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 <linux/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 IP6_SIT_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 unsigned int sit_net_id __read_mostly; 80 struct sit_net { 81 struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE]; 82 struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE]; 83 struct ip_tunnel __rcu *tunnels_l[IP6_SIT_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 && (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 memcpy(dev->dev_addr, &t->parms.iph.saddr, 4); 199 memcpy(dev->broadcast, &t->parms.iph.daddr, 4); 200 201 if ((__force u16)t->parms.i_flags & SIT_ISATAP) 202 dev->priv_flags |= IFF_ISATAP; 203 204 dev->rtnl_link_ops = &sit_link_ops; 205 206 err = register_netdevice(dev); 207 if (err < 0) 208 goto out; 209 210 ipip6_tunnel_clone_6rd(dev, sitn); 211 212 dev_hold(dev); 213 214 ipip6_tunnel_link(sitn, t); 215 return 0; 216 217 out: 218 return err; 219 } 220 221 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, 222 struct ip_tunnel_parm *parms, int create) 223 { 224 __be32 remote = parms->iph.daddr; 225 __be32 local = parms->iph.saddr; 226 struct ip_tunnel *t, *nt; 227 struct ip_tunnel __rcu **tp; 228 struct net_device *dev; 229 char name[IFNAMSIZ]; 230 struct sit_net *sitn = net_generic(net, sit_net_id); 231 232 for (tp = __ipip6_bucket(sitn, parms); 233 (t = rtnl_dereference(*tp)) != NULL; 234 tp = &t->next) { 235 if (local == t->parms.iph.saddr && 236 remote == t->parms.iph.daddr && 237 parms->link == t->parms.link) { 238 if (create) 239 return NULL; 240 else 241 return t; 242 } 243 } 244 if (!create) 245 goto failed; 246 247 if (parms->name[0]) 248 strlcpy(name, parms->name, IFNAMSIZ); 249 else 250 strcpy(name, "sit%d"); 251 252 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 253 ipip6_tunnel_setup); 254 if (!dev) 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 free_netdev(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 | __GFP_NOWARN) : 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 dst_cache_reset(&tunnel->dst_cache); 479 dev_put(dev); 480 } 481 482 static int ipip6_err(struct sk_buff *skb, u32 info) 483 { 484 const struct iphdr *iph = (const struct iphdr *)skb->data; 485 const int type = icmp_hdr(skb)->type; 486 const int code = icmp_hdr(skb)->code; 487 unsigned int data_len = 0; 488 struct ip_tunnel *t; 489 int err; 490 491 switch (type) { 492 default: 493 case ICMP_PARAMETERPROB: 494 return 0; 495 496 case ICMP_DEST_UNREACH: 497 switch (code) { 498 case ICMP_SR_FAILED: 499 /* Impossible event. */ 500 return 0; 501 default: 502 /* All others are translated to HOST_UNREACH. 503 rfc2003 contains "deep thoughts" about NET_UNREACH, 504 I believe they are just ether pollution. --ANK 505 */ 506 break; 507 } 508 break; 509 case ICMP_TIME_EXCEEDED: 510 if (code != ICMP_EXC_TTL) 511 return 0; 512 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */ 513 break; 514 case ICMP_REDIRECT: 515 break; 516 } 517 518 err = -ENOENT; 519 520 t = ipip6_tunnel_lookup(dev_net(skb->dev), 521 skb->dev, 522 iph->daddr, 523 iph->saddr); 524 if (!t) 525 goto out; 526 527 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { 528 ipv4_update_pmtu(skb, dev_net(skb->dev), info, 529 t->parms.link, 0, iph->protocol, 0); 530 err = 0; 531 goto out; 532 } 533 if (type == ICMP_REDIRECT) { 534 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0, 535 iph->protocol, 0); 536 err = 0; 537 goto out; 538 } 539 540 err = 0; 541 if (!ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len)) 542 goto out; 543 544 if (t->parms.iph.daddr == 0) 545 goto out; 546 547 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) 548 goto out; 549 550 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) 551 t->err_count++; 552 else 553 t->err_count = 1; 554 t->err_time = jiffies; 555 out: 556 return err; 557 } 558 559 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr, 560 const struct in6_addr *v6addr) 561 { 562 __be32 v4embed = 0; 563 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed) 564 return true; 565 return false; 566 } 567 568 /* Checks if an address matches an address on the tunnel interface. 569 * Used to detect the NAT of proto 41 packets and let them pass spoofing test. 570 * Long story: 571 * This function is called after we considered the packet as spoofed 572 * in is_spoofed_6rd. 573 * We may have a router that is doing NAT for proto 41 packets 574 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb 575 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd 576 * function will return true, dropping the packet. 577 * But, we can still check if is spoofed against the IP 578 * addresses associated with the interface. 