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                 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         iph = ip_hdr(skb);
 649         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 650                                      iph->saddr, iph->daddr);
 651         if (tunnel != NULL) {
 652                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 653                     tunnel->parms.iph.protocol != 0)
 654                         goto drop;
 655
 656                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 657                         goto drop;
 658                 if (iptunnel_pull_header(skb, 0, tpi.proto))
 659                         goto drop;
 660                 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 661         }
 662
 663         return 1;
 664
 665 drop:
 666         kfree_skb(skb);
 667         return 0;
 668 }
 669
 670 /*
 671  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 672  * stores the embedded IPv4 address in v4dst and returns true.
 673  */
 674 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 675                       __be32 *v4dst)
 676 {
 677 #ifdef CONFIG_IPV6_SIT_6RD
 678         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 679                               tunnel->ip6rd.prefixlen)) {
 680                 unsigned int pbw0, pbi0;
 681                 int pbi1;
 682                 u32 d;
 683
 684                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 685                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 686
 687                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 688                     tunnel->ip6rd.relay_prefixlen;
 689
 690                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 691                 if (pbi1 > 0)
 692                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 693                              (32 - pbi1);
 694
 695                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 696                 return true;
 697         }
 698 #else
 699         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 700                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 701                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 702                 return true;
 703         }
 704 #endif
 705         return false;
 706 }
 707
 708 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 709                              const struct in6_addr *v6dst)
 710 {
 711         __be32 dst = 0;
 712         check_6rd(tunnel, v6dst, &dst);
 713         return dst;
 714 }
 715
 716 /*
 717  *      This function assumes it is being called from dev_queue_xmit()
 718  *      and that skb is filled properly by that function.
 719  */
 720
 721 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 722                                      struct net_device *dev)
 723 {
 724         struct ip_tunnel *tunnel = netdev_priv(dev);
 725         const struct iphdr  *tiph = &tunnel->parms.iph;
 726         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 727         u8     tos = tunnel->parms.iph.tos;
 728         __be16 df = tiph->frag_off;
 729         struct rtable *rt;                      /* Route to the other host */
 730         struct net_device *tdev;                /* Device to other host */
 731         unsigned int max_headroom;              /* The extra header space needed */
 732         __be32 dst = tiph->daddr;
 733         struct flowi4 fl4;
 734         int    mtu;
 735         const struct in6_addr *addr6;
 736         int addr_type;
 737         u8 ttl;
 738         int err;
 739
 740         if (skb->protocol != htons(ETH_P_IPV6))
 741                 goto tx_error;
 742
 743         if (tos == 1)
 744                 tos = ipv6_get_dsfield(iph6);
 745
 746         /* ISATAP (RFC4214) - must come before 6to4 */
 747         if (dev->priv_flags & IFF_ISATAP) {
 748                 struct neighbour *neigh = NULL;
 749                 bool do_tx_error = false;
 750
 751                 if (skb_dst(skb))
 752                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 753
 754                 if (neigh == NULL) {
 755                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 756                         goto tx_error;
 757                 }
 758
 759                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 760                 addr_type = ipv6_addr_type(addr6);
 761
 762                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 763                      ipv6_addr_is_isatap(addr6))
 764                         dst = addr6->s6_addr32[3];
 765                 else
 766                         do_tx_error = true;
 767
 768                 neigh_release(neigh);
 769                 if (do_tx_error)
 770                         goto tx_error;
 771         }
 772
 773         if (!dst)
 774                 dst = try_6rd(tunnel, &iph6->daddr);
 775
 776         if (!dst) {
 777                 struct neighbour *neigh = NULL;
 778                 bool do_tx_error = false;
 779
 780                 if (skb_dst(skb))
