err.no Git - linux-2.6/blob - net/ipv6/ip6_tunnel.c

   1 /*
   2  *      IPv6 over IPv6 tunnel device
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
   7  *
   8  *      $Id$
   9  *
  10  *      Based on:
  11  *      linux/net/ipv6/sit.c
  12  *
  13  *      RFC 2473
  14  *
  15  *      This program is free software; you can redistribute it and/or
  16  *      modify it under the terms of the GNU General Public License
  17  *      as published by the Free Software Foundation; either version
  18  *      2 of the License, or (at your option) any later version.
  19  *
  20  */
  21
  22 #include <linux/module.h>
  23 #include <linux/capability.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/sockios.h>
  27 #include <linux/if.h>
  28 #include <linux/in.h>
  29 #include <linux/ip.h>
  30 #include <linux/if_tunnel.h>
  31 #include <linux/net.h>
  32 #include <linux/in6.h>
  33 #include <linux/netdevice.h>
  34 #include <linux/if_arp.h>
  35 #include <linux/icmpv6.h>
  36 #include <linux/init.h>
  37 #include <linux/route.h>
  38 #include <linux/rtnetlink.h>
  39 #include <linux/netfilter_ipv6.h>
  40
  41 #include <asm/uaccess.h>
  42 #include <asm/atomic.h>
  43
  44 #include <net/ip.h>
  45 #include <net/ipv6.h>
  46 #include <net/ip6_route.h>
  47 #include <net/addrconf.h>
  48 #include <net/ip6_tunnel.h>
  49 #include <net/xfrm.h>
  50 #include <net/dsfield.h>
  51 #include <net/inet_ecn.h>
  52
  53 MODULE_AUTHOR("Ville Nuorvala");
  54 MODULE_DESCRIPTION("IPv6-in-IPv6 tunnel");
  55 MODULE_LICENSE("GPL");
  56
  57 #define IPV6_TLV_TEL_DST_SIZE 8
  58
  59 #ifdef IP6_TNL_DEBUG
  60 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __FUNCTION__)
  61 #else
  62 #define IP6_TNL_TRACE(x...) do {;} while(0)
  63 #endif
  64
  65 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
  66
  67 #define HASH_SIZE  32
  68
  69 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
  70                      (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
  71                     (HASH_SIZE - 1))
  72
  73 static int ip6ip6_fb_tnl_dev_init(struct net_device *dev);
  74 static int ip6ip6_tnl_dev_init(struct net_device *dev);
  75 static void ip6ip6_tnl_dev_setup(struct net_device *dev);
  76
  77 /* the IPv6 tunnel fallback device */
  78 static struct net_device *ip6ip6_fb_tnl_dev;
  79
  80
  81 /* lists for storing tunnels in use */
  82 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
  83 static struct ip6_tnl *tnls_wc[1];
  84 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
  85
  86 /* lock for the tunnel lists */
  87 static DEFINE_RWLOCK(ip6ip6_lock);
  88
  89 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
  90 {
  91         struct dst_entry *dst = t->dst_cache;
  92
  93         if (dst && dst->obsolete &&
  94             dst->ops->check(dst, t->dst_cookie) == NULL) {
  95                 t->dst_cache = NULL;
  96                 dst_release(dst);
  97                 return NULL;
  98         }
  99
 100         return dst;
 101 }
 102
 103 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
 104 {
 105         dst_release(t->dst_cache);
 106         t->dst_cache = NULL;
 107 }
 108
 109 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
 110 {
 111         struct rt6_info *rt = (struct rt6_info *) dst;
 112         t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
 113         dst_release(t->dst_cache);
 114         t->dst_cache = dst;
 115 }
 116
 117 /**
 118  * ip6ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 119  *   @remote: the address of the tunnel exit-point
 120  *   @local: the address of the tunnel entry-point
 121  *
 122  * Return:
 123  *   tunnel matching given end-points if found,
 124  *   else fallback tunnel if its device is up,
 125  *   else %NULL
 126  **/
 127
 128 static struct ip6_tnl *
 129 ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
 130 {
 131         unsigned h0 = HASH(remote);
 132         unsigned h1 = HASH(local);
 133         struct ip6_tnl *t;
 134
 135         for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
 136                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 137                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 138                     (t->dev->flags & IFF_UP))
 139                         return t;
 140         }
 141         if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
 142                 return t;
 143
 144         return NULL;
 145 }
 146
 147 /**
 148  * ip6ip6_bucket - get head of list matching given tunnel parameters
 149  *   @p: parameters containing tunnel end-points
 150  *
 151  * Description:
 152  *   ip6ip6_bucket() returns the head of the list matching the
 153  *   &struct in6_addr entries laddr and raddr in @p.