579 */ 580 static bool only_dnatted(const struct ip_tunnel *tunnel, 581 const struct in6_addr *v6dst) 582 { 583 int prefix_len; 584 585 #ifdef CONFIG_IPV6_SIT_6RD 586 prefix_len = tunnel->ip6rd.prefixlen + 32 587 - tunnel->ip6rd.relay_prefixlen; 588 #else 589 prefix_len = 48; 590 #endif 591 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev); 592 } 593 594 /* Returns true if a packet is spoofed */ 595 static bool packet_is_spoofed(struct sk_buff *skb, 596 const struct iphdr *iph, 597 struct ip_tunnel *tunnel) 598 { 599 const struct ipv6hdr *ipv6h; 600 601 if (tunnel->dev->priv_flags & IFF_ISATAP) { 602 if (!isatap_chksrc(skb, iph, tunnel)) 603 return true; 604 605 return false; 606 } 607 608 if (tunnel->dev->flags & IFF_POINTOPOINT) 609 return false; 610 611 ipv6h = ipv6_hdr(skb); 612 613 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) { 614 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 615 &iph->saddr, &ipv6h->saddr, 616 &iph->daddr, &ipv6h->daddr); 617 return true; 618 } 619 620 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr))) 621 return false; 622 623 if (only_dnatted(tunnel, &ipv6h->daddr)) 624 return false; 625 626 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 627 &iph->saddr, &ipv6h->saddr, 628 &iph->daddr, &ipv6h->daddr); 629 return true; 630 } 631 632 static int ipip6_rcv(struct sk_buff *skb) 633 { 634 const struct iphdr *iph = ip_hdr(skb); 635 struct ip_tunnel *tunnel; 636 int err; 637 638 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 639 iph->saddr, iph->daddr); 640 if (tunnel) { 641 struct pcpu_sw_netstats *tstats; 642 643 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 && 644 tunnel->parms.iph.protocol != 0) 645 goto out; 646 647 skb->mac_header = skb->network_header; 648 skb_reset_network_header(skb); 649 IPCB(skb)->flags = 0; 650 skb->dev = tunnel->dev; 651 652 if (packet_is_spoofed(skb, iph, tunnel)) { 653 tunnel->dev->stats.rx_errors++; 654 goto out; 655 } 656 657 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6), 658 !net_eq(tunnel->net, dev_net(tunnel->dev)))) 659 goto out; 660 661 err = IP_ECN_decapsulate(iph, skb); 662 if (unlikely(err)) { 663 if (log_ecn_error) 664 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 665 &iph->saddr, iph->tos); 666 if (err > 1) { 667 ++tunnel->dev->stats.rx_frame_errors; 668 ++tunnel->dev->stats.rx_errors; 669 goto out; 670 } 671 } 672 673 tstats = this_cpu_ptr(tunnel->dev->tstats); 674 u64_stats_update_begin(&tstats->syncp); 675 tstats->rx_packets++; 676 tstats->rx_bytes += skb->len; 677 u64_stats_update_end(&tstats->syncp); 678 679 netif_rx(skb); 680 681 return 0; 682 } 683 684 /* no tunnel matched, let upstream know, ipsec may handle it */ 685 return 1; 686 out: 687 kfree_skb(skb); 688 return 0; 689 } 690 691 static const struct tnl_ptk_info ipip_tpi = { 692 /* no tunnel info required for ipip. */ 693 .proto = htons(ETH_P_IP), 694 }; 695 696 #if IS_ENABLED(CONFIG_MPLS) 697 static const struct tnl_ptk_info mplsip_tpi = { 698 /* no tunnel info required for mplsip. */ 699 .proto = htons(ETH_P_MPLS_UC), 700 }; 701 #endif 702 703 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto) 704 { 705 const struct iphdr *iph; 706 struct ip_tunnel *tunnel; 707 708 iph = ip_hdr(skb); 709 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 710 iph->saddr, iph->daddr); 711 if (tunnel) { 712 const struct tnl_ptk_info *tpi; 713 714 if (tunnel->parms.iph.protocol != ipproto && 715 tunnel->parms.iph.protocol != 0) 716 goto drop; 717 718 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) 719 goto drop; 720 #if IS_ENABLED(CONFIG_MPLS) 721 if (ipproto == IPPROTO_MPLS) 722 tpi = &mplsip_tpi; 723 else 724 #endif 725 tpi = &ipip_tpi; 726 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 727 goto drop; 728 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 729 } 730 731 return 1; 732 733 drop: 734 kfree_skb(skb); 735 return 0; 736 } 737 738 static int ipip_rcv(struct sk_buff *skb) 739 { 740 return sit_tunnel_rcv(skb, IPPROTO_IPIP); 741 } 742 743 #if IS_ENABLED(CONFIG_MPLS) 744 static int mplsip_rcv(struct sk_buff *skb) 745 { 746 return sit_tunnel_rcv(skb, IPPROTO_MPLS); 747 } 748 #endif 749 750 /* 751 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function 752 * stores the embedded IPv4 address in v4dst and returns true. 753 */ 754 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, 755 __be32 *v4dst) 756 { 757 #ifdef CONFIG_IPV6_SIT_6RD 758 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, 759 tunnel->ip6rd.prefixlen)) { 760 unsigned int pbw0, pbi0; 761 int pbi1; 762 u32 d; 763 764 pbw0 = tunnel->ip6rd.prefixlen >> 5; 765 pbi0 = tunnel->ip6rd.prefixlen & 0x1f; 766 767 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> 768 tunnel->ip6rd.relay_prefixlen; 769 770 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; 771 if (pbi1 > 0) 772 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> 773 (32 - pbi1); 774 775 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d); 776 return true; 777 } 778 #else 779 if (v6dst->s6_addr16[0] == htons(0x2002)) { 780 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ 781 memcpy(v4dst, &v6dst->s6_addr16[1], 4); 782 return true; 783 } 784 #endif 785 return false; 786 } 787 788 static inline __be32 try_6rd(struct ip_tunnel *tunnel, 789 const struct in6_addr *v6dst) 790 { 791 __be32 dst = 0; 792 check_6rd(tunnel, v6dst, &dst); 793 return dst; 794 } 795 796 /* 797 * This function assumes it is being called from dev_queue_xmit() 798 * and that skb is filled properly by that function. 