 781                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 782
 783                 if (neigh == NULL) {
 784                         net_dbg_ratelimited("sit: nexthop == NULL\n");
 785                         goto tx_error;
 786                 }
 787
 788                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 789                 addr_type = ipv6_addr_type(addr6);
 790
 791                 if (addr_type == IPV6_ADDR_ANY) {
 792                         addr6 = &ipv6_hdr(skb)->daddr;
 793                         addr_type = ipv6_addr_type(addr6);
 794                 }
 795
 796                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 797                         dst = addr6->s6_addr32[3];
 798                 else
 799                         do_tx_error = true;
 800
 801                 neigh_release(neigh);
 802                 if (do_tx_error)
 803                         goto tx_error;
 804         }
 805
 806         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 807                                    dst, tiph->saddr,
 808                                    0, 0,
 809                                    IPPROTO_IPV6, RT_TOS(tos),
 810                                    tunnel->parms.link);
 811         if (IS_ERR(rt)) {
 812                 dev->stats.tx_carrier_errors++;
 813                 goto tx_error_icmp;
 814         }
 815         if (rt->rt_type != RTN_UNICAST) {
 816                 ip_rt_put(rt);
 817                 dev->stats.tx_carrier_errors++;
 818                 goto tx_error_icmp;
 819         }
 820         tdev = rt->dst.dev;
 821
 822         if (tdev == dev) {
 823                 ip_rt_put(rt);
 824                 dev->stats.collisions++;
 825                 goto tx_error;
 826         }
 827
 828         if (df) {
 829                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 830
 831                 if (mtu < 68) {
 832                         dev->stats.collisions++;
 833                         ip_rt_put(rt);
 834                         goto tx_error;
 835                 }
 836
 837                 if (mtu < IPV6_MIN_MTU) {
 838                         mtu = IPV6_MIN_MTU;
 839                         df = 0;
 840                 }
 841
 842                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 843                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 844
 845                 if (skb->len > mtu) {
 846                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 847                         ip_rt_put(rt);
 848                         goto tx_error;
 849                 }
 850         }
 851
 852         if (tunnel->err_count > 0) {
 853                 if (time_before(jiffies,
 854                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 855                         tunnel->err_count--;
 856                         dst_link_failure(skb);
 857                 } else
 858                         tunnel->err_count = 0;
 859         }
 860
 861         if (tunnel->net != dev_net(dev))
 862                 skb_scrub_packet(skb);
 863
 864         /*
 865          * Okay, now see if we can stuff it in the buffer as-is.
 866          */
 867         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 868
 869         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 870             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 871                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 872                 if (!new_skb) {
 873                         ip_rt_put(rt);
 874                         dev->stats.tx_dropped++;
 875                         dev_kfree_skb(skb);
 876                         return NETDEV_TX_OK;
 877                 }
 878                 if (skb->sk)
 879                         skb_set_owner_w(new_skb, skb->sk);
 880                 dev_kfree_skb(skb);
 881                 skb = new_skb;
 882                 iph6 = ipv6_hdr(skb);
 883         }
 884         ttl = tiph->ttl;
 885         if (ttl == 0)
 886                 ttl = iph6->hop_limit;
 887         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 888
 889         if (likely(!skb->encapsulation)) {
 890                 skb_reset_inner_headers(skb);
 891                 skb->encapsulation = 1;
 892         }
 893
 894         err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, fl4.daddr,
 895                             IPPROTO_IPV6, tos, ttl, df);
 896         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 897         return NETDEV_TX_OK;
 898
 899 tx_error_icmp:
 900         dst_link_failure(skb);
 901 tx_error:
 902         dev->stats.tx_errors++;
 903         dev_kfree_skb(skb);
 904         return NETDEV_TX_OK;
 905 }
 906
 907 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 908 {
 909         struct ip_tunnel *tunnel = netdev_priv(dev);
 910         const struct iphdr  *tiph = &tunnel->parms.iph;
 911
 912         if (likely(!skb->encapsulation)) {
 913                 skb_reset_inner_headers(skb);
 914                 skb->encapsulation = 1;
 915         }
 916
