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 /* Checks if an address matches an address on the tunnel interface.
 570  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 571  * Long story:
 572  * This function is called after we considered the packet as spoofed
 573  * in is_spoofed_6rd.
 574  * We may have a router that is doing NAT for proto 41 packets
 575  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 576  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 577  * function will return true, dropping the packet.
 578  * But, we can still check if is spoofed against the IP
 579  * addresses associated with the interface.
 580  */
 581 static bool only_dnatted(const struct ip_tunnel *tunnel,
 582         const struct in6_addr *v6dst)
 583 {
 584         int prefix_len;
 585
 586 #ifdef CONFIG_IPV6_SIT_6RD
 587         prefix_len = tunnel->ip6rd.prefixlen + 32
 588                 - tunnel->ip6rd.relay_prefixlen;
 589 #else
 590         prefix_len = 48;
 591 #endif
 592         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 593 }
 594
 595 /* Returns true if a packet is spoofed */
 596 static bool packet_is_spoofed(struct sk_buff *skb,
 597                               const struct iphdr *iph,
 598                               struct ip_tunnel *tunnel)
 599 {
 600         const struct ipv6hdr *ipv6h;
 601
 602         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 603                 if (!isatap_chksrc(skb, iph, tunnel))
 604                         return true;
 605
 606                 return false;
 607         }
 608
 609         if (tunnel->dev->flags & IFF_POINTOPOINT)
 610                 return false;
 611
 612         ipv6h = ipv6_hdr(skb);
 613
 614         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 615                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 616                                      &iph->saddr, &ipv6h->saddr,
 617                                      &iph->daddr, &ipv6h->daddr);
 618                 return true;
 619         }
 620
 621         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 622                 return false;
 623
 624         if (only_dnatted(tunnel, &ipv6h->daddr))
 625                 return false;
 626
 627         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 628                              &iph->saddr, &ipv6h->saddr,
 629                              &iph->daddr, &ipv6h->daddr);
 630         return true;
 631 }
 632
 633 static int ipip6_rcv(struct sk_buff *skb)
 634 {
 635         const struct iphdr *iph = ip_hdr(skb);
 636         struct ip_tunnel *tunnel;
 637         int err;
 638
 639         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 640                                      iph->saddr, iph->daddr);
 641         if (tunnel != NULL) {
 642                 struct pcpu_tstats *tstats;
 643
 644                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 645                     tunnel->parms.iph.protocol != 0)
 646                         goto out;
 647
 648                 skb->mac_header = skb->network_header;
 649                 skb_reset_network_header(skb);
 650                 IPCB(skb)->flags = 0;
 651                 skb->protocol = htons(ETH_P_IPV6);
 652
 653                 if (packet_is_spoofed(skb, iph, tunnel)) {
 654                         tunnel->dev->stats.rx_errors++;
 655                         goto out;
 656                 }
 657
 658                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 659
 660                 err = IP_ECN_decapsulate(iph, skb);
 661                 if (unlikely(err)) {
 662                         if (log_ecn_error)
 663                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 664                                                      &iph->saddr, iph->tos);
 665                         if (err > 1) {
 666                                 ++tunnel->dev->stats.rx_frame_errors;
 667                                 ++tunnel->dev->stats.rx_errors;
 668                                 goto out;
 669                         }
 670                 }
 671
 672                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 673                 tstats->rx_packets++;
 674                 tstats->rx_bytes += skb->len;
 675
 676                 netif_rx(skb);
 677
 678                 return 0;
 679         }
 680
 681         /* no tunnel matched,  let upstream know, ipsec may handle it */
 682         return 1;
 683 out:
 684         kfree_skb(skb);
 685         return 0;
 686 }
 687
 688 static const struct tnl_ptk_info tpi = {
 689         /* no tunnel info required for ipip. */
 690         .proto = htons(ETH_P_IP),
 691 };
 692
 693 static int ipip_rcv(struct sk_buff *skb)
 694 {
 695         const struct iphdr *iph;
 696         struct ip_tunnel *tunnel;
 697
 698         iph = ip_hdr(skb);
 699         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 700                                      iph->saddr, iph->daddr);
 701         if (tunnel != NULL) {
 702                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 703                     tunnel->parms.iph.protocol != 0)
 704                         goto drop;
 705
 706                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 707                         goto drop;
 708                 if (iptunnel_pull_header(skb, 0, tpi.proto))
 709                         goto drop;
 710                 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 711         }
 712
 713         return 1;
 714
 715 drop:
 716         kfree_skb(skb);
 717         return 0;
 718 }
 719
 720 /*
 721  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 722  * stores the embedded IPv4 address in v4dst and returns true.
