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         ip_tunnel_dst_reset_all(tunnel);
 479         dev_put(dev);
 480 }
 481
 482 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH
 483  * if sufficient data bytes are available
 484  */
 485 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb)
 486 {
 487         const struct iphdr *iph = (const struct iphdr *) skb->data;
 488         struct rt6_info *rt;
 489         struct sk_buff *skb2;
 490
 491         if (!pskb_may_pull(skb, iph->ihl * 4 + sizeof(struct ipv6hdr) + 8))
 492                 return 1;
 493
 494         skb2 = skb_clone(skb, GFP_ATOMIC);
 495
 496         if (!skb2)
 497                 return 1;
 498
 499         skb_dst_drop(skb2);
 500         skb_pull(skb2, iph->ihl * 4);
 501         skb_reset_network_header(skb2);
 502
 503         rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
 504
 505         if (rt && rt->dst.dev)
 506                 skb2->dev = rt->dst.dev;
 507
 508         icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
 509
 510         if (rt)
 511                 ip6_rt_put(rt);
 512
 513         kfree_skb(skb2);
 514
 515         return 0;
 516 }
 517
 518 static int ipip6_err(struct sk_buff *skb, u32 info)
 519 {
 520         const struct iphdr *iph = (const struct iphdr *)skb->data;
 521         const int type = icmp_hdr(skb)->type;
 522         const int code = icmp_hdr(skb)->code;
 523         struct ip_tunnel *t;
 524         int err;
 525
 526         switch (type) {
 527         default:
 528         case ICMP_PARAMETERPROB:
 529                 return 0;
 530
 531         case ICMP_DEST_UNREACH:
 532                 switch (code) {
 533                 case ICMP_SR_FAILED:
 534                         /* Impossible event. */
 535                         return 0;
 536                 default:
 537                         /* All others are translated to HOST_UNREACH.
 538                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 539                            I believe they are just ether pollution. --ANK
 540                          */
 541                         break;
 542                 }
 543                 break;
 544         case ICMP_TIME_EXCEEDED:
 545                 if (code != ICMP_EXC_TTL)
 546                         return 0;
 547                 break;
 548         case ICMP_REDIRECT:
 549                 break;
 550         }
 551
 552         err = -ENOENT;
 553
 554         t = ipip6_tunnel_lookup(dev_net(skb->dev),
 555                                 skb->dev,
 556                                 iph->daddr,
 557                                 iph->saddr);
 558         if (t == NULL)
 559                 goto out;
 560
 561         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 562                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 563                                  t->dev->ifindex, 0, IPPROTO_IPV6, 0);
 564                 err = 0;
 565                 goto out;
 566         }
 567         if (type == ICMP_REDIRECT) {
 568                 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0,
 569                               IPPROTO_IPV6, 0);
 570                 err = 0;
 571                 goto out;
 572         }
 573
 574         if (t->parms.iph.daddr == 0)
 575                 goto out;
 576
 577         err = 0;
 578         if (!ipip6_err_gen_icmpv6_unreach(skb))
 579                 goto out;
 580
 581         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 582                 goto out;
 583
 584         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 585                 t->err_count++;
 586         else
 587                 t->err_count = 1;
 588         t->err_time = jiffies;
 589 out:
 590         return err;
 591 }
 592
 593 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 594                                   const struct in6_addr *v6addr)
 595 {
 596         __be32 v4embed = 0;
 597         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 598                 return true;
 599         return false;
 600 }
 601
 602 /* Checks if an address matches an address on the tunnel interface.
 603  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 604  * Long story:
 605  * This function is called after we considered the packet as spoofed
 606  * in is_spoofed_6rd.
 607  * We may have a router that is doing NAT for proto 41 packets
 608  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 609  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 610  * function will return true, dropping the packet.
 611  * But, we can still check if is spoofed against the IP
 612  * addresses associated with the interface.
