Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

at v2.6.27 1516 lines 36 kB view raw
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#include <linux/module.h> 22#include <linux/capability.h> 23#include <linux/errno.h> 24#include <linux/types.h> 25#include <linux/sockios.h> 26#include <linux/icmp.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/icmp.h> 45#include <net/ip.h> 46#include <net/ipv6.h> 47#include <net/ip6_route.h> 48#include <net/addrconf.h> 49#include <net/ip6_tunnel.h> 50#include <net/xfrm.h> 51#include <net/dsfield.h> 52#include <net/inet_ecn.h> 53#include <net/net_namespace.h> 54#include <net/netns/generic.h> 55 56MODULE_AUTHOR("Ville Nuorvala"); 57MODULE_DESCRIPTION("IPv6 tunneling device"); 58MODULE_LICENSE("GPL"); 59 60#define IPV6_TLV_TEL_DST_SIZE 8 61 62#ifdef IP6_TNL_DEBUG 63#define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__) 64#else 65#define IP6_TNL_TRACE(x...) do {;} while(0) 66#endif 67 68#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 69#define IPV6_TCLASS_SHIFT 20 70 71#define HASH_SIZE 32 72 73#define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \ 74 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \ 75 (HASH_SIZE - 1)) 76 77static int ip6_fb_tnl_dev_init(struct net_device *dev); 78static int ip6_tnl_dev_init(struct net_device *dev); 79static void ip6_tnl_dev_setup(struct net_device *dev); 80 81static int ip6_tnl_net_id; 82struct ip6_tnl_net { 83 /* the IPv6 tunnel fallback device */ 84 struct net_device *fb_tnl_dev; 85 /* lists for storing tunnels in use */ 86 struct ip6_tnl *tnls_r_l[HASH_SIZE]; 87 struct ip6_tnl *tnls_wc[1]; 88 struct ip6_tnl **tnls[2]; 89}; 90 91/* lock for the tunnel lists */ 92static DEFINE_RWLOCK(ip6_tnl_lock); 93 94static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t) 95{ 96 struct dst_entry *dst = t->dst_cache; 97 98 if (dst && dst->obsolete && 99 dst->ops->check(dst, t->dst_cookie) == NULL) { 100 t->dst_cache = NULL; 101 dst_release(dst); 102 return NULL; 103 } 104 105 return dst; 106} 107 108static inline void ip6_tnl_dst_reset(struct ip6_tnl *t) 109{ 110 dst_release(t->dst_cache); 111 t->dst_cache = NULL; 112} 113 114static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst) 115{ 116 struct rt6_info *rt = (struct rt6_info *) dst; 117 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0; 118 dst_release(t->dst_cache); 119 t->dst_cache = dst; 120} 121 122/** 123 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses 124 * @remote: the address of the tunnel exit-point 125 * @local: the address of the tunnel entry-point 126 * 127 * Return: 128 * tunnel matching given end-points if found, 129 * else fallback tunnel if its device is up, 130 * else %NULL 131 **/ 132 133static struct ip6_tnl * 134ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local) 135{ 136 unsigned h0 = HASH(remote); 137 unsigned h1 = HASH(local); 138 struct ip6_tnl *t; 139 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 140 141 for (t = ip6n->tnls_r_l[h0 ^ h1]; t; t = t->next) { 142 if (ipv6_addr_equal(local, &t->parms.laddr) && 143 ipv6_addr_equal(remote, &t->parms.raddr) && 144 (t->dev->flags & IFF_UP)) 145 return t; 146 } 147 if ((t = ip6n->tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP)) 148 return t; 149 150 return NULL; 151} 152 153/** 154 * ip6_tnl_bucket - get head of list matching given tunnel parameters 155 * @p: parameters containing tunnel end-points 156 * 157 * Description: 158 * ip6_tnl_bucket() returns the head of the list matching the 159 * &struct in6_addr entries laddr and raddr in @p. 160 * 161 * Return: head of IPv6 tunnel list 162 **/ 163 164static struct ip6_tnl ** 165ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p) 166{ 167 struct in6_addr *remote = &p->raddr; 168 struct in6_addr *local = &p->laddr; 169 unsigned h = 0; 170 int prio = 0; 171 172 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 173 prio = 1; 174 h = HASH(remote) ^ HASH(local); 175 } 176 return &ip6n->tnls[prio][h]; 177} 178 179/** 180 * ip6_tnl_link - add tunnel to hash table 181 * @t: tunnel to be added 182 **/ 183 184static void 185ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 186{ 187 struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms); 188 189 t->next = *tp; 190 write_lock_bh(&ip6_tnl_lock); 191 *tp = t; 192 write_unlock_bh(&ip6_tnl_lock); 193} 194 195/** 196 * ip6_tnl_unlink - remove tunnel from hash table 197 * @t: tunnel to be removed 198 **/ 199 200static void 201ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t) 202{ 203 struct ip6_tnl **tp; 204 205 for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) { 206 if (t == *tp) { 207 write_lock_bh(&ip6_tnl_lock); 208 *tp = t->next; 209 write_unlock_bh(&ip6_tnl_lock); 210 break; 211 } 212 } 213} 214 215/** 216 * ip6_tnl_create() - create a new tunnel 217 * @p: tunnel parameters 218 * @pt: pointer to new tunnel 219 * 220 * Description: 221 * Create tunnel matching given parameters. 