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 default: 531 /* All others are translated to HOST_UNREACH. 532 rfc2003 contains "deep thoughts" about NET_UNREACH, 533 I believe they are just ether pollution. --ANK 534 */ 535 break; 536 } 537 break; 538 case ICMP_TIME_EXCEEDED: 539 if (code != ICMP_EXC_TTL) 540 return 0; 541 break; 542 case ICMP_REDIRECT: 543 break; 544 } 545 546 err = -ENOENT; 547 548 rcu_read_lock(); 549 t = ipip6_tunnel_lookup(dev_net(skb->dev), 550 skb->dev, 551 iph->daddr, 552 iph->saddr); 553 if (t == NULL) 554 goto out; 555 556 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { 557 ipv4_update_pmtu(skb, dev_net(skb->dev), info, 558 t->dev->ifindex, 0, IPPROTO_IPV6, 0); 559 err = 0; 560 goto out; 561 } 562 if (type == ICMP_REDIRECT) { 563 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0, 564 IPPROTO_IPV6, 0); 565 err = 0; 566 goto out; 567 } 568 569 if (t->parms.iph.daddr == 0) 570 goto out; 571 572 err = 0; 573 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) 574 goto out; 575 576 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) 577 t->err_count++; 578 else 579 t->err_count = 1; 580 t->err_time = jiffies; 581 out: 582 rcu_read_unlock(); 583 return err; 584 } 585 586 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb) 587 { 588 if (INET_ECN_is_ce(iph->tos)) 589 IP6_ECN_set_ce(ipv6_hdr(skb)); 590 } 591 592 static int ipip6_rcv(struct sk_buff *skb) 593 { 594 const struct iphdr *iph; 595 struct ip_tunnel *tunnel; 596 597 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) 598 goto out; 599 600 iph = ip_hdr(skb); 601 602 rcu_read_lock(); 603 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 604 iph->saddr, iph->daddr); 605 if (tunnel != NULL) { 606 struct pcpu_tstats *tstats; 607 608 secpath_reset(skb); 609 skb->mac_header = skb->network_header; 610 skb_reset_network_header(skb); 611 IPCB(skb)->flags = 0; 612 skb->protocol = htons(ETH_P_IPV6); 613 skb->pkt_type = PACKET_HOST; 614 615 if ((tunnel->dev->priv_flags & IFF_ISATAP) && 616 !isatap_chksrc(skb, iph, tunnel)) { 617 tunnel->dev->stats.rx_errors++; 618 rcu_read_unlock(); 619 kfree_skb(skb); 620 return 0; 621 } 622 623 tstats = this_cpu_ptr(tunnel->dev->tstats); 624 tstats->rx_packets++; 625 tstats->rx_bytes += skb->len; 626 627 __skb_tunnel_rx(skb, tunnel->dev); 628 629 ipip6_ecn_decapsulate(iph, skb); 630 631 netif_rx(skb); 632 633 rcu_read_unlock(); 634 return 0; 635 } 636 637 /* no tunnel matched, let upstream know, ipsec may handle it */ 638 rcu_read_unlock(); 639 return 1; 640 out: 641 kfree_skb(skb); 642 return 0; 643 } 644 645 /* 646 * Returns the embedded IPv4 address if the IPv6 address 647 * comes from 6rd / 6to4 (RFC 3056) addr space. 648 */ 649 static inline 650 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel) 651 { 652 __be32 dst = 0; 653 654 #ifdef CONFIG_IPV6_SIT_6RD 655 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, 656 tunnel->ip6rd.prefixlen)) { 657 unsigned int pbw0, pbi0; 658 int pbi1; 659 u32 d; 660 661 pbw0 = tunnel->ip6rd.prefixlen >> 5; 662 pbi0 = tunnel->ip6rd.prefixlen & 0x1f; 663 664 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> 665 tunnel->ip6rd.relay_prefixlen; 666 667 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; 668 if (pbi1 > 0) 669 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> 670 (32 - pbi1); 671 672 dst = tunnel->ip6rd.relay_prefix | htonl(d); 673 } 674 #else 675 if (v6dst->s6_addr16[0] == htons(0x2002)) { 676 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ 677 memcpy(&dst, &v6dst->s6_addr16[1], 4); 678 } 679 #endif 680 return dst; 681 } 682 683 /* 684 * This function assumes it is being called from dev_queue_xmit() 685 * and that skb is filled properly by that function. 