799 */ 800 801 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, 802 struct net_device *dev) 803 { 804 struct ip_tunnel *tunnel = netdev_priv(dev); 805 const struct iphdr *tiph = &tunnel->parms.iph; 806 const struct ipv6hdr *iph6 = ipv6_hdr(skb); 807 u8 tos = tunnel->parms.iph.tos; 808 __be16 df = tiph->frag_off; 809 struct rtable *rt; /* Route to the other host */ 810 struct net_device *tdev; /* Device to other host */ 811 unsigned int max_headroom; /* The extra header space needed */ 812 __be32 dst = tiph->daddr; 813 struct flowi4 fl4; 814 int mtu; 815 const struct in6_addr *addr6; 816 int addr_type; 817 u8 ttl; 818 u8 protocol = IPPROTO_IPV6; 819 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 820 821 if (tos == 1) 822 tos = ipv6_get_dsfield(iph6); 823 824 /* ISATAP (RFC4214) - must come before 6to4 */ 825 if (dev->priv_flags & IFF_ISATAP) { 826 struct neighbour *neigh = NULL; 827 bool do_tx_error = false; 828 829 if (skb_dst(skb)) 830 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 831 832 if (!neigh) { 833 net_dbg_ratelimited("nexthop == NULL\n"); 834 goto tx_error; 835 } 836 837 addr6 = (const struct in6_addr *)&neigh->primary_key; 838 addr_type = ipv6_addr_type(addr6); 839 840 if ((addr_type & IPV6_ADDR_UNICAST) && 841 ipv6_addr_is_isatap(addr6)) 842 dst = addr6->s6_addr32[3]; 843 else 844 do_tx_error = true; 845 846 neigh_release(neigh); 847 if (do_tx_error) 848 goto tx_error; 849 } 850 851 if (!dst) 852 dst = try_6rd(tunnel, &iph6->daddr); 853 854 if (!dst) { 855 struct neighbour *neigh = NULL; 856 bool do_tx_error = false; 857 858 if (skb_dst(skb)) 859 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 860 861 if (!neigh) { 862 net_dbg_ratelimited("nexthop == NULL\n"); 863 goto tx_error; 864 } 865 866 addr6 = (const struct in6_addr *)&neigh->primary_key; 867 addr_type = ipv6_addr_type(addr6); 868 869 if (addr_type == IPV6_ADDR_ANY) { 870 addr6 = &ipv6_hdr(skb)->daddr; 871 addr_type = ipv6_addr_type(addr6); 872 } 873 874 if ((addr_type & IPV6_ADDR_COMPATv4) != 0) 875 dst = addr6->s6_addr32[3]; 876 else 877 do_tx_error = true; 878 879 neigh_release(neigh); 880 if (do_tx_error) 881 goto tx_error; 882 } 883 884 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark, 885 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6, 886 0, dst, tiph->saddr, 0, 0, 887 sock_net_uid(tunnel->net, NULL)); 888 rt = ip_route_output_flow(tunnel->net, &fl4, NULL); 889 890 if (IS_ERR(rt)) { 891 dev->stats.tx_carrier_errors++; 892 goto tx_error_icmp; 893 } 894 if (rt->rt_type != RTN_UNICAST) { 895 ip_rt_put(rt); 896 dev->stats.tx_carrier_errors++; 897 goto tx_error_icmp; 898 } 899 tdev = rt->dst.dev; 900 901 if (tdev == dev) { 902 ip_rt_put(rt); 903 dev->stats.collisions++; 904 goto tx_error; 905 } 906 907 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) { 908 ip_rt_put(rt); 909 goto tx_error; 910 } 911 912 if (df) { 913 mtu = dst_mtu(&rt->dst) - t_hlen; 914 915 if (mtu < 68) { 916 dev->stats.collisions++; 917 ip_rt_put(rt); 918 goto tx_error; 919 } 920 921 if (mtu < IPV6_MIN_MTU) { 922 mtu = IPV6_MIN_MTU; 923 df = 0; 924 } 925 926 if (tunnel->parms.iph.daddr && skb_dst(skb)) 927 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu); 928 929 if (skb->len > mtu && !skb_is_gso(skb)) { 930 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 931 ip_rt_put(rt); 932 goto tx_error; 933 } 934 } 935 936 if (tunnel->err_count > 0) { 937 if (time_before(jiffies, 938 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 939 tunnel->err_count--; 940 dst_link_failure(skb); 941 } else 942 tunnel->err_count = 0; 943 } 944 945 /* 946 * Okay, now see if we can stuff it in the buffer as-is. 947 */ 948 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen; 949 950 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 951 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 952 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 953 if (!new_skb) { 954 ip_rt_put(rt); 955 dev->stats.tx_dropped++; 956 kfree_skb(skb); 957 return NETDEV_TX_OK; 958 } 959 if (skb->sk) 960 skb_set_owner_w(new_skb, skb->sk); 961 dev_kfree_skb(skb); 962 skb = new_skb; 963 iph6 = ipv6_hdr(skb); 964 } 965 ttl = tiph->ttl; 966 if (ttl == 0) 967 ttl = iph6->hop_limit; 968 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); 969 970 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) { 971 ip_rt_put(rt); 972 goto tx_error; 973 } 974 975 skb_set_inner_ipproto(skb, IPPROTO_IPV6); 976 977 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl, 978 df, !net_eq(tunnel->net, dev_net(dev))); 979 return NETDEV_TX_OK; 980 981 tx_error_icmp: 982 dst_link_failure(skb); 983 tx_error: 984 kfree_skb(skb); 985 dev->stats.tx_errors++; 986 return NETDEV_TX_OK; 987 } 988 989 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, 990 struct net_device *dev, u8 ipproto) 991 { 992 struct ip_tunnel *tunnel = netdev_priv(dev); 993 const struct iphdr *tiph = &tunnel->parms.iph; 994 995 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) 996 goto tx_error; 997 998 skb_set_inner_ipproto(skb, ipproto); 999 1000 ip_tunnel_xmit(skb, dev, tiph, ipproto); 1001 return NETDEV_TX_OK; 1002 tx_error: 1003 kfree_skb(skb); 1004 dev->stats.tx_errors++; 1005 return NETDEV_TX_OK; 1006 } 1007 1008 