net/ipv6/ip6_tunnel.c

   1 /*
   2  *      IPv6 tunneling device
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
   7  *      Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
   8  *
   9  *      Based on:
  10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  11  *
  12  *      RFC 2473
  13  *
  14  *      This program is free software; you can redistribute it and/or
  15  *      modify it under the terms of the GNU General Public License
  16  *      as published by the Free Software Foundation; either version
  17  *      2 of the License, or (at your option) any later version.
  18  *
  19  */
  20
  21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  22
  23 #include <linux/module.h>
  24 #include <linux/capability.h>
  25 #include <linux/errno.h>
  26 #include <linux/types.h>
  27 #include <linux/sockios.h>
  28 #include <linux/icmp.h>
  29 #include <linux/if.h>
  30 #include <linux/in.h>
  31 #include <linux/ip.h>
  32 #include <linux/net.h>
  33 #include <linux/in6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/icmpv6.h>
  37 #include <linux/init.h>
  38 #include <linux/route.h>
  39 #include <linux/rtnetlink.h>
  40 #include <linux/netfilter_ipv6.h>
  41 #include <linux/slab.h>
  42 #include <linux/hash.h>
  43 #include <linux/etherdevice.h>
  44
  45 #include <asm/uaccess.h>
  46 #include <linux/atomic.h>
  47
  48 #include <net/icmp.h>
  49 #include <net/ip.h>
  50 #include <net/ip_tunnels.h>
  51 #include <net/ipv6.h>
  52 #include <net/ip6_route.h>
  53 #include <net/addrconf.h>
  54 #include <net/ip6_tunnel.h>
  55 #include <net/xfrm.h>
  56 #include <net/dsfield.h>
  57 #include <net/inet_ecn.h>
  58 #include <net/net_namespace.h>
  59 #include <net/netns/generic.h>
  60 #include <net/dst_metadata.h>
  61
  62 MODULE_AUTHOR("Ville Nuorvala");
  63 MODULE_DESCRIPTION("IPv6 tunneling device");
  64 MODULE_LICENSE("GPL");
  65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  66 MODULE_ALIAS_NETDEV("ip6tnl0");
  67
  68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
  69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
  70
  71 static bool log_ecn_error = true;
  72 module_param(log_ecn_error, bool, 0644);
  73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  74
  75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  76 {
  77         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  78
  79         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
  80 }
  81
  82 static int ip6_tnl_dev_init(struct net_device *dev);
  83 static void ip6_tnl_dev_setup(struct net_device *dev);
  84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  85
  86 static int ip6_tnl_net_id __read_mostly;
  87 struct ip6_tnl_net {
  88         /* the IPv6 tunnel fallback device */
  89         struct net_device *fb_tnl_dev;
  90         /* lists for storing tunnels in use */
  91         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
  92         struct ip6_tnl __rcu *tnls_wc[1];
  93         struct ip6_tnl __rcu **tnls[2];
  94         struct ip6_tnl __rcu *collect_md_tun;
  95 };
  96
  97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  98 {
  99         struct pcpu_sw_netstats tmp, sum = { 0 };
 100         int i;
 101
 102         for_each_possible_cpu(i) {
 103                 unsigned int start;
 104                 const struct pcpu_sw_netstats *tstats =
 105                                                    per_cpu_ptr(dev->tstats, i);
 106
 107                 do {
 108                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 109                         tmp.rx_packets = tstats->rx_packets;
 110                         tmp.rx_bytes = tstats->rx_bytes;
 111                         tmp.tx_packets = tstats->tx_packets;
 112                         tmp.tx_bytes =  tstats->tx_bytes;
 113                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 114
 115                 sum.rx_packets += tmp.rx_packets;
 116                 sum.rx_bytes   += tmp.rx_bytes;
 117                 sum.tx_packets += tmp.tx_packets;
 118                 sum.tx_bytes   += tmp.tx_bytes;
 119         }
 120         dev->stats.rx_packets = sum.rx_packets;
 121         dev->stats.rx_bytes   = sum.rx_bytes;
 122         dev->stats.tx_packets = sum.tx_packets;
 123         dev->stats.tx_bytes   = sum.tx_bytes;
 124         return &dev->stats;
 125 }
 126
 127 /**
 128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 129  *   @remote: the address of the tunnel exit-point
 130  *   @local: the address of the tunnel entry-point
 131  *
 132  * Return:
 133  *   tunnel matching given end-points if found,
 134  *   else fallback tunnel if its device is up,
 135  *   else %NULL
 136  **/
 137
 138 #define for_each_ip6_tunnel_rcu(start) \
 139         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 140
 141 static struct ip6_tnl *
 142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 143 {
 144         unsigned int hash = HASH(remote, local);
 145         struct ip6_tnl *t;
 146         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 147         struct in6_addr any;
 148
 149         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 150                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 151                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 152                     (t->dev->flags & IFF_UP))
 153                         return t;
 154         }
 155
 156         memset(&any, 0, sizeof(any));
 157         hash = HASH(&any, local);
 158         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 159                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 160                     ipv6_addr_any(&t->parms.raddr) &&
 161                     (t->dev->flags & IFF_UP))
 162                         return t;
 163         }
 164
 165         hash = HASH(remote, &any);
 166         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 167                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 168                     ipv6_addr_any(&t->parms.laddr) &&
 169                     (t->dev->flags & IFF_UP))
 170                         return t;
 171         }
 172
 173         t = rcu_dereference(ip6n->collect_md_tun);
 174         if (t)
 175                 return t;
 176
 177         t = rcu_dereference(ip6n->tnls_wc[0]);
 178         if (t && (t->dev->flags & IFF_UP))
 179                 return t;
 180
 181         return NULL;
 182 }
 183
 184 /**
 185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 186  *   @p: parameters containing tunnel end-points
 187  *
 188  * Description:
 189  *   ip6_tnl_bucket() returns the head of the list matching the
 190  *   &struct in6_addr entries laddr and raddr in @p.
 191  *
 192  * Return: head of IPv6 tunnel list
 193  **/
 194
 195 static struct ip6_tnl __rcu **
 196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 197 {
 198         const struct in6_addr *remote = &p->raddr;
 199         const struct in6_addr *local = &p->laddr;
 200         unsigned int h = 0;
 201         int prio = 0;
 202
 203         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 204                 prio = 1;
 205                 h = HASH(remote, local);
 206         }
 207         return &ip6n->tnls[prio][h];
 208 }
 209
 210 /**
 211  * ip6_tnl_link - add tunnel to hash table
 212  *   @t: tunnel to be added
 213  **/
 214
 215 static void
 216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 217 {
 218         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 219
 220         if (t->parms.collect_md)
 221                 rcu_assign_pointer(ip6n->collect_md_tun, t);
 222         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 223         rcu_assign_pointer(*tp, t);
 224 }
 225
 226 /**
 227  * ip6_tnl_unlink - remove tunnel from hash table
 228  *   @t: tunnel to be removed
 229  **/
 230
 231 static void
 232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 233 {
 234         struct ip6_tnl __rcu **tp;
 235         struct ip6_tnl *iter;
 236
 237         if (t->parms.collect_md)
 238                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 239
 240         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 241              (iter = rtnl_dereference(*tp)) != NULL;
 242              tp = &iter->next) {
 243                 if (t == iter) {
 244                         rcu_assign_pointer(*tp, t->next);
 245                         break;
 246                 }
 247         }
 248 }
 249
 250 static void ip6_dev_free(struct net_device *dev)
 251 {
 252         struct ip6_tnl *t = netdev_priv(dev);
 253
 254         gro_cells_destroy(&t->gro_cells);
 255         dst_cache_destroy(&t->dst_cache);
 256         free_percpu(dev->tstats);
 257         free_netdev(dev);
 258 }
 259
 260 static int ip6_tnl_create2(struct net_device *dev)
 261 {
 262         struct ip6_tnl *t = netdev_priv(dev);
 263         struct net *net = dev_net(dev);
 264         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 265         int err;
 266
 267         t = netdev_priv(dev);
 268
 269         dev->rtnl_link_ops = &ip6_link_ops;
 270         err = register_netdevice(dev);
 271         if (err < 0)
 272                 goto out;
 273
 274         strcpy(t->parms.name, dev->name);
 275
 276         dev_hold(dev);
 277         ip6_tnl_link(ip6n, t);
 278         return 0;
 279
 280 out:
 281         return err;
 282 }
 283
 284 /**
 285  * ip6_tnl_create - create a new tunnel
 286  *   @p: tunnel parameters
 287  *   @pt: pointer to new tunnel
 288  *
 289  * Description:
 290  *   Create tunnel matching given parameters.
