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 /*
  87  * Must be invoked with rcu_read_lock
  88  */
  89 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  90                                              struct net_device *dev,
  91                                              __be32 remote, __be32 local,
  92                                              int sifindex)
  93 {
  94         unsigned int h0 = HASH(remote);
  95         unsigned int h1 = HASH(local);
  96         struct ip_tunnel *t;
  97         struct sit_net *sitn = net_generic(net, sit_net_id);
  98         int ifindex = dev ? dev->ifindex : 0;
  99
 100         for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 101                 if (local == t->parms.iph.saddr &&
 102                     remote == t->parms.iph.daddr &&
 103                     (!dev || !t->parms.link || ifindex == t->parms.link ||
 104                      sifindex == 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 || 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_l[h1]) {
 116                 if (local == t->parms.iph.saddr &&
 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         t = rcu_dereference(sitn->tunnels_wc[0]);
 123         if (t && (t->dev->flags & IFF_UP))
 124                 return t;
 125         return NULL;
 126 }
 127
 128 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 129                 struct ip_tunnel_parm *parms)
 130 {
 131         __be32 remote = parms->iph.daddr;
 132         __be32 local = parms->iph.saddr;
 133         unsigned int h = 0;
 134         int prio = 0;
 135
 136         if (remote) {
 137                 prio |= 2;
 138                 h ^= HASH(remote);
 139         }
 140         if (local) {
 141                 prio |= 1;
 142                 h ^= HASH(local);
 143         }
 144         return &sitn->tunnels[prio][h];
 145 }
 146
 147 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 148                 struct ip_tunnel *t)
 149 {
 150         return __ipip6_bucket(sitn, &t->parms);
 151 }
 152
 153 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 154 {
 155         struct ip_tunnel __rcu **tp;
 156         struct ip_tunnel *iter;
 157
 158         for (tp = ipip6_bucket(sitn, t);
 159              (iter = rtnl_dereference(*tp)) != NULL;
 160              tp = &iter->next) {
 161                 if (t == iter) {
 162                         rcu_assign_pointer(*tp, t->next);
 163                         break;
 164                 }
 165         }
 166 }
 167
 168 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 169 {
 170         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 171
 172         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 173         rcu_assign_pointer(*tp, t);
 174 }
 175
 176 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 177 {
 178 #ifdef CONFIG_IPV6_SIT_6RD
 179         struct ip_tunnel *t = netdev_priv(dev);
 180
 181         if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 182                 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 183                 t->ip6rd.relay_prefix = 0;
 184                 t->ip6rd.prefixlen = 16;
 185                 t->ip6rd.relay_prefixlen = 0;
 186         } else {
 187                 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 188                 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 189         }
 190 #endif
 191 }
 192
 193 static int ipip6_tunnel_create(struct net_device *dev)
 194 {
 195         struct ip_tunnel *t = netdev_priv(dev);
 196         struct net *net = dev_net(dev);
 197         struct sit_net *sitn = net_generic(net, sit_net_id);
 198         int err;
 199
 200         memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 201         memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 202
 203         if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 204                 dev->priv_flags |= IFF_ISATAP;
 205
 206         dev->rtnl_link_ops = &sit_link_ops;
 207
 208         err = register_netdevice(dev);
 209         if (err < 0)
 210                 goto out;
 211
 212         ipip6_tunnel_clone_6rd(dev, sitn);
 213
 214         ipip6_tunnel_link(sitn, t);
 215         return 0;
 216
 217 out:
 218         return err;
 219 }
 220
 221 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 222                 struct ip_tunnel_parm *parms, int create)
 223 {
 224         __be32 remote = parms->iph.daddr;
 225         __be32 local = parms->iph.saddr;
 226         struct ip_tunnel *t, *nt;
 227         struct ip_tunnel __rcu **tp;
 228         struct net_device *dev;
 229         char name[IFNAMSIZ];
 230         struct sit_net *sitn = net_generic(net, sit_net_id);
 231
 232         for (tp = __ipip6_bucket(sitn, parms);
 233             (t = rtnl_dereference(*tp)) != NULL;
 234              tp = &t->next) {
 235                 if (local == t->parms.iph.saddr &&
 236                     remote == t->parms.iph.daddr &&
 237                     parms->link == t->parms.link) {
 238                         if (create)
 239                                 return NULL;
 240                         else
 241                                 return t;
 242                 }
 243         }
 244         if (!create)
 245                 goto failed;
 246
 247         if (parms->name[0]) {
 248                 if (!dev_valid_name(parms->name))
 249                         goto failed;
 250                 strlcpy(name, parms->name, IFNAMSIZ);
 251         } else {
 252                 strcpy(name, "sit%d");
 253         }
 254         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 255                            ipip6_tunnel_setup);
 256         if (!dev)
 257                 return NULL;
 258
 259         dev_net_set(dev, net);
 260
 261         nt = netdev_priv(dev);
 262
 263         nt->parms = *parms;
 264         if (ipip6_tunnel_create(dev) < 0)
 265                 goto failed_free;
 266
 267         return nt;
 268
 269 failed_free:
 270         free_netdev(dev);
 271 failed:
 272         return NULL;
 273 }
 274
 275 #define for_each_prl_rcu(start)                 \
 276         for (prl = rcu_dereference(start);      \
 277              prl;                               \
 278              prl = rcu_dereference(prl->next))
 279
 280 static struct ip_tunnel_prl_entry *
 281 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 282 {
 283         struct ip_tunnel_prl_entry *prl;
 284
 285         for_each_prl_rcu(t->prl)
 286                 if (prl->addr == addr)
 287                         break;
 288         return prl;
 289
 290 }
 291
 292 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 293                                 struct ip_tunnel_prl __user *a)
 294 {
 295         struct ip_tunnel_prl kprl, *kp;
 296         struct ip_tunnel_prl_entry *prl;
 297         unsigned int cmax, c = 0, ca, len;
 298         int ret = 0;
 299
 300         if (copy_from_user(&kprl, a, sizeof(kprl)))
 301                 return -EFAULT;
 302         cmax = kprl.datalen / sizeof(kprl);
 303         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 304                 cmax = 1;
 305
 306         /* For simple GET or for root users,
 307          * we try harder to allocate.