686 */ 687 688 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, 689 struct net_device *dev) 690 { 691 struct ip_tunnel *tunnel = netdev_priv(dev); 692 struct pcpu_tstats *tstats; 693 const struct iphdr *tiph = &tunnel->parms.iph; 694 const struct ipv6hdr *iph6 = ipv6_hdr(skb); 695 u8 tos = tunnel->parms.iph.tos; 696 __be16 df = tiph->frag_off; 697 struct rtable *rt; /* Route to the other host */ 698 struct net_device *tdev; /* Device to other host */ 699 struct iphdr *iph; /* Our new IP header */ 700 unsigned int max_headroom; /* The extra header space needed */ 701 __be32 dst = tiph->daddr; 702 struct flowi4 fl4; 703 int mtu; 704 const struct in6_addr *addr6; 705 int addr_type; 706 707 if (skb->protocol != htons(ETH_P_IPV6)) 708 goto tx_error; 709 710 if (tos == 1) 711 tos = ipv6_get_dsfield(iph6); 712 713 /* ISATAP (RFC4214) - must come before 6to4 */ 714 if (dev->priv_flags & IFF_ISATAP) { 715 struct neighbour *neigh = NULL; 716 bool do_tx_error = false; 717 718 if (skb_dst(skb)) 719 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 720 721 if (neigh == NULL) { 722 net_dbg_ratelimited("sit: nexthop == NULL\n"); 723 goto tx_error; 724 } 725 726 addr6 = (const struct in6_addr *)&neigh->primary_key; 727 addr_type = ipv6_addr_type(addr6); 728 729 if ((addr_type & IPV6_ADDR_UNICAST) && 730 ipv6_addr_is_isatap(addr6)) 731 dst = addr6->s6_addr32[3]; 732 else 733 do_tx_error = true; 734 735 neigh_release(neigh); 736 if (do_tx_error) 737 goto tx_error; 738 } 739 740 if (!dst) 741 dst = try_6rd(&iph6->daddr, tunnel); 742 743 if (!dst) { 744 struct neighbour *neigh = NULL; 745 bool do_tx_error = false; 746 747 if (skb_dst(skb)) 748 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 749 750 if (neigh == NULL) { 751 net_dbg_ratelimited("sit: nexthop == NULL\n"); 752 goto tx_error; 753 } 754 755 addr6 = (const struct in6_addr *)&neigh->primary_key; 756 addr_type = ipv6_addr_type(addr6); 757 758 if (addr_type == IPV6_ADDR_ANY) { 759 addr6 = &ipv6_hdr(skb)->daddr; 760 addr_type = ipv6_addr_type(addr6); 761 } 762 763 if ((addr_type & IPV6_ADDR_COMPATv4) != 0) 764 dst = addr6->s6_addr32[3]; 765 else 766 do_tx_error = true; 767 768 neigh_release(neigh); 769 if (do_tx_error) 770 goto tx_error; 771 } 772 773 rt = ip_route_output_ports(dev_net(dev), &fl4, NULL, 774 dst, tiph->saddr, 775 0, 0, 776 IPPROTO_IPV6, RT_TOS(tos), 777 tunnel->parms.link); 778 if (IS_ERR(rt)) { 779 dev->stats.tx_carrier_errors++; 780 goto tx_error_icmp; 781 } 782 if (rt->rt_type != RTN_UNICAST) { 783 ip_rt_put(rt); 784 dev->stats.tx_carrier_errors++; 785 goto tx_error_icmp; 786 } 787 tdev = rt->dst.dev; 788 789 if (tdev == dev) { 790 ip_rt_put(rt); 791 dev->stats.collisions++; 792 goto tx_error; 793 } 794 795 if (df) { 796 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr); 797 798 if (mtu < 68) { 799 dev->stats.collisions++; 800 ip_rt_put(rt); 801 goto tx_error; 802 } 803 804 if (mtu < IPV6_MIN_MTU) { 805 mtu = IPV6_MIN_MTU; 806 df = 0; 807 } 808 809 if (tunnel->parms.iph.daddr && skb_dst(skb)) 810 