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 <linux/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 unsigned 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 && t->dev->flags & IFF_UP)
 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 }
 258
 259 static int ip6_tnl_create2(struct net_device *dev)
 260 {
 261         struct ip6_tnl *t = netdev_priv(dev);
 262         struct net *net = dev_net(dev);
 263         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 264         int err;
 265
 266         t = netdev_priv(dev);
 267
 268         dev->rtnl_link_ops = &ip6_link_ops;
 269         err = register_netdevice(dev);
 270         if (err < 0)
 271                 goto out;
 272
 273         strcpy(t->parms.name, dev->name);
 274
 275         dev_hold(dev);
 276         ip6_tnl_link(ip6n, t);
 277         return 0;
 278
 279 out:
 280         return err;
 281 }
 282
 283 /**
 284  * ip6_tnl_create - create a new tunnel
 285  *   @p: tunnel parameters
 286  *   @pt: pointer to new tunnel
 287  *
 288  * Description:
 289  *   Create tunnel matching given parameters.
 290  *
 291  * Return:
 292  *   created tunnel or error pointer
 293  **/
 294
 295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 296 {
 297         struct net_device *dev;
 298         struct ip6_tnl *t;
 299         char name[IFNAMSIZ];
 300         int err = -ENOMEM;
 301
 302         if (p->name[0])
 303                 strlcpy(name, p->name, IFNAMSIZ);
 304         else
 305                 sprintf(name, "ip6tnl%%d");
 306
 307         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 308                            ip6_tnl_dev_setup);
 309         if (!dev)
 310                 goto failed;
 311
 312         dev_net_set(dev, net);
 313
 314         t = netdev_priv(dev);
 315         t->parms = *p;
 316         t->net = dev_net(dev);
 317         err = ip6_tnl_create2(dev);
 318         if (err < 0)
 319                 goto failed_free;
 320
 321         return t;
 322
 323 failed_free:
 324         free_netdev(dev);
 325 failed:
 326         return ERR_PTR(err);
 327 }
 328
 329 /**
 330  * ip6_tnl_locate - find or create tunnel matching given parameters
 331  *   @p: tunnel parameters
 332  *   @create: != 0 if allowed to create new tunnel if no match found
 333  *
 334  * Description:
 335  *   ip6_tnl_locate() first tries to locate an existing tunnel
 336  *   based on @parms. If this is unsuccessful, but @create is set a new
 337  *   tunnel device is created and registered for use.
 338  *
 339  * Return:
 340  *   matching tunnel or error pointer
 341  **/
 342
 343 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 344                 struct __ip6_tnl_parm *p, int create)
 345 {
 346         const struct in6_addr *remote = &p->raddr;
 347         const struct in6_addr *local = &p->laddr;
 348         struct ip6_tnl __rcu **tp;
 349         struct ip6_tnl *t;
 350         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 351
 352         for (tp = ip6_tnl_bucket(ip6n, p);
 353              (t = rtnl_dereference(*tp)) != NULL;
 354              tp = &t->next) {
 355                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 356                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 357                         if (create)
 358                                 return ERR_PTR(-EEXIST);
 359
 360                         return t;
 361                 }
 362         }
 363         if (!create)
 364                 return ERR_PTR(-ENODEV);
 365         return ip6_tnl_create(net, p);
 366 }
 367
 368 /**
 369  * ip6_tnl_dev_uninit - tunnel device uninitializer
 370  *   @dev: the device to be destroyed
 371  *
 372  * Description:
 373  *   ip6_tnl_dev_uninit() removes tunnel from its list
 374  **/
 375
 376 static void
 377 ip6_tnl_dev_uninit(struct net_device *dev)
 378 {
 379         struct ip6_tnl *t = netdev_priv(dev);
 380         struct net *net = t->net;
 381         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 382
 383         if (dev == ip6n->fb_tnl_dev)
 384                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 385         else
 386                 ip6_tnl_unlink(ip6n, t);
 387         dst_cache_reset(&t->dst_cache);
 388         dev_put(dev);
 389 }
 390
 391 /**
 392  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 393  *   @skb: received socket buffer
 394  *
 395  * Return:
 396  *   0 if none was found,
 397  *   else index to encapsulation limit
 398  **/
 399
 400 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 401 {
 402         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 403         unsigned int nhoff = raw - skb->data;
 404         unsigned int off = nhoff + sizeof(*ipv6h);
 405         u8 next, nexthdr = ipv6h->nexthdr;
 406
 407         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 408                 struct ipv6_opt_hdr *hdr;
 409                 u16 optlen;
 410
 411                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 412                         break;
 413
 414                 hdr = (struct ipv6_opt_hdr *)(skb->data + 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                 /* cache hdr->nexthdr, since pskb_may_pull() might
 426                  * invalidate hdr
 427                  */
 428                 next = hdr->nexthdr;
 429                 if (nexthdr == NEXTHDR_DEST) {
 430                         u16 i = 2;
 431
 432                         /* Remember : hdr is no longer valid at this point. */
 433                         if (!pskb_may_pull(skb, off + optlen))
 434                                 break;
 435
 436                         while (1) {
 437                                 struct ipv6_tlv_tnl_enc_lim *tel;
 438
 439                                 /* No more room for encapsulation limit */
 440                                 if (i + sizeof(*tel) > optlen)
 441                                         break;
 442
 443                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 444                                 /* return index of option if found and valid */
 445                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 446                                     tel->length == 1)
 447                                         return i + off - nhoff;
 448                                 /* else jump to next option */
 449                                 if (tel->type)
 450                                         i += tel->length + 2;
 451                                 else
 452                                         i++;
 453                         }
 454                 }
 455                 nexthdr = next;
 456                 off += optlen;
 457         }
 458         return 0;
 459 }
 460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 461
 462 /**
 463  * ip6_tnl_err - tunnel error handler
 464  *
 465  * Description:
 466  *   ip6_tnl_err() should handle errors in the tunnel according
 467  *   to the specifications in RFC 2473.