 613  */
 614 static bool only_dnatted(const struct ip_tunnel *tunnel,
 615         const struct in6_addr *v6dst)
 616 {
 617         int prefix_len;
 618
 619 #ifdef CONFIG_IPV6_SIT_6RD
 620         prefix_len = tunnel->ip6rd.prefixlen + 32
 621                 - tunnel->ip6rd.relay_prefixlen;
 622 #else
 623         prefix_len = 48;
 624 #endif
 625         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 626 }
 627
 628 /* Returns true if a packet is spoofed */
 629 static bool packet_is_spoofed(struct sk_buff *skb,
 630                               const struct iphdr *iph,
 631                               struct ip_tunnel *tunnel)
 632 {
 633         const struct ipv6hdr *ipv6h;
 634
 635         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 636                 if (!isatap_chksrc(skb, iph, tunnel))
 637                         return true;
 638
 639                 return false;
 640         }
 641
 642         if (tunnel->dev->flags & IFF_POINTOPOINT)
 643                 return false;
 644
 645         ipv6h = ipv6_hdr(skb);
 646
 647         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 648                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 649                                      &iph->saddr, &ipv6h->saddr,
 650                                      &iph->daddr, &ipv6h->daddr);
 651                 return true;
 652         }
 653
 654         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 655                 return false;
 656
 657         if (only_dnatted(tunnel, &ipv6h->daddr))
 658                 return false;
 659
 660         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 661                              &iph->saddr, &ipv6h->saddr,
 662                              &iph->daddr, &ipv6h->daddr);
 663         return true;
 664 }
 665
 666 static int ipip6_rcv(struct sk_buff *skb)
 667 {
 668         const struct iphdr *iph = ip_hdr(skb);
 669         struct ip_tunnel *tunnel;
 670         int err;
 671
 672         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 673                                      iph->saddr, iph->daddr);
 674         if (tunnel != NULL) {
 675                 struct pcpu_sw_netstats *tstats;
 676
 677                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 678                     tunnel->parms.iph.protocol != 0)
 679                         goto out;
 680
 681                 skb->mac_header = skb->network_header;
 682                 skb_reset_network_header(skb);
 683                 IPCB(skb)->flags = 0;
 684                 skb->protocol = htons(ETH_P_IPV6);
 685
 686                 if (packet_is_spoofed(skb, iph, tunnel)) {
 687                         tunnel->dev->stats.rx_errors++;
 688                         goto out;
 689                 }
 690
 691                 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 692
 693                 err = IP_ECN_decapsulate(iph, skb);
 694                 if (unlikely(err)) {
 695                         if (log_ecn_error)
 696                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 697                                                      &iph->saddr, iph->tos);
 698                         if (err > 1) {
 699                                 ++tunnel->dev->stats.rx_frame_errors;
 700                                 ++tunnel->dev->stats.rx_errors;
 701                                 goto out;
 702                         }
 703                 }
 704
 705                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 706                 u64_stats_update_begin(&tstats->syncp);
 707                 tstats->rx_packets++;
 708                 tstats->rx_bytes += skb->len;
 709                 u64_stats_update_end(&tstats->syncp);
 710
 711                 netif_rx(skb);
 712
 713                 return 0;
 714         }
 715
 716         /* no tunnel matched,  let upstream know, ipsec may handle it */
 717         return 1;
 718 out:
 719         kfree_skb(skb);
 720         return 0;
 721 }
 722
 723 static const struct tnl_ptk_info tpi = {
 724         /* no tunnel info required for ipip. */
 725         .proto = htons(ETH_P_IP),
 726 };
 727
 728 static int ipip_rcv(struct sk_buff *skb)
 729 {
 730         const struct iphdr *iph;
 731         struct ip_tunnel *tunnel;
 732
 733         iph = ip_hdr(skb);
 734         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 735                                      iph->saddr, iph->daddr);
 736         if (tunnel != NULL) {
 737                 if (tunnel->parms.iph.protocol != IPPROTO_IPIP &&
 738                     tunnel->parms.iph.protocol != 0)
 739                         goto drop;
 740
 741                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 742                         goto drop;
 743                 if (iptunnel_pull_header(skb, 0, tpi.proto))
 744                         goto drop;
 745                 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error);
 746         }
 747
 748         return 1;
 749
 750 drop:
 751         kfree_skb(skb);
 752         return 0;
 753 }
 754
 755 /*
 756  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 757  * stores the embedded IPv4 address in v4dst and returns true.
 758  */
 759 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 760                       __be32 *v4dst)
 761 {
 762 #ifdef CONFIG_IPV6_SIT_6RD
 763         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 764                               tunnel->ip6rd.prefixlen)) {
 765                 unsigned int pbw0, pbi0;
 766                 int pbi1;
 767                 u32 d;
 768
 769                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 770                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 771
 772                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 773                     tunnel->ip6rd.relay_prefixlen;
 774
 775                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 776                 if (pbi1 > 0)
 777                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 778                              (32 - pbi1);
 779
 780                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 781                 return true;
 782         }
 783 #else
 784         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 785                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 786                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 787                 return true;
 788         }
 789 #endif
 790         return false;
 791 }
 792
 793 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 794                              const struct in6_addr *v6dst)
 795 {
 796         __be32 dst = 0;
 797         check_6rd(tunnel, v6dst, &dst);
 798         return dst;
 799 }
 800
 801 /*
 802  *      This function assumes it is being called from dev_queue_xmit()
 803  *      and that skb is filled properly by that function.