 308          */
 309         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 310                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 311                 NULL;
 312
 313         ca = min(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         rcu_read_lock();
 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
 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 out:
 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         dst_cache_reset(&tunnel->dst_cache);
 479         dev_put(dev);
 480 }
 481
 482 static int ipip6_err(struct sk_buff *skb, u32 info)
 483 {
 484         const struct iphdr *iph = (const struct iphdr *)skb->data;
 485         const int type = icmp_hdr(skb)->type;
 486         const int code = icmp_hdr(skb)->code;
 487         unsigned int data_len = 0;
 488         struct ip_tunnel *t;
 489         int sifindex;
 490         int err;
 491
 492         switch (type) {
 493         default:
 494         case ICMP_PARAMETERPROB:
 495                 return 0;
 496
 497         case ICMP_DEST_UNREACH:
 498                 switch (code) {
 499                 case ICMP_SR_FAILED:
 500                         /* Impossible event. */
 501                         return 0;
 502                 default:
 503                         /* All others are translated to HOST_UNREACH.
 504                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 505                            I believe they are just ether pollution. --ANK
 506                          */
 507                         break;
 508                 }
 509                 break;
 510         case ICMP_TIME_EXCEEDED:
 511                 if (code != ICMP_EXC_TTL)
 512                         return 0;
 513                 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 514                 break;
 515         case ICMP_REDIRECT:
 516                 break;
 517         }
 518
 519         err = -ENOENT;
 520
 521         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 522         t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 523                                 iph->daddr, iph->saddr, sifindex);
 524         if (!t)
 525                 goto out;
 526
 527         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 528                 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 529                                  t->parms.link, iph->protocol);
 530                 err = 0;
 531                 goto out;
 532         }
 533         if (type == ICMP_REDIRECT) {
 534                 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 535                               iph->protocol);
 536                 err = 0;
 537                 goto out;
 538         }
 539
 540         err = 0;
 541         if (__in6_dev_get(skb->dev) &&
 542             !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 543                 goto out;
 544
 545         if (t->parms.iph.daddr == 0)
 546                 goto out;
 547
 548         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 549                 goto out;
 550
 551         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 552                 t->err_count++;
 553         else
 554                 t->err_count = 1;
 555         t->err_time = jiffies;
 556 out:
 557         return err;
 558 }
 559
 560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 561                                   const struct in6_addr *v6addr)
 562 {
 563         __be32 v4embed = 0;
 564         if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 565                 return true;
 566         return false;
 567 }
 568
 569 /* Checks if an address matches an address on the tunnel interface.
 570  * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 571  * Long story:
 572  * This function is called after we considered the packet as spoofed
 573  * in is_spoofed_6rd.
 574  * We may have a router that is doing NAT for proto 41 packets
 575  * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 576  * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 577  * function will return true, dropping the packet.
 578  * But, we can still check if is spoofed against the IP
 579  * addresses associated with the interface.
 580  */
 581 static bool only_dnatted(const struct ip_tunnel *tunnel,
 582         const struct in6_addr *v6dst)
 583 {
 584         int prefix_len;
 585
 586 #ifdef CONFIG_IPV6_SIT_6RD
 587         prefix_len = tunnel->ip6rd.prefixlen + 32
 588                 - tunnel->ip6rd.relay_prefixlen;
 589 #else
 590         prefix_len = 48;
 591 #endif
 592         return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 593 }
 594
 595 /* Returns true if a packet is spoofed */
 596 static bool packet_is_spoofed(struct sk_buff *skb,
 597                               const struct iphdr *iph,
 598                               struct ip_tunnel *tunnel)
 599 {
 600         const struct ipv6hdr *ipv6h;
 601
 602         if (tunnel->dev->priv_flags & IFF_ISATAP) {
 603                 if (!isatap_chksrc(skb, iph, tunnel))
 604                         return true;
 605
 606                 return false;
 607         }
 608
 609         if (tunnel->dev->flags & IFF_POINTOPOINT)
 610                 return false;
 611
 612         ipv6h = ipv6_hdr(skb);
 613
 614         if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 615                 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 616                                      &iph->saddr, &ipv6h->saddr,
 617                                      &iph->daddr, &ipv6h->daddr);
 618                 return true;
 619         }
 620
 621         if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 622                 return false;
 623
 624         if (only_dnatted(tunnel, &ipv6h->daddr))
 625                 return false;
 626
 627         net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 628                              &iph->saddr, &ipv6h->saddr,
 629                              &iph->daddr, &ipv6h->daddr);
 630         return true;
 631 }
 632
 633 static int ipip6_rcv(struct sk_buff *skb)
 634 {
 635         const struct iphdr *iph = ip_hdr(skb);
 636         struct ip_tunnel *tunnel;
 637         int sifindex;
 638         int err;
 639
 640         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 641         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 642                                      iph->saddr, iph->daddr, sifindex);
 643         if (tunnel) {
 644                 struct pcpu_sw_netstats *tstats;
 645
 646                 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 647                     tunnel->parms.iph.protocol != 0)
 648                         goto out;
 649
 650                 skb->mac_header = skb->network_header;
 651                 skb_reset_network_header(skb);
 652                 IPCB(skb)->flags = 0;
 653                 skb->dev = tunnel->dev;
 654
 655                 if (packet_is_spoofed(skb, iph, tunnel)) {
 656                         tunnel->dev->stats.rx_errors++;
 657                         goto out;
 658                 }
 659
 660                 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 661                     !net_eq(tunnel->net, dev_net(tunnel->dev))))
 662                         goto out;
 663
 664                 /* skb can be uncloned in iptunnel_pull_header, so
 665                  * old iph is no longer valid
 666                  */
 667                 iph = (const struct iphdr *)skb_mac_header(skb);
 668                 err = IP_ECN_decapsulate(iph, skb);
 669                 if (unlikely(err)) {
 670                         if (log_ecn_error)
 671                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 672                                                      &iph->saddr, iph->tos);
 673                         if (err > 1) {
 674                                 ++tunnel->dev->stats.rx_frame_errors;
 675                                 ++tunnel->dev->stats.rx_errors;
 676                                 goto out;
 677                         }
 678                 }
 679
 680                 tstats = this_cpu_ptr(tunnel->dev->tstats);
 681                 u64_stats_update_begin(&tstats->syncp);
 682                 tstats->rx_packets++;
 683                 tstats->rx_bytes += skb->len;
 684                 u64_stats_update_end(&tstats->syncp);
 685
 686                 netif_rx(skb);
 687
 688                 return 0;
 689         }
 690
 691         /* no tunnel matched,  let upstream know, ipsec may handle it */
 692         return 1;
 693 out:
 694         kfree_skb(skb);
 695         return 0;
 696 }
 697
 698 static const struct tnl_ptk_info ipip_tpi = {
 699         /* no tunnel info required for ipip. */
 700         .proto = htons(ETH_P_IP),
 701 };
 702
 703 #if IS_ENABLED(CONFIG_MPLS)
 704 static const struct tnl_ptk_info mplsip_tpi = {
 705         /* no tunnel info required for mplsip. */
 706         .proto = htons(ETH_P_MPLS_UC),
 707 };
 708 #endif
 709
 710 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 711 {
 712         const struct iphdr *iph;
 713         struct ip_tunnel *tunnel;
 714         int sifindex;
 715
 716         sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 717
 718         iph = ip_hdr(skb);
 719         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 720                                      iph->saddr, iph->daddr, sifindex);
 721         if (tunnel) {
 722                 const struct tnl_ptk_info *tpi;
 723
 724                 if (tunnel->parms.iph.protocol != ipproto &&
 725                     tunnel->parms.iph.protocol != 0)
 726                         goto drop;
 727
 728                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 729                         goto drop;
 730 #if IS_ENABLED(CONFIG_MPLS)
 731                 if (ipproto == IPPROTO_MPLS)
 732                         tpi = &mplsip_tpi;
 733                 else
 734 #endif
 735                         tpi = &ipip_tpi;
 736                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 737                         goto drop;
 738                 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 739         }
 740
 741         return 1;
 742
 743 drop:
 744         kfree_skb(skb);
 745         return 0;
 746 }
 747
 748 static int ipip_rcv(struct sk_buff *skb)
 749 {
 750         return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 751 }
 752
 753 #if IS_ENABLED(CONFIG_MPLS)
 754 static int mplsip_rcv(struct sk_buff *skb)
 755 {
 756         return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 757 }
 758 #endif
 759
 760 /*
 761  * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 762  * stores the embedded IPv4 address in v4dst and returns true.
