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/ip_tunnels.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 bool log_ecn_error = true; 69 module_param(log_ecn_error, bool, 0644); 70 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 71 72 static int ipip6_tunnel_init(struct net_device *dev); 73 static void ipip6_tunnel_setup(struct net_device *dev); 74 static void ipip6_dev_free(struct net_device *dev); 75 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, 76 __be32 *v4dst); 77 static struct rtnl_link_ops sit_link_ops __read_mostly; 78 79 static int sit_net_id __read_mostly; 80 struct sit_net { 81 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE]; 82 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE]; 83 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE]; 84 struct ip_tunnel __rcu *tunnels_wc[1]; 85 struct ip_tunnel __rcu **tunnels[4]; 86 87 struct net_device *fb_tunnel_dev; 88 }; 89 90 /* 91 * Must be invoked with rcu_read_lock 92 */ 93 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net, 94 struct net_device *dev, __be32 remote, __be32 local) 95 { 96 unsigned int h0 = HASH(remote); 97 unsigned int h1 = HASH(local); 98 struct ip_tunnel *t; 99 struct sit_net *sitn = net_generic(net, sit_net_id); 100 101 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) { 102 if (local == t->parms.iph.saddr && 103 remote == t->parms.iph.daddr && 104 (!dev || !t->parms.link || dev->iflink == t->parms.link) && 105 (t->dev->flags & IFF_UP)) 106 return t; 107 } 108 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) { 109 if (remote == t->parms.iph.daddr && 110 (!dev || !t->parms.link || dev->iflink == t->parms.link) && 111 (t->dev->flags & IFF_UP)) 112 return t; 113 } 114 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) { 115 if (local == t->parms.iph.saddr && 116 (!dev || !t->parms.link || dev->iflink == t->parms.link) && 117 (t->dev->flags & IFF_UP)) 118 return t; 119 } 120 t = rcu_dereference(sitn->tunnels_wc[0]); 121 if ((t != NULL) && (t->dev->flags & IFF_UP)) 122 return t; 123 return NULL; 124 } 125 126 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn, 127 struct ip_tunnel_parm *parms) 128 { 129 __be32 remote = parms->iph.daddr; 130 __be32 local = parms->iph.saddr; 131 unsigned int h = 0; 132 int prio = 0; 133 134 if (remote) { 135 prio |= 2; 136 h ^= HASH(remote); 137 } 138 if (local) { 139 prio |= 1; 140 h ^= HASH(local); 141 } 142 return &sitn->tunnels[prio][h]; 143 } 144 145 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn, 146 struct ip_tunnel *t) 147 { 148 return __ipip6_bucket(sitn, &t->parms); 149 } 150 151 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t) 152 { 153 struct ip_tunnel __rcu **tp; 154 struct ip_tunnel *iter; 155 156 for (tp = ipip6_bucket(sitn, t); 157 (iter = rtnl_dereference(*tp)) != NULL; 158 tp = &iter->next) { 159 if (t == iter) { 160 rcu_assign_pointer(*tp, t->next); 161 break; 162 } 163 } 164 } 165 166 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) 167 { 168 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t); 169 170 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 171 rcu_assign_pointer(*tp, t); 172 } 173 174 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn) 175 { 176 #ifdef CONFIG_IPV6_SIT_6RD 177 struct ip_tunnel *t = netdev_priv(dev); 178 179 if (t->dev == sitn->fb_tunnel_dev) { 180 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0); 181 t->ip6rd.relay_prefix = 0; 182 t->ip6rd.prefixlen = 16; 183 t->ip6rd.relay_prefixlen = 0; 184 } else { 185 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev); 186 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd)); 187 } 188 #endif 189 } 190 191 static int ipip6_tunnel_create(struct net_device *dev) 192 { 193 struct ip_tunnel *t = netdev_priv(dev); 194 struct net *net = dev_net(dev); 195 struct sit_net *sitn = net_generic(net, sit_net_id); 196 int err; 197 198 err = ipip6_tunnel_init(dev); 199 if (err < 0) 200 goto out; 201 ipip6_tunnel_clone_6rd(dev, sitn); 202 203 if ((__force u16)t->parms.i_flags & SIT_ISATAP) 204 dev->priv_flags |= IFF_ISATAP; 205 206 err = register_netdevice(dev); 207 if (err < 0) 208 goto out; 209 210 strcpy(t->parms.name, dev->name); 211 dev->rtnl_link_ops = &sit_link_ops; 212 213 dev_hold(dev); 214 215 ipip6_tunnel_link(sitn, t); 216 return 0; 217 218 out: 219 return err; 220 } 221 222 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, 223 struct ip_tunnel_parm *parms, int create) 224 { 225 __be32 remote = parms->iph.daddr; 226 __be32 local = parms->iph.saddr; 227 struct ip_tunnel *t, *nt; 228 struct ip_tunnel __rcu **tp; 229 struct net_device *dev; 230 char name[IFNAMSIZ]; 231 struct sit_net *sitn = net_generic(net, sit_net_id); 232 233 for (tp = __ipip6_bucket(sitn, parms); 234 (t = rtnl_dereference(*tp)) != NULL; 235 tp = &t->next) { 236 if (local == t->parms.iph.saddr && 237 remote == t->parms.iph.daddr && 238 parms->link == t->parms.link) { 239 if (create) 240 return NULL; 241 else 242 return t; 243 } 244 } 245 if (!create) 246 goto failed; 247 248 if (parms->name[0]) 249 strlcpy(name, parms->name, IFNAMSIZ); 250 else 251 strcpy(name, "sit%d"); 252 253 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup); 254 if (dev == NULL) 255 return NULL; 256 257 dev_net_set(dev, net); 258 259 nt = netdev_priv(dev); 260 261 nt->parms = *parms; 262 if (ipip6_tunnel_create(dev) < 0) 263 goto failed_free; 264 265 return nt; 266 267 failed_free: 268 ipip6_dev_free(dev); 269 failed: 270 return NULL; 271 } 272 273 #define for_each_prl_rcu(start) \ 274 for (prl = rcu_dereference(start); \ 275 prl; \ 276 prl = rcu_dereference(prl->next)) 277 278 static struct ip_tunnel_prl_entry * 279 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr) 280 { 281 struct ip_tunnel_prl_entry *prl; 282 283 for_each_prl_rcu(t->prl) 284 if (prl->addr == addr) 285 break; 286 return prl; 287 288 } 289 290 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, 291 struct ip_tunnel_prl __user *a) 292 { 293 struct ip_tunnel_prl kprl, *kp; 294 struct ip_tunnel_prl_entry *prl; 295 unsigned int cmax, c = 0, ca, len; 296 int ret = 0; 297 298 if (copy_from_user(&kprl, a, sizeof(kprl))) 299 return -EFAULT; 300 cmax = kprl.datalen / sizeof(kprl); 301 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY)) 302 cmax = 1; 303 304 /* For simple GET or for root users, 305 * we try harder to allocate. 