net/ipv6/sit.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   9  *
  10  *      Changes:
  11  * Roger Venning <r.venning@telstra.com>:       6to4 support
  12  * Nate Thompson <nate@thebog.net>:             6to4 support
  13  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  14  */
  15
  16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18 #include <linux/module.h>
  19 #include <linux/capability.h>
  20 #include <linux/errno.h>
  21 #include <linux/types.h>
  22 #include <linux/socket.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/in6.h>
  26 #include <linux/netdevice.h>
  27 #include <linux/if_arp.h>
  28 #include <linux/icmp.h>
  29 #include <linux/slab.h>
  30 #include <linux/uaccess.h>
  31 #include <linux/init.h>
  32 #include <linux/netfilter_ipv4.h>
  33 #include <linux/if_ether.h>
  34
  35 #include <net/sock.h>
  36 #include <net/snmp.h>
  37
  38 #include <net/ipv6.h>
  39 #include <net/protocol.h>
  40 #include <net/transp_v6.h>
  41 #include <net/ip6_fib.h>
  42 #include <net/ip6_route.h>
  43 #include <net/ndisc.h>
  44 #include <net/addrconf.h>
  45 #include <net/ip.h>
  46 #include <net/udp.h>
  47 #include <net/icmp.h>
  48 #include <net/ip_tunnels.h>
  49 #include <net/inet_ecn.h>
  50 #include <net/xfrm.h>
  51 #include <net/dsfield.h>
  52 #include <net/net_namespace.h>
  53 #include <net/netns/generic.h>
  54
  55 /*
  56    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  57
  58    For comments look at net/ipv4/ip_gre.c --ANK
  59  */
  60
  61 #define IP6_SIT_HASH_SIZE  16
  62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  63
  64 static bool log_ecn_error = true;
  65 module_param(log_ecn_error, bool, 0644);
  66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  67
  68 static int ipip6_tunnel_init(struct net_device *dev);
  69 static void ipip6_tunnel_setup(struct net_device *dev);
  70 static void ipip6_dev_free(struct net_device *dev);
  71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  72                       __be32 *v4dst);
  73 static struct rtnl_link_ops sit_link_ops __read_mostly;
  74
  75 static unsigned int sit_net_id __read_mostly;
  76 struct sit_net {
  77         struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  78         struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  79         struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
  80         struct ip_tunnel __rcu *tunnels_wc[1];
  81         struct ip_tunnel __rcu **tunnels[4];
  82
  83         struct net_device *fb_tunnel_dev;
  84 };
  85
  86 static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
  87 {
  88         struct ip_tunnel *t = netdev_priv(dev);
  89
  90         return net_generic(t->net, sit_net_id);
  91 }
  92
  93 /*
  94  * Must be invoked with rcu_read_lock
  95  */
  96 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  97                                              struct net_device *dev,
  98                                              __be32 remote, __be32 local,
  99                                              int sifindex)
 100 {
 101         unsigned int h0 = HASH(remote);
 102         unsigned int h1 = HASH(local);
 103         struct ip_tunnel *t;
 104         struct sit_net *sitn = net_generic(net, sit_net_id);
 105         int ifindex = dev ? dev->ifindex : 0;
 106
 107         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 108                 if (local == t->parms.iph.saddr &&
 109                     remote == t->parms.iph.daddr &&
 110                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 111                      sifindex == t->parms.link) &&
 112                     (t->dev->flags & IFF_UP))
 113                         return t;
 114         }
 115         for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 116                 if (remote == t->parms.iph.daddr &&
 117                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 118                      sifindex == t->parms.link) &&
 119                     (t->dev->flags & IFF_UP))
 120                         return t;
 121         }
 122         for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 123                 if (local == t->parms.iph.saddr &&
 124                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 125                      sifindex == t->parms.link) &&
 126                     (t->dev->flags & IFF_UP))
 127                         return t;
 128         }
 129         t = rcu_dereference(sitn->tunnels_wc[0]);
 130         if (t && (t->dev->flags & IFF_UP))
 131                 return t;
 132         return NULL;
 133 }
 134
 135 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 136                 struct ip_tunnel_parm *parms)
 137 {
 138         __be32 remote = parms->iph.daddr;
 139         __be32 local = parms->iph.saddr;
 140         unsigned int h = 0;
 141         int prio = 0;
 142
 143         if (remote) {
 144                 prio |= 2;
 145                 h ^= HASH(remote);
 146         }
 147         if (local) {
 148                 prio |= 1;
 149                 h ^= HASH(local);
 150         }
 151         return &sitn->tunnels[prio][h];
 152 }
 153
 154 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 155                 struct ip_tunnel *t)
 156 {
 157         return __ipip6_bucket(sitn, &t->parms);
 158 }
 159
 160 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 161 {
 162         struct ip_tunnel __rcu **tp;
 163         struct ip_tunnel *iter;
 164
 165         for (tp = ipip6_bucket(sitn, t);
 166              (iter = rtnl_dereference(*tp)) != NULL;
 167              tp = &iter->next) {
 168                 if (t == iter) {
 169                         rcu_assign_pointer(*tp, t->next);
 170                         break;
 171                 }
 172         }
 173 }
 174
 175 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 176 {
 177         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 178
 179         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 180         rcu_assign_pointer(*tp, t);
 181 }
 182
 183 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 184 {
 185 #ifdef CONFIG_IPV6_SIT_6RD
 186         struct ip_tunnel *t = netdev_priv(dev);
 187
 188         if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 189                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 190                 t->ip6rd.relay_prefix = 0;
 191                 t->ip6rd.prefixlen = 16;
 192                 t->ip6rd.relay_prefixlen = 0;
 193         } else {
 194                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 195                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 196         }
 197 #endif
 198 }
 199
 200 static int ipip6_tunnel_create(struct net_device *dev)
 201 {
 202         struct ip_tunnel *t = netdev_priv(dev);
 203         struct net *net = dev_net(dev);
 204         struct sit_net *sitn = net_generic(net, sit_net_id);
 205         int err;
 206
 207         __dev_addr_set(dev, &t->parms.iph.saddr, 4);
 208         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 209
 210         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 211                 dev->priv_flags |= IFF_ISATAP;
 212
 213         dev->rtnl_link_ops = &sit_link_ops;
 214
 215         err = register_netdevice(dev);
 216         if (err < 0)
 217                 goto out;
 218
 219         ipip6_tunnel_clone_6rd(dev, sitn);
 220
 221         ipip6_tunnel_link(sitn, t);
 222         return 0;
 223
 224 out:
 225         return err;
 226 }
 227
 228 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 229                 struct ip_tunnel_parm *parms, int create)
 230 {
 231         __be32 remote = parms->iph.daddr;
 232         __be32 local = parms->iph.saddr;
 233         struct ip_tunnel *t, *nt;
 234         struct ip_tunnel __rcu **tp;
 235         struct net_device *dev;
 236         char name[IFNAMSIZ];
 237         struct sit_net *sitn = net_generic(net, sit_net_id);
 238
 239         for (tp = __ipip6_bucket(sitn, parms);
 240             (t = rtnl_dereference(*tp)) != NULL;
 241              tp = &t->next) {
 242                 if (local == t->parms.iph.saddr &&
 243                     remote == t->parms.iph.daddr &&
 244                     parms->link == t->parms.link) {
 245                         if (create)
 246                                 return NULL;
 247                         else
 248                                 return t;
 249                 }
 250         }
 251         if (!create)
 252                 goto failed;
 253
 254         if (parms->name[0]) {
 255                 if (!dev_valid_name(parms->name))
 256                         goto failed;
 257                 strlcpy(name, parms->name, IFNAMSIZ);
 258         } else {
 259                 strcpy(name, "sit%d");
 260         }
 261         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 262                            ipip6_tunnel_setup);
 263         if (!dev)
 264                 return NULL;
 265
 266         dev_net_set(dev, net);
 267
 268         nt = netdev_priv(dev);
 269
 270         nt->parms = *parms;
 271         if (ipip6_tunnel_create(dev) < 0)
 272                 goto failed_free;
 273
 274         if (!parms->name[0])
 275                 strcpy(parms->name, dev->name);
 276
 277         return nt;
 278
 279 failed_free:
 280         free_netdev(dev);
 281 failed:
 282         return NULL;
 283 }
 284
 285 #define for_each_prl_rcu(start)                 \
 286         for (prl = rcu_dereference(start);      \
 287              prl;                               \
 288              prl = rcu_dereference(prl->next))
 289
 290 static struct ip_tunnel_prl_entry *
 291 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 292 {
 293         struct ip_tunnel_prl_entry *prl;
 294
 295         for_each_prl_rcu(t->prl)
 296                 if (prl->addr == addr)
 297                         break;
 298         return prl;
 299
 300 }
 301
 302 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
 303 {
 304         struct ip_tunnel *t = netdev_priv(dev);
 305         struct ip_tunnel_prl kprl, *kp;
 306         struct ip_tunnel_prl_entry *prl;
 307         unsigned int cmax, c = 0, ca, len;
 308         int ret = 0;
 309
 310         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 311                 return -EINVAL;
 312
 313         if (copy_from_user(&kprl, a, sizeof(kprl)))
 314                 return -EFAULT;
 315         cmax = kprl.datalen / sizeof(kprl);
 316         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 317                 cmax = 1;
 318
 319         /* For simple GET or for root users,
 320          * we try harder to allocate.
