1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT) 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 9 * 10 * Changes: 11 * Roger Venning <r.venning@telstra.com>: 6to4 support 12 * Nate Thompson <nate@thebog.net>: 6to4 support 13 * Fred Templin <fred.l.templin@boeing.com>: isatap support 14 */ 15 16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 17 18 #include <linux/module.h> 19 #include <linux/capability.h> 20 #include <linux/errno.h> 21 #include <linux/types.h> 22 #include <linux/socket.h> 23 #include <linux/sockios.h> 24 #include <linux/net.h> 25 #include <linux/in6.h> 26 #include <linux/netdevice.h> 27 #include <linux/if_arp.h> 28 #include <linux/icmp.h> 29 #include <linux/slab.h> 30 #include <linux/uaccess.h> 31 #include <linux/init.h> 32 #include <linux/netfilter_ipv4.h> 33 #include <linux/if_ether.h> 34 35 #include <net/sock.h> 36 #include <net/snmp.h> 37 38 #include <net/ipv6.h> 39 #include <net/protocol.h> 40 #include <net/transp_v6.h> 41 #include <net/ip6_fib.h> 42 #include <net/ip6_route.h> 43 #include <net/ndisc.h> 44 #include <net/addrconf.h> 45 #include <net/ip.h> 46 #include <net/udp.h> 47 #include <net/icmp.h> 48 #include <net/ip_tunnels.h> 49 #include <net/inet_ecn.h> 50 #include <net/xfrm.h> 51 #include <net/dsfield.h> 52 #include <net/net_namespace.h> 53 #include <net/netns/generic.h> 54 55 /* 56 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c 57 58 For comments look at net/ipv4/ip_gre.c --ANK 59 */ 60 61 #define IP6_SIT_HASH_SIZE 16 62 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF) 63 64 static bool log_ecn_error = true; 65 module_param(log_ecn_error, bool, 0644); 66 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 67 68 static int ipip6_tunnel_init(struct net_device *dev); 69 static void ipip6_tunnel_setup(struct net_device *dev); 70 static void ipip6_dev_free(struct net_device *dev); 71 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, 72 __be32 *v4dst); 73 static struct rtnl_link_ops sit_link_ops __read_mostly; 74 75 static unsigned int sit_net_id __read_mostly; 76 struct sit_net { 77 struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE]; 78 struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE]; 79 struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE]; 80 struct ip_tunnel __rcu *tunnels_wc[1]; 81 struct ip_tunnel __rcu **tunnels[4]; 82 83 struct net_device *fb_tunnel_dev; 84 }; 85 86 static inline struct sit_net *dev_to_sit_net(struct net_device *dev) 87 { 88 struct ip_tunnel *t = netdev_priv(dev); 89 90 return net_generic(t->net, sit_net_id); 91 } 92 93 /* 94 * Must be invoked with rcu_read_lock 95 */ 96 static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net, 97 struct net_device *dev, 98 __be32 remote, __be32 local, 99 int sifindex) 100 { 101 unsigned int h0 = HASH(remote); 102 unsigned int h1 = HASH(local); 103 struct ip_tunnel *t; 104 struct sit_net *sitn = net_generic(net, sit_net_id); 105 int ifindex = dev ? dev->ifindex : 0; 106 107 for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) { 108 if (local == t->parms.iph.saddr && 109 remote == t->parms.iph.daddr && 110 (!dev || !t->parms.link || ifindex == t->parms.link || 111 sifindex == t->parms.link) && 112 (t->dev->flags & IFF_UP)) 113 return t; 114 } 115 for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) { 116 if (remote == t->parms.iph.daddr && 117 (!dev || !t->parms.link || ifindex == t->parms.link || 118 sifindex == t->parms.link) && 119 (t->dev->flags & IFF_UP)) 120 return t; 121 } 122 for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) { 123 if (local == t->parms.iph.saddr && 124 (!dev || !t->parms.link || ifindex == t->parms.link || 125 sifindex == t->parms.link) && 126 (t->dev->flags & IFF_UP)) 127 return t; 128 } 129 t = rcu_dereference(sitn->tunnels_wc[0]); 130 if (t && (t->dev->flags & IFF_UP)) 131 return t; 132 return NULL; 133 } 134 135 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn, 136 struct ip_tunnel_parm *parms) 137 { 138 __be32 remote = parms->iph.daddr; 139 __be32 local = parms->iph.saddr; 140 unsigned int h = 0; 141 int prio = 0; 142 143 if (remote) { 144 prio |= 2; 145 h ^= HASH(remote); 146 } 147 if (local) { 148 prio |= 1; 149 h ^= HASH(local); 150 } 151 return &sitn->tunnels[prio][h]; 152 } 153 154 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn, 155 struct ip_tunnel *t) 156 { 157 return __ipip6_bucket(sitn, &t->parms); 158 } 159 160 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t) 161 { 162 struct ip_tunnel __rcu **tp; 163 struct ip_tunnel *iter; 164 165 for (tp = ipip6_bucket(sitn, t); 166 (iter = rtnl_dereference(*tp)) != NULL; 167 tp = &iter->next) { 168 if (t == iter) { 169 rcu_assign_pointer(*tp, t->next); 170 break; 171 } 172 } 173 } 174 175 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) 176 { 177 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t); 178 179 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 180 rcu_assign_pointer(*tp, t); 181 } 182 183 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn) 184 { 185 #ifdef CONFIG_IPV6_SIT_6RD 186 struct ip_tunnel *t = netdev_priv(dev); 187 188 if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) { 189 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0); 190 t->ip6rd.relay_prefix = 0; 191 t->ip6rd.prefixlen = 16; 192 t->ip6rd.relay_prefixlen = 0; 193 } else { 194 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev); 195 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd)); 196 } 197 #endif 198 } 199 200 static int ipip6_tunnel_create(struct net_device *dev) 201 { 202 struct ip_tunnel *t = netdev_priv(dev); 203 struct net *net = dev_net(dev); 204 struct sit_net *sitn = net_generic(net, sit_net_id); 205 int err; 206 207 __dev_addr_set(dev, &t->parms.iph.saddr, 4); 208 memcpy(dev->broadcast, &t->parms.iph.daddr, 4); 209 210 if ((__force u16)t->parms.i_flags & SIT_ISATAP) 211 dev->priv_flags |= IFF_ISATAP; 212 213 dev->rtnl_link_ops = &sit_link_ops; 214 215 err = register_netdevice(dev); 216 if (err < 0) 217 goto out; 218 219 ipip6_tunnel_clone_6rd(dev, sitn); 220 221 ipip6_tunnel_link(sitn, t); 222 return 0; 223 224 out: 225 return err; 226 } 227 228 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, 229 struct ip_tunnel_parm *parms, int