Pileus Git - ~andy/linux/blob - net/ipv6/sit.c

   1 /*
   2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  *
  14  *      Changes:
  15  * Roger Venning <r.venning@telstra.com>:       6to4 support
  16  * Nate Thompson <nate@thebog.net>:             6to4 support
  17  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  18  */
  19
  20 #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;
 596         struct ip_tunnel *tunnel;
 597         int err;
 598
 599         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 600                 goto out;
 601
 602         iph = ip_hdr(skb);
 603
 604         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 605                                      iph->saddr, iph->daddr);
 606         if (tunnel != NULL) {
 607                 struct pcpu_tstats *tstats;
 608
 609                 secpath_reset(skb);
 610                 skb->mac_header = skb->network_header;
 611                 skb_reset_network_header(skb);
 612                 IPCB(skb)->flags = 0;
 613                 skb->protocol = htons(ETH_P_IPV6);
 614                 skb->pkt_type = PACKET_HOST;
 615
 616                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 617                     !isatap_chksrc(skb, iph, tunnel)) {
 618                         tunnel->dev->stats.rx_errors++;
 619                         goto out;
 620                 }
 621
 622                 __skb_tunnel_rx(skb, tunnel->dev);
 623
 624                 err = IP_ECN_decapsulate(iph, skb);
 625                 if (unlikely(err)) {
 626                         if (log_ecn_error)
 627                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 628                                                      &iph->saddr, iph->tos);
 629                         if (err > 1) {
 630                                 ++tunnel->dev->stats.rx_frame_errors;
 631                                 ++tunnel->dev->stats.rx_errors;
 632                                 goto out;
 633                         }
 634                 }
 635
 636                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 637                 tstats->rx_packets++;
 638                 tstats->rx_bytes += skb->len;
 639
 640                 netif_rx(skb);
 641
 642                 return 0;
 643         }
 644
 645         /* no tunnel matched,  let upstream know, ipsec may handle it */
 646         return 1;
 647 out:
 648         kfree_skb(skb);
 649         return 0;
 650 }
 651
 652 /*
 653  * Returns the embedded IPv4 address if the IPv6 address
 654  * comes from 6rd / 6to4 (RFC 3056) addr space.
 655  */
 656 static inline
 657 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 658 {
 659         __be32 dst = 0;
 660
 661 #ifdef CONFIG_IPV6_SIT_6RD
 662         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 663                               tunnel->ip6rd.prefixlen)) {
 664                 unsigned int pbw0, pbi0;
 665                 int pbi1;
 666                 u32 d;
 667
 668                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 669                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 670
 671                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 672                     tunnel->ip6rd.relay_prefixlen;
 673
 674                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 675                 if (pbi1 > 0)
 676                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 677                              (32 - pbi1);
 678
 679                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
 680         }
 681 #else
 682         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 683                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 684                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 685         }
 686 #endif
 687         return dst;
 688 }
 689
 690 /*
 691  *      This function assumes it is being called from dev_queue_xmit()
 692  *      and that skb is filled properly by that function.