 321          */
 322         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 323                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
 324                 NULL;
 325
 326         ca = min(t->prl_count, cmax);
 327
 328         if (!kp) {
 329                 /* We don't try hard to allocate much memory for
 330                  * non-root users.
 331                  * For root users, retry allocating enough memory for
 332                  * the answer.
 333                  */
 334                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
 335                                               __GFP_NOWARN);
 336                 if (!kp) {
 337                         ret = -ENOMEM;
 338                         goto out;
 339                 }
 340         }
 341
 342         rcu_read_lock();
 343         for_each_prl_rcu(t->prl) {
 344                 if (c >= cmax)
 345                         break;
 346                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 347                         continue;
 348                 kp[c].addr = prl->addr;
 349                 kp[c].flags = prl->flags;
 350                 c++;
 351                 if (kprl.addr != htonl(INADDR_ANY))
 352                         break;
 353         }
 354
 355         rcu_read_unlock();
 356
 357         len = sizeof(*kp) * c;
 358         ret = 0;
 359         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 360                 ret = -EFAULT;
 361
 362         kfree(kp);
 363 out:
 364         return ret;
 365 }
 366
 367 static int
 368 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 369 {
 370         struct ip_tunnel_prl_entry *p;
 371         int err = 0;
 372
 373         if (a->addr == htonl(INADDR_ANY))
 374                 return -EINVAL;
 375
 376         ASSERT_RTNL();
 377
 378         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 379                 if (p->addr == a->addr) {
 380                         if (chg) {
 381                                 p->flags = a->flags;
 382                                 goto out;
 383                         }
 384                         err = -EEXIST;
 385                         goto out;
 386                 }
 387         }
 388
 389         if (chg) {
 390                 err = -ENXIO;
 391                 goto out;
 392         }
 393
 394         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 395         if (!p) {
 396                 err = -ENOBUFS;
 397                 goto out;
 398         }
 399
 400         p->next = t->prl;
 401         p->addr = a->addr;
 402         p->flags = a->flags;
 403         t->prl_count++;
 404         rcu_assign_pointer(t->prl, p);
 405 out:
 406         return err;
 407 }
 408
 409 static void prl_list_destroy_rcu(struct rcu_head *head)
 410 {
 411         struct ip_tunnel_prl_entry *p, *n;
 412
 413         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 414         do {
 415                 n = rcu_dereference_protected(p->next, 1);
 416                 kfree(p);
 417                 p = n;
 418         } while (p);
 419 }
 420
 421 static int
 422 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 423 {
 424         struct ip_tunnel_prl_entry *x;
 425         struct ip_tunnel_prl_entry __rcu **p;
 426         int err = 0;
 427
 428         ASSERT_RTNL();
 429
 430         if (a && a->addr != htonl(INADDR_ANY)) {
 431                 for (p = &t->prl;
 432                      (x = rtnl_dereference(*p)) != NULL;
 433                      p = &x->next) {
 434                         if (x->addr == a->addr) {
 435                                 *p = x->next;
 436                                 kfree_rcu(x, rcu_head);
 437                                 t->prl_count--;
 438                                 goto out;
 439                         }
 440                 }
 441                 err = -ENXIO;
 442         } else {
 443                 x = rtnl_dereference(t->prl);
 444                 if (x) {
 445                         t->prl_count = 0;
 446                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 447                         t->prl = NULL;
 448                 }
 449         }
 450 out:
 451         return err;
 452 }
 453
 454 static int ipip6_tunnel_prl_ctl(struct net_device *dev,
 455                                 struct ip_tunnel_prl __user *data, int cmd)
 456 {
 457         struct ip_tunnel *t = netdev_priv(dev);
 458         struct ip_tunnel_prl prl;
 459         int err;
 460
 461         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
 462                 return -EPERM;
 463         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 464                 return -EINVAL;
 465
 466         if (copy_from_user(&prl, data, sizeof(prl)))
 467                 return -EFAULT;
 468
 469         switch (cmd) {
 470         case SIOCDELPRL:
 471                 err = ipip6_tunnel_del_prl(t, &prl);
 472                 break;
 473         case SIOCADDPRL:
 474         case SIOCCHGPRL:
 475                 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 476                 break;
 477         }
 478         dst_cache_reset(&t->dst_cache);
 479         netdev_state_change(dev);
 480         return err;
 481 }
 482
 483 static int
 484 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 485 {
 486         struct ip_tunnel_prl_entry *p;
 487         int ok = 1;
 488
 489         rcu_read_lock();
 490         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 491         if (p) {
 492                 if (p->flags & PRL_DEFAULT)
 493                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 494                 else
 495                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 496         } else {
 497                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 498
 499                 if (ipv6_addr_is_isatap(addr6) &&
 500                     (addr6->s6_addr32[3] == iph->saddr) &&
 501                     ipv6_chk_prefix(addr6, t->dev))
 502                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 503                 else
 504                         ok = 0;
 505         }
 506         rcu_read_unlock();
 507         return ok;
 508 }
 509
 510 static void ipip6_tunnel_uninit(struct net_device *dev)
 511 {
 512         struct ip_tunnel *tunnel = netdev_priv(dev);
 513         struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 514
 515         if (dev == sitn->fb_tunnel_dev) {
 516                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 517         } else {
 518                 ipip6_tunnel_unlink(sitn, tunnel);
 519                 ipip6_tunnel_del_prl(tunnel, NULL);
 520         }
 521         dst_cache_reset(&tunnel->dst_cache);
 522         dev_put_track(dev, &tunnel->dev_tracker);
 523 }
 524
 525 static int ipip6_err(struct sk_buff *skb, u32 info)
 526 {
 527         const struct iphdr *iph = (const struct iphdr *)skb->data;
 528         const int type = icmp_hdr(skb)->type;
 529         const int code = icmp_hdr(skb)->code;
 530         unsigned int data_len = 0;
 531         struct ip_tunnel *t;
 532         int sifindex;
 533         int err;
 534
 535         switch (type) {
 536         default:
 537         case ICMP_PARAMETERPROB:
 538                 return 0;
 539
 540         case ICMP_DEST_UNREACH:
 541                 switch (code) {
 542                 case ICMP_SR_FAILED:
 543                         /* Impossible event. */
 544                         return 0;
 545                 default:
 546                         /* All others are translated to HOST_UNREACH.