 763  */
 764 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 765                       __be32 *v4dst)
 766 {
 767 #ifdef CONFIG_IPV6_SIT_6RD
 768         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 769                               tunnel->ip6rd.prefixlen)) {
 770                 unsigned int pbw0, pbi0;
 771                 int pbi1;
 772                 u32 d;
 773
 774                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
 775                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 776
 777                 d = tunnel->ip6rd.relay_prefixlen < 32 ?
 778                         (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 779                     tunnel->ip6rd.relay_prefixlen : 0;
 780
 781                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 782                 if (pbi1 > 0)
 783                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 784                              (32 - pbi1);
 785
 786                 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 787                 return true;
 788         }
 789 #else
 790         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 791                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 792                 memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 793                 return true;
 794         }
 795 #endif
 796         return false;
 797 }
 798
 799 static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 800                              const struct in6_addr *v6dst)
 801 {
 802         __be32 dst = 0;
 803         check_6rd(tunnel, v6dst, &dst);
 804         return dst;
 805 }
 806
 807 /*
 808  *      This function assumes it is being called from dev_queue_xmit()
 809  *      and that skb is filled properly by that function.
 810  */
 811
 812 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 813                                      struct net_device *dev)
 814 {
 815         struct ip_tunnel *tunnel = netdev_priv(dev);
 816         const struct iphdr  *tiph = &tunnel->parms.iph;
 817         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 818         u8     tos = tunnel->parms.iph.tos;
 819         __be16 df = tiph->frag_off;
 820         struct rtable *rt;              /* Route to the other host */
 821         struct net_device *tdev;        /* Device to other host */
 822         unsigned int max_headroom;      /* The extra header space needed */
 823         __be32 dst = tiph->daddr;
 824         struct flowi4 fl4;
 825         int    mtu;
 826         const struct in6_addr *addr6;
 827         int addr_type;
 828         u8 ttl;
 829         u8 protocol = IPPROTO_IPV6;
 830         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 831
 832         if (tos == 1)
 833                 tos = ipv6_get_dsfield(iph6);
 834
 835         /* ISATAP (RFC4214) - must come before 6to4 */
 836         if (dev->priv_flags & IFF_ISATAP) {
 837                 struct neighbour *neigh = NULL;
 838                 bool do_tx_error = false;
 839
 840                 if (skb_dst(skb))
 841                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 842
 843                 if (!neigh) {
 844                         net_dbg_ratelimited("nexthop == NULL\n");
 845                         goto tx_error;
 846                 }
 847
 848                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 849                 addr_type = ipv6_addr_type(addr6);
 850
 851                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 852                      ipv6_addr_is_isatap(addr6))
 853                         dst = addr6->s6_addr32[3];
 854                 else
 855                         do_tx_error = true;
 856
 857                 neigh_release(neigh);
 858                 if (do_tx_error)
 859                         goto tx_error;
 860         }
 861
 862         if (!dst)
 863                 dst = try_6rd(tunnel, &iph6->daddr);
 864
 865         if (!dst) {
 866                 struct neighbour *neigh = NULL;
 867                 bool do_tx_error = false;
 868
 869                 if (skb_dst(skb))
 870                         neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 871
 872                 if (!neigh) {
 873                         net_dbg_ratelimited("nexthop == NULL\n");
 874                         goto tx_error;
 875                 }
 876
 877                 addr6 = (const struct in6_addr *)&neigh->primary_key;
 878                 addr_type = ipv6_addr_type(addr6);
 879
 880                 if (addr_type == IPV6_ADDR_ANY) {
 881                         addr6 = &ipv6_hdr(skb)->daddr;
 882                         addr_type = ipv6_addr_type(addr6);
 883                 }
 884
 885                 if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 886                         dst = addr6->s6_addr32[3];
 887                 else
 888                         do_tx_error = true;
 889
 890                 neigh_release(neigh);
 891                 if (do_tx_error)
 892                         goto tx_error;
 893         }
 894
 895         flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 896                            RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 897                            0, dst, tiph->saddr, 0, 0,
 898                            sock_net_uid(tunnel->net, NULL));
 899
 900         rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 901         if (!rt) {
 902                 rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 903                 if (IS_ERR(rt)) {
 904                         dev->stats.tx_carrier_errors++;
 905                         goto tx_error_icmp;
 906                 }
 907                 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 908         }
 909
 910         if (rt->rt_type != RTN_UNICAST) {
 911                 ip_rt_put(rt);
 912                 dev->stats.tx_carrier_errors++;
 913                 goto tx_error_icmp;
 914         }
 915         tdev = rt->dst.dev;
 916
 917         if (tdev == dev) {
 918                 ip_rt_put(rt);
 919                 dev->stats.collisions++;
 920                 goto tx_error;
 921         }
 922
 923         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 924                 ip_rt_put(rt);
 925                 goto tx_error;
 926         }
 927
 928         if (df) {
 929                 mtu = dst_mtu(&rt->dst) - t_hlen;
 930
 931                 if (mtu < 68) {
 932                         dev->stats.collisions++;
 933                         ip_rt_put(rt);
 934                         goto tx_error;
 935                 }
 936
 937                 if (mtu < IPV6_MIN_MTU) {
 938                         mtu = IPV6_MIN_MTU;
 939                         df = 0;
 940                 }
 941
 942                 if (tunnel->parms.iph.daddr)
 943                         skb_dst_update_pmtu_no_confirm(skb, mtu);
 944
 945                 if (skb->len > mtu && !skb_is_gso(skb)) {
 946                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 947                         ip_rt_put(rt);
 948                         goto tx_error;
 949                 }
 950         }
 951
 952         if (tunnel->err_count > 0) {
 953                 if (time_before(jiffies,
 954                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 955                         tunnel->err_count--;
 956                         dst_link_failure(skb);
 957                 } else
 958                         tunnel->err_count = 0;
 959         }
 960
 961         /*
 962          * Okay, now see if we can stuff it in the buffer as-is.