306 */ 307 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ? 308 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) : 309 NULL; 310 311 rcu_read_lock(); 312 313 ca = t->prl_count < cmax ? t->prl_count : cmax; 314 315 if (!kp) { 316 /* We don't try hard to allocate much memory for 317 * non-root users. 318 * For root users, retry allocating enough memory for 319 * the answer. 320 */ 321 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC); 322 if (!kp) { 323 ret = -ENOMEM; 324 goto out; 325 } 326 } 327 328 c = 0; 329 for_each_prl_rcu(t->prl) { 330 if (c >= cmax) 331 break; 332 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr) 333 continue; 334 kp[c].addr = prl->addr; 335 kp[c].flags = prl->flags; 336 c++; 337 if (kprl.addr != htonl(INADDR_ANY)) 338 break; 339 } 340 out: 341 rcu_read_unlock(); 342 343 len = sizeof(*kp) * c; 344 ret = 0; 345 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen)) 346 ret = -EFAULT; 347 348 kfree(kp); 349 350 return ret; 351 } 352 353 static int 354 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) 355 { 356 struct ip_tunnel_prl_entry *p; 357 int err = 0; 358 359 if (a->addr == htonl(INADDR_ANY)) 360 return -EINVAL; 361 362 ASSERT_RTNL(); 363 364 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) { 365 if (p->addr == a->addr) { 366 if (chg) { 367 p->flags = a->flags; 368 goto out; 369 } 370 err = -EEXIST; 371 goto out; 372 } 373 } 374 375 if (chg) { 376 err = -ENXIO; 377 goto out; 378 } 379 380 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL); 381 if (!p) { 382 err = -ENOBUFS; 383 goto out; 384 } 385 386 p->next = t->prl; 387 p->addr = a->addr; 388 p->flags = a->flags; 389 t->prl_count++; 390 rcu_assign_pointer(t->prl, p); 391 out: 392 return err; 393 } 394 395 static void prl_list_destroy_rcu(struct rcu_head *head) 396 { 397 struct ip_tunnel_prl_entry *p, *n; 398 399 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head); 400 do { 401 n = rcu_dereference_protected(p->next, 1); 402 kfree(p); 403 p = n; 404 } while (p); 405 } 406 407 static int 408 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) 409 { 410 struct ip_tunnel_prl_entry *x; 411 struct ip_tunnel_prl_entry __rcu **p; 412 int err = 0; 413 414 ASSERT_RTNL(); 415 416 if (a && a->addr != htonl(INADDR_ANY)) { 417 for (p = &t->prl; 418 (x = rtnl_dereference(*p)) != NULL; 419 p = &x->next) { 420 if (x->addr == a->addr) { 421 *p = x->next; 422 kfree_rcu(x, rcu_head); 423 t->prl_count--; 424 goto out; 425 } 426 } 427 err = -ENXIO; 428 } else { 429 x = rtnl_dereference(t->prl); 430 if (x) { 431 t->prl_count = 0; 432 call_rcu(&x->rcu_head, prl_list_destroy_rcu); 433 t->prl = NULL; 434 } 435 } 436 out: 437 return err; 438 } 439 440 static int 441 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t) 442 { 443 struct ip_tunnel_prl_entry *p; 444 int ok = 1; 445 446 rcu_read_lock(); 447 p = __ipip6_tunnel_locate_prl(t, iph->saddr); 448 if (p) { 449 if (p->flags & PRL_DEFAULT) 450 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT; 451 else 452 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT; 453 } else { 454 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr; 455 456 if (ipv6_addr_is_isatap(addr6) && 457 (addr6->s6_addr32[3] == iph->saddr) && 458 ipv6_chk_prefix(addr6, t->dev)) 459 skb->ndisc_nodetype = NDISC_NODETYPE_HOST; 460 else 461 ok = 0; 462 } 463 rcu_read_unlock(); 464 return ok; 465 } 466 467 static void ipip6_tunnel_uninit(struct net_device *dev) 468 { 469 struct ip_tunnel *tunnel = netdev_priv(dev); 470 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id); 471 472 if (dev == sitn->fb_tunnel_dev) { 473 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL); 474 } else { 475 ipip6_tunnel_unlink(sitn, tunnel); 476 ipip6_tunnel_del_prl(tunnel, NULL); 477 } 478 dev_put(dev); 479 } 480 481 482 static int ipip6_err(struct sk_buff *skb, u32 info) 483 { 484 485 /* All the routers (except for Linux) return only 486 8 bytes of packet payload. It means, that precise relaying of 487 ICMP in the real Internet is absolutely infeasible. 488 */ 489 const struct iphdr *iph = (const struct iphdr *)skb->data; 490 const int type = icmp_hdr(skb)->type; 491 const int code = icmp_hdr(skb)->code; 492 struct ip_tunnel *t; 493 int err; 494 495 switch (type) { 496 default: 497 case ICMP_PARAMETERPROB: 498 return 0; 499 500 case ICMP_DEST_UNREACH: 501 switch (code) { 502 case ICMP_SR_FAILED: 503 case ICMP_PORT_UNREACH: 504 /* Impossible event. */ 505 return 0; 506 default: 507 /* All others are translated to HOST_UNREACH. 