 154  *
 155  * Return: head of IPv6 tunnel list
 156  **/
 157
 158 static struct ip6_tnl **
 159 ip6ip6_bucket(struct ip6_tnl_parm *p)
 160 {
 161         struct in6_addr *remote = &p->raddr;
 162         struct in6_addr *local = &p->laddr;
 163         unsigned h = 0;
 164         int prio = 0;
 165
 166         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 167                 prio = 1;
 168                 h = HASH(remote) ^ HASH(local);
 169         }
 170         return &tnls[prio][h];
 171 }
 172
 173 /**
 174  * ip6ip6_tnl_link - add tunnel to hash table
 175  *   @t: tunnel to be added
 176  **/
 177
 178 static void
 179 ip6ip6_tnl_link(struct ip6_tnl *t)
 180 {
 181         struct ip6_tnl **tp = ip6ip6_bucket(&t->parms);
 182
 183         t->next = *tp;
 184         write_lock_bh(&ip6ip6_lock);
 185         *tp = t;
 186         write_unlock_bh(&ip6ip6_lock);
 187 }
 188
 189 /**
 190  * ip6ip6_tnl_unlink - remove tunnel from hash table
 191  *   @t: tunnel to be removed
 192  **/
 193
 194 static void
 195 ip6ip6_tnl_unlink(struct ip6_tnl *t)
 196 {
 197         struct ip6_tnl **tp;
 198
 199         for (tp = ip6ip6_bucket(&t->parms); *tp; tp = &(*tp)->next) {
 200                 if (t == *tp) {
 201                         write_lock_bh(&ip6ip6_lock);
 202                         *tp = t->next;
 203                         write_unlock_bh(&ip6ip6_lock);
 204                         break;
 205                 }
 206         }
 207 }
 208
 209 /**
 210  * ip6_tnl_create() - create a new tunnel
 211  *   @p: tunnel parameters
 212  *   @pt: pointer to new tunnel
 213  *
 214  * Description:
 215  *   Create tunnel matching given parameters.
 216  *
 217  * Return:
 218  *   created tunnel or NULL
 219  **/
 220
 221 static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
 222 {
 223         struct net_device *dev;
 224         struct ip6_tnl *t;
 225         char name[IFNAMSIZ];
 226         int err;
 227
 228         if (p->name[0]) {
 229                 strlcpy(name, p->name, IFNAMSIZ);
 230         } else {
 231                 int i;
 232                 for (i = 1; i < IP6_TNL_MAX; i++) {
 233                         sprintf(name, "ip6tnl%d", i);
 234                         if (__dev_get_by_name(name) == NULL)
 235                                 break;
 236                 }
 237                 if (i == IP6_TNL_MAX)
 238                         goto failed;
 239         }
 240         dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
 241         if (dev == NULL)
 242                 goto failed;
 243
 244         t = netdev_priv(dev);
 245         dev->init = ip6ip6_tnl_dev_init;
 246         t->parms = *p;
 247
 248         if ((err = register_netdevice(dev)) < 0) {
 249                 free_netdev(dev);
 250                 goto failed;
 251         }
 252         dev_hold(dev);
 253         ip6ip6_tnl_link(t);
 254         return t;
 255 failed:
 256         return NULL;
 257 }
 258
 259 /**
 260  * ip6ip6_tnl_locate - find or create tunnel matching given parameters
 261  *   @p: tunnel parameters
 262  *   @create: != 0 if allowed to create new tunnel if no match found
 263  *
 264  * Description:
 265  *   ip6ip6_tnl_locate() first tries to locate an existing tunnel
 266  *   based on @parms. If this is unsuccessful, but @create is set a new
 267  *   tunnel device is created and registered for use.