 723  */
 724 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 725                       __be32 *v4dst)
 726 {
 727 #ifdef CONFIG_IPV6_SIT_6RD
 728         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 729                               tunnel->ip6rd.prefixlen)) {
 730                 unsigned int pbw0, pbi0;
 731                 int pbi1;
 732                 u32 d;
 733
 734                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 735                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 736
 737                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 738                     tunnel->ip6rd.relay_prefixlen;
 739
 740                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 741                 if (pbi1 > 0)
 742                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 743                              (32 - pbi1);
 744
 745                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 746                 return true;
 747         }
 748 #else
 749         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 750                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 751                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 752                 return true;
 753         }
 754 #endif
 755         return false;
 756 }
 757
 758 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 759                              const struct in6_addr *v6dst)
 760 {
 761         __be32 dst = 0;
 762         check_6rd(tunnel, v6dst, &dst);
 763         return dst;
 764 }
 765
 766 /*
 767  *      This function assumes it is being called from dev_queue_xmit()
 768  *      and that skb is filled properly by that function.
 769  */
 770
 771 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 772                                      struct net_device *dev)
 773 {
 774         struct ip_tunnel *tunnel = netdev_priv(dev);
 775         const struct iphdr  *tiph = &tunnel->parms.iph;
 776         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 777         u8     tos = tunnel->parms.iph.tos;
 778         __be16 df = tiph->frag_off;
 779         struct rtable *rt;                      /* Route to the other host */
 780         struct net_device *tdev;                /* Device to other host */
 781         unsigned int max_headroom;              /* The extra header space needed */
 782         __be32 dst = tiph->daddr;
 783         struct flowi4 fl4;
 784         int    mtu;
 785         const struct in6_addr *addr6;
 786         int addr_type;
 787         u8 ttl;
 788         int err;
 789
 790         if (skb->protocol != htons(ETH_P_IPV6))
 791                 goto tx_error;
 792
 793         if (tos == 1)
 794                 tos = ipv6_get_dsfield(iph6);
 795
 796         /* ISATAP (RFC4214) - must come before 6to4 */
 797         if (dev->priv_flags & IFF_ISATAP) {
 798                 struct neighbour *neigh = NULL;
 799                 bool do_tx_error = false;
 800
 801                 if (skb_dst(skb))
 802                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 803
 804                 if (neigh == NULL) {
 805                         net_dbg_ratelimited("nexthop == NULL\n");
 806                         goto tx_error;
 807                 }
 808
 809                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 810                 addr_type = ipv6_addr_type(addr6);
 811
 812                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 813                      ipv6_addr_is_isatap(addr6))
 814                         dst = addr6->s6_addr32[3];
 815                 else
 816                         do_tx_error = true;
 817
 818                 neigh_release(neigh);
 819                 if (do_tx_error)
 820                         goto tx_error;
 821         }
 822
 823         if (!dst)
 824                 dst = try_6rd(tunnel, &iph6->daddr);
 825
 826         if (!dst) {
 827                 struct neighbour *neigh = NULL;
 828                 bool do_tx_error = false;
 829
 830                 if (skb_dst(skb))
 831                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 832
 833                 if (neigh == NULL) {
 834                         net_dbg_ratelimited("nexthop == NULL\n");
 835                         goto tx_error;
 836                 }
 837
 838                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 839                 addr_type = ipv6_addr_type(addr6);
 840
 841                 if (addr_type == IPV6_ADDR_ANY) {
 842                         addr6 = &ipv6_hdr(skb)->daddr;
 843                         addr_type = ipv6_addr_type(addr6);
 844                 }
 845
 846                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 847                         dst = addr6->s6_addr32[3];
 848                 else
 849                         do_tx_error = true;
 850
 851                 neigh_release(neigh);
 852                 if (do_tx_error)
 853                         goto tx_error;
 854         }
 855
 856         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 857                                    dst, tiph->saddr,
 858                                    0, 0,
 859                                    IPPROTO_IPV6, RT_TOS(tos),
 860                                    tunnel->parms.link);
 861         if (IS_ERR(rt)) {
 862                 dev->stats.tx_carrier_errors++;
 863                 goto tx_error_icmp;
 864         }
 865         if (rt->rt_type != RTN_UNICAST) {
 866                 ip_rt_put(rt);
 867                 dev->stats.tx_carrier_errors++;
 868                 goto tx_error_icmp;
 869         }
 870         tdev = rt->dst.dev;
 871
 872         if (tdev == dev) {
 873                 ip_rt_put(rt);
 874                 dev->stats.collisions++;
 875                 goto tx_error;
 876         }
 877
 878         if (df) {
 879                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 880
 881                 if (mtu < 68) {
 882                         dev->stats.collisions++;
 883                         ip_rt_put(rt);
 884                         goto tx_error;
 885                 }
 886
 887                 if (mtu < IPV6_MIN_MTU) {
 888                         mtu = IPV6_MIN_MTU;
 889                         df = 0;
 890                 }
 891
 892                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 893                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 894
 895                 if (skb->len > mtu) {
 896                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 897                         ip_rt_put(rt);
 898                         goto tx_error;
 899                 }
 900         }
 901
 902         if (tunnel->err_count > 0) {
 903                 if (time_before(jiffies,
 904                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 905                         tunnel->err_count--;
 906                         dst_link_failure(skb);
 907                 } else
 908                         tunnel->err_count = 0;
 909         }
 910
 911         /*
 912          * Okay, now see if we can stuff it in the buffer as-is.
