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/ip_tunnels.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 bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  76                       __be32 *v4dst);
  77 static struct rtnl_link_ops sit_link_ops __read_mostly;
  78
  79 static int sit_net_id __read_mostly;
  80 struct sit_net {
  81         struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
  82         struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
  83         struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
  84         struct ip_tunnel __rcu *tunnels_wc[1];
  85         struct ip_tunnel __rcu **tunnels[4];
  86
  87         struct net_device *fb_tunnel_dev;
  88 };
  89
  90 /*
  91  * Must be invoked with rcu_read_lock
  92  */
  93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  94                 struct net_device *dev, __be32 remote, __be32 local)
  95 {
  96         unsigned int h0 = HASH(remote);
  97         unsigned int h1 = HASH(local);
  98         struct ip_tunnel *t;
  99         struct sit_net *sitn = net_generic(net, sit_net_id);
 100
 101         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 102                 if (local == t->parms.iph.saddr &&
 103                     remote == t->parms.iph.daddr &&
 104                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 105                     (t->dev->flags & IFF_UP))
 106                         return t;
 107         }
 108         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 109                 if (remote == t->parms.iph.daddr &&
 110                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 111                     (t->dev->flags & IFF_UP))
 112                         return t;
 113         }
 114         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 115                 if (local == t->parms.iph.saddr &&
 116                     (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
 117                     (t->dev->flags & IFF_UP))
 118                         return t;
 119         }
 120         t = rcu_dereference(sitn->tunnels_wc[0]);
 121         if ((t != NULL) && (t->dev->flags & IFF_UP))
 122                 return t;
 123         return NULL;
 124 }
 125
 126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 127                 struct ip_tunnel_parm *parms)
 128 {
 129         __be32 remote = parms->iph.daddr;
 130         __be32 local = parms->iph.saddr;
 131         unsigned int h = 0;
 132         int prio = 0;
 133
 134         if (remote) {
 135                 prio |= 2;
 136                 h ^= HASH(remote);
 137         }
 138         if (local) {
 139                 prio |= 1;
 140                 h ^= HASH(local);
 141         }
 142         return &sitn->tunnels[prio][h];
 143 }
 144
 145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 146                 struct ip_tunnel *t)
 147 {
 148         return __ipip6_bucket(sitn, &t->parms);
 149 }
 150
 151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 152 {
 153         struct ip_tunnel __rcu **tp;
 154         struct ip_tunnel *iter;
 155
 156         for (tp = ipip6_bucket(sitn, t);
 157              (iter = rtnl_dereference(*tp)) != NULL;
 158              tp = &iter->next) {
 159                 if (t == iter) {
 160                         rcu_assign_pointer(*tp, t->next);
 161                         break;
 162                 }
 163         }
 164 }
 165
 166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 167 {
 168         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 169
 170         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 171         rcu_assign_pointer(*tp, t);
 172 }
 173
 174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 175 {
 176 #ifdef CONFIG_IPV6_SIT_6RD
 177         struct ip_tunnel *t = netdev_priv(dev);
 178
 179         if (t->dev == sitn->fb_tunnel_dev) {
 180                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 181                 t->ip6rd.relay_prefix = 0;
 182                 t->ip6rd.prefixlen = 16;
 183                 t->ip6rd.relay_prefixlen = 0;
 184         } else {
 185                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 186                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 187         }
 188 #endif
 189 }
 190
 191 static int ipip6_tunnel_create(struct net_device *dev)
 192 {
 193         struct ip_tunnel *t = netdev_priv(dev);
 194         struct net *net = dev_net(dev);
 195         struct sit_net *sitn = net_generic(net, sit_net_id);
 196         int err;
 197
 198         err = ipip6_tunnel_init(dev);
 199         if (err < 0)
 200                 goto out;
 201         ipip6_tunnel_clone_6rd(dev, sitn);
 202
 203         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 204                 dev->priv_flags |= IFF_ISATAP;
 205
 206         err = register_netdevice(dev);
 207         if (err < 0)
 208                 goto out;
 209
 210         strcpy(t->parms.name, dev->name);
 211         dev->rtnl_link_ops = &sit_link_ops;
 212
 213         dev_hold(dev);
 214
 215         ipip6_tunnel_link(sitn, t);
 216         return 0;
 217
 218 out:
 219         return err;
 220 }
 221
 222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 223                 struct ip_tunnel_parm *parms, int create)
 224 {
 225         __be32 remote = parms->iph.daddr;
 226         __be32 local = parms->iph.saddr;
 227         struct ip_tunnel *t, *nt;
 228         struct ip_tunnel __rcu **tp;
 229         struct net_device *dev;
 230         char name[IFNAMSIZ];
 231         struct sit_net *sitn = net_generic(net, sit_net_id);
 232
 233         for (tp = __ipip6_bucket(sitn, parms);
 234             (t = rtnl_dereference(*tp)) != NULL;
 235              tp = &t->next) {
 236                 if (local == t->parms.iph.saddr &&
 237                     remote == t->parms.iph.daddr &&
 238                     parms->link == t->parms.link) {
 239                         if (create)
 240                                 return NULL;
 241                         else
 242                                 return t;
 243                 }
 244         }
 245         if (!create)
 246                 goto failed;
 247
 248         if (parms->name[0])
 249                 strlcpy(name, parms->name, IFNAMSIZ);
 250         else
 251                 strcpy(name, "sit%d");
 252
 253         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 254         if (dev == NULL)
 255                 return NULL;
 256
 257         dev_net_set(dev, net);
 258
 259         nt = netdev_priv(dev);
 260
 261         nt->parms = *parms;
 262         if (ipip6_tunnel_create(dev) < 0)
 263                 goto failed_free;
 264
 265         return nt;
 266
 267 failed_free:
 268         ipip6_dev_free(dev);
 269 failed:
 270         return NULL;
 271 }
 272
 273 #define for_each_prl_rcu(start)                 \
 274         for (prl = rcu_dereference(start);      \
 275              prl;                               \
 276              prl = rcu_dereference(prl->next))
 277
 278 static struct ip_tunnel_prl_entry *
 279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 280 {
 281         struct ip_tunnel_prl_entry *prl;
 282
 283         for_each_prl_rcu(t->prl)
 284                 if (prl->addr == addr)
 285                         break;
 286         return prl;
 287
 288 }
 289
 290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 291                                 struct ip_tunnel_prl __user *a)
 292 {
 293         struct ip_tunnel_prl kprl, *kp;
 294         struct ip_tunnel_prl_entry *prl;
 295         unsigned int cmax, c = 0, ca, len;
 296         int ret = 0;
 297
 298         if (copy_from_user(&kprl, a, sizeof(kprl)))
 299                 return -EFAULT;
 300         cmax = kprl.datalen / sizeof(kprl);
 301         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 302                 cmax = 1;
 303
 304         /* For simple GET or for root users,
 305          * we try harder to allocate.