 291  *
 292  * Return:
 293  *   created tunnel or error pointer
 294  **/
 295
 296 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 297 {
 298         struct net_device *dev;
 299         struct ip6_tnl *t;
 300         char name[IFNAMSIZ];
 301         int err = -ENOMEM;
 302
 303         if (p->name[0])
 304                 strlcpy(name, p->name, IFNAMSIZ);
 305         else
 306                 sprintf(name, "ip6tnl%%d");
 307
 308         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 309                            ip6_tnl_dev_setup);
 310         if (!dev)
 311                 goto failed;
 312
 313         dev_net_set(dev, net);
 314
 315         t = netdev_priv(dev);
 316         t->parms = *p;
 317         t->net = dev_net(dev);
 318         err = ip6_tnl_create2(dev);
 319         if (err < 0)
 320                 goto failed_free;
 321
 322         return t;
 323
 324 failed_free:
 325         ip6_dev_free(dev);
 326 failed:
 327         return ERR_PTR(err);
 328 }
 329
 330 /**
 331  * ip6_tnl_locate - find or create tunnel matching given parameters
 332  *   @p: tunnel parameters
 333  *   @create: != 0 if allowed to create new tunnel if no match found
 334  *
 335  * Description:
 336  *   ip6_tnl_locate() first tries to locate an existing tunnel
 337  *   based on @parms. If this is unsuccessful, but @create is set a new
 338  *   tunnel device is created and registered for use.
 339  *
 340  * Return:
 341  *   matching tunnel or error pointer
 342  **/
 343
 344 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 345                 struct __ip6_tnl_parm *p, int create)
 346 {
 347         const struct in6_addr *remote = &p->raddr;
 348         const struct in6_addr *local = &p->laddr;
 349         struct ip6_tnl __rcu **tp;
 350         struct ip6_tnl *t;
 351         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 352
 353         for (tp = ip6_tnl_bucket(ip6n, p);
 354              (t = rtnl_dereference(*tp)) != NULL;
 355              tp = &t->next) {
 356                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 357                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 358                         if (create)
 359                                 return ERR_PTR(-EEXIST);
 360
 361                         return t;
 362                 }
 363         }
 364         if (!create)
 365                 return ERR_PTR(-ENODEV);
 366         return ip6_tnl_create(net, p);
 367 }
 368
 369 /**
 370  * ip6_tnl_dev_uninit - tunnel device uninitializer
 371  *   @dev: the device to be destroyed
 372  *
 373  * Description:
 374  *   ip6_tnl_dev_uninit() removes tunnel from its list
 375  **/
 376
 377 static void
 378 ip6_tnl_dev_uninit(struct net_device *dev)
 379 {
 380         struct ip6_tnl *t = netdev_priv(dev);
 381         struct net *net = t->net;
 382         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 383
 384         if (dev == ip6n->fb_tnl_dev)
 385                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 386         else
 387                 ip6_tnl_unlink(ip6n, t);
 388         dst_cache_reset(&t->dst_cache);
 389         dev_put(dev);
 390 }
 391
 392 /**
 393  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 394  *   @skb: received socket buffer
 395  *
 396  * Return:
 397  *   0 if none was found,
 398  *   else index to encapsulation limit
 399  **/
 400
 401 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 402 {
 403         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
 404         __u8 nexthdr = ipv6h->nexthdr;
 405         __u16 off = sizeof(*ipv6h);
 406
 407         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 408                 __u16 optlen = 0;
 409                 struct ipv6_opt_hdr *hdr;
 410                 if (raw + off + sizeof(*hdr) > skb->data &&
 411                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 412                         break;
 413
 414                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 415                 if (nexthdr == NEXTHDR_FRAGMENT) {
 416                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 417                         if (frag_hdr->frag_off)
 418                                 break;
 419                         optlen = 8;
 420                 } else if (nexthdr == NEXTHDR_AUTH) {
 421                         optlen = (hdr->hdrlen + 2) << 2;
 422                 } else {
 423                         optlen = ipv6_optlen(hdr);
 424                 }
 425                 if (nexthdr == NEXTHDR_DEST) {
 426                         __u16 i = off + 2;
 427                         while (1) {
 428                                 struct ipv6_tlv_tnl_enc_lim *tel;
 429
 430                                 /* No more room for encapsulation limit */
 431                                 if (i + sizeof (*tel) > off + optlen)
 432                                         break;
 433
 434                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 435                                 /* return index of option if found and valid */
 436                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 437                                     tel->length == 1)
 438                                         return i;
 439                                 /* else jump to next option */
 440                                 if (tel->type)
 441                                         i += tel->length + 2;
 442                                 else
 443                                         i++;
 444                         }
 445                 }
 446                 nexthdr = hdr->nexthdr;
 447                 off += optlen;
 448         }
 449         return 0;
 450 }
 451 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 452
 453 /**
 454  * ip6_tnl_err - tunnel error handler
 455  *
 456  * Description:
 457  *   ip6_tnl_err() should handle errors in the tunnel according
 458  *   to the specifications in RFC 2473.
 459  **/
 460
 461 static int
 462 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 463             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 464 {
 465         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
 466         struct ip6_tnl *t;
 467         int rel_msg = 0;
 468         u8 rel_type = ICMPV6_DEST_UNREACH;
 469         u8 rel_code = ICMPV6_ADDR_UNREACH;
 470         u8 tproto;
 471         __u32 rel_info = 0;
 472         __u16 len;
 473         int err = -ENOENT;
 474
 475         /* If the packet doesn't contain the original IPv6 header we are
 476            in trouble since we might need the source address for further
 477            processing of the error. */
 478
 479         rcu_read_lock();
 480         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 481         if (!t)
 482                 goto out;
 483
 484         tproto = ACCESS_ONCE(t->parms.proto);
 485         if (tproto != ipproto && tproto != 0)
 486                 goto out;
 487
 488         err = 0;
 489
 490         switch (*type) {
 491                 __u32 teli;
 492                 struct ipv6_tlv_tnl_enc_lim *tel;
 493                 __u32 mtu;
 494         case ICMPV6_DEST_UNREACH:
 495                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 496                                     t->parms.name);
 497                 rel_msg = 1;
 498                 break;
 499         case ICMPV6_TIME_EXCEED:
 500                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 501                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 502                                             t->parms.name);
 503                         rel_msg = 1;
 504                 }
 505                 break;
 506         case ICMPV6_PARAMPROB:
 507                 teli = 0;
 508                 if ((*code) == ICMPV6_HDR_FIELD)
 509                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 510
 511                 if (teli && teli == *info - 2) {
 512                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 513                         if (tel->encap_limit == 0) {
 514                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 515                                                     t->parms.name);
 516                                 rel_msg = 1;
 517                         }
 518                 } else {
 519                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 520                                             t->parms.name);
 521                 }
 522                 break;
 523         case ICMPV6_PKT_TOOBIG:
 524                 mtu = *info - offset;
 525                 if (mtu < IPV6_MIN_MTU)
 526                         mtu = IPV6_MIN_MTU;
 527                 t->dev->mtu = mtu;
 528
 529                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 530                 if (len > mtu) {
 531                         rel_type = ICMPV6_PKT_TOOBIG;
 532                         rel_code = 0;
 533                         rel_info = mtu;
 534                         rel_msg = 1;
 535                 }
 536                 break;
 537         }
 538
 539         *type = rel_type;
 540         *code = rel_code;
 541         *info = rel_info;
 542         *msg = rel_msg;
 543
 544 out:
 545         rcu_read_unlock();
 546         return err;
 547 }
 548
 549 static int
 550 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 551            u8 type, u8 code, int offset, __be32 info)
 552 {
 553         int rel_msg = 0;
 554         u8 rel_type = type;
 555         u8 rel_code = code;
 556         __u32 rel_info = ntohl(info);
 557         int err;
 558         struct sk_buff *skb2;
 559         const struct iphdr *eiph;
 560         struct rtable *rt;
 561         struct flowi4 fl4;
 562
 563         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 564                           &rel_msg, &rel_info, offset);
 565         if (err < 0)
 566                 return err;
 567
 568         if (rel_msg == 0)
 569                 return 0;
 570
 571         switch (rel_type) {
 572         case ICMPV6_DEST_UNREACH:
 573                 if (rel_code != ICMPV6_ADDR_UNREACH)
 574                         return 0;
 575                 rel_type = ICMP_DEST_UNREACH;
 576                 rel_code = ICMP_HOST_UNREACH;
 577                 break;
 578         case ICMPV6_PKT_TOOBIG:
 579                 if (rel_code != 0)
 580                         return 0;
 581                 rel_type = ICMP_DEST_UNREACH;
 582                 rel_code = ICMP_FRAG_NEEDED;
 583                 break;
 584         case NDISC_REDIRECT:
 585                 rel_type = ICMP_REDIRECT;
 586                 rel_code = ICMP_REDIR_HOST;
 587         default:
 588                 return 0;
 589         }
 590
 591         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 592                 return 0;
 593
 594         skb2 = skb_clone(skb, GFP_ATOMIC);
 595         if (!skb2)
 596                 return 0;
 597
 598         skb_dst_drop(skb2);
 599
 600         skb_pull(skb2, offset);
 601         skb_reset_network_header(skb2);
 602         eiph = ip_hdr(skb2);
 603
 604         /* Try to guess incoming interface */
 605         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 606                                    eiph->saddr, 0,
 607                                    0, 0,
 608                                    IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 609         if (IS_ERR(rt))
 610                 goto out;
 611
 612         skb2->dev = rt->dst.dev;
 613
 614         /* route "incoming" packet */
 615         if (rt->rt_flags & RTCF_LOCAL) {
 616                 ip_rt_put(rt);
 617                 rt = NULL;
 618                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 619                                            eiph->daddr, eiph->saddr,
 620                                            0, 0,
 621                                            IPPROTO_IPIP,
 622                                            RT_TOS(eiph->tos), 0);
 623                 if (IS_ERR(rt) ||
 624                     rt->dst.dev->type != ARPHRD_TUNNEL) {
 625                         if (!IS_ERR(rt))
 626                                 ip_rt_put(rt);
 627                         goto out;
 628                 }
 629                 skb_dst_set(skb2, &rt->dst);
 630         } else {
 631                 ip_rt_put(rt);
 632                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 633                                    skb2->dev) ||
 634                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 635                         goto out;
 636         }
 637
 638         /* change mtu on this route */
 639         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 640                 if (rel_info > dst_mtu(skb_dst(skb2)))
 641                         goto out;
 642
 643                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
 644         }
 645         if (rel_type == ICMP_REDIRECT)
 646                 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
 647
 648         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 649
 650 out:
 651         kfree_skb(skb2);
 652         return 0;
 653 }
 654
 655 static int
 656 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 657            u8 type, u8 code, int offset, __be32 info)
 658 {
 659         int rel_msg = 0;
 660         u8 rel_type = type;
 661         u8 rel_code = code;
 662         __u32 rel_info = ntohl(info);
 663         int err;
 664
 665         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 666                           &rel_msg, &rel_info, offset);
 667         if (err < 0)
 668                 return err;
 669
 670         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 671                 struct rt6_info *rt;
 672                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 673
 674                 if (!skb2)
 675                         return 0;
 676
 677                 skb_dst_drop(skb2);
 678                 skb_pull(skb2, offset);
 679                 skb_reset_network_header(skb2);
 680
 681                 /* Try to guess incoming interface */
 682                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 683                                 NULL, 0, 0);
 684
 685                 if (rt && rt->dst.dev)
 686                         skb2->dev = rt->dst.dev;
 687
 688                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 689
 690                 ip6_rt_put(rt);
 691
 692                 kfree_skb(skb2);
 693         }
 694
 695         return 0;
 696 }
 697
 698 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 699                                        const struct ipv6hdr *ipv6h,
 700                                        struct sk_buff *skb)
 701 {
 702         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 703
 704         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 705                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 706
 707         return IP6_ECN_decapsulate(ipv6h, skb);
 708 }
 709
 710 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 711                                        const struct ipv6hdr *ipv6h,
 712                                        struct sk_buff *skb)
 713 {
 714         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 715                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 716
 717         return IP6_ECN_decapsulate(ipv6h, skb);
 718 }
 719
 720 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 721                              const struct in6_addr *laddr,
 722                              const struct in6_addr *raddr)
 723 {
 724         struct __ip6_tnl_parm *p = &t->parms;
 725         int ltype = ipv6_addr_type(laddr);
 726         int rtype = ipv6_addr_type(raddr);
 727         __u32 flags = 0;
 728
 729         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 730                 flags = IP6_TNL_F_CAP_PER_PACKET;
 731         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 732                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 733                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 734                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 735                 if (ltype&IPV6_ADDR_UNICAST)
 736                         flags |= IP6_TNL_F_CAP_XMIT;
 737                 if (rtype&IPV6_ADDR_UNICAST)
 738                         flags |= IP6_TNL_F_CAP_RCV;
 739         }
 740         return flags;
 741 }
 742 EXPORT_SYMBOL(ip6_tnl_get_cap);
 743
 744 /* called with rcu_read_lock() */
 745 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 746                                   const struct in6_addr *laddr,
 747                                   const struct in6_addr *raddr)
 748 {
 749         struct __ip6_tnl_parm *p = &t->parms;
 750         int ret = 0;
 751         struct net *net = t->net;
 752
 753         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 754             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 755              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 756                 struct net_device *ldev = NULL;
 757
 758                 if (p->link)
 759                         ldev = dev_get_by_index_rcu(net, p->link);
 760
 761                 if ((ipv6_addr_is_multicast(laddr) ||
 762                      likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
 763                     likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
 764                         ret = 1;
 765         }
 766         return ret;
 767 }
 768 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 769
 770 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 771                          const struct tnl_ptk_info *tpi,
 772                          struct metadata_dst *tun_dst,
 773                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 774                                                 const struct ipv6hdr *ipv6h,
 775                                                 struct sk_buff *skb),
 776                          bool log_ecn_err)
 777 {
 778         struct pcpu_sw_netstats *tstats;
 779         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 780         int err;
 781
 782         if ((!(tpi->flags & TUNNEL_CSUM) &&
 783              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 784             ((tpi->flags & TUNNEL_CSUM) &&
 785              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 786                 tunnel->dev->stats.rx_crc_errors++;
 787                 tunnel->dev->stats.rx_errors++;
 788                 goto drop;
 789         }
 790
 791         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 792                 if (!(tpi->flags & TUNNEL_SEQ) ||
 793                     (tunnel->i_seqno &&
 794                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 795                         tunnel->dev->stats.rx_fifo_errors++;
 796                         tunnel->dev->stats.rx_errors++;
 797                         goto drop;
 798                 }
 799                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 800         }
 801
 802         skb->protocol = tpi->proto;
 803
 804         /* Warning: All skb pointers will be invalidated! */
 805         if (tunnel->dev->type == ARPHRD_ETHER) {
 806                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 807                         tunnel->dev->stats.rx_length_errors++;
 808                         tunnel->dev->stats.rx_errors++;
 809                         goto drop;
 810                 }
 811
 812                 ipv6h = ipv6_hdr(skb);
 813                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 814                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 815         } else {
 816                 skb->dev = tunnel->dev;
 817         }
 818
 819         skb_reset_network_header(skb);
 820         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 821
 822         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 823
 824         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 825         if (unlikely(err)) {
 826                 if (log_ecn_err)
 827                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 828                                              &ipv6h->saddr,
 829                                              ipv6_get_dsfield(ipv6h));
 830                 if (err > 1) {
 831                         ++tunnel->dev->stats.rx_frame_errors;
 832                         ++tunnel->dev->stats.rx_errors;
 833                         goto drop;
 834                 }
 835         }
 836
 837         tstats = this_cpu_ptr(tunnel->dev->tstats);
 838         u64_stats_update_begin(&tstats->syncp);
 839         tstats->rx_packets++;
 840         tstats->rx_bytes += skb->len;
 841         u64_stats_update_end(&tstats->syncp);
 842
 843         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 844
 845         if (tun_dst)
 846                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 847
 848         gro_cells_receive(&tunnel->gro_cells, skb);
 849         return 0;
 850
 851 drop:
 852         kfree_skb(skb);
 853         return 0;
 854 }
 855
 856 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 857                 const struct tnl_ptk_info *tpi,
 858                 struct metadata_dst *tun_dst,
 859                 bool log_ecn_err)
 860 {
 861         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
 862                              log_ecn_err);
 863 }
 864 EXPORT_SYMBOL(ip6_tnl_rcv);
 865
 866 static const struct tnl_ptk_info tpi_v6 = {
 867         /* no tunnel info required for ipxip6. */
 868         .proto = htons(ETH_P_IPV6),
 869 };
 870
 871 static const struct tnl_ptk_info tpi_v4 = {
 872         /* no tunnel info required for ipxip6. */
 873         .proto = htons(ETH_P_IP),
 874 };
 875
 876 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 877                       const struct tnl_ptk_info *tpi,
 878                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 879                                                   const struct ipv6hdr *ipv6h,
 880                                                   struct sk_buff *skb))
 881 {
 882         struct ip6_tnl *t;
 883         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 884         struct metadata_dst *tun_dst = NULL;
 885         int ret = -1;
 886
 887         rcu_read_lock();
 888         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 889
 890         if (t) {
 891                 u8 tproto = ACCESS_ONCE(t->parms.proto);
 892
 893                 if (tproto != ipproto && tproto != 0)
 894                         goto drop;
 895                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 896                         goto drop;
 897                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 898                         goto drop;
 899                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 900                         goto drop;
 901                 if (t->parms.collect_md) {
 902                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 903                         if (!tun_dst)
 904                                 return 0;
 905                 }
 906                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 907                                     log_ecn_error);
 908         }
 909
 910         rcu_read_unlock();
 911
 912         return ret;
 913
 914 drop:
 915         rcu_read_unlock();
 916         kfree_skb(skb);
 917         return 0;
 918 }
 919
 920 static int ip4ip6_rcv(struct sk_buff *skb)
 921 {
 922         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 923                           ip4ip6_dscp_ecn_decapsulate);
 924 }
 925
 926 static int ip6ip6_rcv(struct sk_buff *skb)
 927 {
 928         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 929                           ip6ip6_dscp_ecn_decapsulate);
 930 }
 931
 932 struct ipv6_tel_txoption {
 933         struct ipv6_txoptions ops;
 934         __u8 dst_opt[8];
 935 };
 936
 937 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 938 {
 939         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 940
 941         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 942         opt->dst_opt[3] = 1;
 943         opt->dst_opt[4] = encap_limit;
 944         opt->dst_opt[5] = IPV6_TLV_PADN;
 945         opt->dst_opt[6] = 1;
 946
 947         opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 948         opt->ops.opt_nflen = 8;
 949 }
 950
 951 /**
 952  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 953  *   @t: the outgoing tunnel device
 954  *   @hdr: IPv6 header from the incoming packet
 955  *
 956  * Description:
 957  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 958  *   doesn't match source of incoming packet.
 959  *
 960  * Return:
 961  *   1 if conflict,
 962  *   0 else
 963  **/
 964
 965 static inline bool
 966 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 967 {
 968         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 969 }
 970
 971 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 972                      const struct in6_addr *laddr,
 973                      const struct in6_addr *raddr)
 974 {
 975         struct __ip6_tnl_parm *p = &t->parms;
 976         int ret = 0;
 977         struct net *net = t->net;
 978
 979         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 980             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 981              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 982                 struct net_device *ldev = NULL;
 983
 984                 rcu_read_lock();
 985                 if (p->link)
 986                         ldev = dev_get_by_index_rcu(net, p->link);
 987
 988                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 989                         pr_warn("%s xmit: Local address not yet configured!\n",
 990                                 p->name);
 991                 else if (!ipv6_addr_is_multicast(raddr) &&
 992                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
 993                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 994                                 p->name);
 995                 else
 996                         ret = 1;
 997                 rcu_read_unlock();
 998         }
 999         return ret;
1000 }
1001 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1002
1003 /**
1004  * ip6_tnl_xmit - encapsulate packet and send
1005  *   @skb: the outgoing socket buffer
1006  *   @dev: the outgoing tunnel device
1007  *   @dsfield: dscp code for outer header
1008  *   @fl6: flow of tunneled packet
1009  *   @encap_limit: encapsulation limit
1010  *   @pmtu: Path MTU is stored if packet is too big
1011  *   @proto: next header value
1012  *
1013  * Description:
1014  *   Build new header and do some sanity checks on the packet before sending
1015  *   it.
1016  *
1017  * Return:
1018  *   0 on success
1019  *   -1 fail
1020  *   %-EMSGSIZE message too big. return mtu in this case.
1021  **/
1022
1023 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1024                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1025                  __u8 proto)
1026 {
1027         struct ip6_tnl *t = netdev_priv(dev);
1028         struct net *net = t->net;
1029         struct net_device_stats *stats = &t->dev->stats;
1030         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1031         struct ipv6_tel_txoption opt;
1032         struct dst_entry *dst = NULL, *ndst = NULL;
1033         struct net_device *tdev;
1034         int mtu;
1035         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1036         unsigned int max_headroom = psh_hlen;
1037         bool use_cache = false;
1038         u8 hop_limit;
1039         int err = -1;
1040
1041         if (t->parms.collect_md) {
1042                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1043                 goto route_lookup;
1044         } else {
1045                 hop_limit = t->parms.hop_limit;
1046         }
1047
1048         /* NBMA tunnel */
1049         if (ipv6_addr_any(&t->parms.raddr)) {
1050                 struct in6_addr *addr6;
1051                 struct neighbour *neigh;
1052                 int addr_type;
1053
1054                 if (!skb_dst(skb))
1055                         goto tx_err_link_failure;
1056
1057                 neigh = dst_neigh_lookup(skb_dst(skb),
1058                                          &ipv6_hdr(skb)->daddr);
1059                 if (!neigh)
1060                         goto tx_err_link_failure;
1061
1062                 addr6 = (struct in6_addr *)&neigh->primary_key;
1063                 addr_type = ipv6_addr_type(addr6);
1064
1065                 if (addr_type == IPV6_ADDR_ANY)
1066                         addr6 = &ipv6_hdr(skb)->daddr;
1067
1068                 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1069                 neigh_release(neigh);
1070         } else if (!(t->parms.flags &
1071                      (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1072                 /* enable the cache only only if the routing decision does
1073                  * not depend on the current inner header value
1074                  */
1075                 use_cache = true;
1076         }
1077
1078         if (use_cache)
1079                 dst = dst_cache_get(&t->dst_cache);
1080
1081         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1082                 goto tx_err_link_failure;
1083
1084         if (!dst) {
1085 route_lookup:
1086                 dst = ip6_route_output(net, NULL, fl6);
1087
1088                 if (dst->error)
1089                         goto tx_err_link_failure;
1090                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1091                 if (IS_ERR(dst)) {
1092                         err = PTR_ERR(dst);
1093                         dst = NULL;
1094                         goto tx_err_link_failure;
1095                 }
1096                 if (t->parms.collect_md &&
1097                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1098                                        &fl6->daddr, 0, &fl6->saddr))
1099                         goto tx_err_link_failure;
1100                 ndst = dst;
1101         }
1102
1103         tdev = dst->dev;
1104
1105         if (tdev == dev) {
1106                 stats->collisions++;
1107                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1108                                      t->parms.name);
1109                 goto tx_err_dst_release;
1110         }
1111         mtu = dst_mtu(dst) - psh_hlen;
1112         if (encap_limit >= 0) {
1113                 max_headroom += 8;
1114                 mtu -= 8;
1115         }
1116         if (mtu < IPV6_MIN_MTU)
1117                 mtu = IPV6_MIN_MTU;
1118         if (skb_dst(skb) && !t->parms.collect_md)
1119                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1120         if (skb->len > mtu && !skb_is_gso(skb)) {
1121                 *pmtu = mtu;
1122                 err = -EMSGSIZE;
1123                 goto tx_err_dst_release;
1124         }
1125
1126         if (t->err_count > 0) {
1127                 if (time_before(jiffies,
1128                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1129                         t->err_count--;
1130
1131                         dst_link_failure(skb);
1132                 } else {
1133                         t->err_count = 0;
1134                 }
1135         }
1136
1137         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1138
1139         /*
1140          * Okay, now see if we can stuff it in the buffer as-is.
