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 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
 482  * if sufficient data bytes are available
 483  */
 484 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
 485 {
 486         const struct iphdr *iph = (const struct iphdr *) skb->data;
 487         struct rt6_info *rt;
 488         struct sk_buff *skb2;
 489
 490         if (!pskb_may_pull(skb, iph->ihl * 4 + sizeof(struct ipv6hdr) + 8))
 491                 return 1;
 492
 493         skb2 = skb_clone(skb, GFP_ATOMIC);
 494
 495         if (!skb2)
 496                 return 1;
 497
 498         skb_dst_drop(skb2);
 499         skb_pull(skb2, iph->ihl * 4);
 500         skb_reset_network_header(skb2);
 501
 502         rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
 503
 504         if (rt && rt->dst.dev)
 505                 skb2->dev = rt->dst.dev;
 506
 507         icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
 508
 509         if (rt)
 510                 ip6_rt_put(rt);
 511
 512         kfree_skb(skb2);
 513
 514         return 0;
 515 }
 516
 517 static int ipip6_err(struct sk_buff *skb, u32 info)
 518 {
 519         const struct iphdr *iph = (const struct iphdr *)skb->data;
 520         const int type = icmp_hdr(skb)->type;
 521         const int code = icmp_hdr(skb)->code;
 522         struct ip_tunnel *t;
 523         int err;
 524
 525         switch (type) {
 526         default:
 527         case ICMP_PARAMETERPROB:
 528                 return 0;
 529
 530         case ICMP_DEST_UNREACH:
 531                 switch (code) {
 532                 case ICMP_SR_FAILED:
 533                         /* Impossible event. */
 534                         return 0;
 535                 default:
 536                         /* All others are translated to HOST_UNREACH.
 537                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 538                            I believe they are just ether pollution. --ANK
 539                          */
 540                         break;
 541                 }
 542                 break;
 543         case ICMP_TIME_EXCEEDED:
 544                 if (code != ICMP_EXC_TTL)
 545                         return 0;
 546                 break;
 547         case ICMP_REDIRECT:
 548                 break;
 549         }
 550
 551         err = -ENOENT;
 552
 553         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 554                                 skb->dev,
 555                                 iph->daddr,
 556                                 iph->saddr);
 557         if (t == NULL)
 558                 goto out;
 559
 560         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 561                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 562                                  t->dev->ifindex, 0, IPPROTO_IPV6, 0);
 563                 err = 0;
 564                 goto out;
 565         }
 566         if (type == ICMP_REDIRECT) {
 567                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 568                               IPPROTO_IPV6, 0);
 569                 err = 0;
 570                 goto out;
 571         }
 572
 573         if (t->parms.iph.daddr == 0)
 574                 goto out;
 575
 576         err = 0;
 577         if (!ipip6_err_gen_icmpv6_unreach(skb))
 578                 goto out;
 579
 580         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 581                 goto out;
 582
 583         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 584                 t->err_count++;
 585         else
 586                 t->err_count = 1;
 587         t->err_time = jiffies;
 588 out:
 589         return err;
 590 }
 591
 592 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 593                                   const struct in6_addr *v6addr)
 594 {
 595         __be32 v4embed = 0;
 596         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 597                 return true;
 598         return false;
 599 }
 600
 601 /* Checks if an address matches an address on the tunnel interface.
 602  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 603  * Long story:
 604  * This function is called after we considered the packet as spoofed
 605  * in is_spoofed_6rd.
 606  * We may have a router that is doing NAT for proto 41 packets
 607  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 608  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 609  * function will return true, dropping the packet.
 610  * But, we can still check if is spoofed against the IP
 611  * addresses associated with the interface.