 306          */
 307         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 308                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 309                 NULL;
 310
 311         rcu_read_lock();
 312
 313         ca = t->prl_count < cmax ? t->prl_count : cmax;
 314
 315         if (!kp) {
 316                 /* We don't try hard to allocate much memory for
 317                  * non-root users.
 318                  * For root users, retry allocating enough memory for
 319                  * the answer.
 320                  */
 321                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 322                 if (!kp) {
 323                         ret = -ENOMEM;
 324                         goto out;
 325                 }
 326         }
 327
 328         c = 0;
 329         for_each_prl_rcu(t->prl) {
 330                 if (c >= cmax)
 331                         break;
 332                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 333                         continue;
 334                 kp[c].addr = prl->addr;
 335                 kp[c].flags = prl->flags;
 336                 c++;
 337                 if (kprl.addr != htonl(INADDR_ANY))
 338                         break;
 339         }
 340 out:
 341         rcu_read_unlock();
 342
 343         len = sizeof(*kp) * c;
 344         ret = 0;
 345         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 346                 ret = -EFAULT;
 347
 348         kfree(kp);
 349
 350         return ret;
 351 }
 352
 353 static int
 354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 355 {
 356         struct ip_tunnel_prl_entry *p;
 357         int err = 0;
 358
 359         if (a->addr == htonl(INADDR_ANY))
 360                 return -EINVAL;
 361
 362         ASSERT_RTNL();
 363
 364         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 365                 if (p->addr == a->addr) {
 366                         if (chg) {
 367                                 p->flags = a->flags;
 368                                 goto out;
 369                         }
 370                         err = -EEXIST;
 371                         goto out;
 372                 }
 373         }
 374
 375         if (chg) {
 376                 err = -ENXIO;
 377                 goto out;
 378         }
 379
 380         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 381         if (!p) {
 382                 err = -ENOBUFS;
 383                 goto out;
 384         }
 385
 386         p->next = t->prl;
 387         p->addr = a->addr;
 388         p->flags = a->flags;
 389         t->prl_count++;
 390         rcu_assign_pointer(t->prl, p);
 391 out:
 392         return err;
 393 }
 394
 395 static void prl_list_destroy_rcu(struct rcu_head *head)
 396 {
 397         struct ip_tunnel_prl_entry *p, *n;
 398
 399         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 400         do {
 401                 n = rcu_dereference_protected(p->next, 1);
 402                 kfree(p);
 403                 p = n;
 404         } while (p);
 405 }
 406
 407 static int
 408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 409 {
 410         struct ip_tunnel_prl_entry *x;
 411         struct ip_tunnel_prl_entry __rcu **p;
 412         int err = 0;
 413
 414         ASSERT_RTNL();
 415
 416         if (a && a->addr != htonl(INADDR_ANY)) {
 417                 for (p = &t->prl;
 418                      (x = rtnl_dereference(*p)) != NULL;
 419                      p = &x->next) {
 420                         if (x->addr == a->addr) {
 421                                 *p = x->next;
 422                                 kfree_rcu(x, rcu_head);
 423                                 t->prl_count--;
 424                                 goto out;
 425                         }
 426                 }
 427                 err = -ENXIO;
 428         } else {
 429                 x = rtnl_dereference(t->prl);
 430                 if (x) {
 431                         t->prl_count = 0;
 432                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 433                         t->prl = NULL;
 434                 }
 435         }
 436 out:
 437         return err;
 438 }
 439
 440 static int
 441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 442 {
 443         struct ip_tunnel_prl_entry *p;
 444         int ok = 1;
 445
 446         rcu_read_lock();
 447         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 448         if (p) {
 449                 if (p->flags & PRL_DEFAULT)
 450                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 451                 else
 452                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 453         } else {
 454                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 455
 456                 if (ipv6_addr_is_isatap(addr6) &&
 457                     (addr6->s6_addr32[3] == iph->saddr) &&
 458                     ipv6_chk_prefix(addr6, t->dev))
 459                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 460                 else
 461                         ok = 0;
 462         }
 463         rcu_read_unlock();
 464         return ok;
 465 }
 466
 467 static void ipip6_tunnel_uninit(struct net_device *dev)
 468 {
 469         struct ip_tunnel *tunnel = netdev_priv(dev);
 470         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 471
 472         if (dev == sitn->fb_tunnel_dev) {
 473                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 474         } else {
 475                 ipip6_tunnel_unlink(sitn, tunnel);
 476                 ipip6_tunnel_del_prl(tunnel, NULL);
 477         }
 478         dev_put(dev);
 479 }
 480
 481
 482 static int ipip6_err(struct sk_buff *skb, u32 info)
 483 {
 484
 485 /* All the routers (except for Linux) return only
 486    8 bytes of packet payload. It means, that precise relaying of
 487    ICMP in the real Internet is absolutely infeasible.
 488  */
 489         const struct iphdr *iph = (const struct iphdr *)skb->data;
 490         const int type = icmp_hdr(skb)->type;
 491         const int code = icmp_hdr(skb)->code;
 492         struct ip_tunnel *t;
 493         int err;
 494
 495         switch (type) {
 496         default:
 497         case ICMP_PARAMETERPROB:
 498                 return 0;
 499
 500         case ICMP_DEST_UNREACH:
 501                 switch (code) {
 502                 case ICMP_SR_FAILED:
 503                 case ICMP_PORT_UNREACH:
 504                         /* Impossible event. */
 505                         return 0;
 506                 default:
 507                         /* All others are translated to HOST_UNREACH.