create) 230 { 231 __be32 remote = parms->iph.daddr; 232 __be32 local = parms->iph.saddr; 233 struct ip_tunnel *t, *nt; 234 struct ip_tunnel __rcu **tp; 235 struct net_device *dev; 236 char name[IFNAMSIZ]; 237 struct sit_net *sitn = net_generic(net, sit_net_id); 238 239 for (tp = __ipip6_bucket(sitn, parms); 240 (t = rtnl_dereference(*tp)) != NULL; 241 tp = &t->next) { 242 if (local == t->parms.iph.saddr && 243 remote == t->parms.iph.daddr && 244 parms->link == t->parms.link) { 245 if (create) 246 return NULL; 247 else 248 return t; 249 } 250 } 251 if (!create) 252 goto failed; 253 254 if (parms->name[0]) { 255 if (!dev_valid_name(parms->name)) 256 goto failed; 257 strlcpy(name, parms->name, IFNAMSIZ); 258 } else { 259 strcpy(name, "sit%d"); 260 } 261 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 262 ipip6_tunnel_setup); 263 if (!dev) 264 return NULL; 265 266 dev_net_set(dev, net); 267 268 nt = netdev_priv(dev); 269 270 nt->parms = *parms; 271 if (ipip6_tunnel_create(dev) < 0) 272 goto failed_free; 273 274 if (!parms->name[0]) 275 strcpy(parms->name, dev->name); 276 277 return nt; 278 279 failed_free: 280 free_netdev(dev); 281 failed: 282 return NULL; 283 } 284 285 #define for_each_prl_rcu(start) \ 286 for (prl = rcu_dereference(start); \ 287 prl; \ 288 prl = rcu_dereference(prl->next)) 289 290 static struct ip_tunnel_prl_entry * 291 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr) 292 { 293 struct ip_tunnel_prl_entry *prl; 294 295 for_each_prl_rcu(t->prl) 296 if (prl->addr == addr) 297 break; 298 return prl; 299 300 } 301 302 static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a) 303 { 304 struct ip_tunnel *t = netdev_priv(dev); 305 struct ip_tunnel_prl kprl, *kp; 306 struct ip_tunnel_prl_entry *prl; 307 unsigned int cmax, c = 0, ca, len; 308 int ret = 0; 309 310 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) 311 return -EINVAL; 312 313 if (copy_from_user(&kprl, a, sizeof(kprl))) 314 return -EFAULT; 315 cmax = kprl.datalen / sizeof(kprl); 316 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY)) 317 cmax = 1; 318 319 /* For simple GET or for root users, 320 * we try harder to allocate. 321 */ 322 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ? 323 kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) : 324 NULL; 325 326 rcu_read_lock(); 327 328 ca = min(t->prl_count, cmax); 329 330 if (!kp) { 331 /* We don't try hard to allocate much memory for 332 * non-root users. 333 * For root users, retry allocating enough memory for 334 * the answer. 335 */ 336 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT | 337 __GFP_NOWARN); 338 if (!kp) { 339 ret = -ENOMEM; 340 goto out; 341 } 342 } 343 344 c = 0; 345 for_each_prl_rcu(t->prl) { 346 if (c >= cmax) 347 break; 348 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr) 349 continue; 350 kp[c].addr = prl->addr; 351 kp[c].flags = prl->flags; 352 c++; 353 if (kprl.addr != htonl(INADDR_ANY)) 354 break; 355 } 356 out: 357 rcu_read_unlock(); 358 359 len = sizeof(*kp) * c; 360 ret = 0; 361 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen)) 362 ret = -EFAULT; 363 364 kfree(kp); 365 366 return ret; 367 } 368 369 static int 370 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) 371 { 372 struct ip_tunnel_prl_entry *p; 373 int err = 0; 374 375 if (a->addr == htonl(INADDR_ANY)) 376 return -EINVAL; 377 378 ASSERT_RTNL(); 379 380 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) { 381 if (p->addr == a->addr) { 382 if (chg) { 383 p->flags = a->flags; 384 goto out; 385 } 386 err = -EEXIST; 387 goto out; 388 } 389 } 390 391 if (chg) { 392 err = -ENXIO; 393 goto out; 394 } 395 396 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL); 397 if (!p) { 398 err = -ENOBUFS; 399 goto out; 400 } 401 402 p->next = t->prl; 403 p->addr = a->addr; 404 p->flags = a->flags; 405 t->prl_count++; 406 rcu_assign_pointer(t->prl, p); 407 out: 408 return err; 409 } 410 411 static void prl_list_destroy_rcu(struct rcu_head *head) 412 { 413 struct ip_tunnel_prl_entry *p, *n; 414 415 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head); 416 do { 417 n = rcu_dereference_protected(p->next, 1); 418 kfree(p); 419 p = n; 420 } while (p); 421 } 422 423 static int 424 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) 425 { 426 struct ip_tunnel_prl_entry *x; 427 struct ip_tunnel_prl_entry __rcu **p; 428 int err = 0; 429 430 ASSERT_RTNL(); 431 432 if (a && a->addr != htonl(INADDR_ANY)) { 433 for (p = &t->prl; 434 (x = rtnl_dereference(*p)) != NULL; 435 p = &x->next) { 436 if (x->addr == a->addr) { 437 *p = x->next; 438 kfree_rcu(x, rcu_head); 439 t->prl_count--; 440 goto out; 441 } 442 } 443 err = -ENXIO; 444 } else { 445 x = rtnl_dereference(t->prl); 446 if (x) { 447 t->prl_count = 0; 448 call_rcu(&x->rcu_head, prl_list_destroy_rcu); 449 t->prl = NULL; 450 } 451 } 452 out: 453 return err; 454 } 455 456 static int ipip6_tunnel_prl_ctl(struct net_device *dev, 457 struct ip_tunnel_prl __user *data, int cmd) 458 { 459 struct ip_tunnel *t = netdev_priv(dev); 460 struct ip_tunnel_prl prl; 461 int err; 462 463 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN)) 464 return -EPERM; 465 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) 466 return -EINVAL; 467 468 if (copy_from_user(&prl, data, sizeof(prl))) 469 return -EFAULT; 470 471 switch (cmd) { 472 case SIOCDELPRL: 473 err = ipip6_tunnel_del_prl(t, &prl); 474 break; 475 case SIOCADDPRL: 476 case SIOCCHGPRL: 477 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); 478 break; 479 } 480 dst_cache_reset(&t->dst_cache); 481 netdev_state_change(dev); 482 return err; 483 } 484 485 static int 486 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t) 487 { 488 struct ip_tunnel_prl_entry *p; 489 int ok = 1; 490 491 rcu_read_lock(); 492 p = __ipip6_tunnel_locate_prl(t, iph->saddr); 493 if (p) { 494 if (p->flags & PRL_DEFAULT) 495 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT; 496 else 497 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT; 498 } else { 499 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr; 500 501 if (ipv6_addr_is_isatap(addr6) && 502 (addr6->s6_addr32[3] == iph->saddr) && 503 ipv6_chk_prefix(addr6, t->dev)) 504 skb->ndisc_nodetype = NDISC_NODETYPE_HOST; 505 else 506 ok = 0; 507 } 