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                 secpath_reset(skb);
 585                 skb->mac_header = skb->network_header;
 586                 skb_reset_network_header(skb);
 587                 IPCB(skb)->flags = 0;
 588                 skb->protocol = htons(ETH_P_IPV6);
 589                 skb->pkt_type = PACKET_HOST;
 590
 591                 if (tunnel->dev->priv_flags & IFF_ISATAP) {
 592                         if (!isatap_chksrc(skb, iph, tunnel)) {
 593                                 tunnel->dev->stats.rx_errors++;
 594                                 goto out;
 595                         }
 596                 } else if (!(tunnel->dev->flags&IFF_POINTOPOINT)) {
 597                         if (is_spoofed_6rd(tunnel, iph->saddr,
 598                                            &ipv6_hdr(skb)->saddr) ||
 599                             is_spoofed_6rd(tunnel, iph->daddr,
 600                                            &ipv6_hdr(skb)->daddr)) {
 601                                 tunnel->dev->stats.rx_errors++;
 602                                 goto out;
 603                         }
 604                 }
 605
 606                 __skb_tunnel_rx(skb, tunnel->dev);
 607
 608                 err = IP_ECN_decapsulate(iph, skb);
 609                 if (unlikely(err)) {
 610                         if (log_ecn_error)
 611                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 612                                                      &iph->saddr, iph->tos);
 613                         if (err > 1) {
 614                                 ++tunnel->dev->stats.rx_frame_errors;
 615                                 ++tunnel->dev->stats.rx_errors;
 616                                 goto out;
 617                         }
 618                 }
 619
 620                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 621                 tstats->rx_packets++;
 622                 tstats->rx_bytes += skb->len;
 623
 624                 if (tunnel->net != dev_net(tunnel->dev))
 625                         skb_scrub_packet(skb);
 626                 netif_rx(skb);
 627
 628                 return 0;
 629         }
 630
 631         /* no tunnel matched,  let upstream know, ipsec may handle it */
 632         return 1;
 633 out:
 634         kfree_skb(skb);
 635         return 0;
 636 }
 637
 638 static const struct tnl_ptk_info tpi = {
 639         /* no tunnel info required for ipip. */
 640         .proto = htons(ETH_P_IP),
 641 };
 642
 643 static int ipip_rcv(struct sk_buff *skb)
 644 {
 645         const struct iphdr *iph;
 646         struct ip_tunnel *tunnel;
 647
 648         if (iptunnel_pull_header(skb, 0, tpi.proto))
 649                 goto drop;
 650
 651         iph = ip_hdr(skb);
 652
 653         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 654                                      iph->saddr, iph->daddr);
 655         if (tunnel != NULL) {
 656                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 657                     tunnel->parms.iph.protocol != 0)
 658                         goto drop;
 659
 660                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 661                         goto drop;
 662                 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 663         }
 664
 665         return 1;
 666
 667 drop:
 668         kfree_skb(skb);
 669         return 0;
 670 }
 671
 672 /*
 673  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 674  * stores the embedded IPv4 address in v4dst and returns true.
 675  */
 676 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 677                       __be32 *v4dst)
 678 {
 679 #ifdef CONFIG_IPV6_SIT_6RD
 680         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 681                               tunnel->ip6rd.prefixlen)) {
 682                 unsigned int pbw0, pbi0;
 683                 int pbi1;
 684                 u32 d;
 685
 686                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 687                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 688
 689                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 690                     tunnel->ip6rd.relay_prefixlen;
 691
 692                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 693                 if (pbi1 > 0)
 694                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 695                              (32 - pbi1);
 696
 697                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 698                 return true;
 699         }
 700 #else
 701         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 702                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 703                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 704                 return true;
 705         }
 706 #endif
 707         return false;
 708 }
 709
 710 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 711                              const struct in6_addr *v6dst)
 712 {
 713         __be32 dst = 0;
 714         check_6rd(tunnel, v6dst, &dst);
 715         return dst;
 716 }
 717
 718 /*
 719  *      This function assumes it is being called from dev_queue_xmit()
 720  *      and that skb is filled properly by that function.
