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->ifindex == 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->ifindex == 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->ifindex == 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, NET_NAME_UNKNOWN, 254 ipip6_tunnel_setup); 255 if (dev == NULL) 256 return NULL; 257 258 dev_net_set(dev, net); 259 260 nt = netdev_priv(dev); 261 262 nt->parms = *parms; 263 if (ipip6_tunnel_create(dev) < 0) 264 goto failed_free; 265 266 return nt; 267 268 failed_free: 269 ipip6_dev_free(dev); 270 failed: 271 return NULL; 272 } 273 274 #define for_each_prl_rcu(start) \ 275 for (prl = rcu_dereference(start); \ 276 prl; \ 277 prl = rcu_dereference(prl->next)) 278 279 static struct ip_tunnel_prl_entry * 280 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr) 281 { 282 struct ip_tunnel_prl_entry *prl; 283 284 for_each_prl_rcu(t->prl) 285 if (prl->addr == addr) 286 break; 287 return prl; 288 289 } 290 291 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, 292 struct ip_tunnel_prl __user *a) 293 { 294 struct ip_tunnel_prl kprl, *kp; 295 struct ip_tunnel_prl_entry *prl; 296 unsigned int cmax, c = 0, ca, len; 297 int ret = 0; 298 299 if (copy_from_user(&kprl, a, sizeof(kprl))) 300 return -EFAULT; 301 cmax = kprl.datalen / sizeof(kprl); 302 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY)) 303 cmax = 1; 304 305 /* For simple GET or for root users, 306 * we try harder to allocate. 307 */ 308 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ? 309 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) : 310 NULL; 311 312 rcu_read_lock(); 313 314 ca = t->prl_count < cmax ? t->prl_count : cmax; 315 316 if (!kp) { 317 /* We don't try hard to allocate much memory for 318 * non-root users. 319 * For root users, retry allocating enough memory for 320 * the answer. 321 */ 322 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC); 323 if (!kp) { 324 ret = -ENOMEM; 325 goto out; 326 } 327 } 328 329 c = 0; 330 for_each_prl_rcu(t->prl) { 331 if (c >= cmax) 332 break; 333 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr) 334 continue; 335 kp[c].addr = prl->addr; 336 kp[c].flags = prl->flags; 337 c++; 338 if (kprl.addr != htonl(INADDR_ANY)) 339 break; 340 } 341 out: 342 rcu_read_unlock(); 343 344 len = sizeof(*kp) * c; 345 ret = 0; 346 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen)) 347 ret = -EFAULT; 348 349 kfree(kp); 350 351 return ret; 352 } 353 354 static int 355 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) 356 { 357 struct ip_tunnel_prl_entry *p; 358 int err = 0; 359 360 if (a->addr == htonl(INADDR_ANY)) 361 return -EINVAL; 362 363 ASSERT_RTNL(); 364 365 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) { 366 if (p->addr == a->addr) { 367 if (chg) { 368 p->flags = a->flags; 369 goto out; 370 } 371 err = -EEXIST; 372 goto out; 373 } 374 } 375 376 if (chg) { 377 err = -ENXIO; 378 goto out; 379 } 380 381 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL); 382 if (!p) { 383 err = -ENOBUFS; 384 goto out; 385 } 386 387 p->next = t->prl; 388 p->addr = a->addr; 389 p->flags = a->flags; 390 t->prl_count++; 391 rcu_assign_pointer(t->prl, p); 392 out: 393 return err; 394 } 395 396 static void prl_list_destroy_rcu(struct rcu_head *head) 397 { 398 struct ip_tunnel_prl_entry *p, *n; 399 400 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head); 401 do { 402 n = rcu_dereference_protected(p->next, 1); 403 kfree(p); 404 p = n; 405 } while (p); 406 } 407 408 static int 409 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) 410 { 411 struct ip_tunnel_prl_entry *x; 412 struct ip_tunnel_prl_entry __rcu **p; 413 int err = 0; 414 415 ASSERT_RTNL(); 416 417 if (a && a->addr != htonl(INADDR_ANY)) { 418 for (p = &t->prl; 419 (x = rtnl_dereference(*p)) != NULL; 420 p = &x->next) { 421 if (x->addr == a->addr) { 422 *p = x->next; 423 kfree_rcu(x, rcu_head); 424 t->prl_count--; 425 goto out; 426 } 427 } 428 err = -ENXIO; 429 } else { 430 x = rtnl_dereference(t->prl); 431 if (x) { 432 t->prl_count = 0; 433 call_rcu(&x->rcu_head, prl_list_destroy_rcu); 434 t->prl = NULL; 435 } 436 } 437 out: 438 return err; 439 } 440 441 static int 442 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t) 443 { 444 struct ip_tunnel_prl_entry *p; 445 int ok = 1; 446 447 rcu_read_lock(); 448 p = __ipip6_tunnel_locate_prl(t, iph->saddr); 449 if (p) { 450 if (p->flags & PRL_DEFAULT) 451 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT; 452 else 453 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT; 454 } else { 455 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr; 456 457 if (ipv6_addr_is_isatap(addr6) && 458 (addr6->s6_addr32[3] == iph->saddr) && 459 ipv6_chk_prefix(addr6, t->dev)) 460 skb->ndisc_nodetype = NDISC_NODETYPE_HOST; 461 else 462 ok = 0; 463 } 464 rcu_read_unlock(); 465 return ok; 466 } 467 468 static void ipip6_tunnel_uninit(struct net_device *dev) 469 { 470 struct ip_tunnel *tunnel = netdev_priv(dev); 471 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id); 472 473 if (dev == sitn->fb_tunnel_dev) { 474 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL); 475 } else { 476 ipip6_tunnel_unlink(sitn, tunnel); 477 ipip6_tunnel_del_prl(tunnel, NULL); 478 } 479 ip_tunnel_dst_reset_all(tunnel); 480 dev_put(dev); 481 } 482 483 /* Generate icmpv6 with type/code ICMPV6_DEST_UNREACH/ICMPV6_ADDR_UNREACH 484 * if sufficient data bytes are available 485 */ 486 static int