 268  *
 269  * Return:
 270  *   matching tunnel or NULL
 271  **/
 272
 273 static struct ip6_tnl *ip6ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
 274 {
 275         struct in6_addr *remote = &p->raddr;
 276         struct in6_addr *local = &p->laddr;
 277         struct ip6_tnl *t;
 278
 279         for (t = *ip6ip6_bucket(p); t; t = t->next) {
 280                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 281                     ipv6_addr_equal(remote, &t->parms.raddr))
 282                         return t;
 283         }
 284         if (!create)
 285                 return NULL;
 286         return ip6_tnl_create(p);
 287 }
 288
 289 /**
 290  * ip6ip6_tnl_dev_uninit - tunnel device uninitializer
 291  *   @dev: the device to be destroyed
 292  *
 293  * Description:
 294  *   ip6ip6_tnl_dev_uninit() removes tunnel from its list
 295  **/
 296
 297 static void
 298 ip6ip6_tnl_dev_uninit(struct net_device *dev)
 299 {
 300         struct ip6_tnl *t = netdev_priv(dev);
 301
 302         if (dev == ip6ip6_fb_tnl_dev) {
 303                 write_lock_bh(&ip6ip6_lock);
 304                 tnls_wc[0] = NULL;
 305                 write_unlock_bh(&ip6ip6_lock);
 306         } else {
 307                 ip6ip6_tnl_unlink(t);
 308         }
 309         ip6_tnl_dst_reset(t);
 310         dev_put(dev);
 311 }
 312
 313 /**
 314  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 315  *   @skb: received socket buffer
 316  *
 317  * Return:
 318  *   0 if none was found,
 319  *   else index to encapsulation limit
 320  **/
 321
 322 static __u16
 323 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
 324 {
 325         struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
 326         __u8 nexthdr = ipv6h->nexthdr;
 327         __u16 off = sizeof (*ipv6h);
 328
 329         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 330                 __u16 optlen = 0;
 331                 struct ipv6_opt_hdr *hdr;
 332                 if (raw + off + sizeof (*hdr) > skb->data &&
 333                     !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
 334                         break;
 335
 336                 hdr = (struct ipv6_opt_hdr *) (raw + off);
 337                 if (nexthdr == NEXTHDR_FRAGMENT) {
 338                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 339                         if (frag_hdr->frag_off)
 340                                 break;
 341                         optlen = 8;
 342                 } else if (nexthdr == NEXTHDR_AUTH) {
 343                         optlen = (hdr->hdrlen + 2) << 2;
 344                 } else {
 345                         optlen = ipv6_optlen(hdr);
 346                 }
 347                 if (nexthdr == NEXTHDR_DEST) {
 348                         __u16 i = off + 2;
 349                         while (1) {
 350                                 struct ipv6_tlv_tnl_enc_lim *tel;
 351
 352                                 /* No more room for encapsulation limit */
 353                                 if (i + sizeof (*tel) > off + optlen)
 354                                         break;
 355
 356                                 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
 357                                 /* return index of option if found and valid */
 358                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 359                                     tel->length == 1)
 360                                         return i;
 361                                 /* else jump to next option */
 362                                 if (tel->type)
 363                                         i += tel->length + 2;
 364                                 else
 365                                         i++;
 366                         }
 367                 }
 368                 nexthdr = hdr->nexthdr;
 369                 off += optlen;
 370         }
 371         return 0;
 372 }
 373
 374 /**
 375  * ip6ip6_err - tunnel error handler
 376  *
 377  * Description:
 378  *   ip6ip6_err() should handle errors in the tunnel according
 379  *   to the specifications in RFC 2473.
 380  **/
 381
 382 static int
 383 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 384            int type, int code, int offset, __be32 info)
 385 {
 386         struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
 387         struct ip6_tnl *t;
 388         int rel_msg = 0;
 389         int rel_type = ICMPV6_DEST_UNREACH;
 390         int rel_code = ICMPV6_ADDR_UNREACH;
 391         __u32 rel_info = 0;
 392         __u16 len;
 393         int err = -ENOENT;
 394
 395         /* If the packet doesn't contain the original IPv6 header we are
 396            in trouble since we might need the source address for further
 397            processing of the error. */
 398
 399         read_lock(&ip6ip6_lock);
 400         if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
 401                 goto out;
 402
 403         err = 0;
 404
 405         switch (type) {
 406                 __u32 teli;
 407                 struct ipv6_tlv_tnl_enc_lim *tel;
 408                 __u32 mtu;
 409         case ICMPV6_DEST_UNREACH:
 410                 if (net_ratelimit())
 411                         printk(KERN_WARNING
 412                                "%s: Path to destination invalid "
 413                                "or inactive!\n", t->parms.name);
 414                 rel_msg = 1;
 415                 break;
 416         case ICMPV6_TIME_EXCEED:
 417                 if (code == ICMPV6_EXC_HOPLIMIT) {
 418                         if (net_ratelimit())
 419                                 printk(KERN_WARNING
 420                                        "%s: Too small hop limit or "
 421                                        "routing loop in tunnel!\n",
 422                                        t->parms.name);
 423                         rel_msg = 1;
 424                 }
 425                 break;
 426         case ICMPV6_PARAMPROB:
 427                 /* ignore if parameter problem not caused by a tunnel
 428                    encapsulation limit sub-option */
 429                 if (code != ICMPV6_HDR_FIELD) {
 430                         break;
 431                 }
 432                 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