 468  **/
 469
 470 static int
 471 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 472             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 473 {
 474         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 475         struct net *net = dev_net(skb->dev);
 476         u8 rel_type = ICMPV6_DEST_UNREACH;
 477         u8 rel_code = ICMPV6_ADDR_UNREACH;
 478         __u32 rel_info = 0;
 479         struct ip6_tnl *t;
 480         int err = -ENOENT;
 481         int rel_msg = 0;
 482         u8 tproto;
 483         __u16 len;
 484
 485         /* If the packet doesn't contain the original IPv6 header we are
 486            in trouble since we might need the source address for further
 487            processing of the error. */
 488
 489         rcu_read_lock();
 490         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 491         if (!t)
 492                 goto out;
 493
 494         tproto = READ_ONCE(t->parms.proto);
 495         if (tproto != ipproto && tproto != 0)
 496                 goto out;
 497
 498         err = 0;
 499
 500         switch (*type) {
 501                 struct ipv6_tlv_tnl_enc_lim *tel;
 502                 __u32 mtu, teli;
 503         case ICMPV6_DEST_UNREACH:
 504                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 505                                     t->parms.name);
 506                 rel_msg = 1;
 507                 break;
 508         case ICMPV6_TIME_EXCEED:
 509                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 510                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 511                                             t->parms.name);
 512                         rel_msg = 1;
 513                 }
 514                 break;
 515         case ICMPV6_PARAMPROB:
 516                 teli = 0;
 517                 if ((*code) == ICMPV6_HDR_FIELD)
 518                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 519
 520                 if (teli && teli == *info - 2) {
 521                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 522                         if (tel->encap_limit == 0) {
 523                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 524                                                     t->parms.name);
 525                                 rel_msg = 1;
 526                         }
 527                 } else {
 528                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 529                                             t->parms.name);
 530                 }
 531                 break;
 532         case ICMPV6_PKT_TOOBIG:
 533                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 534                                 sock_net_uid(net, NULL));
 535                 mtu = *info - offset;
 536                 if (mtu < IPV6_MIN_MTU)
 537                         mtu = IPV6_MIN_MTU;
 538                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 539                 if (len > mtu) {
 540                         rel_type = ICMPV6_PKT_TOOBIG;
 541                         rel_code = 0;
 542                         rel_info = mtu;
 543                         rel_msg = 1;
 544                 }
 545                 break;
 546         case NDISC_REDIRECT:
 547                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 548                              sock_net_uid(net, NULL));
 549                 break;
 550         }
 551
 552         *type = rel_type;
 553         *code = rel_code;
 554         *info = rel_info;
 555         *msg = rel_msg;
 556
 557 out:
 558         rcu_read_unlock();
 559         return err;
 560 }
 561
 562 static int
 563 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 564            u8 type, u8 code, int offset, __be32 info)
 565 {
 566         __u32 rel_info = ntohl(info);
 567         const struct iphdr *eiph;
 568         struct sk_buff *skb2;
 569         int err, rel_msg = 0;
 570         u8 rel_type = type;
 571         u8 rel_code = code;
 572         struct rtable *rt;
 573         struct flowi4 fl4;
 574
 575         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 576                           &rel_msg, &rel_info, offset);
 577         if (err < 0)
 578                 return err;
 579
 580         if (rel_msg == 0)
 581                 return 0;
 582
 583         switch (rel_type) {
 584         case ICMPV6_DEST_UNREACH:
 585                 if (rel_code != ICMPV6_ADDR_UNREACH)
 586                         return 0;
 587                 rel_type = ICMP_DEST_UNREACH;
 588                 rel_code = ICMP_HOST_UNREACH;
 589                 break;
 590         case ICMPV6_PKT_TOOBIG:
 591                 if (rel_code != 0)
 592                         return 0;
 593                 rel_type = ICMP_DEST_UNREACH;
 594                 rel_code = ICMP_FRAG_NEEDED;
 595                 break;
 596         default:
 597                 return 0;
 598         }
 599
 600         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 601                 return 0;
 602
 603         skb2 = skb_clone(skb, GFP_ATOMIC);
 604         if (!skb2)
 605                 return 0;
 606
 607         skb_dst_drop(skb2);
 608
 609         skb_pull(skb2, offset);
 610         skb_reset_network_header(skb2);
 611         eiph = ip_hdr(skb2);
 612
 613         /* Try to guess incoming interface */
 614         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 615                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 616         if (IS_ERR(rt))
 617                 goto out;
 618
 619         skb2->dev = rt->dst.dev;
 620         ip_rt_put(rt);
 621
 622         /* route "incoming" packet */
 623         if (rt->rt_flags & RTCF_LOCAL) {
 624                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 625                                            eiph->daddr, eiph->saddr, 0, 0,
 626                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 627                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
 628                         if (!IS_ERR(rt))
 629                                 ip_rt_put(rt);
 630                         goto out;
 631                 }
 632                 skb_dst_set(skb2, &rt->dst);
 633         } else {
 634                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 635                                    skb2->dev) ||
 636                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
 637                         goto out;
 638         }
 639
 640         /* change mtu on this route */
 641         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 642                 if (rel_info > dst_mtu(skb_dst(skb2)))
 643                         goto out;
 644
 645                 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2,
 646                                                 rel_info);
 647         }
 648
 649         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 650
 651 out:
 652         kfree_skb(skb2);
 653         return 0;
 654 }
 655
 656 static int
 657 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 658            u8 type, u8 code, int offset, __be32 info)
 659 {
 660         __u32 rel_info = ntohl(info);
 661         int err, rel_msg = 0;
 662         u8 rel_type = type;
 663         u8 rel_code = code;
 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                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
 764                      likely(!ipv6_chk_addr(net, raddr, NULL, 0))))
 765                         ret = 1;
 766         }
 767         return ret;
 768 }
 769 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 770
 771 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 772                          const struct tnl_ptk_info *tpi,
 773                          struct metadata_dst *tun_dst,
 774                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 775                                                 const struct ipv6hdr *ipv6h,
 776                                                 struct sk_buff *skb),
 777                          bool log_ecn_err)
 778 {
 779         struct pcpu_sw_netstats *tstats;
 780         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 781         int err;
 782
 783         if ((!(tpi->flags & TUNNEL_CSUM) &&
 784              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 785             ((tpi->flags & TUNNEL_CSUM) &&
 786              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 787                 tunnel->dev->stats.rx_crc_errors++;
 788                 tunnel->dev->stats.rx_errors++;
 789                 goto drop;
 790         }
 791
 792         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 793                 if (!(tpi->flags & TUNNEL_SEQ) ||
 794                     (tunnel->i_seqno &&
 795                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 796                         tunnel->dev->stats.rx_fifo_errors++;
 797                         tunnel->dev->stats.rx_errors++;
 798                         goto drop;
 799                 }
 800                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 801         }
 802
 803         skb->protocol = tpi->proto;
 804
 805         /* Warning: All skb pointers will be invalidated! */
 806         if (tunnel->dev->type == ARPHRD_ETHER) {
 807                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 808                         tunnel->dev->stats.rx_length_errors++;
 809                         tunnel->dev->stats.rx_errors++;
 810                         goto drop;
 811                 }
 812
 813                 ipv6h = ipv6_hdr(skb);
 814                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 815                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 816         } else {
 817                 skb->dev = tunnel->dev;
 818         }
 819
 820         skb_reset_network_header(skb);
 821         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 822
 823         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 824
 825         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 826         if (unlikely(err)) {
 827                 if (log_ecn_err)
 828                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 829                                              &ipv6h->saddr,
 830                                              ipv6_get_dsfield(ipv6h));
 831                 if (err > 1) {
 832                         ++tunnel->dev->stats.rx_frame_errors;
 833                         ++tunnel->dev->stats.rx_errors;
 834                         goto drop;
 835                 }
 836         }
 837
 838         tstats = this_cpu_ptr(tunnel->dev->tstats);
 839         u64_stats_update_begin(&tstats->syncp);
 840         tstats->rx_packets++;
 841         tstats->rx_bytes += skb->len;
 842         u64_stats_update_end(&tstats->syncp);
 843
 844         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 845
 846         if (tun_dst)
 847                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 848
 849         gro_cells_receive(&tunnel->gro_cells, skb);
 850         return 0;
 851
 852 drop:
 853         if (tun_dst)