ipip6_err_gen_icmpv6_unreach(struct sk_buff *skb) 487 { 488 int ihl = ((const struct iphdr *)skb->data)->ihl*4; 489 struct rt6_info *rt; 490 struct sk_buff *skb2; 491 492 if (!pskb_may_pull(skb, ihl + sizeof(struct ipv6hdr) + 8)) 493 return 1; 494 495 skb2 = skb_clone(skb, GFP_ATOMIC); 496 497 if (!skb2) 498 return 1; 499 500 skb_dst_drop(skb2); 501 skb_pull(skb2, ihl); 502 skb_reset_network_header(skb2); 503 504 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr, NULL, 0, 0); 505 506 if (rt && rt->dst.dev) 507 skb2->dev = rt->dst.dev; 508 509 icmpv6_send(skb2, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0); 510 511 if (rt) 512 ip6_rt_put(rt); 513 514 kfree_skb(skb2); 515 516 return 0; 517 } 518 519 static int ipip6_err(struct sk_buff *skb, u32 info) 520 { 521 const struct iphdr *iph = (const struct iphdr *)skb->data; 522 const int type = icmp_hdr(skb)->type; 523 const int code = icmp_hdr(skb)->code; 524 struct ip_tunnel *t; 525 int err; 526 527 switch (type) { 528 default: 529 case ICMP_PARAMETERPROB: 530 return 0; 531 532 case ICMP_DEST_UNREACH: 533 switch (code) { 534 case ICMP_SR_FAILED: 535 /* Impossible event. */ 536 return 0; 537 default: 538 /* All others are translated to HOST_UNREACH. 539 rfc2003 contains "deep thoughts" about NET_UNREACH, 540 I believe they are just ether pollution. --ANK 541 */ 542 break; 543 } 544 break; 545 case ICMP_TIME_EXCEEDED: 546 if (code != ICMP_EXC_TTL) 547 return 0; 548 break; 549 case ICMP_REDIRECT: 550 break; 551 } 552 553 err = -ENOENT; 554 555 t = ipip6_tunnel_lookup(dev_net(skb->dev), 556 skb->dev, 557 iph->daddr, 558 iph->saddr); 559 if (t == NULL) 560 goto out; 561 562 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { 563 ipv4_update_pmtu(skb, dev_net(skb->dev), info, 564 t->parms.link, 0, IPPROTO_IPV6, 0); 565 err = 0; 566 goto out; 567 } 568 if (type == ICMP_REDIRECT) { 569 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0, 570 IPPROTO_IPV6, 0); 571 err = 0; 572 goto out; 573 } 574 575 if (t->parms.iph.daddr == 0) 576 goto out; 577 578 err = 0; 579 if (!ipip6_err_gen_icmpv6_unreach(skb)) 580 goto out; 581 582 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) 583 goto out; 584 585 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) 586 t->err_count++; 587 else 588 t->err_count = 1; 589 t->err_time = jiffies; 590 out: 591 return err; 592 } 593 594 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr, 595 const struct in6_addr *v6addr) 596 { 597 __be32 v4embed = 0; 598 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed) 599 return true; 600 return false; 601 } 602 603 /* Checks if an address matches an address on the tunnel interface. 604 * Used to detect the NAT of proto 41 packets and let them pass spoofing test. 605 * Long story: 606 * This function is called after we considered the packet as spoofed 607 * in is_spoofed_6rd. 608 * We may have a router that is doing NAT for proto 41 packets 609 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb 610 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd 611 * function will return true, dropping the packet. 612 * But, we can still check if is spoofed against the IP 613 * addresses associated with the interface. 614 */ 615 static bool only_dnatted(const struct ip_tunnel *tunnel, 616 const struct in6_addr *v6dst) 617 { 618 int prefix_len; 619 620 #ifdef CONFIG_IPV6_SIT_6RD 621 prefix_len = tunnel->ip6rd.prefixlen + 32 622 - tunnel->ip6rd.relay_prefixlen; 623 #else 624 prefix_len = 48; 625 #endif 626 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev); 627 } 628 629 /* Returns true if a packet is spoofed */ 630 static bool packet_is_spoofed(struct sk_buff *skb, 631 const struct iphdr *iph, 632 struct ip_tunnel *tunnel) 633 { 634 const struct ipv6hdr *ipv6h; 635 636 if (tunnel->dev->priv_flags & IFF_ISATAP) { 637 if (!isatap_chksrc(skb, iph, tunnel)) 638 return true; 639 640 return false; 641 } 642 643 if (tunnel->dev->flags & IFF_POINTOPOINT) 644 return false; 645 646 ipv6h = ipv6_hdr(skb); 647 648 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) { 649 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 650 &iph->saddr, &ipv6h->saddr, 651 &iph->daddr, &ipv6h->daddr); 652 return true; 653 } 654 655 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr))) 656 return false; 657 658 if (only_dnatted(tunnel, &ipv6h->daddr)) 659 return false; 660 661 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 662 &iph->saddr, &ipv6h->saddr, 663 &iph->daddr, &ipv6h->daddr); 664 return true; 665 } 666 667 static int ipip6_rcv(struct sk_buff *skb) 668 { 669 const struct iphdr *iph = ip_hdr(skb); 670 struct ip_tunnel *tunnel; 671 int err; 672 673 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 674 iph->saddr, iph->daddr); 675 if (tunnel != NULL) { 676 struct pcpu_sw_netstats *tstats; 677 678 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 && 679 tunnel->parms.iph.protocol != 0) 680 goto out; 681 682 skb->mac_header = skb->network_header; 683 skb_reset_network_header(skb); 684 IPCB(skb)->flags = 0; 685 skb->protocol = htons(ETH_P_IPV6); 686 687 if (packet_is_spoofed(skb, iph, tunnel)) { 688 tunnel->dev->stats.rx_errors++; 689 goto out; 690 } 691 692 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net); 693 694 err = IP_ECN_decapsulate(iph, skb); 695 if (unlikely(err)) { 696 if (log_ecn_error) 697 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 698 &iph->saddr, iph->tos); 699 if (err > 1) { 700 ++tunnel->dev->stats.rx_frame_errors; 