1 /* 2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT) 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 8 * 9 * This program is free software; you can redistribute it and/or 10 * modify it under the terms of the GNU General Public License 11 * as published by the Free Software Foundation; either version 12 * 2 of the License, or (at your option) any later version. 13 * 14 * Changes: 15 * Roger Venning <r.venning@telstra.com>: 6to4 support 16 * Nate Thompson <nate@thebog.net>: 6to4 support 17 * Fred Templin <fred.l.templin@boeing.com>: isatap support 18 */ 19 20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 21 22 #include <linux/module.h> 23 #include <linux/capability.h> 24 #include <linux/errno.h> 25 #include <linux/types.h> 26 #include <linux/socket.h> 27 #include <linux/sockios.h> 28 #include <linux/net.h> 29 #include <linux/in6.h> 30 #include <linux/netdevice.h> 31 #include <linux/if_arp.h> 32 #include <linux/icmp.h> 33 #include <linux/slab.h> 34 #include <asm/uaccess.h> 35 #include <linux/init.h> 36 #include <linux/netfilter_ipv4.h> 37 #include <linux/if_ether.h> 38 39 #include <net/sock.h> 40 #include <net/snmp.h> 41 42 #include <net/ipv6.h> 43 #include <net/protocol.h> 44 #include <net/transp_v6.h> 45 #include <net/ip6_fib.h> 46 #include <net/ip6_route.h> 47 #include <net/ndisc.h> 48 #include <net/addrconf.h> 49 #include <net/ip.h> 50 #include <net/udp.h> 51 #include <net/icmp.h> 52 #include <net/ipip.h> 53 #include <net/inet_ecn.h> 54 #include <net/xfrm.h> 55 #include <net/dsfield.h> 56 #include <net/net_namespace.h> 57 #include <net/netns/generic.h> 58 59 /* 60 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c 61 62 For comments look at net/ipv4/ip_gre.c --ANK 63 */ 64 65 #define HASH_SIZE 16 66 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF) 67 68 static int ipip6_tunnel_init(struct net_device *dev); 69 static void ipip6_tunnel_setup(struct net_device *dev); 70 static void ipip6_dev_free(struct net_device *dev); 71 72 static int sit_net_id __read_mostly; 73 struct sit_net { 74 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE]; 75 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE]; 76 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE]; 77 struct ip_tunnel __rcu *tunnels_wc[1]; 78 struct ip_tunnel __rcu **tunnels[4]; 79 80 struct net_device *fb_tunnel_dev; 81 }; 82 83 /* 84 * Locking : hash tables are protected by RCU and RTNL 85 */ 86 87 #define for_each_ip_tunnel_rcu(start) \ 88 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 89 90 /* often modified stats are per cpu, other are shared (netdev->stats) */ 91 struct pcpu_tstats { 92 u64 rx_packets; 93 u64 rx_bytes; 94 u64 tx_packets; 95 u64 tx_bytes; 96 struct u64_stats_sync syncp; 97 }; 98 99 static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev, 100 struct rtnl_link_stats64 *tot) 101 { 102 int i; 103 104 for_each_possible_cpu(i) { 105 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i); 106 u64 rx_packets, rx_bytes, tx_packets, tx_bytes; 107 unsigned int start; 108 109 do { 110 start = u64_stats_fetch_begin_bh(&tstats->syncp); 111 rx_packets = tstats->rx_packets; 112 tx_packets = tstats->tx_packets; 113 rx_bytes = tstats->rx_bytes; 114 tx_bytes = tstats->tx_bytes; 115 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start)); 116 117 tot->rx_packets += rx_packets; 118 tot->tx_packets += tx_packets; 119 tot->rx_bytes += rx_bytes; 120 tot->tx_bytes += tx_bytes; 121 } 122 123 tot->rx_errors = dev->stats.rx_errors; 124 tot->tx_fifo_errors = dev->stats.tx_fifo_errors; 125 tot->tx_carrier_errors = dev->stats.tx_carrier_errors; 126 tot->tx_dropped = dev->stats.tx_dropped; 127 tot->tx_aborted_errors = dev->stats.tx_aborted_errors; 128 tot->tx_errors = dev->stats.tx_errors; 129 130 return tot; 131 } 132 133 /* 134 * Must be invoked with rcu_read_lock 135 */ 136 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net, 137 struct net_device *dev, __be32 remote, __be32 local) 138 { 139 unsigned int h0 = HASH(remote); 140 unsigned int h1 = HASH(local); 141 struct ip_tunnel *t; 142 struct sit_net *sitn = net_generic(net, sit_net_id); 143 144 