 547                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 548                            I believe they are just ether pollution. --ANK
 549                          */
 550                         break;
 551                 }
 552                 break;
 553         case ICMP_TIME_EXCEEDED:
 554                 if (code != ICMP_EXC_TTL)
 555                         return 0;
 556                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 557                 break;
 558         case ICMP_REDIRECT:
 559                 break;
 560         }
 561
 562         err = -ENOENT;
 563
 564         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 565         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 566                                 iph->daddr, iph->saddr, sifindex);
 567         if (!t)
 568                 goto out;
 569
 570         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 571                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 572                                  t->parms.link, iph->protocol);
 573                 err = 0;
 574                 goto out;
 575         }
 576         if (type == ICMP_REDIRECT) {
 577                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 578                               iph->protocol);
 579                 err = 0;
 580                 goto out;
 581         }
 582
 583         err = 0;
 584         if (__in6_dev_get(skb->dev) &&
 585             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 586                 goto out;
 587
 588         if (t->parms.iph.daddr == 0)
 589                 goto out;
 590
 591         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 592                 goto out;
 593
 594         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 595                 t->err_count++;
 596         else
 597                 t->err_count = 1;
 598         t->err_time = jiffies;
 599 out:
 600         return err;
 601 }
 602
 603 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 604                                   const struct in6_addr *v6addr)
 605 {
 606         __be32 v4embed = 0;
 607         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 608                 return true;
 609         return false;
 610 }
 611
 612 /* Checks if an address matches an address on the tunnel interface.
 613  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 614  * Long story:
 615  * This function is called after we considered the packet as spoofed
 616  * in is_spoofed_6rd.
 617  * We may have a router that is doing NAT for proto 41 packets
 618  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 619  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 620  * function will return true, dropping the packet.
 621  * But, we can still check if is spoofed against the IP
 622  * addresses associated with the interface.
 623  */
 624 static bool only_dnatted(const struct ip_tunnel *tunnel,
 625         const struct in6_addr *v6dst)
 626 {
 627         int prefix_len;
 628
 629 #ifdef CONFIG_IPV6_SIT_6RD
 630         prefix_len = tunnel->ip6rd.prefixlen + 32
 631                 - tunnel->ip6rd.relay_prefixlen;
 632 #else
 633         prefix_len = 48;
 634 #endif
 635         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 636 }
 637
 638 /* Returns true if a packet is spoofed */
 639 static bool packet_is_spoofed(struct sk_buff *skb,
 640                               const struct iphdr *iph,
 641                               struct ip_tunnel *tunnel)
 642 {
 643         const struct ipv6hdr *ipv6h;
 644
 645         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 646                 if (!isatap_chksrc(skb, iph, tunnel))
 647                         return true;
 648
 649                 return false;
 650         }
 651
 652         if (tunnel->dev->flags & IFF_POINTOPOINT)
 653                 return false;
 654
 655         ipv6h = ipv6_hdr(skb);
 656
 657         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 658                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 659                                      &iph->saddr, &ipv6h->saddr,
 660                                      &iph->daddr, &ipv6h->daddr);
 661                 return true;
 662         }
 663
 664         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 665                 return false;
 666
 667         if (only_dnatted(tunnel, &ipv6h->daddr))
 668                 return false;
 669
 670         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 671                              &iph->saddr, &ipv6h->saddr,
 672                              &iph->daddr, &ipv6h->daddr);
 673         return true;
 674 }
 675
 676 static int ipip6_rcv(struct sk_buff *skb)
 677 {
 678         const struct iphdr *iph = ip_hdr(skb);
 679         struct ip_tunnel *tunnel;
 680         int sifindex;
 681         int err;
 682
 683         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 684         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 685                                      iph->saddr, iph->daddr, sifindex);
 686         if (tunnel) {
 687                 struct pcpu_sw_netstats *tstats;
 688
 689                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 690                     tunnel->parms.iph.protocol != 0)
 691                         goto out;
 692
 693                 skb->mac_header = skb->network_header;
 694                 skb_reset_network_header(skb);
 695                 IPCB(skb)->flags = 0;
 696                 skb->dev = tunnel->dev;
 697
 698                 if (packet_is_spoofed(skb, iph, tunnel)) {
 699                         tunnel->dev->stats.rx_errors++;
 700                         goto out;
 701                 }
 702
 703                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 704                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 705                         goto out;
 706
 707                 /* skb can be uncloned in iptunnel_pull_header, so
 708                  * old iph is no longer valid
 709                  */
 710                 iph = (const struct iphdr *)skb_mac_header(skb);
 711                 skb_reset_mac_header(skb);
 712
 713                 err = IP_ECN_decapsulate(iph, skb);
 714                 if (unlikely(err)) {
 715                         if (log_ecn_error)
 716                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 717                                                      &iph->saddr, iph->tos);
 718                         if (err > 1) {
 719                                 ++tunnel->dev->stats.rx_frame_errors;
 720                                 ++tunnel->dev->stats.rx_errors;
 721                                 goto out;
 722                         }
 723                 }
 724
 725                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 726                 u64_stats_update_begin(&tstats->syncp);
 727                 tstats->rx_packets++;
 728                 tstats->rx_bytes += skb->len;
 729                 u64_stats_update_end(&tstats->syncp);
 730
 731                 netif_rx(skb);
 732
 733                 return 0;
 734         }
 735
 736         /* no tunnel matched,  let upstream know, ipsec may handle it */
 737         return 1;
 738 out:
 739         kfree_skb(skb);
 740         return 0;
 741 }
 742
 743 static const struct tnl_ptk_info ipip_tpi = {
 744         /* no tunnel info required for ipip. */
 745         .proto = htons(ETH_P_IP),
 746 };
 747
 748 #if IS_ENABLED(CONFIG_MPLS)
 749 static const struct tnl_ptk_info mplsip_tpi = {
 750         /* no tunnel info required for mplsip. */
 751         .proto = htons(ETH_P_MPLS_UC),
 752 };
 753 #endif
 754
 755 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 756 {
 757         const struct iphdr *iph;
 758         struct ip_tunnel *tunnel;
 759         int sifindex;
 760
 761         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 762
 763         iph = ip_hdr(skb);
 764         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 765                                      iph->saddr, iph->daddr, sifindex);
 766         if (tunnel) {
 767                 const struct tnl_ptk_info *tpi;
 768
 769                 if (tunnel->parms.iph.protocol != ipproto &&
 770                     tunnel->parms.iph.protocol != 0)
 771                         goto drop;
 772
 773                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 774                         goto drop;
 775 #if IS_ENABLED(CONFIG_MPLS)
 776                 if (ipproto == IPPROTO_MPLS)
 777                         tpi = &mplsip_tpi;
 778                 else
 779 #endif
 780                         tpi = &ipip_tpi;
 781                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 782                         goto drop;
 783                 skb_reset_mac_header(skb);
 784
 785                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 786         }
 787
 788         return 1;
 789
 790 drop:
 791         kfree_skb(skb);
 792         return 0;
 793 }
 794
 795 static int ipip_rcv(struct sk_buff *skb)
 796 {
 797         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 798 }
 799
 800 #if IS_ENABLED(CONFIG_MPLS)
 801 static int mplsip_rcv(struct sk_buff *skb)
 802 {
 803         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 804 }
 805 #endif
 806
 807 /*
 808  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 809  * stores the embedded IPv4 address in v4dst and returns true.