 693  */
 694
 695 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 696                                      struct net_device *dev)
 697 {
 698         struct ip_tunnel *tunnel = netdev_priv(dev);
 699         const struct iphdr  *tiph = &tunnel->parms.iph;
 700         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 701         u8     tos = tunnel->parms.iph.tos;
 702         __be16 df = tiph->frag_off;
 703         struct rtable *rt;                      /* Route to the other host */
 704         struct net_device *tdev;                /* Device to other host */
 705         struct iphdr  *iph;                     /* Our new IP header */
 706         unsigned int max_headroom;              /* The extra header space needed */
 707         __be32 dst = tiph->daddr;
 708         struct flowi4 fl4;
 709         int    mtu;
 710         const struct in6_addr *addr6;
 711         int addr_type;
 712
 713         if (skb->protocol != htons(ETH_P_IPV6))
 714                 goto tx_error;
 715
 716         if (tos == 1)
 717                 tos = ipv6_get_dsfield(iph6);
 718
 719         /* ISATAP (RFC4214) - must come before 6to4 */
 720         if (dev->priv_flags & IFF_ISATAP) {
 721                 struct neighbour *neigh = NULL;
 722                 bool do_tx_error = false;
 723
 724                 if (skb_dst(skb))
 725                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 726
 727                 if (neigh == NULL) {
 728                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 729                         goto tx_error;
 730                 }
 731
 732                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 733                 addr_type = ipv6_addr_type(addr6);
 734
 735                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 736                      ipv6_addr_is_isatap(addr6))
 737                         dst = addr6->s6_addr32[3];
 738                 else
 739                         do_tx_error = true;
 740
 741                 neigh_release(neigh);
 742                 if (do_tx_error)
 743                         goto tx_error;
 744         }
 745
 746         if (!dst)
 747                 dst = try_6rd(&iph6->daddr, tunnel);
 748
 749         if (!dst) {
 750                 struct neighbour *neigh = NULL;
 751                 bool do_tx_error = false;
 752
 753                 if (skb_dst(skb))
 754                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 755
 756                 if (neigh == NULL) {
 757                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 758                         goto tx_error;
 759                 }
 760
 761                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 762                 addr_type = ipv6_addr_type(addr6);
 763
 764                 if (addr_type == IPV6_ADDR_ANY) {
 765                         addr6 = &ipv6_hdr(skb)->daddr;
 766                         addr_type = ipv6_addr_type(addr6);
 767                 }
 768
 769                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 770                         dst = addr6->s6_addr32[3];
 771                 else
 772                         do_tx_error = true;
 773
 774                 neigh_release(neigh);
 775                 if (do_tx_error)
 776                         goto tx_error;
 777         }
 778
 779         rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 780                                    dst, tiph->saddr,
 781                                    0, 0,
 782                                    IPPROTO_IPV6, RT_TOS(tos),
 783                                    tunnel->parms.link);
 784         if (IS_ERR(rt)) {
 785                 dev->stats.tx_carrier_errors++;
 786                 goto tx_error_icmp;
 787         }
 788         if (rt->rt_type != RTN_UNICAST) {
 789                 ip_rt_put(rt);
 790                 dev->stats.tx_carrier_errors++;
 791                 goto tx_error_icmp;
 792         }
 793         tdev = rt->dst.dev;
 794
 795         if (tdev == dev) {
 796                 ip_rt_put(rt);
 797                 dev->stats.collisions++;
 798                 goto tx_error;
 799         }
 800
 801         if (df) {
 802                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 803
 804                 if (mtu < 68) {
 805                         dev->stats.collisions++;
 806                         ip_rt_put(rt);
 807                         goto tx_error;
 808                 }
 809
 810                 if (mtu < IPV6_MIN_MTU) {
 811                         mtu = IPV6_MIN_MTU;
 812                         df = 0;
 813                 }
 814
 815                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 816                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 817
 818                 if (skb->len > mtu) {
 819                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 820                         ip_rt_put(rt);
 821                         goto tx_error;
 822                 }
 823         }
 824
 825         if (tunnel->err_count > 0) {
 826                 if (time_before(jiffies,
 827                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 828                         tunnel->err_count--;
 829                         dst_link_failure(skb);
 830                 } else
 831                         tunnel->err_count = 0;
 832         }
 833
 834         /*
 835          * Okay, now see if we can stuff it in the buffer as-is.
 836          */
 837         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 838
 839         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 840             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 841                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 842                 if (!new_skb) {
 843                         ip_rt_put(rt);
 844                         dev->stats.tx_dropped++;
 845                         dev_kfree_skb(skb);
 846                         return NETDEV_TX_OK;
 847                 }
 848                 if (skb->sk)
 849                         skb_set_owner_w(new_skb, skb->sk);
 850                 dev_kfree_skb(skb);
 851                 skb = new_skb;
 852                 iph6 = ipv6_hdr(skb);
 853         }
 854
 855         skb->transport_header = skb->network_header;
 856         skb_push(skb, sizeof(struct iphdr));
 857         skb_reset_network_header(skb);
 858         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 859         IPCB(skb)->flags = 0;
 860         skb_dst_drop(skb);
 861         skb_dst_set(skb, &rt->dst);
 862
 863         /*
 864          *      Push down and install the IPIP header.