 721  */
 722
 723 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 724                                      struct net_device *dev)
 725 {
 726         struct ip_tunnel *tunnel = netdev_priv(dev);
 727         const struct iphdr  *tiph = &tunnel->parms.iph;
 728         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 729         u8     tos = tunnel->parms.iph.tos;
 730         __be16 df = tiph->frag_off;
 731         struct rtable *rt;                      /* Route to the other host */
 732         struct net_device *tdev;                /* Device to other host */
 733         unsigned int max_headroom;              /* The extra header space needed */
 734         __be32 dst = tiph->daddr;
 735         struct flowi4 fl4;
 736         int    mtu;
 737         const struct in6_addr *addr6;
 738         int addr_type;
 739         u8 ttl;
 740         int err;
 741
 742         if (skb->protocol != htons(ETH_P_IPV6))
 743                 goto tx_error;
 744
 745         if (tos == 1)
 746                 tos = ipv6_get_dsfield(iph6);
 747
 748         /* ISATAP (RFC4214) - must come before 6to4 */
 749         if (dev->priv_flags & IFF_ISATAP) {
 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_UNICAST) &&
 765                      ipv6_addr_is_isatap(addr6))
 766                         dst = addr6->s6_addr32[3];
 767                 else
 768                         do_tx_error = true;
 769
 770                 neigh_release(neigh);
 771                 if (do_tx_error)
 772                         goto tx_error;
 773         }
 774
 775         if (!dst)
 776                 dst = try_6rd(tunnel, &iph6->daddr);
 777
 778         if (!dst) {
 779                 struct neighbour *neigh = NULL;
 780                 bool do_tx_error = false;
 781
 782                 if (skb_dst(skb))
 783                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 784
 785                 if (neigh == NULL) {
 786                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 787                         goto tx_error;
 788                 }
 789
 790                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 791                 addr_type = ipv6_addr_type(addr6);
 792
 793                 if (addr_type == IPV6_ADDR_ANY) {
 794                         addr6 = &ipv6_hdr(skb)->daddr;
 795                         addr_type = ipv6_addr_type(addr6);
 796                 }
 797
 798                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 799                         dst = addr6->s6_addr32[3];
 800                 else
 801                         do_tx_error = true;
 802
 803                 neigh_release(neigh);
 804                 if (do_tx_error)
 805                         goto tx_error;
 806         }
 807
 808         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 809                                    dst, tiph->saddr,
 810                                    0, 0,
 811                                    IPPROTO_IPV6, RT_TOS(tos),
 812                                    tunnel->parms.link);
 813         if (IS_ERR(rt)) {
 814                 dev->stats.tx_carrier_errors++;
 815                 goto tx_error_icmp;
 816         }
 817         if (rt->rt_type != RTN_UNICAST) {
 818                 ip_rt_put(rt);
 819                 dev->stats.tx_carrier_errors++;
 820                 goto tx_error_icmp;
 821         }
 822         tdev = rt->dst.dev;
 823
 824         if (tdev == dev) {
 825                 ip_rt_put(rt);
 826                 dev->stats.collisions++;
 827                 goto tx_error;
 828         }
 829
 830         if (df) {
 831                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 832
 833                 if (mtu < 68) {
 834                         dev->stats.collisions++;
 835                         ip_rt_put(rt);
 836                         goto tx_error;
 837                 }
 838
 839                 if (mtu < IPV6_MIN_MTU) {
 840                         mtu = IPV6_MIN_MTU;
 841                         df = 0;
 842                 }
 843
 844                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 845                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 846
 847                 if (skb->len > mtu) {
 848                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 849                         ip_rt_put(rt);
 850                         goto tx_error;
 851                 }
 852         }
 853
 854         if (tunnel->err_count > 0) {
 855                 if (time_before(jiffies,
 856                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 857                         tunnel->err_count--;
 858                         dst_link_failure(skb);
 859                 } else
 860                         tunnel->err_count = 0;
 861         }
 862
 863         if (tunnel->net != dev_net(dev))
 864                 skb_scrub_packet(skb);
 865
 866         /*
 867          * Okay, now see if we can stuff it in the buffer as-is.