 917         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
 918         return NETDEV_TX_OK;
 919 }
 920
 921 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
 922                                    struct net_device *dev)
 923 {
 924         switch (skb->protocol) {
 925         case htons(ETH_P_IP):
 926                 ipip_tunnel_xmit(skb, dev);
 927                 break;
 928         case htons(ETH_P_IPV6):
 929                 ipip6_tunnel_xmit(skb, dev);
 930                 break;
 931         default:
 932                 goto tx_err;
 933         }
 934
 935         return NETDEV_TX_OK;
 936
 937 tx_err:
 938         dev->stats.tx_errors++;
 939         dev_kfree_skb(skb);
 940         return NETDEV_TX_OK;
 941
 942 }
 943
 944 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 945 {
 946         struct net_device *tdev = NULL;
 947         struct ip_tunnel *tunnel;
 948         const struct iphdr *iph;
 949         struct flowi4 fl4;
 950
 951         tunnel = netdev_priv(dev);
 952         iph = &tunnel->parms.iph;
 953
 954         if (iph->daddr) {
 955                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
 956                                                           NULL,
 957                                                           iph->daddr, iph->saddr,
 958                                                           0, 0,
 959                                                           IPPROTO_IPV6,
 960                                                           RT_TOS(iph->tos),
 961                                                           tunnel->parms.link);
 962
 963                 if (!IS_ERR(rt)) {
 964                         tdev = rt->dst.dev;
 965                         ip_rt_put(rt);
 966                 }
 967                 dev->flags |= IFF_POINTOPOINT;
 968         }
 969
 970         if (!tdev && tunnel->parms.link)
 971                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
 972
 973         if (tdev) {
 974                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 975                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 976                 if (dev->mtu < IPV6_MIN_MTU)
 977                         dev->mtu = IPV6_MIN_MTU;
 978         }
 979         dev->iflink = tunnel->parms.link;
 980 }
 981
 982 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
 983 {
 984         struct net *net = t->net;
 985         struct sit_net *sitn = net_generic(net, sit_net_id);
 986
 987         ipip6_tunnel_unlink(sitn, t);
 988         synchronize_net();
 989         t->parms.iph.saddr = p->iph.saddr;
 990         t->parms.iph.daddr = p->iph.daddr;
 991         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
 992         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
 993         ipip6_tunnel_link(sitn, t);
 994         t->parms.iph.ttl = p->iph.ttl;
 995         t->parms.iph.tos = p->iph.tos;
 996         if (t->parms.link != p->link) {
 997                 t->parms.link = p->link;
 998                 ipip6_tunnel_bind_dev(t->dev);
 999         }
1000         netdev_state_change(t->dev);
1001 }
1002
1003 #ifdef CONFIG_IPV6_SIT_6RD
1004 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1005                                    struct ip_tunnel_6rd *ip6rd)
1006 {
1007         struct in6_addr prefix;
1008         __be32 relay_prefix;
1009
1010         if (ip6rd->relay_prefixlen > 32 ||
1011             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1012                 return -EINVAL;
1013
1014         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1015         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1016                 return -EINVAL;
1017         if (ip6rd->relay_prefixlen)
1018                 relay_prefix = ip6rd->relay_prefix &
1019                                htonl(0xffffffffUL <<
1020                                      (32 - ip6rd->relay_prefixlen));
1021         else
1022                 relay_prefix = 0;
1023         if (relay_prefix != ip6rd->relay_prefix)
1024                 return -EINVAL;
1025
1026         t->ip6rd.prefix = prefix;
1027         t->ip6rd.relay_prefix = relay_prefix;
1028         t->ip6rd.prefixlen = ip6rd->prefixlen;
1029         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1030         netdev_state_change(t->dev);
1031         return 0;
1032 }
1033 #endif
1034
1035 static int
1036 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1037 {
1038         int err = 0;
1039         struct ip_tunnel_parm p;
1040         struct ip_tunnel_prl prl;
1041         struct ip_tunnel *t;
1042         struct net *net = dev_net(dev);
1043         struct sit_net *sitn = net_generic(net, sit_net_id);
1044 #ifdef CONFIG_IPV6_SIT_6RD
1045         struct ip_tunnel_6rd ip6rd;
1046 #endif
1047
1048         switch (cmd) {
1049         case SIOCGETTUNNEL:
1050 #ifdef CONFIG_IPV6_SIT_6RD
1051         case SIOCGET6RD:
1052 #endif
1053                 t = NULL;
1054                 if (dev == sitn->fb_tunnel_dev) {
1055                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1056                                 err = -EFAULT;
