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->parms.link, 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->parms.link, 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(rt, skb, fl4.saddr, fl4.daddr, IPPROTO_IPV6, tos,
 978                             ttl, df, !net_eq(tunnel->net, dev_net(dev)));
 979         iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 980         return NETDEV_TX_OK;
 981
 982 tx_error_icmp:
 983         dst_link_failure(skb);
 984 tx_error:
 985         kfree_skb(skb);
 986 out:
 987         dev->stats.tx_errors++;
 988         return NETDEV_TX_OK;
 989 }
 990
 991 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 992 {
 993         struct ip_tunnel *tunnel = netdev_priv(dev);
 994         const struct iphdr  *tiph = &tunnel->parms.iph;
 995
 996         skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
 997         if (IS_ERR(skb))
 998                 goto out;
 999
1000         ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1001         return NETDEV_TX_OK;
1002 out:
1003         dev->stats.tx_errors++;
1004         return NETDEV_TX_OK;
1005 }
1006
1007 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1008                                    struct net_device *dev)
1009 {
1010         switch (skb->protocol) {
1011         case htons(ETH_P_IP):
1012                 ipip_tunnel_xmit(skb, dev);
1013                 break;
1014         case htons(ETH_P_IPV6):
1015                 ipip6_tunnel_xmit(skb, dev);
1016                 break;
1017         default:
1018                 goto tx_err;
1019         }
1020
1021         return NETDEV_TX_OK;
1022
1023 tx_err:
1024         dev->stats.tx_errors++;
1025         kfree_skb(skb);
1026         return NETDEV_TX_OK;
1027
1028 }
1029
1030 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1031 {
1032         struct net_device *tdev = NULL;
1033         struct ip_tunnel *tunnel;
1034         const struct iphdr *iph;
1035         struct flowi4 fl4;
1036
1037         tunnel = netdev_priv(dev);
1038         iph = &tunnel->parms.iph;
1039
1040         if (iph->daddr) {
1041                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1042                                                           NULL,
1043                                                           iph->daddr, iph->saddr,
1044                                                           0, 0,
1045                                                           IPPROTO_IPV6,
1046                                                           RT_TOS(iph->tos),
1047                                                           tunnel->parms.link);
1048
1049                 if (!IS_ERR(rt)) {
1050                         tdev = rt->dst.dev;
1051                         ip_rt_put(rt);
1052                 }
1053                 dev->flags |= IFF_POINTOPOINT;
1054         }
1055
1056         if (!tdev && tunnel->parms.link)
1057                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1058
1059         if (tdev) {
1060                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1061                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
1062                 if (dev->mtu < IPV6_MIN_MTU)
1063                         dev->mtu = IPV6_MIN_MTU;
1064         }
1065         dev->iflink = tunnel->parms.link;
1066 }
1067
1068 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1069 {
1070         struct net *net = t->net;
1071         struct sit_net *sitn = net_generic(net, sit_net_id);
1072
1073         ipip6_tunnel_unlink(sitn, t);
1074         synchronize_net();
1075         t->parms.iph.saddr = p->iph.saddr;
1076         t->parms.iph.daddr = p->iph.daddr;
1077         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1078         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1079         ipip6_tunnel_link(sitn, t);
1080         t->parms.iph.ttl = p->iph.ttl;
1081         t->parms.iph.tos = p->iph.tos;
1082         if (t->parms.link != p->link) {
1083                 t->parms.link = p->link;
1084                 ipip6_tunnel_bind_dev(t->dev);
1085         }
1086         ip_tunnel_dst_reset_all(t);
1087         netdev_state_change(t->dev);
1088 }
1089
1090 #ifdef CONFIG_IPV6_SIT_6RD
1091 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1092                                    struct ip_tunnel_6rd *ip6rd)
1093 {
1094         struct in6_addr prefix;
1095         __be32 relay_prefix;
1096
1097         if (ip6rd->relay_prefixlen > 32 ||
1098             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1099                 return -EINVAL;
1100
1101         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1102         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1103                 return -EINVAL;
1104         if (ip6rd->relay_prefixlen)
1105                 relay_prefix = ip6rd->relay_prefix &