 804  */
 805
 806 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 807                                      struct net_device *dev)
 808 {
 809         struct ip_tunnel *tunnel = netdev_priv(dev);
 810         const struct iphdr  *tiph = &tunnel->parms.iph;
 811         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 812         u8     tos = tunnel->parms.iph.tos;
 813         __be16 df = tiph->frag_off;
 814         struct rtable *rt;                      /* Route to the other host */
 815         struct net_device *tdev;                /* Device to other host */
 816         unsigned int max_headroom;              /* The extra header space needed */
 817         __be32 dst = tiph->daddr;
 818         struct flowi4 fl4;
 819         int    mtu;
 820         const struct in6_addr *addr6;
 821         int addr_type;
 822         u8 ttl;
 823         int err;
 824
 825         if (skb->protocol != htons(ETH_P_IPV6))
 826                 goto tx_error;
 827
 828         if (tos == 1)
 829                 tos = ipv6_get_dsfield(iph6);
 830
 831         /* ISATAP (RFC4214) - must come before 6to4 */
 832         if (dev->priv_flags & IFF_ISATAP) {
 833                 struct neighbour *neigh = NULL;
 834                 bool do_tx_error = false;
 835
 836                 if (skb_dst(skb))
 837                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 838
 839                 if (neigh == NULL) {
 840                         net_dbg_ratelimited("nexthop == NULL\n");
 841                         goto tx_error;
 842                 }
 843
 844                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 845                 addr_type = ipv6_addr_type(addr6);
 846
 847                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 848                      ipv6_addr_is_isatap(addr6))
 849                         dst = addr6->s6_addr32[3];
 850                 else
 851                         do_tx_error = true;
 852
 853                 neigh_release(neigh);
 854                 if (do_tx_error)
 855                         goto tx_error;
 856         }
 857
 858         if (!dst)
 859                 dst = try_6rd(tunnel, &iph6->daddr);
 860
 861         if (!dst) {
 862                 struct neighbour *neigh = NULL;
 863                 bool do_tx_error = false;
 864
 865                 if (skb_dst(skb))
 866                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 867
 868                 if (neigh == NULL) {
 869                         net_dbg_ratelimited("nexthop == NULL\n");
 870                         goto tx_error;
 871                 }
 872
 873                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 874                 addr_type = ipv6_addr_type(addr6);
 875
 876                 if (addr_type == IPV6_ADDR_ANY) {
 877                         addr6 = &ipv6_hdr(skb)->daddr;
 878                         addr_type = ipv6_addr_type(addr6);
 879                 }
 880
 881                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 882                         dst = addr6->s6_addr32[3];
 883                 else
 884                         do_tx_error = true;
 885
 886                 neigh_release(neigh);
 887                 if (do_tx_error)
 888                         goto tx_error;
 889         }
 890
 891         rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 892                                    dst, tiph->saddr,
 893                                    0, 0,
 894                                    IPPROTO_IPV6, RT_TOS(tos),
 895                                    tunnel->parms.link);
 896         if (IS_ERR(rt)) {
 897                 dev->stats.tx_carrier_errors++;
 898                 goto tx_error_icmp;
 899         }
 900         if (rt->rt_type != RTN_UNICAST) {
 901                 ip_rt_put(rt);
 902                 dev->stats.tx_carrier_errors++;
 903                 goto tx_error_icmp;
 904         }
 905         tdev = rt->dst.dev;
 906
 907         if (tdev == dev) {
 908                 ip_rt_put(rt);
 909                 dev->stats.collisions++;
 910                 goto tx_error;
 911         }
 912
 913         if (df) {
 914                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 915
 916                 if (mtu < 68) {
 917                         dev->stats.collisions++;
 918                         ip_rt_put(rt);
 919                         goto tx_error;
 920                 }
 921
 922                 if (mtu < IPV6_MIN_MTU) {
 923                         mtu = IPV6_MIN_MTU;
 924                         df = 0;
 925                 }
 926
 927                 if (tunnel->parms.iph.daddr && skb_dst(skb))
 928                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 929
 930                 if (skb->len > mtu && !skb_is_gso(skb)) {
 931                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 932                         ip_rt_put(rt);
 933                         goto tx_error;
 934                 }
 935         }
 936
 937         if (tunnel->err_count > 0) {
 938                 if (time_before(jiffies,
 939                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 940                         tunnel->err_count--;
 941                         dst_link_failure(skb);
 942                 } else
 943                         tunnel->err_count = 0;
 944         }
 945
 946         /*
 947          * Okay, now see if we can stuff it in the buffer as-is.