701 ++tunnel->dev->stats.rx_errors; 702 goto out; 703 } 704 } 705 706 tstats = this_cpu_ptr(tunnel->dev->tstats); 707 u64_stats_update_begin(&tstats->syncp); 708 tstats->rx_packets++; 709 tstats->rx_bytes += skb->len; 710 u64_stats_update_end(&tstats->syncp); 711 712 netif_rx(skb); 713 714 return 0; 715 } 716 717 /* no tunnel matched, let upstream know, ipsec may handle it */ 718 return 1; 719 out: 720 kfree_skb(skb); 721 return 0; 722 } 723 724 static const struct tnl_ptk_info tpi = { 725 /* no tunnel info required for ipip. */ 726 .proto = htons(ETH_P_IP), 727 }; 728 729 static int ipip_rcv(struct sk_buff *skb) 730 { 731 const struct iphdr *iph; 732 struct ip_tunnel *tunnel; 733 734 iph = ip_hdr(skb); 735 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 736 iph->saddr, iph->daddr); 737 if (tunnel != NULL) { 738 if (tunnel->parms.iph.protocol != IPPROTO_IPIP && 739 tunnel->parms.iph.protocol != 0) 740 goto drop; 741 742 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) 743 goto drop; 744 if (iptunnel_pull_header(skb, 0, tpi.proto)) 745 goto drop; 746 return ip_tunnel_rcv(tunnel, skb, &tpi, log_ecn_error); 747 } 748 749 return 1; 750 751 drop: 752 kfree_skb(skb); 753 return 0; 754 } 755 756 /* 757 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function 758 * stores the embedded IPv4 address in v4dst and returns true. 759 */ 760 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, 761 __be32 *v4dst) 762 { 763 #ifdef CONFIG_IPV6_SIT_6RD 764 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, 765 tunnel->ip6rd.prefixlen)) { 766 unsigned int pbw0, pbi0; 767 int pbi1; 768 u32 d; 769 770 pbw0 = tunnel->ip6rd.prefixlen >> 5; 771 pbi0 = tunnel->ip6rd.prefixlen & 0x1f; 772 773 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> 774 tunnel->ip6rd.relay_prefixlen; 775 776 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; 777 if (pbi1 > 0) 778 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> 779 (32 - pbi1); 780 781 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d); 782 return true; 783 } 784 #else 785 if (v6dst->s6_addr16[0] == htons(0x2002)) { 786 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ 787 memcpy(v4dst, &v6dst->s6_addr16[1], 4); 788 return true; 789 } 790 #endif 791 return false; 792 } 793 794 static inline __be32 try_6rd(struct ip_tunnel *tunnel, 795 const struct in6_addr *v6dst) 796 { 797 __be32 dst = 0; 798 check_6rd(tunnel, v6dst, &dst); 799 return dst; 800 } 801 802 /* 803 * This function assumes it is being called from dev_queue_xmit() 804 * and that skb is filled properly by that function. 805 */ 806 807 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, 808 struct net_device *dev) 809 { 810 struct ip_tunnel *tunnel = netdev_priv(dev); 811 const struct iphdr *tiph = &tunnel->parms.iph; 812 const struct ipv6hdr *iph6 = ipv6_hdr(skb); 813 u8 tos = tunnel->parms.iph.tos; 814 __be16 df = tiph->frag_off; 815 struct rtable *rt; /* Route to the other host */ 816 struct net_device *tdev; /* Device to other host */ 817 unsigned int max_headroom; /* The extra header space needed */ 818 __be32 dst = tiph->daddr; 819 struct flowi4 fl4; 820 int mtu; 821 const struct in6_addr *addr6; 822 int addr_type; 823 u8 ttl; 824 int err; 825 u8 protocol = IPPROTO_IPV6; 826 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 827 828 if (skb->protocol != htons(ETH_P_IPV6)) 829 goto tx_error; 830 831 if (tos == 1) 832 tos = ipv6_get_dsfield(iph6); 833 834 /* ISATAP (RFC4214) - must come before 6to4 */ 835 if (dev->priv_flags & IFF_ISATAP) { 836 struct neighbour *neigh = NULL; 837 bool do_tx_error = false; 838 839 if (skb_dst(skb)) 840 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 841 842 if (neigh == NULL) { 843 net_dbg_ratelimited("nexthop == NULL\n"); 844 goto tx_error; 845 } 846 847 addr6 = (const struct in6_addr *)&neigh->primary_key; 848 addr_type = ipv6_addr_type(addr6); 849 850 if ((addr_type & IPV6_ADDR_UNICAST) && 851 ipv6_addr_is_isatap(addr6)) 852 dst = addr6->s6_addr32[3]; 853 else 854 do_tx_error = true; 855 856 neigh_release(neigh); 857 if (do_tx_error) 858 goto tx_error; 859 } 860 861 if (!dst) 862 dst = try_6rd(tunnel, &iph6->daddr); 863 864 if (!dst) { 865 struct neighbour *neigh = NULL; 866 bool do_tx_error = false; 867 868 if (skb_dst(skb)) 869 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 870 871 if (neigh == NULL) { 872 net_dbg_ratelimited("nexthop == NULL\n"); 873 goto tx_error; 874 } 875 876 addr6 = (const struct in6_addr *)&neigh->primary_key; 877 addr_type = ipv6_addr_type(addr6); 878 879 if (addr_type == IPV6_ADDR_ANY) { 880 addr6 = &ipv6_hdr(skb)->daddr; 881 addr_type = ipv6_addr_type(addr6); 882 } 883 884 if ((addr_type & IPV6_ADDR_COMPATv4) != 0) 885 dst = addr6->s6_addr32[3]; 886 else 887 do_tx_error = true; 888 889 neigh_release(neigh); 890 if (do_tx_error) 891 goto tx_error; 892 } 893 894 rt = ip_route_output_ports(tunnel->net, &fl4, NULL, 895 dst, tiph->saddr, 896 0, 0, 897 IPPROTO_IPV6, RT_TOS(tos), 898 tunnel->parms.link); 899 if (IS_ERR(rt)) { 900 dev->stats.tx_carrier_errors++; 901 goto tx_error_icmp; 902 } 903 if (rt->rt_type != RTN_UNICAST) { 904 ip_rt_put(rt); 905 dev->stats.tx_carrier_errors++; 906 goto tx_error_icmp; 907 } 908 tdev = rt->dst.dev; 909 910 if (tdev == dev) { 911 ip_rt_put(rt); 912 dev->stats.collisions++; 913 goto tx_error; 914 } 915 916 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT); 917 if (IS_ERR(skb)) { 918 ip_rt_put(rt); 919 goto