508 rcu_read_unlock(); 509 return ok; 510 } 511 512 static void ipip6_tunnel_uninit(struct net_device *dev) 513 { 514 struct ip_tunnel *tunnel = netdev_priv(dev); 515 struct sit_net *sitn = net_generic(tunnel->net, sit_net_id); 516 517 if (dev == sitn->fb_tunnel_dev) { 518 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL); 519 } else { 520 ipip6_tunnel_unlink(sitn, tunnel); 521 ipip6_tunnel_del_prl(tunnel, NULL); 522 } 523 dst_cache_reset(&tunnel->dst_cache); 524 netdev_put(dev, &tunnel->dev_tracker); 525 } 526 527 static int ipip6_err(struct sk_buff *skb, u32 info) 528 { 529 const struct iphdr *iph = (const struct iphdr *)skb->data; 530 const int type = icmp_hdr(skb)->type; 531 const int code = icmp_hdr(skb)->code; 532 unsigned int data_len = 0; 533 struct ip_tunnel *t; 534 int sifindex; 535 int err; 536 537 switch (type) { 538 default: 539 case ICMP_PARAMETERPROB: 540 return 0; 541 542 case ICMP_DEST_UNREACH: 543 switch (code) { 544 case ICMP_SR_FAILED: 545 /* Impossible event. */ 546 return 0; 547 default: 548 /* All others are translated to HOST_UNREACH. 549 rfc2003 contains "deep thoughts" about NET_UNREACH, 550 I believe they are just ether pollution. --ANK 551 */ 552 break; 553 } 554 break; 555 case ICMP_TIME_EXCEEDED: 556 if (code != ICMP_EXC_TTL) 557 return 0; 558 data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */ 559 break; 560 case ICMP_REDIRECT: 561 break; 562 } 563 564 err = -ENOENT; 565 566 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0; 567 t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 568 iph->daddr, iph->saddr, sifindex); 569 if (!t) 570 goto out; 571 572 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { 573 ipv4_update_pmtu(skb, dev_net(skb->dev), info, 574 t->parms.link, iph->protocol); 575 err = 0; 576 goto out; 577 } 578 if (type == ICMP_REDIRECT) { 579 ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 580 iph->protocol); 581 err = 0; 582 goto out; 583 } 584 585 err = 0; 586 if (__in6_dev_get(skb->dev) && 587 !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len)) 588 goto out; 589 590 if (t->parms.iph.daddr == 0) 591 goto out; 592 593 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) 594 goto out; 595 596 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) 597 t->err_count++; 598 else 599 t->err_count = 1; 600 t->err_time = jiffies; 601 out: 602 return err; 603 } 604 605 static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr, 606 const struct in6_addr *v6addr) 607 { 608 __be32 v4embed = 0; 609 if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed) 610 return true; 611 return false; 612 } 613 614 /* Checks if an address matches an address on the tunnel interface. 615 * Used to detect the NAT of proto 41 packets and let them pass spoofing test. 616 * Long story: 617 * This function is called after we considered the packet as spoofed 618 * in is_spoofed_6rd. 619 * We may have a router that is doing NAT for proto 41 packets 620 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb 621 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd 622 * function will return true, dropping the packet. 623 * But, we can still check if is spoofed against the IP 624 * addresses associated with the interface. 625 */ 626 static bool only_dnatted(const struct ip_tunnel *tunnel, 627 const struct in6_addr *v6dst) 628 { 629 int prefix_len; 630 631 #ifdef CONFIG_IPV6_SIT_6RD 632 prefix_len = tunnel->ip6rd.prefixlen + 32 633 - tunnel->ip6rd.relay_prefixlen; 634 #else 635 prefix_len = 48; 636 #endif 637 return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev); 638 } 639 640 /* Returns true if a packet is spoofed */ 641 static bool packet_is_spoofed(struct sk_buff *skb, 642 const struct iphdr *iph, 643 struct ip_tunnel *tunnel) 644 { 645 const struct ipv6hdr *ipv6h; 646 647 if (tunnel->dev->priv_flags & IFF_ISATAP) { 648 if (!isatap_chksrc(skb, iph, tunnel)) 649 return true; 650 651 return false; 652 } 653 654 if (tunnel->dev->flags & IFF_POINTOPOINT) 655 return false; 656 657 ipv6h = ipv6_hdr(skb); 658 659 if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) { 660 net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 661 &iph->saddr, &ipv6h->saddr, 662 &iph->daddr, &ipv6h->daddr); 663 return true; 664 } 665 666 if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr))) 667 return false; 668 669 if (only_dnatted(tunnel, &ipv6h->daddr)) 670 return false; 671 672 net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n", 673 &iph->saddr, &ipv6h->saddr, 674 &iph->daddr, &ipv6h->daddr); 675 return true; 676 } 677 678 static int ipip6_rcv(struct sk_buff *skb) 679 { 680 const struct iphdr *iph = ip_hdr(skb); 681 struct ip_tunnel *tunnel; 682 int sifindex; 683 int err; 684 685 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0; 686 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 687 iph->saddr, iph->daddr, sifindex); 688 if (tunnel) { 689 if (tunnel->parms.iph.protocol != IPPROTO_IPV6 && 690 tunnel->parms.iph.protocol != 0) 691 goto out; 692 693 skb->mac_header = skb->network_header; 694 skb_reset_network_header(skb); 695 IPCB(skb)->flags = 0; 696 skb->dev = tunnel->dev; 697 698 if (packet_is_spoofed(skb, iph, tunnel)) { 699 tunnel->dev->stats.rx_errors++; 700 goto out; 701 } 702 703 if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6), 704 !net_eq(tunnel->net, dev_net(tunnel->dev)))) 705 goto out; 706 707 /* skb can be uncloned in iptunnel_pull_header, so 708 * old iph is no longer valid 709 */ 710 iph = (const struct iphdr *)skb_mac_header(skb); 711 skb_reset_mac_header(skb); 712 713 err = IP_ECN_decapsulate(iph, skb); 714 if (unlikely(err)) { 715 if (log_ecn_error) 716 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", 717 &iph->saddr, iph->tos); 718 if (err > 1) { 719 ++tunnel->dev->stats.rx_frame_errors; 720 ++tunnel->dev->stats.rx_errors; 721 goto out; 722 } 723 } 724 725 dev_sw_netstats_rx_add(tunnel->dev, skb->len); 726 727 netif_rx(skb); 728 729 return 0; 730 } 731 732 /* no tunnel matched, let upstream know, ipsec may handle it */ 733 return 1; 734 out: 735 kfree_skb(skb); 736 return 0; 737 } 738 739 static const struct tnl_ptk_info ipip_tpi = { 740 /* no tunnel info required for ipip. */ 741 .proto = htons(ETH_P_IP), 742 }; 743 744 #if