508 rfc2003 contains "deep thoughts" about NET_UNREACH, 509 I believe they are just ether pollution. --ANK 510 */ 511 break; 512 } 513 break; 514 case ICMP_TIME_EXCEEDED: 515 if (code != ICMP_EXC_TTL) 516 return 0; 517 break; 518 case ICMP_REDIRECT: 519 break; 520 } 521 522 err = -ENOENT; 523 524 t = ipip6_tunnel_lookup(dev_net(skb->dev), 525 skb->dev, 526 iph->daddr, 527 iph->saddr); 528 if (t == NULL) 529 goto out; 530 531 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { 532 ipv4_update_pmtu(skb, dev_net(skb->dev), info, 533 t->dev->ifindex, 0, IPPROTO_IPV6, 0); 534 err = 0; 535 goto out; 536 } 537 if (type == ICMP_REDIRECT) { 538 ipv4_redirect(skb, dev_net(skb->dev), t->dev->ifindex, 0, 539 IPPROTO_IPV6, 0); 540 err = 0; 541 goto out; 542 } 543 544 if (t->parms.iph.daddr == 0) 545 goto out; 546 547 err = 0; 548 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) 549 goto out; 550 551 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) 552 t->err_count++; 553 else 554 t->err_count = 1; 555 t->err_time = jiffies; 556 out: 557 return err; 558 } 559 560 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr, 561 const struct in6_addr *v6addr) 562 { 563 __be32 v4embed = 0; 564 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed) 565 return true; 566 return false; 567 } 568 569 /* Checks if an address matches an address on the tunnel interface. 570 * Used to detect the NAT of proto 41 packets and let them pass spoofing test. 571 * Long story: 572 * This function is called after we considered the packet as spoofed 573 * in is_spoofed_6rd. 574 * We may have a router that is doing NAT for proto 41 packets 575 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb 576 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd 577 * function will return true, dropping the packet. 578 * But, we can still check if is spoofed against the IP 579 * addresses associated with the interface. 580 */ 581 static bool only_dnatted(const struct ip_tunnel *tunnel, 582 const struct in6_addr *v6dst) 583 { 584 int prefix_len; 585 586 #ifdef CONFIG_IPV6_SIT_6RD 587 prefix_len = tunnel->ip6rd.prefixlen + 32 588 - tunnel->ip6rd.relay_prefixlen; 589 #else 590 prefix_len = 48; 591 #endif 592 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev); 593 } 594 595 /* Returns true if a packet is spoofed */ 596 static bool packet_is_spoofed(struct sk_buff *skb, 597 const struct iphdr *iph, 598 struct ip_tunnel *tunnel) 599 { 600 const struct ipv6hdr *ipv6h; 601 602 if (tunnel->dev->priv_flags & IFF_ISATAP) { 603 if (!isatap_chksrc(skb, iph, tunnel)) 604 return true; 605 606 return false; 607 } 608 609 if (tunnel->dev->flags & IFF_POINTOPOINT) 610 return false; 611 612 ipv6h = ipv6_hdr(skb); 613 614 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) { 615 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 616 &iph->saddr, &ipv6h->saddr, 617 &iph->daddr, &ipv6h->daddr); 618 return true; 619 } 620 621 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr))) 622 return false; 623 624 if (only_dnatted(tunnel, &ipv6h->daddr)) 625 return false; 626 627 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 628 &iph->saddr, &ipv6h->saddr, 629 &iph->daddr, &ipv6h->daddr); 630 return true; 631 } 632 633 static int ipip6_rcv(struct sk_buff *skb) 634 { 635 const struct iphdr *iph = ip_hdr(skb); 636 struct ip_tunnel *tunnel; 637 int err; 638 639 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 640 iph->saddr, iph->daddr); 641 if (tunnel != NULL) { 642 struct pcpu_tstats *tstats; 643 644 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 && 645 tunnel->parms.iph.protocol != 0) 646 goto out; 647 648 skb->mac_header = skb->network_header; 649 skb_reset_network_header(skb); 650 IPCB(skb)->flags = 0; 651 skb->protocol = htons(ETH_P_IPV6); 652 653 if (packet_is_spoofed(skb, iph, tunnel)) { 654 tunnel->dev->stats.rx_errors++; 655 goto out; 656 } 657 658 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 659 660 err = IP_ECN_decapsulate(iph, skb); 661 if (unlikely(err)) { 662 if (log_ecn_error) 663 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 664 &iph->saddr, iph->tos); 665 if (err > 1) { 666 ++tunnel->dev->stats.rx_frame_errors; 667 ++tunnel->dev->stats.rx_errors; 668 goto out; 669 } 670 } 671 672 tstats = this_cpu_ptr(tunnel->dev->tstats); 673 tstats->rx_packets++; 674 tstats->rx_bytes += skb->len; 675 676 netif_rx(skb); 677 678 return 0; 679 } 680 681 /* no tunnel matched, let upstream know, ipsec may handle it */ 682 return 1; 683 out: 684 kfree_skb(skb); 685 return 0; 686 } 687 688 static const struct tnl_ptk_info tpi = { 689 /* no tunnel info required for ipip. */ 690 .proto = htons(ETH_P_IP), 691 }; 692 693 static int ipip_rcv(struct sk_buff *skb) 694 { 695 const struct iphdr *iph; 696 struct ip_tunnel *tunnel; 697 698 iph = ip_hdr(skb); 699 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 700 iph->saddr, iph->daddr); 701 if (tunnel != NULL) { 702 if (tunnel->parms.iph.protocol != IPPROTO_IPIP && 703 tunnel->parms.iph.protocol != 0) 704 goto drop; 705 706 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) 707 goto drop; 708 if (iptunnel_pull_header(skb, 0, tpi.proto)) 709 goto drop; 710 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error); 711 } 712 713 return 1; 714 715 drop: 716 kfree_skb(skb); 717 return 0; 718 } 719 720 /* 721 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function 722 * stores the embedded IPv4 address in v4dst and returns true. 723 */ 724 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, 725 __be32 *v4dst) 726 { 727 #ifdef CONFIG_IPV6_SIT_6RD 728 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, 729 tunnel->ip6rd.prefixlen)) { 730 unsigned int pbw0, pbi0; 731 int pbi1; 732 