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             <[email protected]>
   7  *      Alexey Kuznetsov        <[email protected]>
   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 <[email protected]>:       6to4 support
  16  * Nate Thompson <[email protected]>:             6to4 support
  17  * Fred Templin <[email protected]>:    isatap support
  18  */
  19
  20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  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 <linux/slab.h>
  34 #include <asm/uaccess.h>
  35 #include <linux/init.h>
  36 #include <linux/netfilter_ipv4.h>
  37 #include <linux/if_ether.h>
  38
  39 #include <net/sock.h>
  40 #include <net/snmp.h>
  41
  42 #include <net/ipv6.h>
  43 #include <net/protocol.h>
  44 #include <net/transp_v6.h>
  45 #include <net/ip6_fib.h>
  46 #include <net/ip6_route.h>
  47 #include <net/ndisc.h>
  48 #include <net/addrconf.h>
  49 #include <net/ip.h>
  50 #include <net/udp.h>
  51 #include <net/icmp.h>
  52 #include <net/ipip.h>
  53 #include <net/inet_ecn.h>
  54 #include <net/xfrm.h>
  55 #include <net/dsfield.h>
  56 #include <net/net_namespace.h>
  57 #include <net/netns/generic.h>
  58
  59 /*
  60    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  61
  62    For comments look at net/ipv4/ip_gre.c --ANK
  63  */
  64
  65 #define HASH_SIZE  16
  66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  67
  68 static bool log_ecn_error = true;
  69 module_param(log_ecn_error, bool, 0644);
  70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  71
  72 static int ipip6_tunnel_init(struct net_device *dev);
  73 static void ipip6_tunnel_setup(struct net_device *dev);
  74 static void ipip6_dev_free(struct net_device *dev);
  75 static struct rtnl_link_ops sit_link_ops __read_mostly;
  76
  77 static int sit_net_id __read_mostly;
  78 struct sit_net {
  79         struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
  80         struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
  81         struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
  82         struct ip_tunnel __rcu *tunnels_wc[1];
  83         struct ip_tunnel __rcu **tunnels[4];
  84
  85         struct net_device *fb_tunnel_dev;
  86 };
  87
  88 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev,
  89                                                    struct rtnl_link_stats64 *tot)
  90 {
  91         int i;
  92
  93         for_each_possible_cpu(i) {
  94                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
  95                 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
  96                 unsigned int start;
  97
  98                 do {
  99                         start = u64_stats_fetch_begin_bh(&tstats->syncp);
 100                         rx_packets = tstats->rx_packets;
 101                         tx_packets = tstats->tx_packets;
 102                         rx_bytes = tstats->rx_bytes;
 103                         tx_bytes = tstats->tx_bytes;
 104                 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
 105
 106                 tot->rx_packets += rx_packets;
 107                 tot->tx_packets += tx_packets;
 108                 tot->rx_bytes   += rx_bytes;
 109                 tot->tx_bytes   += tx_bytes;
 110         }
 111
 112         tot->rx_errors = dev->stats.rx_errors;
 113         tot->rx_frame_errors = dev->stats.rx_frame_errors;
 114         tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
 115         tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
 116         tot->tx_dropped = dev->stats.tx_dropped;
 117         tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
 118         tot->tx_errors = dev->stats.tx_errors;
 119
 120         return tot;
 121 }
 122
 123 /*
 124  * Must be invoked with rcu_read_lock
 125  */
 126 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
 127                 struct net_device *dev, __be32 remote, __be32 local)
 128 {
 129         unsigned int h0 = HASH(remote);
 130         unsigned int h1 = HASH(local);
 131         struct ip_tunnel *t;
 132         struct sit_net *sitn = net_generic(net, sit_net_id);
 133
 134         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 135                 if (local == t->parms.iph.saddr &&
 136                     remote == t->parms.iph.daddr &&
 137                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 138                     (t->dev->flags & IFF_UP))
 139                         return t;
 140         }
 141         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 142                 if (remote == t->parms.iph.daddr &&
 143                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 144                     (t->dev->flags & IFF_UP))
 145                         return t;
 146         }
 147         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 148                 if (local == t->parms.iph.saddr &&
 149                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 150                     (t->dev->flags & IFF_UP))
 151                         return t;
 152         }
 153         t = rcu_dereference(sitn->tunnels_wc[0]);
 154         if ((t != NULL) && (t->dev->flags & IFF_UP))
 155                 return t;
 156         return NULL;
 157 }
 158
 159 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 160                 struct ip_tunnel_parm *parms)
 161 {
 162         __be32 remote = parms->iph.daddr;
 163         __be32 local = parms->iph.saddr;
 164         unsigned int h = 0;
 165         int prio = 0;
 166
 167         if (remote) {
 168                 prio |= 2;
 169                 h ^= HASH(remote);
 170         }
 171         if (local) {
 172                 prio |= 1;
 173                 h ^= HASH(local);
 174         }