 868          */
 869         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 870
 871         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 872             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 873                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 874                 if (!new_skb) {
 875                         ip_rt_put(rt);
 876                         dev->stats.tx_dropped++;
 877                         dev_kfree_skb(skb);
 878                         return NETDEV_TX_OK;
 879                 }
 880                 if (skb->sk)
 881                         skb_set_owner_w(new_skb, skb->sk);
 882                 dev_kfree_skb(skb);
 883                 skb = new_skb;
 884                 iph6 = ipv6_hdr(skb);
 885         }
 886         ttl = tiph->ttl;
 887         if (ttl == 0)
 888                 ttl = iph6->hop_limit;
 889         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 890
 891         err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, fl4.daddr,
 892                             IPPROTO_IPV6, tos, ttl, df);
 893         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 894         return NETDEV_TX_OK;
 895
 896 tx_error_icmp:
 897         dst_link_failure(skb);
 898 tx_error:
 899         dev->stats.tx_errors++;
 900         dev_kfree_skb(skb);
 901         return NETDEV_TX_OK;
 902 }
 903
 904 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 905 {
 906         struct ip_tunnel *tunnel = netdev_priv(dev);
 907         const struct iphdr  *tiph = &tunnel->parms.iph;
 908
 909         if (likely(!skb->encapsulation)) {
 910                 skb_reset_inner_headers(skb);
 911                 skb->encapsulation = 1;
 912         }
 913
 914         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
 915         return NETDEV_TX_OK;
 916 }
 917
 918 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
 919                                    struct net_device *dev)
 920 {
 921         switch (skb->protocol) {
 922         case htons(ETH_P_IP):
 923                 ipip_tunnel_xmit(skb, dev);
 924                 break;
 925         case htons(ETH_P_IPV6):
 926                 ipip6_tunnel_xmit(skb, dev);
 927                 break;
 928         default:
 929                 goto tx_err;
 930         }
 931
 932         return NETDEV_TX_OK;
 933
 934 tx_err:
 935         dev->stats.tx_errors++;
 936         dev_kfree_skb(skb);
 937         return NETDEV_TX_OK;
 938
 939 }
 940
 941 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 942 {
 943         struct net_device *tdev = NULL;
 944         struct ip_tunnel *tunnel;
 945         const struct iphdr *iph;
 946         struct flowi4 fl4;
 947
 948         tunnel = netdev_priv(dev);
 949         iph = &tunnel->parms.iph;
 950
 951         if (iph->daddr) {
 952                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
 953                                                           NULL,
 954                                                           iph->daddr, iph->saddr,
 955                                                           0, 0,
 956                                                           IPPROTO_IPV6,
 957                                                           RT_TOS(iph->tos),
 958                                                           tunnel->parms.link);
 959
 960                 if (!IS_ERR(rt)) {
 961                         tdev = rt->dst.dev;
 962                         ip_rt_put(rt);
 963                 }
 964                 dev->flags |= IFF_POINTOPOINT;
 965         }
 966
 967         if (!tdev && tunnel->parms.link)
 968                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
 969
 970         if (tdev) {
 971                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 972                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 973                 if (dev->mtu < IPV6_MIN_MTU)
 974                         dev->mtu = IPV6_MIN_MTU;
 975         }
 976         dev->iflink = tunnel->parms.link;
 977 }
 978
 979 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 980 {
 981         struct net *net = t->net;
 982         struct sit_net *sitn = net_generic(net, sit_net_id);
 983
 984         ipip6_tunnel_unlink(sitn, t);
 985         synchronize_net();
 986         t->parms.iph.saddr = p->iph.saddr;
 987         t->parms.iph.daddr = p->iph.daddr;
 988         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 989         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 990         ipip6_tunnel_link(sitn, t);
 991         t->parms.iph.ttl = p->iph.ttl;
 992         t->parms.iph.tos = p->iph.tos;
 993         if (t->parms.link != p->link) {
 994                 t->parms.link = p->link;
 995                 ipip6_tunnel_bind_dev(t->dev);
 996         }
 997         netdev_state_change(t->dev);
 998 }
 999
1000 #ifdef CONFIG_IPV6_SIT_6RD
1001 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1002                                    struct ip_tunnel_6rd *ip6rd)
1003 {
1004         struct in6_addr prefix;
1005         __be32 relay_prefix;
1006
1007         if (ip6rd->relay_prefixlen > 32 ||
1008             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1009                 return -EINVAL;
1010
1011         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1012         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1013                 return -EINVAL;