 913          */
 914         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 915
 916         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 917             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 918                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 919                 if (!new_skb) {
 920                         ip_rt_put(rt);
 921                         dev->stats.tx_dropped++;
 922                         dev_kfree_skb(skb);
 923                         return NETDEV_TX_OK;
 924                 }
 925                 if (skb->sk)
 926                         skb_set_owner_w(new_skb, skb->sk);
 927                 dev_kfree_skb(skb);
 928                 skb = new_skb;
 929                 iph6 = ipv6_hdr(skb);
 930         }
 931         ttl = tiph->ttl;
 932         if (ttl == 0)
 933                 ttl = iph6->hop_limit;
 934         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 935
 936         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
 937         if (IS_ERR(skb))
 938                 goto out;
 939
 940         err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, IPPROTO_IPV6, tos,
 941                             ttl, df, !net_eq(tunnel->net, dev_net(dev)));
 942         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 943         return NETDEV_TX_OK;
 944
 945 tx_error_icmp:
 946         dst_link_failure(skb);
 947 tx_error:
 948         dev_kfree_skb(skb);
 949 out:
 950         dev->stats.tx_errors++;
 951         return NETDEV_TX_OK;
 952 }
 953
 954 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 955 {
 956         struct ip_tunnel *tunnel = netdev_priv(dev);
 957         const struct iphdr  *tiph = &tunnel->parms.iph;
 958
 959         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
 960         if (IS_ERR(skb))
 961                 goto out;
 962
 963         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
 964         return NETDEV_TX_OK;
 965 out:
 966         dev->stats.tx_errors++;
 967         return NETDEV_TX_OK;
 968 }
 969
 970 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
 971                                    struct net_device *dev)
 972 {
 973         switch (skb->protocol) {
 974         case htons(ETH_P_IP):
 975                 ipip_tunnel_xmit(skb, dev);
 976                 break;
 977         case htons(ETH_P_IPV6):
 978                 ipip6_tunnel_xmit(skb, dev);
 979                 break;
 980         default:
 981                 goto tx_err;
 982         }
 983
 984         return NETDEV_TX_OK;
 985
 986 tx_err:
 987         dev->stats.tx_errors++;
 988         dev_kfree_skb(skb);
 989         return NETDEV_TX_OK;
 990
 991 }
 992
 993 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 994 {
 995         struct net_device *tdev = NULL;
 996         struct ip_tunnel *tunnel;
 997         const struct iphdr *iph;
 998         struct flowi4 fl4;
 999
1000         tunnel = netdev_priv(dev);
1001         iph = &tunnel->parms.iph;
1002
1003         if (iph->daddr) {
1004                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1005                                                           NULL,
1006                                                           iph->daddr, iph->saddr,
1007                                                           0, 0,
1008                                                           IPPROTO_IPV6,
1009                                                           RT_TOS(iph->tos),
1010                                                           tunnel->parms.link);
1011
1012                 if (!IS_ERR(rt)) {
1013                         tdev = rt->dst.dev;
1014                         ip_rt_put(rt);
1015                 }
1016                 dev->flags |= IFF_POINTOPOINT;
1017         }
1018
1019         if (!tdev && tunnel->parms.link)
1020                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1021
1022         if (tdev) {
1023                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1024                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
1025                 if (dev->mtu < IPV6_MIN_MTU)
1026                         dev->mtu = IPV6_MIN_MTU;
1027         }
1028         dev->iflink = tunnel->parms.link;
1029 }
1030
1031 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1032 {
1033         struct net *net = t->net;
1034         struct sit_net *sitn = net_generic(net, sit_net_id);
1035
1036         ipip6_tunnel_unlink(sitn, t);
1037         synchronize_net();
1038         t->parms.iph.saddr = p->iph.saddr;
1039         t->parms.iph.daddr = p->iph.daddr;
1040         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1041         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1042         ipip6_tunnel_link(sitn, t);
1043         t->parms.iph.ttl = p->iph.ttl;
1044         t->parms.iph.tos = p->iph.tos;
1045         if (t->parms.link != p->link) {
1046                 t->parms.link = p->link;
1047                 ipip6_tunnel_bind_dev(t->dev);
1048         }
1049         netdev_state_change(t->dev);
1050 }
1051
1052 #ifdef CONFIG_IPV6_SIT_6RD
1053 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1054                                    struct ip_tunnel_6rd *ip6rd)
1055 {
1056         struct in6_addr prefix;
1057         __be32 relay_prefix;
1058
1059         if (ip6rd->relay_prefixlen > 32 ||
1060             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1061                 return -EINVAL;
1062
1063         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1064         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1065                 return -EINVAL;
1066         if (ip6rd->relay_prefixlen)
1067                 relay_prefix = ip6rd->relay_prefix &