IS_ENABLED(CONFIG_MPLS) 745 static const struct tnl_ptk_info mplsip_tpi = { 746 /* no tunnel info required for mplsip. */ 747 .proto = htons(ETH_P_MPLS_UC), 748 }; 749 #endif 750 751 static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto) 752 { 753 const struct iphdr *iph; 754 struct ip_tunnel *tunnel; 755 int sifindex; 756 757 sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0; 758 759 iph = ip_hdr(skb); 760 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, 761 iph->saddr, iph->daddr, sifindex); 762 if (tunnel) { 763 const struct tnl_ptk_info *tpi; 764 765 if (tunnel->parms.iph.protocol != ipproto && 766 tunnel->parms.iph.protocol != 0) 767 goto drop; 768 769 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) 770 goto drop; 771 #if IS_ENABLED(CONFIG_MPLS) 772 if (ipproto == IPPROTO_MPLS) 773 tpi = &mplsip_tpi; 774 else 775 #endif 776 tpi = &ipip_tpi; 777 if (iptunnel_pull_header(skb, 0, tpi->proto, false)) 778 goto drop; 779 skb_reset_mac_header(skb); 780 781 return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 782 } 783 784 return 1; 785 786 drop: 787 kfree_skb(skb); 788 return 0; 789 } 790 791 static int ipip_rcv(struct sk_buff *skb) 792 { 793 return sit_tunnel_rcv(skb, IPPROTO_IPIP); 794 } 795 796 #if IS_ENABLED(CONFIG_MPLS) 797 static int mplsip_rcv(struct sk_buff *skb) 798 { 799 return sit_tunnel_rcv(skb, IPPROTO_MPLS); 800 } 801 #endif 802 803 /* 804 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function 805 * stores the embedded IPv4 address in v4dst and returns true. 806 */ 807 static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst, 808 __be32 *v4dst) 809 { 810 #ifdef CONFIG_IPV6_SIT_6RD 811 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, 812 tunnel->ip6rd.prefixlen)) { 813 unsigned int pbw0, pbi0; 814 int pbi1; 815 u32 d; 816 817 pbw0 = tunnel->ip6rd.prefixlen >> 5; 818 pbi0 = tunnel->ip6rd.prefixlen & 0x1f; 819 820 d = tunnel->ip6rd.relay_prefixlen < 32 ? 821 (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> 822 tunnel->ip6rd.relay_prefixlen : 0; 823 824 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; 825 if (pbi1 > 0) 826 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> 827 (32 - pbi1); 828 829 *v4dst = tunnel->ip6rd.relay_prefix | htonl(d); 830 return true; 831 } 832 #else 833 if (v6dst->s6_addr16[0] == htons(0x2002)) { 834 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ 835 memcpy(v4dst, &v6dst->s6_addr16[1], 4); 836 return true; 837 } 838 #endif 839 return false; 840 } 841 842 static inline __be32 try_6rd(struct ip_tunnel *tunnel, 843 const struct in6_addr *v6dst) 844 { 845 __be32 dst = 0; 846 check_6rd(tunnel, v6dst, &dst); 847 return dst; 848 } 849 850 /* 851 * This function assumes it is being called from dev_queue_xmit() 852 * and that skb is filled properly by that function. 853 */ 854 855 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, 856 struct net_device *dev) 857 { 858 struct ip_tunnel *tunnel = netdev_priv(dev); 859 const struct iphdr *tiph = &tunnel->parms.iph; 860 const struct ipv6hdr *iph6 = ipv6_hdr(skb); 861 u8 tos = tunnel->parms.iph.tos; 862 __be16 df = tiph->frag_off; 863 struct rtable *rt; /* Route to the other host */ 864 struct net_device *tdev; /* Device to other host */ 865 unsigned int max_headroom; /* The extra header space needed */ 866 __be32 dst = tiph->daddr; 867 struct flowi4 fl4; 868 int mtu; 869 const struct in6_addr *addr6; 870 int addr_type; 871 u8 ttl; 872 u8 protocol = IPPROTO_IPV6; 873 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 874 875 if (tos == 1) 876 tos = ipv6_get_dsfield(iph6); 877 878 /* ISATAP (RFC4214) - must come before 6to4 */ 879 if (dev->priv_flags & IFF_ISATAP) { 880 struct neighbour *neigh = NULL; 881 bool do_tx_error = false; 882 883 if (skb_dst(skb)) 884 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 885 886 if (!neigh) { 887 net_dbg_ratelimited("nexthop == NULL\n"); 888 goto tx_error; 889 } 890 891 addr6 = (const struct in6_addr *)&neigh->primary_key; 892 addr_type = ipv6_addr_type(addr6); 893 894 if ((addr_type & IPV6_ADDR_UNICAST) && 895 ipv6_addr_is_isatap(addr6)) 896 dst = addr6->s6_addr32[3]; 897 else 898 do_tx_error = true; 899 900 neigh_release(neigh); 901 if (do_tx_error) 902 goto tx_error; 903 } 904 905 if (!dst) 906 dst = try_6rd(tunnel, &iph6->daddr); 907 908 if (!dst) { 909 struct neighbour *neigh = NULL; 910 bool do_tx_error = false; 911 912 if (skb_dst(skb)) 913 neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); 914 915 if (!neigh) { 916 net_dbg_ratelimited("nexthop == NULL\n"); 917 goto tx_error; 918 } 919 920 addr6 = (const struct in6_addr *)&neigh->primary_key; 921 addr_type = ipv6_addr_type(addr6); 922 923 if (addr_type == IPV6_ADDR_ANY) { 924 addr6 = &ipv6_hdr(skb)->daddr; 925 addr_type = ipv6_addr_type(addr6); 926 } 927 928 if ((addr_type & IPV6_ADDR_COMPATv4) != 0) 929 dst = addr6->s6_addr32[3]; 930 else 931 do_tx_error = true; 932 933 neigh_release(neigh); 934 if (do_tx_error) 935 goto tx_error; 936 } 937 938 flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark, 939 RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6, 940 0, dst, tiph->saddr, 0, 0, 941 sock_net_uid(tunnel->net, NULL)); 942 943 rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr); 944 if (!rt) { 945 rt = ip_route_output_flow(tunnel->net, &fl4, NULL); 946 if (IS_ERR(rt)) { 947 dev->stats.tx_carrier_errors++; 948 goto tx_error_icmp; 949 } 950 dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr); 951 } 952 953 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) { 954 ip_rt_put(rt); 955 dev->stats.tx_carrier_errors++; 956 goto tx_error_icmp; 957 } 958 tdev = rt->dst.dev; 959 960 if (tdev == dev) { 961 ip_rt_put(rt); 962 dev->stats.collisions++; 963 goto tx_error; 964 } 965 966 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) { 967 ip_rt_put(rt); 968 goto tx_error; 969 } 970 971 if (df) { 972 mtu = dst_mtu(&rt->dst) - t_hlen; 973 974 if (mtu < IPV4_MIN_MTU) { 975 dev->stats.collisions++; 976 ip_rt_put(rt); 977 goto tx_error; 978 } 979 980 if (mtu < IPV6_MIN_MTU) { 981 mtu = IPV6_MIN_MTU; 982 df = 0; 983 } 984 985 if (tunnel->parms.iph.daddr) 986 skb_dst_update_pmtu_no_confirm(skb, mtu); 987 988 if (skb->len > mtu && !skb_is_gso(skb)) { 989 