1014         if (ip6rd->relay_prefixlen)
1015                 relay_prefix = ip6rd->relay_prefix &
1016                                htonl(0xffffffffUL <<
1017                                      (32 - ip6rd->relay_prefixlen));
1018         else
1019                 relay_prefix = 0;
1020         if (relay_prefix != ip6rd->relay_prefix)
1021                 return -EINVAL;
1022
1023         t->ip6rd.prefix = prefix;
1024         t->ip6rd.relay_prefix = relay_prefix;
1025         t->ip6rd.prefixlen = ip6rd->prefixlen;
1026         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1027         netdev_state_change(t->dev);
1028         return 0;
1029 }
1030 #endif
1031
1032 static int
1033 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1034 {
1035         int err = 0;
1036         struct ip_tunnel_parm p;
1037         struct ip_tunnel_prl prl;
1038         struct ip_tunnel *t;
1039         struct net *net = dev_net(dev);
1040         struct sit_net *sitn = net_generic(net, sit_net_id);
1041 #ifdef CONFIG_IPV6_SIT_6RD
1042         struct ip_tunnel_6rd ip6rd;
1043 #endif
1044
1045         switch (cmd) {
1046         case SIOCGETTUNNEL:
1047 #ifdef CONFIG_IPV6_SIT_6RD
1048         case SIOCGET6RD:
1049 #endif
1050                 t = NULL;
1051                 if (dev == sitn->fb_tunnel_dev) {
1052                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1053                                 err = -EFAULT;
1054                                 break;
1055                         }
1056                         t = ipip6_tunnel_locate(net, &p, 0);
1057                 }
1058                 if (t == NULL)
1059                         t = netdev_priv(dev);
1060
1061                 err = -EFAULT;
1062                 if (cmd == SIOCGETTUNNEL) {
1063                         memcpy(&p, &t->parms, sizeof(p));
1064                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1065                                          sizeof(p)))
1066                                 goto done;
1067 #ifdef CONFIG_IPV6_SIT_6RD
1068                 } else {
1069                         ip6rd.prefix = t->ip6rd.prefix;
1070                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1071                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1072                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1073                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1074                                          sizeof(ip6rd)))
1075                                 goto done;
1076 #endif
1077                 }
1078                 err = 0;
1079                 break;
1080
1081         case SIOCADDTUNNEL:
1082         case SIOCCHGTUNNEL:
1083                 err = -EPERM;
1084                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1085                         goto done;
1086
1087                 err = -EFAULT;
1088                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1089                         goto done;
1090
1091                 err = -EINVAL;
1092                 if (p.iph.protocol != IPPROTO_IPV6 &&
1093                     p.iph.protocol != IPPROTO_IPIP &&
1094                     p.iph.protocol != 0)
1095                         goto done;
1096                 if (p.iph.version != 4 ||
1097                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1098                         goto done;
1099                 if (p.iph.ttl)
1100                         p.iph.frag_off |= htons(IP_DF);
1101
1102                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1103
1104                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1105                         if (t != NULL) {
1106                                 if (t->dev != dev) {
1107                                         err = -EEXIST;
1108                                         break;
1109                                 }
1110                         } else {
1111                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1112                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1113                                         err = -EINVAL;
1114                                         break;
1115                                 }
1116                                 t = netdev_priv(dev);
1117                         }
1118
1119                         ipip6_tunnel_update(t, &p);
1120                 }
1121
1122                 if (t) {
1123                         err = 0;
1124                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1125                                 err = -EFAULT;
1126                 } else
1127                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1128                 break;
1129
1130         case SIOCDELTUNNEL:
1131                 err = -EPERM;
1132                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1133                         goto done;
1134
1135                 if (dev == sitn->fb_tunnel_dev) {
1136                         err = -EFAULT;
1137                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1138                                 goto done;
1139                         err = -ENOENT;
1140                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1141                                 goto done;
1142                         err = -EPERM;
1143                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1144                                 goto done;