222 * 223 * Return: 224 * created tunnel or NULL 225 **/ 226 227static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p) 228{ 229 struct net_device *dev; 230 struct ip6_tnl *t; 231 char name[IFNAMSIZ]; 232 int err; 233 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 234 235 if (p->name[0]) 236 strlcpy(name, p->name, IFNAMSIZ); 237 else 238 sprintf(name, "ip6tnl%%d"); 239 240 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup); 241 if (dev == NULL) 242 goto failed; 243 244 dev_net_set(dev, net); 245 246 if (strchr(name, '%')) { 247 if (dev_alloc_name(dev, name) < 0) 248 goto failed_free; 249 } 250 251 t = netdev_priv(dev); 252 dev->init = ip6_tnl_dev_init; 253 t->parms = *p; 254 255 if ((err = register_netdevice(dev)) < 0) 256 goto failed_free; 257 258 dev_hold(dev); 259 ip6_tnl_link(ip6n, t); 260 return t; 261 262failed_free: 263 free_netdev(dev); 264failed: 265 return NULL; 266} 267 268/** 269 * ip6_tnl_locate - find or create tunnel matching given parameters 270 * @p: tunnel parameters 271 * @create: != 0 if allowed to create new tunnel if no match found 272 * 273 * Description: 274 * ip6_tnl_locate() first tries to locate an existing tunnel 275 * based on @parms. If this is unsuccessful, but @create is set a new 276 * tunnel device is created and registered for use. 277 * 278 * Return: 279 * matching tunnel or NULL 280 **/ 281 282static struct ip6_tnl *ip6_tnl_locate(struct net *net, 283 struct ip6_tnl_parm *p, int create) 284{ 285 struct in6_addr *remote = &p->raddr; 286 struct in6_addr *local = &p->laddr; 287 struct ip6_tnl *t; 288 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 289 290 for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) { 291 if (ipv6_addr_equal(local, &t->parms.laddr) && 292 ipv6_addr_equal(remote, &t->parms.raddr)) 293 return t; 294 } 295 if (!create) 296 return NULL; 297 return ip6_tnl_create(net, p); 298} 299 300/** 301 * ip6_tnl_dev_uninit - tunnel device uninitializer 302 * @dev: the device to be destroyed 303 * 304 * Description: 305 * ip6_tnl_dev_uninit() removes tunnel from its list 306 **/ 307 308static void 309ip6_tnl_dev_uninit(struct net_device *dev) 310{ 311 struct ip6_tnl *t = netdev_priv(dev); 312 struct net *net = dev_net(dev); 313 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 314 315 if (dev == ip6n->fb_tnl_dev) { 316 write_lock_bh(&ip6_tnl_lock); 317 ip6n->tnls_wc[0] = NULL; 318 write_unlock_bh(&ip6_tnl_lock); 319 } else { 320 ip6_tnl_unlink(ip6n, t); 321 } 322 ip6_tnl_dst_reset(t); 323 dev_put(dev); 324} 325 326/** 327 * parse_tvl_tnl_enc_lim - handle encapsulation limit option 328 * @skb: received socket buffer 329 * 330 * Return: 331 * 0 if none was found, 332 * else index to encapsulation limit 333 **/ 334 335static __u16 336parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw) 337{ 338 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw; 339 __u8 nexthdr = ipv6h->nexthdr; 340 __u16 off = sizeof (*ipv6h); 341 342 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) { 343 __u16 optlen = 0; 344 struct ipv6_opt_hdr *hdr; 345 if (raw + off + sizeof (*hdr) > skb->data && 346 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr))) 347 break; 348 349 hdr = (struct ipv6_opt_hdr *) (raw + off); 350 if (nexthdr == NEXTHDR_FRAGMENT) { 351 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr; 352 if (frag_hdr->frag_off) 353 break; 354 optlen = 8; 355 } else if (nexthdr == NEXTHDR_AUTH) { 356 optlen = (hdr->hdrlen + 2) << 2; 357 } else { 358 optlen = ipv6_optlen(hdr); 359 } 360 if (nexthdr == NEXTHDR_DEST) { 361 __u16 i = off + 2; 362 while (1) { 363 struct ipv6_tlv_tnl_enc_lim *tel; 364 365 /* No more room for encapsulation limit */ 366 if (i + sizeof (*tel) > off + optlen) 367 break; 368 369 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i]; 370 /* return index of option if found and valid */ 371 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT && 372 tel->length == 1) 373 return i; 374 /* else jump to next option */ 375 if (tel->type) 376 i += tel->length + 2; 377 else 378 i++; 379 } 380 } 381 nexthdr = hdr->nexthdr; 382 off += optlen; 383 } 384 return 0; 385} 386 387/** 388 * ip6_tnl_err - tunnel error handler 389 * 390 * Description: 391 * ip6_tnl_err() should handle errors in the tunnel according 392 * to the specifications in RFC 2473. 