1106                                htonl(0xffffffffUL <<
1107                                      (32 - ip6rd->relay_prefixlen));
1108         else
1109                 relay_prefix = 0;
1110         if (relay_prefix != ip6rd->relay_prefix)
1111                 return -EINVAL;
1112
1113         t->ip6rd.prefix = prefix;
1114         t->ip6rd.relay_prefix = relay_prefix;
1115         t->ip6rd.prefixlen = ip6rd->prefixlen;
1116         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1117         ip_tunnel_dst_reset_all(t);
1118         netdev_state_change(t->dev);
1119         return 0;
1120 }
1121 #endif
1122
1123 static int
1124 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1125 {
1126         int err = 0;
1127         struct ip_tunnel_parm p;
1128         struct ip_tunnel_prl prl;
1129         struct ip_tunnel *t;
1130         struct net *net = dev_net(dev);
1131         struct sit_net *sitn = net_generic(net, sit_net_id);
1132 #ifdef CONFIG_IPV6_SIT_6RD
1133         struct ip_tunnel_6rd ip6rd;
1134 #endif
1135
1136         switch (cmd) {
1137         case SIOCGETTUNNEL:
1138 #ifdef CONFIG_IPV6_SIT_6RD
1139         case SIOCGET6RD:
1140 #endif
1141                 t = NULL;
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                 }
1149                 if (t == NULL)
1150                         t = netdev_priv(dev);
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 = -ENOENT;
1247                 if (!(t = netdev_priv(dev)))
1248                         goto done;
1249                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1250                 break;
1251
1252         case SIOCADDPRL:
1253         case SIOCDELPRL:
1254         case SIOCCHGPRL:
1255                 err = -EPERM;
1256                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1257                         goto done;
1258                 err = -EINVAL;
1259                 if (dev == sitn->fb_tunnel_dev)
1260                         goto done;
1261                 err = -EFAULT;
1262                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1263                         goto done;
1264                 err = -ENOENT;
1265                 if (!(t = netdev_priv(dev)))
1266                         goto done;
1267
1268                 switch (cmd) {
1269                 case SIOCDELPRL:
1270                         err = ipip6_tunnel_del_prl(t, &prl);
1271                         break;
1272                 case SIOCADDPRL:
1273                 case SIOCCHGPRL:
1274                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1275                         break;
1276                 }
1277                 ip_tunnel_dst_reset_all(t);
1278                 netdev_state_change(dev);
1279                 break;
1280
1281 #ifdef CONFIG_IPV6_SIT_6RD
1282         case SIOCADD6RD:
1283         case SIOCCHG6RD:
1284         case SIOCDEL6RD:
1285                 err = -EPERM;
1286                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1287                         goto done;
1288
1289                 err = -EFAULT;
1290                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1291                                    sizeof(ip6rd)))
1292                         goto done;
1293
1294                 t = netdev_priv(dev);
1295
1296                 if (cmd != SIOCDEL6RD) {
1297                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1298                         if (err < 0)
1299                                 goto done;
1300                 } else
1301                         ipip6_tunnel_clone_6rd(dev, sitn);
1302
1303                 err = 0;
1304                 break;
1305 #endif
1306
1307         default:
1308                 err = -EINVAL;
1309         }
1310
1311 done:
1312         return err;
1313 }
1314
1315 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1316 {
1317         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1318                 return -EINVAL;
1319         dev->mtu = new_mtu;
1320         return 0;
1321 }
1322
1323 static const struct net_device_ops ipip6_netdev_ops = {
1324         .ndo_uninit     = ipip6_tunnel_uninit,
1325         .ndo_start_xmit = sit_tunnel_xmit,
1326         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1327         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1328         .ndo_get_stats64 = ip_tunnel_get_stats64,
1329 };
1330
1331 static void ipip6_dev_free(struct net_device *dev)
1332 {
1333         struct ip_tunnel *tunnel = netdev_priv(dev);
1334
1335         free_percpu(tunnel->dst_cache);
1336         free_percpu(dev->tstats);
1337         free_netdev(dev);
1338 }
1339
1340 #define SIT_FEATURES (NETIF_F_SG           | \
1341                       NETIF_F_FRAGLIST     | \
1342                       NETIF_F_HIGHDMA      | \