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb, 1009 struct net_device *dev) 1010 { 1011 switch (skb->protocol) { 1012 case htons(ETH_P_IP): 1013 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP); 1014 break; 1015 case htons(ETH_P_IPV6): 1016 ipip6_tunnel_xmit(skb, dev); 1017 break; 1018 #if IS_ENABLED(CONFIG_MPLS) 1019 case htons(ETH_P_MPLS_UC): 1020 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS); 1021 break; 1022 #endif 1023 default: 1024 goto tx_err; 1025 } 1026 1027 return NETDEV_TX_OK; 1028 1029 tx_err: 1030 dev->stats.tx_errors++; 1031 kfree_skb(skb); 1032 return NETDEV_TX_OK; 1033 1034 } 1035 1036 static void ipip6_tunnel_bind_dev(struct net_device *dev) 1037 { 1038 struct net_device *tdev = NULL; 1039 struct ip_tunnel *tunnel; 1040 const struct iphdr *iph; 1041 struct flowi4 fl4; 1042 1043 tunnel = netdev_priv(dev); 1044 iph = &tunnel->parms.iph; 1045 1046 if (iph->daddr) { 1047 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4, 1048 NULL, 1049 iph->daddr, iph->saddr, 1050 0, 0, 1051 IPPROTO_IPV6, 1052 RT_TOS(iph->tos), 1053 tunnel->parms.link); 1054 1055 if (!IS_ERR(rt)) { 1056 tdev = rt->dst.dev; 1057 ip_rt_put(rt); 1058 } 1059 dev->flags |= IFF_POINTOPOINT; 1060 } 1061 1062 if (!tdev && tunnel->parms.link) 1063 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link); 1064 1065 if (tdev) { 1066 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1067 1068 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); 1069 dev->mtu = tdev->mtu - t_hlen; 1070 if (dev->mtu < IPV6_MIN_MTU) 1071 dev->mtu = IPV6_MIN_MTU; 1072 } 1073 } 1074 1075 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p, 1076 __u32 fwmark) 1077 { 1078 struct net *net = t->net; 1079 struct sit_net *sitn = net_generic(net, sit_net_id); 1080 1081 ipip6_tunnel_unlink(sitn, t); 1082 synchronize_net(); 1083 t->parms.iph.saddr = p->iph.saddr; 1084 t->parms.iph.daddr = p->iph.daddr; 1085 memcpy(t->dev->dev_addr, &p->iph.saddr, 4); 1086 memcpy(t->dev->broadcast, &p->iph.daddr, 4); 1087 ipip6_tunnel_link(sitn, t); 1088 t->parms.iph.ttl = p->iph.ttl; 1089 t->parms.iph.tos = p->iph.tos; 1090 if (t->parms.link != p->link || t->fwmark != fwmark) { 1091 t->parms.link = p->link; 1092 t->fwmark = fwmark; 1093 ipip6_tunnel_bind_dev(t->dev); 1094 } 1095 dst_cache_reset(&t->dst_cache); 1096 netdev_state_change(t->dev); 1097 } 1098 1099 #ifdef CONFIG_IPV6_SIT_6RD 1100 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t, 1101 struct ip_tunnel_6rd *ip6rd) 1102 { 1103 struct in6_addr prefix; 1104 __be32 relay_prefix; 1105 1106 if (ip6rd->relay_prefixlen > 32 || 1107 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64) 1108 return -EINVAL; 1109 1110 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen); 1111 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix)) 1112 return -EINVAL; 1113 if (ip6rd->relay_prefixlen) 1114 relay_prefix = ip6rd->relay_prefix & 1115 htonl(0xffffffffUL << 1116 (32 - ip6rd->relay_prefixlen)); 1117 else 1118 relay_prefix = 0; 1119 if (relay_prefix != ip6rd->relay_prefix) 1120 return -EINVAL; 1121 1122 t->ip6rd.prefix = prefix; 1123 t->ip6rd.relay_prefix = relay_prefix; 1124 t->ip6rd.prefixlen = ip6rd->prefixlen; 1125 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen; 1126 dst_cache_reset(&t->dst_cache); 1127 netdev_state_change(t->dev); 1128 return 0; 1129 } 1130 #endif 1131 1132 static bool ipip6_valid_ip_proto(u8 ipproto) 1133 { 1134 return ipproto == IPPROTO_IPV6 || 1135 ipproto == IPPROTO_IPIP || 1136 #if IS_ENABLED(CONFIG_MPLS) 1137 ipproto == IPPROTO_MPLS || 1138 #endif 1139 ipproto == 0; 1140 } 1141 1142 static int 1143 ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 1144 { 1145 int err = 0; 1146 struct ip_tunnel_parm p; 1147 struct ip_tunnel_prl prl; 1148 struct ip_tunnel *t = netdev_priv(dev); 1149 struct net *net = t->net; 1150 struct sit_net *sitn = net_generic(net, sit_net_id); 1151 #ifdef CONFIG_IPV6_SIT_6RD 1152 struct ip_tunnel_6rd ip6rd; 1153 #endif 1154 1155 switch (cmd) { 1156 case SIOCGETTUNNEL: 1157 #ifdef CONFIG_IPV6_SIT_6RD 1158 case SIOCGET6RD: 1159 #endif 1160 if (dev == sitn->fb_tunnel_dev) { 1161 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1162 err = -EFAULT; 1163 break; 1164 } 1165 t = ipip6_tunnel_locate(net, &p, 0); 1166 if (!t) 1167 t = netdev_priv(dev); 1168 } 1169 1170 err = -EFAULT; 1171 if (cmd == SIOCGETTUNNEL) { 1172 memcpy(&p, &t->parms, sizeof(p)); 1173 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, 1174 sizeof(p))) 1175 goto done; 1176 #ifdef CONFIG_IPV6_SIT_6RD 1177 } else { 1178 ip6rd.prefix = t->ip6rd.prefix; 1179 ip6rd.relay_prefix = t->ip6rd.relay_prefix; 1180 ip6rd.prefixlen = t->ip6rd.prefixlen; 1181 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; 1182 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, 1183 sizeof(ip6rd))) 1184 goto done; 1185 #endif 1186 } 1187 err = 0; 1188 break; 1189 1190 case SIOCADDTUNNEL: 1191 case SIOCCHGTUNNEL: 1192 err = -EPERM; 1193 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1194 goto done; 1195 1196 err = -EFAULT; 1197 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1198 goto done; 1199 1200 err = -EINVAL; 1201 if (!ipip6_valid_ip_proto(p.iph.protocol)) 