 948          */
 949         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 950
 951         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 952             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 953                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 954                 if (!new_skb) {
 955                         ip_rt_put(rt);
 956                         dev->stats.tx_dropped++;
 957                         kfree_skb(skb);
 958                         return NETDEV_TX_OK;
 959                 }
 960                 if (skb->sk)
 961                         skb_set_owner_w(new_skb, skb->sk);
 962                 dev_kfree_skb(skb);
 963                 skb = new_skb;
 964                 iph6 = ipv6_hdr(skb);
 965         }
 966         ttl = tiph->ttl;
 967         if (ttl == 0)
 968                 ttl = iph6->hop_limit;
 969         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 970
 971         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
 972         if (IS_ERR(skb)) {
 973                 ip_rt_put(rt);
 974                 goto out;
 975         }
 976
 977         err = iptunnel_xmit(skb->sk, rt, skb, fl4.saddr, fl4.daddr,
 978                             IPPROTO_IPV6, tos, ttl, df,
 979                             !net_eq(tunnel->net, dev_net(dev)));
 980         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 981         return NETDEV_TX_OK;
 982
 983 tx_error_icmp:
 984         dst_link_failure(skb);
 985 tx_error:
 986         kfree_skb(skb);
 987 out:
 988         dev->stats.tx_errors++;
 989         return NETDEV_TX_OK;
 990 }
 991
 992 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 993 {
 994         struct ip_tunnel *tunnel = netdev_priv(dev);
 995         const struct iphdr  *tiph = &tunnel->parms.iph;
 996
 997         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
 998         if (IS_ERR(skb))
 999                 goto out;
1000
1001         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1002         return NETDEV_TX_OK;
1003 out:
1004         dev->stats.tx_errors++;
1005         return NETDEV_TX_OK;
1006 }
1007
1008 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1009                                    struct net_device *dev)
1010 {
1011         switch (skb->protocol) {
1012         case htons(ETH_P_IP):
1013                 ipip_tunnel_xmit(skb, dev);
1014                 break;
1015         case htons(ETH_P_IPV6):
1016                 ipip6_tunnel_xmit(skb, dev);
1017                 break;
1018         default:
1019                 goto tx_err;
1020         }
1021
1022         return NETDEV_TX_OK;
1023
1024 tx_err:
1025         dev->stats.tx_errors++;
1026         kfree_skb(skb);
1027         return NETDEV_TX_OK;
1028
1029 }
1030
1031 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1032 {
1033         struct net_device *tdev = NULL;
1034         struct ip_tunnel *tunnel;
1035         const struct iphdr *iph;
1036         struct flowi4 fl4;
1037
1038         tunnel = netdev_priv(dev);
1039         iph = &tunnel->parms.iph;
1040
1041         if (iph->daddr) {
1042                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1043                                                           NULL,
1044                                                           iph->daddr, iph->saddr,
1045                                                           0, 0,
1046                                                           IPPROTO_IPV6,
1047                                                           RT_TOS(iph->tos),
1048                                                           tunnel->parms.link);
1049
1050                 if (!IS_ERR(rt)) {
1051                         tdev = rt->dst.dev;
1052                         ip_rt_put(rt);
1053                 }
1054                 dev->flags |= IFF_POINTOPOINT;
1055         }
1056
1057         if (!tdev && tunnel->parms.link)
1058                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1059
1060         if (tdev) {
1061                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1062                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
1063                 if (dev->mtu < IPV6_MIN_MTU)
1064                         dev->mtu = IPV6_MIN_MTU;
1065         }
1066         dev->iflink = tunnel->parms.link;
1067 }
1068
1069 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1070 {
1071         struct net *net = t->net;
1072         struct sit_net *sitn = net_generic(net, sit_net_id);
1073
1074         ipip6_tunnel_unlink(sitn, t);
1075         synchronize_net();
1076         t->parms.iph.saddr = p->iph.saddr;
1077         t->parms.iph.daddr = p->iph.daddr;
1078         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1079         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1080         ipip6_tunnel_link(sitn, t);
1081         t->parms.iph.ttl = p->iph.ttl;
1082         t->parms.iph.tos = p->iph.tos;
1083         if (t->parms.link != p->link) {
1084                 t->parms.link = p->link;
1085                 ipip6_tunnel_bind_dev(t->dev);
1086         }
1087         ip_tunnel_dst_reset_all(t);
1088         netdev_state_change(t->dev);
1089 }
1090
1091 #ifdef CONFIG_IPV6_SIT_6RD
1092 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1093                                    struct ip_tunnel_6rd *ip6rd)
1094 {
1095         struct in6_addr prefix;
1096         __be32 relay_prefix;
1097
1098         if (ip6rd->relay_prefixlen > 32 ||
1099             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1100                 return -EINVAL;
1101
1102         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1103         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1104                 return -EINVAL;
1105         if (ip6rd->relay_prefixlen)
1106                 relay_prefix = ip6rd->relay_prefix &
1107                                htonl(0xffffffffUL <<
1108                                      (32 - ip6rd->relay_prefixlen));
1109         else
1110                 relay_prefix = 0;
1111         if (relay_prefix != ip6rd->relay_prefix)
1112                 return -EINVAL;
1113
1114         t->ip6rd.prefix = prefix;
1115         t->ip6rd.relay_prefix = relay_prefix;
1116         t->ip6rd.prefixlen = ip6rd->prefixlen;
1117         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1118         ip_tunnel_dst_reset_all(t);
1119         netdev_state_change(t->dev);
1120         return 0;
1121 }
1122 #endif
1123
1124 static int
1125 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1126 {
1127         int err = 0;
1128         struct ip_tunnel_parm p;
1129         struct ip_tunnel_prl prl;
1130         struct ip_tunnel *t = netdev_priv(dev);
1131         struct net *net = t->net;
1132         struct sit_net *sitn = net_generic(net, sit_net_id);
1133 #ifdef CONFIG_IPV6_SIT_6RD
1134         struct ip_tunnel_6rd ip6rd;
1135 #endif
1136
1137         switch (cmd) {
1138         case SIOCGETTUNNEL:
1139 #ifdef CONFIG_IPV6_SIT_6RD
1140         case SIOCGET6RD:
1141 #endif
1142                 if (dev == sitn->fb_tunnel_dev) {
1143                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1144                                 err = -EFAULT;
1145                                 break;
1146                         }
1147                         t = ipip6_tunnel_locate(net, &p, 0);
1148                         if (t == NULL)
1149                                 t = netdev_priv(dev);
1150                 }
1151
1152                 err = -EFAULT;
1153                 if (cmd == SIOCGETTUNNEL) {
1154                         memcpy(&p, &t->parms, sizeof(p));
1155                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1156                                          sizeof(p)))
1157                                 goto done;
1158 #ifdef CONFIG_IPV6_SIT_6RD
1159                 } else {
1160                         ip6rd.prefix = t->ip6rd.prefix;
1161                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1162                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1163                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1164                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1165                                          sizeof(ip6rd)))
1166                                 goto done;
1167 #endif
1168                 }
1169                 err = 0;
1170                 break;
1171
1172         case SIOCADDTUNNEL:
1173         case SIOCCHGTUNNEL:
1174                 err = -EPERM;
1175                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1176                         goto done;
1177
1178                 err = -EFAULT;
1179                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1180                         goto done;
1181
1182                 err = -EINVAL;
1183                 if (p.iph.protocol != IPPROTO_IPV6 &&
1184                     p.iph.protocol != IPPROTO_IPIP &&
1185                     p.iph.protocol != 0)
1186                         goto done;
1187                 if (p.iph.version != 4 ||
1188                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1189                         goto done;
1190                 if (p.iph.ttl)
1191                         p.iph.frag_off |= htons(IP_DF);
1192
1193                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1194
1195                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1196                         if (t != NULL) {
1197                                 if (t->dev != dev) {
1198                                         err = -EEXIST;
1199                                         break;
1200                                 }
1201                         } else {
1202                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1203                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1204                                         err = -EINVAL;
1205                                         break;
1206                                 }
1207                                 t = netdev_priv(dev);
1208                         }
1209
1210                         ipip6_tunnel_update(t, &p);
1211                 }
1212
1213                 if (t) {
1214                         err = 0;
1215                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1216                                 err = -EFAULT;
1217                 } else
1218                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1219                 break;
1220
1221         case SIOCDELTUNNEL:
1222                 err = -EPERM;
1223                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1224                         goto done;
1225
1226                 if (dev == sitn->fb_tunnel_dev) {
1227                         err = -EFAULT;
1228                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1229                                 goto done;
1230                         err = -ENOENT;
1231                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1232                                 goto done;
1233                         err = -EPERM;
1234                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1235                                 goto done;
1236                         dev = t->dev;
1237                 }
1238                 unregister_netdevice(dev);
1239                 err = 0;
1240                 break;
1241
1242         case SIOCGETPRL:
1243                 err = -EINVAL;
1244                 if (dev == sitn->fb_tunnel_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
1262                 switch (cmd) {
1263                 case SIOCDELPRL:
1264                         err = ipip6_tunnel_del_prl(t, &prl);
1265                         break;
1266                 case SIOCADDPRL:
1267                 case SIOCCHGPRL:
1268                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1269                         break;
1270                 }
1271                 ip_tunnel_dst_reset_all(t);
1272                 netdev_state_change(dev);
1273                 break;
1274
1275 #ifdef CONFIG_IPV6_SIT_6RD
1276         case SIOCADD6RD:
1277         case SIOCCHG6RD:
1278         case SIOCDEL6RD:
1279                 err = -EPERM;
1280                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1281                         goto done;
1282
1283                 err = -EFAULT;
1284                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1285                                    sizeof(ip6rd)))
1286                         goto done;
1287
1288                 if (cmd != SIOCDEL6RD) {
1289                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1290                         if (err < 0)
1291                                 goto done;
1292                 } else
1293                         ipip6_tunnel_clone_6rd(dev, sitn);
1294
1295                 err = 0;
1296                 break;
1297 #endif
1298
1299         default:
1300                 err = -EINVAL;
1301         }
1302
1303 done:
1304         return err;
1305 }
1306
1307 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1308 {
1309         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1310                 return -EINVAL;
1311         dev->mtu = new_mtu;
1312         return 0;
1313 }
1314
1315 static const struct net_device_ops ipip6_netdev_ops = {
1316         .ndo_uninit     = ipip6_tunnel_uninit,
1317         .ndo_start_xmit = sit_tunnel_xmit,
1318         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1319         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1320         .ndo_get_stats64 = ip_tunnel_get_stats64,
1321 };
1322
1323 static void ipip6_dev_free(struct net_device *dev)
1324 {
1325         struct ip_tunnel *tunnel = netdev_priv(dev);
1326
1327         free_percpu(tunnel->dst_cache);
1328         free_percpu(dev->tstats);
1329         free_netdev(dev);
1330 }
1331
1332 #define SIT_FEATURES (NETIF_F_SG           | \
1333                       NETIF_F_FRAGLIST     | \