 612  */
 613 static bool only_dnatted(const struct ip_tunnel *tunnel,
 614         const struct in6_addr *v6dst)
 615 {
 616         int prefix_len;
 617
 618 #ifdef CONFIG_IPV6_SIT_6RD
 619         prefix_len = tunnel->ip6rd.prefixlen + 32
 620                 - tunnel->ip6rd.relay_prefixlen;
 621 #else
 622         prefix_len = 48;
 623 #endif
 624         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 625 }
 626
 627 /* Returns true if a packet is spoofed */
 628 static bool packet_is_spoofed(struct sk_buff *skb,
 629                               const struct iphdr *iph,
 630                               struct ip_tunnel *tunnel)
 631 {
 632         const struct ipv6hdr *ipv6h;
 633
 634         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 635                 if (!isatap_chksrc(skb, iph, tunnel))
 636                         return true;
 637
 638                 return false;
 639         }
 640
 641         if (tunnel->dev->flags & IFF_POINTOPOINT)
 642                 return false;
 643
 644         ipv6h = ipv6_hdr(skb);
 645
 646         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 647                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 648                                      &iph->saddr, &ipv6h->saddr,
 649                                      &iph->daddr, &ipv6h->daddr);
 650                 return true;
 651         }
 652
 653         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 654                 return false;
 655
 656         if (only_dnatted(tunnel, &ipv6h->daddr))
 657                 return false;
 658
 659         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 660                              &iph->saddr, &ipv6h->saddr,
 661                              &iph->daddr, &ipv6h->daddr);
 662         return true;
 663 }
 664
 665 static int ipip6_rcv(struct sk_buff *skb)
 666 {
 667         const struct iphdr *iph = ip_hdr(skb);
 668         struct ip_tunnel *tunnel;
 669         int err;
 670
 671         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 672                                      iph->saddr, iph->daddr);
 673         if (tunnel != NULL) {
 674                 struct pcpu_sw_netstats *tstats;
 675
 676                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 677                     tunnel->parms.iph.protocol != 0)
 678                         goto out;
 679
 680                 skb->mac_header = skb->network_header;
 681                 skb_reset_network_header(skb);
 682                 IPCB(skb)->flags = 0;
 683                 skb->protocol = htons(ETH_P_IPV6);
 684
 685                 if (packet_is_spoofed(skb, iph, tunnel)) {
 686                         tunnel->dev->stats.rx_errors++;
 687                         goto out;
 688                 }
 689
 690                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 691
 692                 err = IP_ECN_decapsulate(iph, skb);
 693                 if (unlikely(err)) {
 694                         if (log_ecn_error)
 695                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 696                                                      &iph->saddr, iph->tos);
 697                         if (err > 1) {
 698                                 ++tunnel->dev->stats.rx_frame_errors;
 699                                 ++tunnel->dev->stats.rx_errors;
 700                                 goto out;
 701                         }
 702                 }
 703
 704                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 705                 u64_stats_update_begin(&tstats->syncp);
 706                 tstats->rx_packets++;
 707                 tstats->rx_bytes += skb->len;
 708                 u64_stats_update_end(&tstats->syncp);
 709
 710                 netif_rx(skb);
 711
 712                 return 0;
 713         }
 714
 715         /* no tunnel matched,  let upstream know, ipsec may handle it */
 716         return 1;
 717 out:
 718         kfree_skb(skb);
 719         return 0;
 720 }
 721
 722 static const struct tnl_ptk_info tpi = {
 723         /* no tunnel info required for ipip. */
 724         .proto = htons(ETH_P_IP),
 725 };
 726
 727 static int ipip_rcv(struct sk_buff *skb)
 728 {
 729         const struct iphdr *iph;
 730         struct ip_tunnel *tunnel;
 731
 732         iph = ip_hdr(skb);
 733         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 734                                      iph->saddr, iph->daddr);
 735         if (tunnel != NULL) {
 736                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 737                     tunnel->parms.iph.protocol != 0)
 738                         goto drop;
 739
 740                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 741                         goto drop;
 742                 if (iptunnel_pull_header(skb, 0, tpi.proto))
 743                         goto drop;
 744                 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 745         }
 746
 747         return 1;
 748
 749 drop:
 750         kfree_skb(skb);
 751         return 0;
 752 }
 753
 754 /*
 755  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 756  * stores the embedded IPv4 address in v4dst and returns true.
 757  */
 758 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 759                       __be32 *v4dst)
 760 {
 761 #ifdef CONFIG_IPV6_SIT_6RD
 762         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 763                               tunnel->ip6rd.prefixlen)) {
 764                 unsigned int pbw0, pbi0;
 765                 int pbi1;
 766                 u32 d;
 767
 768                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 769                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 770
 771                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 772                     tunnel->ip6rd.relay_prefixlen;
 773
 774                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 775                 if (pbi1 > 0)
 776                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 777                              (32 - pbi1);
 778
 779                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 780                 return true;
 781         }
 782 #else
 783         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 784                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 785                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 786                 return true;
 787         }
 788 #endif
 789         return false;
 790 }
 791
 792 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 793                              const struct in6_addr *v6dst)
 794 {
 795         __be32 dst = 0;
 796         check_6rd(tunnel, v6dst, &dst);
 797         return dst;
 798 }
 799
 800 /*
 801  *      This function assumes it is being called from dev_queue_xmit()
 802  *      and that skb is filled properly by that function.