393 **/ 394 395static int 396ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt, 397 int *type, int *code, int *msg, __u32 *info, int offset) 398{ 399 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data; 400 struct ip6_tnl *t; 401 int rel_msg = 0; 402 int rel_type = ICMPV6_DEST_UNREACH; 403 int rel_code = ICMPV6_ADDR_UNREACH; 404 __u32 rel_info = 0; 405 __u16 len; 406 int err = -ENOENT; 407 408 /* If the packet doesn't contain the original IPv6 header we are 409 in trouble since we might need the source address for further 410 processing of the error. */ 411 412 read_lock(&ip6_tnl_lock); 413 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, 414 &ipv6h->saddr)) == NULL) 415 goto out; 416 417 if (t->parms.proto != ipproto && t->parms.proto != 0) 418 goto out; 419 420 err = 0; 421 422 switch (*type) { 423 __u32 teli; 424 struct ipv6_tlv_tnl_enc_lim *tel; 425 __u32 mtu; 426 case ICMPV6_DEST_UNREACH: 427 if (net_ratelimit()) 428 printk(KERN_WARNING 429 "%s: Path to destination invalid " 430 "or inactive!\n", t->parms.name); 431 rel_msg = 1; 432 break; 433 case ICMPV6_TIME_EXCEED: 434 if ((*code) == ICMPV6_EXC_HOPLIMIT) { 435 if (net_ratelimit()) 436 printk(KERN_WARNING 437 "%s: Too small hop limit or " 438 "routing loop in tunnel!\n", 439 t->parms.name); 440 rel_msg = 1; 441 } 442 break; 443 case ICMPV6_PARAMPROB: 444 teli = 0; 445 if ((*code) == ICMPV6_HDR_FIELD) 446 teli = parse_tlv_tnl_enc_lim(skb, skb->data); 447 448 if (teli && teli == *info - 2) { 449 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 450 if (tel->encap_limit == 0) { 451 if (net_ratelimit()) 452 printk(KERN_WARNING 453 "%s: Too small encapsulation " 454 "limit or routing loop in " 455 "tunnel!\n", t->parms.name); 456 rel_msg = 1; 457 } 458 } else if (net_ratelimit()) { 459 printk(KERN_WARNING 460 "%s: Recipient unable to parse tunneled " 461 "packet!\n ", t->parms.name); 462 } 463 break; 464 case ICMPV6_PKT_TOOBIG: 465 mtu = *info - offset; 466 if (mtu < IPV6_MIN_MTU) 467 mtu = IPV6_MIN_MTU; 468 t->dev->mtu = mtu; 469 470 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) { 471 rel_type = ICMPV6_PKT_TOOBIG; 472 rel_code = 0; 473 rel_info = mtu; 474 rel_msg = 1; 475 } 476 break; 477 } 478 479 *type = rel_type; 480 *code = rel_code; 481 *info = rel_info; 482 *msg = rel_msg; 483 484out: 485 read_unlock(&ip6_tnl_lock); 486 return err; 487} 488 489static int 490ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 491 int type, int code, int offset, __be32 info) 492{ 493 int rel_msg = 0; 494 int rel_type = type; 495 int rel_code = code; 496 __u32 rel_info = ntohl(info); 497 int err; 498 struct sk_buff *skb2; 499 struct iphdr *eiph; 500 struct flowi fl; 501 struct rtable *rt; 502 503 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code, 504 &rel_msg, &rel_info, offset); 505 if (err < 0) 506 return err; 507 508 if (rel_msg == 0) 509 return 0; 510 511 switch (rel_type) { 512 case ICMPV6_DEST_UNREACH: 513 if (rel_code != ICMPV6_ADDR_UNREACH) 514 return 0; 515 rel_type = ICMP_DEST_UNREACH; 516 rel_code = ICMP_HOST_UNREACH; 517 break; 518 case ICMPV6_PKT_TOOBIG: 519 if (rel_code != 0) 520 return 0; 521 rel_type = ICMP_DEST_UNREACH; 522 rel_code = ICMP_FRAG_NEEDED; 523 break; 524 default: 525 return 0; 526 } 527 528 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr))) 529 return 0; 530 531 skb2 = skb_clone(skb, GFP_ATOMIC); 532 if (!skb2) 533 return 0; 534 535 dst_release(skb2->dst); 536 skb2->dst = NULL; 537 skb_pull(skb2, offset); 538 skb_reset_network_header(skb2); 539 eiph = ip_hdr(skb2); 540 541 /* Try to guess incoming interface */ 542 memset(&fl, 0, sizeof(fl)); 543 fl.fl4_dst = eiph->saddr; 544 fl.fl4_tos = RT_TOS(eiph->tos); 545 fl.proto = IPPROTO_IPIP; 546 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl)) 547 goto out; 548 549 skb2->dev = rt->u.dst.dev; 550 551 /* route "incoming" packet */ 552 if (rt->rt_flags & RTCF_LOCAL) { 553 ip_rt_put(rt); 554 rt = NULL; 555 fl.fl4_dst = eiph->daddr; 556 fl.fl4_src = eiph->saddr; 557 fl.fl4_tos = eiph->tos; 558 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) || 559 rt->u.dst.dev->type != ARPHRD_TUNNEL) { 560 ip_rt_put(rt); 561 goto out; 562 } 563 skb2->dst = (struct dst_entry *)rt; 564 } else { 565 ip_rt_put(rt); 566 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos, 567 skb2->dev) || 568 skb2->dst->dev->type != ARPHRD_TUNNEL) 569 goto out; 570 } 571 572 /* change mtu on this route */ 573 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) { 574 if (rel_info > dst_mtu(skb2->dst)) 575 goto out; 576 577 skb2->dst->ops->update_pmtu(skb2->dst, rel_info); 578 } 579 580 icmp_send(skb2, rel_type, rel_code, htonl(rel_info)); 581 582out: 583 kfree_skb(skb2); 584 return 0; 585} 586 587static int 588ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 589 int type, int code, int offset, __be32 info) 590{ 591 int rel_msg = 0; 592 int rel_type = type; 593 int rel_code = code; 594 __u32 rel_info = ntohl(info); 595 int err; 596 597 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code, 598 &rel_msg, &rel_info, offset); 599 if (err < 0) 600 return err; 601 602 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) { 603 struct rt6_info *rt; 604 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 