1202 goto done; 1203 if (p.iph.version != 4 || 1204 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) 1205 goto done; 1206 if (p.iph.ttl) 1207 p.iph.frag_off |= htons(IP_DF); 1208 1209 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); 1210 1211 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1212 if (t) { 1213 if (t->dev != dev) { 1214 err = -EEXIST; 1215 break; 1216 } 1217 } else { 1218 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || 1219 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { 1220 err = -EINVAL; 1221 break; 1222 } 1223 t = netdev_priv(dev); 1224 } 1225 1226 ipip6_tunnel_update(t, &p, t->fwmark); 1227 } 1228 1229 if (t) { 1230 err = 0; 1231 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) 1232 err = -EFAULT; 1233 } else 1234 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1235 break; 1236 1237 case SIOCDELTUNNEL: 1238 err = -EPERM; 1239 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1240 goto done; 1241 1242 if (dev == sitn->fb_tunnel_dev) { 1243 err = -EFAULT; 1244 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1245 goto done; 1246 err = -ENOENT; 1247 t = ipip6_tunnel_locate(net, &p, 0); 1248 if (!t) 1249 goto done; 1250 err = -EPERM; 1251 if (t == netdev_priv(sitn->fb_tunnel_dev)) 1252 goto done; 1253 dev = t->dev; 1254 } 1255 unregister_netdevice(dev); 1256 err = 0; 1257 break; 1258 1259 case SIOCGETPRL: 1260 err = -EINVAL; 1261 if (dev == sitn->fb_tunnel_dev) 1262 goto done; 1263 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data); 1264 break; 1265 1266 case SIOCADDPRL: 1267 case SIOCDELPRL: 1268 case SIOCCHGPRL: 1269 err = -EPERM; 1270 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1271 goto done; 1272 err = -EINVAL; 1273 if (dev == sitn->fb_tunnel_dev) 1274 goto done; 1275 err = -EFAULT; 1276 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl))) 1277 goto done; 1278 1279 switch (cmd) { 1280 case SIOCDELPRL: 1281 err = ipip6_tunnel_del_prl(t, &prl); 1282 break; 1283 case SIOCADDPRL: 1284 case SIOCCHGPRL: 1285 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); 1286 break; 1287 } 1288 dst_cache_reset(&t->dst_cache); 1289 netdev_state_change(dev); 1290 break; 1291 1292 #ifdef CONFIG_IPV6_SIT_6RD 1293 case SIOCADD6RD: 1294 case SIOCCHG6RD: 1295 case SIOCDEL6RD: 1296 err = -EPERM; 1297 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1298 goto done; 1299 1300 err = -EFAULT; 1301 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, 1302 sizeof(ip6rd))) 1303 goto done; 1304 1305 if (cmd != SIOCDEL6RD) { 1306 err = ipip6_tunnel_update_6rd(t, &ip6rd); 1307 if (err < 0) 1308 goto done; 1309 } else 1310 ipip6_tunnel_clone_6rd(dev, sitn); 1311 1312 err = 0; 1313 break; 1314 #endif 1315 1316 default: 1317 err = -EINVAL; 1318 } 1319 1320 done: 1321 return err; 1322 } 1323 1324 static const struct net_device_ops ipip6_netdev_ops = { 1325 .ndo_init = ipip6_tunnel_init, 1326 .ndo_uninit = ipip6_tunnel_uninit, 1327 .ndo_start_xmit = sit_tunnel_xmit, 1328 .ndo_do_ioctl = ipip6_tunnel_ioctl, 1329 .ndo_get_stats64 = ip_tunnel_get_stats64, 1330 .ndo_get_iflink = ip_tunnel_get_iflink, 1331 }; 1332 1333 static void ipip6_dev_free(struct net_device *dev) 1334 { 1335 struct ip_tunnel *tunnel = netdev_priv(dev); 1336 1337 dst_cache_destroy(&tunnel->dst_cache); 1338 free_percpu(dev->tstats); 1339 } 1340 1341 #define SIT_FEATURES (NETIF_F_SG | \ 1342 NETIF_F_FRAGLIST | \ 1343 NETIF_F_HIGHDMA | \ 1344 NETIF_F_GSO_SOFTWARE | \ 1345 NETIF_F_HW_CSUM) 1346 1347 static void ipip6_tunnel_setup(struct net_device *dev) 1348 { 1349 struct ip_tunnel *tunnel = netdev_priv(dev); 1350 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1351 1352 dev->netdev_ops = &ipip6_netdev_ops; 1353 dev->needs_free_netdev = true; 1354 dev->priv_destructor = ipip6_dev_free; 1355 1356 dev->type = ARPHRD_SIT; 1357 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1358 dev->mtu = ETH_DATA_LEN - t_hlen; 1359 dev->min_mtu = IPV6_MIN_MTU; 1360 dev->max_mtu = 0xFFF8 - t_hlen; 1361 dev->flags = IFF_NOARP; 1362 netif_keep_dst(dev); 1363 dev->addr_len = 4; 1364 dev->features |= NETIF_F_LLTX; 1365 dev->features |= SIT_FEATURES; 1366 dev->hw_features |= SIT_FEATURES; 1367 } 1368 1369 static int ipip6_tunnel_init(struct net_device *dev) 1370 { 1371 struct ip_tunnel *tunnel = netdev_priv(dev); 1372 int err; 1373 1374 tunnel->dev = dev; 1375 tunnel->net = dev_net(dev); 1376 strcpy(tunnel->parms.name, dev->name); 1377 1378 ipip6_tunnel_bind_dev(dev); 1379 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1380 if (!dev->tstats) 1381 return -ENOMEM; 1382 1383 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1384 if (err) { 1385 free_percpu(dev->tstats); 1386 dev->tstats = NULL; 1387 return err; 1388 } 1389 1390 return 0; 1391 } 1392 1393 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev) 1394 { 1395 struct ip_tunnel *tunnel = netdev_priv(dev); 1396 struct iphdr *iph = &tunnel->parms.iph; 1397 struct net *net = dev_net(dev); 1398 struct sit_net *sitn = net_generic(net, sit_net_id); 1399 1400 iph->version = 4; 1401 iph->protocol = IPPROTO_IPV6; 1402 iph->ihl = 5; 1403 iph->ttl = 64; 1404 1405 dev_hold(dev); 1406 