 433
 434                 if (teli && teli == ntohl(info) - 2) {
 435                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 436                         if (tel->encap_limit == 0) {
 437                                 if (net_ratelimit())
 438                                         printk(KERN_WARNING
 439                                                "%s: Too small encapsulation "
 440                                                "limit or routing loop in "
 441                                                "tunnel!\n", t->parms.name);
 442                                 rel_msg = 1;
 443                         }
 444                 }
 445                 break;
 446         case ICMPV6_PKT_TOOBIG:
 447                 mtu = ntohl(info) - offset;
 448                 if (mtu < IPV6_MIN_MTU)
 449                         mtu = IPV6_MIN_MTU;
 450                 t->dev->mtu = mtu;
 451
 452                 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
 453                         rel_type = ICMPV6_PKT_TOOBIG;
 454                         rel_code = 0;
 455                         rel_info = mtu;
 456                         rel_msg = 1;
 457                 }
 458                 break;
 459         }
 460         if (rel_msg &&  pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
 461                 struct rt6_info *rt;
 462                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 463                 if (!skb2)
 464                         goto out;
 465
 466                 dst_release(skb2->dst);
 467                 skb2->dst = NULL;
 468                 skb_pull(skb2, offset);
 469                 skb2->nh.raw = skb2->data;
 470
 471                 /* Try to guess incoming interface */
 472                 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
 473
 474                 if (rt && rt->rt6i_dev)
 475                         skb2->dev = rt->rt6i_dev;
 476
 477                 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 478
 479                 if (rt)
 480                         dst_release(&rt->u.dst);
 481
 482                 kfree_skb(skb2);
 483         }
 484 out:
 485         read_unlock(&ip6ip6_lock);
 486         return err;
 487 }
 488
 489 static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
 490                                           struct sk_buff *skb)
 491 {
 492         struct ipv6hdr *inner_iph = skb->nh.ipv6h;
 493
 494         if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
 495                 IP6_ECN_set_ce(inner_iph);
 496 }
 497 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
 498 {
 499         struct ip6_tnl_parm *p = &t->parms;
 500         int ret = 0;
 501
 502         if (p->flags & IP6_TNL_F_CAP_RCV) {
 503                 struct net_device *ldev = NULL;
 504
 505                 if (p->link)
 506                         ldev = dev_get_by_index(p->link);
 507
 508                 if ((ipv6_addr_is_multicast(&p->laddr) ||
 509                      likely(ipv6_chk_addr(&p->laddr, ldev, 0))) &&
 510                     likely(!ipv6_chk_addr(&p->raddr, NULL, 0)))
 511                         ret = 1;
 512
 513                 if (ldev)
 514                         dev_put(ldev);
 515         }
 516         return ret;
 517 }
 518
 519 /**
 520  * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
 521  *   @skb: received socket buffer
 522  *
 523  * Return: 0
 524  **/
 525
 526 static int
 527 ip6ip6_rcv(struct sk_buff *skb)
 528 {
 529         struct ipv6hdr *ipv6h;
 530         struct ip6_tnl *t;
 531
 532         ipv6h = skb->nh.ipv6h;
 533
 534         read_lock(&ip6ip6_lock);
 535
 536         if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
 537                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 538                         read_unlock(&ip6ip6_lock);
 539                         goto discard;
 540                 }
 541
 542                 if (!ip6_tnl_rcv_ctl(t)) {
 543                         t->stat.rx_dropped++;
 544                         read_unlock(&ip6ip6_lock);
 545                         goto discard;
 546                 }
 547                 secpath_reset(skb);
 548                 skb->mac.raw = skb->nh.raw;
 549                 skb->nh.raw = skb->data;
 550                 skb->protocol = htons(ETH_P_IPV6);
 551                 skb->pkt_type = PACKET_HOST;
 552                 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 553                 skb->dev = t->dev;
 554                 dst_release(skb->dst);
 555                 skb->dst = NULL;
 556                 nf_reset(skb);
 557                 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 558                         ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
 559                 ip6ip6_ecn_decapsulate(ipv6h, skb);
 560                 t->stat.rx_packets++;
 561                 t->stat.rx_bytes += skb->len;
 562                 netif_rx(skb);
 563                 read_unlock(&ip6ip6_lock);
 564                 return 0;
 565         }
 566         read_unlock(&ip6ip6_lock);
 567         return 1;
 568
 569 discard:
 570         kfree_skb(skb);
 571         return 0;
 572 }
 573
 574 static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
 575 {
 576         struct ipv6_tlv_tnl_enc_lim *tel;
 577         struct ipv6_txoptions *opt;
 578         __u8 *raw;
 579
 580         int opt_len = sizeof(*opt) + 8;
 581
 582         if (!(opt = kzalloc(opt_len, GFP_ATOMIC))) {
 583                 return NULL;
 584         }
 585         opt->tot_len = opt_len;
 586         opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
 587         opt->opt_nflen = 8;
 588
 589         tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1);
 590         tel->type = IPV6_TLV_TNL_ENCAP_LIMIT;
 591         tel->length = 1;
 592         tel->encap_limit = encap_limit;
 593
 594         raw = (__u8 *) opt->dst0opt;
 595         raw[5] = IPV6_TLV_PADN;
 596         raw[6] = 1;
 597
 598         return opt;
 599 }
 600
 601 /**
 602  * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 603  *   @t: the outgoing tunnel device
 604  *   @hdr: IPv6 header from the incoming packet
 605  *
 606  * Description:
 607  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 608  *   doesn't match source of incoming packet.