 508                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 509                            I believe they are just ether pollution. --ANK
 510                          */
 511                         break;
 512                 }
 513                 break;
 514         case ICMP_TIME_EXCEEDED:
 515                 if (code != ICMP_EXC_TTL)
 516                         return 0;
 517                 break;
 518         case ICMP_REDIRECT:
 519                 break;
 520         }
 521
 522         err = -ENOENT;
 523
 524         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 525                                 skb->dev,
 526                                 iph->daddr,
 527                                 iph->saddr);
 528         if (t == NULL)
 529                 goto out;
 530
 531         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 532                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 533                                  t->dev->ifindex, 0, IPPROTO_IPV6, 0);
 534                 err = 0;
 535                 goto out;
 536         }
 537         if (type == ICMP_REDIRECT) {
 538                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 539                               IPPROTO_IPV6, 0);
 540                 err = 0;
 541                 goto out;
 542         }
 543
 544         if (t->parms.iph.daddr == 0)
 545                 goto out;
 546
 547         err = 0;
 548         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 549                 goto out;
 550
 551         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 552                 t->err_count++;
 553         else
 554                 t->err_count = 1;
 555         t->err_time = jiffies;
 556 out:
 557         return err;
 558 }
 559
 560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 561                                   const struct in6_addr *v6addr)
 562 {
 563         __be32 v4embed = 0;
 564         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 565                 return true;
 566         return false;
 567 }
 568
 569 static int ipip6_rcv(struct sk_buff *skb)
 570 {
 571         const struct iphdr *iph = ip_hdr(skb);
 572         struct ip_tunnel *tunnel;
 573         int err;
 574
 575         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 576                                      iph->saddr, iph->daddr);
 577         if (tunnel != NULL) {
 578                 struct pcpu_tstats *tstats;
 579
 580                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 581                     tunnel->parms.iph.protocol != 0)
 582                         goto out;
 583
 584                 skb->mac_header = skb->network_header;
 585                 skb_reset_network_header(skb);
 586                 IPCB(skb)->flags = 0;
 587                 skb->protocol = htons(ETH_P_IPV6);
 588
 589                 if (tunnel->dev->priv_flags & IFF_ISATAP) {
 590                         if (!isatap_chksrc(skb, iph, tunnel)) {
 591                                 tunnel->dev->stats.rx_errors++;
 592                                 goto out;
 593                         }
 594                 } else if (!(tunnel->dev->flags&IFF_POINTOPOINT)) {
 595                         if (is_spoofed_6rd(tunnel, iph->saddr,
 596                                            &ipv6_hdr(skb)->saddr) ||
 597                             is_spoofed_6rd(tunnel, iph->daddr,
 598                                            &ipv6_hdr(skb)->daddr)) {
 599                                 tunnel->dev->stats.rx_errors++;
 600                                 goto out;
 601                         }
 602                 }
 603
 604                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 605
 606                 err = IP_ECN_decapsulate(iph, skb);
 607                 if (unlikely(err)) {
 608                         if (log_ecn_error)
 609                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 610                                                      &iph->saddr, iph->tos);
 611                         if (err > 1) {
 612                                 ++tunnel->dev->stats.rx_frame_errors;
 613                                 ++tunnel->dev->stats.rx_errors;
 614                                 goto out;
 615                         }
 616                 }
 617
 618                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 619                 tstats->rx_packets++;
 620                 tstats->rx_bytes += skb->len;
 621
 622                 netif_rx(skb);
 623
 624                 return 0;
 625         }
 626
 627         /* no tunnel matched,  let upstream know, ipsec may handle it */
 628         return 1;
 629 out:
 630         kfree_skb(skb);
 631         return 0;
 632 }
 633
 634 static const struct tnl_ptk_info tpi = {
 635         /* no tunnel info required for ipip. */
 636         .proto = htons(ETH_P_IP),
 637 };
 638
 639 static int ipip_rcv(struct sk_buff *skb)
 640 {
 641         const struct iphdr *iph;
 642         struct ip_tunnel *tunnel;
 643
 644         iph = ip_hdr(skb);
 645         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 646                                      iph->saddr, iph->daddr);
 647         if (tunnel != NULL) {
 648                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 649                     tunnel->parms.iph.protocol != 0)
 650                         goto drop;
 651
 652                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 653                         goto drop;
 654                 if (iptunnel_pull_header(skb, 0, tpi.proto))
 655                         goto drop;
 656                 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 657         }
 658
 659         return 1;
 660
 661 drop:
 662         kfree_skb(skb);
 663         return 0;
 664 }
 665
 666 /*
 667  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 668  * stores the embedded IPv4 address in v4dst and returns true.
 669  */
 670 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 671                       __be32 *v4dst)
 672 {
 673 #ifdef CONFIG_IPV6_SIT_6RD
 674         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 675                               tunnel->ip6rd.prefixlen)) {
 676                 unsigned int pbw0, pbi0;
 677                 int pbi1;
 678                 u32 d;
 679
 680                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 681                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 682
 683                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 684                     tunnel->ip6rd.relay_prefixlen;
 685
 686                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 687                 if (pbi1 > 0)
 688                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 689                              (32 - pbi1);
 690
 691                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 692                 return true;
 693         }
 694 #else
 695         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 696                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 697                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 698                 return true;
 699         }
 700 #endif
 701         return false;
 702 }
 703
 704 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 705                              const struct in6_addr *v6dst)
 706 {
 707         __be32 dst = 0;
 708         check_6rd(tunnel, v6dst, &dst);
 709         return dst;
 710 }
 711
 712 /*
 713  *      This function assumes it is being called from dev_queue_xmit()
 714  *      and that skb is filled properly by that function.