1334                       NETIF_F_HIGHDMA      | \
1335                       NETIF_F_GSO_SOFTWARE | \
1336                       NETIF_F_HW_CSUM)
1337
1338 static void ipip6_tunnel_setup(struct net_device *dev)
1339 {
1340         dev->netdev_ops         = &ipip6_netdev_ops;
1341         dev->destructor         = ipip6_dev_free;
1342
1343         dev->type               = ARPHRD_SIT;
1344         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1345         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1346         dev->flags              = IFF_NOARP;
1347         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1348         dev->iflink             = 0;
1349         dev->addr_len           = 4;
1350         dev->features           |= NETIF_F_LLTX;
1351         dev->features           |= SIT_FEATURES;
1352         dev->hw_features        |= SIT_FEATURES;
1353 }
1354
1355 static int ipip6_tunnel_init(struct net_device *dev)
1356 {
1357         struct ip_tunnel *tunnel = netdev_priv(dev);
1358
1359         tunnel->dev = dev;
1360         tunnel->net = dev_net(dev);
1361
1362         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1363         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1364
1365         ipip6_tunnel_bind_dev(dev);
1366         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1367         if (!dev->tstats)
1368                 return -ENOMEM;
1369
1370         tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1371         if (!tunnel->dst_cache) {
1372                 free_percpu(dev->tstats);
1373                 return -ENOMEM;
1374         }
1375
1376         return 0;
1377 }
1378
1379 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1380 {
1381         struct ip_tunnel *tunnel = netdev_priv(dev);
1382         struct iphdr *iph = &tunnel->parms.iph;
1383         struct net *net = dev_net(dev);
1384         struct sit_net *sitn = net_generic(net, sit_net_id);
1385
1386         tunnel->dev = dev;
1387         tunnel->net = dev_net(dev);
1388         strcpy(tunnel->parms.name, dev->name);
1389
1390         iph->version            = 4;
1391         iph->protocol           = IPPROTO_IPV6;
1392         iph->ihl                = 5;
1393         iph->ttl                = 64;
1394
1395         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1396         if (!dev->tstats)
1397                 return -ENOMEM;
1398
1399         tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1400         if (!tunnel->dst_cache) {
1401                 free_percpu(dev->tstats);
1402                 return -ENOMEM;
1403         }
1404
1405         dev_hold(dev);
1406         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1407         return 0;
1408 }
1409
1410 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1411 {
1412         u8 proto;
1413
1414         if (!data || !data[IFLA_IPTUN_PROTO])
1415                 return 0;
1416
1417         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1418         if (proto != IPPROTO_IPV6 &&
1419             proto != IPPROTO_IPIP &&
1420             proto != 0)
1421                 return -EINVAL;
1422
1423         return 0;
1424 }
1425
1426 static void ipip6_netlink_parms(struct nlattr *data[],
1427                                 struct ip_tunnel_parm *parms)
1428 {
1429         memset(parms, 0, sizeof(*parms));
1430
1431         parms->iph.version = 4;
1432         parms->iph.protocol = IPPROTO_IPV6;
1433         parms->iph.ihl = 5;
1434         parms->iph.ttl = 64;
1435
1436         if (!data)
1437                 return;
1438
1439         if (data[IFLA_IPTUN_LINK])
1440                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1441
1442         if (data[IFLA_IPTUN_LOCAL])
1443                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1444
1445         if (data[IFLA_IPTUN_REMOTE])
1446                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1447
1448         if (data[IFLA_IPTUN_TTL]) {
1449                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1450                 if (parms->iph.ttl)
1451                         parms->iph.frag_off = htons(IP_DF);
1452         }
1453
1454         if (data[IFLA_IPTUN_TOS])
1455                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1456
1457         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1458                 parms->iph.frag_off = htons(IP_DF);
1459
1460         if (data[IFLA_IPTUN_FLAGS])
1461                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1462
1463         if (data[IFLA_IPTUN_PROTO])
1464                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1465
1466 }
1467
1468 #ifdef CONFIG_IPV6_SIT_6RD
1469 /* This function returns true when 6RD attributes are present in the nl msg */
1470 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1471                                     struct ip_tunnel_6rd *ip6rd)
1472 {
1473         bool ret = false;
1474         memset(ip6rd, 0, sizeof(*ip6rd));
1475
1476         if (!data)
1477                 return ret;
1478
1479         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1480                 ret = true;
1481                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1482                            sizeof(struct in6_addr));
1483         }
1484
1485         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1486                 ret = true;
1487                 ip6rd->relay_prefix =
1488                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1489         }
1490
1491         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1492                 ret = true;
1493                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1494         }
1495
1496         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1497                 ret = true;
1498                 ip6rd->relay_prefixlen =
1499                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1500         }
1501
1502         return ret;
1503 }
1504 #endif
1505
1506 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1507                          struct nlattr *tb[], struct nlattr *data[])
1508 {
1509         struct net *net = dev_net(dev);
1510         struct ip_tunnel *nt;
1511 #ifdef CONFIG_IPV6_SIT_6RD
1512         struct ip_tunnel_6rd ip6rd;
1513 #endif
1514         int err;
1515
1516         nt = netdev_priv(dev);
1517         ipip6_netlink_parms(data, &nt->parms);
1518
1519         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1520                 return -EEXIST;
1521
1522         err = ipip6_tunnel_create(dev);
1523         if (err < 0)
1524                 return err;
1525
1526 #ifdef CONFIG_IPV6_SIT_6RD
1527         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1528                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1529 #endif
1530
1531         return err;
1532 }
1533
1534 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1535                           struct nlattr *data[])
1536 {
1537         struct ip_tunnel *t = netdev_priv(dev);