1343                       NETIF_F_GSO_SOFTWARE | \
1344                       NETIF_F_HW_CSUM)
1345
1346 static void ipip6_tunnel_setup(struct net_device *dev)
1347 {
1348         dev->netdev_ops         = &ipip6_netdev_ops;
1349         dev->destructor         = ipip6_dev_free;
1350
1351         dev->type               = ARPHRD_SIT;
1352         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1353         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1354         dev->flags              = IFF_NOARP;
1355         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1356         dev->iflink             = 0;
1357         dev->addr_len           = 4;
1358         dev->features           |= NETIF_F_LLTX;
1359         dev->features           |= SIT_FEATURES;
1360         dev->hw_features        |= SIT_FEATURES;
1361 }
1362
1363 static int ipip6_tunnel_init(struct net_device *dev)
1364 {
1365         struct ip_tunnel *tunnel = netdev_priv(dev);
1366         int i;
1367
1368         tunnel->dev = dev;
1369         tunnel->net = dev_net(dev);
1370
1371         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1372         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1373
1374         ipip6_tunnel_bind_dev(dev);
1375         dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1376         if (!dev->tstats)
1377                 return -ENOMEM;
1378
1379         for_each_possible_cpu(i) {
1380                 struct pcpu_sw_netstats *ipip6_tunnel_stats;
1381                 ipip6_tunnel_stats = per_cpu_ptr(dev->tstats, i);
1382                 u64_stats_init(&ipip6_tunnel_stats->syncp);
1383         }
1384
1385         tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1386         if (!tunnel->dst_cache) {
1387                 free_percpu(dev->tstats);
1388                 return -ENOMEM;
1389         }
1390
1391         return 0;
1392 }
1393
1394 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1395 {
1396         struct ip_tunnel *tunnel = netdev_priv(dev);
1397         struct iphdr *iph = &tunnel->parms.iph;
1398         struct net *net = dev_net(dev);
1399         struct sit_net *sitn = net_generic(net, sit_net_id);
1400         int i;
1401
1402         tunnel->dev = dev;
1403         tunnel->net = dev_net(dev);
1404         strcpy(tunnel->parms.name, dev->name);
1405
1406         iph->version            = 4;
1407         iph->protocol           = IPPROTO_IPV6;
1408         iph->ihl                = 5;
1409         iph->ttl                = 64;
1410
1411         dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1412         if (!dev->tstats)
1413                 return -ENOMEM;
1414
1415         for_each_possible_cpu(i) {
1416                 struct pcpu_sw_netstats *ipip6_fb_stats;
1417                 ipip6_fb_stats = per_cpu_ptr(dev->tstats, i);
1418                 u64_stats_init(&ipip6_fb_stats->syncp);
1419         }
1420
1421         tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1422         if (!tunnel->dst_cache) {
1423                 free_percpu(dev->tstats);
1424                 return -ENOMEM;
1425         }
1426
1427         dev_hold(dev);
1428         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1429         return 0;
1430 }
1431
1432 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1433 {
1434         u8 proto;
1435
1436         if (!data || !data[IFLA_IPTUN_PROTO])
1437                 return 0;
1438
1439         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1440         if (proto != IPPROTO_IPV6 &&
1441             proto != IPPROTO_IPIP &&
1442             proto != 0)
1443                 return -EINVAL;
1444
1445         return 0;
1446 }
1447
1448 static void ipip6_netlink_parms(struct nlattr *data[],
1449                                 struct ip_tunnel_parm *parms)
1450 {
1451         memset(parms, 0, sizeof(*parms));
1452
1453         parms->iph.version = 4;
1454         parms->iph.protocol = IPPROTO_IPV6;
1455         parms->iph.ihl = 5;
1456         parms->iph.ttl = 64;
1457
1458         if (!data)
1459                 return;
1460
1461         if (data[IFLA_IPTUN_LINK])
1462                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1463
1464         if (data[IFLA_IPTUN_LOCAL])
1465                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1466
1467         if (data[IFLA_IPTUN_REMOTE])
1468                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1469
1470         if (data[IFLA_IPTUN_TTL]) {
1471                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1472                 if (parms->iph.ttl)
1473                         parms->iph.frag_off = htons(IP_DF);
1474         }
1475
1476         if (data[IFLA_IPTUN_TOS])
1477                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1478
1479         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1480                 parms->iph.frag_off = htons(IP_DF);