 803  */
 804
 805 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 806                                      struct net_device *dev)
 807 {
 808         struct ip_tunnel *tunnel = netdev_priv(dev);
 809         const struct iphdr  *tiph = &tunnel->parms.iph;
 810         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 811         u8     tos = tunnel->parms.iph.tos;
 812         __be16 df = tiph->frag_off;
 813         struct rtable *rt;                      /* Route to the other host */
 814         struct net_device *tdev;                /* Device to other host */
 815         unsigned int max_headroom;              /* The extra header space needed */
 816         __be32 dst = tiph->daddr;
 817         struct flowi4 fl4;
 818         int    mtu;
 819         const struct in6_addr *addr6;
 820         int addr_type;
 821         u8 ttl;
 822         int err;
 823
 824         if (skb->protocol != htons(ETH_P_IPV6))
 825                 goto tx_error;
 826
 827         if (tos == 1)
 828                 tos = ipv6_get_dsfield(iph6);
 829
 830         /* ISATAP (RFC4214) - must come before 6to4 */
 831         if (dev->priv_flags & IFF_ISATAP) {
 832                 struct neighbour *neigh = NULL;
 833                 bool do_tx_error = false;
 834
 835                 if (skb_dst(skb))
 836                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 837
 838                 if (neigh == NULL) {
 839                         net_dbg_ratelimited("nexthop == NULL\n");
 840                         goto tx_error;
 841                 }
 842
 843                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 844                 addr_type = ipv6_addr_type(addr6);
 845
 846                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 847                      ipv6_addr_is_isatap(addr6))
 848                         dst = addr6->s6_addr32[3];
 849                 else
 850                         do_tx_error = true;
 851
 852                 neigh_release(neigh);
 853                 if (do_tx_error)
 854                         goto tx_error;
 855         }
 856
 857         if (!dst)
 858                 dst = try_6rd(tunnel, &iph6->daddr);
 859
 860         if (!dst) {
 861                 struct neighbour *neigh = NULL;
 862                 bool do_tx_error = false;
 863
 864                 if (skb_dst(skb))
 865                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 866
 867                 if (neigh == NULL) {
 868                         net_dbg_ratelimited("nexthop == NULL\n");
 869                         goto tx_error;
 870                 }
 871
 872                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 873                 addr_type = ipv6_addr_type(addr6);
 874
 875                 if (addr_type == IPV6_ADDR_ANY) {
 876                         addr6 = &ipv6_hdr(skb)->daddr;
 877                         addr_type = ipv6_addr_type(addr6);
 878                 }
 879
 880                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 881                         dst = addr6->s6_addr32[3];
 882                 else
 883                         do_tx_error = true;
 884
 885                 neigh_release(neigh);
 886                 if (do_tx_error)
 887                         goto tx_error;
 888         }
 889
 890         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 891                                    dst, tiph->saddr,
 892                                    0, 0,
 893                                    IPPROTO_IPV6, RT_TOS(tos),
 894                                    tunnel->parms.link);
 895         if (IS_ERR(rt)) {
 896                 dev->stats.tx_carrier_errors++;
 897                 goto tx_error_icmp;
 898         }
 899         if (rt->rt_type != RTN_UNICAST) {
 900                 ip_rt_put(rt);
 901                 dev->stats.tx_carrier_errors++;
 902                 goto tx_error_icmp;
 903         }
 904         tdev = rt->dst.dev;
 905
 906         if (tdev == dev) {
 907                 ip_rt_put(rt);
 908                 dev->stats.collisions++;
 909                 goto tx_error;
 910         }
 911
 912         if (df) {
 913                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 914
 915                 if (mtu < 68) {
 916                         dev->stats.collisions++;
 917                         ip_rt_put(rt);
 918                         goto tx_error;
 919                 }
 920
 921                 if (mtu < IPV6_MIN_MTU) {
 922                         mtu = IPV6_MIN_MTU;
 923                         df = 0;
 924                 }
 925
 926                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 927                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 928
 929                 if (skb->len > mtu && !skb_is_gso(skb)) {
 930                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 931                         ip_rt_put(rt);
 932                         goto tx_error;
 933                 }
 934         }
 935
 936         if (tunnel->err_count > 0) {
 937                 if (time_before(jiffies,
 938                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 939                         tunnel->err_count--;
 940                         dst_link_failure(skb);
 941                 } else
 942                         tunnel->err_count = 0;
 943         }
 944
 945         /*
 946          * Okay, now see if we can stuff it in the buffer as-is.