1057                                 break;
1058                         }
1059                         t = ipip6_tunnel_locate(net, &p, 0);
1060                 }
1061                 if (t == NULL)
1062                         t = netdev_priv(dev);
1063
1064                 err = -EFAULT;
1065                 if (cmd == SIOCGETTUNNEL) {
1066                         memcpy(&p, &t->parms, sizeof(p));
1067                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1068                                          sizeof(p)))
1069                                 goto done;
1070 #ifdef CONFIG_IPV6_SIT_6RD
1071                 } else {
1072                         ip6rd.prefix = t->ip6rd.prefix;
1073                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1074                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1075                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1076                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1077                                          sizeof(ip6rd)))
1078                                 goto done;
1079 #endif
1080                 }
1081                 err = 0;
1082                 break;
1083
1084         case SIOCADDTUNNEL:
1085         case SIOCCHGTUNNEL:
1086                 err = -EPERM;
1087                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1088                         goto done;
1089
1090                 err = -EFAULT;
1091                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1092                         goto done;
1093
1094                 err = -EINVAL;
1095                 if (p.iph.protocol != IPPROTO_IPV6 &&
1096                     p.iph.protocol != IPPROTO_IPIP &&
1097                     p.iph.protocol != 0)
1098                         goto done;
1099                 if (p.iph.version != 4 ||
1100                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1101                         goto done;
1102                 if (p.iph.ttl)
1103                         p.iph.frag_off |= htons(IP_DF);
1104
1105                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1106
1107                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1108                         if (t != NULL) {
1109                                 if (t->dev != dev) {
1110                                         err = -EEXIST;
1111                                         break;
1112                                 }
1113                         } else {
1114                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1115                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1116                                         err = -EINVAL;
1117                                         break;
1118                                 }
1119                                 t = netdev_priv(dev);
1120                         }
1121
1122                         ipip6_tunnel_update(t, &p);
1123                 }
1124
1125                 if (t) {
1126                         err = 0;
1127                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1128                                 err = -EFAULT;
1129                 } else
1130                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1131                 break;
1132
1133         case SIOCDELTUNNEL:
1134                 err = -EPERM;
1135                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1136                         goto done;
1137
1138                 if (dev == sitn->fb_tunnel_dev) {
1139                         err = -EFAULT;
1140                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1141                                 goto done;
1142                         err = -ENOENT;
1143                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1144                                 goto done;
1145                         err = -EPERM;
1146                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1147                                 goto done;
1148                         dev = t->dev;
1149                 }
1150                 unregister_netdevice(dev);
1151                 err = 0;
1152                 break;
1153
1154         case SIOCGETPRL:
1155                 err = -EINVAL;
1156                 if (dev == sitn->fb_tunnel_dev)
1157                         goto done;
1158                 err = -ENOENT;
1159                 if (!(t = netdev_priv(dev)))
1160                         goto done;
1161                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1162                 break;
1163
1164         case SIOCADDPRL:
1165         case SIOCDELPRL:
1166         case SIOCCHGPRL:
1167                 err = -EPERM;
1168                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1169                         goto done;
1170                 err = -EINVAL;
1171                 if (dev == sitn->fb_tunnel_dev)
1172                         goto done;
1173                 err = -EFAULT;
1174                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1175                         goto done;