 609  *
 610  * Return:
 611  *   1 if conflict,
 612  *   0 else
 613  **/
 614
 615 static inline int
 616 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
 617 {
 618         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 619 }
 620
 621 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
 622 {
 623         struct ip6_tnl_parm *p = &t->parms;
 624         int ret = 0;
 625
 626         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 627                 struct net_device *ldev = NULL;
 628
 629                 if (p->link)
 630                         ldev = dev_get_by_index(p->link);
 631
 632                 if (unlikely(!ipv6_chk_addr(&p->laddr, ldev, 0)))
 633                         printk(KERN_WARNING
 634                                "%s xmit: Local address not yet configured!\n",
 635                                p->name);
 636                 else if (!ipv6_addr_is_multicast(&p->raddr) &&
 637                          unlikely(ipv6_chk_addr(&p->raddr, NULL, 0)))
 638                         printk(KERN_WARNING
 639                                "%s xmit: Routing loop! "
 640                                "Remote address found on this node!\n",
 641                                p->name);
 642                 else
 643                         ret = 1;
 644                 if (ldev)
 645                         dev_put(ldev);
 646         }
 647         return ret;
 648 }
 649 /**
 650  * ip6ip6_tnl_xmit - encapsulate packet and send
 651  *   @skb: the outgoing socket buffer
 652  *   @dev: the outgoing tunnel device
 653  *
 654  * Description:
 655  *   Build new header and do some sanity checks on the packet before sending
 656  *   it.
 657  *
 658  * Return:
 659  *   0
 660  **/
 661
 662 static int
 663 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 664 {
 665         struct ip6_tnl *t = netdev_priv(dev);
 666         struct net_device_stats *stats = &t->stat;
 667         struct ipv6hdr *ipv6h = skb->nh.ipv6h;
 668         struct ipv6_txoptions *opt = NULL;
 669         int encap_limit = -1;
 670         __u16 offset;
 671         struct flowi fl;
 672         struct dst_entry *dst;
 673         struct net_device *tdev;
 674         int mtu;
 675         int max_headroom = sizeof(struct ipv6hdr);
 676         u8 proto;
 677         int err;
 678         int pkt_len;
 679         int dsfield;
 680
 681         if (t->recursion++) {
 682                 stats->collisions++;
 683                 goto tx_err;
 684         }
 685         if (skb->protocol != htons(ETH_P_IPV6) ||
 686             !ip6_tnl_xmit_ctl(t) || ip6ip6_tnl_addr_conflict(t, ipv6h))
 687                 goto tx_err;
 688
 689         if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
 690                 struct ipv6_tlv_tnl_enc_lim *tel;
 691                 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
 692                 if (tel->encap_limit == 0) {
 693                         icmpv6_send(skb, ICMPV6_PARAMPROB,
 694                                     ICMPV6_HDR_FIELD, offset + 2, skb->dev);
 695                         goto tx_err;
 696                 }
 697                 encap_limit = tel->encap_limit - 1;
 698         } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
 699                 encap_limit = t->parms.encap_limit;
 700         }
 701         memcpy(&fl, &t->fl, sizeof (fl));
 702         proto = fl.proto;
 703
 704         dsfield = ipv6_get_dsfield(ipv6h);
 705         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
 706                 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 707         if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
 708                 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 709
 710         if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
 711                 goto tx_err;
 712
 713         if ((dst = ip6_tnl_dst_check(t)) != NULL)
 714                 dst_hold(dst);
 715         else {
 716                 dst = ip6_route_output(NULL, &fl);
 717
 718                 if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
 719                         goto tx_err_link_failure;
 720         }
 721
 722         tdev = dst->dev;
 723
 724         if (tdev == dev) {
 725                 stats->collisions++;
 726                 if (net_ratelimit())
 727                         printk(KERN_WARNING
 728                                "%s: Local routing loop detected!\n",
 729                                t->parms.name);
 730                 goto tx_err_dst_release;
 731         }
 732         mtu = dst_mtu(dst) - sizeof (*ipv6h);
 733         if (opt) {
 734                 max_headroom += 8;
 735                 mtu -= 8;
 736         }
 737         if (mtu < IPV6_MIN_MTU)
 738                 mtu = IPV6_MIN_MTU;
 739         if (skb->dst && mtu < dst_mtu(skb->dst)) {
 740                 struct rt6_info *rt = (struct rt6_info *) skb->dst;
 741                 rt->rt6i_flags |= RTF_MODIFIED;
 742                 rt->u.dst.metrics[RTAX_MTU-1] = mtu;
 743         }
 744         if (skb->len > mtu) {
 745                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 746                 goto tx_err_dst_release;
 747         }
 748
 749         /*
 750          * Okay, now see if we can stuff it in the buffer as-is.