 715  */
 716
 717 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 718                                      struct net_device *dev)
 719 {
 720         struct ip_tunnel *tunnel = netdev_priv(dev);
 721         const struct iphdr  *tiph = &tunnel->parms.iph;
 722         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 723         u8     tos = tunnel->parms.iph.tos;
 724         __be16 df = tiph->frag_off;
 725         struct rtable *rt;                      /* Route to the other host */
 726         struct net_device *tdev;                /* Device to other host */
 727         unsigned int max_headroom;              /* The extra header space needed */
 728         __be32 dst = tiph->daddr;
 729         struct flowi4 fl4;
 730         int    mtu;
 731         const struct in6_addr *addr6;
 732         int addr_type;
 733         u8 ttl;
 734         int err;
 735
 736         if (skb->protocol != htons(ETH_P_IPV6))
 737                 goto tx_error;
 738
 739         if (tos == 1)
 740                 tos = ipv6_get_dsfield(iph6);
 741
 742         /* ISATAP (RFC4214) - must come before 6to4 */
 743         if (dev->priv_flags & IFF_ISATAP) {
 744                 struct neighbour *neigh = NULL;
 745                 bool do_tx_error = false;
 746
 747                 if (skb_dst(skb))
 748                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 749
 750                 if (neigh == NULL) {
 751                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 752                         goto tx_error;
 753                 }
 754
 755                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 756                 addr_type = ipv6_addr_type(addr6);
 757
 758                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 759                      ipv6_addr_is_isatap(addr6))
 760                         dst = addr6->s6_addr32[3];
 761                 else
 762                         do_tx_error = true;
 763
 764                 neigh_release(neigh);
 765                 if (do_tx_error)
 766                         goto tx_error;
 767         }
 768
 769         if (!dst)
 770                 dst = try_6rd(tunnel, &iph6->daddr);
 771
 772         if (!dst) {
 773                 struct neighbour *neigh = NULL;
 774                 bool do_tx_error = false;
 775
 776                 if (skb_dst(skb))
 777                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 778
 779                 if (neigh == NULL) {
 780                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 781                         goto tx_error;
 782                 }
 783
 784                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 785                 addr_type = ipv6_addr_type(addr6);
 786
 787                 if (addr_type == IPV6_ADDR_ANY) {
 788                         addr6 = &ipv6_hdr(skb)->daddr;
 789                         addr_type = ipv6_addr_type(addr6);
 790                 }
 791
 792                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 793                         dst = addr6->s6_addr32[3];
 794                 else
 795                         do_tx_error = true;
 796
 797                 neigh_release(neigh);
 798                 if (do_tx_error)
 799                         goto tx_error;
 800         }
 801
 802         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 803                                    dst, tiph->saddr,
 804                                    0, 0,
 805                                    IPPROTO_IPV6, RT_TOS(tos),
 806                                    tunnel->parms.link);
 807         if (IS_ERR(rt)) {
 808                 dev->stats.tx_carrier_errors++;
 809                 goto tx_error_icmp;
 810         }
 811         if (rt->rt_type != RTN_UNICAST) {
 812                 ip_rt_put(rt);
 813                 dev->stats.tx_carrier_errors++;
 814                 goto tx_error_icmp;
 815         }
 816         tdev = rt->dst.dev;
 817
 818         if (tdev == dev) {
 819                 ip_rt_put(rt);
 820                 dev->stats.collisions++;
 821                 goto tx_error;
 822         }
 823
 824         if (df) {
 825                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 826
 827                 if (mtu < 68) {
 828                         dev->stats.collisions++;
 829                         ip_rt_put(rt);
 830                         goto tx_error;
 831                 }
 832
 833                 if (mtu < IPV6_MIN_MTU) {
 834                         mtu = IPV6_MIN_MTU;
 835                         df = 0;
 836                 }
 837
 838                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 839                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 840
 841                 if (skb->len > mtu) {
 842                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 843                         ip_rt_put(rt);
 844                         goto tx_error;
 845                 }
 846         }
 847
 848         if (tunnel->err_count > 0) {
 849                 if (time_before(jiffies,
 850                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 851                         tunnel->err_count--;
 852                         dst_link_failure(skb);
 853                 } else
 854                         tunnel->err_count = 0;
 855         }
 856
 857         /*
 858          * Okay, now see if we can stuff it in the buffer as-is.