u32 d; 733 734 pbw0 = tunnel->ip6rd.prefixlen >> 5; 735 pbi0 = tunnel->ip6rd.prefixlen & 0x1f; 736 737 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> 738 tunnel->ip6rd.relay_prefixlen; 739 740 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; 741 if (pbi1 > 0) 742 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> 743 (32 - pbi1); 744 745 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d); 746 return true; 747 } 748 #else 749 if (v6dst->s6_addr16[0] == htons(0x2002)) { 750 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ 751 memcpy(v4dst, &v6dst->s6_addr16[1], 4); 752 return true; 753 } 754 #endif 755 return false; 756 } 757 758 static inline __be32 try_6rd(struct ip_tunnel *tunnel, 759 const struct in6_addr *v6dst) 760 { 761 __be32 dst = 0; 762 check_6rd(tunnel, v6dst, &dst); 763 return dst; 764 } 765 766 /* 767 * This function assumes it is being called from dev_queue_xmit() 768 * and that skb is filled properly by that function. 769 */ 770 771 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, 772 struct net_device *dev) 773 { 774 struct ip_tunnel *tunnel = netdev_priv(dev); 775 const struct iphdr *tiph = &tunnel->parms.iph; 776 const struct ipv6hdr *iph6 = ipv6_hdr(skb); 777 u8 tos = tunnel->parms.iph.tos; 778 __be16 df = tiph->frag_off; 779 struct rtable *rt; /* Route to the other host */ 780 struct net_device *tdev; /* Device to other host */ 781 unsigned int max_headroom; /* The extra header space needed */ 782 __be32 dst = tiph->daddr; 783 struct flowi4 fl4; 784 int mtu; 785 const struct in6_addr *addr6; 786 int addr_type; 787 u8 ttl; 788 int err; 789 790 if (skb->protocol != htons(ETH_P_IPV6)) 791 goto tx_error; 792 793 if (tos == 1) 794 tos = ipv6_get_dsfield(iph6); 795 796 /* ISATAP (RFC4214) - must come before 6to4 */ 797 if (dev->priv_flags & IFF_ISATAP) { 798 struct neighbour *neigh = NULL; 799 bool do_tx_error = false; 800 801 if (skb_dst(skb)) 802 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 803 804 if (neigh == NULL) { 805 net_dbg_ratelimited("nexthop == NULL\n"); 806 goto tx_error; 807 } 808 809 addr6 = (const struct in6_addr *)&neigh->primary_key; 810 addr_type = ipv6_addr_type(addr6); 811 812 if ((addr_type & IPV6_ADDR_UNICAST) && 813 ipv6_addr_is_isatap(addr6)) 814 dst = addr6->s6_addr32[3]; 815 else 816 do_tx_error = true; 817 818 neigh_release(neigh); 819 if (do_tx_error) 820 goto tx_error; 821 } 822 823 if (!dst) 824 dst = try_6rd(tunnel, &iph6->daddr); 825 826 if (!dst) { 827 struct neighbour *neigh = NULL; 828 bool do_tx_error = false; 829 830 if (skb_dst(skb)) 831 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 832 833 if (neigh == NULL) { 834 net_dbg_ratelimited("nexthop == NULL\n"); 835 goto tx_error; 836 } 837 838 addr6 = (const struct in6_addr *)&neigh->primary_key; 839 addr_type = ipv6_addr_type(addr6); 840 841 if (addr_type == IPV6_ADDR_ANY) { 842 addr6 = &ipv6_hdr(skb)->daddr; 843 addr_type = ipv6_addr_type(addr6); 844 } 845 846 if ((addr_type & IPV6_ADDR_COMPATv4) != 0) 847 dst = addr6->s6_addr32[3]; 848 else 849 do_tx_error = true; 850 851 neigh_release(neigh); 852 if (do_tx_error) 853 goto tx_error; 854 } 855 856 rt = ip_route_output_ports(tunnel->net, &fl4, NULL, 857 dst, tiph->saddr, 858 0, 0, 859 IPPROTO_IPV6, RT_TOS(tos), 860 tunnel->parms.link); 861 if (IS_ERR(rt)) { 862 dev->stats.tx_carrier_errors++; 863 goto tx_error_icmp; 864 } 865 if (rt->rt_type != RTN_UNICAST) { 866 ip_rt_put(rt); 867 dev->stats.tx_carrier_errors++; 868 goto tx_error_icmp; 869 } 870 tdev = rt->dst.dev; 871 872 if (tdev == dev) { 873 ip_rt_put(rt); 874 dev->stats.collisions++; 875 goto tx_error; 876 } 877 878 if (df) { 879 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr); 880 881 if (mtu < 68) { 882 dev->stats.collisions++; 883 ip_rt_put(rt); 884 goto tx_error; 885 } 886 887 if (mtu < IPV6_MIN_MTU) { 888 mtu = IPV6_MIN_MTU; 889 df = 0; 890 } 891 892 if (tunnel->parms.iph.daddr && skb_dst(skb)) 893 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 894 895 if (skb->len > mtu) { 896 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 897 ip_rt_put(rt); 898 goto tx_error; 899 } 900 } 901 902 if (tunnel->err_count > 0) { 903 if (time_before(jiffies, 904 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 905 tunnel->err_count--; 906 dst_link_failure(skb); 907 } else 908 tunnel->err_count = 0; 909 } 910 911 /* 912 * Okay, now see if we can stuff it in the buffer as-is. 913 */ 914 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr); 915 916 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 917 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 918 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 919 if (!new_skb) { 920 ip_rt_put(rt); 921 dev->stats.tx_dropped++; 922 dev_kfree_skb(skb); 923 return NETDEV_TX_OK; 924 } 925 if (skb->sk) 926 skb_set_owner_w(new_skb, skb->sk); 927 dev_kfree_skb(skb); 928 skb = new_skb; 929 iph6 = ipv6_hdr(skb); 930 } 931 ttl = tiph->ttl; 932 if (ttl == 0) 933 ttl = iph6->hop_limit; 934 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); 935 936 if (likely(!skb->encapsulation)) { 937 skb_reset_inner_headers(skb); 938 skb->encapsulation = 1; 939 } 940 941 err = iptunnel_xmit(rt, skb, fl4.saddr, fl4.daddr, IPPROTO_IPV6, tos, 942 ttl, df, !net_eq(tunnel->net, dev_net(dev))); 943 iptunnel_xmit_stats(err, &dev->stats, dev->tstats); 944 return NETDEV_TX_OK; 945 946 tx_error_icmp: 947 dst_link_failure(skb); 948 tx_error: 949 dev->stats.tx_errors++; 950 dev_kfree_skb(skb); 951 return NETDEV_TX_OK; 952 } 953 954 