 963          */
 964         max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 965
 966         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 967             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 968                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 969                 if (!new_skb) {
 970                         ip_rt_put(rt);
 971                         dev->stats.tx_dropped++;
 972                         kfree_skb(skb);
 973                         return NETDEV_TX_OK;
 974                 }
 975                 if (skb->sk)
 976                         skb_set_owner_w(new_skb, skb->sk);
 977                 dev_kfree_skb(skb);
 978                 skb = new_skb;
 979                 iph6 = ipv6_hdr(skb);
 980         }
 981         ttl = tiph->ttl;
 982         if (ttl == 0)
 983                 ttl = iph6->hop_limit;
 984         tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 985
 986         if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 987                 ip_rt_put(rt);
 988                 goto tx_error;
 989         }
 990
 991         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 992
 993         iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 994                       df, !net_eq(tunnel->net, dev_net(dev)));
 995         return NETDEV_TX_OK;
 996
 997 tx_error_icmp:
 998         dst_link_failure(skb);
 999 tx_error:
1000         kfree_skb(skb);
1001         dev->stats.tx_errors++;
1002         return NETDEV_TX_OK;
1003 }
1004
1005 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1006                                      struct net_device *dev, u8 ipproto)
1007 {
1008         struct ip_tunnel *tunnel = netdev_priv(dev);
1009         const struct iphdr  *tiph = &tunnel->parms.iph;
1010
1011         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1012                 goto tx_error;
1013
1014         skb_set_inner_ipproto(skb, ipproto);
1015
1016         ip_tunnel_xmit(skb, dev, tiph, ipproto);
1017         return NETDEV_TX_OK;
1018 tx_error:
1019         kfree_skb(skb);
1020         dev->stats.tx_errors++;
1021         return NETDEV_TX_OK;
1022 }
1023
1024 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1025                                    struct net_device *dev)
1026 {
1027         if (!pskb_inet_may_pull(skb))
1028                 goto tx_err;
1029
1030         switch (skb->protocol) {
1031         case htons(ETH_P_IP):
1032                 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1033                 break;
1034         case htons(ETH_P_IPV6):
1035                 ipip6_tunnel_xmit(skb, dev);
1036                 break;
1037 #if IS_ENABLED(CONFIG_MPLS)
1038         case htons(ETH_P_MPLS_UC):
1039                 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1040                 break;
1041 #endif
1042         default:
1043                 goto tx_err;
1044         }
1045
1046         return NETDEV_TX_OK;
1047
1048 tx_err:
1049         dev->stats.tx_errors++;
1050         kfree_skb(skb);
1051         return NETDEV_TX_OK;
1052
1053 }
1054
1055 static void ipip6_tunnel_bind_dev(struct net_device *dev)
1056 {
1057         struct net_device *tdev = NULL;
1058         struct ip_tunnel *tunnel;
1059         const struct iphdr *iph;
1060         struct flowi4 fl4;
1061
1062         tunnel = netdev_priv(dev);
1063         iph = &tunnel->parms.iph;
1064
1065         if (iph->daddr) {
1066                 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1067                                                           NULL,
1068                                                           iph->daddr, iph->saddr,
1069                                                           0, 0,
1070                                                           IPPROTO_IPV6,
1071                                                           RT_TOS(iph->tos),
1072                                                           tunnel->parms.link);
1073
1074                 if (!IS_ERR(rt)) {
1075                         tdev = rt->dst.dev;
1076                         ip_rt_put(rt);
1077                 }
1078                 dev->flags |= IFF_POINTOPOINT;
1079         }
1080
1081         if (!tdev && tunnel->parms.link)
1082                 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1083
1084         if (tdev && !netif_is_l3_master(tdev)) {
1085                 int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1086
1087                 dev->mtu = tdev->mtu - t_hlen;
1088                 if (dev->mtu < IPV6_MIN_MTU)
1089                         dev->mtu = IPV6_MIN_MTU;
1090         }
1091 }
1092
1093 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1094                                 __u32 fwmark)
1095 {
1096         struct net *net = t->net;
1097         struct sit_net *sitn = net_generic(net, sit_net_id);
1098
1099         ipip6_tunnel_unlink(sitn, t);
1100         synchronize_net();
1101         t->parms.iph.saddr = p->iph.saddr;
1102         t->parms.iph.daddr = p->iph.daddr;
1103         memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1104         memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1105         ipip6_tunnel_link(sitn, t);
1106         t->parms.iph.ttl = p->iph.ttl;
1107         t->parms.iph.tos = p->iph.tos;
1108         t->parms.iph.frag_off = p->iph.frag_off;
1109         if (t->parms.link != p->link || t->fwmark != fwmark) {
1110                 t->parms.link = p->link;
1111                 t->fwmark = fwmark;
1112                 ipip6_tunnel_bind_dev(t->dev);
1113         }
1114         dst_cache_reset(&t->dst_cache);
1115         netdev_state_change(t->dev);
1116 }
1117
1118 #ifdef CONFIG_IPV6_SIT_6RD
1119 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1120                                    struct ip_tunnel_6rd *ip6rd)
1121 {
1122         struct in6_addr prefix;
1123         __be32 relay_prefix;
1124
1125         if (ip6rd->relay_prefixlen > 32 ||
1126             ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1127                 return -EINVAL;
1128
1129         ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1130         if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1131                 return -EINVAL;
1132         if (ip6rd->relay_prefixlen)
1133                 relay_prefix = ip6rd->relay_prefix &
1134                                htonl(0xffffffffUL <<
1135                                      (32 - ip6rd->relay_prefixlen));
1136         else
1137                 relay_prefix = 0;
1138         if (relay_prefix != ip6rd->relay_prefix)
1139                 return -EINVAL;
1140
1141         t->ip6rd.prefix = prefix;
1142         t->ip6rd.relay_prefix = relay_prefix;
1143         t->ip6rd.prefixlen = ip6rd->prefixlen;
1144         t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1145         dst_cache_reset(&t->dst_cache);
1146         netdev_state_change(t->dev);
1147         return 0;
1148 }
1149 #endif
1150
1151 static bool ipip6_valid_ip_proto(u8 ipproto)
1152 {
1153         return ipproto == IPPROTO_IPV6 ||
1154                 ipproto == IPPROTO_IPIP ||
1155 #if IS_ENABLED(CONFIG_MPLS)
1156                 ipproto == IPPROTO_MPLS ||
1157 #endif
1158                 ipproto == 0;
1159 }
1160
1161 static int
1162 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1163 {
1164         int err = 0;
1165         struct ip_tunnel_parm p;
1166         struct ip_tunnel_prl prl;
1167         struct ip_tunnel *t = netdev_priv(dev);
1168         struct net *net = t->net;
1169         struct sit_net *sitn = net_generic(net, sit_net_id);