 175         return &sitn->tunnels[prio][h];
 176 }
 177
 178 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 179                 struct ip_tunnel *t)
 180 {
 181         return __ipip6_bucket(sitn, &t->parms);
 182 }
 183
 184 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 185 {
 186         struct ip_tunnel __rcu **tp;
 187         struct ip_tunnel *iter;
 188
 189         for (tp = ipip6_bucket(sitn, t);
 190              (iter = rtnl_dereference(*tp)) != NULL;
 191              tp = &iter->next) {
 192                 if (t == iter) {
 193                         rcu_assign_pointer(*tp, t->next);
 194                         break;
 195                 }
 196         }
 197 }
 198
 199 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 200 {
 201         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 202
 203         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 204         rcu_assign_pointer(*tp, t);
 205 }
 206
 207 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 208 {
 209 #ifdef CONFIG_IPV6_SIT_6RD
 210         struct ip_tunnel *t = netdev_priv(dev);
 211
 212         if (t->dev == sitn->fb_tunnel_dev) {
 213                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 214                 t->ip6rd.relay_prefix = 0;
 215                 t->ip6rd.prefixlen = 16;
 216                 t->ip6rd.relay_prefixlen = 0;
 217         } else {
 218                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 219                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 220         }
 221 #endif
 222 }
 223
 224 static int ipip6_tunnel_create(struct net_device *dev)
 225 {
 226         struct ip_tunnel *t = netdev_priv(dev);
 227         struct net *net = dev_net(dev);
 228         struct sit_net *sitn = net_generic(net, sit_net_id);
 229         int err;
 230
 231         err = ipip6_tunnel_init(dev);
 232         if (err < 0)
 233                 goto out;
 234         ipip6_tunnel_clone_6rd(dev, sitn);
 235
 236         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 237                 dev->priv_flags |= IFF_ISATAP;
 238
 239         err = register_netdevice(dev);
 240         if (err < 0)
 241                 goto out;
 242
 243         strcpy(t->parms.name, dev->name);
 244         dev->rtnl_link_ops = &sit_link_ops;
 245
 246         dev_hold(dev);
 247
 248         ipip6_tunnel_link(sitn, t);
 249         return 0;
 250
 251 out:
 252         return err;
 253 }
 254
 255 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 256                 struct ip_tunnel_parm *parms, int create)
 257 {
 258         __be32 remote = parms->iph.daddr;
 259         __be32 local = parms->iph.saddr;
 260         struct ip_tunnel *t, *nt;
 261         struct ip_tunnel __rcu **tp;
 262         struct net_device *dev;
 263         char name[IFNAMSIZ];
 264         struct sit_net *sitn = net_generic(net, sit_net_id);
 265
 266         for (tp = __ipip6_bucket(sitn, parms);
 267             (t = rtnl_dereference(*tp)) != NULL;
 268              tp = &t->next) {
 269                 if (local == t->parms.iph.saddr &&
 270                     remote == t->parms.iph.daddr &&
 271                     parms->link == t->parms.link) {
 272                         if (create)
 273                                 return NULL;
 274                         else
 275                                 return t;
 276                 }
 277         }
 278         if (!create)
 279                 goto failed;
 280
 281         if (parms->name[0])
 282                 strlcpy(name, parms->name, IFNAMSIZ);
 283         else
 284                 strcpy(name, "sit%d");
 285
 286         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 287         if (dev == NULL)
 288                 return NULL;
 289
 290         dev_net_set(dev, net);
 291
 292         nt = netdev_priv(dev);
 293
 294         nt->parms = *parms;
 295         if (ipip6_tunnel_create(dev) < 0)
 296                 goto failed_free;
 297
 298         return nt;
 299
 300 failed_free:
 301         ipip6_dev_free(dev);
 302 failed:
 303         return NULL;
 304 }
 305
 306 #define for_each_prl_rcu(start)                 \
 307         for (prl = rcu_dereference(start);      \
 308              prl;                               \
 309              prl = rcu_dereference(prl->next))
 310
 311 static struct ip_tunnel_prl_entry *
 312 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 313 {
 314         struct ip_tunnel_prl_entry *prl;
 315
 316         for_each_prl_rcu(t->prl)
 317                 if (prl->addr == addr)
 318                         break;
 319         return prl;
 320
 321 }
 322
 323 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 324                                 struct ip_tunnel_prl __user *a)
 325 {
 326         struct ip_tunnel_prl kprl, *kp;
 327         struct ip_tunnel_prl_entry *prl;
 328         unsigned int cmax, c = 0, ca, len;
 329         int ret = 0;
 330
 331         if (copy_from_user(&kprl, a, sizeof(kprl)))
 332                 return -EFAULT;
 333         cmax = kprl.datalen / sizeof(kprl);
 334         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 335                 cmax = 1;
 336
 337         /* For simple GET or for root users,
 338          * we try harder to allocate.
 339          */
 340         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 341                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 342                 NULL;
 343
 344         rcu_read_lock();
 345
 346         ca = t->prl_count < cmax ? t->prl_count : cmax;
 347
 348         if (!kp) {
 349                 /* We don't try hard to allocate much memory for
 350                  * non-root users.
 351                  * For root users, retry allocating enough memory for
 352                  * the answer.