1068                                htonl(0xffffffffUL <<
1069                                      (32 - ip6rd->relay_prefixlen));
1070         else
1071                 relay_prefix = 0;
1072         if (relay_prefix != ip6rd->relay_prefix)
1073                 return -EINVAL;
1074
1075         t->ip6rd.prefix = prefix;
1076         t->ip6rd.relay_prefix = relay_prefix;
1077         t->ip6rd.prefixlen = ip6rd->prefixlen;
1078         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1079         netdev_state_change(t->dev);
1080         return 0;
1081 }
1082 #endif
1083
1084 static int
1085 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1086 {
1087         int err = 0;
1088         struct ip_tunnel_parm p;
1089         struct ip_tunnel_prl prl;
1090         struct ip_tunnel *t;
1091         struct net *net = dev_net(dev);
1092         struct sit_net *sitn = net_generic(net, sit_net_id);
1093 #ifdef CONFIG_IPV6_SIT_6RD
1094         struct ip_tunnel_6rd ip6rd;
1095 #endif
1096
1097         switch (cmd) {
1098         case SIOCGETTUNNEL:
1099 #ifdef CONFIG_IPV6_SIT_6RD
1100         case SIOCGET6RD:
1101 #endif
1102                 t = NULL;
1103                 if (dev == sitn->fb_tunnel_dev) {
1104                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1105                                 err = -EFAULT;
1106                                 break;
1107                         }
1108                         t = ipip6_tunnel_locate(net, &p, 0);
1109                 }
1110                 if (t == NULL)
1111                         t = netdev_priv(dev);
1112
1113                 err = -EFAULT;
1114                 if (cmd == SIOCGETTUNNEL) {
1115                         memcpy(&p, &t->parms, sizeof(p));
1116                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1117                                          sizeof(p)))
1118                                 goto done;
1119 #ifdef CONFIG_IPV6_SIT_6RD
1120                 } else {
1121                         ip6rd.prefix = t->ip6rd.prefix;
1122                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1123                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1124                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1125                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1126                                          sizeof(ip6rd)))
1127                                 goto done;
1128 #endif
1129                 }
1130                 err = 0;
1131                 break;
1132
1133         case SIOCADDTUNNEL:
1134         case SIOCCHGTUNNEL:
1135                 err = -EPERM;
1136                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1137                         goto done;
1138
1139                 err = -EFAULT;
1140                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1141                         goto done;
1142
1143                 err = -EINVAL;
1144                 if (p.iph.protocol != IPPROTO_IPV6 &&
1145                     p.iph.protocol != IPPROTO_IPIP &&
1146                     p.iph.protocol != 0)
1147                         goto done;
1148                 if (p.iph.version != 4 ||
1149                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1150                         goto done;
1151                 if (p.iph.ttl)
1152                         p.iph.frag_off |= htons(IP_DF);
1153
1154                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1155
1156                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1157                         if (t != NULL) {
1158                                 if (t->dev != dev) {
1159                                         err = -EEXIST;
1160                                         break;
1161                                 }
1162                         } else {
1163                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1164                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1165                                         err = -EINVAL;
1166                                         break;
1167                                 }
1168                                 t = netdev_priv(dev);
1169                         }
1170
1171                         ipip6_tunnel_update(t, &p);
1172                 }
1173
1174                 if (t) {
1175                         err = 0;
1176                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1177                                 err = -EFAULT;
1178                 } else
1179                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1180                 break;
1181
1182         case SIOCDELTUNNEL:
1183                 err = -EPERM;
1184                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1185                         goto done;
1186
1187                 if (dev == sitn->fb_tunnel_dev) {
1188                         err = -EFAULT;
1189                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1190                                 goto done;
1191                         err = -ENOENT;
1192                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1193                                 goto done;
1194                         err = -EPERM;
1195                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1196                                 goto done;
1197                         dev = t->dev;
1198                 }
1199                 unregister_netdevice(dev);
1200                 err = 0;
1201                 break;
1202
1203         case SIOCGETPRL:
1204                 err = -EINVAL;
1205                 if (dev == sitn->fb_tunnel_dev)
1206                         goto done;
1207                 err = -ENOENT;
1208                 if (!(t = netdev_priv(dev)))