1170 #ifdef CONFIG_IPV6_SIT_6RD
1171         struct ip_tunnel_6rd ip6rd;
1172 #endif
1173
1174         switch (cmd) {
1175         case SIOCGETTUNNEL:
1176 #ifdef CONFIG_IPV6_SIT_6RD
1177         case SIOCGET6RD:
1178 #endif
1179                 if (dev == sitn->fb_tunnel_dev) {
1180                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1181                                 err = -EFAULT;
1182                                 break;
1183                         }
1184                         t = ipip6_tunnel_locate(net, &p, 0);
1185                         if (!t)
1186                                 t = netdev_priv(dev);
1187                 }
1188
1189                 err = -EFAULT;
1190                 if (cmd == SIOCGETTUNNEL) {
1191                         memcpy(&p, &t->parms, sizeof(p));
1192                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1193                                          sizeof(p)))
1194                                 goto done;
1195 #ifdef CONFIG_IPV6_SIT_6RD
1196                 } else {
1197                         ip6rd.prefix = t->ip6rd.prefix;
1198                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1199                         ip6rd.prefixlen = t->ip6rd.prefixlen;
1200                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1201                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1202                                          sizeof(ip6rd)))
1203                                 goto done;
1204 #endif
1205                 }
1206                 err = 0;
1207                 break;
1208
1209         case SIOCADDTUNNEL:
1210         case SIOCCHGTUNNEL:
1211                 err = -EPERM;
1212                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1213                         goto done;
1214
1215                 err = -EFAULT;
1216                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1217                         goto done;
1218
1219                 err = -EINVAL;
1220                 if (!ipip6_valid_ip_proto(p.iph.protocol))
1221                         goto done;
1222                 if (p.iph.version != 4 ||
1223                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1224                         goto done;
1225                 if (p.iph.ttl)
1226                         p.iph.frag_off |= htons(IP_DF);
1227
1228                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1229
1230                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1231                         if (t) {
1232                                 if (t->dev != dev) {
1233                                         err = -EEXIST;
1234                                         break;
1235                                 }
1236                         } else {
1237                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1238                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1239                                         err = -EINVAL;
1240                                         break;
1241                                 }
1242                                 t = netdev_priv(dev);
1243                         }
1244
1245                         ipip6_tunnel_update(t, &p, t->fwmark);
1246                 }
1247
1248                 if (t) {
1249                         err = 0;
1250                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1251                                 err = -EFAULT;
1252                 } else
1253                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1254                 break;
1255
1256         case SIOCDELTUNNEL:
1257                 err = -EPERM;
1258                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1259                         goto done;
1260
1261                 if (dev == sitn->fb_tunnel_dev) {
1262                         err = -EFAULT;
1263                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1264                                 goto done;
1265                         err = -ENOENT;
1266                         t = ipip6_tunnel_locate(net, &p, 0);
1267                         if (!t)
1268                                 goto done;
1269                         err = -EPERM;
1270                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1271                                 goto done;
1272                         dev = t->dev;
1273                 }
1274                 unregister_netdevice(dev);
1275                 err = 0;
1276                 break;
1277
1278         case SIOCGETPRL:
1279                 err = -EINVAL;
1280                 if (dev == sitn->fb_tunnel_dev)
1281                         goto done;
1282                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1283                 break;
1284
1285         case SIOCADDPRL:
1286         case SIOCDELPRL:
1287         case SIOCCHGPRL:
1288                 err = -EPERM;
1289                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1290                         goto done;
1291                 err = -EINVAL;
1292                 if (dev == sitn->fb_tunnel_dev)
1293                         goto done;
1294                 err = -EFAULT;
1295                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1296                         goto done;
1297
1298                 switch (cmd) {
1299                 case SIOCDELPRL:
1300                         err = ipip6_tunnel_del_prl(t, &prl);
1301                         break;
1302                 case SIOCADDPRL:
1303                 case SIOCCHGPRL:
1304                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1305                         break;
1306                 }
1307                 dst_cache_reset(&t->dst_cache);
1308                 netdev_state_change(dev);
1309                 break;
1310
1311 #ifdef CONFIG_IPV6_SIT_6RD
1312         case SIOCADD6RD:
1313         case SIOCCHG6RD:
1314         case SIOCDEL6RD:
1315                 err = -EPERM;
1316                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1317                         goto done;
1318
1319                 err = -EFAULT;
1320                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1321                                    sizeof(ip6rd)))
1322                         goto done;
1323
1324                 if (cmd != SIOCDEL6RD) {
1325                         err = ipip6_tunnel_update_6rd(t, &ip6rd);
1326                         if (err < 0)
1327                                 goto done;
1328                 } else
1329                         ipip6_tunnel_clone_6rd(dev, sitn);
1330
1331                 err = 0;
1332                 break;
1333 #endif
1334
1335         default:
1336                 err = -EINVAL;
1337         }
1338
1339 done:
1340         return err;
1341 }
1342
1343 static const struct net_device_ops ipip6_netdev_ops = {
1344         .ndo_init       = ipip6_tunnel_init,
1345         .ndo_uninit     = ipip6_tunnel_uninit,
1346         .ndo_start_xmit = sit_tunnel_xmit,
1347         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1348         .ndo_get_stats64 = ip_tunnel_get_stats64,
1349         .ndo_get_iflink = ip_tunnel_get_iflink,
1350 };
1351
1352 static void ipip6_dev_free(struct net_device *dev)
1353 {
1354         struct ip_tunnel *tunnel = netdev_priv(dev);
1355
1356         dst_cache_destroy(&tunnel->dst_cache);
1357         free_percpu(dev->tstats);
1358 }
1359
1360 #define SIT_FEATURES (NETIF_F_SG           | \
1361                       NETIF_F_FRAGLIST     | \