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) 955 { 956 struct ip_tunnel *tunnel = netdev_priv(dev); 957 const struct iphdr *tiph = &tunnel->parms.iph; 958 959 if (likely(!skb->encapsulation)) { 960 skb_reset_inner_headers(skb); 961 skb->encapsulation = 1; 962 } 963 964 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP); 965 return NETDEV_TX_OK; 966 } 967 968 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb, 969 struct net_device *dev) 970 { 971 switch (skb->protocol) { 972 case htons(ETH_P_IP): 973 ipip_tunnel_xmit(skb, dev); 974 break; 975 case htons(ETH_P_IPV6): 976 ipip6_tunnel_xmit(skb, dev); 977 break; 978 default: 979 goto tx_err; 980 } 981 982 return NETDEV_TX_OK; 983 984 tx_err: 985 dev->stats.tx_errors++; 986 dev_kfree_skb(skb); 987 return NETDEV_TX_OK; 988 989 } 990 991 static void ipip6_tunnel_bind_dev(struct net_device *dev) 992 { 993 struct net_device *tdev = NULL; 994 struct ip_tunnel *tunnel; 995 const struct iphdr *iph; 996 struct flowi4 fl4; 997 998 tunnel = netdev_priv(dev); 999 iph = &tunnel->parms.iph; 1000 1001 if (iph->daddr) { 1002 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4, 1003 NULL, 1004 iph->daddr, iph->saddr, 1005 0, 0, 1006 IPPROTO_IPV6, 1007 RT_TOS(iph->tos), 1008 tunnel->parms.link); 1009 1010 if (!IS_ERR(rt)) { 1011 tdev = rt->dst.dev; 1012 ip_rt_put(rt); 1013 } 1014 dev->flags |= IFF_POINTOPOINT; 1015 } 1016 1017 if (!tdev && tunnel->parms.link) 1018 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link); 1019 1020 if (tdev) { 1021 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); 1022 dev->mtu = tdev->mtu - sizeof(struct iphdr); 1023 if (dev->mtu < IPV6_MIN_MTU) 1024 dev->mtu = IPV6_MIN_MTU; 1025 } 1026 dev->iflink = tunnel->parms.link; 1027 } 1028 1029 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p) 1030 { 1031 struct net *net = t->net; 1032 struct sit_net *sitn = net_generic(net, sit_net_id); 1033 1034 ipip6_tunnel_unlink(sitn, t); 1035 synchronize_net(); 1036 t->parms.iph.saddr = p->iph.saddr; 1037 t->parms.iph.daddr = p->iph.daddr; 1038 memcpy(t->dev->dev_addr, &p->iph.saddr, 4); 1039 memcpy(t->dev->broadcast, &p->iph.daddr, 4); 1040 ipip6_tunnel_link(sitn, t); 1041 t->parms.iph.ttl = p->iph.ttl; 1042 t->parms.iph.tos = p->iph.tos; 1043 if (t->parms.link != p->link) { 1044 t->parms.link = p->link; 1045 ipip6_tunnel_bind_dev(t->dev); 1046 } 1047 netdev_state_change(t->dev); 1048 } 1049 1050 #ifdef CONFIG_IPV6_SIT_6RD 1051 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t, 1052 struct ip_tunnel_6rd *ip6rd) 1053 { 1054 struct in6_addr prefix; 1055 __be32 relay_prefix; 1056 1057 if (ip6rd->relay_prefixlen > 32 || 1058 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64) 1059 return -EINVAL; 1060 1061 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen); 1062 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix)) 1063 return -EINVAL; 1064 if (ip6rd->relay_prefixlen) 1065 relay_prefix = ip6rd->relay_prefix & 1066 htonl(0xffffffffUL << 1067 (32 - ip6rd->relay_prefixlen)); 1068 else 1069 relay_prefix = 0; 1070 if (relay_prefix != ip6rd->relay_prefix) 1071 return -EINVAL; 1072 1073 t->ip6rd.prefix = prefix; 1074 t->ip6rd.relay_prefix = relay_prefix; 1075 t->ip6rd.prefixlen = ip6rd->prefixlen; 1076 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen; 1077 netdev_state_change(t->dev); 1078 return 0; 1079 } 1080 #endif 1081 1082 static int 1083 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) 1084 { 1085 int err = 0; 1086 struct ip_tunnel_parm p; 1087 struct ip_tunnel_prl prl; 1088 struct ip_tunnel *t; 1089 struct net *net = dev_net(dev); 1090 struct sit_net *sitn = net_generic(net, sit_net_id); 1091 #ifdef CONFIG_IPV6_SIT_6RD 1092 struct ip_tunnel_6rd ip6rd; 1093 #endif 1094 1095 switch (cmd) { 1096 case SIOCGETTUNNEL: 1097 #ifdef CONFIG_IPV6_SIT_6RD 1098 case SIOCGET6RD: 1099 #endif 1100 t = NULL; 1101 if (dev == sitn->fb_tunnel_dev) { 1102 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1103 err = -EFAULT; 1104 break; 1105 } 1106 t = ipip6_tunnel_locate(net, &p, 0); 1107 } 1108 if (t == NULL) 1109 t = netdev_priv(dev); 1110 1111 err = -EFAULT; 1112 if (cmd == SIOCGETTUNNEL) { 1113 memcpy(&p, &t->parms, sizeof(p)); 1114 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, 1115 sizeof(p))) 1116 goto done; 1117 #ifdef CONFIG_IPV6_SIT_6RD 1118 } else { 1119 ip6rd.prefix = t->ip6rd.prefix; 1120 ip6rd.relay_prefix = t->ip6rd.relay_prefix; 1121 ip6rd.prefixlen = t->ip6rd.prefixlen; 1122 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; 1123 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, 1124 sizeof(ip6rd))) 1125 goto done; 1126 #endif 1127 } 1128 err = 0; 1129 break; 1130 1131 case SIOCADDTUNNEL: 1132 case SIOCCHGTUNNEL: 1133 err = -EPERM; 1134 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1135 goto done; 1136 1137 err = -EFAULT; 1138 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1139 goto done; 1140 1141 err = -EINVAL; 1142 if (p.iph.protocol != IPPROTO_IPV6 && 1143 p.iph.protocol != IPPROTO_IPIP && 1144 p.iph.protocol != 0) 1145 goto done; 1146 if (p.iph.version != 4 || 1147 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) 1148 goto done; 1149 if (p.iph.ttl) 1150 p.iph.frag_off |= htons(IP_DF); 1151 1152 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); 1153 1154 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1155 if (t != NULL) { 1156 if (t->dev != dev) { 1157 err = -EEXIST; 1158 break; 1159 } 1160 } else { 1161 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || 1162 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { 1163 err = -EINVAL; 1164 break; 1165 } 1166 t = netdev_priv(dev); 1167 } 1168 1169 ipip6_tunnel_update(t, &p); 1170 } 1171 1172 if (t) { 1173 err = 0; 1174 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) 1175 err = -EFAULT; 1176 } else 1177 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1178 break; 1179 1180 case SIOCDELTUNNEL: 1181 err = -EPERM; 1182 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1183 goto done; 1184 1185 if (dev == sitn->fb_tunnel_dev) { 1186 err = -EFAULT; 1187 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1188 goto done; 1189 err = -ENOENT; 1190 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL) 1191 goto done; 1192 err = -EPERM; 1193 if (t == netdev_priv(sitn->fb_tunnel_dev)) 1194 goto done; 1195 dev = t->dev; 1196 } 1197 unregister_netdevice(dev); 1198 err = 0; 1199 break; 1200 1201 case SIOCGETPRL: 1202 err = -EINVAL; 1203 if (dev == sitn->fb_tunnel_dev) 1204 goto done; 1205 err = -ENOENT; 1206 if (!(t = netdev_priv(dev))) 1207 goto done; 1208 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data); 1209 break; 1210 1211 case SIOCADDPRL: 1212 case SIOCDELPRL: 1213 case SIOCCHGPRL: 1214 err = -EPERM; 1215 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1216 goto done; 1217 err = -EINVAL; 1218 if (dev == sitn->fb_tunnel_dev) 1219 goto done; 1220 err = -EFAULT; 1221 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl))) 1222 goto done; 1223 err = -ENOENT; 1224 if (!(t = netdev_priv(dev))) 1225 goto done; 1226 1227 switch (cmd) { 1228 case SIOCDELPRL: 1229 err = ipip6_tunnel_del_prl(t, &prl); 1230 break; 1231 case SIOCADDPRL: 1232 case SIOCCHGPRL: 1233 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); 1234 break; 1235 } 1236 netdev_state_change(dev); 1237 break; 1238 1239 #ifdef CONFIG_IPV6_SIT_6RD 1240 case SIOCADD6RD: 1241 case SIOCCHG6RD: 1242 case SIOCDEL6RD: 1243 err = -EPERM; 1244 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1245 goto done; 1246 1247 err = -EFAULT; 1248 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, 1249 sizeof(ip6rd))) 1250 goto done; 1251 1252 t = netdev_priv(dev); 1253 1254 if (cmd != SIOCDEL6RD) { 1255 err = ipip6_tunnel_update_6rd(t, &ip6rd); 1256 if (err < 0) 1257 goto done; 1258 } else 1259 ipip6_tunnel_clone_6rd(dev, sitn); 1260 1261 err = 0; 1262 break; 1263 #endif 1264 1265 default: 1266 err = -EINVAL; 1267 } 1268 1269 done: 1270 return err; 1271 } 1272 1273 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu) 1274 { 1275 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr)) 1276 return -EINVAL; 1277 dev->mtu = new_mtu; 1278 return 0; 1279 } 1280 1281 static const struct net_device_ops ipip6_netdev_ops = { 1282 .ndo_uninit = ipip6_tunnel_uninit, 1283 .ndo_start_xmit = sit_tunnel_xmit, 1284 .ndo_do_ioctl = ipip6_tunnel_ioctl, 1285 .ndo_change_mtu = ipip6_tunnel_change_mtu, 1286 .ndo_get_stats64 = ip_tunnel_get_stats64, 1287 }; 1288 1289 static void ipip6_dev_free(struct net_device *dev) 1290 { 1291 free_percpu(dev->tstats); 1292 free_netdev(dev); 1293 } 1294 1295 static void ipip6_tunnel_setup(struct net_device *dev) 1296 { 1297 dev->netdev_ops = &ipip6_netdev_ops; 1298 dev->destructor = ipip6_dev_free; 1299 1300 dev->type = ARPHRD_SIT; 1301 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr); 1302 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr); 1303 dev->flags = IFF_NOARP; 1304 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; 1305 dev->iflink = 0; 1306 dev->addr_len = 4; 1307 dev->features |= NETIF_F_LLTX; 1308 } 1309 1310 static int ipip6_tunnel_init(struct net_device *dev) 1311 { 1312 struct ip_tunnel *tunnel = netdev_priv(dev); 1313 1314 tunnel->dev = dev; 1315 tunnel->net = dev_net(dev); 1316 1317 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); 1318 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); 1319 1320 ipip6_tunnel_bind_dev(dev); 1321 dev->tstats = alloc_percpu(struct pcpu_tstats); 1322 if (!dev->tstats) 1323 return -ENOMEM; 1324 1325 return 0; 1326 } 1327 1328 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev) 1329 { 1330 struct ip_tunnel *tunnel = netdev_priv(dev); 1331 struct iphdr *iph = &tunnel->parms.iph; 1332 struct net *net = dev_net(dev); 1333 struct sit_net *sitn = net_generic(net, sit_net_id); 1334 1335 tunnel->dev = dev; 1336 tunnel->net = dev_net(dev); 1337 strcpy(tunnel->parms.name, dev->name); 1338 1339 iph->version = 4; 1340 iph->protocol = IPPROTO_IPV6; 1341 iph->ihl = 5; 1342 iph->ttl = 64; 1343 1344 dev->tstats = alloc_percpu(struct pcpu_tstats); 1345 if (!dev->tstats) 1346 return -ENOMEM; 1347 dev_hold(dev); 1348 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); 1349 return 0; 1350 } 1351 1352 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[]) 1353 { 1354 u8 proto; 1355 1356 if (!data || !data[IFLA_IPTUN_PROTO]) 1357 return 0; 1358 1359 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1360 if (proto != IPPROTO_IPV6 && 1361 proto != IPPROTO_IPIP && 1362 proto != 0) 1363 return -EINVAL; 1364 1365 return 0; 1366 } 1367 1368 static void ipip6_netlink_parms(struct nlattr *data[], 1369 struct ip_tunnel_parm *parms) 1370 { 1371 memset(parms, 0, sizeof(*parms)); 1372 1373 parms->iph.version = 4; 1374 parms->iph.protocol = IPPROTO_IPV6; 1375 parms->iph.ihl = 5; 1376 parms->iph.ttl = 64; 1377 1378 if (!data) 1379 return; 1380 1381 if (data[IFLA_IPTUN_LINK]) 1382 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1383 1384 if (data[IFLA_IPTUN_LOCAL]) 1385 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]); 1386 1387 if (data[IFLA_IPTUN_REMOTE]) 1388 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]); 1389 1390 if (data[IFLA_IPTUN_TTL]) { 1391 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]); 1392 if (parms->iph.ttl) 1393 parms->iph.frag_off = htons(IP_DF); 1394 } 1395 1396 if (data[IFLA_IPTUN_TOS]) 1397 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]); 1398 1399 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC])) 1400 parms->iph.frag_off = htons(IP_DF); 1401 1402 if (data[IFLA_IPTUN_FLAGS]) 1403 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]); 1404 1405 if (data[IFLA_IPTUN_PROTO]) 1406 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1407 1408 } 1409 1410 #ifdef CONFIG_IPV6_SIT_6RD 1411 /* This function returns true when 6RD attributes are present in the nl msg */ 1412 static bool ipip6_netlink_6rd_parms(struct nlattr *data[], 1413 struct ip_tunnel_6rd *ip6rd) 1414 { 1415 bool ret = false; 1416 memset(ip6rd, 0, sizeof(*ip6rd)); 1417 1418 if (!data) 1419 return ret; 1420 1421 if (data[IFLA_IPTUN_6RD_PREFIX]) { 1422 ret = true; 1423 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX], 1424 sizeof(struct in6_addr)); 1425 } 1426 1427 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) { 1428 ret = true; 1429 ip6rd->relay_prefix = 1430 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]); 1431 } 1432 1433 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) { 1434 ret = true; 1435 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]); 1436 } 1437 1438 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) { 1439 ret = true; 1440 ip6rd->relay_prefixlen = 1441 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]); 1442 } 1443 1444 return ret; 1445 } 1446 #endif 1447 1448 static int ipip6_newlink(struct net *src_net, struct net_device *dev, 1449 struct nlattr *tb[], struct nlattr *data[]) 1450 { 1451 struct net *net = dev_net(dev); 1452 struct ip_tunnel *nt; 1453 #ifdef CONFIG_IPV6_SIT_6RD 1454 struct ip_tunnel_6rd ip6rd; 1455 #endif 1456 int err; 1457 1458 nt = netdev_priv(dev); 1459 ipip6_netlink_parms(data, &nt->parms); 1460 1461 if (ipip6_tunnel_locate(net, &nt->parms, 0)) 1462 return -EEXIST; 1463 1464 err = ipip6_tunnel_create(dev); 1465 if (err < 0) 1466 return err; 1467 1468 #ifdef CONFIG_IPV6_SIT_6RD 1469 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1470 err = ipip6_tunnel_update_6rd(nt, &ip6rd); 1471 #endif 1472 1473 return err; 1474 } 1475 1476 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[], 1477 struct nlattr *data[]) 1478 { 1479 struct ip_tunnel *t = netdev_priv(dev); 1480 struct ip_tunnel_parm p; 1481 struct net *net = t->net; 1482 struct sit_net *sitn = net_generic(net, sit_net_id); 1483 #ifdef CONFIG_IPV6_SIT_6RD 1484 struct ip_tunnel_6rd ip6rd; 1485 #endif 1486 1487 if (dev == sitn->fb_tunnel_dev) 1488 return -EINVAL; 1489 1490 ipip6_netlink_parms(data, &p); 1491 1492 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) || 1493 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr)) 1494 return -EINVAL; 1495 1496 t = ipip6_tunnel_locate(net, &p, 0); 1497 1498 if (t) { 1499 if (t->dev != dev) 1500 return -EEXIST; 1501 } else 1502 t = netdev_priv(dev); 1503 1504 ipip6_tunnel_update(t, &p); 1505 1506 #ifdef CONFIG_IPV6_SIT_6RD 1507 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1508 return ipip6_tunnel_update_6rd(t, &ip6rd); 1509 #endif 1510 1511 return 0; 1512 } 1513 1514 static size_t ipip6_get_size(const struct net_device *dev) 1515 { 1516 return 1517 /* IFLA_IPTUN_LINK */ 1518 nla_total_size(4) + 1519 /* IFLA_IPTUN_LOCAL */ 1520 nla_total_size(4) + 1521 /* IFLA_IPTUN_REMOTE */ 1522 nla_total_size(4) + 1523 /* IFLA_IPTUN_TTL */ 1524 nla_total_size(1) + 1525 /* IFLA_IPTUN_TOS */ 1526 nla_total_size(1) + 1527 /* IFLA_IPTUN_PMTUDISC */ 1528 nla_total_size(1) + 1529 /* IFLA_IPTUN_FLAGS */ 1530 nla_total_size(2) + 1531 /* IFLA_IPTUN_PROTO */ 1532 nla_total_size(1) + 1533 #ifdef CONFIG_IPV6_SIT_6RD 1534 /* IFLA_IPTUN_6RD_PREFIX */ 1535 nla_total_size(sizeof(struct in6_addr)) + 1536 /* IFLA_IPTUN_6RD_RELAY_PREFIX */ 1537 nla_total_size(4) + 1538 /* IFLA_IPTUN_6RD_PREFIXLEN */ 1539 nla_total_size(2) + 1540 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */ 1541 nla_total_size(2) + 1542 #endif 1543 0; 1544 } 1545 1546 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1547 { 1548 struct ip_tunnel *tunnel = netdev_priv(dev); 1549 struct ip_tunnel_parm *parm = &tunnel->parms; 1550 1551 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1552 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) || 1553 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) || 1554 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) || 1555 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) || 1556 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC, 1557 !!