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 811 812 if (skb->len > mtu) { 813 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 814 ip_rt_put(rt); 815 goto tx_error; 816 } 817 } 818 819 if (tunnel->err_count > 0) { 820 if (time_before(jiffies, 821 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 822 tunnel->err_count--; 823 dst_link_failure(skb); 824 } else 825 tunnel->err_count = 0; 826 } 827 828 /* 829 * Okay, now see if we can stuff it in the buffer as-is. 830 */ 831 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr); 832 833 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 834 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 835 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 836 if (!new_skb) { 837 ip_rt_put(rt); 838 dev->stats.tx_dropped++; 839 dev_kfree_skb(skb); 840 return NETDEV_TX_OK; 841 } 842 if (skb->sk) 843 skb_set_owner_w(new_skb, skb->sk); 844 dev_kfree_skb(skb); 845 skb = new_skb; 846 iph6 = ipv6_hdr(skb); 847 } 848 849 skb->transport_header = skb->network_header; 850 skb_push(skb, sizeof(struct iphdr)); 851 skb_reset_network_header(skb); 852 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 853 IPCB(skb)->flags = 0; 854 skb_dst_drop(skb); 855 skb_dst_set(skb, &rt->dst); 856 857 /* 858 * Push down and install the IPIP header. 859 */ 860 861 iph = ip_hdr(skb); 862 iph->version = 4; 863 iph->ihl = sizeof(struct iphdr)>>2; 864 iph->frag_off = df; 865 iph->protocol = IPPROTO_IPV6; 866 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); 867 iph->daddr = fl4.daddr; 868 iph->saddr = fl4.saddr; 869 870 if ((iph->ttl = tiph->ttl) == 0) 871 iph->ttl = iph6->hop_limit; 872 873 nf_reset(skb); 874 tstats = this_cpu_ptr(dev->tstats); 875 __IPTUNNEL_XMIT(tstats, &dev->stats); 876 return NETDEV_TX_OK; 877 878 tx_error_icmp: 879 dst_link_failure(skb); 880 tx_error: 881 dev->stats.tx_errors++; 882 dev_kfree_skb(skb); 883 return NETDEV_TX_OK; 884 } 885 886 static void ipip6_tunnel_bind_dev(struct net_device *dev) 887 { 888 struct net_device *tdev = NULL; 889 struct ip_tunnel *tunnel; 890 const struct iphdr *iph; 891 struct flowi4 fl4; 892 893 tunnel = netdev_priv(dev); 894 iph = &tunnel->parms.iph; 895 896 if (iph->daddr) { 897 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL, 898 iph->daddr, iph->saddr, 899 0, 0, 900 IPPROTO_IPV6, 901 RT_TOS(iph->tos), 902 tunnel->parms.link); 903 904 if (!IS_ERR(rt)) { 905 tdev = rt->dst.dev; 906 ip_rt_put(rt); 907 } 908 dev->flags |= IFF_POINTOPOINT; 909 } 910 911 if (!tdev && tunnel->parms.link) 912 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link); 913 914 if (tdev) { 915 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); 916 dev->mtu = tdev->mtu - sizeof(struct iphdr); 917 if (dev->mtu < IPV6_MIN_MTU) 918 dev->mtu = IPV6_MIN_MTU; 919 } 920 dev->iflink = tunnel->parms.link; 921 } 922 923 static int 924 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) 925 { 926 int err = 0; 927 struct ip_tunnel_parm p; 928 struct ip_tunnel_prl prl; 929 struct ip_tunnel *t; 930 struct net *net = dev_net(dev); 931 struct sit_net *sitn = net_generic(net, sit_net_id); 932 #ifdef CONFIG_IPV6_SIT_6RD 933 struct ip_tunnel_6rd ip6rd; 934 #endif 935 936 switch (cmd) { 937 case SIOCGETTUNNEL: 938 #ifdef CONFIG_IPV6_SIT_6RD 939 case SIOCGET6RD: 940 #endif 941 t = NULL; 942 if (dev == sitn->fb_tunnel_dev) { 943 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 