1145                         dev = t->dev;
1146                 }
1147                 unregister_netdevice(dev);
1148                 err = 0;
1149                 break;
1150
1151         case SIOCGETPRL:
1152                 err = -EINVAL;
1153                 if (dev == sitn->fb_tunnel_dev)
1154                         goto done;
1155                 err = -ENOENT;
1156                 if (!(t = netdev_priv(dev)))
1157                         goto done;
1158                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1159                 break;
1160
1161         case SIOCADDPRL:
1162         case SIOCDELPRL:
1163         case SIOCCHGPRL:
1164                 err = -EPERM;
1165                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1166                         goto done;
1167                 err = -EINVAL;
1168                 if (dev == sitn->fb_tunnel_dev)
1169                         goto done;
1170                 err = -EFAULT;
1171                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1172                         goto done;
1173                 err = -ENOENT;
1174                 if (!(t = netdev_priv(dev)))
1175                         goto done;
1176
1177                 switch (cmd) {
1178                 case SIOCDELPRL:
1179                         err = ipip6_tunnel_del_prl(t, &prl);
1180                         break;
1181                 case SIOCADDPRL:
1182                 case SIOCCHGPRL:
1183                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1184                         break;
1185                 }
1186                 netdev_state_change(dev);
1187                 break;
1188
1189 #ifdef CONFIG_IPV6_SIT_6RD
1190         case SIOCADD6RD:
1191         case SIOCCHG6RD:
1192         case SIOCDEL6RD:
1193                 err = -EPERM;
1194                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1195                         goto done;
1196
1197                 err = -EFAULT;
1198                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1199                                    sizeof(ip6rd)))
1200                         goto done;
1201
1202                 t = netdev_priv(dev);
1203
1204                 if (cmd != SIOCDEL6RD) {
1205                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1206                         if (err < 0)
1207                                 goto done;
1208                 } else
1209                         ipip6_tunnel_clone_6rd(dev, sitn);
1210
1211                 err = 0;
1212                 break;
1213 #endif
1214
1215         default:
1216                 err = -EINVAL;
1217         }
1218
1219 done:
1220         return err;
1221 }
1222
1223 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1224 {
1225         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1226                 return -EINVAL;
1227         dev->mtu = new_mtu;
1228         return 0;
1229 }
1230
1231 static const struct net_device_ops ipip6_netdev_ops = {
1232         .ndo_uninit     = ipip6_tunnel_uninit,
1233         .ndo_start_xmit = sit_tunnel_xmit,
1234         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1235         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1236         .ndo_get_stats64 = ip_tunnel_get_stats64,
1237 };
1238
1239 static void ipip6_dev_free(struct net_device *dev)
1240 {
1241         free_percpu(dev->tstats);
1242         free_netdev(dev);
1243 }
1244
1245 static void ipip6_tunnel_setup(struct net_device *dev)
1246 {
1247         dev->netdev_ops         = &ipip6_netdev_ops;
1248         dev->destructor         = ipip6_dev_free;
1249
1250         dev->type               = ARPHRD_SIT;
1251         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1252         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1253         dev->flags              = IFF_NOARP;
1254         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1255         dev->iflink             = 0;
1256         dev->addr_len           = 4;
1257         dev->features           |= NETIF_F_LLTX;
1258 }
1259
1260 static int ipip6_tunnel_init(struct net_device *dev)
1261 {
1262         struct ip_tunnel *tunnel = netdev_priv(dev);
1263
1264         tunnel->dev = dev;
1265         tunnel->net = dev_net(dev);
1266
1267         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1268         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1269
1270         ipip6_tunnel_bind_dev(dev);
1271         dev->tstats = alloc_percpu(struct pcpu_tstats);
1272         if (!dev->tstats)
1273                 return -ENOMEM;
1274
1275         return 0;
1276 }
1277
1278 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1279 {
1280         struct ip_tunnel *tunnel = netdev_priv(dev);
1281         struct iphdr *iph = &tunnel->parms.iph;
1282         struct net *net = dev_net(dev);
1283         struct sit_net *sitn = net_generic(net, sit_net_id);
1284
1285         tunnel->dev = dev;
1286         tunnel->net = dev_net(dev);
1287         strcpy(tunnel->parms.name, dev->name);
1288
1289         iph->version            = 4;
1290         iph->protocol           = IPPROTO_IPV6;
1291         iph->ihl                = 5;
1292         iph->ttl                = 64;
1293
1294         dev->tstats = alloc_percpu(struct pcpu_tstats);
1295         if (!dev->tstats)
1296                 return -ENOMEM;
1297         dev_hold(dev);
1298         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1299         return 0;