 353                  */
 354                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 355                 if (!kp) {
 356                         ret = -ENOMEM;
 357                         goto out;
 358                 }
 359         }
 360
 361         c = 0;
 362         for_each_prl_rcu(t->prl) {
 363                 if (c >= cmax)
 364                         break;
 365                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 366                         continue;
 367                 kp[c].addr = prl->addr;
 368                 kp[c].flags = prl->flags;
 369                 c++;
 370                 if (kprl.addr != htonl(INADDR_ANY))
 371                         break;
 372         }
 373 out:
 374         rcu_read_unlock();
 375
 376         len = sizeof(*kp) * c;
 377         ret = 0;
 378         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 379                 ret = -EFAULT;
 380
 381         kfree(kp);
 382
 383         return ret;
 384 }
 385
 386 static int
 387 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 388 {
 389         struct ip_tunnel_prl_entry *p;
 390         int err = 0;
 391
 392         if (a->addr == htonl(INADDR_ANY))
 393                 return -EINVAL;
 394
 395         ASSERT_RTNL();
 396
 397         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 398                 if (p->addr == a->addr) {
 399                         if (chg) {
 400                                 p->flags = a->flags;
 401                                 goto out;
 402                         }
 403                         err = -EEXIST;
 404                         goto out;
 405                 }
 406         }
 407
 408         if (chg) {
 409                 err = -ENXIO;
 410                 goto out;
 411         }
 412
 413         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 414         if (!p) {
 415                 err = -ENOBUFS;
 416                 goto out;
 417         }
 418
 419         p->next = t->prl;
 420         p->addr = a->addr;
 421         p->flags = a->flags;
 422         t->prl_count++;
 423         rcu_assign_pointer(t->prl, p);
 424 out:
 425         return err;
 426 }
 427
 428 static void prl_list_destroy_rcu(struct rcu_head *head)
 429 {
 430         struct ip_tunnel_prl_entry *p, *n;
 431
 432         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 433         do {
 434                 n = rcu_dereference_protected(p->next, 1);
 435                 kfree(p);
 436                 p = n;
 437         } while (p);
 438 }
 439
 440 static int
 441 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 442 {
 443         struct ip_tunnel_prl_entry *x;
 444         struct ip_tunnel_prl_entry __rcu **p;
 445         int err = 0;
 446
 447         ASSERT_RTNL();
 448
 449         if (a && a->addr != htonl(INADDR_ANY)) {
 450                 for (p = &t->prl;
 451                      (x = rtnl_dereference(*p)) != NULL;
 452                      p = &x->next) {
 453                         if (x->addr == a->addr) {
 454                                 *p = x->next;
 455                                 kfree_rcu(x, rcu_head);
 456                                 t->prl_count--;
 457                                 goto out;
 458                         }
 459                 }
 460                 err = -ENXIO;
 461         } else {
 462                 x = rtnl_dereference(t->prl);
 463                 if (x) {
 464                         t->prl_count = 0;
 465                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 466                         t->prl = NULL;
 467                 }
 468         }
 469 out:
 470         return err;
 471 }
 472
 473 static int
 474 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 475 {
 476         struct ip_tunnel_prl_entry *p;
 477         int ok = 1;
 478
 479         rcu_read_lock();
 480         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 481         if (p) {
 482                 if (p->flags & PRL_DEFAULT)
 483                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 484                 else
 485                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 486         } else {
 487                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 488
 489                 if (ipv6_addr_is_isatap(addr6) &&
 490                     (addr6->s6_addr32[3] == iph->saddr) &&
 491                     ipv6_chk_prefix(addr6, t->dev))
 492                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 493                 else
 494                         ok = 0;
 495         }
 496         rcu_read_unlock();
 497         return ok;
 498 }
 499
 500 static void ipip6_tunnel_uninit(struct net_device *dev)
 501 {
 502         struct net *net = dev_net(dev);
 503         struct sit_net *sitn = net_generic(net, sit_net_id);
 504
 505         if (dev == sitn->fb_tunnel_dev) {
 506                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 507         } else {
 508                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
 509                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 510         }
 511         dev_put(dev);
 512 }
 513
 514
 515 static int ipip6_err(struct sk_buff *skb, u32 info)
 516 {
 517
 518 /* All the routers (except for Linux) return only
 519    8 bytes of packet payload. It means, that precise relaying of
 520    ICMP in the real Internet is absolutely infeasible.
 521  */
 522         const struct iphdr *iph = (const struct iphdr *)skb->data;
 523         const int type = icmp_hdr(skb)->type;
 524         const int code = icmp_hdr(skb)->code;
 525         struct ip_tunnel *t;
 526         int err;
 527
 528         switch (type) {
 529         default:
 530         case ICMP_PARAMETERPROB:
 531                 return 0;
 532
 533         case ICMP_DEST_UNREACH:
 534                 switch (code) {
 535                 case ICMP_SR_FAILED:
 536                 case ICMP_PORT_UNREACH:
 537                         /* Impossible event. */
 538                         return 0;
 539                 default:
 540                         /* All others are translated to HOST_UNREACH.
 541                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 542                            I believe they are just ether pollution. --ANK
 543                          */
 544                         break;
 545                 }
 546                 break;
 547         case ICMP_TIME_EXCEEDED:
 548                 if (code != ICMP_EXC_TTL)
 549                         return 0;
 550                 break;
 551         case ICMP_REDIRECT:
 552                 break;
 553         }
 554
 555         err = -ENOENT;
 556
 557         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 558                                 skb->dev,
 559                                 iph->daddr,
 560                                 iph->saddr);
 561         if (t == NULL)
 562                 goto out;
 563
 564         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 565                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 566                                  t->dev->ifindex, 0, IPPROTO_IPV6, 0);
 567                 err = 0;
 568                 goto out;
 569         }
 570         if (type == ICMP_REDIRECT) {
 571                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 572                               IPPROTO_IPV6, 0);
 573                 err = 0;
 574                 goto out;
 575         }
 576
 577         if (t->parms.iph.daddr == 0)
 578                 goto out;
 579
 580         err = 0;
 581         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 582                 goto out;
 583
 584         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 585                 t->err_count++;
 586         else
 587                 t->err_count = 1;
 588         t->err_time = jiffies;
 589 out:
 590         return err;
 591 }
 592
 593 static int ipip6_rcv(struct sk_buff *skb)
 594 {
 595         const struct iphdr *iph = ip_hdr(skb);
 596         struct ip_tunnel *tunnel;
 597         int err;
 598
 599         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 600                                      iph->saddr, iph->daddr);
 601         if (tunnel != NULL) {
 602                 struct pcpu_tstats *tstats;
 603
 604                 secpath_reset(skb);
 605                 skb->mac_header = skb->network_header;
 606                 skb_reset_network_header(skb);
 607                 IPCB(skb)->flags = 0;
 608                 skb->protocol = htons(ETH_P_IPV6);
 609                 skb->pkt_type = PACKET_HOST;
 610
 611                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 612                     !isatap_chksrc(skb, iph, tunnel)) {
 613                         tunnel->dev->stats.rx_errors++;
 614                         goto out;
 615                 }
 616
 617                 __skb_tunnel_rx(skb, tunnel->dev);
 618
 619                 err = IP_ECN_decapsulate(iph, skb);
 620                 if (unlikely(err)) {
 621                         if (log_ecn_error)
 622                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 623                                                      &iph->saddr, iph->tos);
 624                         if (err > 1) {
 625                                 ++tunnel->dev->stats.rx_frame_errors;
 626                                 ++tunnel->dev->stats.rx_errors;
 627                                 goto out;
 628                         }
 629                 }
 630
 631                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 632                 tstats->rx_packets++;
 633                 tstats->rx_bytes += skb->len;
 634
 635                 netif_rx(skb);
 636
 637                 return 0;
 638         }
 639
 640         /* no tunnel matched,  let upstream know, ipsec may handle it */
 641         return 1;
 642 out:
 643         kfree_skb(skb);
 644         return 0;
 645 }
 646
 647 /*
 648  * Returns the embedded IPv4 address if the IPv6 address
 649  * comes from 6rd / 6to4 (RFC 3056) addr space.