944 err = -EFAULT; 945 break; 946 } 947 t = ipip6_tunnel_locate(net, &p, 0); 948 } 949 if (t == NULL) 950 t = netdev_priv(dev); 951 952 err = -EFAULT; 953 if (cmd == SIOCGETTUNNEL) { 954 memcpy(&p, &t->parms, sizeof(p)); 955 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, 956 sizeof(p))) 957 goto done; 958 #ifdef CONFIG_IPV6_SIT_6RD 959 } else { 960 ip6rd.prefix = t->ip6rd.prefix; 961 ip6rd.relay_prefix = t->ip6rd.relay_prefix; 962 ip6rd.prefixlen = t->ip6rd.prefixlen; 963 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; 964 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, 965 sizeof(ip6rd))) 966 goto done; 967 #endif 968 } 969 err = 0; 970 break; 971 972 case SIOCADDTUNNEL: 973 case SIOCCHGTUNNEL: 974 err = -EPERM; 975 if (!capable(CAP_NET_ADMIN)) 976 goto done; 977 978 err = -EFAULT; 979 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 980 goto done; 981 982 err = -EINVAL; 983 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 || 984 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) 985 goto done; 986 if (p.iph.ttl) 987 p.iph.frag_off |= htons(IP_DF); 988 989 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); 990 991 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 992 if (t != NULL) { 993 if (t->dev != dev) { 994 err = -EEXIST; 995 break; 996 } 997 } else { 998 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || 999 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { 1000 err = -EINVAL; 1001 break; 1002 } 1003 t = netdev_priv(dev); 1004 ipip6_tunnel_unlink(sitn, t); 1005 synchronize_net(); 1006 t->parms.iph.saddr = p.iph.saddr; 1007 t->parms.iph.daddr = p.iph.daddr; 1008 memcpy(dev->dev_addr, &p.iph.saddr, 4); 1009 memcpy(dev->broadcast, &p.iph.daddr, 4); 1010 ipip6_tunnel_link(sitn, t); 1011 netdev_state_change(dev); 1012 } 1013 } 1014 1015 if (t) { 1016 err = 0; 1017 if (cmd == SIOCCHGTUNNEL) { 1018 t->parms.iph.ttl = p.iph.ttl; 1019 t->parms.iph.tos = p.iph.tos; 1020 if (t->parms.link != p.link) { 1021 t->parms.link = p.link; 1022 ipip6_tunnel_bind_dev(dev); 1023 netdev_state_change(dev); 1024 } 1025 } 1026 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) 1027 err = -EFAULT; 1028 } else 1029 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1030 break; 1031 1032 case SIOCDELTUNNEL: 1033 err = -EPERM; 1034 if (!capable(CAP_NET_ADMIN)) 1035 goto done; 1036 1037 if (dev == sitn->fb_tunnel_dev) { 1038 err = -EFAULT; 1039 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1040 goto done; 1041 err = -ENOENT; 1042 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL) 1043 goto done; 1044 err = -EPERM; 1045 if (t == netdev_priv(sitn->fb_tunnel_dev)) 1046 goto done; 1047 dev = t->dev; 1048 } 1049 unregister_netdevice(dev); 1050 err = 0; 1051 break; 1052 1053 case SIOCGETPRL: 1054 err = -EINVAL; 1055 if (dev == sitn->fb_tunnel_dev) 1056 goto done; 1057 err = -ENOENT; 1058 if (!(t = netdev_priv(dev))) 1059 goto done; 1060 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data); 1061 break; 1062 1063 case SIOCADDPRL: 1064 case SIOCDELPRL: 1065 case SIOCCHGPRL: 1066 err = -EPERM; 1067 if (!capable(CAP_NET_ADMIN)) 1068 goto done; 1069 err = -EINVAL; 1070 if (dev == sitn->fb_tunnel_dev) 1071 goto done; 1072 err = -EFAULT; 1073 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl))) 1074 goto done; 1075 err = -ENOENT; 1076 if (!