 810  */
 811 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 812                       __be32 *v4dst)
 813 {
 814 #ifdef CONFIG_IPV6_SIT_6RD
 815         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 816                               tunnel->ip6rd.prefixlen)) {
 817                 unsigned int pbw0, pbi0;
 818                 int pbi1;
 819                 u32 d;
 820
 821                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 822                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 823
 824                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 825                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 826                     tunnel->ip6rd.relay_prefixlen : 0;
 827
 828                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 829                 if (pbi1 > 0)
 830                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 831                              (32 - pbi1);
 832
 833                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 834                 return true;
 835         }
 836 #else
 837         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 838                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 839                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 840                 return true;
 841         }
 842 #endif
 843         return false;
 844 }
 845
 846 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 847                              const struct in6_addr *v6dst)
 848 {
 849         __be32 dst = 0;
 850         check_6rd(tunnel, v6dst, &dst);
 851         return dst;
 852 }
 853
 854 /*
 855  *      This function assumes it is being called from dev_queue_xmit()
 856  *      and that skb is filled properly by that function.
 857  */
 858
 859 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 860                                      struct net_device *dev)
 861 {
 862         struct ip_tunnel *tunnel = netdev_priv(dev);
 863         const struct iphdr  *tiph = &tunnel->parms.iph;
 864         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 865         u8     tos = tunnel->parms.iph.tos;
 866         __be16 df = tiph->frag_off;
 867         struct rtable *rt;              /* Route to the other host */
 868         struct net_device *tdev;        /* Device to other host */
 869         unsigned int max_headroom;      /* The extra header space needed */
 870         __be32 dst = tiph->daddr;
 871         struct flowi4 fl4;
 872         int    mtu;
 873         const struct in6_addr *addr6;
 874         int addr_type;
 875         u8 ttl;
 876         u8 protocol = IPPROTO_IPV6;
 877         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 878
 879         if (tos == 1)
 880                 tos = ipv6_get_dsfield(iph6);
 881
 882         /* ISATAP (RFC4214) - must come before 6to4 */
 883         if (dev->priv_flags & IFF_ISATAP) {
 884                 struct neighbour *neigh = NULL;
 885                 bool do_tx_error = false;
 886
 887                 if (skb_dst(skb))
 888                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 889
 890                 if (!neigh) {
 891                         net_dbg_ratelimited("nexthop == NULL\n");
 892                         goto tx_error;
 893                 }
 894
 895                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 896                 addr_type = ipv6_addr_type(addr6);
 897
 898                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 899                      ipv6_addr_is_isatap(addr6))
 900                         dst = addr6->s6_addr32[3];
 901                 else
 902                         do_tx_error = true;
 903
 904                 neigh_release(neigh);
 905                 if (do_tx_error)
 906                         goto tx_error;
 907         }
 908
 909         if (!dst)
 910                 dst = try_6rd(tunnel, &iph6->daddr);
 911
 912         if (!dst) {
 913                 struct neighbour *neigh = NULL;
 914                 bool do_tx_error = false;
 915
 916                 if (skb_dst(skb))
 917                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 918
 919                 if (!neigh) {
 920                         net_dbg_ratelimited("nexthop == NULL\n");
 921                         goto tx_error;
 922                 }
 923
 924                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 925                 addr_type = ipv6_addr_type(addr6);
 926
 927                 if (addr_type == IPV6_ADDR_ANY) {
 928                         addr6 = &ipv6_hdr(skb)->daddr;
 929                         addr_type = ipv6_addr_type(addr6);
 930                 }
 931
 932                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 933                         dst = addr6->s6_addr32[3];
 934                 else
 935                         do_tx_error = true;
 936
 937                 neigh_release(neigh);
 938                 if (do_tx_error)
 939                         goto tx_error;
 940         }
 941
 942         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 943                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 944                            0, dst, tiph->saddr, 0, 0,
 945                            sock_net_uid(tunnel->net, NULL));
 946
 947         rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 948         if (!rt) {
 949                 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 950                 if (IS_ERR(rt)) {
 951                         dev->stats.tx_carrier_errors++;
 952                         goto tx_error_icmp;
 953                 }
 954                 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 955         }
 956
 957         if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 958                 ip_rt_put(rt);
 959                 dev->stats.tx_carrier_errors++;
 960                 goto tx_error_icmp;
 961         }
 962         tdev = rt->dst.dev;
 963
 964         if (tdev == dev) {
 965                 ip_rt_put(rt);
 966                 dev->stats.collisions++;
 967                 goto tx_error;
 968         }
 969
 970         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 971                 ip_rt_put(rt);
 972                 goto tx_error;
 973         }
 974
 975         if (df) {
 976                 mtu = dst_mtu(&rt->dst) - t_hlen;
 977
 978                 if (mtu < IPV4_MIN_MTU) {
 979                         dev->stats.collisions++;
 980                         ip_rt_put(rt);
 981                         goto tx_error;
 982                 }
 983
 984                 if (mtu < IPV6_MIN_MTU) {
 985                         mtu = IPV6_MIN_MTU;
 986                         df = 0;
 987                 }
 988
 989                 if (tunnel->parms.iph.daddr)
 990                         skb_dst_update_pmtu_no_confirm(skb, mtu);
 991
 992                 if (skb->len > mtu && !skb_is_gso(skb)) {
 993                         icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 994                         ip_rt_put(rt);
 995                         goto tx_error;
 996                 }
 997         }
 998
 999         if (tunnel->err_count > 0) {
1000                 if (time_before(jiffies,
1001                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
1002                         tunnel->err_count--;
1003                         dst_link_failure(skb);
1004                 } else
1005                         tunnel->err_count = 0;
1006         }
1007
1008         /*
1009          * Okay, now see if we can stuff it in the buffer as-is.