1141          */
1142         max_headroom += LL_RESERVED_SPACE(tdev);
1143
1144         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1145             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1146                 struct sk_buff *new_skb;
1147
1148                 new_skb = skb_realloc_headroom(skb, max_headroom);
1149                 if (!new_skb)
1150                         goto tx_err_dst_release;
1151
1152                 if (skb->sk)
1153                         skb_set_owner_w(new_skb, skb->sk);
1154                 consume_skb(skb);
1155                 skb = new_skb;
1156         }
1157
1158         if (t->parms.collect_md) {
1159                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1160                         goto tx_err_dst_release;
1161         } else {
1162                 if (use_cache && ndst)
1163                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1164         }
1165         skb_dst_set(skb, dst);
1166
1167         if (encap_limit >= 0) {
1168                 init_tel_txopt(&opt, encap_limit);
1169                 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1170         }
1171
1172         /* Calculate max headroom for all the headers and adjust
1173          * needed_headroom if necessary.
1174          */
1175         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1176                         + dst->header_len + t->hlen;
1177         if (max_headroom > dev->needed_headroom)
1178                 dev->needed_headroom = max_headroom;
1179
1180         err = ip6_tnl_encap(skb, t, &proto, fl6);
1181         if (err)
1182                 return err;
1183
1184         skb->protocol = htons(ETH_P_IPV6);
1185         skb_push(skb, sizeof(struct ipv6hdr));
1186         skb_reset_network_header(skb);
1187         ipv6h = ipv6_hdr(skb);
1188         ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1189                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1190         ipv6h->hop_limit = hop_limit;
1191         ipv6h->nexthdr = proto;
1192         ipv6h->saddr = fl6->saddr;
1193         ipv6h->daddr = fl6->daddr;
1194         ip6tunnel_xmit(NULL, skb, dev);
1195         return 0;
1196 tx_err_link_failure:
1197         stats->tx_carrier_errors++;
1198         dst_link_failure(skb);
1199 tx_err_dst_release:
1200         dst_release(dst);
1201         return err;
1202 }
1203 EXPORT_SYMBOL(ip6_tnl_xmit);
1204
1205 static inline int
1206 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1207 {
1208         struct ip6_tnl *t = netdev_priv(dev);
1209         const struct iphdr  *iph = ip_hdr(skb);
1210         int encap_limit = -1;
1211         struct flowi6 fl6;
1212         __u8 dsfield;
1213         __u32 mtu;
1214         u8 tproto;
1215         int err;
1216
1217         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1218
1219         tproto = ACCESS_ONCE(t->parms.proto);
1220         if (tproto != IPPROTO_IPIP && tproto != 0)
1221                 return -1;
1222
1223         dsfield = ipv4_get_dsfield(iph);
1224
1225         if (t->parms.collect_md) {
1226                 struct ip_tunnel_info *tun_info;
1227                 const struct ip_tunnel_key *key;
1228
1229                 tun_info = skb_tunnel_info(skb);
1230                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1231                              ip_tunnel_info_af(tun_info) != AF_INET6))
1232                         return -1;
1233                 key = &tun_info->key;
1234                 memset(&fl6, 0, sizeof(fl6));
1235                 fl6.flowi6_proto = IPPROTO_IPIP;
1236                 fl6.daddr = key->u.ipv6.dst;
1237                 fl6.flowlabel = key->label;
1238         } else {
1239                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1240                         encap_limit = t->parms.encap_limit;
1241
1242                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1243                 fl6.flowi6_proto = IPPROTO_IPIP;
1244
1245                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1246                         fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1247                                          & IPV6_TCLASS_MASK;
1248                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1249                         fl6.flowi6_mark = skb->mark;
1250         }
1251
1252         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1253                 return -1;
1254
1255         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1256
1257         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1258                            IPPROTO_IPIP);
1259         if (err != 0) {
1260                 /* XXX: send ICMP error even if DF is not set. */
1261                 if (err == -EMSGSIZE)
1262                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1263                                   htonl(mtu));
1264                 return -1;
1265         }
1266
1267         return 0;
1268 }
1269
1270 static inline int
1271 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1272 {
1273         struct ip6_tnl *t = netdev_priv(dev);
1274         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1275         int encap_limit = -1;
1276         __u16 offset;
1277         struct flowi6 fl6;
1278         __u8 dsfield;
1279         __u32 mtu;
1280         u8 tproto;
1281         int err;
1282
1283         tproto = ACCESS_ONCE(t->parms.proto);
1284         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1285             ip6_tnl_addr_conflict(t, ipv6h))
1286                 return -1;
1287
1288         dsfield = ipv6_get_dsfield(ipv6h);
1289
1290         if (t->parms.collect_md) {
1291                 struct ip_tunnel_info *tun_info;
1292                 const struct ip_tunnel_key *key;
1293
1294                 tun_info = skb_tunnel_info(skb);
1295                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1296                              ip_tunnel_info_af(tun_info) != AF_INET6))
1297                         return -1;
1298                 key = &tun_info->key;
1299                 memset(&fl6, 0, sizeof(fl6));
1300                 fl6.flowi6_proto = IPPROTO_IPV6;
1301                 fl6.daddr = key->u.ipv6.dst;
1302                 fl6.flowlabel = key->label;
1303         } else {
1304                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1305                 if (offset > 0) {
1306                         struct ipv6_tlv_tnl_enc_lim *tel;
1307
1308                         tel = (void *)&skb_network_header(skb)[offset];
1309                         if (tel->encap_limit == 0) {
1310                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1311                                             ICMPV6_HDR_FIELD, offset + 2);
1312                                 return -1;
1313                         }
1314                         encap_limit = tel->encap_limit - 1;
1315                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1316                         encap_limit = t->parms.encap_limit;
1317                 }
1318
1319                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1320                 fl6.flowi6_proto = IPPROTO_IPV6;
1321
1322                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1323                         fl6.flowlabel |= (*(__be32 *)ipv6h & IPV6_TCLASS_MASK);
1324                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1325                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1326                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1327                         fl6.flowi6_mark = skb->mark;
1328         }
1329
1330         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1331                 return -1;
1332
1333         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1334
1335         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1336                            IPPROTO_IPV6);
1337         if (err != 0) {
1338                 if (err == -EMSGSIZE)
1339                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1340                 return -1;
1341         }
1342
1343         return 0;
1344 }
1345
1346 static netdev_tx_t
1347 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1348 {
1349         struct ip6_tnl *t = netdev_priv(dev);
1350         struct net_device_stats *stats = &t->dev->stats;
1351         int ret;
1352
1353         switch (skb->protocol) {
1354         case htons(ETH_P_IP):
1355                 ret = ip4ip6_tnl_xmit(skb, dev);
1356                 break;
1357         case htons(ETH_P_IPV6):
1358                 ret = ip6ip6_tnl_xmit(skb, dev);
1359                 break;
1360         default:
1361                 goto tx_err;
1362         }
1363
1364         if (ret < 0)
1365                 goto tx_err;
1366
1367         return NETDEV_TX_OK;
1368
1369 tx_err:
1370         stats->tx_errors++;
1371         stats->tx_dropped++;
1372         kfree_skb(skb);
1373         return NETDEV_TX_OK;
1374 }
1375
1376 static void ip6_tnl_link_config(struct ip6_tnl *t)
1377 {
1378         struct net_device *dev = t->dev;
1379         struct __ip6_tnl_parm *p = &t->parms;
1380         struct flowi6 *fl6 = &t->fl.u.ip6;
1381         int t_hlen;
1382
1383         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1384         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1385
1386         /* Set up flowi template */
1387         fl6->saddr = p->laddr;
1388         fl6->daddr = p->raddr;
1389         fl6->flowi6_oif = p->link;
1390         fl6->flowlabel = 0;
1391
1392         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1393                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1394         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1395                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1396
1397         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1398         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1399
1400         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1401                 dev->flags |= IFF_POINTOPOINT;
1402         else
1403                 dev->flags &= ~IFF_POINTOPOINT;
1404