 650  */
 651 static inline
 652 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 653 {
 654         __be32 dst = 0;
 655
 656 #ifdef CONFIG_IPV6_SIT_6RD
 657         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 658                               tunnel->ip6rd.prefixlen)) {
 659                 unsigned int pbw0, pbi0;
 660                 int pbi1;
 661                 u32 d;
 662
 663                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 664                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 665
 666                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 667                     tunnel->ip6rd.relay_prefixlen;
 668
 669                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 670                 if (pbi1 > 0)
 671                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 672                              (32 - pbi1);
 673
 674                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
 675         }
 676 #else
 677         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 678                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 679                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 680         }
 681 #endif
 682         return dst;
 683 }
 684
 685 /*
 686  *      This function assumes it is being called from dev_queue_xmit()
 687  *      and that skb is filled properly by that function.
 688  */
 689
 690 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 691                                      struct net_device *dev)
 692 {
 693         struct ip_tunnel *tunnel = netdev_priv(dev);
 694         const struct iphdr  *tiph = &tunnel->parms.iph;
 695         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 696         u8     tos = tunnel->parms.iph.tos;
 697         __be16 df = tiph->frag_off;
 698         struct rtable *rt;                      /* Route to the other host */
 699         struct net_device *tdev;                /* Device to other host */
 700         struct iphdr  *iph;                     /* Our new IP header */
 701         unsigned int max_headroom;              /* The extra header space needed */
 702         __be32 dst = tiph->daddr;
 703         struct flowi4 fl4;
 704         int    mtu;
 705         const struct in6_addr *addr6;
 706         int addr_type;
 707
 708         if (skb->protocol != htons(ETH_P_IPV6))
 709                 goto tx_error;
 710
 711         if (tos == 1)
 712                 tos = ipv6_get_dsfield(iph6);
 713
 714         /* ISATAP (RFC4214) - must come before 6to4 */
 715         if (dev->priv_flags & IFF_ISATAP) {
 716                 struct neighbour *neigh = NULL;
 717                 bool do_tx_error = false;
 718
 719                 if (skb_dst(skb))
 720                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 721
 722                 if (neigh == NULL) {
 723                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 724                         goto tx_error;
 725                 }
 726
 727                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 728                 addr_type = ipv6_addr_type(addr6);
 729
 730                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 731                      ipv6_addr_is_isatap(addr6))
 732                         dst = addr6->s6_addr32[3];
 733                 else
 734                         do_tx_error = true;
 735
 736                 neigh_release(neigh);
 737                 if (do_tx_error)
 738                         goto tx_error;
 739         }
 740
 741         if (!dst)
 742                 dst = try_6rd(&iph6->daddr, tunnel);
 743
 744         if (!dst) {
 745                 struct neighbour *neigh = NULL;
 746                 bool do_tx_error = false;
 747
 748                 if (skb_dst(skb))
 749                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 750
 751                 if (neigh == NULL) {
 752                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 753                         goto tx_error;
 754                 }
 755
 756                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 757                 addr_type = ipv6_addr_type(addr6);
 758
 759                 if (addr_type == IPV6_ADDR_ANY) {
 760                         addr6 = &ipv6_hdr(skb)->daddr;
 761                         addr_type = ipv6_addr_type(addr6);
 762                 }
 763
 764                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 765                         dst = addr6->s6_addr32[3];
 766                 else
 767                         do_tx_error = true;
 768
 769                 neigh_release(neigh);
 770                 if (do_tx_error)
 771                         goto tx_error;
 772         }
 773
 774         rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 775                                    dst, tiph->saddr,
 776                                    0, 0,
 777                                    IPPROTO_IPV6, RT_TOS(tos),
 778                                    tunnel->parms.link);
 779         if (IS_ERR(rt)) {
 780                 dev->stats.tx_carrier_errors++;
 781                 goto tx_error_icmp;
 782         }
 783         if (rt->rt_type != RTN_UNICAST) {
 784                 ip_rt_put(rt);
 785                 dev->stats.tx_carrier_errors++;
 786                 goto tx_error_icmp;
 787         }
 788         tdev = rt->dst.dev;
 789
 790         if (tdev == dev) {
 791                 ip_rt_put(rt);
 792                 dev->stats.collisions++;
 793                 goto tx_error;
 794         }
 795
 796         if (df) {
 797                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 798
 799                 if (mtu < 68) {
 800                         dev->stats.collisions++;
 801                         ip_rt_put(rt);
 802                         goto tx_error;
 803                 }
 804
 805                 if (mtu < IPV6_MIN_MTU) {
 806                         mtu = IPV6_MIN_MTU;
 807                         df = 0;
 808                 }
 809
 810                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 811                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 812
 813                 if (skb->len > mtu) {
 814                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 815                         ip_rt_put(rt);
 816                         goto tx_error;
 817                 }
 818         }
 819
 820         if (tunnel->err_count > 0) {
 821                 if (time_before(jiffies,
 822                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 823                         tunnel->err_count--;
 824                         dst_link_failure(skb);
 825                 } else
 826                         tunnel->err_count = 0;
 827         }
 828
 829         /*
 830          * Okay, now see if we can stuff it in the buffer as-is.