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) { 145 if (local == t->parms.iph.saddr && 146 remote == t->parms.iph.daddr && 147 (!dev || !t->parms.link || dev->iflink == t->parms.link) && 148 (t->dev->flags & IFF_UP)) 149 return t; 150 } 151 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) { 152 if (remote == t->parms.iph.daddr && 153 (!dev || !t->parms.link || dev->iflink == t->parms.link) && 154 (t->dev->flags & IFF_UP)) 155 return t; 156 } 157 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) { 158 if (local == t->parms.iph.saddr && 159 (!dev || !t->parms.link || dev->iflink == t->parms.link) && 160 (t->dev->flags & IFF_UP)) 161 return t; 162 } 163 t = rcu_dereference(sitn->tunnels_wc[0]); 164 if ((t != NULL) && (t->dev->flags & IFF_UP)) 165 return t; 166 return NULL; 167 } 168 169 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn, 170 struct ip_tunnel_parm *parms) 171 { 172 __be32 remote = parms->iph.daddr; 173 __be32 local = parms->iph.saddr; 174 unsigned int h = 0; 175 int prio = 0; 176 177 if (remote) { 178 prio |= 2; 179 h ^= HASH(remote); 180 } 181 if (local) { 182 prio |= 1; 183 h ^= HASH(local); 184 } 185 return &sitn->tunnels[prio][h]; 186 } 187 188 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn, 189 struct ip_tunnel *t) 190 { 191 return __ipip6_bucket(sitn, &t->parms); 192 } 193 194 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t) 195 { 196 struct ip_tunnel __rcu **tp; 197 struct ip_tunnel *iter; 198 199 for (tp = ipip6_bucket(sitn, t); 200 (iter = rtnl_dereference(*tp)) != NULL; 201 tp = &iter->next) { 202 if (t == iter) { 203 rcu_assign_pointer(*tp, t->next); 204 break; 205 } 206 } 207 } 208 209 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) 210 { 211 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t); 212 213 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 214 rcu_assign_pointer(*tp, t); 215 } 216 217 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn) 218 { 219 #ifdef CONFIG_IPV6_SIT_6RD 220 struct ip_tunnel *t = netdev_priv(dev); 221 222 if (t->dev == sitn->fb_tunnel_dev) { 223 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0); 224 t->ip6rd.relay_prefix = 0; 225 t->ip6rd.prefixlen = 16; 226 t->ip6rd.relay_prefixlen = 0; 227 } else { 228 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev); 229 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd)); 230 } 231 #endif 232 } 233 234 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, 235 struct ip_tunnel_parm *parms, int create) 236 { 237 __be32 remote = parms->iph.daddr; 238 __be32 local = parms->iph.saddr; 239 struct ip_tunnel *t, *nt; 240 struct ip_tunnel __rcu **tp; 241 struct net_device *dev; 242 char name[IFNAMSIZ]; 243 struct sit_net *sitn = net_generic(net, sit_net_id); 244 245 for (tp = __ipip6_bucket(sitn, parms); 246 (t = rtnl_dereference(*tp)) != NULL; 247 tp = &t->next) { 248 if (local == t->parms.iph.saddr && 249 remote == t->parms.iph.daddr && 250 parms->link == t->parms.link) { 251 if (create) 252 return NULL; 253 else 254 return t; 255 } 256 } 257 if (!create) 258 goto failed; 259 260 if (parms->name[0]) 261 strlcpy(name, parms->name, IFNAMSIZ); 262 else 263 strcpy(name, "sit%d"); 264 265 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup); 266 if (dev == NULL) 267 return NULL; 268 269 dev_net_set(dev, net); 270 271 nt = netdev_priv(dev); 272 273 nt->parms = *parms; 274 if (ipip6_tunnel_init(dev) < 0) 275 goto failed_free; 276 ipip6_tunnel_clone_6rd(dev, sitn); 277 278 if (parms->i_flags & SIT_ISATAP) 279 dev->priv_flags |= IFF_ISATAP; 280 281 if (register_netdevice(dev) < 0) 282 goto failed_free; 283 284 strcpy(nt->parms.name, dev->name); 285 286 dev_hold(dev); 287 288 ipip6_tunnel_link(sitn, nt); 289 return nt; 290 291 failed_free: 292 ipip6_dev_free(dev); 293 failed: 294 return NULL; 295 } 296 297 #define for_each_prl_rcu(start) \ 298 for (prl = rcu_dereference(start); \ 299 prl; \ 300 prl = rcu_dereference(prl->next)) 301 302 static struct ip_tunnel_prl_entry * 303 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr) 304 { 305 struct ip_tunnel_prl_entry *prl; 306 307 for_each_prl_rcu(t->prl) 308 if (prl->addr == addr) 309 break; 310 return prl; 311 312 } 313 314 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, 315 struct ip_tunnel_prl __user *a) 316 { 317 struct ip_tunnel_prl kprl, *kp; 318 struct ip_tunnel_prl_entry *prl; 319 unsigned int cmax, c = 0, ca, len; 320 int ret = 0; 321 322 if (copy_from_user(&kprl, a, sizeof(kprl))) 323 return -EFAULT; 324 cmax = kprl.datalen / sizeof(kprl); 325 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY)) 326 cmax = 1; 327 328 /* For simple GET or for root users, 329 * we try harder to allocate. 