 859          */
 860         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 861
 862         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 863             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 864                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 865                 if (!new_skb) {
 866                         ip_rt_put(rt);
 867                         dev->stats.tx_dropped++;
 868                         dev_kfree_skb(skb);
 869                         return NETDEV_TX_OK;
 870                 }
 871                 if (skb->sk)
 872                         skb_set_owner_w(new_skb, skb->sk);
 873                 dev_kfree_skb(skb);
 874                 skb = new_skb;
 875                 iph6 = ipv6_hdr(skb);
 876         }
 877         ttl = tiph->ttl;
 878         if (ttl == 0)
 879                 ttl = iph6->hop_limit;
 880         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 881
 882         if (likely(!skb->encapsulation)) {
 883                 skb_reset_inner_headers(skb);
 884                 skb->encapsulation = 1;
 885         }
 886
 887         err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, IPPROTO_IPV6, tos,
 888                             ttl, df, !net_eq(tunnel->net, dev_net(dev)));
 889         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 890         return NETDEV_TX_OK;
 891
 892 tx_error_icmp:
 893         dst_link_failure(skb);
 894 tx_error:
 895         dev->stats.tx_errors++;
 896         dev_kfree_skb(skb);
 897         return NETDEV_TX_OK;
 898 }
 899
 900 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 901 {
 902         struct ip_tunnel *tunnel = netdev_priv(dev);
 903         const struct iphdr  *tiph = &tunnel->parms.iph;
 904
 905         if (likely(!skb->encapsulation)) {
 906                 skb_reset_inner_headers(skb);
 907                 skb->encapsulation = 1;
 908         }
 909
 910         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
 911         return NETDEV_TX_OK;
 912 }
 913
 914 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
 915                                    struct net_device *dev)
 916 {
 917         switch (skb->protocol) {
 918         case htons(ETH_P_IP):
 919                 ipip_tunnel_xmit(skb, dev);
 920                 break;
 921         case htons(ETH_P_IPV6):
 922                 ipip6_tunnel_xmit(skb, dev);
 923                 break;
 924         default:
 925                 goto tx_err;
 926         }
 927
 928         return NETDEV_TX_OK;
 929
 930 tx_err:
 931         dev->stats.tx_errors++;
 932         dev_kfree_skb(skb);
 933         return NETDEV_TX_OK;
 934
 935 }
 936
 937 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 938 {
 939         struct net_device *tdev = NULL;
 940         struct ip_tunnel *tunnel;
 941         const struct iphdr *iph;
 942         struct flowi4 fl4;
 943
 944         tunnel = netdev_priv(dev);
 945         iph = &tunnel->parms.iph;
 946
 947         if (iph->daddr) {
 948                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
 949                                                           NULL,
 950                                                           iph->daddr, iph->saddr,
 951                                                           0, 0,
 952                                                           IPPROTO_IPV6,
 953                                                           RT_TOS(iph->tos),
 954                                                           tunnel->parms.link);
 955
 956                 if (!IS_ERR(rt)) {
 957                         tdev = rt->dst.dev;
 958                         ip_rt_put(rt);
 959                 }
 960                 dev->flags |= IFF_POINTOPOINT;
 961         }
 962
 963         if (!tdev && tunnel->parms.link)
 964                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
 965
 966         if (tdev) {
 967                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 968                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 969                 if (dev->mtu < IPV6_MIN_MTU)
 970                         dev->mtu = IPV6_MIN_MTU;
 971         }
 972         dev->iflink = tunnel->parms.link;
 973 }
 974
 975 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 976 {
 977         struct net *net = t->net;
 978         struct sit_net *sitn = net_generic(net, sit_net_id);
 979
 980         ipip6_tunnel_unlink(sitn, t);
 981         synchronize_net();
 982         t->parms.iph.saddr = p->iph.saddr;
 983         t->parms.iph.daddr = p->iph.daddr;
 984         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 985         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 986         ipip6_tunnel_link(sitn, t);
 987         t->parms.iph.ttl = p->iph.ttl;
 988         t->parms.iph.tos = p->iph.tos;
 989         if (t->parms.link != p->link) {
 990                 t->parms.link = p->link;
 991                 ipip6_tunnel_bind_dev(t->dev);
 992         }
 993         netdev_state_change(t->dev);
 994 }
 995
 996 #ifdef CONFIG_IPV6_SIT_6RD
 997 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
 998                                    struct ip_tunnel_6rd *ip6rd)
 999 {
1000         struct in6_addr prefix;
1001         __be32 relay_prefix;
1002
1003         if (ip6rd->relay_prefixlen > 32 ||
1004             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1005                 return -EINVAL;
1006
1007         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1008         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1009                 return -EINVAL;
1010         if (ip6rd->relay_prefixlen)
1011                 relay_prefix = ip6rd->relay_prefix &
1012                                htonl(0xffffffffUL <<
1013                                      (32 - ip6rd->relay_prefixlen));
1014         else
1015                 relay_prefix = 0;
1016         if (relay_prefix != ip6rd->relay_prefix)
1017                 return -EINVAL;
1018
1019         t->ip6rd.prefix = prefix;
1020         t->ip6rd.relay_prefix = relay_prefix;
1021         t->ip6rd.prefixlen = ip6rd->prefixlen;
1022         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1023         netdev_state_change(t->dev);
1024         return 0;
1025 }
1026 #endif
1027
1028 static int
1029 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1030 {
1031         int err = 0;
1032         struct ip_tunnel_parm p;
1033         struct ip_tunnel_prl prl;
1034         struct ip_tunnel *t;
1035         struct net *net = dev_net(dev);
1036         struct sit_net *sitn = net_generic(net, sit_net_id);
1037 #ifdef CONFIG_IPV6_SIT_6RD
1038         struct ip_tunnel_6rd ip6rd;
1039 #endif
1040
1041         switch (cmd) {
1042         case SIOCGETTUNNEL:
1043 #ifdef CONFIG_IPV6_SIT_6RD
1044         case SIOCGET6RD:
1045 #endif
1046                 t = NULL;
1047                 if (dev == sitn->fb_tunnel_dev) {
1048                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1049                                 err = -EFAULT;
1050                                 break;
1051                         }
1052                         t = ipip6_tunnel_locate(net, &p, 0);
1053                 }
1054                 if (t == NULL)
1055                         t = netdev_priv(dev);
1056
1057                 err = -EFAULT;
1058                 if (cmd == SIOCGETTUNNEL) {
1059                         memcpy(&p, &t->parms, sizeof(p));
1060                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1061                                          sizeof(p)))
1062                                 goto done;
1063 #ifdef CONFIG_IPV6_SIT_6RD
1064                 } else {
1065                         ip6rd.prefix = t->ip6rd.prefix;