1300 }
1301
1302 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1303 {
1304         u8 proto;
1305
1306         if (!data || !data[IFLA_IPTUN_PROTO])
1307                 return 0;
1308
1309         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1310         if (proto != IPPROTO_IPV6 &&
1311             proto != IPPROTO_IPIP &&
1312             proto != 0)
1313                 return -EINVAL;
1314
1315         return 0;
1316 }
1317
1318 static void ipip6_netlink_parms(struct nlattr *data[],
1319                                 struct ip_tunnel_parm *parms)
1320 {
1321         memset(parms, 0, sizeof(*parms));
1322
1323         parms->iph.version = 4;
1324         parms->iph.protocol = IPPROTO_IPV6;
1325         parms->iph.ihl = 5;
1326         parms->iph.ttl = 64;
1327
1328         if (!data)
1329                 return;
1330
1331         if (data[IFLA_IPTUN_LINK])
1332                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1333
1334         if (data[IFLA_IPTUN_LOCAL])
1335                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1336
1337         if (data[IFLA_IPTUN_REMOTE])
1338                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1339
1340         if (data[IFLA_IPTUN_TTL]) {
1341                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1342                 if (parms->iph.ttl)
1343                         parms->iph.frag_off = htons(IP_DF);
1344         }
1345
1346         if (data[IFLA_IPTUN_TOS])
1347                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1348
1349         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1350                 parms->iph.frag_off = htons(IP_DF);
1351
1352         if (data[IFLA_IPTUN_FLAGS])
1353                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1354
1355         if (data[IFLA_IPTUN_PROTO])
1356                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1357
1358 }
1359
1360 #ifdef CONFIG_IPV6_SIT_6RD
1361 /* This function returns true when 6RD attributes are present in the nl msg */
1362 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1363                                     struct ip_tunnel_6rd *ip6rd)
1364 {
1365         bool ret = false;
1366         memset(ip6rd, 0, sizeof(*ip6rd));
1367
1368         if (!data)
1369                 return ret;
1370
1371         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1372                 ret = true;
1373                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1374                            sizeof(struct in6_addr));
1375         }
1376
1377         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1378                 ret = true;
1379                 ip6rd->relay_prefix =
1380                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1381         }
1382
1383         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1384                 ret = true;
1385                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1386         }
1387
1388         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1389                 ret = true;
1390                 ip6rd->relay_prefixlen =
1391                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1392         }
1393
1394         return ret;
1395 }
1396 #endif
1397
1398 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1399                          struct nlattr *tb[], struct nlattr *data[])
1400 {
1401         struct net *net = dev_net(dev);
1402         struct ip_tunnel *nt;
1403 #ifdef CONFIG_IPV6_SIT_6RD
1404         struct ip_tunnel_6rd ip6rd;
1405 #endif
1406         int err;
1407
1408         nt = netdev_priv(dev);
1409         ipip6_netlink_parms(data, &nt->parms);
1410
1411         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1412                 return -EEXIST;
1413
1414         err = ipip6_tunnel_create(dev);
1415         if (err < 0)
1416                 return err;
1417
1418 #ifdef CONFIG_IPV6_SIT_6RD
1419         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1420                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1421 #endif
1422
1423         return err;
1424 }
1425
1426 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1427                           struct nlattr *data[])
1428 {
1429         struct ip_tunnel *t = netdev_priv(dev);
1430         struct ip_tunnel_parm p;
1431         struct net *net = t->net;
1432         struct sit_net *sitn = net_generic(net, sit_net_id);
1433 #ifdef CONFIG_IPV6_SIT_6RD
1434         struct ip_tunnel_6rd ip6rd;
1435 #endif
1436
1437         if (dev == sitn->fb_tunnel_dev)
1438                 return -EINVAL;
1439
1440         ipip6_netlink_parms(data, &p);
1441
1442         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1443             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1444                 return -EINVAL;
1445
1446         t = ipip6_tunnel_locate(net, &p, 0);
1447
1448         if (t) {
1449                 if (t->dev != dev)
1450                         return -EEXIST;
1451         } else
1452                 t = netdev_priv(dev);
1453
1454         ipip6_tunnel_update(t, &p);
1455
1456 #ifdef CONFIG_IPV6_SIT_6RD
1457         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1458                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1459 #endif
1460
1461         return 0;
1462 }
1463
1464 static size_t ipip6_get_size(const struct net_device *dev)
1465 {
1466         return