1010          */
1011         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1012
1013         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1014             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1015                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1016                 if (!new_skb) {
1017                         ip_rt_put(rt);
1018                         dev->stats.tx_dropped++;
1019                         kfree_skb(skb);
1020                         return NETDEV_TX_OK;
1021                 }
1022                 if (skb->sk)
1023                         skb_set_owner_w(new_skb, skb->sk);
1024                 dev_kfree_skb(skb);
1025                 skb = new_skb;
1026                 iph6 = ipv6_hdr(skb);
1027         }
1028         ttl = tiph->ttl;
1029         if (ttl == 0)
1030                 ttl = iph6->hop_limit;
1031         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1032
1033         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
1034                 ip_rt_put(rt);
1035                 goto tx_error;
1036         }
1037
1038         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1039
1040         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1041                       df, !net_eq(tunnel->net, dev_net(dev)));
1042         return NETDEV_TX_OK;
1043
1044 tx_error_icmp:
1045         dst_link_failure(skb);
1046 tx_error:
1047         kfree_skb(skb);
1048         dev->stats.tx_errors++;
1049         return NETDEV_TX_OK;
1050 }
1051
1052 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1053                                      struct net_device *dev, u8 ipproto)
1054 {
1055         struct ip_tunnel *tunnel = netdev_priv(dev);
1056         const struct iphdr  *tiph = &tunnel->parms.iph;
1057
1058         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1059                 goto tx_error;
1060
1061         skb_set_inner_ipproto(skb, ipproto);
1062
1063         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1064         return NETDEV_TX_OK;
1065 tx_error:
1066         kfree_skb(skb);
1067         dev->stats.tx_errors++;
1068         return NETDEV_TX_OK;
1069 }
1070
1071 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1072                                    struct net_device *dev)
1073 {
1074         if (!pskb_inet_may_pull(skb))
1075                 goto tx_err;
1076
1077         switch (skb->protocol) {
1078         case htons(ETH_P_IP):
1079                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1080                 break;
1081         case htons(ETH_P_IPV6):
1082                 ipip6_tunnel_xmit(skb, dev);
1083                 break;
1084 #if IS_ENABLED(CONFIG_MPLS)
1085         case htons(ETH_P_MPLS_UC):
1086                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1087                 break;
1088 #endif
1089         default:
1090                 goto tx_err;
1091         }
1092
1093         return NETDEV_TX_OK;
1094
1095 tx_err:
1096         dev->stats.tx_errors++;
1097         kfree_skb(skb);
1098         return NETDEV_TX_OK;
1099
1100 }
1101
1102 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1103 {
1104         struct net_device *tdev = NULL;
1105         struct ip_tunnel *tunnel;
1106         const struct iphdr *iph;
1107         struct flowi4 fl4;
1108
1109         tunnel = netdev_priv(dev);
1110         iph = &tunnel->parms.iph;
1111
1112         if (iph->daddr) {
1113                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1114                                                           NULL,
1115                                                           iph->daddr, iph->saddr,
1116                                                           0, 0,
1117                                                           IPPROTO_IPV6,
1118                                                           RT_TOS(iph->tos),
1119                                                           tunnel->parms.link);
1120
1121                 if (!IS_ERR(rt)) {
1122                         tdev = rt->dst.dev;
1123                         ip_rt_put(rt);
1124                 }
1125                 dev->flags |= IFF_POINTOPOINT;
1126         }
1127
1128         if (!tdev && tunnel->parms.link)
1129                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1130
1131         if (tdev && !netif_is_l3_master(tdev)) {
1132                 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1133
1134                 dev->mtu = tdev->mtu - t_hlen;
1135                 if (dev->mtu < IPV6_MIN_MTU)
1136                         dev->mtu = IPV6_MIN_MTU;
1137         }
1138 }
1139
1140 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1141                                 __u32 fwmark)
1142 {
1143         struct net *net = t->net;
1144         struct sit_net *sitn = net_generic(net, sit_net_id);
1145
1146         ipip6_tunnel_unlink(sitn, t);
1147         synchronize_net();
1148         t->parms.iph.saddr = p->iph.saddr;
1149         t->parms.iph.daddr = p->iph.daddr;
1150         __dev_addr_set(t->dev, &p->iph.saddr, 4);
1151         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1152         ipip6_tunnel_link(sitn, t);
1153         t->parms.iph.ttl = p->iph.ttl;
1154         t->parms.iph.tos = p->iph.tos;
1155         t->parms.iph.frag_off = p->iph.frag_off;
1156         if (t->parms.link != p->link || t->fwmark != fwmark) {
1157                 t->parms.link = p->link;
1158                 t->fwmark = fwmark;
1159                 ipip6_tunnel_bind_dev(t->dev);
1160         }
1161         dst_cache_reset(&t->dst_cache);
1162         netdev_state_change(t->dev);
1163 }
1164
1165 #ifdef CONFIG_IPV6_SIT_6RD
1166 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1167                                    struct ip_tunnel_6rd *ip6rd)
1168 {
1169         struct in6_addr prefix;
1170         __be32 relay_prefix;
1171
1172         if (ip6rd->relay_prefixlen > 32 ||
1173             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1174                 return -EINVAL;
1175
1176         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1177         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1178                 return -EINVAL;
1179         if (ip6rd->relay_prefixlen)
1180                 relay_prefix = ip6rd->relay_prefix &
1181                                htonl(0xffffffffUL <<
1182                                      (32 - ip6rd->relay_prefixlen));
1183         else
1184                 relay_prefix = 0;
1185         if (relay_prefix != ip6rd->relay_prefix)
1186                 return -EINVAL;
1187
1188         t->ip6rd.prefix = prefix;
1189         t->ip6rd.relay_prefix = relay_prefix;
1190         t->ip6rd.prefixlen = ip6rd->prefixlen;
1191         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1192         dst_cache_reset(&t->dst_cache);
1193         netdev_state_change(t->dev);
1194         return 0;
1195 }
1196
1197 static int
1198 ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1199 {
1200         struct ip_tunnel *t = netdev_priv(dev);
1201         struct ip_tunnel_6rd ip6rd;
1202         struct ip_tunnel_parm p;
1203
1204         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1205                 if (copy_from_user(&p, data, sizeof(p)))
1206                         return -EFAULT;
1207                 t = ipip6_tunnel_locate(t->net, &p, 0);
1208         }
1209         if (!t)
1210                 t = netdev_priv(dev);
1211
1212         ip6rd.prefix = t->ip6rd.prefix;
1213         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1214         ip6rd.prefixlen = t->ip6rd.prefixlen;
1215         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1216         if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1217                 return -EFAULT;
1218         return 0;
1219 }
1220
1221 static int
1222 ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1223                     int cmd)
1224 {
1225         struct ip_tunnel *t = netdev_priv(dev);
1226         struct ip_tunnel_6rd ip6rd;
1227         int err;
1228
1229         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1230                 return -EPERM;
1231         if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1232                 return -EFAULT;
1233
1234         if (cmd != SIOCDEL6RD) {
1235                 err = ipip6_tunnel_update_6rd(t, &ip6rd);
1236                 if (err < 0)
1237                         return err;
1238         } else
1239                 ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1240         return 0;
1241 }
1242
1243 #endif /* CONFIG_IPV6_SIT_6RD */
1244
1245 static bool ipip6_valid_ip_proto(u8 ipproto)
1246 {
1247         return ipproto == IPPROTO_IPV6 ||
1248                 ipproto == IPPROTO_IPIP ||
1249 #if IS_ENABLED(CONFIG_MPLS)
1250                 ipproto == IPPROTO_MPLS ||
1251 #endif
1252                 ipproto == 0;
1253 }
1254
1255 static int
1256 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1257 {
1258         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1259                 return -EPERM;
1260
1261         if (!ipip6_valid_ip_proto(p->iph.protocol))
1262                 return -EINVAL;
1263         if (p->iph.version != 4 ||
1264             p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1265                 return -EINVAL;