 831          */
 832         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 833
 834         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 835             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 836                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 837                 if (!new_skb) {
 838                         ip_rt_put(rt);
 839                         dev->stats.tx_dropped++;
 840                         dev_kfree_skb(skb);
 841                         return NETDEV_TX_OK;
 842                 }
 843                 if (skb->sk)
 844                         skb_set_owner_w(new_skb, skb->sk);
 845                 dev_kfree_skb(skb);
 846                 skb = new_skb;
 847                 iph6 = ipv6_hdr(skb);
 848         }
 849
 850         skb->transport_header = skb->network_header;
 851         skb_push(skb, sizeof(struct iphdr));
 852         skb_reset_network_header(skb);
 853         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 854         IPCB(skb)->flags = 0;
 855         skb_dst_drop(skb);
 856         skb_dst_set(skb, &rt->dst);
 857
 858         /*
 859          *      Push down and install the IPIP header.
 860          */
 861
 862         iph                     =       ip_hdr(skb);
 863         iph->version            =       4;
 864         iph->ihl                =       sizeof(struct iphdr)>>2;
 865         iph->frag_off           =       df;
 866         iph->protocol           =       IPPROTO_IPV6;
 867         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 868         iph->daddr              =       fl4.daddr;
 869         iph->saddr              =       fl4.saddr;
 870
 871         if ((iph->ttl = tiph->ttl) == 0)
 872                 iph->ttl        =       iph6->hop_limit;
 873
 874         iptunnel_xmit(skb, dev);
 875         return NETDEV_TX_OK;
 876
 877 tx_error_icmp:
 878         dst_link_failure(skb);
 879 tx_error:
 880         dev->stats.tx_errors++;
 881         dev_kfree_skb(skb);
 882         return NETDEV_TX_OK;
 883 }
 884
 885 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 886 {
 887         struct net_device *tdev = NULL;
 888         struct ip_tunnel *tunnel;
 889         const struct iphdr *iph;
 890         struct flowi4 fl4;
 891
 892         tunnel = netdev_priv(dev);
 893         iph = &tunnel->parms.iph;
 894
 895         if (iph->daddr) {
 896                 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 897                                                           iph->daddr, iph->saddr,
 898                                                           0, 0,
 899                                                           IPPROTO_IPV6,
 900                                                           RT_TOS(iph->tos),
 901                                                           tunnel->parms.link);
 902
 903                 if (!IS_ERR(rt)) {
 904                         tdev = rt->dst.dev;
 905                         ip_rt_put(rt);
 906                 }
 907                 dev->flags |= IFF_POINTOPOINT;
 908         }
 909
 910         if (!tdev && tunnel->parms.link)
 911                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 912
 913         if (tdev) {
 914                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 915                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 916                 if (dev->mtu < IPV6_MIN_MTU)
 917                         dev->mtu = IPV6_MIN_MTU;
 918         }
 919         dev->iflink = tunnel->parms.link;
 920 }
 921
 922 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 923 {
 924         struct net *net = dev_net(t->dev);
 925         struct sit_net *sitn = net_generic(net, sit_net_id);
 926
 927         ipip6_tunnel_unlink(sitn, t);
 928         synchronize_net();
 929         t->parms.iph.saddr = p->iph.saddr;
 930         t->parms.iph.daddr = p->iph.daddr;
 931         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 932         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 933         ipip6_tunnel_link(sitn, t);
 934         t->parms.iph.ttl = p->iph.ttl;
 935         t->parms.iph.tos = p->iph.tos;
 936         if (t->parms.link != p->link) {
 937                 t->parms.link = p->link;
 938                 ipip6_tunnel_bind_dev(t->dev);
 939         }
 940         netdev_state_change(t->dev);
 941 }
 942
 943 #ifdef CONFIG_IPV6_SIT_6RD
 944 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
 945                                    struct ip_tunnel_6rd *ip6rd)
 946 {
 947         struct in6_addr prefix;
 948         __be32 relay_prefix;
 949
 950         if (ip6rd->relay_prefixlen > 32 ||
 951             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
 952                 return -EINVAL;
 953
 954         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
 955         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
 956                 return -EINVAL;
 957         if (ip6rd->relay_prefixlen)
 958                 relay_prefix = ip6rd->relay_prefix &
 959                                htonl(0xffffffffUL <<
 960                                      (32 - ip6rd->relay_prefixlen));
 961         else
 962                 relay_prefix = 0;
 963         if (relay_prefix != ip6rd->relay_prefix)
 964                 return -EINVAL;
 965
 966         t->ip6rd.prefix = prefix;
 967         t->ip6rd.relay_prefix = relay_prefix;
 968         t->ip6rd.prefixlen = ip6rd->prefixlen;
 969         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
 970         netdev_state_change(t->dev);
 971         return 0;
 972 }
 973 #endif
 974
 975 static int
 976 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 977 {
 978         int err = 0;
 979         struct ip_tunnel_parm p;
 980         struct ip_tunnel_prl prl;
 981         struct ip_tunnel *t;
 982         struct net *net = dev_net(dev);
 983         struct sit_net *sitn = net_generic(net, sit_net_id);
 984 #ifdef CONFIG_IPV6_SIT_6RD
 985         struct ip_tunnel_6rd ip6rd;
 986 #endif
 987
 988         switch (cmd) {
 989         case SIOCGETTUNNEL:
 990 #ifdef CONFIG_IPV6_SIT_6RD
 991         case SIOCGET6RD:
 992 #endif
 993                 t = NULL;
 994                 if (dev == sitn->fb_tunnel_dev) {
 995                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 996                                 err = -EFAULT;
 997                                 break;
 998                         }
 999                         t = ipip6_tunnel_locate(net, &p, 0);
1000                 }
1001                 if (t == NULL)
1002                         t = netdev_priv(dev);
1003
1004                 err = -EFAULT;
1005                 if (cmd == SIOCGETTUNNEL) {
1006                         memcpy(&p, &t->parms, sizeof(p));
1007                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1008                                          sizeof(p)))
1009                                 goto done;
1010 #ifdef CONFIG_IPV6_SIT_6RD
1011                 } else {
1012                         ip6rd.prefix = t->ip6rd.prefix;