1362                       NETIF_F_HIGHDMA      | \
1363                       NETIF_F_GSO_SOFTWARE | \
1364                       NETIF_F_HW_CSUM)
1365
1366 static void ipip6_tunnel_setup(struct net_device *dev)
1367 {
1368         struct ip_tunnel *tunnel = netdev_priv(dev);
1369         int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1370
1371         dev->netdev_ops         = &ipip6_netdev_ops;
1372         dev->needs_free_netdev  = true;
1373         dev->priv_destructor    = ipip6_dev_free;
1374
1375         dev->type               = ARPHRD_SIT;
1376         dev->mtu                = ETH_DATA_LEN - t_hlen;
1377         dev->min_mtu            = IPV6_MIN_MTU;
1378         dev->max_mtu            = IP6_MAX_MTU - t_hlen;
1379         dev->flags              = IFF_NOARP;
1380         netif_keep_dst(dev);
1381         dev->addr_len           = 4;
1382         dev->features           |= NETIF_F_LLTX;
1383         dev->features           |= SIT_FEATURES;
1384         dev->hw_features        |= SIT_FEATURES;
1385 }
1386
1387 static int ipip6_tunnel_init(struct net_device *dev)
1388 {
1389         struct ip_tunnel *tunnel = netdev_priv(dev);
1390         int err;
1391
1392         tunnel->dev = dev;
1393         tunnel->net = dev_net(dev);
1394         strcpy(tunnel->parms.name, dev->name);
1395
1396         ipip6_tunnel_bind_dev(dev);
1397         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1398         if (!dev->tstats)
1399                 return -ENOMEM;
1400
1401         err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1402         if (err) {
1403                 free_percpu(dev->tstats);
1404                 dev->tstats = NULL;
1405                 return err;
1406         }
1407         dev_hold(dev);
1408         return 0;
1409 }
1410
1411 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1412 {
1413         struct ip_tunnel *tunnel = netdev_priv(dev);
1414         struct iphdr *iph = &tunnel->parms.iph;
1415         struct net *net = dev_net(dev);
1416         struct sit_net *sitn = net_generic(net, sit_net_id);
1417
1418         iph->version            = 4;
1419         iph->protocol           = IPPROTO_IPV6;
1420         iph->ihl                = 5;
1421         iph->ttl                = 64;
1422
1423         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1424 }
1425
1426 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1427                           struct netlink_ext_ack *extack)
1428 {
1429         u8 proto;
1430
1431         if (!data || !data[IFLA_IPTUN_PROTO])
1432                 return 0;
1433
1434         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1435         if (!ipip6_valid_ip_proto(proto))
1436                 return -EINVAL;
1437
1438         return 0;
1439 }
1440
1441 static void ipip6_netlink_parms(struct nlattr *data[],
1442                                 struct ip_tunnel_parm *parms,
1443                                 __u32 *fwmark)
1444 {
1445         memset(parms, 0, sizeof(*parms));
1446
1447         parms->iph.version = 4;
1448         parms->iph.protocol = IPPROTO_IPV6;
1449         parms->iph.ihl = 5;
1450         parms->iph.ttl = 64;
1451
1452         if (!data)
1453                 return;
1454
1455         if (data[IFLA_IPTUN_LINK])
1456                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1457
1458         if (data[IFLA_IPTUN_LOCAL])
1459                 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1460
1461         if (data[IFLA_IPTUN_REMOTE])
1462                 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1463
1464         if (data[IFLA_IPTUN_TTL]) {
1465                 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1466                 if (parms->iph.ttl)
1467                         parms->iph.frag_off = htons(IP_DF);
1468         }
1469
1470         if (data[IFLA_IPTUN_TOS])
1471                 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1472
1473         if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1474                 parms->iph.frag_off = htons(IP_DF);
1475
1476         if (data[IFLA_IPTUN_FLAGS])
1477                 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1478
1479         if (data[IFLA_IPTUN_PROTO])
1480                 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1481
1482         if (data[IFLA_IPTUN_FWMARK])
1483                 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1484 }
1485
1486 /* This function returns true when ENCAP attributes are present in the nl msg */
1487 static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1488                                       struct ip_tunnel_encap *ipencap)
1489 {
1490         bool ret = false;
1491
1492         memset(ipencap, 0, sizeof(*ipencap));
1493
1494         if (!data)
1495                 return ret;
1496
1497         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1498                 ret = true;
1499                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1500         }
1501
1502         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1503                 ret = true;
1504                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1505         }
1506
1507         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1508                 ret = true;
1509                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1510         }
1511
1512         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1513                 ret = true;
1514                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1515         }
1516
1517         return ret;
1518 }
1519
1520 #ifdef CONFIG_IPV6_SIT_6RD
1521 /* This function returns true when 6RD attributes are present in the nl msg */
1522 static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1523                                     struct ip_tunnel_6rd *ip6rd)
1524 {
1525         bool ret = false;
1526         memset(ip6rd, 0, sizeof(*ip6rd));
1527
1528         if (!data)
1529                 return ret;
1530
1531         if (data[IFLA_IPTUN_6RD_PREFIX]) {
1532                 ret = true;
1533                 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1534         }
1535
1536         if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1537                 ret = true;
1538                 ip6rd->relay_prefix =
1539                         nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1540         }
1541
1542         if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1543                 ret = true;
1544                 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1545         }
1546
1547         if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1548                 ret = true;
1549                 ip6rd->relay_prefixlen =
1550                         nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1551         }
1552
1553         return ret;
1554 }
1555 #endif
1556
1557 static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1558                          struct nlattr *tb[], struct nlattr *data[],
1559                          struct netlink_ext_ack *extack)
1560 {
1561         struct net *net = dev_net(dev);
1562         struct ip_tunnel *nt;
1563         struct ip_tunnel_encap ipencap;
1564 #ifdef CONFIG_IPV6_SIT_6RD
1565         struct ip_tunnel_6rd ip6rd;
1566 #endif
1567         int err;
1568
1569         nt = netdev_priv(dev);
1570
1571         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1572                 err = ip_tunnel_encap_setup(nt, &ipencap);
1573                 if (err < 0)
1574                         return err;