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 990 ip_rt_put(rt); 991 goto tx_error; 992 } 993 } 994 995 if (tunnel->err_count > 0) { 996 if (time_before(jiffies, 997 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { 998 tunnel->err_count--; 999 dst_link_failure(skb); 1000 } else 1001 tunnel->err_count = 0; 1002 } 1003 1004 /* 1005 * Okay, now see if we can stuff it in the buffer as-is. 1006 */ 1007 max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen; 1008 1009 if (skb_headroom(skb) < max_headroom || skb_shared(skb) || 1010 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { 1011 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); 1012 if (!new_skb) { 1013 ip_rt_put(rt); 1014 dev->stats.tx_dropped++; 1015 kfree_skb(skb); 1016 return NETDEV_TX_OK; 1017 } 1018 if (skb->sk) 1019 skb_set_owner_w(new_skb, skb->sk); 1020 dev_kfree_skb(skb); 1021 skb = new_skb; 1022 iph6 = ipv6_hdr(skb); 1023 } 1024 ttl = tiph->ttl; 1025 if (ttl == 0) 1026 ttl = iph6->hop_limit; 1027 tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); 1028 1029 if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) { 1030 ip_rt_put(rt); 1031 goto tx_error; 1032 } 1033 1034 skb_set_inner_ipproto(skb, IPPROTO_IPV6); 1035 1036 iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl, 1037 df, !net_eq(tunnel->net, dev_net(dev))); 1038 return NETDEV_TX_OK; 1039 1040 tx_error_icmp: 1041 dst_link_failure(skb); 1042 tx_error: 1043 kfree_skb(skb); 1044 dev->stats.tx_errors++; 1045 return NETDEV_TX_OK; 1046 } 1047 1048 static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb, 1049 struct net_device *dev, u8 ipproto) 1050 { 1051 struct ip_tunnel *tunnel = netdev_priv(dev); 1052 const struct iphdr *tiph = &tunnel->parms.iph; 1053 1054 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) 1055 goto tx_error; 1056 1057 skb_set_inner_ipproto(skb, ipproto); 1058 1059 ip_tunnel_xmit(skb, dev, tiph, ipproto); 1060 return NETDEV_TX_OK; 1061 tx_error: 1062 kfree_skb(skb); 1063 dev->stats.tx_errors++; 1064 return NETDEV_TX_OK; 1065 } 1066 1067 static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb, 1068 struct net_device *dev) 1069 { 1070 if (!pskb_inet_may_pull(skb)) 1071 goto tx_err; 1072 1073 switch (skb->protocol) { 1074 case htons(ETH_P_IP): 1075 sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP); 1076 break; 1077 case htons(ETH_P_IPV6): 1078 ipip6_tunnel_xmit(skb, dev); 1079 break; 1080 #if IS_ENABLED(CONFIG_MPLS) 1081 case htons(ETH_P_MPLS_UC): 1082 sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS); 1083 break; 1084 #endif 1085 default: 1086 goto tx_err; 1087 } 1088 1089 return NETDEV_TX_OK; 1090 1091 tx_err: 1092 dev->stats.tx_errors++; 1093 kfree_skb(skb); 1094 return NETDEV_TX_OK; 1095 1096 } 1097 1098 static void ipip6_tunnel_bind_dev(struct net_device *dev) 1099 { 1100 struct net_device *tdev = NULL; 1101 struct ip_tunnel *tunnel; 1102 const struct iphdr *iph; 1103 struct flowi4 fl4; 1104 1105 tunnel = netdev_priv(dev); 1106 iph = &tunnel->parms.iph; 1107 1108 if (iph->daddr) { 1109 struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4, 1110 NULL, 1111 iph->daddr, iph->saddr, 1112 0, 0, 1113 IPPROTO_IPV6, 1114 RT_TOS(iph->tos), 1115 tunnel->parms.link); 1116 1117 if (!IS_ERR(rt)) { 1118 tdev = rt->dst.dev; 1119 ip_rt_put(rt); 1120 } 1121 dev->flags |= IFF_POINTOPOINT; 1122 } 1123 1124 if (!tdev && tunnel->parms.link) 1125 tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link); 1126 1127 if (tdev && !netif_is_l3_master(tdev)) { 1128 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1129 1130 dev->mtu = tdev->mtu - t_hlen; 1131 if (dev->mtu < IPV6_MIN_MTU) 1132 dev->mtu = IPV6_MIN_MTU; 1133 } 1134 } 1135 1136 static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p, 1137 __u32 fwmark) 1138 { 1139 struct net *net = t->net; 1140 struct sit_net *sitn = net_generic(net, sit_net_id); 1141 1142 ipip6_tunnel_unlink(sitn, t); 1143 synchronize_net(); 1144 t->parms.iph.saddr = p->iph.saddr; 1145 t->parms.iph.daddr = p->iph.daddr; 1146 __dev_addr_set(t->dev, &p->iph.saddr, 4); 1147 memcpy(t->dev->broadcast, &p->iph.daddr, 4); 1148 ipip6_tunnel_link(sitn, t); 1149 t->parms.iph.ttl = p->iph.ttl; 1150 t->parms.iph.tos = p->iph.tos; 1151 t->parms.iph.frag_off = p->iph.frag_off; 1152 if (t->parms.link != p->link || t->fwmark != fwmark) { 1153 t->parms.link = p->link; 1154 t->fwmark = fwmark; 1155 ipip6_tunnel_bind_dev(t->dev); 1156 } 1157 dst_cache_reset(&t->dst_cache); 1158 netdev_state_change(t->dev); 1159 } 1160 1161 #ifdef CONFIG_IPV6_SIT_6RD 1162 static int ipip6_tunnel_update_6rd(struct ip_tunnel *t, 1163 struct ip_tunnel_6rd *ip6rd) 1164 { 1165 struct in6_addr prefix; 1166 __be32 relay_prefix; 1167 1168 if (ip6rd->relay_prefixlen > 32 || 1169 ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64) 1170 return -EINVAL; 1171 1172 ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen); 1173 if (!ipv6_addr_equal(&prefix, &ip6rd->prefix)) 1174 return -EINVAL; 1175 if (ip6rd->relay_prefixlen) 1176 relay_prefix = ip6rd->relay_prefix & 1177 htonl(0xffffffffUL << 1178 (32 - ip6rd->relay_prefixlen)); 1179 else 1180 relay_prefix = 0; 1181 if (relay_prefix != ip6rd->relay_prefix) 1182 return -EINVAL; 1183 1184 t->ip6rd.prefix = prefix; 1185 t->ip6rd.relay_prefix = relay_prefix; 1186 t->ip6rd.prefixlen = ip6rd->prefixlen; 1187 t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen; 1188 dst_cache_reset(&t->dst_cache); 1189 netdev_state_change(t->dev); 1190 return 0; 1191 } 1192 1193 static int 1194 ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data) 1195 { 1196 struct ip_tunnel *t = netdev_priv(dev); 1197 struct ip_tunnel_6rd ip6rd; 1198 struct ip_tunnel_parm p; 1199 1200 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) { 1201 if (copy_from_user(&p, data, sizeof(p))) 1202 return -EFAULT; 1203 t = ipip6_tunnel_locate(t->net, &p, 0); 1204 } 1205 if (!t) 1206 t = netdev_priv(dev); 1207 1208 ip6rd.prefix = t->ip6rd.prefix; 1209 ip6rd.relay_prefix = t->ip6rd.relay_prefix; 1210 ip6rd.prefixlen = t->ip6rd.prefixlen; 1211 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; 1212 if (copy_to_user(data, &ip6rd, sizeof(ip6rd))) 1213 return -EFAULT; 1214 return 0; 1215 } 1216 1217 static int 1218 ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data, 