 751          */
 752         max_headroom += LL_RESERVED_SPACE(tdev);
 753
 754         if (skb_headroom(skb) < max_headroom ||
 755             skb_cloned(skb) || skb_shared(skb)) {
 756                 struct sk_buff *new_skb;
 757
 758                 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
 759                         goto tx_err_dst_release;
 760
 761                 if (skb->sk)
 762                         skb_set_owner_w(new_skb, skb->sk);
 763                 kfree_skb(skb);
 764                 skb = new_skb;
 765         }
 766         dst_release(skb->dst);
 767         skb->dst = dst_clone(dst);
 768
 769         skb->h.raw = skb->nh.raw;
 770
 771         if (opt)
 772                 ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
 773
 774         skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
 775         ipv6h = skb->nh.ipv6h;
 776         *(__be32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
 777         dsfield = INET_ECN_encapsulate(0, dsfield);
 778         ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
 779         ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
 780         ipv6h->hop_limit = t->parms.hop_limit;
 781         ipv6h->nexthdr = proto;
 782         ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
 783         ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
 784         nf_reset(skb);
 785         pkt_len = skb->len;
 786         err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
 787                       skb->dst->dev, dst_output);
 788
 789         if (net_xmit_eval(err) == 0) {
 790                 stats->tx_bytes += pkt_len;
 791                 stats->tx_packets++;
 792         } else {
 793                 stats->tx_errors++;
 794                 stats->tx_aborted_errors++;
 795         }
 796         ip6_tnl_dst_store(t, dst);
 797
 798         kfree(opt);
 799
 800         t->recursion--;
 801         return 0;
 802 tx_err_link_failure:
 803         stats->tx_carrier_errors++;
 804         dst_link_failure(skb);
 805 tx_err_dst_release:
 806         dst_release(dst);
 807         kfree(opt);
 808 tx_err:
 809         stats->tx_errors++;
 810         stats->tx_dropped++;
 811         kfree_skb(skb);
 812         t->recursion--;
 813         return 0;
 814 }
 815
 816 static void ip6_tnl_set_cap(struct ip6_tnl *t)
 817 {
 818         struct ip6_tnl_parm *p = &t->parms;
 819         int ltype = ipv6_addr_type(&p->laddr);
 820         int rtype = ipv6_addr_type(&p->raddr);
 821
 822         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
 823
 824         if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 825             rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 826             !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 827             !((ltype|rtype) & IPV6_ADDR_LINKLOCAL)) {
 828                 if (ltype&IPV6_ADDR_UNICAST)
 829                         p->flags |= IP6_TNL_F_CAP_XMIT;
 830                 if (rtype&IPV6_ADDR_UNICAST)
 831                         p->flags |= IP6_TNL_F_CAP_RCV;
 832         }
 833 }
 834
 835 static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
 836 {
 837         struct net_device *dev = t->dev;
 838         struct ip6_tnl_parm *p = &t->parms;
 839         struct flowi *fl = &t->fl;
 840
 841         memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 842         memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 843
 844         /* Set up flowi template */
 845         ipv6_addr_copy(&fl->fl6_src, &p->laddr);
 846         ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
 847         fl->oif = p->link;
 848         fl->fl6_flowlabel = 0;
 849
 850         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
 851                 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
 852         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
 853                 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
 854
 855         ip6_tnl_set_cap(t);
 856
 857         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
 858                 dev->flags |= IFF_POINTOPOINT;
 859         else
 860                 dev->flags &= ~IFF_POINTOPOINT;
 861
 862         dev->iflink = p->link;
 863
 864         if (p->flags & IP6_TNL_F_CAP_XMIT) {
 865                 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
 866                                                  p->link, 0);
 867
 868                 if (rt == NULL)
 869                         return;
 870
 871                 if (rt->rt6i_dev) {
 872                         dev->hard_header_len = rt->rt6i_dev->hard_header_len +
 873                                 sizeof (struct ipv6hdr);
 874
 875                         dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
 876
 877                         if (dev->mtu < IPV6_MIN_MTU)
 878                                 dev->mtu = IPV6_MIN_MTU;
 879                 }
 880                 dst_release(&rt->u.dst);
 881         }
 882 }
 883
 884 /**
 885  * ip6ip6_tnl_change - update the tunnel parameters
 886  *   @t: tunnel to be changed
 887  *   @p: tunnel configuration parameters
 888  *   @active: != 0 if tunnel is ready for use
 889  *
 890  * Description:
 891  *   ip6ip6_tnl_change() updates the tunnel parameters
 892  **/
 893
 894 static int
 895 ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
 896 {
 897         ipv6_addr_copy(&t->parms.laddr, &p->laddr);
 898         ipv6_addr_copy(&t->parms.raddr, &p->raddr);
 899         t->parms.flags = p->flags;
 900         t->parms.hop_limit = p->hop_limit;
 901         t->parms.encap_limit = p->encap_limit;
 902         t->parms.flowinfo = p->flowinfo;
 903         t->parms.link = p->link;
 904         ip6_tnl_dst_reset(t);
 905         ip6ip6_tnl_link_config(t);
 906         return 0;
 907 }
 908
 909 /**
 910  * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
 911  *   @dev: virtual device associated with tunnel
 912  *   @ifr: parameters passed from userspace
 913  *   @cmd: command to be performed
 914  *
 915  * Description:
 916  *   ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
 917  *   from userspace.