1467                 /* IFLA_IPTUN_LINK */
1468                 nla_total_size(4) +
1469                 /* IFLA_IPTUN_LOCAL */
1470                 nla_total_size(4) +
1471                 /* IFLA_IPTUN_REMOTE */
1472                 nla_total_size(4) +
1473                 /* IFLA_IPTUN_TTL */
1474                 nla_total_size(1) +
1475                 /* IFLA_IPTUN_TOS */
1476                 nla_total_size(1) +
1477                 /* IFLA_IPTUN_PMTUDISC */
1478                 nla_total_size(1) +
1479                 /* IFLA_IPTUN_FLAGS */
1480                 nla_total_size(2) +
1481                 /* IFLA_IPTUN_PROTO */
1482                 nla_total_size(1) +
1483 #ifdef CONFIG_IPV6_SIT_6RD
1484                 /* IFLA_IPTUN_6RD_PREFIX */
1485                 nla_total_size(sizeof(struct in6_addr)) +
1486                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1487                 nla_total_size(4) +
1488                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1489                 nla_total_size(2) +
1490                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1491                 nla_total_size(2) +
1492 #endif
1493                 0;
1494 }
1495
1496 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1497 {
1498         struct ip_tunnel *tunnel = netdev_priv(dev);
1499         struct ip_tunnel_parm *parm = &tunnel->parms;
1500
1501         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1502             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1503             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1504             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1505             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1506             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1507                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1508             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1509             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1510                 goto nla_put_failure;
1511
1512 #ifdef CONFIG_IPV6_SIT_6RD
1513         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1514                     &tunnel->ip6rd.prefix) ||
1515             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1516                          tunnel->ip6rd.relay_prefix) ||
1517             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1518                         tunnel->ip6rd.prefixlen) ||
1519             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1520                         tunnel->ip6rd.relay_prefixlen))
1521                 goto nla_put_failure;
1522 #endif
1523
1524         return 0;
1525
1526 nla_put_failure:
1527         return -EMSGSIZE;
1528 }
1529
1530 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1531         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1532         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1533         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1534         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1535         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1536         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1537         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1538         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1539 #ifdef CONFIG_IPV6_SIT_6RD
1540         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1541         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1542         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1543         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1544 #endif
1545 };
1546
1547 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1548         .kind           = "sit",
1549         .maxtype        = IFLA_IPTUN_MAX,
1550         .policy         = ipip6_policy,
1551         .priv_size      = sizeof(struct ip_tunnel),
1552         .setup          = ipip6_tunnel_setup,
1553         .validate       = ipip6_validate,
1554         .newlink        = ipip6_newlink,
1555         .changelink     = ipip6_changelink,
1556         .get_size       = ipip6_get_size,
1557         .fill_info      = ipip6_fill_info,
1558 };
1559
1560 static struct xfrm_tunnel sit_handler __read_mostly = {
1561         .handler        =       ipip6_rcv,
1562         .err_handler    =       ipip6_err,
1563         .priority       =       1,
1564 };
1565
1566 static struct xfrm_tunnel ipip_handler __read_mostly = {
1567         .handler        =       ipip_rcv,
1568         .err_handler    =       ipip6_err,
1569         .priority       =       2,
1570 };
1571
1572 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1573 {
1574         struct net *net = dev_net(sitn->fb_tunnel_dev);
1575         struct net_device *dev, *aux;
1576         int prio;
1577
1578         for_each_netdev_safe(net, dev, aux)
1579                 if (dev->rtnl_link_ops == &sit_link_ops)
1580                         unregister_netdevice_queue(dev, head);
1581
1582         for (prio = 1; prio < 4; prio++) {
1583                 int h;
1584                 for (h = 0; h < HASH_SIZE; h++) {
1585                         struct ip_tunnel *t;
1586
1587                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1588                         while (t != NULL) {
1589                                 /* If dev is in the same netns, it has already
1590                                  * been added to the list by the previous loop.