 865          */
 866
 867         iph                     =       ip_hdr(skb);
 868         iph->version            =       4;
 869         iph->ihl                =       sizeof(struct iphdr)>>2;
 870         iph->frag_off           =       df;
 871         iph->protocol           =       IPPROTO_IPV6;
 872         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 873         iph->daddr              =       fl4.daddr;
 874         iph->saddr              =       fl4.saddr;
 875
 876         if ((iph->ttl = tiph->ttl) == 0)
 877                 iph->ttl        =       iph6->hop_limit;
 878
 879         iptunnel_xmit(skb, dev);
 880         return NETDEV_TX_OK;
 881
 882 tx_error_icmp:
 883         dst_link_failure(skb);
 884 tx_error:
 885         dev->stats.tx_errors++;
 886         dev_kfree_skb(skb);
 887         return NETDEV_TX_OK;
 888 }
 889
 890 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 891 {
 892         struct net_device *tdev = NULL;
 893         struct ip_tunnel *tunnel;
 894         const struct iphdr *iph;
 895         struct flowi4 fl4;
 896
 897         tunnel = netdev_priv(dev);
 898         iph = &tunnel->parms.iph;
 899
 900         if (iph->daddr) {
 901                 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
 902                                                           iph->daddr, iph->saddr,
 903                                                           0, 0,
 904                                                           IPPROTO_IPV6,
 905                                                           RT_TOS(iph->tos),
 906                                                           tunnel->parms.link);
 907
 908                 if (!IS_ERR(rt)) {
 909                         tdev = rt->dst.dev;
 910                         ip_rt_put(rt);
 911                 }
 912                 dev->flags |= IFF_POINTOPOINT;
 913         }
 914
 915         if (!tdev && tunnel->parms.link)
 916                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 917
 918         if (tdev) {
 919                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 920                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 921                 if (dev->mtu < IPV6_MIN_MTU)
 922                         dev->mtu = IPV6_MIN_MTU;
 923         }
 924         dev->iflink = tunnel->parms.link;
 925 }
 926
 927 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 928 {
 929         struct net *net = dev_net(t->dev);
 930         struct sit_net *sitn = net_generic(net, sit_net_id);
 931
 932         ipip6_tunnel_unlink(sitn, t);
 933         synchronize_net();
 934         t->parms.iph.saddr = p->iph.saddr;
 935         t->parms.iph.daddr = p->iph.daddr;
 936         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 937         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 938         ipip6_tunnel_link(sitn, t);
 939         t->parms.iph.ttl = p->iph.ttl;
 940         t->parms.iph.tos = p->iph.tos;
 941         if (t->parms.link != p->link) {
 942                 t->parms.link = p->link;
 943                 ipip6_tunnel_bind_dev(t->dev);
 944         }
 945         netdev_state_change(t->dev);
 946 }
 947
 948 #ifdef CONFIG_IPV6_SIT_6RD
 949 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
 950                                    struct ip_tunnel_6rd *ip6rd)
 951 {
 952         struct in6_addr prefix;
 953         __be32 relay_prefix;
 954
 955         if (ip6rd->relay_prefixlen > 32 ||
 956             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
 957                 return -EINVAL;
 958
 959         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
 960         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
 961                 return -EINVAL;
 962         if (ip6rd->relay_prefixlen)
 963                 relay_prefix = ip6rd->relay_prefix &
 964                                htonl(0xffffffffUL <<
 965                                      (32 - ip6rd->relay_prefixlen));
 966         else
 967                 relay_prefix = 0;
 968         if (relay_prefix != ip6rd->relay_prefix)
 969                 return -EINVAL;
 970
 971         t->ip6rd.prefix = prefix;
 972         t->ip6rd.relay_prefix = relay_prefix;
 973         t->ip6rd.prefixlen = ip6rd->prefixlen;
 974         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
 975         netdev_state_change(t->dev);
 976         return 0;
 977 }
 978 #endif
 979
 980 static int
 981 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 982 {
 983         int err = 0;
 984         struct ip_tunnel_parm p;
 985         struct ip_tunnel_prl prl;
 986         struct ip_tunnel *t;
 987         struct net *net = dev_net(dev);
 988         struct sit_net *sitn = net_generic(net, sit_net_id);
 989 #ifdef CONFIG_IPV6_SIT_6RD
 990         struct ip_tunnel_6rd ip6rd;
 991 #endif
 992
 993         switch (cmd) {
 994         case SIOCGETTUNNEL:
 995 #ifdef CONFIG_IPV6_SIT_6RD
 996         case SIOCGET6RD:
 997 #endif
 998                 t = NULL;
 999                 if (dev == sitn->fb_tunnel_dev) {
1000                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1001                                 err = -EFAULT;
1002                                 break;
1003                         }
1004                         t = ipip6_tunnel_locate(net, &p, 0);
1005                 }
1006                 if (t == NULL)
1007                         t = netdev_priv(dev);
1008