 918  *
 919  *   The possible commands are the following:
 920  *     %SIOCGETTUNNEL: get tunnel parameters for device
 921  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
 922  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
 923  *     %SIOCDELTUNNEL: delete tunnel
 924  *
 925  *   The fallback device "ip6tnl0", created during module
 926  *   initialization, can be used for creating other tunnel devices.
 927  *
 928  * Return:
 929  *   0 on success,
 930  *   %-EFAULT if unable to copy data to or from userspace,
 931  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
 932  *   %-EINVAL if passed tunnel parameters are invalid,
 933  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
 934  *   %-ENODEV if attempting to change or delete a nonexisting device
 935  **/
 936
 937 static int
 938 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 939 {
 940         int err = 0;
 941         struct ip6_tnl_parm p;
 942         struct ip6_tnl *t = NULL;
 943
 944         switch (cmd) {
 945         case SIOCGETTUNNEL:
 946                 if (dev == ip6ip6_fb_tnl_dev) {
 947                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
 948                                 err = -EFAULT;
 949                                 break;
 950                         }
 951                         t = ip6ip6_tnl_locate(&p, 0);
 952                 }
 953                 if (t == NULL)
 954                         t = netdev_priv(dev);
 955                 memcpy(&p, &t->parms, sizeof (p));
 956                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
 957                         err = -EFAULT;
 958                 }
 959                 break;
 960         case SIOCADDTUNNEL:
 961         case SIOCCHGTUNNEL:
 962                 err = -EPERM;
 963                 if (!capable(CAP_NET_ADMIN))
 964                         break;
 965                 err = -EFAULT;
 966                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
 967                         break;
 968                 err = -EINVAL;
 969                 if (p.proto != IPPROTO_IPV6)
 970                         break;
 971                 t = ip6ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
 972                 if (dev != ip6ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
 973                         if (t != NULL) {
 974                                 if (t->dev != dev) {
 975                                         err = -EEXIST;
 976                                         break;
 977                                 }
 978                         } else
 979                                 t = netdev_priv(dev);
 980
 981                         ip6ip6_tnl_unlink(t);
 982                         err = ip6ip6_tnl_change(t, &p);
 983                         ip6ip6_tnl_link(t);
 984                         netdev_state_change(dev);
 985                 }
 986                 if (t) {
 987                         err = 0;
 988                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
 989                                 err = -EFAULT;
 990
 991                 } else
 992                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 993                 break;
 994         case SIOCDELTUNNEL:
 995                 err = -EPERM;
 996                 if (!capable(CAP_NET_ADMIN))
 997                         break;
 998
 999                 if (dev == ip6ip6_fb_tnl_dev) {
1000                         err = -EFAULT;
1001                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1002                                 break;
1003                         err = -ENOENT;
1004                         if ((t = ip6ip6_tnl_locate(&p, 0)) == NULL)
1005                                 break;
1006                         err = -EPERM;
1007                         if (t->dev == ip6ip6_fb_tnl_dev)
1008                                 break;
1009                         dev = t->dev;
1010                 }
1011                 err = unregister_netdevice(dev);
1012                 break;
1013         default:
1014                 err = -EINVAL;
1015         }
1016         return err;
1017 }
1018
1019 /**
1020  * ip6ip6_tnl_get_stats - return the stats for tunnel device
1021  *   @dev: virtual device associated with tunnel
1022  *
1023  * Return: stats for device
1024  **/
1025
1026 static struct net_device_stats *
1027 ip6ip6_tnl_get_stats(struct net_device *dev)
1028 {
1029         return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1030 }
1031
1032 /**