1591                                  */
1592                                 if (dev_net(t->dev) != net)
1593                                         unregister_netdevice_queue(t->dev,
1594                                                                    head);
1595                                 t = rtnl_dereference(t->next);
1596                         }
1597                 }
1598         }
1599 }
1600
1601 static int __net_init sit_init_net(struct net *net)
1602 {
1603         struct sit_net *sitn = net_generic(net, sit_net_id);
1604         struct ip_tunnel *t;
1605         int err;
1606
1607         sitn->tunnels[0] = sitn->tunnels_wc;
1608         sitn->tunnels[1] = sitn->tunnels_l;
1609         sitn->tunnels[2] = sitn->tunnels_r;
1610         sitn->tunnels[3] = sitn->tunnels_r_l;
1611
1612         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1613                                            ipip6_tunnel_setup);
1614         if (!sitn->fb_tunnel_dev) {
1615                 err = -ENOMEM;
1616                 goto err_alloc_dev;
1617         }
1618         dev_net_set(sitn->fb_tunnel_dev, net);
1619         /* FB netdevice is special: we have one, and only one per netns.
1620          * Allowing to move it to another netns is clearly unsafe.
1621          */
1622         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1623
1624         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1625         if (err)
1626                 goto err_dev_free;
1627
1628         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1629
1630         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1631                 goto err_reg_dev;
1632
1633         t = netdev_priv(sitn->fb_tunnel_dev);
1634
1635         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1636         return 0;
1637
1638 err_reg_dev:
1639         dev_put(sitn->fb_tunnel_dev);
1640 err_dev_free:
1641         ipip6_dev_free(sitn->fb_tunnel_dev);
1642 err_alloc_dev:
1643         return err;
1644 }
1645
1646 static void __net_exit sit_exit_net(struct net *net)
1647 {
1648         struct sit_net *sitn = net_generic(net, sit_net_id);
1649         LIST_HEAD(list);
1650
1651         rtnl_lock();
1652         sit_destroy_tunnels(sitn, &list);
1653         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1654         unregister_netdevice_many(&list);
1655         rtnl_unlock();
1656 }
1657
1658 static struct pernet_operations sit_net_ops = {
1659         .init = sit_init_net,
1660         .exit = sit_exit_net,
1661         .id   = &sit_net_id,
1662         .size = sizeof(struct sit_net),
1663 };
1664
1665 static void __exit sit_cleanup(void)
1666 {
1667         rtnl_link_unregister(&sit_link_ops);
1668         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1669         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1670
1671         unregister_pernet_device(&sit_net_ops);
1672         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1673 }
1674
1675 static int __init sit_init(void)
1676 {
1677         int err;
1678
1679         pr_info("IPv6 over IPv4 tunneling driver\n");
1680
1681         err = register_pernet_device(&sit_net_ops);
1682         if (err < 0)
1683                 return err;
1684         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1685         if (err < 0) {
1686                 pr_info("%s: can't register ip6ip4\n", __func__);
1687                 goto xfrm_tunnel_failed;
1688         }
1689         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1690         if (err < 0) {
1691                 pr_info("%s: can't register ip4ip4\n", __func__);
1692                 goto xfrm_tunnel4_failed;
1693         }
1694         err = rtnl_link_register(&sit_link_ops);
1695         if (err < 0)
1696                 goto rtnl_link_failed;
1697
1698 out:
1699         return err;
1700
1701 rtnl_link_failed:
1702         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1703 xfrm_tunnel4_failed:
1704         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1705 xfrm_tunnel_failed:
1706         unregister_pernet_device(&sit_net_ops);
1707         goto out;
1708 }
1709
1710 module_init(sit_init);
1711 module_exit(sit_cleanup);
1712 MODULE_LICENSE("GPL");
1713 MODULE_ALIAS_NETDEV("sit0");