out; 920 } 921 922 if (df) { 923 mtu = dst_mtu(&rt->dst) - t_hlen; 924 925 if (mtu < 68) { 926 dev->stats.collisions++; 927 ip_rt_put(rt); 928 goto tx_error; 929 } 930 931 if (mtu < IPV6_MIN_MTU) { 932 mtu = IPV6_MIN_MTU; 933 df = 0; 934 } 935 936 if (tunnel->parms.iph.daddr && skb_dst(skb)) 937 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 938 939 if (skb->len > mtu && !skb_is_gso(skb)) { 940 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 941 ip_rt_put(rt); 942 goto tx_error; 943 } 944 } 945 946 if (tunnel->err_count > 0) { 947 if (time_before(jiffies, 948 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 949 tunnel->err_count--; 950 dst_link_failure(skb); 951 } else 952 tunnel->err_count = 0; 953 } 954 955 /* 956 * Okay, now see if we can stuff it in the buffer as-is. 957 */ 958 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen; 959 960 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 961 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 962 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 963 if (!new_skb) { 964 ip_rt_put(rt); 965 dev->stats.tx_dropped++; 966 kfree_skb(skb); 967 return NETDEV_TX_OK; 968 } 969 if (skb->sk) 970 skb_set_owner_w(new_skb, skb->sk); 971 dev_kfree_skb(skb); 972 skb = new_skb; 973 iph6 = ipv6_hdr(skb); 974 } 975 ttl = tiph->ttl; 976 if (ttl == 0) 977 ttl = iph6->hop_limit; 978 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); 979 980 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) { 981 ip_rt_put(rt); 982 goto tx_error; 983 } 984 985 skb_set_inner_ipproto(skb, IPPROTO_IPV6); 986 987 err = iptunnel_xmit(skb->sk, rt, skb, fl4.saddr, fl4.daddr, 988 protocol, tos, ttl, df, 989 !net_eq(tunnel->net, dev_net(dev))); 990 iptunnel_xmit_stats(err, &dev->stats, dev->tstats); 991 return NETDEV_TX_OK; 992 993 tx_error_icmp: 994 dst_link_failure(skb); 995 tx_error: 996 kfree_skb(skb); 997 out: 998 dev->stats.tx_errors++; 999 return NETDEV_TX_OK; 1000 } 1001 1002 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) 1003 { 1004 struct ip_tunnel *tunnel = netdev_priv(dev); 1005 const struct iphdr *tiph = &tunnel->parms.iph; 1006 1007 skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP); 1008 if (IS_ERR(skb)) 1009 goto out; 1010 1011 skb_set_inner_ipproto(skb, IPPROTO_IPIP); 1012 1013 ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP); 1014 return NETDEV_TX_OK; 1015 out: 1016 dev->stats.tx_errors++; 1017 return NETDEV_TX_OK; 1018 } 1019 1020 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb, 1021 struct net_device *dev) 1022 { 1023 switch (skb->protocol) { 1024 case htons(ETH_P_IP): 1025 ipip_tunnel_xmit(skb, dev); 1026 break; 1027 case htons(ETH_P_IPV6): 1028 ipip6_tunnel_xmit(skb, dev); 1029 break; 1030 default: 1031 goto tx_err; 1032 } 1033 1034 return NETDEV_TX_OK; 1035 1036 tx_err: 1037 dev->stats.tx_errors++; 1038 kfree_skb(skb); 1039 return NETDEV_TX_OK; 1040 1041 } 1042 1043 static void ipip6_tunnel_bind_dev(struct net_device *dev) 1044 { 1045 struct net_device *tdev = NULL; 1046 struct ip_tunnel *tunnel; 1047 const struct iphdr *iph; 1048 struct flowi4 fl4; 1049 1050 tunnel = netdev_priv(dev); 1051 iph = &tunnel->parms.iph; 1052 1053 if (iph->daddr) { 1054 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4, 1055 NULL, 1056 iph->daddr, iph->saddr, 1057 0, 0, 1058 IPPROTO_IPV6, 1059 RT_TOS(iph->tos), 1060 tunnel->parms.link); 1061 1062 if (!IS_ERR(rt)) { 1063 tdev = rt->dst.dev; 1064 ip_rt_put(rt); 1065 } 1066 dev->flags |= IFF_POINTOPOINT; 1067 } 1068 1069 if (!tdev && tunnel->parms.link) 1070 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link); 1071 1072 if (tdev) { 1073 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1074 1075 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); 1076 dev->mtu = tdev->mtu - t_hlen; 1077 if (dev->mtu < IPV6_MIN_MTU) 1078 dev->mtu = IPV6_MIN_MTU; 1079 } 1080 dev->iflink = tunnel->parms.link; 1081 } 1082 1083 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p) 1084 { 1085 struct net *net = t->net; 1086 struct sit_net *sitn = net_generic(net, sit_net_id); 1087 1088 ipip6_tunnel_unlink(sitn, t); 1089 synchronize_net(); 1090 t->parms.iph.saddr = p->iph.saddr; 1091 t->parms.iph.daddr = p->iph.daddr; 1092 memcpy(t->dev->dev_addr, &p->iph.saddr, 4); 1093 memcpy(t->dev->broadcast, &p->iph.daddr, 4); 1094 ipip6_tunnel_link(sitn, t); 1095 t->parms.iph.ttl = p->iph.ttl; 1096 t->parms.iph.tos = p->iph.tos; 1097 if (t->parms.link != p->link) { 1098 t->parms.link = p->link; 1099 ipip6_tunnel_bind_dev(t->dev); 1100 } 1101 ip_tunnel_dst_reset_all(t); 1102 netdev_state_change(t->dev); 1103 } 1104 1105 #ifdef CONFIG_IPV6_SIT_6RD 1106 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t, 1107 struct ip_tunnel_6rd *ip6rd) 1108 { 1109 struct in6_addr prefix; 1110 __be32 relay_prefix; 1111 1112 if (ip6rd->relay_prefixlen > 32 || 1113 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64) 1114 return -EINVAL; 1115 1116 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen); 1117 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix)) 1118 return -EINVAL; 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 return -EINVAL; 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 ip_tunnel_dst_reset_all(t); 1133 netdev_state_change(t->dev); 1134 return 0; 1135 } 1136 #endif 1137 1138 static int 1139 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1140 { 1141 int err = 0; 1142 struct ip_tunnel_parm p; 1143 struct ip_tunnel_prl