1405         t->tun_hlen = 0;
1406         t->hlen = t->encap_hlen + t->tun_hlen;
1407         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1408
1409         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1410                 int strict = (ipv6_addr_type(&p->raddr) &
1411                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1412
1413                 struct rt6_info *rt = rt6_lookup(t->net,
1414                                                  &p->raddr, &p->laddr,
1415                                                  p->link, strict);
1416
1417                 if (!rt)
1418                         return;
1419
1420                 if (rt->dst.dev) {
1421                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1422                                 t_hlen;
1423
1424                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1425                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1426                                 dev->mtu -= 8;
1427
1428                         if (dev->mtu < IPV6_MIN_MTU)
1429                                 dev->mtu = IPV6_MIN_MTU;
1430                 }
1431                 ip6_rt_put(rt);
1432         }
1433 }
1434
1435 /**
1436  * ip6_tnl_change - update the tunnel parameters
1437  *   @t: tunnel to be changed
1438  *   @p: tunnel configuration parameters
1439  *
1440  * Description:
1441  *   ip6_tnl_change() updates the tunnel parameters
1442  **/
1443
1444 static int
1445 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1446 {
1447         t->parms.laddr = p->laddr;
1448         t->parms.raddr = p->raddr;
1449         t->parms.flags = p->flags;
1450         t->parms.hop_limit = p->hop_limit;
1451         t->parms.encap_limit = p->encap_limit;
1452         t->parms.flowinfo = p->flowinfo;
1453         t->parms.link = p->link;
1454         t->parms.proto = p->proto;
1455         dst_cache_reset(&t->dst_cache);
1456         ip6_tnl_link_config(t);
1457         return 0;
1458 }
1459
1460 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1461 {
1462         struct net *net = t->net;
1463         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1464         int err;
1465
1466         ip6_tnl_unlink(ip6n, t);
1467         synchronize_net();
1468         err = ip6_tnl_change(t, p);
1469         ip6_tnl_link(ip6n, t);
1470         netdev_state_change(t->dev);
1471         return err;
1472 }
1473
1474 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1475 {
1476         /* for default tnl0 device allow to change only the proto */
1477         t->parms.proto = p->proto;
1478         netdev_state_change(t->dev);
1479         return 0;
1480 }
1481
1482 static void
1483 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1484 {
1485         p->laddr = u->laddr;
1486         p->raddr = u->raddr;
1487         p->flags = u->flags;
1488         p->hop_limit = u->hop_limit;
1489         p->encap_limit = u->encap_limit;
1490         p->flowinfo = u->flowinfo;
1491         p->link = u->link;
1492         p->proto = u->proto;
1493         memcpy(p->name, u->name, sizeof(u->name));
1494 }
1495
1496 static void
1497 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1498 {
1499         u->laddr = p->laddr;
1500         u->raddr = p->raddr;
1501         u->flags = p->flags;
1502         u->hop_limit = p->hop_limit;
1503         u->encap_limit = p->encap_limit;
1504         u->flowinfo = p->flowinfo;
1505         u->link = p->link;
1506         u->proto = p->proto;
1507         memcpy(u->name, p->name, sizeof(u->name));
1508 }
1509
1510 /**
1511  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1512  *   @dev: virtual device associated with tunnel
1513  *   @ifr: parameters passed from userspace
1514  *   @cmd: command to be performed
1515  *
1516  * Description:
1517  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1518  *   from userspace.
1519  *
1520  *   The possible commands are the following:
1521  *     %SIOCGETTUNNEL: get tunnel parameters for device
1522  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1523  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1524  *     %SIOCDELTUNNEL: delete tunnel
1525  *
1526  *   The fallback device "ip6tnl0", created during module
1527  *   initialization, can be used for creating other tunnel devices.
1528  *
1529  * Return:
1530  *   0 on success,
1531  *   %-EFAULT if unable to copy data to or from userspace,
1532  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1533  *   %-EINVAL if passed tunnel parameters are invalid,
1534  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1535  *   %-ENODEV if attempting to change or delete a nonexisting device
1536  **/
1537
1538 static int
1539 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1540 {
1541         int err = 0;
1542         struct ip6_tnl_parm p;
1543         struct __ip6_tnl_parm p1;
1544         struct ip6_tnl *t = netdev_priv(dev);
1545         struct net *net = t->net;
1546         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1547
1548         memset(&p1, 0, sizeof(p1));
1549
1550         switch (cmd) {
1551         case SIOCGETTUNNEL:
1552                 if (dev == ip6n->fb_tnl_dev) {
1553                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1554                                 err = -EFAULT;
1555                                 break;
1556                         }
1557                         ip6_tnl_parm_from_user(&p1, &p);
1558                         t = ip6_tnl_locate(net, &p1, 0);
1559                         if (IS_ERR(t))
1560                                 t = netdev_priv(dev);
1561                 } else {
1562                         memset(&p, 0, sizeof(p));
1563                 }
1564                 ip6_tnl_parm_to_user(&p, &t->parms);
1565                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1566                         err = -EFAULT;
1567                 }
1568                 break;
1569         case SIOCADDTUNNEL:
1570         case SIOCCHGTUNNEL:
1571                 err = -EPERM;
1572                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1573                         break;
1574                 err = -EFAULT;
1575                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1576                         break;
1577                 err = -EINVAL;
1578                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1579                     p.proto != 0)
1580                         break;
1581                 ip6_tnl_parm_from_user(&p1, &p);
1582                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1583                 if (cmd == SIOCCHGTUNNEL) {
1584                         if (!IS_ERR(t)) {
1585                                 if (t->dev != dev) {
1586                                         err = -EEXIST;
1587                                         break;
1588                                 }
1589                         } else
1590                                 t = netdev_priv(dev);
1591                         if (dev == ip6n->fb_tnl_dev)
1592                                 err = ip6_tnl0_update(t, &p1);
1593                         else
1594                                 err = ip6_tnl_update(t, &p1);
1595                 }
1596                 if (!IS_ERR(t)) {
1597                         err = 0;
1598                         ip6_tnl_parm_to_user(&p, &t->parms);
1599                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1600                                 err = -EFAULT;
1601
1602                 } else {
1603                         err = PTR_ERR(t);
1604                 }
1605                 break;
1606         case SIOCDELTUNNEL:
1607                 err = -EPERM;
1608                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1609                         break;
1610
1611                 if (dev == ip6n->fb_tnl_dev) {
1612                         err = -EFAULT;
1613                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1614                                 break;
1615                         err = -ENOENT;
1616                         ip6_tnl_parm_from_user(&p1, &p);
1617                         t = ip6_tnl_locate(net, &p1, 0);
1618                         if (IS_ERR(t))
1619                                 break;
1620                         err = -EPERM;
1621                         if (t->dev == ip6n->fb_tnl_dev)
1622                                 break;
1623                         dev = t->dev;
1624                 }
1625                 err = 0;
1626                 unregister_netdevice(dev);
1627                 break;
1628         default:
1629                 err = -EINVAL;
1630         }
1631         return err;
1632 }
1633
1634 /**
1635  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1636  *   @dev: virtual device associated with tunnel
1637  *   @new_mtu: the new mtu
1638  *
1639  * Return:
1640  *   0 on success,
1641  *   %-EINVAL if mtu too small
1642  **/
1643
1644 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1645 {
1646         struct ip6_tnl *tnl = netdev_priv(dev);
1647
1648         if (tnl->parms.proto == IPPROTO_IPIP) {
1649                 if (new_mtu < 68)
1650                         return -EINVAL;
1651         } else {
1652                 if (new_mtu < IPV6_MIN_MTU)
1653                         return -EINVAL;
1654         }
1655         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1656                 return -EINVAL;
1657         dev->mtu = new_mtu;
1658         return 0;
1659 }
1660 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1661
1662 int ip6_tnl_get_iflink(const struct net_device *dev)
1663 {
1664         struct ip6_tnl *t = netdev_priv(dev);
1665
1666         return t->parms.link;
1667 }
1668 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1669
1670 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1671                           unsigned int num)
1672 {
1673         if (num >= MAX_IPTUN_ENCAP_OPS)
1674                 return -ERANGE;
1675
1676         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1677                         &ip6tun_encaps[num],
1678                         NULL, ops) ? 0 : -1;
1679 }