1209                         goto done;
1210                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1211                 break;
1212
1213         case SIOCADDPRL:
1214         case SIOCDELPRL:
1215         case SIOCCHGPRL:
1216                 err = -EPERM;
1217                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1218                         goto done;
1219                 err = -EINVAL;
1220                 if (dev == sitn->fb_tunnel_dev)
1221                         goto done;
1222                 err = -EFAULT;
1223                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1224                         goto done;
1225                 err = -ENOENT;
1226                 if (!(t = netdev_priv(dev)))
1227                         goto done;
1228
1229                 switch (cmd) {
1230                 case SIOCDELPRL:
1231                         err = ipip6_tunnel_del_prl(t, &prl);
1232                         break;
1233                 case SIOCADDPRL:
1234                 case SIOCCHGPRL:
1235                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1236                         break;
1237                 }
1238                 netdev_state_change(dev);
1239                 break;
1240
1241 #ifdef CONFIG_IPV6_SIT_6RD
1242         case SIOCADD6RD:
1243         case SIOCCHG6RD:
1244         case SIOCDEL6RD:
1245                 err = -EPERM;
1246                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1247                         goto done;
1248
1249                 err = -EFAULT;
1250                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1251                                    sizeof(ip6rd)))
1252                         goto done;
1253
1254                 t = netdev_priv(dev);
1255
1256                 if (cmd != SIOCDEL6RD) {
1257                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1258                         if (err < 0)
1259                                 goto done;
1260                 } else
1261                         ipip6_tunnel_clone_6rd(dev, sitn);
1262
1263                 err = 0;
1264                 break;
1265 #endif
1266
1267         default:
1268                 err = -EINVAL;
1269         }
1270
1271 done:
1272         return err;
1273 }
1274
1275 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1276 {
1277         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1278                 return -EINVAL;
1279         dev->mtu = new_mtu;
1280         return 0;
1281 }
1282
1283 static const struct net_device_ops ipip6_netdev_ops = {
1284         .ndo_uninit     = ipip6_tunnel_uninit,
1285         .ndo_start_xmit = sit_tunnel_xmit,
1286         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1287         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1288         .ndo_get_stats64 = ip_tunnel_get_stats64,
1289 };
1290
1291 static void ipip6_dev_free(struct net_device *dev)
1292 {
1293         free_percpu(dev->tstats);
1294         free_netdev(dev);
1295 }
1296
1297 #define SIT_FEATURES (NETIF_F_SG           | \
1298                       NETIF_F_FRAGLIST     | \
1299                       NETIF_F_HIGHDMA      | \
1300                       NETIF_F_GSO_SOFTWARE | \
1301                       NETIF_F_HW_CSUM)
1302
1303 static void ipip6_tunnel_setup(struct net_device *dev)
1304 {
1305         dev->netdev_ops         = &ipip6_netdev_ops;
1306         dev->destructor         = ipip6_dev_free;
1307
1308         dev->type               = ARPHRD_SIT;
1309         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1310         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1311         dev->flags              = IFF_NOARP;
1312         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1313         dev->iflink             = 0;
1314         dev->addr_len           = 4;
1315         dev->features           |= NETIF_F_LLTX;
1316         dev->features           |= SIT_FEATURES;
1317         dev->hw_features        |= SIT_FEATURES;
1318 }
1319
1320 static int ipip6_tunnel_init(struct net_device *dev)
1321 {
1322         struct ip_tunnel *tunnel = netdev_priv(dev);
1323         int i;
1324
1325         tunnel->dev = dev;
1326         tunnel->net = dev_net(dev);
1327
1328         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1329         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1330
1331         ipip6_tunnel_bind_dev(dev);
1332         dev->tstats = alloc_percpu(struct pcpu_tstats);
1333         if (!dev->tstats)
1334                 return -ENOMEM;
1335
1336         for_each_possible_cpu(i) {
1337                 struct pcpu_tstats *ipip6_tunnel_stats;
1338                 ipip6_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1339                 u64_stats_init(&ipip6_tunnel_stats->syncp);
1340         }
1341
1342         return 0;
1343 }
1344
1345 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1346 {
1347         struct ip_tunnel *tunnel = netdev_priv(dev);
1348         struct iphdr *iph = &tunnel->parms.iph;
1349         struct net *net = dev_net(dev);
1350         struct sit_net *sitn = net_generic(net, sit_net_id);
1351         int i;
1352
1353         tunnel->dev = dev;
1354         tunnel->net = dev_net(dev);
1355         strcpy(tunnel->parms.name, dev->name);
1356
1357         iph->version            = 4;
1358         iph->protocol           = IPPROTO_IPV6;
1359         iph->ihl                = 5;
1360         iph->ttl                = 64;
1361
1362         dev->tstats = alloc_percpu(struct pcpu_tstats);
1363         if (!dev->tstats)
1364                 return -ENOMEM;
1365
1366         for_each_possible_cpu(i) {
1367                 struct pcpu_tstats *ipip6_fb_stats;
1368                 ipip6_fb_stats = per_cpu_ptr(dev->tstats, i);
