitn;
1173
1174         sitn = net_generic(net, sit_net_id);
1175         rtnl_lock();
1176         sit_destroy_tunnels(sitn);
1177         unregister_netdevice(sitn->fb_tunnel_dev);
1178         rtnl_unlock();
1179         kfree(sitn);
1180 }
1181
1182 static struct pernet_operations sit_net_ops = {
1183         .init = sit_init_net,
1184         .exit = sit_exit_net,
1185 };
1186
1187 static void __exit sit_cleanup(void)
1188 {
1189         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1190
1191         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1192 }
1193
1194 static int __init sit_init(void)
1195 {
1196         int err;
1197
1198         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1199
1200         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1201                 printk(KERN_INFO "sit init: Can't add protocol\n");
1202                 return -EAGAIN;
1203         }
1204
1205         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1206         if (err < 0)
1207                 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1208
1209         return err;
1210 }
1211
1212 module_init(sit_init);
1213 module_exit(sit_cleanup);
1214 MODULE_LICENSE("GPL");
1215 MODULE_ALIAS("sit0");