330 */ 331 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ? 332 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) : 333 NULL; 334 335 rcu_read_lock(); 336 337 ca = t->prl_count < cmax ? t->prl_count : cmax; 338 339 if (!kp) { 340 /* We don't try hard to allocate much memory for 341 * non-root users. 342 * For root users, retry allocating enough memory for 343 * the answer. 344 */ 345 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC); 346 if (!kp) { 347 ret = -ENOMEM; 348 goto out; 349 } 350 } 351 352 c = 0; 353 for_each_prl_rcu(t->prl) { 354 if (c >= cmax) 355 break; 356 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr) 357 continue; 358 kp[c].addr = prl->addr; 359 kp[c].flags = prl->flags; 360 c++; 361 if (kprl.addr != htonl(INADDR_ANY)) 362 break; 363 } 364 out: 365 rcu_read_unlock(); 366 367 len = sizeof(*kp) * c; 368 ret = 0; 369 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen)) 370 ret = -EFAULT; 371 372 kfree(kp); 373 374 return ret; 375 } 376 377 static int 378 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) 379 { 380 struct ip_tunnel_prl_entry *p; 381 int err = 0; 382 383 if (a->addr == htonl(INADDR_ANY)) 384 return -EINVAL; 385 386 ASSERT_RTNL(); 387 388 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) { 389 if (p->addr == a->addr) { 390 if (chg) { 391 p->flags = a->flags; 392 goto out; 393 } 394 err = -EEXIST; 395 goto out; 396 } 397 } 398 399 if (chg) { 400 err = -ENXIO; 401 goto out; 402 } 403 404 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL); 405 if (!p) { 406 err = -ENOBUFS; 407 goto out; 408 } 409 410 p->next = t->prl; 411 p->addr = a->addr; 412 p->flags = a->flags; 413 t->prl_count++; 414 rcu_assign_pointer(t->prl, p); 415 out: 416 return err; 417 } 418 419 static void prl_list_destroy_rcu(struct rcu_head *head) 420 { 421 struct ip_tunnel_prl_entry *p, *n; 422 423 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head); 424 do { 425 n = rcu_dereference_protected(p->next, 1); 426 kfree(p); 427 p = n; 428 } while (p); 429 } 430 431 static int 432 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) 433 { 434 struct ip_tunnel_prl_entry *x; 435 struct ip_tunnel_prl_entry __rcu **p; 436 int err = 0; 437 438 ASSERT_RTNL(); 439 440 if (a && a->addr != htonl(INADDR_ANY)) { 441 for (p = &t->prl; 442 (x = rtnl_dereference(*p)) != NULL; 443 p = &x->next) { 444 if (x->addr == a->addr) { 445 *p = x->next; 446 kfree_rcu(x, rcu_head); 447 t->prl_count--; 448 goto out; 449 } 450 } 451 err = -ENXIO; 452 } else { 453 x = rtnl_dereference(t->prl); 454 if (x) { 455 t->prl_count = 0; 456 call_rcu(&x->rcu_head, prl_list_destroy_rcu); 457 t->prl = NULL; 458 } 459 } 460 out: 461 return err; 462 } 463 464 static int 465 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t) 466 { 467 struct ip_tunnel_prl_entry *p; 468 int ok = 1; 469 470 rcu_read_lock(); 471 p = __ipip6_tunnel_locate_prl(t, iph->saddr); 472 if (p) { 473 if (p->flags & PRL_DEFAULT) 474 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT; 475 else 476 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT; 477 } else { 478 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr; 479 480 if (ipv6_addr_is_isatap(addr6) && 481 (addr6->s6_addr32[3] == iph->saddr) && 482 ipv6_chk_prefix(addr6, t->dev)) 483 skb->ndisc_nodetype = NDISC_NODETYPE_HOST; 484 else 485 ok = 0; 486 } 487 rcu_read_unlock(); 488 return ok; 489 } 490 491 static void ipip6_tunnel_uninit(struct net_device *dev) 492 { 493 struct net *net = dev_net(dev); 494 struct sit_net *sitn = net_generic(net, sit_net_id); 495 496 if (dev == sitn->fb_tunnel_dev) { 497 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL); 498 } else { 499 ipip6_tunnel_unlink(sitn, netdev_priv(dev)); 500 ipip6_tunnel_del_prl(netdev_priv(dev), NULL); 501 } 502 dev_put(dev); 503 } 504 505 506 static int ipip6_err(struct sk_buff *skb, u32 info) 507 { 508 509 /* All the routers (except for Linux) return only 510 8 bytes of packet payload. It means, that precise relaying of 511 ICMP in the real Internet is absolutely infeasible. 