prl; 1144 struct ip_tunnel *t = netdev_priv(dev); 1145 struct net *net = t->net; 1146 struct sit_net *sitn = net_generic(net, sit_net_id); 1147 #ifdef CONFIG_IPV6_SIT_6RD 1148 struct ip_tunnel_6rd ip6rd; 1149 #endif 1150 1151 switch (cmd) { 1152 case SIOCGETTUNNEL: 1153 #ifdef CONFIG_IPV6_SIT_6RD 1154 case SIOCGET6RD: 1155 #endif 1156 if (dev == sitn->fb_tunnel_dev) { 1157 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1158 err = -EFAULT; 1159 break; 1160 } 1161 t = ipip6_tunnel_locate(net, &p, 0); 1162 if (t == NULL) 1163 t = netdev_priv(dev); 1164 } 1165 1166 err = -EFAULT; 1167 if (cmd == SIOCGETTUNNEL) { 1168 memcpy(&p, &t->parms, sizeof(p)); 1169 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, 1170 sizeof(p))) 1171 goto done; 1172 #ifdef CONFIG_IPV6_SIT_6RD 1173 } else { 1174 ip6rd.prefix = t->ip6rd.prefix; 1175 ip6rd.relay_prefix = t->ip6rd.relay_prefix; 1176 ip6rd.prefixlen = t->ip6rd.prefixlen; 1177 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; 1178 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, 1179 sizeof(ip6rd))) 1180 goto done; 1181 #endif 1182 } 1183 err = 0; 1184 break; 1185 1186 case SIOCADDTUNNEL: 1187 case SIOCCHGTUNNEL: 1188 err = -EPERM; 1189 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1190 goto done; 1191 1192 err = -EFAULT; 1193 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1194 goto done; 1195 1196 err = -EINVAL; 1197 if (p.iph.protocol != IPPROTO_IPV6 && 1198 p.iph.protocol != IPPROTO_IPIP && 1199 p.iph.protocol != 0) 1200 goto done; 1201 if (p.iph.version != 4 || 1202 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) 1203 goto done; 1204 if (p.iph.ttl) 1205 p.iph.frag_off |= htons(IP_DF); 1206 1207 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); 1208 1209 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1210 if (t != NULL) { 1211 if (t->dev != dev) { 1212 err = -EEXIST; 1213 break; 1214 } 1215 } else { 1216 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || 1217 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { 1218 err = -EINVAL; 1219 break; 1220 } 1221 t = netdev_priv(dev); 1222 } 1223 1224 ipip6_tunnel_update(t, &p); 1225 } 1226 1227 if (t) { 1228 err = 0; 1229 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) 1230 err = -EFAULT; 1231 } else 1232 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1233 break; 1234 1235 case SIOCDELTUNNEL: 1236 err = -EPERM; 1237 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1238 goto done; 1239 1240 if (dev == sitn->fb_tunnel_dev) { 1241 err = -EFAULT; 1242 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1243 goto done; 1244 err = -ENOENT; 1245 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL) 1246 goto done; 1247 err = -EPERM; 1248 if (t == netdev_priv(sitn->fb_tunnel_dev)) 1249 goto done; 1250 dev = t->dev; 1251 } 1252 unregister_netdevice(dev); 1253 err = 0; 1254 break; 1255 1256 case SIOCGETPRL: 1257 err = -EINVAL; 1258 if (dev == sitn->fb_tunnel_dev) 1259 goto done; 1260 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data); 1261 break; 1262 1263 case SIOCADDPRL: 1264 case SIOCDELPRL: 1265 case SIOCCHGPRL: 1266 err = -EPERM; 1267 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1268 goto done; 1269 err = -EINVAL; 1270 if (dev == sitn->fb_tunnel_dev) 1271 goto done; 1272 err = -EFAULT; 1273 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl))) 1274 goto done; 1275 1276 switch (cmd) { 1277 case SIOCDELPRL: 1278 err = ipip6_tunnel_del_prl(t, &prl); 1279 break; 1280 case SIOCADDPRL: 1281 case SIOCCHGPRL: 1282 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); 1283 break; 1284 } 1285 ip_tunnel_dst_reset_all(t); 1286 netdev_state_change(dev); 1287 break; 1288 1289 #ifdef CONFIG_IPV6_SIT_6RD 1290 case SIOCADD6RD: 1291 case SIOCCHG6RD: 1292 case SIOCDEL6RD: 1293 err = -EPERM; 1294 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1295 goto done; 1296 1297 err = -EFAULT; 1298 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, 1299 sizeof(ip6rd))) 1300 goto done; 1301 1302 if (cmd != SIOCDEL6RD) { 1303 err = ipip6_tunnel_update_6rd(t, &ip6rd); 1304 if (err < 0) 1305 goto done; 1306 } else 1307 ipip6_tunnel_clone_6rd(dev, sitn); 1308 1309 err = 0; 1310 break; 1311 #endif 1312 1313 default: 1314 err = -EINVAL; 1315 } 1316 1317 done: 1318 return err; 1319 } 1320 1321 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu) 1322 { 1323 struct ip_tunnel *tunnel = netdev_priv(dev); 1324 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1325 1326 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - t_hlen) 1327 return -EINVAL; 1328 dev->mtu = new_mtu; 1329 return 0; 1330 } 1331 1332 static const struct net_device_ops ipip6_netdev_ops = { 1333 .ndo_uninit = ipip6_tunnel_uninit, 1334 .ndo_start_xmit = sit_tunnel_xmit, 1335 .ndo_do_ioctl = ipip6_tunnel_ioctl, 1336 .ndo_change_mtu = ipip6_tunnel_change_mtu, 1337 .ndo_get_stats64 = ip_tunnel_get_stats64, 1338 }; 1339 1340 static void ipip6_dev_free(struct net_device *dev) 1341 { 1342 struct ip_tunnel *tunnel = netdev_priv(dev); 1343 1344 free_percpu(tunnel->dst_cache); 1345 free_percpu(dev->tstats); 1346 free_netdev(dev); 1347 } 1348 1349 #define SIT_FEATURES (NETIF_F_SG | \ 1350 NETIF_F_FRAGLIST | \ 1351 NETIF_F_HIGHDMA | \ 1352 NETIF_F_GSO_SOFTWARE | \ 1353 NETIF_F_HW_CSUM) 1354 1355 static void ipip6_tunnel_setup(struct net_device *dev) 1356 { 1357 struct ip_tunnel *tunnel = netdev_priv(dev); 1358 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1359 