1013                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1014                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1015                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1016                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1017                                          sizeof(ip6rd)))
1018                                 goto done;
1019 #endif
1020                 }
1021                 err = 0;
1022                 break;
1023
1024         case SIOCADDTUNNEL:
1025         case SIOCCHGTUNNEL:
1026                 err = -EPERM;
1027                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1028                         goto done;
1029
1030                 err = -EFAULT;
1031                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1032                         goto done;
1033
1034                 err = -EINVAL;
1035                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1036                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1037                         goto done;
1038                 if (p.iph.ttl)
1039                         p.iph.frag_off |= htons(IP_DF);
1040
1041                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1042
1043                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1044                         if (t != NULL) {
1045                                 if (t->dev != dev) {
1046                                         err = -EEXIST;
1047                                         break;
1048                                 }
1049                         } else {
1050                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1051                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1052                                         err = -EINVAL;
1053                                         break;
1054                                 }
1055                                 t = netdev_priv(dev);
1056                         }
1057
1058                         ipip6_tunnel_update(t, &p);
1059                 }
1060
1061                 if (t) {
1062                         err = 0;
1063                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1064                                 err = -EFAULT;
1065                 } else
1066                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1067                 break;
1068
1069         case SIOCDELTUNNEL:
1070                 err = -EPERM;
1071                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1072                         goto done;
1073
1074                 if (dev == sitn->fb_tunnel_dev) {
1075                         err = -EFAULT;
1076                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1077                                 goto done;
1078                         err = -ENOENT;
1079                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1080                                 goto done;
1081                         err = -EPERM;
1082                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1083                                 goto done;
1084                         dev = t->dev;
1085                 }
1086                 unregister_netdevice(dev);
1087                 err = 0;
1088                 break;
1089
1090         case SIOCGETPRL:
1091                 err = -EINVAL;
1092                 if (dev == sitn->fb_tunnel_dev)
1093                         goto done;
1094                 err = -ENOENT;
1095                 if (!(t = netdev_priv(dev)))
1096                         goto done;
1097                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1098                 break;
1099
1100         case SIOCADDPRL:
1101         case SIOCDELPRL:
1102         case SIOCCHGPRL:
1103                 err = -EPERM;
1104                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1105                         goto done;
1106                 err = -EINVAL;
1107                 if (dev == sitn->fb_tunnel_dev)
1108                         goto done;
1109                 err = -EFAULT;
1110                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1111                         goto done;
1112                 err = -ENOENT;
1113                 if (!(t = netdev_priv(dev)))
1114                         goto done;
1115
1116                 switch (cmd) {
1117                 case SIOCDELPRL:
1118                         err = ipip6_tunnel_del_prl(t, &prl);
1119                         break;
1120                 case SIOCADDPRL:
1121                 case SIOCCHGPRL:
1122                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1123                         break;
1124                 }
1125                 netdev_state_change(dev);
1126                 break;
1127
1128 #ifdef CONFIG_IPV6_SIT_6RD
1129         case SIOCADD6RD:
1130         case SIOCCHG6RD:
1131         case SIOCDEL6RD:
1132                 err = -EPERM;
1133                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1134                         goto done;
1135
1136                 err = -EFAULT;
1137                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1138                                    sizeof(ip6rd)))
1139                         goto done;
1140
1141                 t = netdev_priv(dev);
1142
1143                 if (cmd != SIOCDEL6RD) {
1144                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1145                         if (err < 0)
1146                                 goto done;
1147                 } else
1148                         ipip6_tunnel_clone_6rd(dev, sitn);
1149
1150                 err = 0;
1151                 break;
1152 #endif
1153
1154         default:
1155                 err = -EINVAL;
1156         }
1157
1158 done:
1159         return err;
1160 }
1161
1162 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1163 {
1164         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1165                 return -EINVAL;
1166         dev->mtu = new_mtu;
1167         return 0;
1168 }
1169
1170 static const struct net_device_ops ipip6_netdev_ops = {
1171         .ndo_uninit     = ipip6_tunnel_uninit,
1172         .ndo_start_xmit = ipip6_tunnel_xmit,
1173         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1174         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1175         .ndo_get_stats64= ipip6_get_stats64,
1176 };
1177
1178 static void ipip6_dev_free(struct net_device *dev)
1179 {
1180         free_percpu(dev->tstats);
1181         free_netdev(dev);
1182 }
1183
1184 static void ipip6_tunnel_setup(struct net_device *dev)
1185 {
1186         dev->netdev_ops         = &ipip6_netdev_ops;
1187         dev->destructor         = ipip6_dev_free;
1188
1189         dev->type               = ARPHRD_SIT;
1190         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1191         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1192         dev->flags              = IFF_NOARP;
1193         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1194         dev->iflink             = 0;
1195         dev->addr_len           = 4;