 854                 dst_release((struct dst_entry *)tun_dst);
 855         kfree_skb(skb);
 856         return 0;
 857 }
 858
 859 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 860                 const struct tnl_ptk_info *tpi,
 861                 struct metadata_dst *tun_dst,
 862                 bool log_ecn_err)
 863 {
 864         return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
 865                              log_ecn_err);
 866 }
 867 EXPORT_SYMBOL(ip6_tnl_rcv);
 868
 869 static const struct tnl_ptk_info tpi_v6 = {
 870         /* no tunnel info required for ipxip6. */
 871         .proto = htons(ETH_P_IPV6),
 872 };
 873
 874 static const struct tnl_ptk_info tpi_v4 = {
 875         /* no tunnel info required for ipxip6. */
 876         .proto = htons(ETH_P_IP),
 877 };
 878
 879 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 880                       const struct tnl_ptk_info *tpi,
 881                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 882                                                   const struct ipv6hdr *ipv6h,
 883                                                   struct sk_buff *skb))
 884 {
 885         struct ip6_tnl *t;
 886         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 887         struct metadata_dst *tun_dst = NULL;
 888         int ret = -1;
 889
 890         rcu_read_lock();
 891         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 892
 893         if (t) {
 894                 u8 tproto = READ_ONCE(t->parms.proto);
 895
 896                 if (tproto != ipproto && tproto != 0)
 897                         goto drop;
 898                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 899                         goto drop;
 900                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 901                         goto drop;
 902                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 903                         goto drop;
 904                 if (t->parms.collect_md) {
 905                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 906                         if (!tun_dst)
 907                                 goto drop;
 908                 }
 909                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 910                                     log_ecn_error);
 911         }
 912
 913         rcu_read_unlock();
 914
 915         return ret;
 916
 917 drop:
 918         rcu_read_unlock();
 919         kfree_skb(skb);
 920         return 0;
 921 }
 922
 923 static int ip4ip6_rcv(struct sk_buff *skb)
 924 {
 925         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 926                           ip4ip6_dscp_ecn_decapsulate);
 927 }
 928
 929 static int ip6ip6_rcv(struct sk_buff *skb)
 930 {
 931         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 932                           ip6ip6_dscp_ecn_decapsulate);
 933 }
 934
 935 struct ipv6_tel_txoption {
 936         struct ipv6_txoptions ops;
 937         __u8 dst_opt[8];
 938 };
 939
 940 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 941 {
 942         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 943
 944         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 945         opt->dst_opt[3] = 1;
 946         opt->dst_opt[4] = encap_limit;
 947         opt->dst_opt[5] = IPV6_TLV_PADN;
 948         opt->dst_opt[6] = 1;
 949
 950         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 951         opt->ops.opt_nflen = 8;
 952 }
 953
 954 /**
 955  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 956  *   @t: the outgoing tunnel device
 957  *   @hdr: IPv6 header from the incoming packet
 958  *
 959  * Description:
 960  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 961  *   doesn't match source of incoming packet.
 962  *
 963  * Return:
 964  *   1 if conflict,
 965  *   0 else
 966  **/
 967
 968 static inline bool
 969 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 970 {
 971         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 972 }
 973
 974 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 975                      const struct in6_addr *laddr,
 976                      const struct in6_addr *raddr)
 977 {
 978         struct __ip6_tnl_parm *p = &t->parms;
 979         int ret = 0;
 980         struct net *net = t->net;
 981
 982         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 983             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 984              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 985                 struct net_device *ldev = NULL;
 986
 987                 rcu_read_lock();
 988                 if (p->link)
 989                         ldev = dev_get_by_index_rcu(net, p->link);
 990
 991                 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
 992                         pr_warn("%s xmit: Local address not yet configured!\n",
 993                                 p->name);
 994                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
 995                          !ipv6_addr_is_multicast(raddr) &&
 996                          unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
 997                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
 998                                 p->name);
 999                 else
1000                         ret = 1;
1001                 rcu_read_unlock();
1002         }
1003         return ret;
1004 }
1005 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1006
1007 /**
1008  * ip6_tnl_xmit - encapsulate packet and send
1009  *   @skb: the outgoing socket buffer
1010  *   @dev: the outgoing tunnel device
1011  *   @dsfield: dscp code for outer header
1012  *   @fl6: flow of tunneled packet
1013  *   @encap_limit: encapsulation limit
1014  *   @pmtu: Path MTU is stored if packet is too big
1015  *   @proto: next header value
1016  *
1017  * Description:
1018  *   Build new header and do some sanity checks on the packet before sending
1019  *   it.
1020  *
1021  * Return:
1022  *   0 on success
1023  *   -1 fail
1024  *   %-EMSGSIZE message too big. return mtu in this case.
1025  **/
1026
1027 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1028                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1029                  __u8 proto)
1030 {
1031         struct ip6_tnl *t = netdev_priv(dev);
1032         struct net *net = t->net;
1033         struct net_device_stats *stats = &t->dev->stats;
1034         struct ipv6hdr *ipv6h;
1035         struct ipv6_tel_txoption opt;
1036         struct dst_entry *dst = NULL, *ndst = NULL;
1037         struct net_device *tdev;
1038         int mtu;
1039         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1040         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1041         unsigned int max_headroom = psh_hlen;
1042         bool use_cache = false;
1043         u8 hop_limit;
1044         int err = -1;
1045
1046         if (t->parms.collect_md) {
1047                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1048                 goto route_lookup;
1049         } else {
1050                 hop_limit = t->parms.hop_limit;
1051         }
1052
1053         /* NBMA tunnel */
1054         if (ipv6_addr_any(&t->parms.raddr)) {
1055                 if (skb->protocol == htons(ETH_P_IPV6)) {
1056                         struct in6_addr *addr6;
1057                         struct neighbour *neigh;
1058                         int addr_type;
1059
1060                         if (!skb_dst(skb))
1061                                 goto tx_err_link_failure;
1062
1063                         neigh = dst_neigh_lookup(skb_dst(skb),
1064                                                  &ipv6_hdr(skb)->daddr);
1065                         if (!neigh)
1066                                 goto tx_err_link_failure;
1067
1068                         addr6 = (struct in6_addr *)&neigh->primary_key;
1069                         addr_type = ipv6_addr_type(addr6);
1070
1071                         if (addr_type == IPV6_ADDR_ANY)
1072                                 addr6 = &ipv6_hdr(skb)->daddr;
1073
1074                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1075                         neigh_release(neigh);
1076                 }
1077         } else if (!(t->parms.flags &
1078                      (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1079                 /* enable the cache only only if the routing decision does
1080                  * not depend on the current inner header value
1081                  */
1082                 use_cache = true;
1083         }
1084
1085         if (use_cache)
1086                 dst = dst_cache_get(&t->dst_cache);
1087
1088         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1089                 goto tx_err_link_failure;
1090
1091         if (!dst) {
1092 route_lookup:
1093                 /* add dsfield to flowlabel for route lookup */
1094                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1095
1096                 dst = ip6_route_output(net, NULL, fl6);
1097
1098                 if (dst->error)
1099                         goto tx_err_link_failure;
1100                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1101                 if (IS_ERR(dst)) {
1102                         err = PTR_ERR(dst);
1103                         dst = NULL;
1104                         goto tx_err_link_failure;
1105                 }
1106                 if (t->parms.collect_md &&
1107                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1108                                        &fl6->daddr, 0, &fl6->saddr))
1109                         goto tx_err_link_failure;
1110                 ndst = dst;
1111         }
1112
1113         tdev = dst->dev;
1114
1115         if (tdev == dev) {
1116                 stats->collisions++;
1117                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1118                                      t->parms.name);
1119                 goto tx_err_dst_release;
1120         }
1121         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1122         if (encap_limit >= 0) {
1123                 max_headroom += 8;
1124                 mtu -= 8;
1125         }
1126         if (mtu < IPV6_MIN_MTU)
1127                 mtu = IPV6_MIN_MTU;
1128         if (skb_dst(skb) && !t->parms.collect_md)
1129                 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1130         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1131                 *pmtu = mtu;
1132                 err = -EMSGSIZE;
1133                 goto tx_err_dst_release;
1134         }
1135
1136         if (t->err_count > 0) {
1137                 if (time_before(jiffies,
1138                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1139                         t->err_count--;
1140
1141                         dst_link_failure(skb);
1142                 } else {
1143                         t->err_count = 0;
1144                 }
1145         }
1146
1147         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1148
1149         /*
1150          * Okay, now see if we can stuff it in the buffer as-is.