 947          */
 948         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 949
 950         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 951             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 952                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 953                 if (!new_skb) {
 954                         ip_rt_put(rt);
 955                         dev->stats.tx_dropped++;
 956                         kfree_skb(skb);
 957                         return NETDEV_TX_OK;
 958                 }
 959                 if (skb->sk)
 960                         skb_set_owner_w(new_skb, skb->sk);
 961                 dev_kfree_skb(skb);
 962                 skb = new_skb;
 963                 iph6 = ipv6_hdr(skb);
 964         }
 965         ttl = tiph->ttl;
 966         if (ttl == 0)
 967                 ttl = iph6->hop_limit;
 968         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 969
 970         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
 971         if (IS_ERR(skb)) {
 972                 ip_rt_put(rt);
 973                 goto out;
 974         }
 975
 976         err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, IPPROTO_IPV6, tos,
 977                             ttl, df, !net_eq(tunnel->net, dev_net(dev)));
 978         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 979         return NETDEV_TX_OK;
 980
 981 tx_error_icmp:
 982         dst_link_failure(skb);
 983 tx_error:
 984         kfree_skb(skb);
 985 out:
 986         dev->stats.tx_errors++;
 987         return NETDEV_TX_OK;
 988 }
 989
 990 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 991 {
 992         struct ip_tunnel *tunnel = netdev_priv(dev);
 993         const struct iphdr  *tiph = &tunnel->parms.iph;
 994
 995         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
 996         if (IS_ERR(skb))
 997                 goto out;
 998
 999         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1000         return NETDEV_TX_OK;
1001 out:
1002         dev->stats.tx_errors++;
1003         return NETDEV_TX_OK;
1004 }
1005
1006 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1007                                    struct net_device *dev)
1008 {
1009         switch (skb->protocol) {
1010         case htons(ETH_P_IP):
1011                 ipip_tunnel_xmit(skb, dev);
1012                 break;
1013         case htons(ETH_P_IPV6):
1014                 ipip6_tunnel_xmit(skb, dev);
1015                 break;
1016         default:
1017                 goto tx_err;
1018         }
1019
1020         return NETDEV_TX_OK;
1021
1022 tx_err:
1023         dev->stats.tx_errors++;
1024         kfree_skb(skb);
1025         return NETDEV_TX_OK;
1026
1027 }
1028
1029 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1030 {
1031         struct net_device *tdev = NULL;
1032         struct ip_tunnel *tunnel;
1033         const struct iphdr *iph;
1034         struct flowi4 fl4;
1035
1036         tunnel = netdev_priv(dev);
1037         iph = &tunnel->parms.iph;
1038
1039         if (iph->daddr) {
1040                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1041                                                           NULL,
1042                                                           iph->daddr, iph->saddr,
1043                                                           0, 0,
1044                                                           IPPROTO_IPV6,
1045                                                           RT_TOS(iph->tos),
1046                                                           tunnel->parms.link);
1047
1048                 if (!IS_ERR(rt)) {
1049                         tdev = rt->dst.dev;
1050                         ip_rt_put(rt);
1051                 }
1052                 dev->flags |= IFF_POINTOPOINT;
1053         }
1054
1055         if (!tdev && tunnel->parms.link)
1056                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1057
1058         if (tdev) {
1059                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1060                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
1061                 if (dev->mtu < IPV6_MIN_MTU)
1062                         dev->mtu = IPV6_MIN_MTU;
1063         }
1064         dev->iflink = tunnel->parms.link;
1065 }
1066
1067 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1068 {
1069         struct net *net = t->net;
1070         struct sit_net *sitn = net_generic(net, sit_net_id);
1071
1072         ipip6_tunnel_unlink(sitn, t);
1073         synchronize_net();
1074         t->parms.iph.saddr = p->iph.saddr;
1075         t->parms.iph.daddr = p->iph.daddr;
1076         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1077         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1078         ipip6_tunnel_link(sitn, t);
1079         t->parms.iph.ttl = p->iph.ttl;
1080         t->parms.iph.tos = p->iph.tos;
1081         if (t->parms.link != p->link) {
1082                 t->parms.link = p->link;
1083                 ipip6_tunnel_bind_dev(t->dev);
1084         }
1085         netdev_state_change(t->dev);
1086 }
1087
1088 #ifdef CONFIG_IPV6_SIT_6RD
1089 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1090                                    struct ip_tunnel_6rd *ip6rd)
1091 {
1092         struct in6_addr prefix;
1093         __be32 relay_prefix;
1094
1095         if (ip6rd->relay_prefixlen > 32 ||
1096             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1097                 return -EINVAL;
1098
1099         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1100         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1101                 return -EINVAL;
1102         if (ip6rd->relay_prefixlen)
1103                 relay_prefix = ip6rd->relay_prefix &
1104                                htonl(0xffffffffUL <<
1105                                      (32 - ip6rd->relay_prefixlen));
1106         else
1107                 relay_prefix = 0;
1108         if (relay_prefix != ip6rd->relay_prefix)
1109                 return -EINVAL;
1110
1111         t->ip6rd.prefix = prefix;
1112         t->ip6rd.relay_prefix = relay_prefix;
1113         t->ip6rd.prefixlen = ip6rd->prefixlen;
1114         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1115         netdev_state_change(t->dev);
1116         return 0;
1117 }
1118 #endif
1119
1120 static int
1121 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1122 {
1123         int err = 0;
1124         struct ip_tunnel_parm p;
1125         struct ip_tunnel_prl prl;
1126         struct ip_tunnel *t;
1127         struct net *net = dev_net(dev);
1128         struct sit_net *sitn = net_generic(net, sit_net_id);
1129 #ifdef CONFIG_IPV6_SIT_6RD
1130         struct ip_tunnel_6rd ip6rd;
1131 #endif
1132
1133         switch (cmd) {
1134         case SIOCGETTUNNEL:
1135 #ifdef CONFIG_IPV6_SIT_6RD
1136         case SIOCGET6RD:
1137 #endif
1138                 t = NULL;
1139                 if (dev == sitn->fb_tunnel_dev) {
1140                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1141                                 err = -EFAULT;