1575         }
1576
1577         ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1578
1579         if (ipip6_tunnel_locate(net, &nt->parms, 0))
1580                 return -EEXIST;
1581
1582         err = ipip6_tunnel_create(dev);
1583         if (err < 0)
1584                 return err;
1585
1586         if (tb[IFLA_MTU]) {
1587                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1588
1589                 if (mtu >= IPV6_MIN_MTU &&
1590                     mtu <= IP6_MAX_MTU - dev->hard_header_len)
1591                         dev->mtu = mtu;
1592         }
1593
1594 #ifdef CONFIG_IPV6_SIT_6RD
1595         if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1596                 err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1597                 if (err < 0)
1598                         unregister_netdevice_queue(dev, NULL);
1599         }
1600 #endif
1601
1602         return err;
1603 }
1604
1605 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1606                             struct nlattr *data[],
1607                             struct netlink_ext_ack *extack)
1608 {
1609         struct ip_tunnel *t = netdev_priv(dev);
1610         struct ip_tunnel_parm p;
1611         struct ip_tunnel_encap ipencap;
1612         struct net *net = t->net;
1613         struct sit_net *sitn = net_generic(net, sit_net_id);
1614 #ifdef CONFIG_IPV6_SIT_6RD
1615         struct ip_tunnel_6rd ip6rd;
1616 #endif
1617         __u32 fwmark = t->fwmark;
1618         int err;
1619
1620         if (dev == sitn->fb_tunnel_dev)
1621                 return -EINVAL;
1622
1623         if (ipip6_netlink_encap_parms(data, &ipencap)) {
1624                 err = ip_tunnel_encap_setup(t, &ipencap);
1625                 if (err < 0)
1626                         return err;
1627         }
1628
1629         ipip6_netlink_parms(data, &p, &fwmark);
1630
1631         if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1632             (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1633                 return -EINVAL;
1634
1635         t = ipip6_tunnel_locate(net, &p, 0);
1636
1637         if (t) {
1638                 if (t->dev != dev)
1639                         return -EEXIST;
1640         } else
1641                 t = netdev_priv(dev);
1642
1643         ipip6_tunnel_update(t, &p, fwmark);
1644
1645 #ifdef CONFIG_IPV6_SIT_6RD
1646         if (ipip6_netlink_6rd_parms(data, &ip6rd))
1647                 return ipip6_tunnel_update_6rd(t, &ip6rd);
1648 #endif
1649
1650         return 0;
1651 }
1652
1653 static size_t ipip6_get_size(const struct net_device *dev)
1654 {
1655         return
1656                 /* IFLA_IPTUN_LINK */
1657                 nla_total_size(4) +
1658                 /* IFLA_IPTUN_LOCAL */
1659                 nla_total_size(4) +
1660                 /* IFLA_IPTUN_REMOTE */
1661                 nla_total_size(4) +
1662                 /* IFLA_IPTUN_TTL */
1663                 nla_total_size(1) +
1664                 /* IFLA_IPTUN_TOS */
1665                 nla_total_size(1) +
1666                 /* IFLA_IPTUN_PMTUDISC */
1667                 nla_total_size(1) +
1668                 /* IFLA_IPTUN_FLAGS */
1669                 nla_total_size(2) +
1670                 /* IFLA_IPTUN_PROTO */
1671                 nla_total_size(1) +
1672 #ifdef CONFIG_IPV6_SIT_6RD
1673                 /* IFLA_IPTUN_6RD_PREFIX */
1674                 nla_total_size(sizeof(struct in6_addr)) +
1675                 /* IFLA_IPTUN_6RD_RELAY_PREFIX */
1676                 nla_total_size(4) +
1677                 /* IFLA_IPTUN_6RD_PREFIXLEN */
1678                 nla_total_size(2) +
1679                 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1680                 nla_total_size(2) +
1681 #endif
1682                 /* IFLA_IPTUN_ENCAP_TYPE */
1683                 nla_total_size(2) +
1684                 /* IFLA_IPTUN_ENCAP_FLAGS */
1685                 nla_total_size(2) +
1686                 /* IFLA_IPTUN_ENCAP_SPORT */
1687                 nla_total_size(2) +
1688                 /* IFLA_IPTUN_ENCAP_DPORT */
1689                 nla_total_size(2) +
1690                 /* IFLA_IPTUN_FWMARK */
1691                 nla_total_size(4) +
1692                 0;
1693 }
1694
1695 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1696 {
1697         struct ip_tunnel *tunnel = netdev_priv(dev);
1698         struct ip_tunnel_parm *parm = &tunnel->parms;
1699
1700         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1701             nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1702             nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1703             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1704             nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1705             nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1706                        !!(parm->iph.frag_off & htons(IP_DF))) ||
1707             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1708             nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1709             nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1710                 goto nla_put_failure;
1711
1712 #ifdef CONFIG_IPV6_SIT_6RD
1713         if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1714                              &tunnel->ip6rd.prefix) ||
1715             nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1716                             tunnel->ip6rd.relay_prefix) ||
1717             nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1718                         tunnel->ip6rd.prefixlen) ||
1719             nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1720                         tunnel->ip6rd.relay_prefixlen))
1721                 goto nla_put_failure;
1722 #endif
1723
1724         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1725                         tunnel->encap.type) ||
1726             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1727                         tunnel->encap.sport) ||
1728             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1729                         tunnel->encap.dport) ||
1730             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1731                         tunnel->encap.flags))
1732                 goto nla_put_failure;
1733
1734         return 0;
1735
1736 nla_put_failure:
1737         return -EMSGSIZE;
1738 }
1739
1740 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1741         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
1742         [IFLA_IPTUN_LOCAL]              = { .type = NLA_U32 },
1743         [IFLA_IPTUN_REMOTE]             = { .type = NLA_U32 },
1744         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
1745         [IFLA_IPTUN_TOS]                = { .type = NLA_U8 },
1746         [IFLA_IPTUN_PMTUDISC]           = { .type = NLA_U8 },
1747         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U16 },
1748         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
1749 #ifdef CONFIG_IPV6_SIT_6RD
1750         [IFLA_IPTUN_6RD_PREFIX]         = { .len = sizeof(struct in6_addr) },
1751         [IFLA_IPTUN_6RD_RELAY_PREFIX]   = { .type = NLA_U32 },
1752         [IFLA_IPTUN_6RD_PREFIXLEN]      = { .type = NLA_U16 },
1753         [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1754 #endif
1755         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
1756         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
1757         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
1758         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