512 */ 513 const struct iphdr *iph = (const struct iphdr *)skb->data; 514 const int type = icmp_hdr(skb)->type; 515 const int code = icmp_hdr(skb)->code; 516 struct ip_tunnel *t; 517 int err; 518 519 switch (type) { 520 default: 521 case ICMP_PARAMETERPROB: 522 return 0; 523 524 case ICMP_DEST_UNREACH: 525 switch (code) { 526 case ICMP_SR_FAILED: 527 case ICMP_PORT_UNREACH: 528 /* Impossible event. */ 529 return 0; 530 case ICMP_FRAG_NEEDED: 531 /* Soft state for pmtu is maintained by IP core. */ 532 return 0; 533 default: 534 /* All others are translated to HOST_UNREACH. 535 rfc2003 contains "deep thoughts" about NET_UNREACH, 536 I believe they are just ether pollution. --ANK 537 */ 538 break; 539 } 540 break; 541 case ICMP_TIME_EXCEEDED: 542 if (code != ICMP_EXC_TTL) 543 return 0; 544 break; 545 } 546 547 err = -ENOENT; 548 549 rcu_read_lock(); 550 t = ipip6_tunnel_lookup(dev_net(skb->dev), 551 skb->dev, 552 iph->daddr, 553 iph->saddr); 554 if (t == NULL || t->parms.iph.daddr == 0) 555 goto out; 556 557 err = 0; 558 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) 559 goto out; 560 561 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) 562 t->err_count++; 563 else 564 t->err_count = 1; 565 t->err_time = jiffies; 566 out: 567 rcu_read_unlock(); 568 return err; 569 } 570 571 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb) 572 { 573 if (INET_ECN_is_ce(iph->tos)) 574 IP6_ECN_set_ce(ipv6_hdr(skb)); 575 } 576 577 static int ipip6_rcv(struct sk_buff *skb) 578 { 579 const struct iphdr *iph; 580 struct ip_tunnel *tunnel; 581 582 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 583 goto out; 584 585 iph = ip_hdr(skb); 586 587 rcu_read_lock(); 588 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 589 iph->saddr, iph->daddr); 590 if (tunnel != NULL) { 591 struct pcpu_tstats *tstats; 592 593 secpath_reset(skb); 594 skb->mac_header = skb->network_header; 595 skb_reset_network_header(skb); 596 IPCB(skb)->flags = 0; 597 skb->protocol = htons(ETH_P_IPV6); 598 skb->pkt_type = PACKET_HOST; 599 600 if ((tunnel->dev->priv_flags & IFF_ISATAP) && 601 !isatap_chksrc(skb, iph, tunnel)) { 602 tunnel->dev->stats.rx_errors++; 603 rcu_read_unlock(); 604 kfree_skb(skb); 605 return 0; 606 } 607 608 tstats = this_cpu_ptr(tunnel->dev->tstats); 609 tstats->rx_packets++; 610 tstats->rx_bytes += skb->len; 611 612 __skb_tunnel_rx(skb, tunnel->dev); 613 614 ipip6_ecn_decapsulate(iph, skb); 615 616 netif_rx(skb); 617 618 rcu_read_unlock(); 619 return 0; 620 } 621 622 /* no tunnel matched, let upstream know, ipsec may handle it */ 623 rcu_read_unlock(); 624 return 1; 625 out: 626 kfree_skb(skb); 627 return 0; 628 } 629 630 /* 631 * Returns the embedded IPv4 address if the IPv6 address 632 * comes from 6rd / 6to4 (RFC 3056) addr space. 633 */ 634 static inline 635 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel) 636 { 637 __be32 dst = 0; 638 639 #ifdef CONFIG_IPV6_SIT_6RD 640 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, 641 tunnel->ip6rd.prefixlen)) { 642 unsigned int pbw0, pbi0; 643 int pbi1; 644 u32 d; 645 646 pbw0 = tunnel->ip6rd.prefixlen >> 5; 647 pbi0 = tunnel->ip6rd.prefixlen & 0x1f; 648 649 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> 650 tunnel->ip6rd.relay_prefixlen; 651 652 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; 653 if (pbi1 > 0) 654 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> 655 (32 - pbi1); 656 657 dst = tunnel->ip6rd.relay_prefix | htonl(d); 658 } 659 #else 660 if (v6dst->s6_addr16[0] == htons(0x2002)) { 661 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ 662 memcpy(&dst, &v6dst->s6_addr16[1], 4); 663 } 664 #endif 665 return dst; 666 } 667 668 /* 669 * This function assumes it is being called from dev_queue_xmit() 670 * and that skb is filled properly by that function. 