1009                 err = -EFAULT;
1010                 if (cmd == SIOCGETTUNNEL) {
1011                         memcpy(&p, &t->parms, sizeof(p));
1012                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1013                                          sizeof(p)))
1014                                 goto done;
1015 #ifdef CONFIG_IPV6_SIT_6RD
1016                 } else {
1017                         ip6rd.prefix = t->ip6rd.prefix;
1018                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1019                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1020                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1021                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1022                                          sizeof(ip6rd)))
1023                                 goto done;
1024 #endif
1025                 }
1026                 err = 0;
1027                 break;
1028
1029         case SIOCADDTUNNEL:
1030         case SIOCCHGTUNNEL:
1031                 err = -EPERM;
1032                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1033                         goto done;
1034
1035                 err = -EFAULT;
1036                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1037                         goto done;
1038
1039                 err = -EINVAL;
1040                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
1041                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1042                         goto done;
1043                 if (p.iph.ttl)
1044                         p.iph.frag_off |= htons(IP_DF);
1045
1046                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1047
1048                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1049                         if (t != NULL) {
1050                                 if (t->dev != dev) {
1051                                         err = -EEXIST;
1052                                         break;
1053                                 }
1054                         } else {
1055                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1056                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1057                                         err = -EINVAL;
1058                                         break;
1059                                 }
1060                                 t = netdev_priv(dev);
1061                         }
1062
1063                         ipip6_tunnel_update(t, &p);
1064                 }
1065
1066                 if (t) {
1067                         err = 0;
1068                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1069                                 err = -EFAULT;
1070                 } else
1071                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1072                 break;
1073
1074         case SIOCDELTUNNEL:
1075                 err = -EPERM;
1076                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1077                         goto done;
1078
1079                 if (dev == sitn->fb_tunnel_dev) {
1080                         err = -EFAULT;
1081                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1082                                 goto done;
1083                         err = -ENOENT;
1084                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1085                                 goto done;
1086                         err = -EPERM;
1087                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1088                                 goto done;
1089                         dev = t->dev;
1090                 }
1091                 unregister_netdevice(dev);
1092                 err = 0;
1093                 break;
1094
1095         case SIOCGETPRL:
1096                 err = -EINVAL;
1097                 if (dev == sitn->fb_tunnel_dev)
1098                         goto done;
1099                 err = -ENOENT;
1100                 if (!(t = netdev_priv(dev)))
1101                         goto done;
1102                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1103                 break;
1104
1105         case SIOCADDPRL:
1106         case SIOCDELPRL:
1107         case SIOCCHGPRL:
1108                 err = -EPERM;
1109                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1110                         goto done;
1111                 err = -EINVAL;
1112                 if (dev == sitn->fb_tunnel_dev)
1113                         goto done;
1114                 err = -EFAULT;
1115                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1116                         goto done;
1117                 err = -ENOENT;
1118                 if (!(t = netdev_priv(dev)))
1119                         goto done;
1120
1121                 switch (cmd) {
1122                 case SIOCDELPRL:
1123                         err = ipip6_tunnel_del_prl(t, &prl);
1124                         break;
1125                 case SIOCADDPRL:
1126                 case SIOCCHGPRL:
1127                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1128                         break;
1129                 }
1130                 netdev_state_change(dev);
1131                 break;
1132
1133 #ifdef CONFIG_IPV6_SIT_6RD
1134         case SIOCADD6RD:
1135         case SIOCCHG6RD:
1136         case SIOCDEL6RD:
1137                 err = -EPERM;
1138                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1139                         goto done;