1369                 u64_stats_init(&ipip6_fb_stats->syncp);
1370         }
1371
1372         dev_hold(dev);
1373         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1374         return 0;
1375 }
1376
1377 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1378 {
1379         u8 proto;
1380
1381         if (!data || !data[IFLA_IPTUN_PROTO])
1382                 return 0;
1383
1384         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1385         if (proto != IPPROTO_IPV6 &&
1386             proto != IPPROTO_IPIP &&
1387             proto != 0)
1388                 return -EINVAL;
1389
1390         return 0;
1391 }
1392
1393 static void ipip6_netlink_parms(struct nlattr *data[],
1394                                 struct ip_tunnel_parm *parms)
1395 {
1396         memset(parms, 0, sizeof(*parms));
1397
1398         parms->iph.version = 4;
1399         parms->iph.protocol = IPPROTO_IPV6;
1400         parms->iph.ihl = 5;
1401         parms->iph.ttl = 64;
1402
1403         if (!data)
1404                 return;
1405
1406         if (data[IFLA_IPTUN_LINK])
1407                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1408
1409         if (data[IFLA_IPTUN_LOCAL])
1410                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1411
1412         if (data[IFLA_IPTUN_REMOTE])
1413                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1414
1415         if (data[IFLA_IPTUN_TTL]) {
1416                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1417                 if (parms->iph.ttl)
1418                         parms->iph.frag_off = htons(IP_DF);
1419         }
1420
1421         if (data[IFLA_IPTUN_TOS])
1422                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1423
1424         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1425                 parms->iph.frag_off = htons(IP_DF);
1426
1427         if (data[IFLA_IPTUN_FLAGS])
1428                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1429
1430         if (data[IFLA_IPTUN_PROTO])
1431                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1432
1433 }
1434
1435 #ifdef CONFIG_IPV6_SIT_6RD
1436 /* This function returns true when 6RD attributes are present in the nl msg */
1437 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1438                                     struct ip_tunnel_6rd *ip6rd)
1439 {
1440         bool ret = false;
1441         memset(ip6rd, 0, sizeof(*ip6rd));
1442
1443         if (!data)
1444                 return ret;
1445
1446         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1447                 ret = true;
1448                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1449                            sizeof(struct in6_addr));
1450         }
1451
1452         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1453                 ret = true;
1454                 ip6rd->relay_prefix =
1455                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1456         }
1457
1458         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1459                 ret = true;
1460                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1461         }
1462
1463         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1464                 ret = true;
1465                 ip6rd->relay_prefixlen =
1466                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1467         }
1468
1469         return ret;
1470 }
1471 #endif
1472
1473 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1474                          struct nlattr *tb[], struct nlattr *data[])
1475 {
1476         struct net *net = dev_net(dev);
1477         struct ip_tunnel *nt;
1478 #ifdef CONFIG_IPV6_SIT_6RD
1479         struct ip_tunnel_6rd ip6rd;
1480 #endif
1481         int err;
1482
1483         nt = netdev_priv(dev);
1484         ipip6_netlink_parms(data, &nt->parms);
1485
1486         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1487                 return -EEXIST;
1488
1489         err = ipip6_tunnel_create(dev);
1490         if (err < 0)
1491                 return err;
1492
1493 #ifdef CONFIG_IPV6_SIT_6RD
1494         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1495                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1496 #endif
1497
1498         return err;
1499 }
1500
1501 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1502                           struct nlattr *data[])
1503 {
1504         struct ip_tunnel *t = netdev_priv(dev);
1505         struct ip_tunnel_parm p;
1506         struct net *net = t->net;
1507         struct sit_net *sitn = net_generic(net, sit_net_id);
1508 #ifdef CONFIG_IPV6_SIT_6RD
1509         struct ip_tunnel_6rd ip6rd;
1510 #endif
1511
1512         if (dev == sitn->fb_tunnel_dev)
1513                 return -EINVAL;
1514
1515         ipip6_netlink_parms(data, &p);
1516
1517         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1518             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1519                 return -EINVAL;
1520
1521         t = ipip6_tunnel_locate(net, &p, 0);
1522
1523         if (t) {
1524                 if (t->dev != dev)
1525                         return -EEXIST;
1526         } else
1527                 t = netdev_priv(dev);
1528
1529         ipip6_tunnel_update(t, &p);
1530
1531 #ifdef CONFIG_IPV6_SIT_6RD
1532         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1533                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1534 #endif
1535
1536         return 0;
1537 }
1538
1539 static size_t ipip6_get_size(const struct net_device *dev)
1540 {
1541         return
1542                 /* IFLA_IPTUN_LINK */
1543                 nla_total_size(4) +
1544                 /* IFLA_IPTUN_LOCAL */