1142                                 break;
1143                         }
1144                         t = ipip6_tunnel_locate(net, &p, 0);
1145                 }
1146                 if (t == NULL)
1147                         t = netdev_priv(dev);
1148
1149                 err = -EFAULT;
1150                 if (cmd == SIOCGETTUNNEL) {
1151                         memcpy(&p, &t->parms, sizeof(p));
1152                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1153                                          sizeof(p)))
1154                                 goto done;
1155 #ifdef CONFIG_IPV6_SIT_6RD
1156                 } else {
1157                         ip6rd.prefix = t->ip6rd.prefix;
1158                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1159                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1160                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1161                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1162                                          sizeof(ip6rd)))
1163                                 goto done;
1164 #endif
1165                 }
1166                 err = 0;
1167                 break;
1168
1169         case SIOCADDTUNNEL:
1170         case SIOCCHGTUNNEL:
1171                 err = -EPERM;
1172                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1173                         goto done;
1174
1175                 err = -EFAULT;
1176                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1177                         goto done;
1178
1179                 err = -EINVAL;
1180                 if (p.iph.protocol != IPPROTO_IPV6 &&
1181                     p.iph.protocol != IPPROTO_IPIP &&
1182                     p.iph.protocol != 0)
1183                         goto done;
1184                 if (p.iph.version != 4 ||
1185                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1186                         goto done;
1187                 if (p.iph.ttl)
1188                         p.iph.frag_off |= htons(IP_DF);
1189
1190                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1191
1192                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1193                         if (t != NULL) {
1194                                 if (t->dev != dev) {
1195                                         err = -EEXIST;
1196                                         break;
1197                                 }
1198                         } else {
1199                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1200                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1201                                         err = -EINVAL;
1202                                         break;
1203                                 }
1204                                 t = netdev_priv(dev);
1205                         }
1206
1207                         ipip6_tunnel_update(t, &p);
1208                 }
1209
1210                 if (t) {
1211                         err = 0;
1212                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1213                                 err = -EFAULT;
1214                 } else
1215                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1216                 break;
1217
1218         case SIOCDELTUNNEL:
1219                 err = -EPERM;
1220                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1221                         goto done;
1222
1223                 if (dev == sitn->fb_tunnel_dev) {
1224                         err = -EFAULT;
1225                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1226                                 goto done;
1227                         err = -ENOENT;
1228                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1229                                 goto done;
1230                         err = -EPERM;
1231                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1232                                 goto done;
1233                         dev = t->dev;
1234                 }
1235                 unregister_netdevice(dev);
1236                 err = 0;
1237                 break;
1238
1239         case SIOCGETPRL:
1240                 err = -EINVAL;
1241                 if (dev == sitn->fb_tunnel_dev)
1242                         goto done;
1243                 err = -ENOENT;
1244                 if (!(t = netdev_priv(dev)))
1245                         goto done;
1246                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1247                 break;
1248
1249         case SIOCADDPRL:
1250         case SIOCDELPRL:
1251         case SIOCCHGPRL:
1252                 err = -EPERM;
1253                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1254                         goto done;
1255                 err = -EINVAL;
1256                 if (dev == sitn->fb_tunnel_dev)
1257                         goto done;
1258                 err = -EFAULT;
1259                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1260                         goto done;
1261                 err = -ENOENT;
1262                 if (!(t = netdev_priv(dev)))
1263                         goto done;
1264
1265                 switch (cmd) {
1266                 case SIOCDELPRL:
1267                         err = ipip6_tunnel_del_prl(t, &prl);
1268                         break;
1269                 case SIOCADDPRL:
1270                 case SIOCCHGPRL:
1271                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1272                         break;
1273                 }
1274                 netdev_state_change(dev);
1275                 break;
1276
1277 #ifdef CONFIG_IPV6_SIT_6RD
1278         case SIOCADD6RD:
1279         case SIOCCHG6RD:
1280         case SIOCDEL6RD:
1281                 err = -EPERM;
1282                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1283                         goto done;
1284
1285                 err = -EFAULT;
1286                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1287                                    sizeof(ip6rd)))
1288                         goto done;
1289
1290                 t = netdev_priv(dev);
1291
1292                 if (cmd != SIOCDEL6RD) {
1293                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1294                         if (err < 0)
1295                                 goto done;
1296                 } else
1297                         ipip6_tunnel_clone_6rd(dev, sitn);
1298
1299                 err = 0;
1300                 break;
1301 #endif
1302
1303         default:
1304                 err = -EINVAL;
1305         }
1306
1307 done:
1308         return err;
1309 }
1310
1311 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1312 {
1313         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1314                 return -EINVAL;
1315         dev->mtu = new_mtu;
1316         return 0;
1317 }
1318
1319 static const struct net_device_ops ipip6_netdev_ops = {