1140
1141                 err = -EFAULT;
1142                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1143                                    sizeof(ip6rd)))
1144                         goto done;
1145
1146                 t = netdev_priv(dev);
1147
1148                 if (cmd != SIOCDEL6RD) {
1149                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1150                         if (err < 0)
1151                                 goto done;
1152                 } else
1153                         ipip6_tunnel_clone_6rd(dev, sitn);
1154
1155                 err = 0;
1156                 break;
1157 #endif
1158
1159         default:
1160                 err = -EINVAL;
1161         }
1162
1163 done:
1164         return err;
1165 }
1166
1167 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1168 {
1169         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1170                 return -EINVAL;
1171         dev->mtu = new_mtu;
1172         return 0;
1173 }
1174
1175 static const struct net_device_ops ipip6_netdev_ops = {
1176         .ndo_uninit     = ipip6_tunnel_uninit,
1177         .ndo_start_xmit = ipip6_tunnel_xmit,
1178         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1179         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1180         .ndo_get_stats64= ipip6_get_stats64,
1181 };
1182
1183 static void ipip6_dev_free(struct net_device *dev)
1184 {
1185         free_percpu(dev->tstats);
1186         free_netdev(dev);
1187 }
1188
1189 static void ipip6_tunnel_setup(struct net_device *dev)
1190 {
1191         dev->netdev_ops         = &ipip6_netdev_ops;
1192         dev->destructor         = ipip6_dev_free;
1193
1194         dev->type               = ARPHRD_SIT;
1195         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1196         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1197         dev->flags              = IFF_NOARP;
1198         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1199         dev->iflink             = 0;
1200         dev->addr_len           = 4;
1201         dev->features           |= NETIF_F_NETNS_LOCAL;
1202         dev->features           |= NETIF_F_LLTX;
1203 }
1204
1205 static int ipip6_tunnel_init(struct net_device *dev)
1206 {
1207         struct ip_tunnel *tunnel = netdev_priv(dev);
1208
1209         tunnel->dev = dev;
1210
1211         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1212         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1213
1214         ipip6_tunnel_bind_dev(dev);
1215         dev->tstats = alloc_percpu(struct pcpu_tstats);
1216         if (!dev->tstats)
1217                 return -ENOMEM;
1218
1219         return 0;
1220 }
1221
1222 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1223 {
1224         struct ip_tunnel *tunnel = netdev_priv(dev);
1225         struct iphdr *iph = &tunnel->parms.iph;
1226         struct net *net = dev_net(dev);
1227         struct sit_net *sitn = net_generic(net, sit_net_id);
1228
1229         tunnel->dev = dev;
1230         strcpy(tunnel->parms.name, dev->name);
1231
1232         iph->version            = 4;
1233         iph->protocol           = IPPROTO_IPV6;
1234         iph->ihl                = 5;
1235         iph->ttl                = 64;
1236
1237         dev->tstats = alloc_percpu(struct pcpu_tstats);
1238         if (!dev->tstats)
1239                 return -ENOMEM;
1240         dev_hold(dev);
1241         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1242         return 0;
1243 }
1244
1245 static void ipip6_netlink_parms(struct nlattr *data[],
1246                                 struct ip_tunnel_parm *parms)
1247 {
1248         memset(parms, 0, sizeof(*parms));
1249
1250         parms->iph.version = 4;
1251         parms->iph.protocol = IPPROTO_IPV6;
1252         parms->iph.ihl = 5;
1253         parms->iph.ttl = 64;
1254
1255         if (!data)
1256                 return;
1257
1258         if (data[IFLA_IPTUN_LINK])
1259                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1260
1261         if (data[IFLA_IPTUN_LOCAL])
1262                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1263
1264         if (data[IFLA_IPTUN_REMOTE])
1265                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1266
1267         if (data[IFLA_IPTUN_TTL]) {
1268                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1269                 if (parms->iph.ttl)
1270                         parms->iph.frag_off = htons(IP_DF);
1271         }
1272
1273         if (data[IFLA_IPTUN_TOS])
1274                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1275
1276         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1277                 parms->iph.frag_off = htons(IP_DF);
1278
1279         if (data[IFLA_IPTUN_FLAGS])
1280                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1281 }
1282
1283 #ifdef CONFIG_IPV6_SIT_6RD
1284 /* This function returns true when 6RD attributes are present in the nl msg */
1285 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1286                                     struct ip_tunnel_6rd *ip6rd)
1287 {
1288         bool ret = false;
1289         memset(ip6rd, 0, sizeof(*ip6rd));
1290
1291         if (!data)
1292                 return ret;
1293
1294         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1295                 ret = true;