1151          */
1152         max_headroom += LL_RESERVED_SPACE(tdev);
1153
1154         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1155             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1156                 struct sk_buff *new_skb;
1157
1158                 new_skb = skb_realloc_headroom(skb, max_headroom);
1159                 if (!new_skb)
1160                         goto tx_err_dst_release;
1161
1162                 if (skb->sk)
1163                         skb_set_owner_w(new_skb, skb->sk);
1164                 consume_skb(skb);
1165                 skb = new_skb;
1166         }
1167
1168         if (t->parms.collect_md) {
1169                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1170                         goto tx_err_dst_release;
1171         } else {
1172                 if (use_cache && ndst)
1173                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1174         }
1175         skb_dst_set(skb, dst);
1176
1177         if (encap_limit >= 0) {
1178                 init_tel_txopt(&opt, encap_limit);
1179                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1180         }
1181         hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1182
1183         /* Calculate max headroom for all the headers and adjust
1184          * needed_headroom if necessary.
1185          */
1186         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1187                         + dst->header_len + t->hlen;
1188         if (max_headroom > dev->needed_headroom)
1189                 dev->needed_headroom = max_headroom;
1190
1191         err = ip6_tnl_encap(skb, t, &proto, fl6);
1192         if (err)
1193                 return err;
1194
1195         skb_push(skb, sizeof(struct ipv6hdr));
1196         skb_reset_network_header(skb);
1197         ipv6h = ipv6_hdr(skb);
1198         ip6_flow_hdr(ipv6h, dsfield,
1199                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1200         ipv6h->hop_limit = hop_limit;
1201         ipv6h->nexthdr = proto;
1202         ipv6h->saddr = fl6->saddr;
1203         ipv6h->daddr = fl6->daddr;
1204         ip6tunnel_xmit(NULL, skb, dev);
1205         return 0;
1206 tx_err_link_failure:
1207         stats->tx_carrier_errors++;
1208         dst_link_failure(skb);
1209 tx_err_dst_release:
1210         dst_release(dst);
1211         return err;
1212 }
1213 EXPORT_SYMBOL(ip6_tnl_xmit);
1214
1215 static inline int
1216 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1217 {
1218         struct ip6_tnl *t = netdev_priv(dev);
1219         const struct iphdr  *iph = ip_hdr(skb);
1220         int encap_limit = -1;
1221         struct flowi6 fl6;
1222         __u8 dsfield;
1223         __u32 mtu;
1224         u8 tproto;
1225         int err;
1226
1227         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1228
1229         tproto = READ_ONCE(t->parms.proto);
1230         if (tproto != IPPROTO_IPIP && tproto != 0)
1231                 return -1;
1232
1233         if (t->parms.collect_md) {
1234                 struct ip_tunnel_info *tun_info;
1235                 const struct ip_tunnel_key *key;
1236
1237                 tun_info = skb_tunnel_info(skb);
1238                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1239                              ip_tunnel_info_af(tun_info) != AF_INET6))
1240                         return -1;
1241                 key = &tun_info->key;
1242                 memset(&fl6, 0, sizeof(fl6));
1243                 fl6.flowi6_proto = IPPROTO_IPIP;
1244                 fl6.daddr = key->u.ipv6.dst;
1245                 fl6.flowlabel = key->label;
1246                 dsfield =  key->tos;
1247         } else {
1248                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1249                         encap_limit = t->parms.encap_limit;
1250
1251                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1252                 fl6.flowi6_proto = IPPROTO_IPIP;
1253
1254                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1255                         dsfield = ipv4_get_dsfield(iph);
1256                 else
1257                         dsfield = ip6_tclass(t->parms.flowinfo);
1258                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1259                         fl6.flowi6_mark = skb->mark;
1260                 else
1261                         fl6.flowi6_mark = t->parms.fwmark;
1262         }
1263
1264         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1265
1266         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1267                 return -1;
1268
1269         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1270
1271         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1272
1273         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1274                            IPPROTO_IPIP);
1275         if (err != 0) {
1276                 /* XXX: send ICMP error even if DF is not set. */
1277                 if (err == -EMSGSIZE)
1278                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1279                                   htonl(mtu));
1280                 return -1;
1281         }
1282
1283         return 0;
1284 }
1285
1286 static inline int
1287 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1288 {
1289         struct ip6_tnl *t = netdev_priv(dev);
1290         struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1291         int encap_limit = -1;
1292         __u16 offset;
1293         struct flowi6 fl6;
1294         __u8 dsfield;
1295         __u32 mtu;
1296         u8 tproto;
1297         int err;
1298
1299         tproto = READ_ONCE(t->parms.proto);
1300         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1301             ip6_tnl_addr_conflict(t, ipv6h))
1302                 return -1;
1303
1304         if (t->parms.collect_md) {
1305                 struct ip_tunnel_info *tun_info;
1306                 const struct ip_tunnel_key *key;
1307
1308                 tun_info = skb_tunnel_info(skb);
1309                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1310                              ip_tunnel_info_af(tun_info) != AF_INET6))
1311                         return -1;
1312                 key = &tun_info->key;
1313                 memset(&fl6, 0, sizeof(fl6));
1314                 fl6.flowi6_proto = IPPROTO_IPV6;
1315                 fl6.daddr = key->u.ipv6.dst;
1316                 fl6.flowlabel = key->label;
1317                 dsfield = key->tos;
1318         } else {
1319                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1320                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1321                 ipv6h = ipv6_hdr(skb);
1322                 if (offset > 0) {
1323                         struct ipv6_tlv_tnl_enc_lim *tel;
1324
1325                         tel = (void *)&skb_network_header(skb)[offset];
1326                         if (tel->encap_limit == 0) {
1327                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1328                                             ICMPV6_HDR_FIELD, offset + 2);
1329                                 return -1;
1330                         }
1331                         encap_limit = tel->encap_limit - 1;
1332                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1333                         encap_limit = t->parms.encap_limit;
1334                 }
1335
1336                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1337                 fl6.flowi6_proto = IPPROTO_IPV6;
1338