1266
1267         if (p->iph.ttl)
1268                 p->iph.frag_off |= htons(IP_DF);
1269         return 0;
1270 }
1271
1272 static int
1273 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1274 {
1275         struct ip_tunnel *t = netdev_priv(dev);
1276
1277         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1278                 t = ipip6_tunnel_locate(t->net, p, 0);
1279         if (!t)
1280                 t = netdev_priv(dev);
1281         memcpy(p, &t->parms, sizeof(*p));
1282         return 0;
1283 }
1284
1285 static int
1286 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1287 {
1288         struct ip_tunnel *t = netdev_priv(dev);
1289         int err;
1290
1291         err = __ipip6_tunnel_ioctl_validate(t->net, p);
1292         if (err)
1293                 return err;
1294
1295         t = ipip6_tunnel_locate(t->net, p, 1);
1296         if (!t)
1297                 return -ENOBUFS;
1298         return 0;
1299 }
1300
1301 static int
1302 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1303 {
1304         struct ip_tunnel *t = netdev_priv(dev);
1305         int err;
1306
1307         err = __ipip6_tunnel_ioctl_validate(t->net, p);
1308         if (err)
1309                 return err;
1310
1311         t = ipip6_tunnel_locate(t->net, p, 0);
1312         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1313                 if (!t)
1314                         return -ENOENT;
1315         } else {
1316                 if (t) {
1317                         if (t->dev != dev)
1318                                 return -EEXIST;
1319                 } else {
1320                         if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1321                             (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1322                                 return -EINVAL;
1323                         t = netdev_priv(dev);
1324                 }
1325
1326                 ipip6_tunnel_update(t, p, t->fwmark);
1327         }
1328
1329         return 0;
1330 }
1331
1332 static int
1333 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1334 {
1335         struct ip_tunnel *t = netdev_priv(dev);
1336
1337         if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1338                 return -EPERM;
1339
1340         if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1341                 t = ipip6_tunnel_locate(t->net, p, 0);
1342                 if (!t)
1343                         return -ENOENT;
1344                 if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1345                         return -EPERM;
1346                 dev = t->dev;
1347         }
1348         unregister_netdevice(dev);
1349         return 0;
1350 }
1351
1352 static int
1353 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
1354 {
1355         switch (cmd) {
1356         case SIOCGETTUNNEL:
1357                 return ipip6_tunnel_get(dev, p);
1358         case SIOCADDTUNNEL:
1359                 return ipip6_tunnel_add(dev, p);
1360         case SIOCCHGTUNNEL:
1361                 return ipip6_tunnel_change(dev, p);
1362         case SIOCDELTUNNEL:
1363                 return ipip6_tunnel_del(dev, p);
1364         default:
1365                 return -EINVAL;
1366         }
1367 }
1368
1369 static int
1370 ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1371                             void __user *data, int cmd)
1372 {
1373         switch (cmd) {
1374         case SIOCGETTUNNEL:
1375         case SIOCADDTUNNEL:
1376         case SIOCCHGTUNNEL:
1377         case SIOCDELTUNNEL:
1378                 return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1379         case SIOCGETPRL:
1380                 return ipip6_tunnel_get_prl(dev, data);
1381         case SIOCADDPRL:
1382         case SIOCDELPRL:
1383         case SIOCCHGPRL:
1384                 return ipip6_tunnel_prl_ctl(dev, data, cmd);
1385 #ifdef CONFIG_IPV6_SIT_6RD
1386         case SIOCGET6RD:
1387                 return ipip6_tunnel_get6rd(dev, data);
1388         case SIOCADD6RD:
1389         case SIOCCHG6RD:
1390         case SIOCDEL6RD:
1391                 return ipip6_tunnel_6rdctl(dev, data, cmd);
1392 #endif
1393         default:
1394                 return -EINVAL;
1395         }
1396 }
1397
1398 static const struct net_device_ops ipip6_netdev_ops = {
1399         .ndo_init       = ipip6_tunnel_init,
1400         .ndo_uninit     = ipip6_tunnel_uninit,
1401         .ndo_start_xmit = sit_tunnel_xmit,
1402         .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1403         .ndo_get_stats64 = dev_get_tstats64,
1404         .ndo_get_iflink = ip_tunnel_get_iflink,
1405         .ndo_tunnel_ctl = ipip6_tunnel_ctl,
1406 };
1407
1408 static void ipip6_dev_free(struct net_device *dev)
1409 {
1410         struct ip_tunnel *tunnel = netdev_priv(dev);
1411
1412         dst_cache_destroy(&tunnel->dst_cache);
1413         free_percpu(dev->tstats);
1414 }
1415
1416 #define SIT_FEATURES (NETIF_F_SG           | \
1417                       NETIF_F_FRAGLIST     | \
1418                       NETIF_F_HIGHDMA      | \
1419                       NETIF_F_GSO_SOFTWARE | \
1420                       NETIF_F_HW_CSUM)
1421
1422 static void ipip6_tunnel_setup(struct net_device *dev)
1423 {
1424         struct ip_tunnel *tunnel = netdev_priv(dev);
1425         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1426
1427         dev->netdev_ops         = &ipip6_netdev_ops;
1428         dev->header_ops         = &ip_tunnel_header_ops;
1429         dev->needs_free_netdev  = true;
1430         dev->priv_destructor    = ipip6_dev_free;
1431
1432         dev->type               = ARPHRD_SIT;
1433         dev->mtu                = ETH_DATA_LEN - t_hlen;
1434         dev->min_mtu            = IPV6_MIN_MTU;
1435         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1436         dev->flags              = IFF_NOARP;
1437         netif_keep_dst(dev);
1438         dev->addr_len           = 4;
1439         dev->features           |= NETIF_F_LLTX;
1440         dev->features           |= SIT_FEATURES;
1441         dev->hw_features        |= SIT_FEATURES;
1442 }
1443
1444 static int ipip6_tunnel_init(struct net_device *dev)
1445 {
1446         struct ip_tunnel *tunnel = netdev_priv(dev);
1447         int err;
1448
1449         tunnel->dev = dev;
1450         tunnel->net = dev_net(dev);
1451         strcpy(tunnel->parms.name, dev->name);
1452
1453         ipip6_tunnel_bind_dev(dev);
1454         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1455         if (!dev->tstats)
1456                 return -ENOMEM;
1457
1458         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1459         if (err) {
1460                 free_percpu(dev->tstats);
1461                 dev->tstats = NULL;
1462                 return err;
1463         }
1464         dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL);
1465         return 0;
1466 }
1467
1468 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1469 {
1470         struct ip_tunnel *tunnel = netdev_priv(dev);
1471         struct iphdr *iph = &tunnel->parms.iph;
1472         struct net *net = dev_net(dev);
1473         struct sit_net *sitn = net_generic(net, sit_net_id);
1474
1475         iph->version            = 4;
1476         iph->protocol           = IPPROTO_IPV6;
1477         iph->ihl                = 5;
1478         iph->ttl                = 64;
1479
1480         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1481 }
1482
1483 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1484                           struct netlink_ext_ack *extack)
1485 {
1486         u8 proto;
1487
1488         if (!data || !data[IFLA_IPTUN_PROTO])
1489                 return 0;
1490
1491         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1492         if (!ipip6_valid_ip_proto(proto))
1493                 return -EINVAL;
1494
1495         return 0;
1496 }
1497
1498 static void ipip6_netlink_parms(struct nlattr *data[],
1499                                 struct ip_tunnel_parm *parms,
1500                                 __u32 *fwmark)
1501 {
1502         memset(parms, 0, sizeof(*parms));
1503
1504         parms->iph.version = 4;
1505         parms->iph.protocol = IPPROTO_IPV6;
1506         parms->iph.ihl = 5;
1507         parms->iph.ttl = 64;
1508
1509         if (!data)
1510                 return;
1511
1512         if (data[IFLA_IPTUN_LINK])
1513                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1514
1515         if (data[IFLA_IPTUN_LOCAL])
1516                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1517
1518         if (data[IFLA_IPTUN_REMOTE])
1519                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1520
1521         if (data[IFLA_IPTUN_TTL]) {
1522                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1523                 if (parms->iph.ttl)
1524                         parms->iph.frag_off = htons(IP_DF);
1525         }
1526
1527         if (data[IFLA_IPTUN_TOS])
1528                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1529
1530         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1531                 parms->iph.frag_off = htons(IP_DF);
1532
1533         if (data[IFLA_IPTUN_FLAGS])