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); 1407 } 1408 1409 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[], 1410 struct netlink_ext_ack *extack) 1411 { 1412 u8 proto; 1413 1414 if (!data || !data[IFLA_IPTUN_PROTO]) 1415 return 0; 1416 1417 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1418 if (!ipip6_valid_ip_proto(proto)) 1419 return -EINVAL; 1420 1421 return 0; 1422 } 1423 1424 static void ipip6_netlink_parms(struct nlattr *data[], 1425 struct ip_tunnel_parm *parms, 1426 __u32 *fwmark) 1427 { 1428 memset(parms, 0, sizeof(*parms)); 1429 1430 parms->iph.version = 4; 1431 parms->iph.protocol = IPPROTO_IPV6; 1432 parms->iph.ihl = 5; 1433 parms->iph.ttl = 64; 1434 1435 if (!data) 1436 return; 1437 1438 if (data[IFLA_IPTUN_LINK]) 1439 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1440 1441 if (data[IFLA_IPTUN_LOCAL]) 1442 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]); 1443 1444 if (data[IFLA_IPTUN_REMOTE]) 1445 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]); 1446 1447 if (data[IFLA_IPTUN_TTL]) { 1448 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]); 1449 if (parms->iph.ttl) 1450 parms->iph.frag_off = htons(IP_DF); 1451 } 1452 1453 if (data[IFLA_IPTUN_TOS]) 1454 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]); 1455 1456 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC])) 1457 parms->iph.frag_off = htons(IP_DF); 1458 1459 if (data[IFLA_IPTUN_FLAGS]) 1460 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]); 1461 1462 if (data[IFLA_IPTUN_PROTO]) 1463 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1464 1465 if (data[IFLA_IPTUN_FWMARK]) 1466 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 1467 } 1468 1469 /* This function returns true when ENCAP attributes are present in the nl msg */ 1470 static bool ipip6_netlink_encap_parms(struct nlattr *data[], 1471 struct ip_tunnel_encap *ipencap) 1472 { 1473 bool ret = false; 1474 1475 memset(ipencap, 0, sizeof(*ipencap)); 1476 1477 if (!data) 1478 return ret; 1479 1480 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 1481 ret = true; 1482 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 1483 } 1484 1485 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 1486 ret = true; 1487 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 1488 } 1489 1490 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 1491 ret = true; 1492 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 1493 } 1494 1495 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 1496 ret = true; 1497 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 1498 } 1499 1500 return ret; 1501 } 1502 1503 #ifdef CONFIG_IPV6_SIT_6RD 1504 /* This function returns true when 6RD attributes are present in the nl msg */ 1505 static bool ipip6_netlink_6rd_parms(struct nlattr *data[], 1506 struct ip_tunnel_6rd *ip6rd) 1507 { 1508 bool ret = false; 1509 memset(ip6rd, 0, sizeof(*ip6rd)); 1510 1511 if (!data) 1512 return ret; 1513 1514 if (data[IFLA_IPTUN_6RD_PREFIX]) { 1515 ret = true; 1516 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]); 1517 } 1518 1519 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) { 1520 ret = true; 1521 ip6rd->relay_prefix = 1522 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]); 1523 } 1524 1525 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) { 1526 ret = true; 1527 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]); 1528 } 1529 1530 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) { 1531 ret = true; 1532 ip6rd->relay_prefixlen = 1533 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]); 1534 } 1535 1536 return ret; 1537 } 1538 #endif 1539 1540 static int ipip6_newlink(struct net *src_net, struct net_device *dev, 1541 struct nlattr *tb[], struct nlattr *data[], 1542 struct netlink_ext_ack *extack) 1543 { 1544 struct net *net = dev_net(dev); 1545 struct ip_tunnel *nt; 1546 struct ip_tunnel_encap ipencap; 1547 #ifdef CONFIG_IPV6_SIT_6RD 1548 struct ip_tunnel_6rd ip6rd; 1549 #endif 1550 int err; 1551 1552 nt = netdev_priv(dev); 1553 1554 if (ipip6_netlink_encap_parms(data, &ipencap)) { 1555 err = ip_tunnel_encap_setup(nt, &ipencap); 1556 if (err < 0) 1557 return err; 1558 } 1559 1560 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark); 1561 1562 if (ipip6_tunnel_locate(net, &nt->parms, 0)) 1563 return -EEXIST; 1564 1565 err = ipip6_tunnel_create(dev); 1566 if (err < 0) 1567 return err; 1568 1569 #ifdef CONFIG_IPV6_SIT_6RD 1570 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1571 err = ipip6_tunnel_update_6rd(nt, &ip6rd); 1572 #endif 1573 1574 return err; 1575 } 1576 1577 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[], 1578 struct nlattr *data[], 1579 struct netlink_ext_ack *extack) 1580 { 1581 struct ip_tunnel *t = netdev_priv(dev); 1582 struct ip_tunnel_parm p; 1583 struct ip_tunnel_encap ipencap; 1584 struct net *net = t->net; 1585 struct sit_net *sitn = net_generic(net, sit_net_id); 1586 #ifdef CONFIG_IPV6_SIT_6RD 1587 struct ip_tunnel_6rd ip6rd; 1588 #endif 1589 __u32 fwmark = t->fwmark; 1590 int err; 1591 1592 if (dev == sitn->fb_tunnel_dev) 1593 return -EINVAL; 1594 1595 if (ipip6_netlink_encap_parms(data, &ipencap)) { 1596 err = ip_tunnel_encap_setup(t, &ipencap); 1597 if (err < 0) 1598 return err; 1599 } 1600 1601 ipip6_netlink_parms(data, &p, &fwmark); 1602 1603 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) || 1604 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr)) 1605 return -EINVAL; 1606 1607 t = ipip6_tunnel_locate(net, &p, 0); 1608 1609 if (t) { 1610 if (t->dev != dev) 1611 return -EEXIST; 1612 } else 1613 t = netdev_priv(dev); 1614 1615 ipip6_tunnel_update(t, &p, fwmark); 1616 1617 #ifdef CONFIG_IPV6_SIT_6RD 1618 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1619 return ipip6_tunnel_update_6rd(t, &ip6rd); 1620 #endif 1621 1622 return 0; 1623 } 1624 1625 static size_t ipip6_get_size(const struct net_device *dev) 1626 { 1627 return 1628 /* IFLA_IPTUN_LINK */ 1629 nla_total_size(4) + 1630 /* IFLA_IPTUN_LOCAL */ 1631 nla_total_size(4) + 1632 /* IFLA_IPTUN_REMOTE */ 1633 nla_total_size(4) + 1634 /* IFLA_IPTUN_TTL */ 1635 nla_total_size(1) + 1636 /* IFLA_IPTUN_TOS */ 1637 nla_total_size(1) + 1638 /* IFLA_IPTUN_PMTUDISC */ 1639 nla_total_size(1) + 1640 /* IFLA_IPTUN_FLAGS */ 1641 nla_total_size(2) + 1642 /* IFLA_IPTUN_PROTO */ 1643 nla_total_size(1) + 1644 #ifdef CONFIG_IPV6_SIT_6RD 1645 /* IFLA_IPTUN_6RD_PREFIX */ 1646 nla_total_size(sizeof(struct in6_addr)) + 1647 /* IFLA_IPTUN_6RD_RELAY_PREFIX */ 1648 nla_total_size(4) + 1649 /* IFLA_IPTUN_6RD_PREFIXLEN */ 1650 nla_total_size(2) + 1651 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */ 1652 nla_total_size(2) + 1653 #endif 1654 /* IFLA_IPTUN_ENCAP_TYPE */ 1655 nla_total_size(2) + 1656 /* IFLA_IPTUN_ENCAP_FLAGS */ 1657 nla_total_size(2) + 1658 /* IFLA_IPTUN_ENCAP_SPORT */ 1659 nla_total_size(2) + 1660 /* IFLA_IPTUN_ENCAP_DPORT */ 1661 nla_total_size(2) + 1662 /* IFLA_IPTUN_FWMARK */ 1663 nla_total_size(4) + 1664 0; 1665 } 1666 1667 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1668 { 1669 struct ip_tunnel *tunnel = netdev_priv(dev); 1670 struct ip_tunnel_parm *parm = &tunnel->parms; 1671 1672 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1673 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) || 1674 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) || 1675 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) || 1676 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) || 1677 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC, 1678 !!(parm->iph.frag_off & htons(IP_DF))) || 1679 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) || 1680 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) || 1681 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark)) 1682 goto nla_put_failure; 1683 1684 #ifdef CONFIG_IPV6_SIT_6RD 1685 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX, 1686 &tunnel->ip6rd.prefix) || 1687 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX, 1688 tunnel->ip6rd.relay_prefix) || 1689 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN, 1690 tunnel->ip6rd.prefixlen) || 1691 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN, 1692 tunnel->ip6rd.relay_prefixlen)) 1693 goto nla_put_failure; 1694 #endif 1695 1696 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, 1697 tunnel->encap.type) || 1698 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, 1699 tunnel->encap.sport) || 1700 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, 1701 tunnel->encap.dport) || 1702 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, 1703 tunnel->encap.flags)) 1704 goto nla_put_failure; 1705 1706 return 0; 1707 1708 nla_put_failure: 1709 return -EMSGSIZE; 1710 } 1711 1712 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = { 1713 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1714 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 }, 1715 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 }, 1716 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1717 [IFLA_IPTUN_TOS] = { .type = NLA_U8 }, 1718 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 }, 1719 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 }, 1720 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1721 #ifdef CONFIG_IPV6_SIT_6RD 1722 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) }, 1723 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 }, 1724 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 }, 1725 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 }, 1726 #endif 1727 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 1728 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 1729 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 1730 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 1731 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 1732 }; 1733 1734 static void ipip6_dellink(struct net_device *dev, struct list_head *head) 1735 { 1736 struct net *net = dev_net(dev); 1737 struct sit_net *sitn = net_generic(net, sit_net_id); 1738 1739 if (dev != sitn->fb_tunnel_dev) 1740 unregister_netdevice_queue(dev, head); 1741 } 1742 1743 static struct rtnl_link_ops sit_link_ops __read_mostly = { 1744 .kind = "sit", 1745 .maxtype = IFLA_IPTUN_MAX, 1746 .policy = ipip6_policy, 1747 .priv_size = sizeof(struct ip_tunnel), 1748 .setup = ipip6_tunnel_setup, 1749 .validate = ipip6_validate, 1750 .newlink = ipip6_newlink, 1751 .changelink = ipip6_changelink, 1752 .get_size = ipip6_get_size, 1753 .fill_info = ipip6_fill_info, 1754 .dellink = ipip6_dellink, 1755 .get_link_net = ip_tunnel_get_link_net, 1756 }; 1757 1758 static struct xfrm_tunnel sit_handler __read_mostly = { 1759 .handler = ipip6_rcv, 1760 .err_handler = ipip6_err, 1761 .priority = 1, 1762 }; 1763 1764 static struct xfrm_tunnel ipip_handler __read_mostly = { 1765 .handler = ipip_rcv, 1766 .err_handler = ipip6_err, 1767 .priority = 2, 1768 }; 1769 1770 #if IS_ENABLED(CONFIG_MPLS) 1771 static struct xfrm_tunnel mplsip_handler __read_mostly = { 1772 .handler = mplsip_rcv, 1773 .err_handler = ipip6_err, 1774 .priority = 2, 1775 }; 1776 #endif 1777 1778 static void __net_exit sit_destroy_tunnels(struct net *net, 1779 struct list_head *head) 1780 { 1781 struct sit_net *sitn = net_generic(net, sit_net_id); 1782 struct net_device *dev, *aux; 1783 int prio; 1784 1785 for_each_netdev_safe(net, dev, aux) 1786 if (dev->rtnl_link_ops == &sit_link_ops) 1787 unregister_netdevice_queue(dev, head); 1788 1789 for (prio = 1; prio < 4; prio++) { 1790 int h; 1791 for (h = 0; h < IP6_SIT_HASH_SIZE; h++) { 1792 struct ip_tunnel *t; 1793 1794 t = rtnl_dereference(sitn->tunnels[prio][h]); 1795 while (t) { 1796 /* If dev is in the same netns, it has already 1797 * been added to the list by the previous loop. 1798 */ 1799 if (!net_eq(dev_net(t->dev), net)) 1800 unregister_netdevice_queue(t->dev, 1801 head); 1802 t = rtnl_dereference(t->next); 1803 } 1804 } 1805 } 1806 } 1807 1808 static int __net_init sit_init_net(struct net *net) 1809 { 1810 struct sit_net *sitn = net_generic(net, sit_net_id); 1811 struct ip_tunnel *t; 1812 int err; 1813 1814 sitn->tunnels[0] = sitn->tunnels_wc; 1815 sitn->tunnels[1] = sitn->tunnels_l; 1816 sitn->tunnels[2] = sitn->tunnels_r; 1817 sitn->tunnels[3] = sitn->tunnels_r_l; 1818 1819 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", 1820 NET_NAME_UNKNOWN, 1821 ipip6_tunnel_setup); 1822 if (!sitn->fb_tunnel_dev) { 1823 err = -ENOMEM; 1824 goto err_alloc_dev; 1825 } 1826 dev_net_set(sitn->fb_tunnel_dev, net); 1827 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops; 1828 /* FB netdevice is special: we have one, and only one per netns. 1829 * Allowing to move it to another netns is clearly unsafe. 1830 */ 1831 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1832 1833 err = register_netdev(sitn->fb_tunnel_dev); 1834 if (err) 1835 goto err_reg_dev; 1836 1837 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); 1838 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); 1839 1840 t = netdev_priv(sitn->fb_tunnel_dev); 1841 1842 strcpy(t->parms.name, sitn->fb_tunnel_dev->name); 1843 return 0; 1844 1845 err_reg_dev: 1846 ipip6_dev_free(sitn->fb_tunnel_dev); 1847 err_alloc_dev: 1848 return err; 1849 } 1850 1851 static void __net_exit sit_exit_net(struct net *net) 1852 { 1853 LIST_HEAD(list); 1854 1855 rtnl_lock(); 1856 sit_destroy_tunnels(net, &list); 1857 unregister_netdevice_many(&list); 1858 rtnl_unlock(); 1859 } 1860 1861 static struct pernet_operations sit_net_ops = { 1862 .init = sit_init_net, 1863 .exit = sit_exit_net, 1864 .id = &sit_net_id, 1865 .size = sizeof(struct sit_net), 1866 }; 1867 1868 static void __exit sit_cleanup(void) 1869 { 1870 rtnl_link_unregister(&sit_link_ops); 1871 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1872 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1873 #if IS_ENABLED(CONFIG_MPLS) 1874 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS); 1875 #endif 1876 1877 unregister_pernet_device(&sit_net_ops); 1878 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1879 } 1880 1881 static int __init sit_init(void) 1882 { 1883 int err; 1884 1885 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n"); 1886 1887 err = register_pernet_device(&sit_net_ops); 1888 if (err < 0) 1889 return err; 1890 err = xfrm4_tunnel_register(&sit_handler, AF_INET6); 1891 if (err < 0) { 1892 pr_info("%s: can't register ip6ip4\n", __func__); 1893 goto xfrm_tunnel_failed; 1894 } 1895 err = xfrm4_tunnel_register(&ipip_handler, AF_INET); 1896 if (err < 0) { 1897 pr_info("%s: can't register ip4ip4\n", __func__); 1898 goto xfrm_tunnel4_failed; 1899 } 1900 #if IS_ENABLED(CONFIG_MPLS) 1901 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS); 1902 if (err < 0) { 1903 pr_info("%s: can't register mplsip\n", __func__); 1904 goto xfrm_tunnel_mpls_failed; 1905 } 1906 #endif 1907 err = rtnl_link_register(&sit_link_ops); 1908 if (err < 0) 1909 goto rtnl_link_failed; 1910 1911 out: 1912 return err; 1913 1914 rtnl_link_failed: 1915 #if IS_ENABLED(CONFIG_MPLS) 1916 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS); 1917 xfrm_tunnel_mpls_failed: 1918 #endif 1919 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1920 xfrm_tunnel4_failed: 1921 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1922 xfrm_tunnel_failed: 1923 unregister_pernet_device(&sit_net_ops); 1924 goto out; 1925 } 1926 1927 module_init(sit_init); 1928 module_exit(sit_cleanup); 1929 MODULE_LICENSE("GPL"); 1930 MODULE_ALIAS_RTNL_LINK("sit"); 1931 MODULE_ALIAS_NETDEV("sit0"); 1932