1481
1482         if (data[IFLA_IPTUN_FLAGS])
1483                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1484
1485         if (data[IFLA_IPTUN_PROTO])
1486                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1487
1488 }
1489
1490 #ifdef CONFIG_IPV6_SIT_6RD
1491 /* This function returns true when 6RD attributes are present in the nl msg */
1492 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1493                                     struct ip_tunnel_6rd *ip6rd)
1494 {
1495         bool ret = false;
1496         memset(ip6rd, 0, sizeof(*ip6rd));
1497
1498         if (!data)
1499                 return ret;
1500
1501         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1502                 ret = true;
1503                 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1504                            sizeof(struct in6_addr));
1505         }
1506
1507         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1508                 ret = true;
1509                 ip6rd->relay_prefix =
1510                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1511         }
1512
1513         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1514                 ret = true;
1515                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1516         }
1517
1518         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1519                 ret = true;
1520                 ip6rd->relay_prefixlen =
1521                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1522         }
1523
1524         return ret;
1525 }
1526 #endif
1527
1528 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1529                          struct nlattr *tb[], struct nlattr *data[])
1530 {
1531         struct net *net = dev_net(dev);
1532         struct ip_tunnel *nt;
1533 #ifdef CONFIG_IPV6_SIT_6RD
1534         struct ip_tunnel_6rd ip6rd;
1535 #endif
1536         int err;
1537
1538         nt = netdev_priv(dev);
1539         ipip6_netlink_parms(data, &nt->parms);
1540
1541         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1542                 return -EEXIST;
1543
1544         err = ipip6_tunnel_create(dev);
1545         if (err < 0)
1546                 return err;
1547
1548 #ifdef CONFIG_IPV6_SIT_6RD
1549         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1550                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1551 #endif
1552
1553         return err;
1554 }
1555
1556 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1557                           struct nlattr *data[])
1558 {
1559         struct ip_tunnel *t = netdev_priv(dev);
1560         struct ip_tunnel_parm p;
1561         struct net *net = t->net;
1562         struct sit_net *sitn = net_generic(net, sit_net_id);
1563 #ifdef CONFIG_IPV6_SIT_6RD
1564         struct ip_tunnel_6rd ip6rd;
1565 #endif
1566
1567         if (dev == sitn->fb_tunnel_dev)
1568                 return -EINVAL;
1569
1570         ipip6_netlink_parms(data, &p);
1571
1572         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1573             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1574                 return -EINVAL;
1575
1576         t = ipip6_tunnel_locate(net, &p, 0);
1577
1578         if (t) {
1579                 if (t->dev != dev)
1580                         return -EEXIST;
1581         } else
1582                 t = netdev_priv(dev);
1583
1584         ipip6_tunnel_update(t, &p);
1585
1586 #ifdef CONFIG_IPV6_SIT_6RD
1587         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1588                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1589 #endif
1590
1591         return 0;
1592 }
1593
1594 static size_t ipip6_get_size(const struct net_device *dev)
1595 {
1596         return
1597                 /* IFLA_IPTUN_LINK */
1598                 nla_total_size(4) +
1599                 /* IFLA_IPTUN_LOCAL */
1600                 nla_total_size(4) +
1601                 /* IFLA_IPTUN_REMOTE */
1602                 nla_total_size(4) +
1603                 /* IFLA_IPTUN_TTL */
1604                 nla_total_size(1) +
1605                 /* IFLA_IPTUN_TOS */
1606                 nla_total_size(1) +
1607                 /* IFLA_IPTUN_PMTUDISC */
1608                 nla_total_size(1) +
1609                 /* IFLA_IPTUN_FLAGS */
1610                 nla_total_size(2) +
1611                 /* IFLA_IPTUN_PROTO */
1612                 nla_total_size(1) +
1613 #ifdef CONFIG_IPV6_SIT_6RD
1614                 /* IFLA_IPTUN_6RD_PREFIX */
1615                 nla_total_size(sizeof(struct in6_addr)) +
1616                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1617                 nla_total_size(4) +
1618                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1619                 nla_total_size(2) +