671 */ 672 673 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, 674 struct net_device *dev) 675 { 676 struct ip_tunnel *tunnel = netdev_priv(dev); 677 struct pcpu_tstats *tstats; 678 const struct iphdr *tiph = &tunnel->parms.iph; 679 const struct ipv6hdr *iph6 = ipv6_hdr(skb); 680 u8 tos = tunnel->parms.iph.tos; 681 __be16 df = tiph->frag_off; 682 struct rtable *rt; /* Route to the other host */ 683 struct net_device *tdev; /* Device to other host */ 684 struct iphdr *iph; /* Our new IP header */ 685 unsigned int max_headroom; /* The extra header space needed */ 686 __be32 dst = tiph->daddr; 687 struct flowi4 fl4; 688 int mtu; 689 const struct in6_addr *addr6; 690 int addr_type; 691 692 if (skb->protocol != htons(ETH_P_IPV6)) 693 goto tx_error; 694 695 if (tos == 1) 696 tos = ipv6_get_dsfield(iph6); 697 698 /* ISATAP (RFC4214) - must come before 6to4 */ 699 if (dev->priv_flags & IFF_ISATAP) { 700 struct neighbour *neigh = NULL; 701 bool do_tx_error = false; 702 703 if (skb_dst(skb)) 704 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 705 706 if (neigh == NULL) { 707 net_dbg_ratelimited("sit: nexthop == NULL\n"); 708 goto tx_error; 709 } 710 711 addr6 = (const struct in6_addr *)&neigh->primary_key; 712 addr_type = ipv6_addr_type(addr6); 713 714 if ((addr_type & IPV6_ADDR_UNICAST) && 715 ipv6_addr_is_isatap(addr6)) 716 dst = addr6->s6_addr32[3]; 717 else 718 do_tx_error = true; 719 720 neigh_release(neigh); 721 if (do_tx_error) 722 goto tx_error; 723 } 724 725 if (!dst) 726 dst = try_6rd(&iph6->daddr, tunnel); 727 728 if (!dst) { 729 struct neighbour *neigh = NULL; 730 bool do_tx_error = false; 731 732 if (skb_dst(skb)) 733 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 734 735 if (neigh == NULL) { 736 net_dbg_ratelimited("sit: nexthop == NULL\n"); 737 goto tx_error; 738 } 739 740 addr6 = (const struct in6_addr *)&neigh->primary_key; 741 addr_type = ipv6_addr_type(addr6); 742 743 if (addr_type == IPV6_ADDR_ANY) { 744 addr6 = &ipv6_hdr(skb)->daddr; 745 addr_type = ipv6_addr_type(addr6); 746 } 747 748 if ((addr_type & IPV6_ADDR_COMPATv4) != 0) 749 dst = addr6->s6_addr32[3]; 750 else 751 do_tx_error = true; 752 753 neigh_release(neigh); 754 if (do_tx_error) 755 goto tx_error; 756 } 757 758 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL, 759 dst, tiph->saddr, 760 0, 0, 761 IPPROTO_IPV6, RT_TOS(tos), 762 tunnel->parms.link); 763 if (IS_ERR(rt)) { 764 dev->stats.tx_carrier_errors++; 765 goto tx_error_icmp; 766 } 767 if (rt->rt_type != RTN_UNICAST) { 768 ip_rt_put(rt); 769 dev->stats.tx_carrier_errors++; 770 goto tx_error_icmp; 771 } 772 tdev = rt->dst.dev; 773 774 if (tdev == dev) { 775 ip_rt_put(rt); 776 dev->stats.collisions++; 777 goto tx_error; 778 } 779 780 if (df) { 781 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr); 782 783 if (mtu < 68) { 784 dev->stats.collisions++; 785 ip_rt_put(rt); 786 goto tx_error; 787 } 788 789 if (mtu < IPV6_MIN_MTU) { 790 mtu = IPV6_MIN_MTU; 791 df = 0; 792 } 793 794 if (tunnel->parms.iph.daddr && skb_dst(skb)) 795 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu); 796 797 if (skb->len > mtu) { 798 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 799 ip_rt_put(rt); 800 goto tx_error; 801 } 802 } 803 804 if (tunnel->err_count > 0) { 805 if (time_before(jiffies, 806 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 807 tunnel->err_count--; 808 dst_link_failure(skb); 809 } else 810 tunnel->err_count = 0; 811 } 812 813 /* 814 * Okay, now see if we can stuff it in the buffer as-is. 815 */ 816 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr); 817 818 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 819 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 820 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 821 if (!new_skb) { 822 ip_rt_put(rt); 823 dev->stats.tx_dropped++; 824 dev_kfree_skb(skb); 825 return NETDEV_TX_OK; 826 } 827 if (skb->sk) 828 skb_set_owner_w(new_skb, skb->sk); 829 dev_kfree_skb(skb); 830 skb = new_skb; 831 iph6 = ipv6_hdr(skb); 832 } 833 834 skb->transport_header = skb->network_header; 835 skb_push(skb, sizeof(struct iphdr)); 836 skb_reset_network_header(skb); 837 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 838 IPCB(skb)->flags = 0; 839 skb_dst_drop(skb); 840 skb_dst_set(skb, &rt->dst); 841 842 /* 843 * Push down and install the IPIP header. 