1545                 nla_total_size(4) +
1546                 /* IFLA_IPTUN_REMOTE */
1547                 nla_total_size(4) +
1548                 /* IFLA_IPTUN_TTL */
1549                 nla_total_size(1) +
1550                 /* IFLA_IPTUN_TOS */
1551                 nla_total_size(1) +
1552                 /* IFLA_IPTUN_PMTUDISC */
1553                 nla_total_size(1) +
1554                 /* IFLA_IPTUN_FLAGS */
1555                 nla_total_size(2) +
1556                 /* IFLA_IPTUN_PROTO */
1557                 nla_total_size(1) +
1558 #ifdef CONFIG_IPV6_SIT_6RD
1559                 /* IFLA_IPTUN_6RD_PREFIX */
1560                 nla_total_size(sizeof(struct in6_addr)) +
1561                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1562                 nla_total_size(4) +
1563                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1564                 nla_total_size(2) +
1565                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1566                 nla_total_size(2) +
1567 #endif
1568                 0;
1569 }
1570
1571 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1572 {
1573         struct ip_tunnel *tunnel = netdev_priv(dev);
1574         struct ip_tunnel_parm *parm = &tunnel->parms;
1575
1576         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1577             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1578             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1579             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1580             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1581             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1582                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1583             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1584             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1585                 goto nla_put_failure;
1586
1587 #ifdef CONFIG_IPV6_SIT_6RD
1588         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1589                     &tunnel->ip6rd.prefix) ||
1590             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1591                          tunnel->ip6rd.relay_prefix) ||
1592             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1593                         tunnel->ip6rd.prefixlen) ||
1594             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1595                         tunnel->ip6rd.relay_prefixlen))
1596                 goto nla_put_failure;
1597 #endif
1598
1599         return 0;
1600
1601 nla_put_failure:
1602         return -EMSGSIZE;
1603 }
1604
1605 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1606         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1607         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1608         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1609         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1610         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1611         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1612         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1613         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1614 #ifdef CONFIG_IPV6_SIT_6RD
1615         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1616         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1617         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1618         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1619 #endif
1620 };
1621
1622 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1623 {
1624         struct net *net = dev_net(dev);
1625         struct sit_net *sitn = net_generic(net, sit_net_id);
1626
1627         if (dev != sitn->fb_tunnel_dev)
1628                 unregister_netdevice_queue(dev, head);
1629 }
1630
1631 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1632         .kind           = "sit",
1633         .maxtype        = IFLA_IPTUN_MAX,
1634         .policy         = ipip6_policy,
1635         .priv_size      = sizeof(struct ip_tunnel),
1636         .setup          = ipip6_tunnel_setup,
1637         .validate       = ipip6_validate,
1638         .newlink        = ipip6_newlink,
1639         .changelink     = ipip6_changelink,
1640         .get_size       = ipip6_get_size,
1641         .fill_info      = ipip6_fill_info,
1642         .dellink        = ipip6_dellink,
1643 };
1644
1645 static struct xfrm_tunnel sit_handler __read_mostly = {
1646         .handler        =       ipip6_rcv,
1647         .err_handler    =       ipip6_err,
1648         .priority       =       1,
1649 };
1650
1651 static struct xfrm_tunnel ipip_handler __read_mostly = {
1652         .handler        =       ipip_rcv,
1653         .err_handler    =       ipip6_err,
1654         .priority       =       2,
1655 };
1656
1657 static void __net_exit sit_destroy_tunnels(struct net *net,
1658                                            struct list_head *head)
1659 {
1660         struct sit_net *sitn = net_generic(net, sit_net_id);
1661         struct net_device *dev, *aux;
1662         int prio;
1663
1664         for_each_netdev_safe(net, dev, aux)
1665                 if (dev->rtnl_link_ops == &sit_link_ops)
1666                         unregister_netdevice_queue(dev, head);
1667
1668         for (prio = 1; prio < 4; prio++) {
1669                 int h;
1670                 for (h = 0; h < HASH_SIZE; h++) {
1671                         struct ip_tunnel *t;
1672
1673                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1674                         while (t != NULL) {
1675                                 /* If dev is in the same netns, it has already
1676                                  * been added to the list by the previous loop.