1680 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1681
1682 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1683                           unsigned int num)
1684 {
1685         int ret;
1686
1687         if (num >= MAX_IPTUN_ENCAP_OPS)
1688                 return -ERANGE;
1689
1690         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1691                        &ip6tun_encaps[num],
1692                        ops, NULL) == ops) ? 0 : -1;
1693
1694         synchronize_net();
1695
1696         return ret;
1697 }
1698 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1699
1700 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1701                         struct ip_tunnel_encap *ipencap)
1702 {
1703         int hlen;
1704
1705         memset(&t->encap, 0, sizeof(t->encap));
1706
1707         hlen = ip6_encap_hlen(ipencap);
1708         if (hlen < 0)
1709                 return hlen;
1710
1711         t->encap.type = ipencap->type;
1712         t->encap.sport = ipencap->sport;
1713         t->encap.dport = ipencap->dport;
1714         t->encap.flags = ipencap->flags;
1715
1716         t->encap_hlen = hlen;
1717         t->hlen = t->encap_hlen + t->tun_hlen;
1718
1719         return 0;
1720 }
1721 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1722
1723 static const struct net_device_ops ip6_tnl_netdev_ops = {
1724         .ndo_init       = ip6_tnl_dev_init,
1725         .ndo_uninit     = ip6_tnl_dev_uninit,
1726         .ndo_start_xmit = ip6_tnl_start_xmit,
1727         .ndo_do_ioctl   = ip6_tnl_ioctl,
1728         .ndo_change_mtu = ip6_tnl_change_mtu,
1729         .ndo_get_stats  = ip6_get_stats,
1730         .ndo_get_iflink = ip6_tnl_get_iflink,
1731 };
1732
1733 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1734                          NETIF_F_FRAGLIST |     \
1735                          NETIF_F_HIGHDMA |      \
1736                          NETIF_F_GSO_SOFTWARE | \
1737                          NETIF_F_HW_CSUM)
1738
1739 /**
1740  * ip6_tnl_dev_setup - setup virtual tunnel device
1741  *   @dev: virtual device associated with tunnel
1742  *
1743  * Description:
1744  *   Initialize function pointers and device parameters
1745  **/
1746
1747 static void ip6_tnl_dev_setup(struct net_device *dev)
1748 {
1749         dev->netdev_ops = &ip6_tnl_netdev_ops;
1750         dev->destructor = ip6_dev_free;
1751
1752         dev->type = ARPHRD_TUNNEL6;
1753         dev->flags |= IFF_NOARP;
1754         dev->addr_len = sizeof(struct in6_addr);
1755         dev->features |= NETIF_F_LLTX;
1756         netif_keep_dst(dev);
1757
1758         dev->features           |= IPXIPX_FEATURES;
1759         dev->hw_features        |= IPXIPX_FEATURES;
1760
1761         /* This perm addr will be used as interface identifier by IPv6 */
1762         dev->addr_assign_type = NET_ADDR_RANDOM;
1763         eth_random_addr(dev->perm_addr);
1764 }
1765
1766
1767 /**
1768  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1769  *   @dev: virtual device associated with tunnel
1770  **/
1771
1772 static inline int
1773 ip6_tnl_dev_init_gen(struct net_device *dev)
1774 {
1775         struct ip6_tnl *t = netdev_priv(dev);
1776         int ret;
1777         int t_hlen;
1778
1779         t->dev = dev;
1780         t->net = dev_net(dev);
1781         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1782         if (!dev->tstats)
1783                 return -ENOMEM;
1784
1785         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1786         if (ret)
1787                 goto free_stats;
1788
1789         ret = gro_cells_init(&t->gro_cells, dev);
1790         if (ret)
1791                 goto destroy_dst;
1792
1793         t->tun_hlen = 0;
1794         t->hlen = t->encap_hlen + t->tun_hlen;
1795         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1796
1797         dev->type = ARPHRD_TUNNEL6;
1798         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1799         dev->mtu = ETH_DATA_LEN - t_hlen;
1800         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1801                 dev->mtu -= 8;
1802
1803         return 0;
1804
1805 destroy_dst:
1806         dst_cache_destroy(&t->dst_cache);
1807 free_stats:
1808         free_percpu(dev->tstats);
1809         dev->tstats = NULL;
1810
1811         return ret;
1812 }
1813
1814 /**
1815  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1816  *   @dev: virtual device associated with tunnel
1817  **/
1818
1819 static int ip6_tnl_dev_init(struct net_device *dev)
1820 {
1821         struct ip6_tnl *t = netdev_priv(dev);
1822         int err = ip6_tnl_dev_init_gen(dev);
1823
1824         if (err)
1825                 return err;
1826         ip6_tnl_link_config(t);
1827         if (t->parms.collect_md) {
1828                 dev->features |= NETIF_F_NETNS_LOCAL;
1829                 netif_keep_dst(dev);
1830         }
1831         return 0;
1832 }
1833
1834 /**
1835  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1836  *   @dev: fallback device
1837  *
1838  * Return: 0
1839  **/
1840
1841 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1842 {
1843         struct ip6_tnl *t = netdev_priv(dev);
1844         struct net *net = dev_net(dev);
1845         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1846
1847         t->parms.proto = IPPROTO_IPV6;
1848         dev_hold(dev);
1849
1850         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1851         return 0;
1852 }
1853
1854 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1855 {
1856         u8 proto;
1857
1858         if (!data || !data[IFLA_IPTUN_PROTO])
1859                 return 0;
1860
1861         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1862         if (proto != IPPROTO_IPV6 &&
1863             proto != IPPROTO_IPIP &&
1864             proto != 0)
1865                 return -EINVAL;
1866
1867         return 0;
1868 }
1869
1870 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1871                                   struct __ip6_tnl_parm *parms)
1872 {
1873         memset(parms, 0, sizeof(*parms));
1874
1875         if (!data)
1876                 return;
1877
1878         if (data[IFLA_IPTUN_LINK])
1879                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1880
1881         if (data[IFLA_IPTUN_LOCAL])
1882                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1883
1884         if (data[IFLA_IPTUN_REMOTE])
1885                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1886
1887         if (data[IFLA_IPTUN_TTL])
1888                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1889
1890         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1891                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1892
1893         if (data[IFLA_IPTUN_FLOWINFO])
1894                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1895
1896         if (data[IFLA_IPTUN_FLAGS])
1897                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1898
1899         if (data[IFLA_IPTUN_PROTO])
1900                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1901
1902         if (data[IFLA_IPTUN_COLLECT_METADATA])
1903                 parms->collect_md = true;
1904 }
1905
1906 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1907                                         struct ip_tunnel_encap *ipencap)
1908 {
1909         bool ret = false;
1910
1911         memset(ipencap, 0, sizeof(*ipencap));
1912
1913         if (!data)
1914                 return ret;
1915
1916         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1917                 ret = true;
1918                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1919         }
1920
1921         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1922                 ret = true;
1923                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1924         }
1925
1926         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1927                 ret = true;
1928                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1929         }
1930
1931         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1932                 ret = true;
1933                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1934         }
1935
1936         return ret;
1937 }
1938
1939 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1940                            struct nlattr *tb[], struct nlattr *data[])
1941 {
1942         struct net *net = dev_net(dev);
1943         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1944         struct ip6_tnl *nt, *t;
1945         struct ip_tunnel_encap ipencap;
1946
1947         nt = netdev_priv(dev);
1948
1949         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1950                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1951
1952                 if (err < 0)
1953                         return err;
1954         }
1955
1956         ip6_tnl_netlink_parms(data, &nt->parms);
1957
1958         if (nt->parms.collect_md) {
1959                 if (rtnl_dereference(ip6n->collect_md_tun))
1960                         return -EEXIST;
1961         } else {
1962                 t = ip6_tnl_locate(net, &nt->parms, 0);
1963                 if (!IS_ERR(t))
1964                         return -EEXIST;
1965         }
1966
1967         return ip6_tnl_create2(dev);
1968 }
1969
1970 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1971                               struct nlattr *data[])
1972 {
1973         struct ip6_tnl *t = netdev_priv(dev);
1974         struct __ip6_tnl_parm p;
1975         struct net *net = t->net;
1976         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1977         struct ip_tunnel_encap ipencap;
1978
1979         if (dev == ip6n->fb_tnl_dev)
1980                 return -EINVAL;
1981
1982         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1983                 int err = ip6_tnl_encap_setup(t, &ipencap);
1984
1985                 if (err < 0)
1986                         return err;
1987         }
1988         ip6_tnl_netlink_parms(data, &p);
1989         if (p.collect_md)
1990                 return -EINVAL;
1991
1992         t = ip6_tnl_locate(net, &p, 0);