844 */ 845 846 iph = ip_hdr(skb); 847 iph->version = 4; 848 iph->ihl = sizeof(struct iphdr)>>2; 849 iph->frag_off = df; 850 iph->protocol = IPPROTO_IPV6; 851 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); 852 iph->daddr = fl4.daddr; 853 iph->saddr = fl4.saddr; 854 855 if ((iph->ttl = tiph->ttl) == 0) 856 iph->ttl = iph6->hop_limit; 857 858 nf_reset(skb); 859 tstats = this_cpu_ptr(dev->tstats); 860 __IPTUNNEL_XMIT(tstats, &dev->stats); 861 return NETDEV_TX_OK; 862 863 tx_error_icmp: 864 dst_link_failure(skb); 865 tx_error: 866 dev->stats.tx_errors++; 867 dev_kfree_skb(skb); 868 return NETDEV_TX_OK; 869 } 870 871 static void ipip6_tunnel_bind_dev(struct net_device *dev) 872 { 873 struct net_device *tdev = NULL; 874 struct ip_tunnel *tunnel; 875 const struct iphdr *iph; 876 struct flowi4 fl4; 877 878 tunnel = netdev_priv(dev); 879 iph = &tunnel->parms.iph; 880 881 if (iph->daddr) { 882 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL, 883 iph->daddr, iph->saddr, 884 0, 0, 885 IPPROTO_IPV6, 886 RT_TOS(iph->tos), 887 tunnel->parms.link); 888 889 if (!IS_ERR(rt)) { 890 tdev = rt->dst.dev; 891 ip_rt_put(rt); 892 } 893 dev->flags |= IFF_POINTOPOINT; 894 } 895 896 if (!tdev && tunnel->parms.link) 897 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link); 898 899 if (tdev) { 900 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); 901 dev->mtu = tdev->mtu - sizeof(struct iphdr); 902 if (dev->mtu < IPV6_MIN_MTU) 903 dev->mtu = IPV6_MIN_MTU; 904 } 905 dev->iflink = tunnel->parms.link; 906 } 907 908 static int 909 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) 910 { 911 int err = 0; 912 struct ip_tunnel_parm p; 913 struct ip_tunnel_prl prl; 914 struct ip_tunnel *t; 915 struct net *net = dev_net(dev); 916 struct sit_net *sitn = net_generic(net, sit_net_id); 917 #ifdef CONFIG_IPV6_SIT_6RD 918 struct ip_tunnel_6rd ip6rd; 919 #endif 920 921 switch (cmd) { 922 case SIOCGETTUNNEL: 923 #ifdef CONFIG_IPV6_SIT_6RD 924 case SIOCGET6RD: 925 #endif 926 t = NULL; 927 if (dev == sitn->fb_tunnel_dev) { 928 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 929 err = -EFAULT; 930 break; 931 } 932 t = ipip6_tunnel_locate(net, &p, 0); 933 } 934 if (t == NULL) 935 t = netdev_priv(dev); 936 937 err = -EFAULT; 938 if (cmd == SIOCGETTUNNEL) { 939 memcpy(&p, &t->parms, sizeof(p)); 940 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, 941 sizeof(p))) 942 goto done; 943 #ifdef CONFIG_IPV6_SIT_6RD 944 } else { 945 ip6rd.prefix = t->ip6rd.prefix; 946 ip6rd.relay_prefix = t->ip6rd.relay_prefix; 947 ip6rd.prefixlen = t->ip6rd.prefixlen; 948 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; 949 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, 950 sizeof(ip6rd))) 951 goto done; 952 #endif 953 } 954 err = 0; 955 break; 956 957 case SIOCADDTUNNEL: 958 case SIOCCHGTUNNEL: 959 err = -EPERM; 960 if (!capable(CAP_NET_ADMIN)) 961 goto done; 962 963 err = -EFAULT; 964 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 965 goto done; 966 967 err = -EINVAL; 968 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 || 969 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) 970 goto done; 971 if (p.iph.ttl) 972 p.iph.frag_off |= htons(IP_DF); 973 974 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); 975 976 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 977 if (t != NULL) { 978 if (t->dev != dev) { 979 err = -EEXIST; 980 break; 981 } 982 } else { 983 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || 984 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { 985 err = -EINVAL; 986 break; 987 } 988 t = netdev_priv(dev); 989 ipip6_tunnel_unlink(sitn, t); 990 synchronize_net(); 991 t->parms.iph.saddr = p.iph.saddr; 992 t->parms.iph.daddr = p.iph.daddr; 993 memcpy(dev->dev_addr, &p.iph.saddr, 4); 994 memcpy(dev->broadcast, &p.iph.daddr, 4); 995 ipip6_tunnel_link(sitn, t); 996 netdev_state_change(dev); 997 } 998 } 999 1000 if (t) { 1001 err = 0; 1002 if (cmd == SIOCCHGTUNNEL) { 1003 t->parms.iph.ttl = p.iph.ttl; 1004 t->parms.iph.tos = p.iph.tos; 1005 if (t->parms.link != p.link) { 1006 t->parms.link = p.link; 1007 ipip6_tunnel_bind_dev(dev); 1008 netdev_state_change(dev); 1009 } 1010 } 1011 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) 1012 err = -EFAULT; 1013 } else 1014 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1015 break; 1016 1017 case SIOCDELTUNNEL: 1018 err = -EPERM; 1019 if (!capable(CAP_NET_ADMIN)) 1020 goto done; 1021 1022 if (dev == sitn->fb_tunnel_dev) { 1023 err = -EFAULT; 1024 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1025 goto done; 1026 err = -ENOENT; 1027 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL) 1028 goto done; 1029 err = -EPERM; 1030 if (t == netdev_priv(sitn->fb_tunnel_dev)) 1031 goto done; 1032 dev = t->dev; 1033 } 1034 unregister_netdevice(dev); 1035 err = 0; 1036 break; 1037 1038 case SIOCGETPRL: 1039 err = -EINVAL; 1040 if (dev == sitn->fb_tunnel_dev) 1041 goto done; 1042 err = -ENOENT; 1043 if (!(t = netdev_priv(dev))) 1044 goto done; 1045 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data); 1046 break; 1047 1048 case SIOCADDPRL: 1049 case SIOCDELPRL: 1050 case SIOCCHGPRL: 1051 err = -EPERM; 1052 if (!capable(CAP_NET_ADMIN)) 1053 goto done; 1054 err = -EINVAL; 1055 if (dev == sitn->fb_tunnel_dev) 1056 goto done; 1057 err = -EFAULT; 1058 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl))) 1059 goto done; 1060 err = -ENOENT; 1061 if (!(t = netdev_priv(dev))) 1062 goto done; 1063 1064 switch (cmd) { 1065 case SIOCDELPRL: 1066 err = ipip6_tunnel_del_prl(t, &prl); 1067 break; 1068 case SIOCADDPRL: 1069 case SIOCCHGPRL: 1070 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); 1071 break; 1072 } 1073 netdev_state_change(dev); 1074 break; 1075 1076 #ifdef CONFIG_IPV6_SIT_6RD 1077 case SIOCADD6RD: 1078 case SIOCCHG6RD: 1079 case SIOCDEL6RD: 1080 err = -EPERM; 1081 if (!capable(CAP_NET_ADMIN)) 1082 goto done; 1083 1084 err = -EFAULT; 1085 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, 1086 sizeof(ip6rd))) 1087 goto done; 1088 1089 t = netdev_priv(dev); 1090 1091 if (cmd != SIOCDEL6RD) { 1092 struct in6_addr prefix; 1093 __be32 relay_prefix; 1094 1095 err = -EINVAL; 1096 if (ip6rd.relay_prefixlen > 32 || 1097 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64) 1098 goto done; 1099 1100 ipv6_addr_prefix(&prefix, &ip6rd.prefix, 1101 ip6rd.prefixlen); 1102 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix)) 1103 goto done; 1104 if (ip6rd.relay_prefixlen) 1105 relay_prefix = ip6rd.relay_prefix & 1106 htonl(0xffffffffUL << 1107 (32 - ip6rd.relay_prefixlen)); 1108 else 1109 relay_prefix = 0; 1110 if (relay_prefix != ip6rd.relay_prefix) 1111 goto done; 1112 1113 t->ip6rd.prefix = prefix; 1114 t->ip6rd.relay_prefix = relay_prefix; 1115 t->ip6rd.prefixlen = ip6rd.prefixlen; 1116 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen; 1117 } else 1118 ipip6_tunnel_clone_6rd(dev, sitn); 1119 1120 err = 0; 1121 break; 1122 #endif 1123 1124 default: 1125 err = -EINVAL; 1126 } 1127 1128 done: 1129 return err; 1130 } 1131 1132 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu) 1133 { 1134 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr)) 1135 return -EINVAL; 1136 dev->mtu = new_mtu; 1137 return 0; 1138 } 1139 1140 static const struct net_device_ops ipip6_netdev_ops = { 1141 .ndo_uninit = ipip6_tunnel_uninit, 1142 .ndo_start_xmit = ipip6_tunnel_xmit, 1143 .ndo_do_ioctl = ipip6_tunnel_ioctl, 1144 .ndo_change_mtu = ipip6_tunnel_change_mtu, 1145 .ndo_get_stats64= ipip6_get_stats64, 1146 }; 1147 1148 static void ipip6_dev_free(struct net_device *dev) 1149 { 1150 free_percpu(dev->tstats); 1151 free_netdev(dev); 1152 } 1153 1154 static void ipip6_tunnel_setup(struct net_device *dev) 1155 { 1156 dev->netdev_ops = &ipip6_netdev_ops; 1157 dev->destructor = ipip6_dev_free; 1158 1159 dev->type = ARPHRD_SIT; 1160 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr); 1161 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr); 1162 dev->flags = IFF_NOARP; 1163 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1164 