1219 int cmd) 1220 { 1221 struct ip_tunnel *t = netdev_priv(dev); 1222 struct ip_tunnel_6rd ip6rd; 1223 int err; 1224 1225 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN)) 1226 return -EPERM; 1227 if (copy_from_user(&ip6rd, data, sizeof(ip6rd))) 1228 return -EFAULT; 1229 1230 if (cmd != SIOCDEL6RD) { 1231 err = ipip6_tunnel_update_6rd(t, &ip6rd); 1232 if (err < 0) 1233 return err; 1234 } else 1235 ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev)); 1236 return 0; 1237 } 1238 1239 #endif /* CONFIG_IPV6_SIT_6RD */ 1240 1241 static bool ipip6_valid_ip_proto(u8 ipproto) 1242 { 1243 return ipproto == IPPROTO_IPV6 || 1244 ipproto == IPPROTO_IPIP || 1245 #if IS_ENABLED(CONFIG_MPLS) 1246 ipproto == IPPROTO_MPLS || 1247 #endif 1248 ipproto == 0; 1249 } 1250 1251 static int 1252 __ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p) 1253 { 1254 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1255 return -EPERM; 1256 1257 if (!ipip6_valid_ip_proto(p->iph.protocol)) 1258 return -EINVAL; 1259 if (p->iph.version != 4 || 1260 p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF))) 1261 return -EINVAL; 1262 1263 if (p->iph.ttl) 1264 p->iph.frag_off |= htons(IP_DF); 1265 return 0; 1266 } 1267 1268 static int 1269 ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p) 1270 { 1271 struct ip_tunnel *t = netdev_priv(dev); 1272 1273 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) 1274 t = ipip6_tunnel_locate(t->net, p, 0); 1275 if (!t) 1276 t = netdev_priv(dev); 1277 memcpy(p, &t->parms, sizeof(*p)); 1278 return 0; 1279 } 1280 1281 static int 1282 ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p) 1283 { 1284 struct ip_tunnel *t = netdev_priv(dev); 1285 int err; 1286 1287 err = __ipip6_tunnel_ioctl_validate(t->net, p); 1288 if (err) 1289 return err; 1290 1291 t = ipip6_tunnel_locate(t->net, p, 1); 1292 if (!t) 1293 return -ENOBUFS; 1294 return 0; 1295 } 1296 1297 static int 1298 ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p) 1299 { 1300 struct ip_tunnel *t = netdev_priv(dev); 1301 int err; 1302 1303 err = __ipip6_tunnel_ioctl_validate(t->net, p); 1304 if (err) 1305 return err; 1306 1307 t = ipip6_tunnel_locate(t->net, p, 0); 1308 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) { 1309 if (!t) 1310 return -ENOENT; 1311 } else { 1312 if (t) { 1313 if (t->dev != dev) 1314 return -EEXIST; 1315 } else { 1316 if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) || 1317 (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr)) 1318 return -EINVAL; 1319 t = netdev_priv(dev); 1320 } 1321 1322 ipip6_tunnel_update(t, p, t->fwmark); 1323 } 1324 1325 return 0; 1326 } 1327 1328 static int 1329 ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p) 1330 { 1331 struct ip_tunnel *t = netdev_priv(dev); 1332 1333 if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN)) 1334 return -EPERM; 1335 1336 if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) { 1337 t = ipip6_tunnel_locate(t->net, p, 0); 1338 if (!t) 1339 return -ENOENT; 1340 if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev)) 1341 return -EPERM; 1342 dev = t->dev; 1343 } 1344 unregister_netdevice(dev); 1345 return 0; 1346 } 1347 1348 static int 1349 ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd) 1350 { 1351 switch (cmd) { 1352 case SIOCGETTUNNEL: 1353 return ipip6_tunnel_get(dev, p); 1354 case SIOCADDTUNNEL: 1355 return ipip6_tunnel_add(dev, p); 1356 case SIOCCHGTUNNEL: 1357 return ipip6_tunnel_change(dev, p); 1358 case SIOCDELTUNNEL: 1359 return ipip6_tunnel_del(dev, p); 1360 default: 1361 return -EINVAL; 1362 } 1363 } 1364 1365 static int 1366 ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr, 1367 void __user *data, int cmd) 1368 { 1369 switch (cmd) { 1370 case SIOCGETTUNNEL: 1371 case SIOCADDTUNNEL: 1372 case SIOCCHGTUNNEL: 1373 case SIOCDELTUNNEL: 1374 return ip_tunnel_siocdevprivate(dev, ifr, data, cmd); 1375 case SIOCGETPRL: 1376 return ipip6_tunnel_get_prl(dev, data); 1377 case SIOCADDPRL: 1378 case SIOCDELPRL: 1379 case SIOCCHGPRL: 1380 return ipip6_tunnel_prl_ctl(dev, data, cmd); 1381 #ifdef CONFIG_IPV6_SIT_6RD 1382 case SIOCGET6RD: 1383 return ipip6_tunnel_get6rd(dev, data); 1384 case SIOCADD6RD: 1385 case SIOCCHG6RD: 1386 case SIOCDEL6RD: 1387 return ipip6_tunnel_6rdctl(dev, data, cmd); 1388 #endif 1389 default: 1390 return -EINVAL; 1391 } 1392 } 1393 1394 static const struct net_device_ops ipip6_netdev_ops = { 1395 .ndo_init = ipip6_tunnel_init, 1396 .ndo_uninit = ipip6_tunnel_uninit, 1397 .ndo_start_xmit = sit_tunnel_xmit, 1398 .ndo_siocdevprivate = ipip6_tunnel_siocdevprivate, 1399 .ndo_get_stats64 = dev_get_tstats64, 1400 .ndo_get_iflink = ip_tunnel_get_iflink, 1401 .ndo_tunnel_ctl = ipip6_tunnel_ctl, 1402 }; 1403 1404 static void ipip6_dev_free(struct net_device *dev) 1405 { 1406 struct ip_tunnel *tunnel = netdev_priv(dev); 1407 1408 dst_cache_destroy(&tunnel->dst_cache); 1409 free_percpu(dev->tstats); 1410 } 1411 1412 #define SIT_FEATURES (NETIF_F_SG | \ 1413 NETIF_F_FRAGLIST | \ 1414 NETIF_F_HIGHDMA | \ 1415 NETIF_F_GSO_SOFTWARE | \ 1416 NETIF_F_HW_CSUM) 1417 1418 static void ipip6_tunnel_setup(struct net_device *dev) 1419 { 1420 struct ip_tunnel *tunnel = netdev_priv(dev); 1421 int t_hlen = tunnel->hlen + sizeof(struct iphdr); 1422 1423 dev->netdev_ops = &ipip6_netdev_ops; 1424 dev->header_ops = &ip_tunnel_header_ops; 1425 dev->needs_free_netdev = true; 1426 dev->priv_destructor = ipip6_dev_free; 1427 1428 dev->type = ARPHRD_SIT; 1429 dev->mtu = ETH_DATA_LEN - t_hlen; 1430 dev->min_mtu = IPV6_MIN_MTU; 1431 dev->max_mtu = IP6_MAX_MTU - t_hlen; 1432 dev->flags = IFF_NOARP; 1433 netif_keep_dst(dev); 1434 dev->addr_len = 4; 1435 dev->features |= NETIF_F_LLTX; 1436 dev->features |= SIT_FEATURES; 1437 dev->hw_features |= SIT_FEATURES; 1438 } 1439 1440 static int ipip6_tunnel_init(struct net_device *dev) 1441 { 1442 struct ip_tunnel *tunnel = netdev_priv(dev); 1443 int err; 1444 1445 tunnel->dev = dev; 1446 tunnel->net = dev_net(dev); 1447 strcpy(tunnel->parms.name, dev->name); 1448 1449 ipip6_tunnel_bind_dev(dev); 1450 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1451 if (!dev->tstats) 1452 return -ENOMEM; 1453 1454 err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1455 if (err) { 1456 