605 606 if (!skb2) 607 return 0; 608 609 dst_release(skb2->dst); 610 skb2->dst = NULL; 611 skb_pull(skb2, offset); 612 skb_reset_network_header(skb2); 613 614 /* Try to guess incoming interface */ 615 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, 616 NULL, 0, 0); 617 618 if (rt && rt->rt6i_dev) 619 skb2->dev = rt->rt6i_dev; 620 621 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev); 622 623 if (rt) 624 dst_release(&rt->u.dst); 625 626 kfree_skb(skb2); 627 } 628 629 return 0; 630} 631 632static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t, 633 struct ipv6hdr *ipv6h, 634 struct sk_buff *skb) 635{ 636 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK; 637 638 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 639 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield); 640 641 if (INET_ECN_is_ce(dsfield)) 642 IP_ECN_set_ce(ip_hdr(skb)); 643} 644 645static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t, 646 struct ipv6hdr *ipv6h, 647 struct sk_buff *skb) 648{ 649 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY) 650 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb)); 651 652 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h))) 653 IP6_ECN_set_ce(ipv6_hdr(skb)); 654} 655 656static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t) 657{ 658 struct ip6_tnl_parm *p = &t->parms; 659 int ret = 0; 660 struct net *net = dev_net(t->dev); 661 662 if (p->flags & IP6_TNL_F_CAP_RCV) { 663 struct net_device *ldev = NULL; 664 665 if (p->link) 666 ldev = dev_get_by_index(net, p->link); 667 668 if ((ipv6_addr_is_multicast(&p->laddr) || 669 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) && 670 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0))) 671 ret = 1; 672 673 if (ldev) 674 dev_put(ldev); 675 } 676 return ret; 677} 678 679/** 680 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally 681 * @skb: received socket buffer 682 * @protocol: ethernet protocol ID 683 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN 684 * 685 * Return: 0 686 **/ 687 688static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol, 689 __u8 ipproto, 690 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t, 691 struct ipv6hdr *ipv6h, 692 struct sk_buff *skb)) 693{ 694 struct ip6_tnl *t; 695 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 696 697 read_lock(&ip6_tnl_lock); 698 699 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, 700 &ipv6h->daddr)) != NULL) { 701 if (t->parms.proto != ipproto && t->parms.proto != 0) { 702 read_unlock(&ip6_tnl_lock); 703 goto discard; 704 } 705 706 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 707 read_unlock(&ip6_tnl_lock); 708 goto discard; 709 } 710 711 if (!ip6_tnl_rcv_ctl(t)) { 712 t->dev->stats.rx_dropped++; 713 read_unlock(&ip6_tnl_lock); 714 goto discard; 715 } 716 secpath_reset(skb); 717 skb->mac_header = skb->network_header; 718 skb_reset_network_header(skb); 719 skb->protocol = htons(protocol); 720 skb->pkt_type = PACKET_HOST; 721 memset(skb->cb, 0, sizeof(struct inet6_skb_parm)); 722 skb->dev = t->dev; 723 dst_release(skb->dst); 724 skb->dst = NULL; 725 nf_reset(skb); 726 727 dscp_ecn_decapsulate(t, ipv6h, skb); 728 729 t->dev->stats.rx_packets++; 730 t->dev->stats.rx_bytes += skb->len; 731 netif_rx(skb); 732 read_unlock(&ip6_tnl_lock); 733 return 0; 734 } 735 read_unlock(&ip6_tnl_lock); 736 return 1; 737 738discard: 739 kfree_skb(skb); 740 return 0; 741} 742 743static int ip4ip6_rcv(struct sk_buff *skb) 744{ 745 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP, 746 ip4ip6_dscp_ecn_decapsulate); 747} 748 749static int ip6ip6_rcv(struct sk_buff *skb) 750{ 751 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6, 752 ip6ip6_dscp_ecn_decapsulate); 753} 754 755struct ipv6_tel_txoption { 756 struct ipv6_txoptions ops; 757 __u8 dst_opt[8]; 758}; 759 760static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) 761{ 762 memset(opt, 0, sizeof(struct ipv6_tel_txoption)); 763 764 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; 765 opt->dst_opt[3] = 1; 766 opt->dst_opt[4] = encap_limit; 767 opt->dst_opt[5] = IPV6_TLV_PADN; 768 opt->dst_opt[6] = 1; 769 770 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; 771 opt->ops.opt_nflen = 8; 772} 773 774/** 775 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 776 * @t: the outgoing tunnel device 777 * @hdr: IPv6 header from the incoming packet 778 * 779 * Description: 780 * Avoid trivial tunneling loop by checking that tunnel exit-point 781 * doesn't match source of incoming packet. 