1360 dev->netdev_ops = &ipip6_netdev_ops; 1361 dev->destructor = ipip6_dev_free; 1362 1363 dev->type = ARPHRD_SIT; 1364 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1365 dev->mtu = ETH_DATA_LEN - t_hlen; 1366 dev->flags = IFF_NOARP; 1367 netif_keep_dst(dev); 1368 dev->iflink = 0; 1369 dev->addr_len = 4; 1370 dev->features |= NETIF_F_LLTX; 1371 dev->features |= SIT_FEATURES; 1372 dev->hw_features |= SIT_FEATURES; 1373 } 1374 1375 static int ipip6_tunnel_init(struct net_device *dev) 1376 { 1377 struct ip_tunnel *tunnel = netdev_priv(dev); 1378 1379 tunnel->dev = dev; 1380 tunnel->net = dev_net(dev); 1381 1382 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); 1383 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); 1384 1385 ipip6_tunnel_bind_dev(dev); 1386 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1387 if (!dev->tstats) 1388 return -ENOMEM; 1389 1390 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst); 1391 if (!tunnel->dst_cache) { 1392 free_percpu(dev->tstats); 1393 return -ENOMEM; 1394 } 1395 1396 return 0; 1397 } 1398 1399 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev) 1400 { 1401 struct ip_tunnel *tunnel = netdev_priv(dev); 1402 struct iphdr *iph = &tunnel->parms.iph; 1403 struct net *net = dev_net(dev); 1404 struct sit_net *sitn = net_generic(net, sit_net_id); 1405 1406 tunnel->dev = dev; 1407 tunnel->net = dev_net(dev); 1408 strcpy(tunnel->parms.name, dev->name); 1409 1410 iph->version = 4; 1411 iph->protocol = IPPROTO_IPV6; 1412 iph->ihl = 5; 1413 iph->ttl = 64; 1414 1415 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1416 if (!dev->tstats) 1417 return -ENOMEM; 1418 1419 tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst); 1420 if (!tunnel->dst_cache) { 1421 free_percpu(dev->tstats); 1422 return -ENOMEM; 1423 } 1424 1425 dev_hold(dev); 1426 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); 1427 return 0; 1428 } 1429 1430 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[]) 1431 { 1432 u8 proto; 1433 1434 if (!data || !data[IFLA_IPTUN_PROTO]) 1435 return 0; 1436 1437 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1438 if (proto != IPPROTO_IPV6 && 1439 proto != IPPROTO_IPIP && 1440 proto != 0) 1441 return -EINVAL; 1442 1443 return 0; 1444 } 1445 1446 static void ipip6_netlink_parms(struct nlattr *data[], 1447 struct ip_tunnel_parm *parms) 1448 { 1449 memset(parms, 0, sizeof(*parms)); 1450 1451 parms->iph.version = 4; 1452 parms->iph.protocol = IPPROTO_IPV6; 1453 parms->iph.ihl = 5; 1454 parms->iph.ttl = 64; 1455 1456 if (!data) 1457 return; 1458 1459 if (data[IFLA_IPTUN_LINK]) 1460 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1461 1462 if (data[IFLA_IPTUN_LOCAL]) 1463 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]); 1464 1465 if (data[IFLA_IPTUN_REMOTE]) 1466 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]); 1467 1468 if (data[IFLA_IPTUN_TTL]) { 1469 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]); 1470 if (parms->iph.ttl) 1471 parms->iph.frag_off = htons(IP_DF); 1472 } 1473 1474 if (data[IFLA_IPTUN_TOS]) 1475 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]); 1476 1477 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC])) 1478 parms->iph.frag_off = htons(IP_DF); 1479 1480 if (data[IFLA_IPTUN_FLAGS]) 1481 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]); 1482 1483 if (data[IFLA_IPTUN_PROTO]) 1484 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1485 1486 } 1487 1488 /* This function returns true when ENCAP attributes are present in the nl msg */ 1489 static bool ipip6_netlink_encap_parms(struct nlattr *data[], 1490 struct ip_tunnel_encap *ipencap) 1491 { 1492 bool ret = false; 1493 1494 memset(ipencap, 0, sizeof(*ipencap)); 1495 1496 if (!data) 1497 return ret; 1498 1499 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 1500 ret = true; 1501 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 1502 } 1503 1504 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 1505 ret = true; 1506 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 1507 } 1508 1509 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 1510 ret = true; 1511 ipencap->sport = nla_get_u16(data[IFLA_IPTUN_ENCAP_SPORT]); 1512 } 1513 1514 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 1515 ret = true; 1516 ipencap->dport = nla_get_u16(data[IFLA_IPTUN_ENCAP_DPORT]); 1517 } 1518 1519 return ret; 1520 } 1521 1522 #ifdef CONFIG_IPV6_SIT_6RD 1523 /* This function returns true when 6RD attributes are present in the nl msg */ 1524 static bool ipip6_netlink_6rd_parms(struct nlattr *data[], 1525 struct ip_tunnel_6rd *ip6rd) 1526 { 1527 bool ret = false; 1528 memset(ip6rd, 0, sizeof(*ip6rd)); 1529 1530 if (!data) 1531 return ret; 1532 1533 if (data[IFLA_IPTUN_6RD_PREFIX]) { 1534 ret = true; 1535 nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX], 1536 sizeof(struct in6_addr)); 1537 } 1538 1539 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) { 1540 ret = true; 1541 ip6rd->relay_prefix = 1542 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]); 1543 } 1544 1545 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) { 1546 ret = true; 1547 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]); 1548 } 1549 1550 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) { 1551 ret = true; 1552 ip6rd->relay_prefixlen = 1553 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]); 1554 } 1555 1556 return ret; 1557 } 1558 #endif 1559 1560 