1538         struct ip_tunnel_parm p;
1539         struct net *net = t->net;
1540         struct sit_net *sitn = net_generic(net, sit_net_id);
1541 #ifdef CONFIG_IPV6_SIT_6RD
1542         struct ip_tunnel_6rd ip6rd;
1543 #endif
1544
1545         if (dev == sitn->fb_tunnel_dev)
1546                 return -EINVAL;
1547
1548         ipip6_netlink_parms(data, &p);
1549
1550         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1551             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1552                 return -EINVAL;
1553
1554         t = ipip6_tunnel_locate(net, &p, 0);
1555
1556         if (t) {
1557                 if (t->dev != dev)
1558                         return -EEXIST;
1559         } else
1560                 t = netdev_priv(dev);
1561
1562         ipip6_tunnel_update(t, &p);
1563
1564 #ifdef CONFIG_IPV6_SIT_6RD
1565         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1566                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1567 #endif
1568
1569         return 0;
1570 }
1571
1572 static size_t ipip6_get_size(const struct net_device *dev)
1573 {
1574         return
1575                 /* IFLA_IPTUN_LINK */
1576                 nla_total_size(4) +
1577                 /* IFLA_IPTUN_LOCAL */
1578                 nla_total_size(4) +
1579                 /* IFLA_IPTUN_REMOTE */
1580                 nla_total_size(4) +
1581                 /* IFLA_IPTUN_TTL */
1582                 nla_total_size(1) +
1583                 /* IFLA_IPTUN_TOS */
1584                 nla_total_size(1) +
1585                 /* IFLA_IPTUN_PMTUDISC */
1586                 nla_total_size(1) +
1587                 /* IFLA_IPTUN_FLAGS */
1588                 nla_total_size(2) +
1589                 /* IFLA_IPTUN_PROTO */
1590                 nla_total_size(1) +
1591 #ifdef CONFIG_IPV6_SIT_6RD
1592                 /* IFLA_IPTUN_6RD_PREFIX */
1593                 nla_total_size(sizeof(struct in6_addr)) +
1594                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1595                 nla_total_size(4) +
1596                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1597                 nla_total_size(2) +
1598                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1599                 nla_total_size(2) +
1600 #endif
1601                 0;
1602 }
1603
1604 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1605 {
1606         struct ip_tunnel *tunnel = netdev_priv(dev);
1607         struct ip_tunnel_parm *parm = &tunnel->parms;
1608
1609         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1610             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1611             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1612             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1613             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1614             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1615                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1616             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1617             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1618                 goto nla_put_failure;
1619
1620 #ifdef CONFIG_IPV6_SIT_6RD
1621         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1622                     &tunnel->ip6rd.prefix) ||
1623             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1624                          tunnel->ip6rd.relay_prefix) ||
1625             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1626                         tunnel->ip6rd.prefixlen) ||
1627             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1628                         tunnel->ip6rd.relay_prefixlen))
1629                 goto nla_put_failure;
1630 #endif
1631
1632         return 0;
1633
1634 nla_put_failure:
1635         return -EMSGSIZE;
1636 }
1637
1638 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1639         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1640         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1641         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1642         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1643         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1644         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1645         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1646         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1647 #ifdef CONFIG_IPV6_SIT_6RD
1648         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1649         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1650         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1651         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1652 #endif
1653 };
1654
1655 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1656 {
1657         struct net *net = dev_net(dev);
1658         struct sit_net *sitn = net_generic(net, sit_net_id);
1659
1660         if (dev != sitn->fb_tunnel_dev)
1661                 unregister_netdevice_queue(dev, head);
1662 }
1663
1664 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1665         .kind           = "sit",
1666         .maxtype        = IFLA_IPTUN_MAX,
1667         .policy         = ipip6_policy,
1668         .priv_size      = sizeof(struct ip_tunnel),
1669         .setup          = ipip6_tunnel_setup,
1670         .validate       = ipip6_validate,
1671         .newlink        = ipip6_newlink,
1672         .changelink     = ipip6_changelink,
1673         .get_size       = ipip6_get_size,
1674         .fill_info      = ipip6_fill_info,
1675         .dellink        = ipip6_dellink,
1676 };
1677
1678 static struct xfrm_tunnel sit_handler __read_mostly = {
1679         .handler        =       ipip6_rcv,
1680         .err_handler    =       ipip6_err,
1681         .priority       =       1,
1682 };
1683
1684 static struct xfrm_tunnel ipip_handler __read_mostly = {
1685         .handler        =       ipip_rcv,
1686         .err_handler    =       ipip6_err,
1687         .priority       =       2,
1688 };
1689
1690 static void __net_exit sit_destroy_tunnels(struct net *net,
1691                                            struct list_head *head)
1692 {
1693         struct sit_net *sitn = net_generic(net, sit_net_id);
1694         struct net_device *dev, *aux;
1695         int prio;
1696
1697         for_each_netdev_safe(net, dev, aux)
1698                 if (dev->rtnl_link_ops == &sit_link_ops)
1699                         unregister_netdevice_queue(dev, head);
1700
1701         for (prio = 1; prio < 4; prio++) {
1702                 int h;
1703                 for (h = 0; h < HASH_SIZE; h++) {
1704                         struct ip_tunnel *t;
1705
1706                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1707                         while (t != NULL) {
1708                                 /* If dev is in the same netns, it has already
1709                                  * been added to the list by the previous loop.