782 * 783 * Return: 784 * 1 if conflict, 785 * 0 else 786 **/ 787 788static inline int 789ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr) 790{ 791 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 792} 793 794static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t) 795{ 796 struct ip6_tnl_parm *p = &t->parms; 797 int ret = 0; 798 struct net *net = dev_net(t->dev); 799 800 if (p->flags & IP6_TNL_F_CAP_XMIT) { 801 struct net_device *ldev = NULL; 802 803 if (p->link) 804 ldev = dev_get_by_index(net, p->link); 805 806 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0))) 807 printk(KERN_WARNING 808 "%s xmit: Local address not yet configured!\n", 809 p->name); 810 else if (!ipv6_addr_is_multicast(&p->raddr) && 811 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0))) 812 printk(KERN_WARNING 813 "%s xmit: Routing loop! " 814 "Remote address found on this node!\n", 815 p->name); 816 else 817 ret = 1; 818 if (ldev) 819 dev_put(ldev); 820 } 821 return ret; 822} 823/** 824 * ip6_tnl_xmit2 - encapsulate packet and send 825 * @skb: the outgoing socket buffer 826 * @dev: the outgoing tunnel device 827 * @dsfield: dscp code for outer header 828 * @fl: flow of tunneled packet 829 * @encap_limit: encapsulation limit 830 * @pmtu: Path MTU is stored if packet is too big 831 * 832 * Description: 833 * Build new header and do some sanity checks on the packet before sending 834 * it. 835 * 836 * Return: 837 * 0 on success 838 * -1 fail 839 * %-EMSGSIZE message too big. return mtu in this case. 840 **/ 841 842static int ip6_tnl_xmit2(struct sk_buff *skb, 843 struct net_device *dev, 844 __u8 dsfield, 845 struct flowi *fl, 846 int encap_limit, 847 __u32 *pmtu) 848{ 849 struct ip6_tnl *t = netdev_priv(dev); 850 struct net_device_stats *stats = &t->dev->stats; 851 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 852 struct ipv6_tel_txoption opt; 853 struct dst_entry *dst; 854 struct net_device *tdev; 855 int mtu; 856 unsigned int max_headroom = sizeof(struct ipv6hdr); 857 u8 proto; 858 int err = -1; 859 int pkt_len; 860 861 if ((dst = ip6_tnl_dst_check(t)) != NULL) 862 dst_hold(dst); 863 else { 864 dst = ip6_route_output(dev_net(dev), NULL, fl); 865 866 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0) 867 goto tx_err_link_failure; 868 } 869 870 tdev = dst->dev; 871 872 if (tdev == dev) { 873 stats->collisions++; 874 if (net_ratelimit()) 875 printk(KERN_WARNING 876 "%s: Local routing loop detected!\n", 877 t->parms.name); 878 goto tx_err_dst_release; 879 } 880 mtu = dst_mtu(dst) - sizeof (*ipv6h); 881 if (encap_limit >= 0) { 882 max_headroom += 8; 883 mtu -= 8; 884 } 885 if (mtu < IPV6_MIN_MTU) 886 mtu = IPV6_MIN_MTU; 887 if (skb->dst) 888 skb->dst->ops->update_pmtu(skb->dst, mtu); 889 if (skb->len > mtu) { 890 *pmtu = mtu; 891 err = -EMSGSIZE; 892 goto tx_err_dst_release; 893 } 894 895 /* 896 * Okay, now see if we can stuff it in the buffer as-is. 897 */ 898 max_headroom += LL_RESERVED_SPACE(tdev); 899 900 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 901 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 902 struct sk_buff *new_skb; 903 904 if (!(new_skb = skb_realloc_headroom(skb, max_headroom))) 905 goto tx_err_dst_release; 906 907 if (skb->sk) 908 skb_set_owner_w(new_skb, skb->sk); 909 kfree_skb(skb); 910 skb = new_skb; 911 } 912 dst_release(skb->dst); 913 skb->dst = dst_clone(dst); 914 915 skb->transport_header = skb->network_header; 916 917 proto = fl->proto; 918 if (encap_limit >= 0) { 919 init_tel_txopt(&opt, encap_limit); 920 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); 921 } 922 skb_push(skb, sizeof(struct ipv6hdr)); 923 skb_reset_network_header(skb); 924 ipv6h = ipv6_hdr(skb); 925 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000); 926 dsfield = INET_ECN_encapsulate(0, dsfield); 927 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield); 928 ipv6h->hop_limit = t->parms.hop_limit; 929 ipv6h->nexthdr = proto; 930 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src); 931 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst); 932 nf_reset(skb); 933 pkt_len = skb->len; 934 err = ip6_local_out(skb); 935 936 if (net_xmit_eval(err) == 0) { 937 stats->tx_bytes += pkt_len; 938 stats->tx_packets++; 939 } else { 940 stats->tx_errors++; 941 stats->tx_aborted_errors++; 942 } 943 ip6_tnl_dst_store(t, dst); 944 return 0; 945tx_err_link_failure: 946 stats->tx_carrier_errors++; 947 dst_link_failure(skb); 948tx_err_dst_release: 949 dst_release(dst); 950 return err; 951} 952 953static inline int 954ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 955{ 956 struct ip6_tnl *t = netdev_priv(dev); 957 struct iphdr *iph = ip_hdr(skb); 958 int encap_limit = -1; 959 struct flowi fl; 960 __u8 dsfield; 961 __u32 mtu; 962 int err; 963 964 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) || 965 !ip6_tnl_xmit_ctl(t)) 966 return -1; 967 968 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 969 encap_limit = t->parms.encap_limit; 970 971 memcpy(&fl, &t->fl, sizeof (fl)); 972 fl.proto = IPPROTO_IPIP; 973 974 dsfield = ipv4_get_dsfield(iph); 975 976 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 977 