1339                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1340                         dsfield = ipv6_get_dsfield(ipv6h);
1341                 else
1342                         dsfield = ip6_tclass(t->parms.flowinfo);
1343                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1344                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1345                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1346                         fl6.flowi6_mark = skb->mark;
1347                 else
1348                         fl6.flowi6_mark = t->parms.fwmark;
1349         }
1350
1351         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1352
1353         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1354                 return -1;
1355
1356         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1357
1358         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1359
1360         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1361                            IPPROTO_IPV6);
1362         if (err != 0) {
1363                 if (err == -EMSGSIZE)
1364                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1365                 return -1;
1366         }
1367
1368         return 0;
1369 }
1370
1371 static netdev_tx_t
1372 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1373 {
1374         struct ip6_tnl *t = netdev_priv(dev);
1375         struct net_device_stats *stats = &t->dev->stats;
1376         int ret;
1377
1378         switch (skb->protocol) {
1379         case htons(ETH_P_IP):
1380                 ret = ip4ip6_tnl_xmit(skb, dev);
1381                 break;
1382         case htons(ETH_P_IPV6):
1383                 ret = ip6ip6_tnl_xmit(skb, dev);
1384                 break;
1385         default:
1386                 goto tx_err;
1387         }
1388
1389         if (ret < 0)
1390                 goto tx_err;
1391
1392         return NETDEV_TX_OK;
1393
1394 tx_err:
1395         stats->tx_errors++;
1396         stats->tx_dropped++;
1397         kfree_skb(skb);
1398         return NETDEV_TX_OK;
1399 }
1400
1401 static void ip6_tnl_link_config(struct ip6_tnl *t)
1402 {
1403         struct net_device *dev = t->dev;
1404         struct __ip6_tnl_parm *p = &t->parms;
1405         struct flowi6 *fl6 = &t->fl.u.ip6;
1406         int t_hlen;
1407
1408         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1409         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1410
1411         /* Set up flowi template */
1412         fl6->saddr = p->laddr;
1413         fl6->daddr = p->raddr;
1414         fl6->flowi6_oif = p->link;
1415         fl6->flowlabel = 0;
1416
1417         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1418                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1419         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1420                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1421
1422         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1423         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1424
1425         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1426                 dev->flags |= IFF_POINTOPOINT;
1427         else
1428                 dev->flags &= ~IFF_POINTOPOINT;
1429
1430         t->tun_hlen = 0;
1431         t->hlen = t->encap_hlen + t->tun_hlen;
1432         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1433
1434         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1435                 int strict = (ipv6_addr_type(&p->raddr) &
1436                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1437
1438                 struct rt6_info *rt = rt6_lookup(t->net,
1439                                                  &p->raddr, &p->laddr,
1440                                                  p->link, strict);
1441
1442                 if (!rt)
1443                         return;
1444
1445                 if (rt->dst.dev) {
1446                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1447                                 t_hlen;
1448
1449                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1450                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1451                                 dev->mtu -= 8;
1452
1453                         if (dev->mtu < IPV6_MIN_MTU)
1454                                 dev->mtu = IPV6_MIN_MTU;
1455                 }
1456                 ip6_rt_put(rt);
1457         }
1458 }
1459
1460 /**
1461  * ip6_tnl_change - update the tunnel parameters
1462  *   @t: tunnel to be changed
1463  *   @p: tunnel configuration parameters
1464  *
1465  * Description:
1466  *   ip6_tnl_change() updates the tunnel parameters
1467  **/
1468
1469 static int
1470 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1471 {
1472         t->parms.laddr = p->laddr;
1473         t->parms.raddr = p->raddr;
1474         t->parms.flags = p->flags;
1475         t->parms.hop_limit = p->hop_limit;
1476         t->parms.encap_limit = p->encap_limit;
1477         t->parms.flowinfo = p->flowinfo;
1478         t->parms.link = p->link;
1479         t->parms.proto = p->proto;
1480         t->parms.fwmark = p->fwmark;
1481         dst_cache_reset(&t->dst_cache);
1482         ip6_tnl_link_config(t);
1483         return 0;
1484 }
1485
1486 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1487 {
1488         struct net *net = t->net;
1489         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1490         int err;
1491
1492         ip6_tnl_unlink(ip6n, t);
1493         synchronize_net();
1494         err = ip6_tnl_change(t, p);
1495         ip6_tnl_link(ip6n, t);
1496         netdev_state_change(t->dev);
1497         return err;
1498 }
1499
1500 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1501 {
1502         /* for default tnl0 device allow to change only the proto */
1503         t->parms.proto = p->proto;
1504         netdev_state_change(t->dev);
1505         return 0;
1506 }
1507
1508 static void
1509 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1510 {
1511         p->laddr = u->laddr;
1512         p->raddr = u->raddr;
1513         p->flags = u->flags;
1514         p->hop_limit = u->hop_limit;
1515         p->encap_limit = u->encap_limit;
1516         p->flowinfo = u->flowinfo;
1517         p->link = u->link;
1518         p->proto = u->proto;
1519         memcpy(p->name, u->name, sizeof(u->name));
1520 }
1521
1522 static void
1523 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1524 {
1525         u->laddr = p->laddr;
1526         u->raddr = p->raddr;
1527         u->flags = p->flags;
1528         u->hop_limit = p->hop_limit;
1529         u->encap_limit = p->encap_limit;
1530         u->flowinfo = p->flowinfo;
1531         u->link = p->link;
1532         u->proto = p->proto;
1533         memcpy(u->name, p->name, sizeof(u->name));
1534 }
1535
1536 /**
1537  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1538  *   @dev: virtual device associated with tunnel
1539  *   @ifr: parameters passed from userspace
1540  *   @cmd: command to be performed
1541  *
1542  * Description:
1543  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1544  *   from userspace.
1545  *
1546  *   The possible commands are the following:
1547  *     %SIOCGETTUNNEL: get tunnel parameters for device
1548  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1549  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1550  *     %SIOCDELTUNNEL: delete tunnel
1551  *
1552  *   The fallback device "ip6tnl0", created during module
1553  *   initialization, can be used for creating other tunnel devices.