1320         .ndo_uninit     = ipip6_tunnel_uninit,
1321         .ndo_start_xmit = sit_tunnel_xmit,
1322         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1323         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1324         .ndo_get_stats64 = ip_tunnel_get_stats64,
1325 };
1326
1327 static void ipip6_dev_free(struct net_device *dev)
1328 {
1329         free_percpu(dev->tstats);
1330         free_netdev(dev);
1331 }
1332
1333 #define SIT_FEATURES (NETIF_F_SG           | \
1334                       NETIF_F_FRAGLIST     | \
1335                       NETIF_F_HIGHDMA      | \
1336                       NETIF_F_GSO_SOFTWARE | \
1337                       NETIF_F_HW_CSUM)
1338
1339 static void ipip6_tunnel_setup(struct net_device *dev)
1340 {
1341         dev->netdev_ops         = &ipip6_netdev_ops;
1342         dev->destructor         = ipip6_dev_free;
1343
1344         dev->type               = ARPHRD_SIT;
1345         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1346         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1347         dev->flags              = IFF_NOARP;
1348         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1349         dev->iflink             = 0;
1350         dev->addr_len           = 4;
1351         dev->features           |= NETIF_F_LLTX;
1352         dev->features           |= SIT_FEATURES;
1353         dev->hw_features        |= SIT_FEATURES;
1354 }
1355
1356 static int ipip6_tunnel_init(struct net_device *dev)
1357 {
1358         struct ip_tunnel *tunnel = netdev_priv(dev);
1359
1360         tunnel->dev = dev;
1361         tunnel->net = dev_net(dev);
1362
1363         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1364         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1365
1366         ipip6_tunnel_bind_dev(dev);
1367         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1368         if (!dev->tstats)
1369                 return -ENOMEM;
1370
1371         return 0;
1372 }
1373
1374 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1375 {
1376         struct ip_tunnel *tunnel = netdev_priv(dev);
1377         struct iphdr *iph = &tunnel->parms.iph;
1378         struct net *net = dev_net(dev);
1379         struct sit_net *sitn = net_generic(net, sit_net_id);
1380
1381         tunnel->dev = dev;
1382         tunnel->net = dev_net(dev);
1383         strcpy(tunnel->parms.name, dev->name);
1384
1385         iph->version            = 4;
1386         iph->protocol           = IPPROTO_IPV6;
1387         iph->ihl                = 5;
1388         iph->ttl                = 64;
1389
1390         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1391         if (!dev->tstats)
1392                 return -ENOMEM;
1393
1394         dev_hold(dev);
1395         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1396         return 0;
1397 }
1398
1399 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1400 {
1401         u8 proto;
1402
1403         if (!data || !data[IFLA_IPTUN_PROTO])
1404                 return 0;
1405
1406         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1407         if (proto != IPPROTO_IPV6 &&
1408             proto != IPPROTO_IPIP &&
1409             proto != 0)
1410                 return -EINVAL;
1411
1412         return 0;
1413 }
1414
1415 static void ipip6_netlink_parms(struct nlattr *data[],
1416                                 struct ip_tunnel_parm *parms)
1417 {
1418         memset(parms, 0, sizeof(*parms));
1419
1420         parms->iph.version = 4;
1421         parms->iph.protocol = IPPROTO_IPV6;
1422         parms->iph.ihl = 5;
1423         parms->iph.ttl = 64;
1424
1425         if (!data)
1426                 return;
1427
1428         if (data[IFLA_IPTUN_LINK])
1429                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1430
1431         if (data[IFLA_IPTUN_LOCAL])
1432                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1433
1434         if (data[IFLA_IPTUN_REMOTE])
1435                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1436
1437         if (data[IFLA_IPTUN_TTL]) {
1438                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1439                 if (parms->iph.ttl)
1440                         parms->iph.frag_off = htons(IP_DF);
1441         }
1442
1443         if (data[IFLA_IPTUN_TOS])
1444                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1445
1446         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1447                 parms->iph.frag_off = htons(IP_DF);
1448
1449         if (data[IFLA_IPTUN_FLAGS])
1450                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1451
1452         if (data[IFLA_IPTUN_PROTO])
1453                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1454
1455 }
1456
1457 #ifdef CONFIG_IPV6_SIT_6RD
1458 /* This function returns true when 6RD attributes are present in the nl msg */
1459 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1460                                     struct ip_tunnel_6rd *ip6rd)
1461 {
1462         bool ret = false;
1463         memset(ip6rd, 0, sizeof(*ip6rd));
1464
1465         if (!data)
1466                 return ret;
1467
1468         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1469                 ret = true;
1470                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1471                            sizeof(struct in6_addr));
1472         }
1473
1474         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1475                 ret = true;
1476                 ip6rd->relay_prefix =
1477                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1478         }
1479
1480         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1481                 ret = true;
1482                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1483         }
1484
1485         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1486                 ret = true;
1487                 ip6rd->relay_prefixlen =
1488                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1489         }
1490
1491         return ret;
1492 }
1493 #endif
1494
1495 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1496                          struct nlattr *tb[], struct nlattr *data[])
1497 {
1498         struct net *net = dev_net(dev);
1499         struct ip_tunnel *nt;
1500 #ifdef CONFIG_IPV6_SIT_6RD
1501         struct ip_tunnel_6rd ip6rd;
1502 #endif
1503         int err;
1504
1505         nt = netdev_priv(dev);
1506         ipip6_netlink_parms(data, &nt->parms);
1507
1508         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1509                 return -EEXIST;
1510
1511         err = ipip6_tunnel_create(dev);
1512         if (err < 0)
1513                 return err;
1514
1515 #ifdef CONFIG_IPV6_SIT_6RD
1516         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1517                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1518 #endif
1519
1520         return err;
1521 }
1522
1523 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1524                           struct nlattr *data[])