1296                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1297                            sizeof(struct in6_addr));
1298         }
1299
1300         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1301                 ret = true;
1302                 ip6rd->relay_prefix =
1303                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1304         }
1305
1306         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1307                 ret = true;
1308                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1309         }
1310
1311         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1312                 ret = true;
1313                 ip6rd->relay_prefixlen =
1314                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1315         }
1316
1317         return ret;
1318 }
1319 #endif
1320
1321 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1322                          struct nlattr *tb[], struct nlattr *data[])
1323 {
1324         struct net *net = dev_net(dev);
1325         struct ip_tunnel *nt;
1326 #ifdef CONFIG_IPV6_SIT_6RD
1327         struct ip_tunnel_6rd ip6rd;
1328 #endif
1329         int err;
1330
1331         nt = netdev_priv(dev);
1332         ipip6_netlink_parms(data, &nt->parms);
1333
1334         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1335                 return -EEXIST;
1336
1337         err = ipip6_tunnel_create(dev);
1338         if (err < 0)
1339                 return err;
1340
1341 #ifdef CONFIG_IPV6_SIT_6RD
1342         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1343                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1344 #endif
1345
1346         return err;
1347 }
1348
1349 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1350                           struct nlattr *data[])
1351 {
1352         struct ip_tunnel *t;
1353         struct ip_tunnel_parm p;
1354         struct net *net = dev_net(dev);
1355         struct sit_net *sitn = net_generic(net, sit_net_id);
1356 #ifdef CONFIG_IPV6_SIT_6RD
1357         struct ip_tunnel_6rd ip6rd;
1358 #endif
1359
1360         if (dev == sitn->fb_tunnel_dev)
1361                 return -EINVAL;
1362
1363         ipip6_netlink_parms(data, &p);
1364
1365         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1366             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1367                 return -EINVAL;
1368
1369         t = ipip6_tunnel_locate(net, &p, 0);
1370
1371         if (t) {
1372                 if (t->dev != dev)
1373                         return -EEXIST;
1374         } else
1375                 t = netdev_priv(dev);
1376
1377         ipip6_tunnel_update(t, &p);
1378
1379 #ifdef CONFIG_IPV6_SIT_6RD
1380         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1381                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1382 #endif
1383
1384         return 0;
1385 }
1386
1387 static size_t ipip6_get_size(const struct net_device *dev)
1388 {
1389         return
1390                 /* IFLA_IPTUN_LINK */
1391                 nla_total_size(4) +
1392                 /* IFLA_IPTUN_LOCAL */
1393                 nla_total_size(4) +
1394                 /* IFLA_IPTUN_REMOTE */
1395                 nla_total_size(4) +
1396                 /* IFLA_IPTUN_TTL */
1397                 nla_total_size(1) +
1398                 /* IFLA_IPTUN_TOS */
1399                 nla_total_size(1) +
1400                 /* IFLA_IPTUN_PMTUDISC */
1401                 nla_total_size(1) +
1402                 /* IFLA_IPTUN_FLAGS */
1403                 nla_total_size(2) +
1404 #ifdef CONFIG_IPV6_SIT_6RD
1405                 /* IFLA_IPTUN_6RD_PREFIX */
1406                 nla_total_size(sizeof(struct in6_addr)) +
1407                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1408                 nla_total_size(4) +
1409                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1410                 nla_total_size(2) +
1411                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1412                 nla_total_size(2) +
1413 #endif
1414                 0;
1415 }
1416
1417 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1418 {
1419         struct ip_tunnel *tunnel = netdev_priv(dev);
1420         struct ip_tunnel_parm *parm = &tunnel->parms;
1421
1422         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1423             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1424             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1425             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1426             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1427             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1428                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1429             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1430                 goto nla_put_failure;
1431
1432 #ifdef CONFIG_IPV6_SIT_6RD
1433         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1434                     &tunnel->ip6rd.prefix) ||
1435             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1436                          tunnel->ip6rd.relay_prefix) ||
1437             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1438                         tunnel->ip6rd.prefixlen) ||
1439             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1440                         tunnel->ip6rd.relay_prefixlen))
1441                 goto nla_put_failure;
1442 #endif
1443