1534                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1535
1536         if (data[IFLA_IPTUN_PROTO])
1537                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1538
1539         if (data[IFLA_IPTUN_FWMARK])
1540                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1541 }
1542
1543 /* This function returns true when ENCAP attributes are present in the nl msg */
1544 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1545                                       struct ip_tunnel_encap *ipencap)
1546 {
1547         bool ret = false;
1548
1549         memset(ipencap, 0, sizeof(*ipencap));
1550
1551         if (!data)
1552                 return ret;
1553
1554         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1555                 ret = true;
1556                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1557         }
1558
1559         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1560                 ret = true;
1561                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1562         }
1563
1564         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1565                 ret = true;
1566                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1567         }
1568
1569         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1570                 ret = true;
1571                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1572         }
1573
1574         return ret;
1575 }
1576
1577 #ifdef CONFIG_IPV6_SIT_6RD
1578 /* This function returns true when 6RD attributes are present in the nl msg */
1579 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1580                                     struct ip_tunnel_6rd *ip6rd)
1581 {
1582         bool ret = false;
1583         memset(ip6rd, 0, sizeof(*ip6rd));
1584
1585         if (!data)
1586                 return ret;
1587
1588         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1589                 ret = true;
1590                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1591         }
1592
1593         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1594                 ret = true;
1595                 ip6rd->relay_prefix =
1596                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1597         }
1598
1599         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1600                 ret = true;
1601                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1602         }
1603
1604         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1605                 ret = true;
1606                 ip6rd->relay_prefixlen =
1607                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1608         }
1609
1610         return ret;
1611 }
1612 #endif
1613
1614 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1615                          struct nlattr *tb[], struct nlattr *data[],
1616                          struct netlink_ext_ack *extack)
1617 {
1618         struct net *net = dev_net(dev);
1619         struct ip_tunnel *nt;
1620         struct ip_tunnel_encap ipencap;
1621 #ifdef CONFIG_IPV6_SIT_6RD
1622         struct ip_tunnel_6rd ip6rd;
1623 #endif
1624         int err;
1625
1626         nt = netdev_priv(dev);
1627
1628         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1629                 err = ip_tunnel_encap_setup(nt, &ipencap);
1630                 if (err < 0)
1631                         return err;
1632         }
1633
1634         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1635
1636         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1637                 return -EEXIST;
1638
1639         err = ipip6_tunnel_create(dev);
1640         if (err < 0)
1641                 return err;
1642
1643         if (tb[IFLA_MTU]) {
1644                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1645
1646                 if (mtu >= IPV6_MIN_MTU &&
1647                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1648                         dev->mtu = mtu;
1649         }
1650
1651 #ifdef CONFIG_IPV6_SIT_6RD
1652         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1653                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1654                 if (err < 0)
1655                         unregister_netdevice_queue(dev, NULL);
1656         }
1657 #endif
1658
1659         return err;
1660 }
1661
1662 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1663                             struct nlattr *data[],
1664                             struct netlink_ext_ack *extack)
1665 {
1666         struct ip_tunnel *t = netdev_priv(dev);
1667         struct ip_tunnel_parm p;
1668         struct ip_tunnel_encap ipencap;
1669         struct net *net = t->net;
1670         struct sit_net *sitn = net_generic(net, sit_net_id);
1671 #ifdef CONFIG_IPV6_SIT_6RD
1672         struct ip_tunnel_6rd ip6rd;
1673 #endif
1674         __u32 fwmark = t->fwmark;
1675         int err;
1676
1677         if (dev == sitn->fb_tunnel_dev)
1678                 return -EINVAL;
1679
1680         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1681                 err = ip_tunnel_encap_setup(t, &ipencap);
1682                 if (err < 0)
1683                         return err;
1684         }
1685
1686         ipip6_netlink_parms(data, &p, &fwmark);
1687
1688         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1689             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1690                 return -EINVAL;
1691
1692         t = ipip6_tunnel_locate(net, &p, 0);
1693
1694         if (t) {
1695                 if (t->dev != dev)
1696                         return -EEXIST;
1697         } else
1698                 t = netdev_priv(dev);
1699
1700         ipip6_tunnel_update(t, &p, fwmark);
1701
1702 #ifdef CONFIG_IPV6_SIT_6RD
1703         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1704                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1705 #endif
1706
1707         return 0;
1708 }
1709
1710 static size_t ipip6_get_size(const struct net_device *dev)
1711 {
1712         return
1713                 /* IFLA_IPTUN_LINK */
1714                 nla_total_size(4) +
1715                 /* IFLA_IPTUN_LOCAL */
1716                 nla_total_size(4) +
1717                 /* IFLA_IPTUN_REMOTE */
1718                 nla_total_size(4) +
1719                 /* IFLA_IPTUN_TTL */
1720                 nla_total_size(1) +
1721                 /* IFLA_IPTUN_TOS */
1722                 nla_total_size(1) +
1723                 /* IFLA_IPTUN_PMTUDISC */
1724                 nla_total_size(1) +
1725                 /* IFLA_IPTUN_FLAGS */
1726                 nla_total_size(2) +
1727                 /* IFLA_IPTUN_PROTO */
1728                 nla_total_size(1) +
1729 #ifdef CONFIG_IPV6_SIT_6RD
1730                 /* IFLA_IPTUN_6RD_PREFIX */
1731                 nla_total_size(sizeof(struct in6_addr)) +
1732                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1733                 nla_total_size(4) +
1734                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1735                 nla_total_size(2) +
1736                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1737                 nla_total_size(2) +
1738 #endif
1739                 /* IFLA_IPTUN_ENCAP_TYPE */
1740                 nla_total_size(2) +
1741                 /* IFLA_IPTUN_ENCAP_FLAGS */
1742                 nla_total_size(2) +
1743                 /* IFLA_IPTUN_ENCAP_SPORT */
1744                 nla_total_size(2) +
1745                 /* IFLA_IPTUN_ENCAP_DPORT */
1746                 nla_total_size(2) +
1747                 /* IFLA_IPTUN_FWMARK */
1748                 nla_total_size(4) +
1749                 0;
1750 }
1751
1752 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1753 {
1754         struct ip_tunnel *tunnel = netdev_priv(dev);
1755         struct ip_tunnel_parm *parm = &tunnel->parms;
1756
1757         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1758             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1759             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1760             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1761             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1762             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1763                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1764             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1765             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1766             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1767                 goto nla_put_failure;
1768
1769 #ifdef CONFIG_IPV6_SIT_6RD
1770         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1771                              &tunnel->ip6rd.prefix) ||
1772             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1773                             tunnel->ip6rd.relay_prefix) ||
1774             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1775                         tunnel->ip6rd.prefixlen) ||
1776             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1777                         tunnel->ip6rd.relay_prefixlen))
1778                 goto nla_put_failure;
1779 #endif
1780