1677                                  */
1678                                 if (!net_eq(dev_net(t->dev), net))
1679                                         unregister_netdevice_queue(t->dev,
1680                                                                    head);
1681                                 t = rtnl_dereference(t->next);
1682                         }
1683                 }
1684         }
1685 }
1686
1687 static int __net_init sit_init_net(struct net *net)
1688 {
1689         struct sit_net *sitn = net_generic(net, sit_net_id);
1690         struct ip_tunnel *t;
1691         int err;
1692
1693         sitn->tunnels[0] = sitn->tunnels_wc;
1694         sitn->tunnels[1] = sitn->tunnels_l;
1695         sitn->tunnels[2] = sitn->tunnels_r;
1696         sitn->tunnels[3] = sitn->tunnels_r_l;
1697
1698         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1699                                            ipip6_tunnel_setup);
1700         if (!sitn->fb_tunnel_dev) {
1701                 err = -ENOMEM;
1702                 goto err_alloc_dev;
1703         }
1704         dev_net_set(sitn->fb_tunnel_dev, net);
1705         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1706         /* FB netdevice is special: we have one, and only one per netns.
1707          * Allowing to move it to another netns is clearly unsafe.
1708          */
1709         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1710
1711         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1712         if (err)
1713                 goto err_dev_free;
1714
1715         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1716
1717         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1718                 goto err_reg_dev;
1719
1720         t = netdev_priv(sitn->fb_tunnel_dev);
1721
1722         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1723         return 0;
1724
1725 err_reg_dev:
1726         dev_put(sitn->fb_tunnel_dev);
1727 err_dev_free:
1728         ipip6_dev_free(sitn->fb_tunnel_dev);
1729 err_alloc_dev:
1730         return err;
1731 }
1732
1733 static void __net_exit sit_exit_net(struct net *net)
1734 {
1735         LIST_HEAD(list);
1736
1737         rtnl_lock();
1738         sit_destroy_tunnels(net, &list);
1739         unregister_netdevice_many(&list);
1740         rtnl_unlock();
1741 }
1742
1743 static struct pernet_operations sit_net_ops = {
1744         .init = sit_init_net,
1745         .exit = sit_exit_net,
1746         .id   = &sit_net_id,
1747         .size = sizeof(struct sit_net),
1748 };
1749
1750 static void __exit sit_cleanup(void)
1751 {
1752         rtnl_link_unregister(&sit_link_ops);
1753         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1754         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1755
1756         unregister_pernet_device(&sit_net_ops);
1757         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1758 }
1759
1760 static int __init sit_init(void)
1761 {
1762         int err;
1763
1764         pr_info("IPv6 over IPv4 tunneling driver\n");
1765
1766         err = register_pernet_device(&sit_net_ops);
1767         if (err < 0)
1768                 return err;
1769         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1770         if (err < 0) {
1771                 pr_info("%s: can't register ip6ip4\n", __func__);
1772                 goto xfrm_tunnel_failed;
1773         }
1774         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1775         if (err < 0) {
1776                 pr_info("%s: can't register ip4ip4\n", __func__);
1777                 goto xfrm_tunnel4_failed;
1778         }
1779         err = rtnl_link_register(&sit_link_ops);
1780         if (err < 0)
1781                 goto rtnl_link_failed;
1782
1783 out:
1784         return err;
1785
1786 rtnl_link_failed:
1787         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1788 xfrm_tunnel4_failed:
1789         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1790 xfrm_tunnel_failed:
1791         unregister_pernet_device(&sit_net_ops);
1792         goto out;
1793 }
1794
1795 module_init(sit_init);
1796 module_exit(sit_cleanup);
1797 MODULE_LICENSE("GPL");
1798 MODULE_ALIAS_NETDEV("sit0");