(parm->iph.frag_off & htons(IP_DF))) || 1558 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) || 1559 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags)) 1560 goto nla_put_failure; 1561 1562 #ifdef CONFIG_IPV6_SIT_6RD 1563 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr), 1564 &tunnel->ip6rd.prefix) || 1565 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX, 1566 tunnel->ip6rd.relay_prefix) || 1567 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN, 1568 tunnel->ip6rd.prefixlen) || 1569 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN, 1570 tunnel->ip6rd.relay_prefixlen)) 1571 goto nla_put_failure; 1572 #endif 1573 1574 return 0; 1575 1576 nla_put_failure: 1577 return -EMSGSIZE; 1578 } 1579 1580 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = { 1581 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1582 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 }, 1583 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 }, 1584 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1585 [IFLA_IPTUN_TOS] = { .type = NLA_U8 }, 1586 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 }, 1587 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 }, 1588 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1589 #ifdef CONFIG_IPV6_SIT_6RD 1590 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) }, 1591 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 }, 1592 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 }, 1593 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 }, 1594 #endif 1595 }; 1596 1597 static struct rtnl_link_ops sit_link_ops __read_mostly = { 1598 .kind = "sit", 1599 .maxtype = IFLA_IPTUN_MAX, 1600 .policy = ipip6_policy, 1601 .priv_size = sizeof(struct ip_tunnel), 1602 .setup = ipip6_tunnel_setup, 1603 .validate = ipip6_validate, 1604 .newlink = ipip6_newlink, 1605 .changelink = ipip6_changelink, 1606 .get_size = ipip6_get_size, 1607 .fill_info = ipip6_fill_info, 1608 }; 1609 1610 static struct xfrm_tunnel sit_handler __read_mostly = { 1611 .handler = ipip6_rcv, 1612 .err_handler = ipip6_err, 1613 .priority = 1, 1614 }; 1615 1616 static struct xfrm_tunnel ipip_handler __read_mostly = { 1617 .handler = ipip_rcv, 1618 .err_handler = ipip6_err, 1619 .priority = 2, 1620 }; 1621 1622 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head) 1623 { 1624 struct net *net = dev_net(sitn->fb_tunnel_dev); 1625 struct net_device *dev, *aux; 1626 int prio; 1627 1628 for_each_netdev_safe(net, dev, aux) 1629 if (dev->rtnl_link_ops == &sit_link_ops) 1630 unregister_netdevice_queue(dev, head); 1631 1632 for (prio = 1; prio < 4; prio++) { 1633 int h; 1634 for (h = 0; h < HASH_SIZE; h++) { 1635 struct ip_tunnel *t; 1636 1637 t = rtnl_dereference(sitn->tunnels[prio][h]); 1638 while (t != NULL) { 1639 /* If dev is in the same netns, it has already 1640 * been added to the list by the previous loop. 1641 */ 1642 if (!net_eq(dev_net(t->dev), net)) 1643 unregister_netdevice_queue(t->dev, 1644 head); 1645 t = rtnl_dereference(t->next); 1646 } 1647 } 1648 } 1649 } 1650 1651 static int __net_init sit_init_net(struct net *net) 1652 { 1653 struct sit_net *sitn = net_generic(net, sit_net_id); 1654 struct ip_tunnel *t; 1655 int err; 1656 1657 sitn->tunnels[0] = sitn->tunnels_wc; 1658 sitn->tunnels[1] = sitn->tunnels_l; 1659 sitn->tunnels[2] = sitn->tunnels_r; 1660 sitn->tunnels[3] = sitn->tunnels_r_l; 1661 1662 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", 1663 ipip6_tunnel_setup); 1664 if (!sitn->fb_tunnel_dev) { 1665 err = -ENOMEM; 1666 goto err_alloc_dev; 1667 } 1668 dev_net_set(sitn->fb_tunnel_dev, net); 1669 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops; 1670 /* FB netdevice is special: we have one, and only one per netns. 1671 * Allowing to move it to another netns is clearly unsafe. 1672 */ 1673 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1674 1675 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); 1676 if (err) 1677 goto err_dev_free; 1678 1679 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); 1680 1681 if ((err = register_netdev(sitn->fb_tunnel_dev))) 1682 goto err_reg_dev; 1683 1684 t = netdev_priv(sitn->fb_tunnel_dev); 1685 1686 strcpy(t->parms.name, sitn->fb_tunnel_dev->name); 1687 return 0; 1688 1689 err_reg_dev: 1690 dev_put(sitn->fb_tunnel_dev); 1691 err_dev_free: 1692 ipip6_dev_free(sitn->fb_tunnel_dev); 1693 err_alloc_dev: 1694 return err; 1695 } 1696 1697 static void __net_exit sit_exit_net(struct net *net) 1698 { 1699 struct sit_net *sitn = net_generic(net, sit_net_id); 1700 LIST_HEAD(list); 1701 1702 rtnl_lock(); 1703 sit_destroy_tunnels(sitn, &list); 1704 unregister_netdevice_many(&list); 1705 rtnl_unlock(); 1706 } 1707 1708 static struct pernet_operations sit_net_ops = { 1709 .init = sit_init_net, 1710 .exit = sit_exit_net, 1711 .id = &sit_net_id, 1712 .size = sizeof(struct sit_net), 1713 }; 1714 1715 static void __exit sit_cleanup(void) 1716 { 1717 rtnl_link_unregister(&sit_link_ops); 1718 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1719 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1720 1721 unregister_pernet_device(&sit_net_ops); 1722 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1723 } 1724 1725 static int __init sit_init(void) 1726 { 1727 int err; 1728 1729 pr_info("IPv6 over IPv4 tunneling driver\n"); 1730 1731 err = register_pernet_device(&sit_net_ops); 1732 if (err < 0) 1733 return err; 1734 err = xfrm4_tunnel_register(&sit_handler, AF_INET6); 1735 if (err < 0) { 1736 pr_info("%s: can't register ip6ip4\n", __func__); 1737 goto xfrm_tunnel_failed; 1738 } 1739 err = xfrm4_tunnel_register(&ipip_handler, AF_INET); 1740 if (err < 0) { 1741 pr_info("%s: can't register ip4ip4\n", __func__); 1742 goto xfrm_tunnel4_failed; 1743 } 1744 err = rtnl_link_register(&sit_link_ops); 1745 if (err < 0) 1746 goto rtnl_link_failed; 1747 1748 out: 1749 return err; 1750 1751 rtnl_link_failed: 1752 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1753 xfrm_tunnel4_failed: 1754 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1755 xfrm_tunnel_failed: 1756 unregister_pernet_device(&sit_net_ops); 1757 goto out; 1758 } 1759 1760 module_init(sit_init); 1761 module_exit(sit_cleanup); 1762 MODULE_LICENSE("GPL"); 1763 MODULE_ALIAS_NETDEV("sit0"); 1764