1993         if (!IS_ERR(t)) {
1994                 if (t->dev != dev)
1995                         return -EEXIST;
1996         } else
1997                 t = netdev_priv(dev);
1998
1999         return ip6_tnl_update(t, &p);
2000 }
2001
2002 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2003 {
2004         struct net *net = dev_net(dev);
2005         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2006
2007         if (dev != ip6n->fb_tnl_dev)
2008                 unregister_netdevice_queue(dev, head);
2009 }
2010
2011 static size_t ip6_tnl_get_size(const struct net_device *dev)
2012 {
2013         return
2014                 /* IFLA_IPTUN_LINK */
2015                 nla_total_size(4) +
2016                 /* IFLA_IPTUN_LOCAL */
2017                 nla_total_size(sizeof(struct in6_addr)) +
2018                 /* IFLA_IPTUN_REMOTE */
2019                 nla_total_size(sizeof(struct in6_addr)) +
2020                 /* IFLA_IPTUN_TTL */
2021                 nla_total_size(1) +
2022                 /* IFLA_IPTUN_ENCAP_LIMIT */
2023                 nla_total_size(1) +
2024                 /* IFLA_IPTUN_FLOWINFO */
2025                 nla_total_size(4) +
2026                 /* IFLA_IPTUN_FLAGS */
2027                 nla_total_size(4) +
2028                 /* IFLA_IPTUN_PROTO */
2029                 nla_total_size(1) +
2030                 /* IFLA_IPTUN_ENCAP_TYPE */
2031                 nla_total_size(2) +
2032                 /* IFLA_IPTUN_ENCAP_FLAGS */
2033                 nla_total_size(2) +
2034                 /* IFLA_IPTUN_ENCAP_SPORT */
2035                 nla_total_size(2) +
2036                 /* IFLA_IPTUN_ENCAP_DPORT */
2037                 nla_total_size(2) +
2038                 /* IFLA_IPTUN_COLLECT_METADATA */
2039                 nla_total_size(0) +
2040                 0;
2041 }
2042
2043 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2044 {
2045         struct ip6_tnl *tunnel = netdev_priv(dev);
2046         struct __ip6_tnl_parm *parm = &tunnel->parms;
2047
2048         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2049             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2050             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2051             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2052             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2053             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2054             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2055             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
2056                 goto nla_put_failure;
2057
2058         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2059             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2060             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2061             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2062                 goto nla_put_failure;
2063
2064         if (parm->collect_md)
2065                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2066                         goto nla_put_failure;
2067         return 0;
2068
2069 nla_put_failure:
2070         return -EMSGSIZE;
2071 }
2072
2073 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2074 {
2075         struct ip6_tnl *tunnel = netdev_priv(dev);
2076
2077         return tunnel->net;
2078 }
2079 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2080
2081 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2082         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2083         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2084         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2085         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2086         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2087         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2088         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2089         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2090         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2091         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2092         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2093         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2094         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2095 };
2096
2097 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2098         .kind           = "ip6tnl",
2099         .maxtype        = IFLA_IPTUN_MAX,
2100         .policy         = ip6_tnl_policy,
2101         .priv_size      = sizeof(struct ip6_tnl),
2102         .setup          = ip6_tnl_dev_setup,
2103         .validate       = ip6_tnl_validate,
2104         .newlink        = ip6_tnl_newlink,
2105         .changelink     = ip6_tnl_changelink,
2106         .dellink        = ip6_tnl_dellink,
2107         .get_size       = ip6_tnl_get_size,
2108         .fill_info      = ip6_tnl_fill_info,
2109         .get_link_net   = ip6_tnl_get_link_net,
2110 };
2111
2112 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2113         .handler        = ip4ip6_rcv,
2114         .err_handler    = ip4ip6_err,
2115         .priority       =       1,
2116 };
2117
2118 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2119         .handler        = ip6ip6_rcv,
2120         .err_handler    = ip6ip6_err,
2121         .priority       =       1,
2122 };
2123
2124 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2125 {
2126         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2127         struct net_device *dev, *aux;
2128         int h;
2129         struct ip6_tnl *t;
2130         LIST_HEAD(list);
2131
2132         for_each_netdev_safe(net, dev, aux)
2133                 if (dev->rtnl_link_ops == &ip6_link_ops)
2134                         unregister_netdevice_queue(dev, &list);
2135
2136         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2137                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2138                 while (t) {
2139                         /* If dev is in the same netns, it has already
2140                          * been added to the list by the previous loop.
2141                          */
2142                         if (!net_eq(dev_net(t->dev), net))
2143                                 unregister_netdevice_queue(t->dev, &list);
2144                         t = rtnl_dereference(t->next);
2145                 }
2146         }
2147
2148         unregister_netdevice_many(&list);
2149 }
2150
2151 static int __net_init ip6_tnl_init_net(struct net *net)
2152 {
2153         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2154         struct ip6_tnl *t = NULL;
2155         int err;
2156
2157         ip6n->tnls[0] = ip6n->tnls_wc;
2158         ip6n->tnls[1] = ip6n->tnls_r_l;
2159
2160         err = -ENOMEM;
2161         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2162                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2163
2164         if (!ip6n->fb_tnl_dev)
2165                 goto err_alloc_dev;
2166         dev_net_set(ip6n->fb_tnl_dev, net);
2167         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2168         /* FB netdevice is special: we have one, and only one per netns.
2169          * Allowing to move it to another netns is clearly unsafe.
2170          */
2171         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2172
2173         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2174         if (err < 0)
2175                 goto err_register;
2176
2177         err = register_netdev(ip6n->fb_tnl_dev);
2178         if (err < 0)
2179                 goto err_register;
2180
2181         t = netdev_priv(ip6n->fb_tnl_dev);
2182
2183         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2184         return 0;
2185
2186 err_register:
2187         ip6_dev_free(ip6n->fb_tnl_dev);
2188 err_alloc_dev:
2189         return err;
2190 }
2191
2192 static void __net_exit ip6_tnl_exit_net(struct net *net)
2193 {
2194         rtnl_lock();
2195         ip6_tnl_destroy_tunnels(net);
2196         rtnl_unlock();
2197 }
2198
2199 static struct pernet_operations ip6_tnl_net_ops = {
2200         .init = ip6_tnl_init_net,
2201         .exit = ip6_tnl_exit_net,
2202         .id   = &ip6_tnl_net_id,
2203         .size = sizeof(struct ip6_tnl_net),
2204 };
2205
2206 /**
2207  * ip6_tunnel_init - register protocol and reserve needed resources
2208  *
2209  * Return: 0 on success
2210  **/
2211
2212 static int __init ip6_tunnel_init(void)
2213 {
2214         int  err;
2215
2216         err = register_pernet_device(&ip6_tnl_net_ops);
2217         if (err < 0)
2218                 goto out_pernet;
2219
2220         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2221         if (err < 0) {
2222                 pr_err("%s: can't register ip4ip6\n", __func__);
2223                 goto out_ip4ip6;
2224         }
2225
2226         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2227         if (err < 0) {
2228                 pr_err("%s: can't register ip6ip6\n", __func__);
2229                 goto out_ip6ip6;
2230         }
2231         err = rtnl_link_register(&ip6_link_ops);
2232         if (err < 0)
2233                 goto rtnl_link_failed;
2234
2235         return 0;
2236
2237 rtnl_link_failed:
2238         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2239 out_ip6ip6:
2240         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2241 out_ip4ip6:
2242         unregister_pernet_device(&ip6_tnl_net_ops);
2243 out_pernet:
2244         return err;
2245 }
2246
2247 /**
2248  * ip6_tunnel_cleanup - free resources and unregister protocol
2249  **/
2250
2251 static void __exit ip6_tunnel_cleanup(void)
2252 {
2253         rtnl_link_unregister(&ip6_link_ops);
2254         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2255                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2256
2257         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2258                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2259
2260         unregister_pernet_device(&ip6_tnl_net_ops);
2261 }
2262
2263 module_init(ip6_tunnel_init);
2264 module_exit(ip6_tunnel_cleanup);