(t = netdev_priv(dev))) 1077 goto done; 1078 1079 switch (cmd) { 1080 case SIOCDELPRL: 1081 err = ipip6_tunnel_del_prl(t, &prl); 1082 break; 1083 case SIOCADDPRL: 1084 case SIOCCHGPRL: 1085 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); 1086 break; 1087 } 1088 netdev_state_change(dev); 1089 break; 1090 1091 #ifdef CONFIG_IPV6_SIT_6RD 1092 case SIOCADD6RD: 1093 case SIOCCHG6RD: 1094 case SIOCDEL6RD: 1095 err = -EPERM; 1096 if (!capable(CAP_NET_ADMIN)) 1097 goto done; 1098 1099 err = -EFAULT; 1100 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, 1101 sizeof(ip6rd))) 1102 goto done; 1103 1104 t = netdev_priv(dev); 1105 1106 if (cmd != SIOCDEL6RD) { 1107 struct in6_addr prefix; 1108 __be32 relay_prefix; 1109 1110 err = -EINVAL; 1111 if (ip6rd.relay_prefixlen > 32 || 1112 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64) 1113 goto done; 1114 1115 ipv6_addr_prefix(&prefix, &ip6rd.prefix, 1116 ip6rd.prefixlen); 1117 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix)) 1118 goto done; 1119 if (ip6rd.relay_prefixlen) 1120 relay_prefix = ip6rd.relay_prefix & 1121 htonl(0xffffffffUL << 1122 (32 - ip6rd.relay_prefixlen)); 1123 else 1124 relay_prefix = 0; 1125 if (relay_prefix != ip6rd.relay_prefix) 1126 goto done; 1127 1128 t->ip6rd.prefix = prefix; 1129 t->ip6rd.relay_prefix = relay_prefix; 1130 t->ip6rd.prefixlen = ip6rd.prefixlen; 1131 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen; 1132 } else 1133 ipip6_tunnel_clone_6rd(dev, sitn); 1134 1135 err = 0; 1136 break; 1137 #endif 1138 1139 default: 1140 err = -EINVAL; 1141 } 1142 1143 done: 1144 return err; 1145 } 1146 1147 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu) 1148 { 1149 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr)) 1150 return -EINVAL; 1151 dev->mtu = new_mtu; 1152 return 0; 1153 } 1154 1155 static const struct net_device_ops ipip6_netdev_ops = { 1156 .ndo_uninit = ipip6_tunnel_uninit, 1157 .ndo_start_xmit = ipip6_tunnel_xmit, 1158 .ndo_do_ioctl = ipip6_tunnel_ioctl, 1159 .ndo_change_mtu = ipip6_tunnel_change_mtu, 1160 .ndo_get_stats64= ipip6_get_stats64, 1161 }; 1162 1163 static void ipip6_dev_free(struct net_device *dev) 1164 { 1165 free_percpu(dev->tstats); 1166 free_netdev(dev); 1167 } 1168 1169 static void ipip6_tunnel_setup(struct net_device *dev) 1170 { 1171 dev->netdev_ops = &ipip6_netdev_ops; 1172 dev->destructor = ipip6_dev_free; 1173 1174 dev->type = ARPHRD_SIT; 1175 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr); 1176 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr); 1177 dev->flags = IFF_NOARP; 1178 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1179 dev->iflink = 0; 1180 dev->addr_len = 4; 1181 dev->features |= NETIF_F_NETNS_LOCAL; 1182 dev->features |= NETIF_F_LLTX; 1183 } 1184 1185 static int ipip6_tunnel_init(struct net_device *dev) 1186 { 1187 struct ip_tunnel *tunnel = netdev_priv(dev); 1188 1189 tunnel->dev = dev; 1190 1191 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); 1192 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); 1193 1194 ipip6_tunnel_bind_dev(dev); 1195 dev->tstats = alloc_percpu(struct pcpu_tstats); 1196 if (!dev->tstats) 1197 return -ENOMEM; 1198 1199 return 0; 1200 } 1201 1202 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev) 1203 { 1204 struct ip_tunnel *tunnel = netdev_priv(dev); 1205 struct iphdr *iph = &tunnel->parms.iph; 1206 struct net *net = dev_net(dev); 1207 