1196         dev->features           |= NETIF_F_NETNS_LOCAL;
1197         dev->features           |= NETIF_F_LLTX;
1198 }
1199
1200 static int ipip6_tunnel_init(struct net_device *dev)
1201 {
1202         struct ip_tunnel *tunnel = netdev_priv(dev);
1203
1204         tunnel->dev = dev;
1205
1206         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1207         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1208
1209         ipip6_tunnel_bind_dev(dev);
1210         dev->tstats = alloc_percpu(struct pcpu_tstats);
1211         if (!dev->tstats)
1212                 return -ENOMEM;
1213
1214         return 0;
1215 }
1216
1217 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1218 {
1219         struct ip_tunnel *tunnel = netdev_priv(dev);
1220         struct iphdr *iph = &tunnel->parms.iph;
1221         struct net *net = dev_net(dev);
1222         struct sit_net *sitn = net_generic(net, sit_net_id);
1223
1224         tunnel->dev = dev;
1225         strcpy(tunnel->parms.name, dev->name);
1226
1227         iph->version            = 4;
1228         iph->protocol           = IPPROTO_IPV6;
1229         iph->ihl                = 5;
1230         iph->ttl                = 64;
1231
1232         dev->tstats = alloc_percpu(struct pcpu_tstats);
1233         if (!dev->tstats)
1234                 return -ENOMEM;
1235         dev_hold(dev);
1236         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1237         return 0;
1238 }
1239
1240 static void ipip6_netlink_parms(struct nlattr *data[],
1241                                 struct ip_tunnel_parm *parms)
1242 {
1243         memset(parms, 0, sizeof(*parms));
1244
1245         parms->iph.version = 4;
1246         parms->iph.protocol = IPPROTO_IPV6;
1247         parms->iph.ihl = 5;
1248         parms->iph.ttl = 64;
1249
1250         if (!data)
1251                 return;
1252
1253         if (data[IFLA_IPTUN_LINK])
1254                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1255
1256         if (data[IFLA_IPTUN_LOCAL])
1257                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1258
1259         if (data[IFLA_IPTUN_REMOTE])
1260                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1261
1262         if (data[IFLA_IPTUN_TTL]) {
1263                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1264                 if (parms->iph.ttl)
1265                         parms->iph.frag_off = htons(IP_DF);
1266         }
1267
1268         if (data[IFLA_IPTUN_TOS])
1269                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1270
1271         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1272                 parms->iph.frag_off = htons(IP_DF);
1273
1274         if (data[IFLA_IPTUN_FLAGS])
1275                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1276 }
1277
1278 #ifdef CONFIG_IPV6_SIT_6RD
1279 /* This function returns true when 6RD attributes are present in the nl msg */
1280 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1281                                     struct ip_tunnel_6rd *ip6rd)
1282 {
1283         bool ret = false;
1284         memset(ip6rd, 0, sizeof(*ip6rd));
1285
1286         if (!data)
1287                 return ret;
1288
1289         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1290                 ret = true;
1291                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1292                            sizeof(struct in6_addr));
1293         }
1294
1295         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1296                 ret = true;
1297                 ip6rd->relay_prefix =
1298                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1299         }
1300
1301         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1302                 ret = true;
1303                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1304         }
1305
1306         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1307                 ret = true;
1308                 ip6rd->relay_prefixlen =
1309                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1310         }
1311
1312         return ret;
1313 }
1314 #endif
1315
1316 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1317                          struct nlattr *tb[], struct nlattr *data[])
1318 {
1319         struct net *net = dev_net(dev);
1320         struct ip_tunnel *nt;
1321 #ifdef CONFIG_IPV6_SIT_6RD
1322         struct ip_tunnel_6rd ip6rd;
1323 #endif
1324         int err;
1325
1326         nt = netdev_priv(dev);
1327         ipip6_netlink_parms(data, &nt->parms);
1328
1329         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1330                 return -EEXIST;
1331
1332         err = ipip6_tunnel_create(dev);
1333         if (err < 0)
1334                 return err;
1335
1336 #ifdef CONFIG_IPV6_SIT_6RD
1337         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1338                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1339 #endif
1340
1341         return err;
1342 }
1343
1344 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1345                           struct nlattr *data[])
1346 {
1347         struct ip_tunnel *t;
1348         struct ip_tunnel_parm p;
1349         struct net *net = dev_net(dev);
1350         struct sit_net *sitn = net_generic(net, sit_net_id);
1351 #ifdef CONFIG_IPV6_SIT_6RD
1352         struct ip_tunnel_6rd ip6rd;
1353 #endif
1354
1355         if (dev == sitn->fb_tunnel_dev)
1356                 return -EINVAL;
1357
1358         ipip6_netlink_parms(data, &p);
1359
1360         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1361             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1362                 return -EINVAL;
1363
1364         t = ipip6_tunnel_locate(net, &p, 0);
1365
1366         if (t) {
1367                 if (t->dev != dev)
1368                         return -EEXIST;
1369         } else
1370                 t = netdev_priv(dev);
1371
1372         ipip6_tunnel_update(t, &p);
1373
1374 #ifdef CONFIG_IPV6_SIT_6RD
1375         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1376                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1377 #endif
1378
1379         return 0;
1380 }
1381
1382 static size_t ipip6_get_size(const struct net_device *dev)
1383 {
1384         return
1385                 /* IFLA_IPTUN_LINK */
1386                 nla_total_size(4) +
1387                 /* IFLA_IPTUN_LOCAL */
1388                 nla_total_size(4) +
1389                 /* IFLA_IPTUN_REMOTE */
1390                 nla_total_size(4) +
1391                 /* IFLA_IPTUN_TTL */
1392                 nla_total_size(1) +
1393                 /* IFLA_IPTUN_TOS */
1394                 nla_total_size(1) +
1395                 /* IFLA_IPTUN_PMTUDISC */
1396                 nla_total_size(1) +
1397                 /* IFLA_IPTUN_FLAGS */
1398                 nla_total_size(2) +
1399 #ifdef CONFIG_IPV6_SIT_6RD
1400                 /* IFLA_IPTUN_6RD_PREFIX */
1401                 nla_total_size(sizeof(struct in6_addr)) +
1402                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1403                 nla_total_size(4) +