1620                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1621                 nla_total_size(2) +
1622 #endif
1623                 0;
1624 }
1625
1626 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1627 {
1628         struct ip_tunnel *tunnel = netdev_priv(dev);
1629         struct ip_tunnel_parm *parm = &tunnel->parms;
1630
1631         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1632             nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1633             nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1634             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1635             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1636             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1637                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1638             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1639             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1640                 goto nla_put_failure;
1641
1642 #ifdef CONFIG_IPV6_SIT_6RD
1643         if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1644                     &tunnel->ip6rd.prefix) ||
1645             nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1646                          tunnel->ip6rd.relay_prefix) ||
1647             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1648                         tunnel->ip6rd.prefixlen) ||
1649             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1650                         tunnel->ip6rd.relay_prefixlen))
1651                 goto nla_put_failure;
1652 #endif
1653
1654         return 0;
1655
1656 nla_put_failure:
1657         return -EMSGSIZE;
1658 }
1659
1660 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1661         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1662         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1663         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1664         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1665         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1666         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1667         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1668         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1669 #ifdef CONFIG_IPV6_SIT_6RD
1670         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1671         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1672         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1673         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1674 #endif
1675 };
1676
1677 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1678 {
1679         struct net *net = dev_net(dev);
1680         struct sit_net *sitn = net_generic(net, sit_net_id);
1681
1682         if (dev != sitn->fb_tunnel_dev)
1683                 unregister_netdevice_queue(dev, head);
1684 }
1685
1686 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1687         .kind           = "sit",
1688         .maxtype        = IFLA_IPTUN_MAX,
1689         .policy         = ipip6_policy,
1690         .priv_size      = sizeof(struct ip_tunnel),
1691         .setup          = ipip6_tunnel_setup,
1692         .validate       = ipip6_validate,
1693         .newlink        = ipip6_newlink,
1694         .changelink     = ipip6_changelink,
1695         .get_size       = ipip6_get_size,
1696         .fill_info      = ipip6_fill_info,
1697         .dellink        = ipip6_dellink,
1698 };
1699
1700 static struct xfrm_tunnel sit_handler __read_mostly = {
1701         .handler        =       ipip6_rcv,
1702         .err_handler    =       ipip6_err,
1703         .priority       =       1,
1704 };
1705
1706 static struct xfrm_tunnel ipip_handler __read_mostly = {
1707         .handler        =       ipip_rcv,
1708         .err_handler    =       ipip6_err,
1709         .priority       =       2,
1710 };
1711
1712 static void __net_exit sit_destroy_tunnels(struct net *net,
1713                                            struct list_head *head)
1714 {
1715         struct sit_net *sitn = net_generic(net, sit_net_id);
1716         struct net_device *dev, *aux;
1717         int prio;
1718
1719         for_each_netdev_safe(net, dev, aux)
1720                 if (dev->rtnl_link_ops == &sit_link_ops)
1721                         unregister_netdevice_queue(dev, head);
1722
1723         for (prio = 1; prio < 4; prio++) {
1724                 int h;
1725                 for (h = 0; h < HASH_SIZE; h++) {
1726                         struct ip_tunnel *t;
1727
1728                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1729                         while (t != NULL) {
1730                                 /* If dev is in the same netns, it has already
1731                                  * been added to the list by the previous loop.