free_percpu(dev->tstats); 1457 dev->tstats = NULL; 1458 return err; 1459 } 1460 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1461 return 0; 1462 } 1463 1464 static void __net_init ipip6_fb_tunnel_init(struct net_device *dev) 1465 { 1466 struct ip_tunnel *tunnel = netdev_priv(dev); 1467 struct iphdr *iph = &tunnel->parms.iph; 1468 struct net *net = dev_net(dev); 1469 struct sit_net *sitn = net_generic(net, sit_net_id); 1470 1471 iph->version = 4; 1472 iph->protocol = IPPROTO_IPV6; 1473 iph->ihl = 5; 1474 iph->ttl = 64; 1475 1476 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); 1477 } 1478 1479 static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[], 1480 struct netlink_ext_ack *extack) 1481 { 1482 u8 proto; 1483 1484 if (!data || !data[IFLA_IPTUN_PROTO]) 1485 return 0; 1486 1487 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1488 if (!ipip6_valid_ip_proto(proto)) 1489 return -EINVAL; 1490 1491 return 0; 1492 } 1493 1494 static void ipip6_netlink_parms(struct nlattr *data[], 1495 struct ip_tunnel_parm *parms, 1496 __u32 *fwmark) 1497 { 1498 memset(parms, 0, sizeof(*parms)); 1499 1500 parms->iph.version = 4; 1501 parms->iph.protocol = IPPROTO_IPV6; 1502 parms->iph.ihl = 5; 1503 parms->iph.ttl = 64; 1504 1505 if (!data) 1506 return; 1507 1508 if (data[IFLA_IPTUN_LINK]) 1509 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]); 1510 1511 if (data[IFLA_IPTUN_LOCAL]) 1512 parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]); 1513 1514 if (data[IFLA_IPTUN_REMOTE]) 1515 parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]); 1516 1517 if (data[IFLA_IPTUN_TTL]) { 1518 parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]); 1519 if (parms->iph.ttl) 1520 parms->iph.frag_off = htons(IP_DF); 1521 } 1522 1523 if (data[IFLA_IPTUN_TOS]) 1524 parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]); 1525 1526 if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC])) 1527 parms->iph.frag_off = htons(IP_DF); 1528 1529 if (data[IFLA_IPTUN_FLAGS]) 1530 parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]); 1531 1532 if (data[IFLA_IPTUN_PROTO]) 1533 parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]); 1534 1535 if (data[IFLA_IPTUN_FWMARK]) 1536 *fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]); 1537 } 1538 1539 /* This function returns true when ENCAP attributes are present in the nl msg */ 1540 static bool ipip6_netlink_encap_parms(struct nlattr *data[], 1541 struct ip_tunnel_encap *ipencap) 1542 { 1543 bool ret = false; 1544 1545 memset(ipencap, 0, sizeof(*ipencap)); 1546 1547 if (!data) 1548 return ret; 1549 1550 if (data[IFLA_IPTUN_ENCAP_TYPE]) { 1551 ret = true; 1552 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]); 1553 } 1554 1555 if (data[IFLA_IPTUN_ENCAP_FLAGS]) { 1556 ret = true; 1557 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]); 1558 } 1559 1560 if (data[IFLA_IPTUN_ENCAP_SPORT]) { 1561 ret = true; 1562 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]); 1563 } 1564 1565 if (data[IFLA_IPTUN_ENCAP_DPORT]) { 1566 ret = true; 1567 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]); 1568 } 1569 1570 return ret; 1571 } 1572 1573 #ifdef CONFIG_IPV6_SIT_6RD 1574 /* This function returns true when 6RD attributes are present in the nl msg */ 1575 static bool ipip6_netlink_6rd_parms(struct nlattr *data[], 1576 struct ip_tunnel_6rd *ip6rd) 1577 { 1578 bool ret = false; 1579 memset(ip6rd, 0, sizeof(*ip6rd)); 1580 1581 if (!data) 1582 return ret; 1583 1584 if (data[IFLA_IPTUN_6RD_PREFIX]) { 1585 ret = true; 1586 ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]); 1587 } 1588 1589 if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) { 1590 ret = true; 1591 ip6rd->relay_prefix = 1592 nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]); 1593 } 1594 1595 if (data[IFLA_IPTUN_6RD_PREFIXLEN]) { 1596 ret = true; 1597 ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]); 1598 } 1599 1600 if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) { 1601 ret = true; 1602 ip6rd->relay_prefixlen = 1603 nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]); 1604 } 1605 1606 return ret; 1607 } 1608 #endif 1609 1610 static int ipip6_newlink(struct net *src_net, struct net_device *dev, 1611 struct nlattr *tb[], struct nlattr *data[], 1612 struct netlink_ext_ack *extack) 1613 { 1614 struct net *net = dev_net(dev); 1615 struct ip_tunnel *nt; 1616 struct ip_tunnel_encap ipencap; 1617 #ifdef CONFIG_IPV6_SIT_6RD 1618 struct ip_tunnel_6rd ip6rd; 1619 #endif 1620 int err; 1621 1622 nt = netdev_priv(dev); 1623 1624 if (ipip6_netlink_encap_parms(data, &ipencap)) { 1625 err = ip_tunnel_encap_setup(nt, &ipencap); 1626 if (err < 0) 1627 return err; 1628 } 1629 1630 ipip6_netlink_parms(data, &nt->parms, &nt->fwmark); 1631 1632 if (ipip6_tunnel_locate(net, &nt->parms, 0)) 1633 return -EEXIST; 1634 1635 err = ipip6_tunnel_create(dev); 1636 if (err < 0) 1637 return err; 1638 1639 if (tb[IFLA_MTU]) { 1640 u32 mtu = nla_get_u32(tb[IFLA_MTU]); 1641 1642 if (mtu >= IPV6_MIN_MTU && 1643 mtu <= IP6_MAX_MTU - dev->hard_header_len) 1644 dev->mtu = mtu; 1645 } 1646 1647 #ifdef CONFIG_IPV6_SIT_6RD 1648 if (ipip6_netlink_6rd_parms(data, &ip6rd)) { 1649 err = ipip6_tunnel_update_6rd(nt, &ip6rd); 1650 if (err < 0) 1651 unregister_netdevice_queue(dev, NULL); 1652 } 1653 #endif 1654 1655 return err; 1656 } 1657 1658 static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[], 1659 struct nlattr *data[], 1660 struct netlink_ext_ack *extack) 1661 { 1662 struct ip_tunnel *t = netdev_priv(dev); 1663 struct ip_tunnel_parm p; 1664 struct ip_tunnel_encap ipencap; 1665 struct net *net = t->net; 1666 struct sit_net *sitn = net_generic(net, sit_net_id); 1667 #ifdef CONFIG_IPV6_SIT_6RD 1668 struct ip_tunnel_6rd ip6rd; 1669 #endif 1670 __u32 fwmark = t->fwmark; 1671 int err; 1672 1673 if (dev == sitn->fb_tunnel_dev) 1674 return -EINVAL; 1675 1676 if (ipip6_netlink_encap_parms(data, &ipencap)) { 1677 err = ip_tunnel_encap_setup(t, &ipencap); 1678 if (err < 0) 1679 return err; 1680 } 1681 1682 ipip6_netlink_parms(data, &p, &fwmark); 1683 1684 if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) || 1685 (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr)) 1686 return -EINVAL; 1687 1688 t = ipip6_tunnel_locate(net, &p, 