1444         return 0;
1445
1446 nla_put_failure:
1447         return -EMSGSIZE;
1448 }
1449
1450 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1451         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1452         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1453         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1454         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1455         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1456         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1457         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1458 #ifdef CONFIG_IPV6_SIT_6RD
1459         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1460         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1461         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1462         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1463 #endif
1464 };
1465
1466 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1467         .kind           = "sit",
1468         .maxtype        = IFLA_IPTUN_MAX,
1469         .policy         = ipip6_policy,
1470         .priv_size      = sizeof(struct ip_tunnel),
1471         .setup          = ipip6_tunnel_setup,
1472         .newlink        = ipip6_newlink,
1473         .changelink     = ipip6_changelink,
1474         .get_size       = ipip6_get_size,
1475         .fill_info      = ipip6_fill_info,
1476 };
1477
1478 static struct xfrm_tunnel sit_handler __read_mostly = {
1479         .handler        =       ipip6_rcv,
1480         .err_handler    =       ipip6_err,
1481         .priority       =       1,
1482 };
1483
1484 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1485 {
1486         int prio;
1487
1488         for (prio = 1; prio < 4; prio++) {
1489                 int h;
1490                 for (h = 0; h < HASH_SIZE; h++) {
1491                         struct ip_tunnel *t;
1492
1493                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1494                         while (t != NULL) {
1495                                 unregister_netdevice_queue(t->dev, head);
1496                                 t = rtnl_dereference(t->next);
1497                         }
1498                 }
1499         }
1500 }
1501
1502 static int __net_init sit_init_net(struct net *net)
1503 {
1504         struct sit_net *sitn = net_generic(net, sit_net_id);
1505         struct ip_tunnel *t;
1506         int err;
1507
1508         sitn->tunnels[0] = sitn->tunnels_wc;
1509         sitn->tunnels[1] = sitn->tunnels_l;
1510         sitn->tunnels[2] = sitn->tunnels_r;
1511         sitn->tunnels[3] = sitn->tunnels_r_l;
1512
1513         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1514                                            ipip6_tunnel_setup);
1515         if (!sitn->fb_tunnel_dev) {
1516                 err = -ENOMEM;
1517                 goto err_alloc_dev;
1518         }
1519         dev_net_set(sitn->fb_tunnel_dev, net);
1520
1521         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1522         if (err)
1523                 goto err_dev_free;
1524
1525         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1526
1527         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1528                 goto err_reg_dev;
1529
1530         t = netdev_priv(sitn->fb_tunnel_dev);
1531
1532         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1533         return 0;
1534
1535 err_reg_dev:
1536         dev_put(sitn->fb_tunnel_dev);
1537 err_dev_free:
1538         ipip6_dev_free(sitn->fb_tunnel_dev);
1539 err_alloc_dev:
1540         return err;
1541 }
1542
1543 static void __net_exit sit_exit_net(struct net *net)
1544 {
1545         struct sit_net *sitn = net_generic(net, sit_net_id);
1546         LIST_HEAD(list);
1547
1548         rtnl_lock();
1549         sit_destroy_tunnels(sitn, &list);
1550         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1551         unregister_netdevice_many(&list);
1552         rtnl_unlock();
1553 }
1554
1555 static struct pernet_operations sit_net_ops = {
1556         .init = sit_init_net,
1557         .exit = sit_exit_net,
1558         .id   = &sit_net_id,
1559         .size = sizeof(struct sit_net),
1560 };
1561
1562 static void __exit sit_cleanup(void)
1563 {
1564         rtnl_link_unregister(&sit_link_ops);
1565         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1566
1567         unregister_pernet_device(&sit_net_ops);
1568         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1569 }
1570
1571 static int __init sit_init(void)
1572 {
1573         int err;
1574
1575         pr_info("IPv6 over IPv4 tunneling driver\n");
1576
1577         err = register_pernet_device(&sit_net_ops);
1578         if (err < 0)
1579                 return err;
1580         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1581         if (err < 0) {
1582                 pr_info("%s: can't add protocol\n", __func__);
1583                 goto xfrm_tunnel_failed;
1584         }
1585         err = rtnl_link_register(&sit_link_ops);
1586         if (err < 0)
1587                 goto rtnl_link_failed;
1588
1589 out:
1590         return err;
1591
1592 rtnl_link_failed:
1593         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1594 xfrm_tunnel_failed:
1595         unregister_pernet_device(&sit_net_ops);
1596         goto out;
1597 }
1598
1599 module_init(sit_init);
1600 module_exit(sit_cleanup);
1601 MODULE_LICENSE("GPL");
1602 MODULE_ALIAS_NETDEV("sit0");