1176                 err = -ENOENT;
1177                 if (!(t = netdev_priv(dev)))
1178                         goto done;
1179
1180                 switch (cmd) {
1181                 case SIOCDELPRL:
1182                         err = ipip6_tunnel_del_prl(t, &prl);
1183                         break;
1184                 case SIOCADDPRL:
1185                 case SIOCCHGPRL:
1186                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1187                         break;
1188                 }
1189                 netdev_state_change(dev);
1190                 break;
1191
1192 #ifdef CONFIG_IPV6_SIT_6RD
1193         case SIOCADD6RD:
1194         case SIOCCHG6RD:
1195         case SIOCDEL6RD:
1196                 err = -EPERM;
1197                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1198                         goto done;
1199
1200                 err = -EFAULT;
1201                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1202                                    sizeof(ip6rd)))
1203                         goto done;
1204
1205                 t = netdev_priv(dev);
1206
1207                 if (cmd != SIOCDEL6RD) {
1208                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1209                         if (err < 0)
1210                                 goto done;
1211                 } else
1212                         ipip6_tunnel_clone_6rd(dev, sitn);
1213
1214                 err = 0;
1215                 break;
1216 #endif
1217
1218         default:
1219                 err = -EINVAL;
1220         }
1221
1222 done:
1223         return err;
1224 }
1225
1226 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1227 {
1228         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1229                 return -EINVAL;
1230         dev->mtu = new_mtu;
1231         return 0;
1232 }
1233
1234 static const struct net_device_ops ipip6_netdev_ops = {
1235         .ndo_uninit     = ipip6_tunnel_uninit,
1236         .ndo_start_xmit = sit_tunnel_xmit,
1237         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1238         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1239         .ndo_get_stats64 = ip_tunnel_get_stats64,
1240 };
1241
1242 static void ipip6_dev_free(struct net_device *dev)
1243 {
1244         free_percpu(dev->tstats);
1245         free_netdev(dev);
1246 }
1247
1248 static void ipip6_tunnel_setup(struct net_device *dev)
1249 {
1250         dev->netdev_ops         = &ipip6_netdev_ops;
1251         dev->destructor         = ipip6_dev_free;
1252
1253         dev->type               = ARPHRD_SIT;
1254         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1255         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1256         dev->flags              = IFF_NOARP;
1257         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1258         dev->iflink             = 0;
1259         dev->addr_len           = 4;
1260         dev->features           |= NETIF_F_LLTX;
1261 }
1262
1263 static int ipip6_tunnel_init(struct net_device *dev)
1264 {
1265         struct ip_tunnel *tunnel = netdev_priv(dev);
1266
1267         tunnel->dev = dev;
1268         tunnel->net = dev_net(dev);
1269
1270         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1271         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1272
1273         ipip6_tunnel_bind_dev(dev);
1274         dev->tstats = alloc_percpu(struct pcpu_tstats);
1275         if (!dev->tstats)
1276                 return -ENOMEM;
1277
1278         return 0;
1279 }
1280
1281 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1282 {
1283         struct ip_tunnel *tunnel = netdev_priv(dev);
1284         struct iphdr *iph = &tunnel->parms.iph;
1285         struct net *net = dev_net(dev);
1286         struct sit_net *sitn = net_generic(net, sit_net_id);
1287
1288         tunnel->dev = dev;
1289         tunnel->net = dev_net(dev);
1290         strcpy(tunnel->parms.name, dev->name);
1291
1292         iph->version            = 4;
1293         iph->protocol           = IPPROTO_IPV6;
1294         iph->ihl                = 5;
1295         iph->ttl                = 64;
1296
1297         dev->tstats = alloc_percpu(struct pcpu_tstats);
1298         if (!dev->tstats)
1299                 return -ENOMEM;
1300         dev_hold(dev);
1301         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1302         return 0;
1303 }
1304
1305 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1306 {
1307         u8 proto;
1308
1309         if (!data || !data[IFLA_IPTUN_PROTO])
1310                 return 0;
1311
1312         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1313         if (proto != IPPROTO_IPV6 &&
1314             proto != IPPROTO_IPIP &&
1315             proto != 0)
1316                 return -EINVAL;
1317
1318         return 0;
1319 }
1320
1321 static void ipip6_netlink_parms(struct nlattr *data[],
1322                                 struct ip_tunnel_parm *parms)
1323 {
1324         memset(parms, 0, sizeof(*parms));
1325
1326         parms->iph.version = 4;