1554  *
1555  * Return:
1556  *   0 on success,
1557  *   %-EFAULT if unable to copy data to or from userspace,
1558  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1559  *   %-EINVAL if passed tunnel parameters are invalid,
1560  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1561  *   %-ENODEV if attempting to change or delete a nonexisting device
1562  **/
1563
1564 static int
1565 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1566 {
1567         int err = 0;
1568         struct ip6_tnl_parm p;
1569         struct __ip6_tnl_parm p1;
1570         struct ip6_tnl *t = netdev_priv(dev);
1571         struct net *net = t->net;
1572         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1573
1574         memset(&p1, 0, sizeof(p1));
1575
1576         switch (cmd) {
1577         case SIOCGETTUNNEL:
1578                 if (dev == ip6n->fb_tnl_dev) {
1579                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1580                                 err = -EFAULT;
1581                                 break;
1582                         }
1583                         ip6_tnl_parm_from_user(&p1, &p);
1584                         t = ip6_tnl_locate(net, &p1, 0);
1585                         if (IS_ERR(t))
1586                                 t = netdev_priv(dev);
1587                 } else {
1588                         memset(&p, 0, sizeof(p));
1589                 }
1590                 ip6_tnl_parm_to_user(&p, &t->parms);
1591                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1592                         err = -EFAULT;
1593                 }
1594                 break;
1595         case SIOCADDTUNNEL:
1596         case SIOCCHGTUNNEL:
1597                 err = -EPERM;
1598                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1599                         break;
1600                 err = -EFAULT;
1601                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1602                         break;
1603                 err = -EINVAL;
1604                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1605                     p.proto != 0)
1606                         break;
1607                 ip6_tnl_parm_from_user(&p1, &p);
1608                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1609                 if (cmd == SIOCCHGTUNNEL) {
1610                         if (!IS_ERR(t)) {
1611                                 if (t->dev != dev) {
1612                                         err = -EEXIST;
1613                                         break;
1614                                 }
1615                         } else
1616                                 t = netdev_priv(dev);
1617                         if (dev == ip6n->fb_tnl_dev)
1618                                 err = ip6_tnl0_update(t, &p1);
1619                         else
1620                                 err = ip6_tnl_update(t, &p1);
1621                 }
1622                 if (!IS_ERR(t)) {
1623                         err = 0;
1624                         ip6_tnl_parm_to_user(&p, &t->parms);
1625                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1626                                 err = -EFAULT;
1627
1628                 } else {
1629                         err = PTR_ERR(t);
1630                 }
1631                 break;
1632         case SIOCDELTUNNEL:
1633                 err = -EPERM;
1634                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1635                         break;
1636
1637                 if (dev == ip6n->fb_tnl_dev) {
1638                         err = -EFAULT;
1639                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1640                                 break;
1641                         err = -ENOENT;
1642                         ip6_tnl_parm_from_user(&p1, &p);
1643                         t = ip6_tnl_locate(net, &p1, 0);
1644                         if (IS_ERR(t))
1645                                 break;
1646                         err = -EPERM;
1647                         if (t->dev == ip6n->fb_tnl_dev)
1648                                 break;
1649                         dev = t->dev;
1650                 }
1651                 err = 0;
1652                 unregister_netdevice(dev);
1653                 break;
1654         default:
1655                 err = -EINVAL;
1656         }
1657         return err;
1658 }
1659
1660 /**
1661  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1662  *   @dev: virtual device associated with tunnel
1663  *   @new_mtu: the new mtu
1664  *
1665  * Return:
1666  *   0 on success,
1667  *   %-EINVAL if mtu too small
1668  **/
1669
1670 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1671 {
1672         struct ip6_tnl *tnl = netdev_priv(dev);
1673
1674         if (tnl->parms.proto == IPPROTO_IPIP) {
1675                 if (new_mtu < ETH_MIN_MTU)
1676                         return -EINVAL;
1677         } else {
1678                 if (new_mtu < IPV6_MIN_MTU)
1679                         return -EINVAL;
1680         }
1681         if (new_mtu > 0xFFF8 - dev->hard_header_len)
1682                 return -EINVAL;
1683         dev->mtu = new_mtu;
1684         return 0;
1685 }
1686 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1687
1688 int ip6_tnl_get_iflink(const struct net_device *dev)
1689 {
1690         struct ip6_tnl *t = netdev_priv(dev);
1691
1692         return t->parms.link;
1693 }
1694 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1695
1696 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1697                           unsigned int num)
1698 {
1699         if (num >= MAX_IPTUN_ENCAP_OPS)
1700                 return -ERANGE;
1701
1702         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1703                         &ip6tun_encaps[num],
1704                         NULL, ops) ? 0 : -1;
1705 }
1706 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1707
1708 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1709                           unsigned int num)
1710 {
1711         int ret;
1712
1713         if (num >= MAX_IPTUN_ENCAP_OPS)
1714                 return -ERANGE;
1715
1716         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1717                        &ip6tun_encaps[num],
1718                        ops, NULL) == ops) ? 0 : -1;
1719
1720         synchronize_net();
1721
1722         return ret;
1723 }
1724 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1725
1726 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1727                         struct ip_tunnel_encap *ipencap)
1728 {
1729         int hlen;
1730
1731         memset(&t->encap, 0, sizeof(t->encap));
1732
1733         hlen = ip6_encap_hlen(ipencap);
1734         if (hlen < 0)
1735                 return hlen;
1736
1737         t->encap.type = ipencap->type;
1738         t->encap.sport = ipencap->sport;
1739         t->encap.dport = ipencap->dport;
1740         t->encap.flags = ipencap->flags;
1741
1742         t->encap_hlen = hlen;
1743         t->hlen = t->encap_hlen + t->tun_hlen;
1744
1745         return 0;
1746 }
1747 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1748
1749 static const struct net_device_ops ip6_tnl_netdev_ops = {
1750         .ndo_init       = ip6_tnl_dev_init,
1751         .ndo_uninit     = ip6_tnl_dev_uninit,
1752         .ndo_start_xmit = ip6_tnl_start_xmit,
1753         .ndo_do_ioctl   = ip6_tnl_ioctl,
1754         .ndo_change_mtu = ip6_tnl_change_mtu,
1755         .ndo_get_stats  = ip6_get_stats,
1756         .ndo_get_iflink = ip6_tnl_get_iflink,
1757 };
1758
1759 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1760                          NETIF_F_FRAGLIST |     \
1761                          NETIF_F_HIGHDMA |      \
1762                          NETIF_F_GSO_SOFTWARE | \
1763                          NETIF_F_HW_CSUM)
1764
1765 /**
1766  * ip6_tnl_dev_setup - setup virtual tunnel device
1767  *   @dev: virtual device associated with tunnel
1768  *
1769  * Description:
1770  *   Initialize function pointers and device parameters
1771  **/
1772
1773 static void ip6_tnl_dev_setup(struct net_device *dev)
1774 {
1775         dev->netdev_ops = &ip6_tnl_netdev_ops;
1776         dev->needs_free_netdev = true;
1777         dev->priv_destructor = ip6_dev_free;
1778
1779         dev->type = ARPHRD_TUNNEL6;
1780         dev->flags |= IFF_NOARP;
1781         dev->addr_len = sizeof(struct in6_addr);
1782         dev->features |= NETIF_F_LLTX;
1783         netif_keep_dst(dev);
1784
1785         dev->features           |= IPXIPX_FEATURES;
1786         dev->hw_features        |= IPXIPX_FEATURES;
1787
1788         /* This perm addr will be used as interface identifier by IPv6 */
1789         dev->addr_assign_type = NET_ADDR_RANDOM;
1790         eth_random_addr(dev->perm_addr);
1791 }
1792
1793
1794 /**
1795  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1796  *   @dev: virtual device associated with tunnel
1797  **/
1798
1799 static inline int
1800 ip6_tnl_dev_init_gen(struct net_device *dev)
1801 {