1781         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1782                         tunnel->encap.type) ||
1783             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1784                         tunnel->encap.sport) ||
1785             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1786                         tunnel->encap.dport) ||
1787             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1788                         tunnel->encap.flags))
1789                 goto nla_put_failure;
1790
1791         return 0;
1792
1793 nla_put_failure:
1794         return -EMSGSIZE;
1795 }
1796
1797 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1798         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1799         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1800         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1801         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1802         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1803         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1804         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1805         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1806 #ifdef CONFIG_IPV6_SIT_6RD
1807         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1808         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1809         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1810         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1811 #endif
1812         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1813         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1814         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1815         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1816         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1817 };
1818
1819 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1820 {
1821         struct net *net = dev_net(dev);
1822         struct sit_net *sitn = net_generic(net, sit_net_id);
1823
1824         if (dev != sitn->fb_tunnel_dev)
1825                 unregister_netdevice_queue(dev, head);
1826 }
1827
1828 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1829         .kind           = "sit",
1830         .maxtype        = IFLA_IPTUN_MAX,
1831         .policy         = ipip6_policy,
1832         .priv_size      = sizeof(struct ip_tunnel),
1833         .setup          = ipip6_tunnel_setup,
1834         .validate       = ipip6_validate,
1835         .newlink        = ipip6_newlink,
1836         .changelink     = ipip6_changelink,
1837         .get_size       = ipip6_get_size,
1838         .fill_info      = ipip6_fill_info,
1839         .dellink        = ipip6_dellink,
1840         .get_link_net   = ip_tunnel_get_link_net,
1841 };
1842
1843 static struct xfrm_tunnel sit_handler __read_mostly = {
1844         .handler        =       ipip6_rcv,
1845         .err_handler    =       ipip6_err,
1846         .priority       =       1,
1847 };
1848
1849 static struct xfrm_tunnel ipip_handler __read_mostly = {
1850         .handler        =       ipip_rcv,
1851         .err_handler    =       ipip6_err,
1852         .priority       =       2,
1853 };
1854
1855 #if IS_ENABLED(CONFIG_MPLS)
1856 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1857         .handler        =       mplsip_rcv,
1858         .err_handler    =       ipip6_err,
1859         .priority       =       2,
1860 };
1861 #endif
1862
1863 static void __net_exit sit_destroy_tunnels(struct net *net,
1864                                            struct list_head *head)
1865 {
1866         struct sit_net *sitn = net_generic(net, sit_net_id);
1867         struct net_device *dev, *aux;
1868         int prio;
1869
1870         for_each_netdev_safe(net, dev, aux)
1871                 if (dev->rtnl_link_ops == &sit_link_ops)
1872                         unregister_netdevice_queue(dev, head);
1873
1874         for (prio = 0; prio < 4; prio++) {
1875                 int h;
1876                 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1877                         struct ip_tunnel *t;
1878
1879                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1880                         while (t) {
1881                                 /* If dev is in the same netns, it has already
1882                                  * been added to the list by the previous loop.
1883                                  */
1884                                 if (!net_eq(dev_net(t->dev), net))
1885                                         unregister_netdevice_queue(t->dev,
1886                                                                    head);
1887                                 t = rtnl_dereference(t->next);
1888                         }
1889                 }
1890         }
1891 }
1892
1893 static int __net_init sit_init_net(struct net *net)
1894 {
1895         struct sit_net *sitn = net_generic(net, sit_net_id);
1896         struct ip_tunnel *t;
1897         int err;
1898
1899         sitn->tunnels[0] = sitn->tunnels_wc;
1900         sitn->tunnels[1] = sitn->tunnels_l;
1901         sitn->tunnels[2] = sitn->tunnels_r;
1902         sitn->tunnels[3] = sitn->tunnels_r_l;
1903
1904         if (!net_has_fallback_tunnels(net))
1905                 return 0;
1906
1907         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1908                                            NET_NAME_UNKNOWN,
1909                                            ipip6_tunnel_setup);
1910         if (!sitn->fb_tunnel_dev) {
1911                 err = -ENOMEM;
1912                 goto err_alloc_dev;
1913         }
1914         dev_net_set(sitn->fb_tunnel_dev, net);
1915         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1916         /* FB netdevice is special: we have one, and only one per netns.
1917          * Allowing to move it to another netns is clearly unsafe.
1918          */
1919         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1920
1921         err = register_netdev(sitn->fb_tunnel_dev);
1922         if (err)
1923                 goto err_reg_dev;
1924
1925         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1926         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1927
1928         t = netdev_priv(sitn->fb_tunnel_dev);
1929
1930         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1931         return 0;
1932
1933 err_reg_dev:
1934         free_netdev(sitn->fb_tunnel_dev);
1935 err_alloc_dev:
1936         return err;
1937 }
1938
1939 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1940 {
1941         LIST_HEAD(list);
1942         struct net *net;
1943
1944         rtnl_lock();
1945         list_for_each_entry(net, net_list, exit_list)
1946                 sit_destroy_tunnels(net, &list);
1947
1948         unregister_netdevice_many(&list);
1949         rtnl_unlock();
1950 }
1951
1952 static struct pernet_operations sit_net_ops = {
1953         .init = sit_init_net,
1954         .exit_batch = sit_exit_batch_net,
1955         .id   = &sit_net_id,
1956         .size = sizeof(struct sit_net),
1957 };
1958
1959 static void __exit sit_cleanup(void)
1960 {
1961         rtnl_link_unregister(&sit_link_ops);
1962         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1963         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1964 #if IS_ENABLED(CONFIG_MPLS)
1965         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1966 #endif
1967
1968         unregister_pernet_device(&sit_net_ops);
1969         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1970 }
1971
1972 static int __init sit_init(void)
1973 {
1974         int err;
1975
1976         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1977
1978         err = register_pernet_device(&sit_net_ops);
1979         if (err < 0)
1980                 return err;
1981         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1982         if (err < 0) {
1983                 pr_info("%s: can't register ip6ip4\n", __func__);
1984                 goto xfrm_tunnel_failed;
1985         }
1986         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1987         if (err < 0) {
1988                 pr_info("%s: can't register ip4ip4\n", __func__);
1989                 goto xfrm_tunnel4_failed;
1990         }
1991 #if IS_ENABLED(CONFIG_MPLS)
1992         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1993         if (err < 0) {
1994                 pr_info("%s: can't register mplsip\n", __func__);
1995                 goto xfrm_tunnel_mpls_failed;
1996         }
1997 #endif
1998         err = rtnl_link_register(&sit_link_ops);
1999         if (err < 0)
2000                 goto rtnl_link_failed;
2001
2002 out:
2003         return err;
2004
2005 rtnl_link_failed:
2006 #if IS_ENABLED(CONFIG_MPLS)
2007         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
2008 xfrm_tunnel_mpls_failed:
2009 #endif
2010         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
2011 xfrm_tunnel4_failed:
2012         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
2013 xfrm_tunnel_failed:
2014         unregister_pernet_device(&sit_net_ops);
2015         goto out;
2016 }
2017
2018 module_init(sit_init);
2019 module_exit(sit_cleanup);
2020 MODULE_LICENSE("GPL");
2021 MODULE_ALIAS_RTNL_LINK("sit");
2022 MODULE_ALIAS_NETDEV("sit0");