1066                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1067                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1068                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1069                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1070                                          sizeof(ip6rd)))
1071                                 goto done;
1072 #endif
1073                 }
1074                 err = 0;
1075                 break;
1076
1077         case SIOCADDTUNNEL:
1078         case SIOCCHGTUNNEL:
1079                 err = -EPERM;
1080                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1081                         goto done;
1082
1083                 err = -EFAULT;
1084                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1085                         goto done;
1086
1087                 err = -EINVAL;
1088                 if (p.iph.protocol != IPPROTO_IPV6 &&
1089                     p.iph.protocol != IPPROTO_IPIP &&
1090                     p.iph.protocol != 0)
1091                         goto done;
1092                 if (p.iph.version != 4 ||
1093                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1094                         goto done;
1095                 if (p.iph.ttl)
1096                         p.iph.frag_off |= htons(IP_DF);
1097
1098                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1099
1100                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1101                         if (t != NULL) {
1102                                 if (t->dev != dev) {
1103                                         err = -EEXIST;
1104                                         break;
1105                                 }
1106                         } else {
1107                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1108                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1109                                         err = -EINVAL;
1110                                         break;
1111                                 }
1112                                 t = netdev_priv(dev);
1113                         }
1114
1115                         ipip6_tunnel_update(t, &p);
1116                 }
1117
1118                 if (t) {
1119                         err = 0;
1120                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1121                                 err = -EFAULT;
1122                 } else
1123                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1124                 break;
1125
1126         case SIOCDELTUNNEL:
1127                 err = -EPERM;
1128                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1129                         goto done;
1130
1131                 if (dev == sitn->fb_tunnel_dev) {
1132                         err = -EFAULT;
1133                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1134                                 goto done;
1135                         err = -ENOENT;
1136                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1137                                 goto done;
1138                         err = -EPERM;
1139                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1140                                 goto done;
1141                         dev = t->dev;
1142                 }
1143                 unregister_netdevice(dev);
1144                 err = 0;
1145                 break;
1146
1147         case SIOCGETPRL:
1148                 err = -EINVAL;
1149                 if (dev == sitn->fb_tunnel_dev)
1150                         goto done;
1151                 err = -ENOENT;
1152                 if (!(t = netdev_priv(dev)))
1153                         goto done;
1154                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1155                 break;
1156
1157         case SIOCADDPRL:
1158         case SIOCDELPRL:
1159         case SIOCCHGPRL:
1160                 err = -EPERM;
1161                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1162                         goto done;
1163                 err = -EINVAL;
1164                 if (dev == sitn->fb_tunnel_dev)
1165                         goto done;
1166                 err = -EFAULT;
1167                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1168                         goto done;
1169                 err = -ENOENT;
1170                 if (!(t = netdev_priv(dev)))
1171                         goto done;
1172
1173                 switch (cmd) {
1174                 case SIOCDELPRL:
1175                         err = ipip6_tunnel_del_prl(t, &prl);
1176                         break;
1177                 case SIOCADDPRL:
1178                 case SIOCCHGPRL:
1179                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1180                         break;
1181                 }
1182                 netdev_state_change(dev);
1183                 break;
1184
1185 #ifdef CONFIG_IPV6_SIT_6RD
1186         case SIOCADD6RD:
1187         case SIOCCHG6RD:
1188         case SIOCDEL6RD:
1189                 err = -EPERM;
1190                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1191                         goto done;
1192
1193                 err = -EFAULT;
1194                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1195                                    sizeof(ip6rd)))
1196                         goto done;
1197
1198                 t = netdev_priv(dev);
1199
1200                 if (cmd != SIOCDEL6RD) {
1201                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1202                         if (err < 0)
1203                                 goto done;
1204                 } else
1205                         ipip6_tunnel_clone_6rd(dev, sitn);
1206
1207                 err = 0;
1208                 break;
1209 #endif
1210
1211         default:
1212                 err = -EINVAL;
1213         }
1214
1215 done:
1216         return err;
1217 }
1218
1219 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1220 {
1221         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1222                 return -EINVAL;
1223         dev->mtu = new_mtu;
1224         return 0;
1225 }
1226
1227 static const struct net_device_ops ipip6_netdev_ops = {
1228         .ndo_uninit     = ipip6_tunnel_uninit,
1229         .ndo_start_xmit = sit_tunnel_xmit,
1230         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1231         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1232         .ndo_get_stats64 = ip_tunnel_get_stats64,
1233 };
1234
1235 static void ipip6_dev_free(struct net_device *dev)
1236 {
1237         free_percpu(dev->tstats);
1238         free_netdev(dev);
1239 }
1240
1241 static void ipip6_tunnel_setup(struct net_device *dev)
1242 {
1243         dev->netdev_ops         = &ipip6_netdev_ops;
1244         dev->destructor         = ipip6_dev_free;
1245
1246         dev->type               = ARPHRD_SIT;
1247         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1248         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1249         dev->flags              = IFF_NOARP;
1250         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1251         dev->iflink             = 0;
1252         dev->addr_len           = 4;
1253         dev->features           |= NETIF_F_LLTX;
1254 }
1255
1256 static int ipip6_tunnel_init(struct net_device *dev)
1257 {
1258         struct ip_tunnel *tunnel = netdev_priv(dev);
1259
1260         tunnel->dev = dev;
1261         tunnel->net = dev_net(dev);
1262
1263         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1264         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1265
1266         ipip6_tunnel_bind_dev(dev);
1267         dev->tstats = alloc_percpu(struct pcpu_tstats);
1268         if (!dev->tstats)
1269                 return -ENOMEM;
1270
1271         return 0;
1272 }
1273
1274 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1275 {
1276         struct ip_tunnel *tunnel = netdev_priv(dev);