1802         struct ip6_tnl *t = netdev_priv(dev);
1803         int ret;
1804         int t_hlen;
1805
1806         t->dev = dev;
1807         t->net = dev_net(dev);
1808         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1809         if (!dev->tstats)
1810                 return -ENOMEM;
1811
1812         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1813         if (ret)
1814                 goto free_stats;
1815
1816         ret = gro_cells_init(&t->gro_cells, dev);
1817         if (ret)
1818                 goto destroy_dst;
1819
1820         t->tun_hlen = 0;
1821         t->hlen = t->encap_hlen + t->tun_hlen;
1822         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1823
1824         dev->type = ARPHRD_TUNNEL6;
1825         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1826         dev->mtu = ETH_DATA_LEN - t_hlen;
1827         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1828                 dev->mtu -= 8;
1829         dev->min_mtu = ETH_MIN_MTU;
1830         dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1831
1832         return 0;
1833
1834 destroy_dst:
1835         dst_cache_destroy(&t->dst_cache);
1836 free_stats:
1837         free_percpu(dev->tstats);
1838         dev->tstats = NULL;
1839
1840         return ret;
1841 }
1842
1843 /**
1844  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1845  *   @dev: virtual device associated with tunnel
1846  **/
1847
1848 static int ip6_tnl_dev_init(struct net_device *dev)
1849 {
1850         struct ip6_tnl *t = netdev_priv(dev);
1851         int err = ip6_tnl_dev_init_gen(dev);
1852
1853         if (err)
1854                 return err;
1855         ip6_tnl_link_config(t);
1856         if (t->parms.collect_md) {
1857                 dev->features |= NETIF_F_NETNS_LOCAL;
1858                 netif_keep_dst(dev);
1859         }
1860         return 0;
1861 }
1862
1863 /**
1864  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1865  *   @dev: fallback device
1866  *
1867  * Return: 0
1868  **/
1869
1870 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1871 {
1872         struct ip6_tnl *t = netdev_priv(dev);
1873         struct net *net = dev_net(dev);
1874         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1875
1876         t->parms.proto = IPPROTO_IPV6;
1877         dev_hold(dev);
1878
1879         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1880         return 0;
1881 }
1882
1883 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1884                             struct netlink_ext_ack *extack)
1885 {
1886         u8 proto;
1887
1888         if (!data || !data[IFLA_IPTUN_PROTO])
1889                 return 0;
1890
1891         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1892         if (proto != IPPROTO_IPV6 &&
1893             proto != IPPROTO_IPIP &&
1894             proto != 0)
1895                 return -EINVAL;
1896
1897         return 0;
1898 }
1899
1900 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1901                                   struct __ip6_tnl_parm *parms)
1902 {
1903         memset(parms, 0, sizeof(*parms));
1904
1905         if (!data)
1906                 return;
1907
1908         if (data[IFLA_IPTUN_LINK])
1909                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1910
1911         if (data[IFLA_IPTUN_LOCAL])
1912                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1913
1914         if (data[IFLA_IPTUN_REMOTE])
1915                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1916
1917         if (data[IFLA_IPTUN_TTL])
1918                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1919
1920         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1921                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1922
1923         if (data[IFLA_IPTUN_FLOWINFO])
1924                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1925
1926         if (data[IFLA_IPTUN_FLAGS])
1927                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1928
1929         if (data[IFLA_IPTUN_PROTO])
1930                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1931
1932         if (data[IFLA_IPTUN_COLLECT_METADATA])
1933                 parms->collect_md = true;
1934
1935         if (data[IFLA_IPTUN_FWMARK])
1936                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1937 }
1938
1939 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1940                                         struct ip_tunnel_encap *ipencap)
1941 {
1942         bool ret = false;
1943
1944         memset(ipencap, 0, sizeof(*ipencap));
1945
1946         if (!data)
1947                 return ret;
1948
1949         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1950                 ret = true;
1951                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1952         }
1953
1954         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1955                 ret = true;
1956                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1957         }
1958
1959         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1960                 ret = true;
1961                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1962         }
1963
1964         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1965                 ret = true;
1966                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1967         }
1968
1969         return ret;
1970 }
1971
1972 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1973                            struct nlattr *tb[], struct nlattr *data[],
1974                            struct netlink_ext_ack *extack)
1975 {
1976         struct net *net = dev_net(dev);
1977         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1978         struct ip6_tnl *nt, *t;
1979         struct ip_tunnel_encap ipencap;
1980
1981         nt = netdev_priv(dev);
1982
1983         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1984                 int err = ip6_tnl_encap_setup(nt, &ipencap);
1985
1986                 if (err < 0)
1987                         return err;
1988         }
1989
1990         ip6_tnl_netlink_parms(data, &nt->parms);
1991
1992         if (nt->parms.collect_md) {
1993                 if (rtnl_dereference(ip6n->collect_md_tun))
1994                         return -EEXIST;
1995         } else {
1996                 t = ip6_tnl_locate(net, &nt->parms, 0);
1997                 if (!IS_ERR(t))
1998                         return -EEXIST;
1999         }
2000
2001         return ip6_tnl_create2(dev);
2002 }
2003
2004 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2005                               struct nlattr *data[],
2006                               struct netlink_ext_ack *extack)
2007 {
2008         struct ip6_tnl *t = netdev_priv(dev);
2009         struct __ip6_tnl_parm p;
2010         struct net *net = t->net;
2011         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2012         struct ip_tunnel_encap ipencap;
2013
2014         if (dev == ip6n->fb_tnl_dev)
2015                 return -EINVAL;
2016
2017         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2018                 int err = ip6_tnl_encap_setup(t, &ipencap);
2019
2020                 if (err < 0)
2021                         return err;
2022         }
2023         ip6_tnl_netlink_parms(data, &p);
2024         if (p.collect_md)
2025                 return -EINVAL;
2026
2027         t = ip6_tnl_locate(net, &p, 0);
2028         if (!IS_ERR(t)) {
2029                 if (t->dev != dev)
2030                         return -EEXIST;
2031         } else
2032                 t = netdev_priv(dev);
2033
2034         return ip6_tnl_update(t, &p);
2035 }
2036
2037 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2038 {
2039         struct net *net = dev_net(dev);
2040         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2041
2042         if (dev != ip6n->fb_tnl_dev)
2043                 unregister_netdevice_queue(dev, head);
2044 }
2045
2046 static size_t ip6_tnl_get_size(const struct net_device *dev)
2047 {
2048         return
2049                 /* IFLA_IPTUN_LINK */
2050                 nla_total_size(4) +
2051                 /* IFLA_IPTUN_LOCAL */
2052                 nla_total_size(sizeof(struct in6_addr)) +
2053                 /* IFLA_IPTUN_REMOTE */
2054                 nla_total_size(sizeof(struct in6_addr)) +
2055                 /* IFLA_IPTUN_TTL */
2056                 nla_total_size(1) +
2057                 /* IFLA_IPTUN_ENCAP_LIMIT */
2058                 nla_total_size(1) +
2059                 /* IFLA_IPTUN_FLOWINFO */
2060                 nla_total_size(4) +
2061                 /* IFLA_IPTUN_FLAGS */
2062                 nla_total_size(4) +
2063                 /* IFLA_IPTUN_PROTO */
2064                 nla_total_size(1) +
2065                 /* IFLA_IPTUN_ENCAP_TYPE */
2066                 nla_total_size(2) +
2067                 /* IFLA_IPTUN_ENCAP_FLAGS */
2068                 nla_total_size(2) +
2069                 /* IFLA_IPTUN_ENCAP_SPORT */