1033  * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1034  *   @dev: virtual device associated with tunnel
1035  *   @new_mtu: the new mtu
1036  *
1037  * Return:
1038  *   0 on success,
1039  *   %-EINVAL if mtu too small
1040  **/
1041
1042 static int
1043 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1044 {
1045         if (new_mtu < IPV6_MIN_MTU) {
1046                 return -EINVAL;
1047         }
1048         dev->mtu = new_mtu;
1049         return 0;
1050 }
1051
1052 /**
1053  * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1054  *   @dev: virtual device associated with tunnel
1055  *
1056  * Description:
1057  *   Initialize function pointers and device parameters
1058  **/
1059
1060 static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1061 {
1062         SET_MODULE_OWNER(dev);
1063         dev->uninit = ip6ip6_tnl_dev_uninit;
1064         dev->destructor = free_netdev;
1065         dev->hard_start_xmit = ip6ip6_tnl_xmit;
1066         dev->get_stats = ip6ip6_tnl_get_stats;
1067         dev->do_ioctl = ip6ip6_tnl_ioctl;
1068         dev->change_mtu = ip6ip6_tnl_change_mtu;
1069
1070         dev->type = ARPHRD_TUNNEL6;
1071         dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1072         dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1073         dev->flags |= IFF_NOARP;
1074         dev->addr_len = sizeof(struct in6_addr);
1075 }
1076
1077
1078 /**
1079  * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1080  *   @dev: virtual device associated with tunnel
1081  **/
1082
1083 static inline void
1084 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1085 {
1086         struct ip6_tnl *t = netdev_priv(dev);
1087         t->fl.proto = IPPROTO_IPV6;
1088         t->dev = dev;
1089         strcpy(t->parms.name, dev->name);
1090 }
1091
1092 /**
1093  * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1094  *   @dev: virtual device associated with tunnel
1095  **/
1096
1097 static int
1098 ip6ip6_tnl_dev_init(struct net_device *dev)
1099 {
1100         struct ip6_tnl *t = netdev_priv(dev);
1101         ip6ip6_tnl_dev_init_gen(dev);
1102         ip6ip6_tnl_link_config(t);
1103         return 0;
1104 }
1105
1106 /**
1107  * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1108  *   @dev: fallback device
1109  *
1110  * Return: 0
1111  **/
1112
1113 static int
1114 ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1115 {
1116         struct ip6_tnl *t = netdev_priv(dev);
1117         ip6ip6_tnl_dev_init_gen(dev);
1118         dev_hold(dev);
1119         tnls_wc[0] = t;
1120         return 0;
1121 }
1122
1123 static struct xfrm6_tunnel ip6ip6_handler = {
1124         .handler        = ip6ip6_rcv,
1125         .err_handler    = ip6ip6_err,
1126         .priority       =       1,
1127 };
1128
1129 /**
1130  * ip6_tunnel_init - register protocol and reserve needed resources
1131  *
1132  * Return: 0 on success
1133  **/
1134
1135 static int __init ip6_tunnel_init(void)
1136 {
1137         int  err;
1138
1139         if (xfrm6_tunnel_register(&ip6ip6_handler)) {
1140                 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1141                 return -EAGAIN;
1142         }
1143         ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1144                                          ip6ip6_tnl_dev_setup);
1145
1146         if (!ip6ip6_fb_tnl_dev) {
1147                 err = -ENOMEM;
1148                 goto fail;
1149         }
1150         ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1151
1152         if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1153                 free_netdev(ip6ip6_fb_tnl_dev);
1154                 goto fail;
1155         }
1156         return 0;
1157 fail:
1158         xfrm6_tunnel_deregister(&ip6ip6_handler);
1159         return err;
1160 }
1161
1162 static void __exit ip6ip6_destroy_tunnels(void)
1163 {
1164         int h;
1165         struct ip6_tnl *t;
1166
1167         for (h = 0; h < HASH_SIZE; h++) {
1168                 while ((t = tnls_r_l[h]) != NULL)
1169                         unregister_netdevice(t->dev);
1170         }
1171
1172         t = tnls_wc[0];
1173         unregister_netdevice(t->dev);
1174 }
1175
1176 /**
1177  * ip6_tunnel_cleanup - free resources and unregister protocol
1178  **/
1179
1180 static void __exit ip6_tunnel_cleanup(void)
1181 {
1182         if (xfrm6_tunnel_deregister(&ip6ip6_handler))
1183                 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1184
1185         rtnl_lock();
1186         ip6ip6_destroy_tunnels();
1187         rtnl_unlock();
1188 }
1189
1190 module_init(ip6_tunnel_init);
1191 module_exit(ip6_tunnel_cleanup);