struct sit_net *sitn = net_generic(net, sit_net_id); 1208 1209 tunnel->dev = dev; 1210 strcpy(tunnel->parms.name, dev->name); 1211 1212 iph->version = 4; 1213 iph->protocol = IPPROTO_IPV6; 1214 iph->ihl = 5; 1215 iph->ttl = 64; 1216 1217 dev->tstats = alloc_percpu(struct pcpu_tstats); 1218 if (!dev->tstats) 1219 return -ENOMEM; 1220 dev_hold(dev); 1221 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); 1222 return 0; 1223 } 1224 1225 static struct xfrm_tunnel sit_handler __read_mostly = { 1226 .handler = ipip6_rcv, 1227 .err_handler = ipip6_err, 1228 .priority = 1, 1229 }; 1230 1231 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head) 1232 { 1233 int prio; 1234 1235 for (prio = 1; prio < 4; prio++) { 1236 int h; 1237 for (h = 0; h < HASH_SIZE; h++) { 1238 struct ip_tunnel *t; 1239 1240 t = rtnl_dereference(sitn->tunnels[prio][h]); 1241 while (t != NULL) { 1242 unregister_netdevice_queue(t->dev, head); 1243 t = rtnl_dereference(t->next); 1244 } 1245 } 1246 } 1247 } 1248 1249 static int __net_init sit_init_net(struct net *net) 1250 { 1251 struct sit_net *sitn = net_generic(net, sit_net_id); 1252 struct ip_tunnel *t; 1253 int err; 1254 1255 sitn->tunnels[0] = sitn->tunnels_wc; 1256 sitn->tunnels[1] = sitn->tunnels_l; 1257 sitn->tunnels[2] = sitn->tunnels_r; 1258 sitn->tunnels[3] = sitn->tunnels_r_l; 1259 1260 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", 1261 ipip6_tunnel_setup); 1262 if (!sitn->fb_tunnel_dev) { 1263 err = -ENOMEM; 1264 goto err_alloc_dev; 1265 } 1266 dev_net_set(sitn->fb_tunnel_dev, net); 1267 1268 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); 1269 if (err) 1270 goto err_dev_free; 1271 1272 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); 1273 1274 if ((err = register_netdev(sitn->fb_tunnel_dev))) 1275 goto err_reg_dev; 1276 1277 t = netdev_priv(sitn->fb_tunnel_dev); 1278 1279 strcpy(t->parms.name, sitn->fb_tunnel_dev->name); 1280 return 0; 1281 1282 err_reg_dev: 1283 dev_put(sitn->fb_tunnel_dev); 1284 err_dev_free: 1285 ipip6_dev_free(sitn->fb_tunnel_dev); 1286 err_alloc_dev: 1287 return err; 1288 } 1289 1290 static void __net_exit sit_exit_net(struct net *net) 1291 { 1292 struct sit_net *sitn = net_generic(net, sit_net_id); 1293 LIST_HEAD(list); 1294 1295 rtnl_lock(); 1296 sit_destroy_tunnels(sitn, &list); 1297 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list); 1298 unregister_netdevice_many(&list); 1299 rtnl_unlock(); 1300 } 1301 1302 static struct pernet_operations sit_net_ops = { 1303 .init = sit_init_net, 1304 .exit = sit_exit_net, 1305 .id = &sit_net_id, 1306 .size = sizeof(struct sit_net), 1307 }; 1308 1309 static void __exit sit_cleanup(void) 1310 { 1311 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1312 1313 unregister_pernet_device(&sit_net_ops); 1314 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1315 } 1316 1317 static int __init sit_init(void) 1318 { 1319 int err; 1320 1321 pr_info("IPv6 over IPv4 tunneling driver\n"); 1322 1323 err = register_pernet_device(&sit_net_ops); 1324 if (err < 0) 1325 return err; 1326 err = xfrm4_tunnel_register(&sit_handler, AF_INET6); 1327 if (err < 0) { 1328 unregister_pernet_device(&sit_net_ops); 1329 pr_info("%s: can't add protocol\n", __func__); 1330 } 1331 return err; 1332 } 1333 1334 module_init(sit_init); 1335 module_exit(sit_cleanup); 1336 MODULE_LICENSE("GPL"); 1337 MODULE_ALIAS_NETDEV("sit0"); 1338