static int ipip6_newlink(struct net *src_net, struct net_device *dev, 1561 struct nlattr *tb[], struct nlattr *data[]) 1562 { 1563 struct net *net = dev_net(dev); 1564 struct ip_tunnel *nt; 1565 struct ip_tunnel_encap ipencap; 1566 #ifdef CONFIG_IPV6_SIT_6RD 1567 struct ip_tunnel_6rd ip6rd; 1568 #endif 1569 int err; 1570 1571 nt = netdev_priv(dev); 1572 1573 if (ipip6_netlink_encap_parms(data, &ipencap)) { 1574 err = ip_tunnel_encap_setup(nt, &ipencap); 1575 if (err < 0) 1576 return err; 1577 } 1578 1579 ipip6_netlink_parms(data, &nt->parms); 1580 1581 if (ipip6_tunnel_locate(net, &nt->parms, 0)) 1582 return -EEXIST; 1583 1584 err = ipip6_tunnel_create(dev); 1585 if (err < 0) 1586 return err; 1587 1588 #ifdef CONFIG_IPV6_SIT_6RD 1589 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1590 err = ipip6_tunnel_update_6rd(nt, &ip6rd); 1591 #endif 1592 1593 return err; 1594 } 1595 1596 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[], 1597 struct nlattr *data[]) 1598 { 1599 struct ip_tunnel *t = netdev_priv(dev); 1600 struct ip_tunnel_parm p; 1601 struct ip_tunnel_encap ipencap; 1602 struct net *net = t->net; 1603 struct sit_net *sitn = net_generic(net, sit_net_id); 1604 #ifdef CONFIG_IPV6_SIT_6RD 1605 struct ip_tunnel_6rd ip6rd; 1606 #endif 1607 int err; 1608 1609 if (dev == sitn->fb_tunnel_dev) 1610 return -EINVAL; 1611 1612 if (ipip6_netlink_encap_parms(data, &ipencap)) { 1613 err = ip_tunnel_encap_setup(t, &ipencap); 1614 if (err < 0) 1615 return err; 1616 } 1617 1618 ipip6_netlink_parms(data, &p); 1619 1620 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) || 1621 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr)) 1622 return -EINVAL; 1623 1624 t = ipip6_tunnel_locate(net, &p, 0); 1625 1626 if (t) { 1627 if (t->dev != dev) 1628 return -EEXIST; 1629 } else 1630 t = netdev_priv(dev); 1631 1632 ipip6_tunnel_update(t, &p); 1633 1634 #ifdef CONFIG_IPV6_SIT_6RD 1635 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1636 return ipip6_tunnel_update_6rd(t, &ip6rd); 1637 #endif 1638 1639 return 0; 1640 } 1641 1642 static size_t ipip6_get_size(const struct net_device *dev) 1643 { 1644 return 1645 /* IFLA_IPTUN_LINK */ 1646 nla_total_size(4) + 1647 /* IFLA_IPTUN_LOCAL */ 1648 nla_total_size(4) + 1649 /* IFLA_IPTUN_REMOTE */ 1650 nla_total_size(4) + 1651 /* IFLA_IPTUN_TTL */ 1652 nla_total_size(1) + 1653 /* IFLA_IPTUN_TOS */ 1654 nla_total_size(1) + 1655 /* IFLA_IPTUN_PMTUDISC */ 1656 nla_total_size(1) + 1657 /* IFLA_IPTUN_FLAGS */ 1658 nla_total_size(2) + 1659 /* IFLA_IPTUN_PROTO */ 1660 nla_total_size(1) + 1661 #ifdef CONFIG_IPV6_SIT_6RD 1662 /* IFLA_IPTUN_6RD_PREFIX */ 1663 nla_total_size(sizeof(struct in6_addr)) + 1664 /* IFLA_IPTUN_6RD_RELAY_PREFIX */ 1665 nla_total_size(4) + 1666 /* IFLA_IPTUN_6RD_PREFIXLEN */ 1667 nla_total_size(2) + 1668 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */ 1669 nla_total_size(2) + 1670 #endif 1671 /* IFLA_IPTUN_ENCAP_TYPE */ 1672 nla_total_size(2) + 1673 /* IFLA_IPTUN_ENCAP_FLAGS */ 1674 nla_total_size(2) + 1675 /* IFLA_IPTUN_ENCAP_SPORT */ 1676 nla_total_size(2) + 1677 /* IFLA_IPTUN_ENCAP_DPORT */ 1678 nla_total_size(2) + 1679 0; 1680 } 1681 1682 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1683 { 1684 struct ip_tunnel *tunnel = netdev_priv(dev); 1685 struct ip_tunnel_parm *parm = &tunnel->parms; 1686 1687 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1688 nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) || 1689 nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) || 1690 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) || 1691 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) || 1692 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC, 1693 !!(parm->iph.frag_off & htons(IP_DF))) || 1694 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) || 1695 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags)) 1696 goto nla_put_failure; 1697 1698 #ifdef CONFIG_IPV6_SIT_6RD 1699 if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr), 1700 &tunnel->ip6rd.prefix) || 1701 nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX, 1702 tunnel->ip6rd.relay_prefix) || 1703 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN, 1704 tunnel->ip6rd.prefixlen) || 1705 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN, 1706 tunnel->ip6rd.relay_prefixlen)) 1707 goto nla_put_failure; 1708 #endif 1709 1710 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, 1711 tunnel->encap.type) || 1712 nla_put_u16(skb, IFLA_IPTUN_ENCAP_SPORT, 1713 tunnel->encap.sport) || 1714 nla_put_u16(skb, IFLA_IPTUN_ENCAP_DPORT, 1715 tunnel->encap.dport) || 1716 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, 1717 tunnel->encap.dport)) 1718 goto nla_put_failure; 1719 1720 return 0; 1721 1722 nla_put_failure: 1723 return -EMSGSIZE; 1724 } 1725 1726 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = { 1727 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1728 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 }, 1729 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 }, 1730 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1731 [IFLA_IPTUN_TOS] = { .type = NLA_U8 }, 1732 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 }, 1733 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 }, 1734 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1735 #ifdef CONFIG_IPV6_SIT_6RD 1736 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) }, 1737 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 }, 1738 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 }, 1739 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 }, 1740 #endif 1741 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 1742 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 1743 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 1744 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 1745 }; 1746 1747 static void ipip6_dellink(struct net_device *dev, struct list_head *head) 1748 { 1749 struct net *net = dev_net(dev); 1750 struct sit_net *sitn = net_generic(net, sit_net_id); 1751 1752 if (dev != sitn->fb_tunnel_dev) 1753 unregister_netdevice_queue(dev, head); 1754 } 1755 1756 static struct rtnl_link_ops sit_link_ops __read_mostly = { 1757 .kind = "sit", 1758 .maxtype = IFLA_IPTUN_MAX, 1759 .policy = ipip6_policy, 1760 .priv_size = sizeof(struct ip_tunnel), 1761 .setup = ipip6_tunnel_setup, 1762 .validate = ipip6_validate, 1763 .newlink = ipip6_newlink, 1764 .changelink = ipip6_changelink, 1765 .get_size = ipip6_get_size, 1766 .fill_info = ipip6_fill_info, 1767 .dellink = ipip6_dellink, 1768 }; 1769 1770 static struct xfrm_tunnel sit_handler __read_mostly = { 1771 .handler = ipip6_rcv, 1772 .err_handler = ipip6_err, 1773 .priority = 1, 1774 }; 1775 1776 static struct xfrm_tunnel ipip_handler __read_mostly = { 1777 .handler = ipip_rcv, 1778 .err_handler = ipip6_err, 1779 .priority = 2, 1780 }; 1781 1782 static void __net_exit sit_destroy_tunnels(struct net *net, 1783 struct list_head *head) 1784 { 1785 struct sit_net *sitn = net_generic(net, sit_net_id); 1786 struct net_device *dev, *aux; 1787 int prio; 1788 1789 for_each_netdev_safe(net, dev, aux) 1790 if (dev->rtnl_link_ops == &sit_link_ops) 1791 unregister_netdevice_queue(dev, head); 1792 1793 for (prio = 1; prio < 4; prio++) { 1794 int h; 1795 for (h = 0; h < HASH_SIZE; h++) { 1796 struct ip_tunnel *t; 1797 1798 t = rtnl_dereference(sitn->tunnels[prio][h]); 1799 while (t != NULL) { 1800 /* If dev is in the same netns, it has already 1801 * been added to the list by the previous loop. 1802 */ 1803 if (!net_eq(dev_net(t->dev), net)) 1804 unregister_netdevice_queue(t->dev, 1805 head); 1806 t = rtnl_dereference(t->next); 1807 } 1808 } 1809 } 1810 } 1811 1812 static int __net_init sit_init_net(struct net *net) 1813 { 1814 struct sit_net *sitn = net_generic(net, sit_net_id); 1815 struct ip_tunnel *t; 1816 int err; 1817 1818 sitn->tunnels[0] = sitn->tunnels_wc; 1819 sitn->tunnels[1] = sitn->tunnels_l; 1820 sitn->tunnels[2] = sitn->tunnels_r; 1821 sitn->tunnels[3] = sitn->tunnels_r_l; 1822 1823 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", 1824 NET_NAME_UNKNOWN, 1825 ipip6_tunnel_setup); 1826 if (!sitn->fb_tunnel_dev) { 1827 err = -ENOMEM; 1828 goto err_alloc_dev; 1829 } 1830 dev_net_set(sitn->fb_tunnel_dev, net); 1831 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops; 1832 /* FB netdevice is special: we have one, and only one per netns. 1833 * Allowing to move it to another netns is clearly unsafe. 1834 */ 1835 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1836 1837 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); 1838 if (err) 1839 goto err_dev_free; 1840 1841 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); 1842 1843 if ((err = register_netdev(sitn->fb_tunnel_dev))) 1844 goto err_reg_dev; 1845 1846 t = netdev_priv(sitn->fb_tunnel_dev); 1847 1848 strcpy(t->parms.name, sitn->fb_tunnel_dev->name); 1849 return 0; 1850 1851 err_reg_dev: 1852 dev_put(sitn->fb_tunnel_dev); 1853 err_dev_free: 1854 ipip6_dev_free(sitn->fb_tunnel_dev); 1855 err_alloc_dev: 1856 return err; 1857 } 1858 1859 static void __net_exit sit_exit_net(struct net *net) 1860 { 1861 LIST_HEAD(list); 1862 1863 rtnl_lock(); 1864 sit_destroy_tunnels(net, &list); 1865 unregister_netdevice_many(&list); 1866 rtnl_unlock(); 1867 } 1868 1869 static struct pernet_operations sit_net_ops = { 1870 .init = sit_init_net, 1871 .exit = sit_exit_net, 1872 .id = &sit_net_id, 1873 .size = sizeof(struct sit_net), 1874 }; 1875 1876 static void __exit sit_cleanup(void) 1877 { 1878 rtnl_link_unregister(&sit_link_ops); 1879 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1880 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1881 1882 unregister_pernet_device(&sit_net_ops); 1883 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1884 } 1885 1886 static int __init sit_init(void) 1887 { 1888 int err; 1889 1890 pr_info("IPv6 over IPv4 tunneling driver\n"); 1891 1892 err = register_pernet_device(&sit_net_ops); 1893 if (err < 0) 1894 return err; 1895 err = xfrm4_tunnel_register(&sit_handler, AF_INET6); 1896 if (err < 0) { 1897 pr_info("%s: can't register ip6ip4\n", __func__); 1898 goto xfrm_tunnel_failed; 1899 } 1900 err = xfrm4_tunnel_register(&ipip_handler, AF_INET); 1901 if (err < 0) { 1902 pr_info("%s: can't register ip4ip4\n", __func__); 1903 goto xfrm_tunnel4_failed; 1904 } 1905 err = rtnl_link_register(&sit_link_ops); 1906 if (err < 0) 1907 goto rtnl_link_failed; 1908 1909 out: 1910 return err; 1911 1912 rtnl_link_failed: 1913 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1914 xfrm_tunnel4_failed: 1915 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1916 xfrm_tunnel_failed: 1917 unregister_pernet_device(&sit_net_ops); 1918 goto out; 1919 } 1920 1921 module_init(sit_init); 1922 module_exit(sit_cleanup); 1923 MODULE_LICENSE("GPL"); 1924 MODULE_ALIAS_RTNL_LINK("sit"); 1925 MODULE_ALIAS_NETDEV("sit0"); 1926