1327         parms->iph.protocol = IPPROTO_IPV6;
1328         parms->iph.ihl = 5;
1329         parms->iph.ttl = 64;
1330
1331         if (!data)
1332                 return;
1333
1334         if (data[IFLA_IPTUN_LINK])
1335                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1336
1337         if (data[IFLA_IPTUN_LOCAL])
1338                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1339
1340         if (data[IFLA_IPTUN_REMOTE])
1341                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1342
1343         if (data[IFLA_IPTUN_TTL]) {
1344                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1345                 if (parms->iph.ttl)
1346                         parms->iph.frag_off = htons(IP_DF);
1347         }
1348
1349         if (data[IFLA_IPTUN_TOS])
1350                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1351
1352         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1353                 parms->iph.frag_off = htons(IP_DF);
1354
1355         if (data[IFLA_IPTUN_FLAGS])
1356                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1357
1358         if (data[IFLA_IPTUN_PROTO])
1359                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1360
1361 }
1362
1363 #ifdef CONFIG_IPV6_SIT_6RD
1364 /* This function returns true when 6RD attributes are present in the nl msg */
1365 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1366                                     struct ip_tunnel_6rd *ip6rd)
1367 {
1368         bool ret = false;
1369         memset(ip6rd, 0, sizeof(*ip6rd));
1370
1371         if (!data)
1372                 return ret;
1373
1374         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1375                 ret = true;
1376                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1377                            sizeof(struct in6_addr));
1378         }
1379
1380         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1381                 ret = true;
1382                 ip6rd->relay_prefix =
1383                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1384         }
1385
1386         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1387                 ret = true;
1388                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1389         }
1390
1391         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1392                 ret = true;
1393                 ip6rd->relay_prefixlen =
1394                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1395         }
1396
1397         return ret;
1398 }
1399 #endif
1400
1401 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1402                          struct nlattr *tb[], struct nlattr *data[])
1403 {
1404         struct net *net = dev_net(dev);
1405         struct ip_tunnel *nt;
1406 #ifdef CONFIG_IPV6_SIT_6RD
1407         struct ip_tunnel_6rd ip6rd;
1408 #endif
1409         int err;
1410
1411         nt = netdev_priv(dev);
1412         ipip6_netlink_parms(data, &nt->parms);
1413
1414         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1415                 return -EEXIST;
1416
1417         err = ipip6_tunnel_create(dev);
1418         if (err < 0)
1419                 return err;
1420
1421 #ifdef CONFIG_IPV6_SIT_6RD
1422         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1423                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1424 #endif
1425
1426         return err;
1427 }
1428
1429 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1430                           struct nlattr *data[])
1431 {
1432         struct ip_tunnel *t = netdev_priv(dev);
1433         struct ip_tunnel_parm p;
1434         struct net *net = t->net;
1435         struct sit_net *sitn = net_generic(net, sit_net_id);
1436 #ifdef CONFIG_IPV6_SIT_6RD
1437         struct ip_tunnel_6rd ip6rd;
1438 #endif
1439
1440         if (dev == sitn->fb_tunnel_dev)
1441                 return -EINVAL;
1442
1443         ipip6_netlink_parms(data, &p);
1444
1445         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1446             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1447                 return -EINVAL;
1448
1449         t = ipip6_tunnel_locate(net, &p, 0);
1450
1451         if (t) {
1452                 if (t->dev != dev)
1453                         return -EEXIST;
1454         } else
1455                 t = netdev_priv(dev);
1456
1457         ipip6_tunnel_update(t, &p);
1458
1459 #ifdef CONFIG_IPV6_SIT_6RD
1460         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1461                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1462 #endif
1463
1464         return 0;
1465 }
1466
1467 static size_t ipip6_get_size(const struct net_device *dev)
1468 {
1469         return
1470                 /* IFLA_IPTUN_LINK */
1471                 nla_total_size(4) +
1472                 /* IFLA_IPTUN_LOCAL */
1473                 nla_total_size(4) +
1474                 /* IFLA_IPTUN_REMOTE */
1475                 nla_total_size(4) +
1476                 /* IFLA_IPTUN_TTL */
1477                 nla_total_size(1) +