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 978 & IPV6_TCLASS_MASK; 979 980 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu); 981 if (err != 0) { 982 /* XXX: send ICMP error even if DF is not set. */ 983 if (err == -EMSGSIZE) 984 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 985 htonl(mtu)); 986 return -1; 987 } 988 989 return 0; 990} 991 992static inline int 993ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 994{ 995 struct ip6_tnl *t = netdev_priv(dev); 996 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 997 int encap_limit = -1; 998 __u16 offset; 999 struct flowi fl; 1000 __u8 dsfield; 1001 __u32 mtu; 1002 int err; 1003 1004 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 1005 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h)) 1006 return -1; 1007 1008 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb)); 1009 if (offset > 0) { 1010 struct ipv6_tlv_tnl_enc_lim *tel; 1011 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 1012 if (tel->encap_limit == 0) { 1013 icmpv6_send(skb, ICMPV6_PARAMPROB, 1014 ICMPV6_HDR_FIELD, offset + 2, skb->dev); 1015 return -1; 1016 } 1017 encap_limit = tel->encap_limit - 1; 1018 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1019 encap_limit = t->parms.encap_limit; 1020 1021 memcpy(&fl, &t->fl, sizeof (fl)); 1022 fl.proto = IPPROTO_IPV6; 1023 1024 dsfield = ipv6_get_dsfield(ipv6h); 1025 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)) 1026 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 1027 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1028 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); 1029 1030 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu); 1031 if (err != 0) { 1032 if (err == -EMSGSIZE) 1033 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev); 1034 return -1; 1035 } 1036 1037 return 0; 1038} 1039 1040static int 1041ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 1042{ 1043 struct ip6_tnl *t = netdev_priv(dev); 1044 struct net_device_stats *stats = &t->dev->stats; 1045 int ret; 1046 1047 if (t->recursion++) { 1048 stats->collisions++; 1049 goto tx_err; 1050 } 1051 1052 switch (skb->protocol) { 1053 case __constant_htons(ETH_P_IP): 1054 ret = ip4ip6_tnl_xmit(skb, dev); 1055 break; 1056 case __constant_htons(ETH_P_IPV6): 1057 ret = ip6ip6_tnl_xmit(skb, dev); 1058 break; 1059 default: 1060 goto tx_err; 1061 } 1062 1063 if (ret < 0) 1064 goto tx_err; 1065 1066 t->recursion--; 1067 return 0; 1068 1069tx_err: 1070 stats->tx_errors++; 1071 stats->tx_dropped++; 1072 kfree_skb(skb); 1073 t->recursion--; 1074 return 0; 1075} 1076 1077static void ip6_tnl_set_cap(struct ip6_tnl *t) 1078{ 1079 struct ip6_tnl_parm *p = &t->parms; 1080 int ltype = ipv6_addr_type(&p->laddr); 1081 int rtype = ipv6_addr_type(&p->raddr); 1082 1083 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV); 1084 1085 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1086 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) && 1087 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) && 1088 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) { 1089 if (ltype&IPV6_ADDR_UNICAST) 1090 p->flags |= IP6_TNL_F_CAP_XMIT; 1091 if (rtype&IPV6_ADDR_UNICAST) 1092 p->flags |= IP6_TNL_F_CAP_RCV; 1093 } 1094} 1095 1096static void ip6_tnl_link_config(struct ip6_tnl *t) 1097{ 1098 struct net_device *dev = t->dev; 1099 struct ip6_tnl_parm *p = &t->parms; 1100 struct flowi *fl = &t->fl; 1101 1102 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1103 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1104 1105 /* Set up flowi template */ 1106 ipv6_addr_copy(&fl->fl6_src, &p->laddr); 1107 ipv6_addr_copy(&fl->fl6_dst, &p->raddr); 1108 fl->oif = p->link; 1109 fl->fl6_flowlabel = 0; 1110 1111 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1112 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1113 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1114 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1115 1116 ip6_tnl_set_cap(t); 1117 1118 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV) 1119 dev->flags |= IFF_POINTOPOINT; 1120 else 1121 dev->flags &= ~IFF_POINTOPOINT; 1122 1123 dev->iflink = p->link; 1124 1125 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1126 int strict = (ipv6_addr_type(&p->raddr) & 1127 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1128 1129 struct rt6_info *rt = rt6_lookup(dev_net(dev), 1130 &p->raddr, &p->laddr, 1131 p->link, strict); 1132 1133 if (rt == NULL) 1134 return; 1135 1136 if (rt->rt6i_dev) { 1137 dev->hard_header_len = rt->rt6i_dev->hard_header_len + 1138 sizeof (struct ipv6hdr); 1139 1140 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr); 1141 1142 if (dev->mtu < IPV6_MIN_MTU) 1143 dev->mtu = IPV6_MIN_MTU; 1144 } 1145 dst_release(&rt->u.dst); 1146 } 1147} 1148 1149/** 1150 * ip6_tnl_change - update the tunnel parameters 1151 * @t: tunnel to be changed 1152 * @p: tunnel configuration parameters 1153 * @active: != 0 if tunnel is ready for use 1154 * 1155 * Description: 1156 * ip6_tnl_change() updates the tunnel parameters 1157 **/ 1158 1159static int 1160ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p) 1161{ 1162 ipv6_addr_copy(&t->parms.laddr, &p->laddr); 1163 ipv6_addr_copy(&t->parms.raddr, &p->raddr); 1164 t->parms.flags = p->flags; 1165 t->parms.hop_limit = p->hop_limit; 1166 t->parms.encap_limit = p->encap_limit; 1167 t->parms.flowinfo = p->flowinfo; 1168 t->parms.link = p->link; 1169 t->parms.proto = p->proto; 1170 ip6_tnl_dst_reset(t); 1171 ip6_tnl_link_config(t); 1172 return 0; 1173} 1174 1175/** 1176 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace 1177 * @dev: virtual device associated with tunnel 1178 * @ifr: parameters passed from userspace 1179 * @cmd: command to be performed 1180 * 1181 * Description: 1182 * ip6_tnl_ioctl() is used for managing IPv6 tunnels 1183 * from userspace. 1184 * 1185 * The possible commands are the following: 1186 * %SIOCGETTUNNEL: get tunnel parameters for device 1187 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 1188 * %SIOCCHGTUNNEL: change tunnel parameters to those given 1189 * %SIOCDELTUNNEL: delete tunnel 1190 * 1191 * The fallback device "ip6tnl0", created during module 1192 * initialization, can be used for creating other tunnel devices. 1193 * 1194 * Return: 1195 * 0 on success, 1196 * %-EFAULT if unable to copy data to or from userspace, 1197 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 1198 * %-EINVAL if passed tunnel parameters are invalid, 1199 * %-EEXIST if changing a tunnel's parameters would cause a conflict 1200 * %-ENODEV if attempting to change or delete a nonexisting device 1201 **/ 1202 1203static int 1204ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1205{ 1206 int err = 0; 1207 struct ip6_tnl_parm p; 1208 struct ip6_tnl *t = NULL; 1209 struct net *net = dev_net(dev); 1210 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1211 1212 switch (cmd) { 1213 case SIOCGETTUNNEL: 1214 if (dev == ip6n->fb_tnl_dev) { 1215 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) { 1216 err = -EFAULT; 1217 break; 1218 } 1219 t = ip6_tnl_locate(net, &p, 0); 1220 } 1221 if (t == NULL) 1222 t = netdev_priv(dev); 1223 memcpy(&p, &t->parms, sizeof (p)); 1224 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) { 1225 err = -EFAULT; 1226 } 1227 break; 1228 case SIOCADDTUNNEL: 1229 case SIOCCHGTUNNEL: 1230 err = -EPERM; 1231 if (!capable(CAP_NET_ADMIN)) 1232 break; 1233 err = -EFAULT; 1234 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1235 break; 1236 err = -EINVAL; 1237 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP && 1238 p.proto != 0) 1239 break; 1240 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL); 1241 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 1242 if (t != NULL) { 1243 if (t->dev != dev) { 1244 err = -EEXIST; 1245 break; 1246 } 1247 } else 1248 t = netdev_priv(dev); 1249 1250 ip6_tnl_unlink(ip6n, t); 1251 err = ip6_tnl_change(t, &p); 1252 ip6_tnl_link(ip6n, t); 1253 netdev_state_change(dev); 1254 } 1255 if (t) { 1256 err = 0; 1257 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p))) 1258 err = -EFAULT; 1259 1260 } else 1261 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1262 break; 1263 case SIOCDELTUNNEL: 1264 err = -EPERM; 1265 if (!capable(CAP_NET_ADMIN)) 1266 break; 1267 1268 if (dev == ip6n->fb_tnl_dev) { 1269 err = -EFAULT; 1270 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) 1271 break; 1272 err = -ENOENT; 1273 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL) 1274 break; 1275 err = -EPERM; 1276 if (t->dev == ip6n->fb_tnl_dev) 1277 break; 1278 dev = t->dev; 1279 } 1280 err = 0; 1281 unregister_netdevice(dev); 1282 break; 1283 default: 1284 err = -EINVAL; 1285 } 1286 return err; 1287} 1288 1289/** 1290 * ip6_tnl_change_mtu - change mtu manually for tunnel device 1291 * @dev: virtual device associated with tunnel 1292 * @new_mtu: the new mtu 1293 * 1294 * Return: 1295 * 0 on success, 1296 * %-EINVAL if mtu too small 1297 **/ 1298 1299static int 1300ip6_tnl_change_mtu(struct net_device *dev, int new_mtu) 1301{ 1302 if (new_mtu < IPV6_MIN_MTU) { 1303 return -EINVAL; 1304 } 1305 dev->mtu = new_mtu; 1306 return 0; 1307} 1308 1309/** 1310 * ip6_tnl_dev_setup - setup virtual tunnel device 1311 * @dev: virtual device associated with tunnel 1312 * 1313 * Description: 1314 * Initialize function pointers and device parameters 1315 **/ 1316 1317static void ip6_tnl_dev_setup(struct net_device *dev) 1318{ 1319 dev->uninit = ip6_tnl_dev_uninit; 1320 dev->destructor = free_netdev; 1321 dev->hard_start_xmit = ip6_tnl_xmit; 1322 dev->do_ioctl = ip6_tnl_ioctl; 1323 dev->change_mtu = ip6_tnl_change_mtu; 1324 1325 dev->type = ARPHRD_TUNNEL6; 1326 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr); 1327 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr); 1328 dev->flags |= IFF_NOARP; 1329 dev->addr_len = sizeof(struct in6_addr); 1330 dev->features |= NETIF_F_NETNS_LOCAL; 1331} 1332 1333 1334/** 1335 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices 1336 * @dev: virtual device associated with tunnel 1337 **/ 1338 1339static inline void 1340ip6_tnl_dev_init_gen(struct net_device *dev) 1341{ 1342 struct ip6_tnl *t = netdev_priv(dev); 1343 t->dev = dev; 1344 strcpy(t->parms.name, dev->name); 1345} 1346 1347/** 1348 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices 1349 * @dev: virtual device associated with tunnel 1350 **/ 1351 1352static int 1353ip6_tnl_dev_init(struct net_device *dev) 1354{ 1355 struct ip6_tnl *t = netdev_priv(dev); 1356 ip6_tnl_dev_init_gen(dev); 1357 ip6_tnl_link_config(t); 1358 return 0; 1359} 1360 1361/** 1362 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device 1363 * @dev: fallback device 1364 * 1365 * Return: 0 1366 **/ 1367 1368static int 1369ip6_fb_tnl_dev_init(struct net_device *dev) 1370{ 1371 struct ip6_tnl *t = netdev_priv(dev); 1372 struct net *net = dev_net(dev); 1373 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id); 1374 1375 ip6_tnl_dev_init_gen(dev); 1376 t->parms.proto = IPPROTO_IPV6; 1377 dev_hold(dev); 1378 ip6n->tnls_wc[0] = t; 1379 return 0; 1380} 1381 1382static struct xfrm6_tunnel ip4ip6_handler = { 1383 .handler = ip4ip6_rcv, 1384 .err_handler = ip4ip6_err, 1385 .priority = 1, 1386}; 1387 1388static struct xfrm6_tunnel ip6ip6_handler = { 1389 .handler = ip6ip6_rcv, 1390 .err_handler = ip6ip6_err, 1391 .priority = 1, 1392}; 1393 1394static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n) 1395{ 1396 int h; 1397 struct ip6_tnl *t; 1398 1399 for (h = 0; h < HASH_SIZE; h++) { 1400 while ((t = ip6n->tnls_r_l[h]) != NULL) 1401 unregister_netdevice(t->dev); 1402 } 1403 1404 t = ip6n->tnls_wc[0]; 1405 unregister_netdevice(t->dev); 1406} 1407 1408static int ip6_tnl_init_net(struct net *net) 1409{ 1410 int err; 1411 struct ip6_tnl_net *ip6n; 1412 1413 err = -ENOMEM; 1414 ip6n = kzalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL); 1415 if (ip6n == NULL) 1416 goto err_alloc; 1417 1418 err = net_assign_generic(net, ip6_tnl_net_id, ip6n); 1419 if (err < 0) 1420 goto err_assign; 1421 1422 ip6n->tnls[0] = ip6n->tnls_wc; 1423 ip6n->tnls[1] = ip6n->tnls_r_l; 1424 1425 err = -ENOMEM; 1426 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0", 1427 ip6_tnl_dev_setup); 1428 1429 if (!ip6n->fb_tnl_dev) 1430 goto err_alloc_dev; 1431 1432 ip6n->fb_tnl_dev->init = ip6_fb_tnl_dev_init; 1433 dev_net_set(ip6n->fb_tnl_dev, net); 1434 1435 err = register_netdev(ip6n->fb_tnl_dev); 1436 if (err < 0) 1437 goto err_register; 1438 return 0; 1439 1440err_register: 1441 free_netdev(ip6n->fb_tnl_dev); 1442err_alloc_dev: 1443 /* nothing */ 1444err_assign: 1445 kfree(ip6n); 1446err_alloc: 1447 return err; 1448} 1449 1450static void ip6_tnl_exit_net(struct net *net) 1451{ 1452 struct ip6_tnl_net *ip6n; 1453 1454 ip6n = net_generic(net, ip6_tnl_net_id); 1455 rtnl_lock(); 1456 ip6_tnl_destroy_tunnels(ip6n); 1457 rtnl_unlock(); 1458 kfree(ip6n); 1459} 1460 1461static struct pernet_operations ip6_tnl_net_ops = { 1462 .init = ip6_tnl_init_net, 1463 .exit = ip6_tnl_exit_net, 1464}; 1465 1466/** 1467 * ip6_tunnel_init - register protocol and reserve needed resources 1468 * 1469 * Return: 0 on success 1470 **/ 1471 1472static int __init ip6_tunnel_init(void) 1473{ 1474 int err; 1475 1476 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) { 1477 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n"); 1478 err = -EAGAIN; 1479 goto out; 1480 } 1481 1482 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) { 1483 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n"); 1484 err = -EAGAIN; 1485 goto unreg_ip4ip6; 1486 } 1487 1488 err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops); 1489 if (err < 0) 1490 goto err_pernet; 1491 return 0; 1492err_pernet: 1493 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6); 1494unreg_ip4ip6: 1495 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET); 1496out: 1497 return err; 1498} 1499 1500/** 1501 * ip6_tunnel_cleanup - free resources and unregister protocol 1502 **/ 1503 1504static void __exit ip6_tunnel_cleanup(void) 1505{ 1506 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET)) 1507 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n"); 1508 1509 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6)) 1510 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n"); 1511 1512 unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops); 1513} 1514 1515module_init(ip6_tunnel_init); 1516module_exit(ip6_tunnel_cleanup);