1759         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
1760 };
1761
1762 static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1763 {
1764         struct net *net = dev_net(dev);
1765         struct sit_net *sitn = net_generic(net, sit_net_id);
1766
1767         if (dev != sitn->fb_tunnel_dev)
1768                 unregister_netdevice_queue(dev, head);
1769 }
1770
1771 static struct rtnl_link_ops sit_link_ops __read_mostly = {
1772         .kind           = "sit",
1773         .maxtype        = IFLA_IPTUN_MAX,
1774         .policy         = ipip6_policy,
1775         .priv_size      = sizeof(struct ip_tunnel),
1776         .setup          = ipip6_tunnel_setup,
1777         .validate       = ipip6_validate,
1778         .newlink        = ipip6_newlink,
1779         .changelink     = ipip6_changelink,
1780         .get_size       = ipip6_get_size,
1781         .fill_info      = ipip6_fill_info,
1782         .dellink        = ipip6_dellink,
1783         .get_link_net   = ip_tunnel_get_link_net,
1784 };
1785
1786 static struct xfrm_tunnel sit_handler __read_mostly = {
1787         .handler        =       ipip6_rcv,
1788         .err_handler    =       ipip6_err,
1789         .priority       =       1,
1790 };
1791
1792 static struct xfrm_tunnel ipip_handler __read_mostly = {
1793         .handler        =       ipip_rcv,
1794         .err_handler    =       ipip6_err,
1795         .priority       =       2,
1796 };
1797
1798 #if IS_ENABLED(CONFIG_MPLS)
1799 static struct xfrm_tunnel mplsip_handler __read_mostly = {
1800         .handler        =       mplsip_rcv,
1801         .err_handler    =       ipip6_err,
1802         .priority       =       2,
1803 };
1804 #endif
1805
1806 static void __net_exit sit_destroy_tunnels(struct net *net,
1807                                            struct list_head *head)
1808 {
1809         struct sit_net *sitn = net_generic(net, sit_net_id);
1810         struct net_device *dev, *aux;
1811         int prio;
1812
1813         for_each_netdev_safe(net, dev, aux)
1814                 if (dev->rtnl_link_ops == &sit_link_ops)
1815                         unregister_netdevice_queue(dev, head);
1816
1817         for (prio = 0; prio < 4; prio++) {
1818                 int h;
1819                 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1820                         struct ip_tunnel *t;
1821
1822                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1823                         while (t) {
1824                                 /* If dev is in the same netns, it has already
1825                                  * been added to the list by the previous loop.
1826                                  */
1827                                 if (!net_eq(dev_net(t->dev), net))
1828                                         unregister_netdevice_queue(t->dev,
1829                                                                    head);
1830                                 t = rtnl_dereference(t->next);
1831                         }
1832                 }
1833         }
1834 }
1835
1836 static int __net_init sit_init_net(struct net *net)
1837 {
1838         struct sit_net *sitn = net_generic(net, sit_net_id);
1839         struct ip_tunnel *t;
1840         int err;
1841
1842         sitn->tunnels[0] = sitn->tunnels_wc;
1843         sitn->tunnels[1] = sitn->tunnels_l;
1844         sitn->tunnels[2] = sitn->tunnels_r;
1845         sitn->tunnels[3] = sitn->tunnels_r_l;
1846
1847         if (!net_has_fallback_tunnels(net))
1848                 return 0;
1849
1850         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1851                                            NET_NAME_UNKNOWN,
1852                                            ipip6_tunnel_setup);
1853         if (!sitn->fb_tunnel_dev) {
1854                 err = -ENOMEM;
1855                 goto err_alloc_dev;
1856         }
1857         dev_net_set(sitn->fb_tunnel_dev, net);
1858         sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1859         /* FB netdevice is special: we have one, and only one per netns.
1860          * Allowing to move it to another netns is clearly unsafe.
1861          */
1862         sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1863
1864         err = register_netdev(sitn->fb_tunnel_dev);
1865         if (err)
1866                 goto err_reg_dev;
1867
1868         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1869         ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1870
1871         t = netdev_priv(sitn->fb_tunnel_dev);
1872
1873         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1874         return 0;
1875
1876 err_reg_dev:
1877         free_netdev(sitn->fb_tunnel_dev);
1878 err_alloc_dev:
1879         return err;
1880 }
1881
1882 static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1883 {
1884         LIST_HEAD(list);
1885         struct net *net;
1886
1887         rtnl_lock();
1888         list_for_each_entry(net, net_list, exit_list)
1889                 sit_destroy_tunnels(net, &list);
1890
1891         unregister_netdevice_many(&list);
1892         rtnl_unlock();
1893 }
1894
1895 static struct pernet_operations sit_net_ops = {
1896         .init = sit_init_net,
1897         .exit_batch = sit_exit_batch_net,
1898         .id   = &sit_net_id,
1899         .size = sizeof(struct sit_net),
1900 };
1901
1902 static void __exit sit_cleanup(void)
1903 {
1904         rtnl_link_unregister(&sit_link_ops);
1905         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1906         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1907 #if IS_ENABLED(CONFIG_MPLS)
1908         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1909 #endif
1910
1911         unregister_pernet_device(&sit_net_ops);
1912         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1913 }
1914
1915 static int __init sit_init(void)
1916 {
1917         int err;
1918
1919         pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1920
1921         err = register_pernet_device(&sit_net_ops);
1922         if (err < 0)
1923                 return err;
1924         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1925         if (err < 0) {
1926                 pr_info("%s: can't register ip6ip4\n", __func__);
1927                 goto xfrm_tunnel_failed;
1928         }
1929         err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1930         if (err < 0) {
1931                 pr_info("%s: can't register ip4ip4\n", __func__);
1932                 goto xfrm_tunnel4_failed;
1933         }
1934 #if IS_ENABLED(CONFIG_MPLS)
1935         err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1936         if (err < 0) {
1937                 pr_info("%s: can't register mplsip\n", __func__);
1938                 goto xfrm_tunnel_mpls_failed;
1939         }
1940 #endif
1941         err = rtnl_link_register(&sit_link_ops);
1942         if (err < 0)
1943                 goto rtnl_link_failed;
1944
1945 out:
1946         return err;
1947
1948 rtnl_link_failed:
1949 #if IS_ENABLED(CONFIG_MPLS)
1950         xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1951 xfrm_tunnel_mpls_failed:
1952 #endif
1953         xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1954 xfrm_tunnel4_failed:
1955         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1956 xfrm_tunnel_failed:
1957         unregister_pernet_device(&sit_net_ops);
1958         goto out;
1959 }
1960
1961 module_init(sit_init);
1962 module_exit(sit_cleanup);
1963 MODULE_LICENSE("GPL");
1964 MODULE_ALIAS_RTNL_LINK("sit");
1965 MODULE_ALIAS_NETDEV("sit0");