1404                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1405                 nla_total_size(2) +
1406                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1407                 nla_total_size(2) +
1408 #endif
1409                 0;
1410 }
1411
1412 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1413 {
1414         struct ip_tunnel *tunnel = netdev_priv(dev);
1415         struct ip_tunnel_parm *parm = &tunnel->parms;
1416
1417         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1418             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1419             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1420             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1421             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1422             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1423                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1424             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1425                 goto nla_put_failure;
1426
1427 #ifdef CONFIG_IPV6_SIT_6RD
1428         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1429                     &tunnel->ip6rd.prefix) ||
1430             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1431                          tunnel->ip6rd.relay_prefix) ||
1432             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1433                         tunnel->ip6rd.prefixlen) ||
1434             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1435                         tunnel->ip6rd.relay_prefixlen))
1436                 goto nla_put_failure;
1437 #endif
1438
1439         return 0;
1440
1441 nla_put_failure:
1442         return -EMSGSIZE;
1443 }
1444
1445 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1446         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1447         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1448         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1449         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1450         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1451         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1452         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1453 #ifdef CONFIG_IPV6_SIT_6RD
1454         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1455         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1456         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1457         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1458 #endif
1459 };
1460
1461 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1462         .kind           = "sit",
1463         .maxtype        = IFLA_IPTUN_MAX,
1464         .policy         = ipip6_policy,
1465         .priv_size      = sizeof(struct ip_tunnel),
1466         .setup          = ipip6_tunnel_setup,
1467         .newlink        = ipip6_newlink,
1468         .changelink     = ipip6_changelink,
1469         .get_size       = ipip6_get_size,
1470         .fill_info      = ipip6_fill_info,
1471 };
1472
1473 static struct xfrm_tunnel sit_handler __read_mostly = {
1474         .handler        =       ipip6_rcv,
1475         .err_handler    =       ipip6_err,
1476         .priority       =       1,
1477 };
1478
1479 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1480 {
1481         int prio;
1482
1483         for (prio = 1; prio < 4; prio++) {
1484                 int h;
1485                 for (h = 0; h < HASH_SIZE; h++) {
1486                         struct ip_tunnel *t;
1487
1488                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1489                         while (t != NULL) {
1490                                 unregister_netdevice_queue(t->dev, head);
1491                                 t = rtnl_dereference(t->next);
1492                         }
1493                 }
1494         }
1495 }
1496
1497 static int __net_init sit_init_net(struct net *net)
1498 {
1499         struct sit_net *sitn = net_generic(net, sit_net_id);
1500         struct ip_tunnel *t;
1501         int err;
1502
1503         sitn->tunnels[0] = sitn->tunnels_wc;
1504         sitn->tunnels[1] = sitn->tunnels_l;
1505         sitn->tunnels[2] = sitn->tunnels_r;
1506         sitn->tunnels[3] = sitn->tunnels_r_l;
1507
1508         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1509                                            ipip6_tunnel_setup);
1510         if (!sitn->fb_tunnel_dev) {
1511                 err = -ENOMEM;
1512                 goto err_alloc_dev;
1513         }
1514         dev_net_set(sitn->fb_tunnel_dev, net);
1515
1516         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1517         if (err)
1518                 goto err_dev_free;
1519
1520         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1521
1522         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1523                 goto err_reg_dev;
1524
1525         t = netdev_priv(sitn->fb_tunnel_dev);
1526
1527         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1528         return 0;
1529
1530 err_reg_dev:
1531         dev_put(sitn->fb_tunnel_dev);
1532 err_dev_free:
1533         ipip6_dev_free(sitn->fb_tunnel_dev);
1534 err_alloc_dev:
1535         return err;
1536 }
1537
1538 static void __net_exit sit_exit_net(struct net *net)
1539 {
1540         struct sit_net *sitn = net_generic(net, sit_net_id);
1541         LIST_HEAD(list);
1542
1543         rtnl_lock();
1544         sit_destroy_tunnels(sitn, &list);
1545         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1546         unregister_netdevice_many(&list);
1547         rtnl_unlock();
1548 }
1549
1550 static struct pernet_operations sit_net_ops = {
1551         .init = sit_init_net,
1552         .exit = sit_exit_net,
1553         .id   = &sit_net_id,
1554         .size = sizeof(struct sit_net),
1555 };
1556
1557 static void __exit sit_cleanup(void)
1558 {
1559         rtnl_link_unregister(&sit_link_ops);
1560         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1561
1562         unregister_pernet_device(&sit_net_ops);
1563         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1564 }
1565
1566 static int __init sit_init(void)
1567 {
1568         int err;
1569
1570         pr_info("IPv6 over IPv4 tunneling driver\n");
1571
1572         err = register_pernet_device(&sit_net_ops);
1573         if (err < 0)
1574                 return err;
1575         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1576         if (err < 0) {
1577                 pr_info("%s: can't add protocol\n", __func__);
1578                 goto xfrm_tunnel_failed;
1579         }
1580         err = rtnl_link_register(&sit_link_ops);
1581         if (err < 0)
1582                 goto rtnl_link_failed;
1583
1584 out:
1585         return err;
1586
1587 rtnl_link_failed:
1588         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1589 xfrm_tunnel_failed:
1590         unregister_pernet_device(&sit_net_ops);
1591         goto out;
1592 }
1593
1594 module_init(sit_init);
1595 module_exit(sit_cleanup);
1596 MODULE_LICENSE("GPL");
1597 MODULE_ALIAS_NETDEV("sit0");