1277         struct iphdr *iph = &tunnel->parms.iph;
1278         struct net *net = dev_net(dev);
1279         struct sit_net *sitn = net_generic(net, sit_net_id);
1280
1281         tunnel->dev = dev;
1282         tunnel->net = dev_net(dev);
1283         strcpy(tunnel->parms.name, dev->name);
1284
1285         iph->version            = 4;
1286         iph->protocol           = IPPROTO_IPV6;
1287         iph->ihl                = 5;
1288         iph->ttl                = 64;
1289
1290         dev->tstats = alloc_percpu(struct pcpu_tstats);
1291         if (!dev->tstats)
1292                 return -ENOMEM;
1293         dev_hold(dev);
1294         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1295         return 0;
1296 }
1297
1298 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1299 {
1300         u8 proto;
1301
1302         if (!data || !data[IFLA_IPTUN_PROTO])
1303                 return 0;
1304
1305         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1306         if (proto != IPPROTO_IPV6 &&
1307             proto != IPPROTO_IPIP &&
1308             proto != 0)
1309                 return -EINVAL;
1310
1311         return 0;
1312 }
1313
1314 static void ipip6_netlink_parms(struct nlattr *data[],
1315                                 struct ip_tunnel_parm *parms)
1316 {
1317         memset(parms, 0, sizeof(*parms));
1318
1319         parms->iph.version = 4;
1320         parms->iph.protocol = IPPROTO_IPV6;
1321         parms->iph.ihl = 5;
1322         parms->iph.ttl = 64;
1323
1324         if (!data)
1325                 return;
1326
1327         if (data[IFLA_IPTUN_LINK])
1328                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1329
1330         if (data[IFLA_IPTUN_LOCAL])
1331                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1332
1333         if (data[IFLA_IPTUN_REMOTE])
1334                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1335
1336         if (data[IFLA_IPTUN_TTL]) {
1337                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1338                 if (parms->iph.ttl)
1339                         parms->iph.frag_off = htons(IP_DF);
1340         }
1341
1342         if (data[IFLA_IPTUN_TOS])
1343                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1344
1345         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1346                 parms->iph.frag_off = htons(IP_DF);
1347
1348         if (data[IFLA_IPTUN_FLAGS])
1349                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1350
1351         if (data[IFLA_IPTUN_PROTO])
1352                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1353
1354 }
1355
1356 #ifdef CONFIG_IPV6_SIT_6RD
1357 /* This function returns true when 6RD attributes are present in the nl msg */
1358 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1359                                     struct ip_tunnel_6rd *ip6rd)
1360 {
1361         bool ret = false;
1362         memset(ip6rd, 0, sizeof(*ip6rd));
1363
1364         if (!data)
1365                 return ret;
1366
1367         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1368                 ret = true;
1369                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1370                            sizeof(struct in6_addr));
1371         }
1372
1373         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1374                 ret = true;
1375                 ip6rd->relay_prefix =
1376                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1377         }
1378
1379         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1380                 ret = true;
1381                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1382         }
1383
1384         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1385                 ret = true;
1386                 ip6rd->relay_prefixlen =
1387                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1388         }
1389
1390         return ret;
1391 }
1392 #endif
1393
1394 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1395                          struct nlattr *tb[], struct nlattr *data[])
1396 {
1397         struct net *net = dev_net(dev);
1398         struct ip_tunnel *nt;
1399 #ifdef CONFIG_IPV6_SIT_6RD
1400         struct ip_tunnel_6rd ip6rd;
1401 #endif
1402         int err;
1403
1404         nt = netdev_priv(dev);
1405         ipip6_netlink_parms(data, &nt->parms);
1406
1407         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1408                 return -EEXIST;
1409
1410         err = ipip6_tunnel_create(dev);
1411         if (err < 0)
1412                 return err;
1413
1414 #ifdef CONFIG_IPV6_SIT_6RD
1415         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1416                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1417 #endif
1418
1419         return err;
1420 }
1421
1422 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1423                           struct nlattr *data[])
1424 {
1425         struct ip_tunnel *t = netdev_priv(dev);
1426         struct ip_tunnel_parm p;
1427         struct net *net = t->net;
1428         struct sit_net *sitn = net_generic(net, sit_net_id);
1429 #ifdef CONFIG_IPV6_SIT_6RD
1430         struct ip_tunnel_6rd ip6rd;
1431 #endif
1432
1433         if (dev == sitn->fb_tunnel_dev)
1434                 return -EINVAL;
1435
1436         ipip6_netlink_parms(data, &p);
1437
1438         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1439             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1440                 return -EINVAL;
1441
1442         t = ipip6_tunnel_locate(net, &p, 0);
1443
1444         if (t) {
1445                 if (t->dev != dev)
1446                         return -EEXIST;
1447         } else
1448                 t = netdev_priv(dev);
1449
1450         ipip6_tunnel_update(t, &p);
1451
1452 #ifdef CONFIG_IPV6_SIT_6RD
1453         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1454                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1455 #endif
1456
1457         return 0;
1458 }
1459
1460 static size_t ipip6_get_size(const struct net_device *dev)
1461 {
1462         return
1463                 /* IFLA_IPTUN_LINK */
1464                 nla_total_size(4) +
1465                 /* IFLA_IPTUN_LOCAL */
1466                 nla_total_size(4) +
1467                 /* IFLA_IPTUN_REMOTE */
1468                 nla_total_size(4) +
1469                 /* IFLA_IPTUN_TTL */
1470                 nla_total_size(1) +
1471                 /* IFLA_IPTUN_TOS */
1472                 nla_total_size(1) +
1473                 /* IFLA_IPTUN_PMTUDISC */
1474                 nla_total_size(1) +
1475                 /* IFLA_IPTUN_FLAGS */
1476                 nla_total_size(2) +
1477                 /* IFLA_IPTUN_PROTO */
1478                 nla_total_size(1) +
1479 #ifdef CONFIG_IPV6_SIT_6RD
1480                 /* IFLA_IPTUN_6RD_PREFIX */
1481                 nla_total_size(sizeof(struct in6_addr)) +
1482                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1483                 nla_total_size(4) +
1484                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1485                 nla_total_size(2) +
1486                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1487                 nla_total_size(2) +
1488 #endif
1489                 0;
1490 }
1491
1492 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1493 {
1494         struct ip_tunnel *tunnel = netdev_priv(dev);
1495         struct ip_tunnel_parm *parm = &tunnel->parms;
1496
1497         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1498             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1499             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1500             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1501             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1502             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1503                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1504             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1505             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1506                 goto nla_put_failure;