1525 {
1526         struct ip_tunnel *t = netdev_priv(dev);
1527         struct ip_tunnel_parm p;
1528         struct net *net = t->net;
1529         struct sit_net *sitn = net_generic(net, sit_net_id);
1530 #ifdef CONFIG_IPV6_SIT_6RD
1531         struct ip_tunnel_6rd ip6rd;
1532 #endif
1533
1534         if (dev == sitn->fb_tunnel_dev)
1535                 return -EINVAL;
1536
1537         ipip6_netlink_parms(data, &p);
1538
1539         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1540             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1541                 return -EINVAL;
1542
1543         t = ipip6_tunnel_locate(net, &p, 0);
1544
1545         if (t) {
1546                 if (t->dev != dev)
1547                         return -EEXIST;
1548         } else
1549                 t = netdev_priv(dev);
1550
1551         ipip6_tunnel_update(t, &p);
1552
1553 #ifdef CONFIG_IPV6_SIT_6RD
1554         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1555                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1556 #endif
1557
1558         return 0;
1559 }
1560
1561 static size_t ipip6_get_size(const struct net_device *dev)
1562 {
1563         return
1564                 /* IFLA_IPTUN_LINK */
1565                 nla_total_size(4) +
1566                 /* IFLA_IPTUN_LOCAL */
1567                 nla_total_size(4) +
1568                 /* IFLA_IPTUN_REMOTE */
1569                 nla_total_size(4) +
1570                 /* IFLA_IPTUN_TTL */
1571                 nla_total_size(1) +
1572                 /* IFLA_IPTUN_TOS */
1573                 nla_total_size(1) +
1574                 /* IFLA_IPTUN_PMTUDISC */
1575                 nla_total_size(1) +
1576                 /* IFLA_IPTUN_FLAGS */
1577                 nla_total_size(2) +
1578                 /* IFLA_IPTUN_PROTO */
1579                 nla_total_size(1) +
1580 #ifdef CONFIG_IPV6_SIT_6RD
1581                 /* IFLA_IPTUN_6RD_PREFIX */
1582                 nla_total_size(sizeof(struct in6_addr)) +
1583                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1584                 nla_total_size(4) +
1585                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1586                 nla_total_size(2) +
1587                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1588                 nla_total_size(2) +
1589 #endif
1590                 0;
1591 }
1592
1593 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1594 {
1595         struct ip_tunnel *tunnel = netdev_priv(dev);
1596         struct ip_tunnel_parm *parm = &tunnel->parms;
1597
1598         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1599             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1600             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1601             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1602             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1603             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1604                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1605             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1606             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1607                 goto nla_put_failure;
1608
1609 #ifdef CONFIG_IPV6_SIT_6RD
1610         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1611                     &tunnel->ip6rd.prefix) ||
1612             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1613                          tunnel->ip6rd.relay_prefix) ||
1614             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1615                         tunnel->ip6rd.prefixlen) ||
1616             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1617                         tunnel->ip6rd.relay_prefixlen))
1618                 goto nla_put_failure;
1619 #endif
1620
1621         return 0;
1622
1623 nla_put_failure:
1624         return -EMSGSIZE;
1625 }
1626
1627 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1628         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1629         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1630         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1631         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1632         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1633         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1634         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1635         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1636 #ifdef CONFIG_IPV6_SIT_6RD
1637         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1638         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1639         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1640         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1641 #endif
1642 };
1643
1644 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1645 {
1646         struct net *net = dev_net(dev);
1647         struct sit_net *sitn = net_generic(net, sit_net_id);
1648
1649         if (dev != sitn->fb_tunnel_dev)
1650                 unregister_netdevice_queue(dev, head);
1651 }
1652
1653 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1654         .kind           = "sit",
1655         .maxtype        = IFLA_IPTUN_MAX,
1656         .policy         = ipip6_policy,
1657         .priv_size      = sizeof(struct ip_tunnel),
1658         .setup          = ipip6_tunnel_setup,
1659         .validate       = ipip6_validate,
1660         .newlink        = ipip6_newlink,
1661         .changelink     = ipip6_changelink,
1662         .get_size       = ipip6_get_size,
1663         .fill_info      = ipip6_fill_info,
1664         .dellink        = ipip6_dellink,
1665 };
1666
1667 static struct xfrm_tunnel sit_handler __read_mostly = {
1668         .handler        =       ipip6_rcv,
1669         .err_handler    =       ipip6_err,
1670         .priority       =       1,
1671 };
1672
1673 static struct xfrm_tunnel ipip_handler __read_mostly = {
1674         .handler        =       ipip_rcv,
1675         .err_handler    =       ipip6_err,
1676         .priority       =       2,
1677 };
1678
1679 static void __net_exit sit_destroy_tunnels(struct net *net,
1680                                            struct list_head *head)
1681 {
1682         struct sit_net *sitn = net_generic(net, sit_net_id);
1683         struct net_device *dev, *aux;
1684         int prio;
1685
1686         for_each_netdev_safe(net, dev, aux)
1687                 if (dev->rtnl_link_ops == &sit_link_ops)
1688                         unregister_netdevice_queue(dev, head);
1689
1690         for (prio = 1; prio < 4; prio++) {
1691                 int h;
1692                 for (h = 0; h < HASH_SIZE; h++) {
1693                         struct ip_tunnel *t;
1694
1695                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1696                         while (t != NULL) {
1697                                 /* If dev is in the same netns, it has already
1698                                  * been added to the list by the previous loop.