1478                 /* IFLA_IPTUN_TOS */
1479                 nla_total_size(1) +
1480                 /* IFLA_IPTUN_PMTUDISC */
1481                 nla_total_size(1) +
1482                 /* IFLA_IPTUN_FLAGS */
1483                 nla_total_size(2) +
1484                 /* IFLA_IPTUN_PROTO */
1485                 nla_total_size(1) +
1486 #ifdef CONFIG_IPV6_SIT_6RD
1487                 /* IFLA_IPTUN_6RD_PREFIX */
1488                 nla_total_size(sizeof(struct in6_addr)) +
1489                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1490                 nla_total_size(4) +
1491                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1492                 nla_total_size(2) +
1493                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1494                 nla_total_size(2) +
1495 #endif
1496                 0;
1497 }
1498
1499 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1500 {
1501         struct ip_tunnel *tunnel = netdev_priv(dev);
1502         struct ip_tunnel_parm *parm = &tunnel->parms;
1503
1504         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1505             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1506             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1507             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1508             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1509             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1510                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1511             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1512             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1513                 goto nla_put_failure;
1514
1515 #ifdef CONFIG_IPV6_SIT_6RD
1516         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1517                     &tunnel->ip6rd.prefix) ||
1518             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1519                          tunnel->ip6rd.relay_prefix) ||
1520             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1521                         tunnel->ip6rd.prefixlen) ||
1522             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1523                         tunnel->ip6rd.relay_prefixlen))
1524                 goto nla_put_failure;
1525 #endif
1526
1527         return 0;
1528
1529 nla_put_failure:
1530         return -EMSGSIZE;
1531 }
1532
1533 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1534         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1535         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1536         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1537         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1538         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1539         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1540         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1541         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1542 #ifdef CONFIG_IPV6_SIT_6RD
1543         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1544         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1545         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1546         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1547 #endif
1548 };
1549
1550 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1551         .kind           = "sit",
1552         .maxtype        = IFLA_IPTUN_MAX,
1553         .policy         = ipip6_policy,
1554         .priv_size      = sizeof(struct ip_tunnel),
1555         .setup          = ipip6_tunnel_setup,
1556         .validate       = ipip6_validate,
1557         .newlink        = ipip6_newlink,
1558         .changelink     = ipip6_changelink,
1559         .get_size       = ipip6_get_size,
1560         .fill_info      = ipip6_fill_info,
1561 };
1562
1563 static struct xfrm_tunnel sit_handler __read_mostly = {
1564         .handler        =       ipip6_rcv,
1565         .err_handler    =       ipip6_err,
1566         .priority       =       1,
1567 };
1568
1569 static struct xfrm_tunnel ipip_handler __read_mostly = {
1570         .handler        =       ipip_rcv,
1571         .err_handler    =       ipip6_err,
1572         .priority       =       2,
1573 };
1574
1575 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1576 {
1577         struct net *net = dev_net(sitn->fb_tunnel_dev);
1578         struct net_device *dev, *aux;
1579         int prio;
1580
1581         for_each_netdev_safe(net, dev, aux)
1582                 if (dev->rtnl_link_ops == &sit_link_ops)
1583                         unregister_netdevice_queue(dev, head);
1584
1585         for (prio = 1; prio < 4; prio++) {
1586                 int h;
1587                 for (h = 0; h < HASH_SIZE; h++) {
1588                         struct ip_tunnel *t;
1589
1590                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1591                         while (t != NULL) {
1592                                 /* If dev is in the same netns, it has already
1593                                  * been added to the list by the previous loop.