1710                                  */
1711                                 if (!net_eq(dev_net(t->dev), net))
1712                                         unregister_netdevice_queue(t->dev,
1713                                                                    head);
1714                                 t = rtnl_dereference(t->next);
1715                         }
1716                 }
1717         }
1718 }
1719
1720 static int __net_init sit_init_net(struct net *net)
1721 {
1722         struct sit_net *sitn = net_generic(net, sit_net_id);
1723         struct ip_tunnel *t;
1724         int err;
1725
1726         sitn->tunnels[0] = sitn->tunnels_wc;
1727         sitn->tunnels[1] = sitn->tunnels_l;
1728         sitn->tunnels[2] = sitn->tunnels_r;
1729         sitn->tunnels[3] = sitn->tunnels_r_l;
1730
1731         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1732                                            ipip6_tunnel_setup);
1733         if (!sitn->fb_tunnel_dev) {
1734                 err = -ENOMEM;
1735                 goto err_alloc_dev;
1736         }
1737         dev_net_set(sitn->fb_tunnel_dev, net);
1738         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1739         /* FB netdevice is special: we have one, and only one per netns.
1740          * Allowing to move it to another netns is clearly unsafe.
1741          */
1742         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1743
1744         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1745         if (err)
1746                 goto err_dev_free;
1747
1748         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1749
1750         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1751                 goto err_reg_dev;
1752
1753         t = netdev_priv(sitn->fb_tunnel_dev);
1754
1755         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1756         return 0;
1757
1758 err_reg_dev:
1759         dev_put(sitn->fb_tunnel_dev);
1760 err_dev_free:
1761         ipip6_dev_free(sitn->fb_tunnel_dev);
1762 err_alloc_dev:
1763         return err;
1764 }
1765
1766 static void __net_exit sit_exit_net(struct net *net)
1767 {
1768         LIST_HEAD(list);
1769
1770         rtnl_lock();
1771         sit_destroy_tunnels(net, &list);
1772         unregister_netdevice_many(&list);
1773         rtnl_unlock();
1774 }
1775
1776 static struct pernet_operations sit_net_ops = {
1777         .init = sit_init_net,
1778         .exit = sit_exit_net,
1779         .id   = &sit_net_id,
1780         .size = sizeof(struct sit_net),
1781 };
1782
1783 static void __exit sit_cleanup(void)
1784 {
1785         rtnl_link_unregister(&sit_link_ops);
1786         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1787         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1788
1789         unregister_pernet_device(&sit_net_ops);
1790         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1791 }
1792
1793 static int __init sit_init(void)
1794 {
1795         int err;
1796
1797         pr_info("IPv6 over IPv4 tunneling driver\n");
1798
1799         err = register_pernet_device(&sit_net_ops);
1800         if (err < 0)
1801                 return err;
1802         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1803         if (err < 0) {
1804                 pr_info("%s: can't register ip6ip4\n", __func__);
1805                 goto xfrm_tunnel_failed;
1806         }
1807         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1808         if (err < 0) {
1809                 pr_info("%s: can't register ip4ip4\n", __func__);
1810                 goto xfrm_tunnel4_failed;
1811         }
1812         err = rtnl_link_register(&sit_link_ops);
1813         if (err < 0)
1814                 goto rtnl_link_failed;
1815
1816 out:
1817         return err;
1818
1819 rtnl_link_failed:
1820         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1821 xfrm_tunnel4_failed:
1822         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1823 xfrm_tunnel_failed:
1824         unregister_pernet_device(&sit_net_ops);
1825         goto out;
1826 }
1827
1828 module_init(sit_init);
1829 module_exit(sit_cleanup);
1830 MODULE_LICENSE("GPL");
1831 MODULE_ALIAS_NETDEV("sit0");