1732                                  */
1733                                 if (!net_eq(dev_net(t->dev), net))
1734                                         unregister_netdevice_queue(t->dev,
1735                                                                    head);
1736                                 t = rtnl_dereference(t->next);
1737                         }
1738                 }
1739         }
1740 }
1741
1742 static int __net_init sit_init_net(struct net *net)
1743 {
1744         struct sit_net *sitn = net_generic(net, sit_net_id);
1745         struct ip_tunnel *t;
1746         int err;
1747
1748         sitn->tunnels[0] = sitn->tunnels_wc;
1749         sitn->tunnels[1] = sitn->tunnels_l;
1750         sitn->tunnels[2] = sitn->tunnels_r;
1751         sitn->tunnels[3] = sitn->tunnels_r_l;
1752
1753         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1754                                            ipip6_tunnel_setup);
1755         if (!sitn->fb_tunnel_dev) {
1756                 err = -ENOMEM;
1757                 goto err_alloc_dev;
1758         }
1759         dev_net_set(sitn->fb_tunnel_dev, net);
1760         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1761         /* FB netdevice is special: we have one, and only one per netns.
1762          * Allowing to move it to another netns is clearly unsafe.
1763          */
1764         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1765
1766         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1767         if (err)
1768                 goto err_dev_free;
1769
1770         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1771
1772         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1773                 goto err_reg_dev;
1774
1775         t = netdev_priv(sitn->fb_tunnel_dev);
1776
1777         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1778         return 0;
1779
1780 err_reg_dev:
1781         dev_put(sitn->fb_tunnel_dev);
1782 err_dev_free:
1783         ipip6_dev_free(sitn->fb_tunnel_dev);
1784 err_alloc_dev:
1785         return err;
1786 }
1787
1788 static void __net_exit sit_exit_net(struct net *net)
1789 {
1790         LIST_HEAD(list);
1791
1792         rtnl_lock();
1793         sit_destroy_tunnels(net, &list);
1794         unregister_netdevice_many(&list);
1795         rtnl_unlock();
1796 }
1797
1798 static struct pernet_operations sit_net_ops = {
1799         .init = sit_init_net,
1800         .exit = sit_exit_net,
1801         .id   = &sit_net_id,
1802         .size = sizeof(struct sit_net),
1803 };
1804
1805 static void __exit sit_cleanup(void)
1806 {
1807         rtnl_link_unregister(&sit_link_ops);
1808         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1809         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1810
1811         unregister_pernet_device(&sit_net_ops);
1812         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1813 }
1814
1815 static int __init sit_init(void)
1816 {
1817         int err;
1818
1819         pr_info("IPv6 over IPv4 tunneling driver\n");
1820
1821         err = register_pernet_device(&sit_net_ops);
1822         if (err < 0)
1823                 return err;
1824         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1825         if (err < 0) {
1826                 pr_info("%s: can't register ip6ip4\n", __func__);
1827                 goto xfrm_tunnel_failed;
1828         }
1829         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1830         if (err < 0) {
1831                 pr_info("%s: can't register ip4ip4\n", __func__);
1832                 goto xfrm_tunnel4_failed;
1833         }
1834         err = rtnl_link_register(&sit_link_ops);
1835         if (err < 0)
1836                 goto rtnl_link_failed;
1837
1838 out:
1839         return err;
1840
1841 rtnl_link_failed:
1842         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1843 xfrm_tunnel4_failed:
1844         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1845 xfrm_tunnel_failed:
1846         unregister_pernet_device(&sit_net_ops);
1847         goto out;
1848 }
1849
1850 module_init(sit_init);
1851 module_exit(sit_cleanup);
1852 MODULE_LICENSE("GPL");
1853 MODULE_ALIAS_RTNL_LINK("sit");
1854 MODULE_ALIAS_NETDEV("sit0");