1594                                  */
1595                                 if (dev_net(t->dev) != net)
1596                                         unregister_netdevice_queue(t->dev,
1597                                                                    head);
1598                                 t = rtnl_dereference(t->next);
1599                         }
1600                 }
1601         }
1602 }
1603
1604 static int __net_init sit_init_net(struct net *net)
1605 {
1606         struct sit_net *sitn = net_generic(net, sit_net_id);
1607         struct ip_tunnel *t;
1608         int err;
1609
1610         sitn->tunnels[0] = sitn->tunnels_wc;
1611         sitn->tunnels[1] = sitn->tunnels_l;
1612         sitn->tunnels[2] = sitn->tunnels_r;
1613         sitn->tunnels[3] = sitn->tunnels_r_l;
1614
1615         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1616                                            ipip6_tunnel_setup);
1617         if (!sitn->fb_tunnel_dev) {
1618                 err = -ENOMEM;
1619                 goto err_alloc_dev;
1620         }
1621         dev_net_set(sitn->fb_tunnel_dev, net);
1622         /* FB netdevice is special: we have one, and only one per netns.
1623          * Allowing to move it to another netns is clearly unsafe.
1624          */
1625         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1626
1627         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1628         if (err)
1629                 goto err_dev_free;
1630
1631         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1632
1633         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1634                 goto err_reg_dev;
1635
1636         t = netdev_priv(sitn->fb_tunnel_dev);
1637
1638         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1639         return 0;
1640
1641 err_reg_dev:
1642         dev_put(sitn->fb_tunnel_dev);
1643 err_dev_free:
1644         ipip6_dev_free(sitn->fb_tunnel_dev);
1645 err_alloc_dev:
1646         return err;
1647 }
1648
1649 static void __net_exit sit_exit_net(struct net *net)
1650 {
1651         struct sit_net *sitn = net_generic(net, sit_net_id);
1652         LIST_HEAD(list);
1653
1654         rtnl_lock();
1655         sit_destroy_tunnels(sitn, &list);
1656         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1657         unregister_netdevice_many(&list);
1658         rtnl_unlock();
1659 }
1660
1661 static struct pernet_operations sit_net_ops = {
1662         .init = sit_init_net,
1663         .exit = sit_exit_net,
1664         .id   = &sit_net_id,
1665         .size = sizeof(struct sit_net),
1666 };
1667
1668 static void __exit sit_cleanup(void)
1669 {
1670         rtnl_link_unregister(&sit_link_ops);
1671         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1672         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1673
1674         unregister_pernet_device(&sit_net_ops);
1675         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1676 }
1677
1678 static int __init sit_init(void)
1679 {
1680         int err;
1681
1682         pr_info("IPv6 over IPv4 tunneling driver\n");
1683
1684         err = register_pernet_device(&sit_net_ops);
1685         if (err < 0)
1686                 return err;
1687         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1688         if (err < 0) {
1689                 pr_info("%s: can't register ip6ip4\n", __func__);
1690                 goto xfrm_tunnel_failed;
1691         }
1692         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1693         if (err < 0) {
1694                 pr_info("%s: can't register ip4ip4\n", __func__);
1695                 goto xfrm_tunnel4_failed;
1696         }
1697         err = rtnl_link_register(&sit_link_ops);
1698         if (err < 0)
1699                 goto rtnl_link_failed;
1700
1701 out:
1702         return err;
1703
1704 rtnl_link_failed:
1705         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1706 xfrm_tunnel4_failed:
1707         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1708 xfrm_tunnel_failed:
1709         unregister_pernet_device(&sit_net_ops);
1710         goto out;
1711 }
1712
1713 module_init(sit_init);
1714 module_exit(sit_cleanup);
1715 MODULE_LICENSE("GPL");
1716 MODULE_ALIAS_NETDEV("sit0");