2070                 nla_total_size(2) +
2071                 /* IFLA_IPTUN_ENCAP_DPORT */
2072                 nla_total_size(2) +
2073                 /* IFLA_IPTUN_COLLECT_METADATA */
2074                 nla_total_size(0) +
2075                 /* IFLA_IPTUN_FWMARK */
2076                 nla_total_size(4) +
2077                 0;
2078 }
2079
2080 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2081 {
2082         struct ip6_tnl *tunnel = netdev_priv(dev);
2083         struct __ip6_tnl_parm *parm = &tunnel->parms;
2084
2085         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2086             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2087             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2088             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2089             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2090             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2091             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2092             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2093             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2094                 goto nla_put_failure;
2095
2096         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2097             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2098             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2099             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2100                 goto nla_put_failure;
2101
2102         if (parm->collect_md)
2103                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2104                         goto nla_put_failure;
2105
2106         return 0;
2107
2108 nla_put_failure:
2109         return -EMSGSIZE;
2110 }
2111
2112 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2113 {
2114         struct ip6_tnl *tunnel = netdev_priv(dev);
2115
2116         return tunnel->net;
2117 }
2118 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2119
2120 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2121         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2122         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2123         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2124         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2125         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2126         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2127         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2128         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2129         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2130         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2131         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2132         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2133         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2134         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2135 };
2136
2137 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2138         .kind           = "ip6tnl",
2139         .maxtype        = IFLA_IPTUN_MAX,
2140         .policy         = ip6_tnl_policy,
2141         .priv_size      = sizeof(struct ip6_tnl),
2142         .setup          = ip6_tnl_dev_setup,
2143         .validate       = ip6_tnl_validate,
2144         .newlink        = ip6_tnl_newlink,
2145         .changelink     = ip6_tnl_changelink,
2146         .dellink        = ip6_tnl_dellink,
2147         .get_size       = ip6_tnl_get_size,
2148         .fill_info      = ip6_tnl_fill_info,
2149         .get_link_net   = ip6_tnl_get_link_net,
2150 };
2151
2152 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2153         .handler        = ip4ip6_rcv,
2154         .err_handler    = ip4ip6_err,
2155         .priority       =       1,
2156 };
2157
2158 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2159         .handler        = ip6ip6_rcv,
2160         .err_handler    = ip6ip6_err,
2161         .priority       =       1,
2162 };
2163
2164 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2165 {
2166         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2167         struct net_device *dev, *aux;
2168         int h;
2169         struct ip6_tnl *t;
2170
2171         for_each_netdev_safe(net, dev, aux)
2172                 if (dev->rtnl_link_ops == &ip6_link_ops)
2173                         unregister_netdevice_queue(dev, list);
2174
2175         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2176                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2177                 while (t) {
2178                         /* If dev is in the same netns, it has already
2179                          * been added to the list by the previous loop.
2180                          */
2181                         if (!net_eq(dev_net(t->dev), net))
2182                                 unregister_netdevice_queue(t->dev, list);
2183                         t = rtnl_dereference(t->next);
2184                 }
2185         }
2186 }
2187
2188 static int __net_init ip6_tnl_init_net(struct net *net)
2189 {
2190         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2191         struct ip6_tnl *t = NULL;
2192         int err;
2193
2194         ip6n->tnls[0] = ip6n->tnls_wc;
2195         ip6n->tnls[1] = ip6n->tnls_r_l;
2196
2197         err = -ENOMEM;
2198         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2199                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2200
2201         if (!ip6n->fb_tnl_dev)
2202                 goto err_alloc_dev;
2203         dev_net_set(ip6n->fb_tnl_dev, net);
2204         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2205         /* FB netdevice is special: we have one, and only one per netns.
2206          * Allowing to move it to another netns is clearly unsafe.
2207          */
2208         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2209
2210         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2211         if (err < 0)
2212                 goto err_register;
2213
2214         err = register_netdev(ip6n->fb_tnl_dev);
2215         if (err < 0)
2216                 goto err_register;
2217
2218         t = netdev_priv(ip6n->fb_tnl_dev);
2219
2220         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2221         return 0;
2222
2223 err_register:
2224         free_netdev(ip6n->fb_tnl_dev);
2225 err_alloc_dev:
2226         return err;
2227 }
2228
2229 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2230 {
2231         struct net *net;
2232         LIST_HEAD(list);
2233
2234         rtnl_lock();
2235         list_for_each_entry(net, net_list, exit_list)
2236                 ip6_tnl_destroy_tunnels(net, &list);
2237         unregister_netdevice_many(&list);
2238         rtnl_unlock();
2239 }
2240
2241 static struct pernet_operations ip6_tnl_net_ops = {
2242         .init = ip6_tnl_init_net,
2243         .exit_batch = ip6_tnl_exit_batch_net,
2244         .id   = &ip6_tnl_net_id,
2245         .size = sizeof(struct ip6_tnl_net),
2246 };
2247
2248 /**
2249  * ip6_tunnel_init - register protocol and reserve needed resources
2250  *
2251  * Return: 0 on success
2252  **/
2253
2254 static int __init ip6_tunnel_init(void)
2255 {
2256         int  err;
2257
2258         if (!ipv6_mod_enabled())
2259                 return -EOPNOTSUPP;
2260
2261         err = register_pernet_device(&ip6_tnl_net_ops);
2262         if (err < 0)
2263                 goto out_pernet;
2264
2265         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2266         if (err < 0) {
2267                 pr_err("%s: can't register ip4ip6\n", __func__);
2268                 goto out_ip4ip6;
2269         }
2270
2271         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2272         if (err < 0) {
2273                 pr_err("%s: can't register ip6ip6\n", __func__);
2274                 goto out_ip6ip6;
2275         }
2276         err = rtnl_link_register(&ip6_link_ops);
2277         if (err < 0)
2278                 goto rtnl_link_failed;
2279
2280         return 0;
2281
2282 rtnl_link_failed:
2283         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2284 out_ip6ip6:
2285         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2286 out_ip4ip6:
2287         unregister_pernet_device(&ip6_tnl_net_ops);
2288 out_pernet:
2289         return err;
2290 }
2291
2292 /**
2293  * ip6_tunnel_cleanup - free resources and unregister protocol
2294  **/
2295
2296 static void __exit ip6_tunnel_cleanup(void)
2297 {
2298         rtnl_link_unregister(&ip6_link_ops);
2299         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2300                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2301
2302         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2303                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2304
2305         unregister_pernet_device(&ip6_tnl_net_ops);
2306 }
2307
2308 module_init(ip6_tunnel_init);
2309 module_exit(ip6_tunnel_cleanup);