1507
1508 #ifdef CONFIG_IPV6_SIT_6RD
1509         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1510                     &tunnel->ip6rd.prefix) ||
1511             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1512                          tunnel->ip6rd.relay_prefix) ||
1513             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1514                         tunnel->ip6rd.prefixlen) ||
1515             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1516                         tunnel->ip6rd.relay_prefixlen))
1517                 goto nla_put_failure;
1518 #endif
1519
1520         return 0;
1521
1522 nla_put_failure:
1523         return -EMSGSIZE;
1524 }
1525
1526 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1527         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1528         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1529         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1530         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1531         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1532         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1533         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1534         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1535 #ifdef CONFIG_IPV6_SIT_6RD
1536         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1537         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1538         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1539         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1540 #endif
1541 };
1542
1543 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1544         .kind           = "sit",
1545         .maxtype        = IFLA_IPTUN_MAX,
1546         .policy         = ipip6_policy,
1547         .priv_size      = sizeof(struct ip_tunnel),
1548         .setup          = ipip6_tunnel_setup,
1549         .validate       = ipip6_validate,
1550         .newlink        = ipip6_newlink,
1551         .changelink     = ipip6_changelink,
1552         .get_size       = ipip6_get_size,
1553         .fill_info      = ipip6_fill_info,
1554 };
1555
1556 static struct xfrm_tunnel sit_handler __read_mostly = {
1557         .handler        =       ipip6_rcv,
1558         .err_handler    =       ipip6_err,
1559         .priority       =       1,
1560 };
1561
1562 static struct xfrm_tunnel ipip_handler __read_mostly = {
1563         .handler        =       ipip_rcv,
1564         .err_handler    =       ipip6_err,
1565         .priority       =       2,
1566 };
1567
1568 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1569 {
1570         struct net *net = dev_net(sitn->fb_tunnel_dev);
1571         struct net_device *dev, *aux;
1572         int prio;
1573
1574         for_each_netdev_safe(net, dev, aux)
1575                 if (dev->rtnl_link_ops == &sit_link_ops)
1576                         unregister_netdevice_queue(dev, head);
1577
1578         for (prio = 1; prio < 4; prio++) {
1579                 int h;
1580                 for (h = 0; h < HASH_SIZE; h++) {
1581                         struct ip_tunnel *t;
1582
1583                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1584                         while (t != NULL) {
1585                                 /* If dev is in the same netns, it has already
1586                                  * been added to the list by the previous loop.
1587                                  */
1588                                 if (!net_eq(dev_net(t->dev), net))
1589                                         unregister_netdevice_queue(t->dev,
1590                                                                    head);
1591                                 t = rtnl_dereference(t->next);
1592                         }
1593                 }
1594         }
1595 }
1596
1597 static int __net_init sit_init_net(struct net *net)
1598 {
1599         struct sit_net *sitn = net_generic(net, sit_net_id);
1600         struct ip_tunnel *t;
1601         int err;
1602
1603         sitn->tunnels[0] = sitn->tunnels_wc;
1604         sitn->tunnels[1] = sitn->tunnels_l;
1605         sitn->tunnels[2] = sitn->tunnels_r;
1606         sitn->tunnels[3] = sitn->tunnels_r_l;
1607
1608         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1609                                            ipip6_tunnel_setup);
1610         if (!sitn->fb_tunnel_dev) {
1611                 err = -ENOMEM;
1612                 goto err_alloc_dev;
1613         }
1614         dev_net_set(sitn->fb_tunnel_dev, net);
1615         /* FB netdevice is special: we have one, and only one per netns.
1616          * Allowing to move it to another netns is clearly unsafe.
1617          */
1618         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1619
1620         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1621         if (err)
1622                 goto err_dev_free;
1623
1624         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1625
1626         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1627                 goto err_reg_dev;
1628
1629         t = netdev_priv(sitn->fb_tunnel_dev);
1630
1631         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1632         return 0;
1633
1634 err_reg_dev:
1635         dev_put(sitn->fb_tunnel_dev);
1636 err_dev_free:
1637         ipip6_dev_free(sitn->fb_tunnel_dev);
1638 err_alloc_dev:
1639         return err;
1640 }
1641
1642 static void __net_exit sit_exit_net(struct net *net)
1643 {
1644         struct sit_net *sitn = net_generic(net, sit_net_id);
1645         LIST_HEAD(list);
1646
1647         rtnl_lock();
1648         sit_destroy_tunnels(sitn, &list);
1649         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1650         unregister_netdevice_many(&list);
1651         rtnl_unlock();
1652 }
1653
1654 static struct pernet_operations sit_net_ops = {
1655         .init = sit_init_net,
1656         .exit = sit_exit_net,
1657         .id   = &sit_net_id,
1658         .size = sizeof(struct sit_net),
1659 };
1660
1661 static void __exit sit_cleanup(void)
1662 {
1663         rtnl_link_unregister(&sit_link_ops);
1664         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1665         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1666
1667         unregister_pernet_device(&sit_net_ops);
1668         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1669 }
1670
1671 static int __init sit_init(void)
1672 {
1673         int err;
1674
1675         pr_info("IPv6 over IPv4 tunneling driver\n");
1676
1677         err = register_pernet_device(&sit_net_ops);
1678         if (err < 0)
1679                 return err;
1680         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1681         if (err < 0) {
1682                 pr_info("%s: can't register ip6ip4\n", __func__);
1683                 goto xfrm_tunnel_failed;
1684         }
1685         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1686         if (err < 0) {
1687                 pr_info("%s: can't register ip4ip4\n", __func__);
1688                 goto xfrm_tunnel4_failed;
1689         }
1690         err = rtnl_link_register(&sit_link_ops);
1691         if (err < 0)
1692                 goto rtnl_link_failed;
1693
1694 out:
1695         return err;
1696
1697 rtnl_link_failed:
1698         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1699 xfrm_tunnel4_failed:
1700         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1701 xfrm_tunnel_failed:
1702         unregister_pernet_device(&sit_net_ops);
1703         goto out;
1704 }
1705
1706 module_init(sit_init);
1707 module_exit(sit_cleanup);
1708 MODULE_LICENSE("GPL");
1709 MODULE_ALIAS_NETDEV("sit0");