dev->iflink = 0; 1165 dev->addr_len = 4; 1166 dev->features |= NETIF_F_NETNS_LOCAL; 1167 dev->features |= NETIF_F_LLTX; 1168 } 1169 1170 static int ipip6_tunnel_init(struct net_device *dev) 1171 { 1172 struct ip_tunnel *tunnel = netdev_priv(dev); 1173 1174 tunnel->dev = dev; 1175 1176 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); 1177 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); 1178 1179 ipip6_tunnel_bind_dev(dev); 1180 dev->tstats = alloc_percpu(struct pcpu_tstats); 1181 if (!dev->tstats) 1182 return -ENOMEM; 1183 1184 return 0; 1185 } 1186 1187 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev) 1188 { 1189 struct ip_tunnel *tunnel = netdev_priv(dev); 1190 struct iphdr *iph = &tunnel->parms.iph; 1191 struct net *net = dev_net(dev); 1192 struct sit_net *sitn = net_generic(net, sit_net_id); 1193 1194 tunnel->dev = dev; 1195 strcpy(tunnel->parms.name, dev->name); 1196 1197 iph->version = 4; 1198 iph->protocol = IPPROTO_IPV6; 1199 iph->ihl = 5; 1200 iph->ttl = 64; 1201 1202 dev->tstats = alloc_percpu(struct pcpu_tstats); 1203 if (!dev->tstats) 1204 return -ENOMEM; 1205 dev_hold(dev); 1206 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); 1207 return 0; 1208 } 1209 1210 static struct xfrm_tunnel sit_handler __read_mostly = { 1211 .handler = ipip6_rcv, 1212 .err_handler = ipip6_err, 1213 .priority = 1, 1214 }; 1215 1216 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head) 1217 { 1218 int prio; 1219 1220 for (prio = 1; prio < 4; prio++) { 1221 int h; 1222 for (h = 0; h < HASH_SIZE; h++) { 1223 struct ip_tunnel *t; 1224 1225 t = rtnl_dereference(sitn->tunnels[prio][h]); 1226 while (t != NULL) { 1227 unregister_netdevice_queue(t->dev, head); 1228 t = rtnl_dereference(t->next); 1229 } 1230 } 1231 } 1232 } 1233 1234 static int __net_init sit_init_net(struct net *net) 1235 { 1236 struct sit_net *sitn = net_generic(net, sit_net_id); 1237 struct ip_tunnel *t; 1238 int err; 1239 1240 sitn->tunnels[0] = sitn->tunnels_wc; 1241 sitn->tunnels[1] = sitn->tunnels_l; 1242 sitn->tunnels[2] = sitn->tunnels_r; 1243 sitn->tunnels[3] = sitn->tunnels_r_l; 1244 1245 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", 1246 ipip6_tunnel_setup); 1247 if (!sitn->fb_tunnel_dev) { 1248 err = -ENOMEM; 1249 goto err_alloc_dev; 1250 } 1251 dev_net_set(sitn->fb_tunnel_dev, net); 1252 1253 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); 1254 if (err) 1255 goto err_dev_free; 1256 1257 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); 1258 1259 if ((err = register_netdev(sitn->fb_tunnel_dev))) 1260 goto err_reg_dev; 1261 1262 t = netdev_priv(sitn->fb_tunnel_dev); 1263 1264 strcpy(t->parms.name, sitn->fb_tunnel_dev->name); 1265 return 0; 1266 1267 err_reg_dev: 1268 dev_put(sitn->fb_tunnel_dev); 1269 err_dev_free: 1270 ipip6_dev_free(sitn->fb_tunnel_dev); 1271 err_alloc_dev: 1272 return err; 1273 } 1274 1275 static void __net_exit sit_exit_net(struct net *net) 1276 { 1277 struct sit_net *sitn = net_generic(net, sit_net_id); 1278 LIST_HEAD(list); 1279 1280 rtnl_lock(); 1281 sit_destroy_tunnels(sitn, &list); 1282 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list); 1283 unregister_netdevice_many(&list); 1284 rtnl_unlock(); 1285 } 1286 1287 static struct pernet_operations sit_net_ops = { 1288 .init = sit_init_net, 1289 .exit = sit_exit_net, 1290 .id = &sit_net_id, 1291 .size = sizeof(struct sit_net), 1292 }; 1293 1294 static void __exit sit_cleanup(void) 1295 { 1296 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1297 1298 unregister_pernet_device(&sit_net_ops); 1299 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1300 } 1301 1302 static int __init sit_init(void) 1303 { 1304 int err; 1305 1306 pr_info("IPv6 over IPv4 tunneling driver\n"); 1307 1308 err = register_pernet_device(&sit_net_ops); 1309 if (err < 0) 1310 return err; 1311 err = xfrm4_tunnel_register(&sit_handler, AF_INET6); 1312 if (err < 0) { 1313 unregister_pernet_device(&sit_net_ops); 1314 pr_info("%s: can't add protocol\n", __func__); 1315 } 1316 return err; 1317 } 1318 1319 module_init(sit_init); 1320 module_exit(sit_cleanup); 1321 MODULE_LICENSE("GPL"); 1322 MODULE_ALIAS_NETDEV("sit0"); 1323