1699                                  */
1700                                 if (!net_eq(dev_net(t->dev), net))
1701                                         unregister_netdevice_queue(t->dev,
1702                                                                    head);
1703                                 t = rtnl_dereference(t->next);
1704                         }
1705                 }
1706         }
1707 }
1708
1709 static int __net_init sit_init_net(struct net *net)
1710 {
1711         struct sit_net *sitn = net_generic(net, sit_net_id);
1712         struct ip_tunnel *t;
1713         int err;
1714
1715         sitn->tunnels[0] = sitn->tunnels_wc;
1716         sitn->tunnels[1] = sitn->tunnels_l;
1717         sitn->tunnels[2] = sitn->tunnels_r;
1718         sitn->tunnels[3] = sitn->tunnels_r_l;
1719
1720         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1721                                            ipip6_tunnel_setup);
1722         if (!sitn->fb_tunnel_dev) {
1723                 err = -ENOMEM;
1724                 goto err_alloc_dev;
1725         }
1726         dev_net_set(sitn->fb_tunnel_dev, net);
1727         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1728         /* FB netdevice is special: we have one, and only one per netns.
1729          * Allowing to move it to another netns is clearly unsafe.
1730          */
1731         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1732
1733         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1734         if (err)
1735                 goto err_dev_free;
1736
1737         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1738
1739         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1740                 goto err_reg_dev;
1741
1742         t = netdev_priv(sitn->fb_tunnel_dev);
1743
1744         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1745         return 0;
1746
1747 err_reg_dev:
1748         dev_put(sitn->fb_tunnel_dev);
1749 err_dev_free:
1750         ipip6_dev_free(sitn->fb_tunnel_dev);
1751 err_alloc_dev:
1752         return err;
1753 }
1754
1755 static void __net_exit sit_exit_net(struct net *net)
1756 {
1757         LIST_HEAD(list);
1758
1759         rtnl_lock();
1760         sit_destroy_tunnels(net, &list);
1761         unregister_netdevice_many(&list);
1762         rtnl_unlock();
1763 }
1764
1765 static struct pernet_operations sit_net_ops = {
1766         .init = sit_init_net,
1767         .exit = sit_exit_net,
1768         .id   = &sit_net_id,
1769         .size = sizeof(struct sit_net),
1770 };
1771
1772 static void __exit sit_cleanup(void)
1773 {
1774         rtnl_link_unregister(&sit_link_ops);
1775         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1776         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1777
1778         unregister_pernet_device(&sit_net_ops);
1779         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1780 }
1781
1782 static int __init sit_init(void)
1783 {
1784         int err;
1785
1786         pr_info("IPv6 over IPv4 tunneling driver\n");
1787
1788         err = register_pernet_device(&sit_net_ops);
1789         if (err < 0)
1790                 return err;
1791         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1792         if (err < 0) {
1793                 pr_info("%s: can't register ip6ip4\n", __func__);
1794                 goto xfrm_tunnel_failed;
1795         }
1796         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1797         if (err < 0) {
1798                 pr_info("%s: can't register ip4ip4\n", __func__);
1799                 goto xfrm_tunnel4_failed;
1800         }
1801         err = rtnl_link_register(&sit_link_ops);
1802         if (err < 0)
1803                 goto rtnl_link_failed;
1804
1805 out:
1806         return err;
1807
1808 rtnl_link_failed:
1809         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1810 xfrm_tunnel4_failed:
1811         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1812 xfrm_tunnel_failed:
1813         unregister_pernet_device(&sit_net_ops);
1814         goto out;
1815 }
1816
1817 module_init(sit_init);
1818 module_exit(sit_cleanup);
1819 MODULE_LICENSE("GPL");
1820 MODULE_ALIAS_NETDEV("sit0");