0); 1689 1690 if (t) { 1691 if (t->dev != dev) 1692 return -EEXIST; 1693 } else 1694 t = netdev_priv(dev); 1695 1696 ipip6_tunnel_update(t, &p, fwmark); 1697 1698 #ifdef CONFIG_IPV6_SIT_6RD 1699 if (ipip6_netlink_6rd_parms(data, &ip6rd)) 1700 return ipip6_tunnel_update_6rd(t, &ip6rd); 1701 #endif 1702 1703 return 0; 1704 } 1705 1706 static size_t ipip6_get_size(const struct net_device *dev) 1707 { 1708 return 1709 /* IFLA_IPTUN_LINK */ 1710 nla_total_size(4) + 1711 /* IFLA_IPTUN_LOCAL */ 1712 nla_total_size(4) + 1713 /* IFLA_IPTUN_REMOTE */ 1714 nla_total_size(4) + 1715 /* IFLA_IPTUN_TTL */ 1716 nla_total_size(1) + 1717 /* IFLA_IPTUN_TOS */ 1718 nla_total_size(1) + 1719 /* IFLA_IPTUN_PMTUDISC */ 1720 nla_total_size(1) + 1721 /* IFLA_IPTUN_FLAGS */ 1722 nla_total_size(2) + 1723 /* IFLA_IPTUN_PROTO */ 1724 nla_total_size(1) + 1725 #ifdef CONFIG_IPV6_SIT_6RD 1726 /* IFLA_IPTUN_6RD_PREFIX */ 1727 nla_total_size(sizeof(struct in6_addr)) + 1728 /* IFLA_IPTUN_6RD_RELAY_PREFIX */ 1729 nla_total_size(4) + 1730 /* IFLA_IPTUN_6RD_PREFIXLEN */ 1731 nla_total_size(2) + 1732 /* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */ 1733 nla_total_size(2) + 1734 #endif 1735 /* IFLA_IPTUN_ENCAP_TYPE */ 1736 nla_total_size(2) + 1737 /* IFLA_IPTUN_ENCAP_FLAGS */ 1738 nla_total_size(2) + 1739 /* IFLA_IPTUN_ENCAP_SPORT */ 1740 nla_total_size(2) + 1741 /* IFLA_IPTUN_ENCAP_DPORT */ 1742 nla_total_size(2) + 1743 /* IFLA_IPTUN_FWMARK */ 1744 nla_total_size(4) + 1745 0; 1746 } 1747 1748 static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1749 { 1750 struct ip_tunnel *tunnel = netdev_priv(dev); 1751 struct ip_tunnel_parm *parm = &tunnel->parms; 1752 1753 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) || 1754 nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) || 1755 nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) || 1756 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) || 1757 nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) || 1758 nla_put_u8(skb, IFLA_IPTUN_PMTUDISC, 1759 !!(parm->iph.frag_off & htons(IP_DF))) || 1760 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) || 1761 nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) || 1762 nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark)) 1763 goto nla_put_failure; 1764 1765 #ifdef CONFIG_IPV6_SIT_6RD 1766 if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX, 1767 &tunnel->ip6rd.prefix) || 1768 nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX, 1769 tunnel->ip6rd.relay_prefix) || 1770 nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN, 1771 tunnel->ip6rd.prefixlen) || 1772 nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN, 1773 tunnel->ip6rd.relay_prefixlen)) 1774 goto nla_put_failure; 1775 #endif 1776 1777 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, 1778 tunnel->encap.type) || 1779 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, 1780 tunnel->encap.sport) || 1781 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, 1782 tunnel->encap.dport) || 1783 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, 1784 tunnel->encap.flags)) 1785 goto nla_put_failure; 1786 1787 return 0; 1788 1789 nla_put_failure: 1790 return -EMSGSIZE; 1791 } 1792 1793 static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = { 1794 [IFLA_IPTUN_LINK] = { .type = NLA_U32 }, 1795 [IFLA_IPTUN_LOCAL] = { .type = NLA_U32 }, 1796 [IFLA_IPTUN_REMOTE] = { .type = NLA_U32 }, 1797 [IFLA_IPTUN_TTL] = { .type = NLA_U8 }, 1798 [IFLA_IPTUN_TOS] = { .type = NLA_U8 }, 1799 [IFLA_IPTUN_PMTUDISC] = { .type = NLA_U8 }, 1800 [IFLA_IPTUN_FLAGS] = { .type = NLA_U16 }, 1801 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 }, 1802 #ifdef CONFIG_IPV6_SIT_6RD 1803 [IFLA_IPTUN_6RD_PREFIX] = { .len = sizeof(struct in6_addr) }, 1804 [IFLA_IPTUN_6RD_RELAY_PREFIX] = { .type = NLA_U32 }, 1805 [IFLA_IPTUN_6RD_PREFIXLEN] = { .type = NLA_U16 }, 1806 [IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 }, 1807 #endif 1808 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 }, 1809 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 }, 1810 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 }, 1811 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 }, 1812 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 }, 1813 }; 1814 1815 static void ipip6_dellink(struct net_device *dev, struct list_head *head) 1816 { 1817 struct net *net = dev_net(dev); 1818 struct sit_net *sitn = net_generic(net, sit_net_id); 1819 1820 if (dev != sitn->fb_tunnel_dev) 1821 unregister_netdevice_queue(dev, head); 1822 } 1823 1824 static struct rtnl_link_ops sit_link_ops __read_mostly = { 1825 .kind = "sit", 1826 .maxtype = IFLA_IPTUN_MAX, 1827 .policy = ipip6_policy, 1828 .priv_size = sizeof(struct ip_tunnel), 1829 .setup = ipip6_tunnel_setup, 1830 .validate = ipip6_validate, 1831 .newlink = ipip6_newlink, 1832 .changelink = ipip6_changelink, 1833 .get_size = ipip6_get_size, 1834 .fill_info = ipip6_fill_info, 1835 .dellink = ipip6_dellink, 1836 .get_link_net = ip_tunnel_get_link_net, 1837 }; 1838 1839 static struct xfrm_tunnel sit_handler __read_mostly = { 1840 .handler = ipip6_rcv, 1841 .err_handler = ipip6_err, 1842 .priority = 1, 1843 }; 1844 1845 static struct xfrm_tunnel ipip_handler __read_mostly = { 1846 .handler = ipip_rcv, 1847 .err_handler = ipip6_err, 1848 .priority = 2, 1849 }; 1850 1851 #if IS_ENABLED(CONFIG_MPLS) 1852 static struct xfrm_tunnel mplsip_handler __read_mostly = { 1853 .handler = mplsip_rcv, 1854 .err_handler = ipip6_err, 1855 .priority = 2, 1856 }; 1857 #endif 1858 1859 static void __net_exit sit_destroy_tunnels(struct net *net, 1860 struct list_head *head) 1861 { 1862 struct sit_net *sitn = net_generic(net, sit_net_id); 1863 struct net_device *dev, *aux; 1864 int prio; 1865 1866 for_each_netdev_safe(net, dev, aux) 1867 if (dev->rtnl_link_ops == &sit_link_ops) 1868 unregister_netdevice_queue(dev, head); 1869 1870 for (prio = 0; prio < 4; prio++) { 1871 int h; 1872 for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) { 1873 struct ip_tunnel *t; 1874 1875 t = rtnl_dereference(sitn->tunnels[prio][h]); 1876 while (t) { 1877 /* If dev is in the same netns, it has already 1878 * been added to the list by the previous loop. 1879 */ 1880 if (!net_eq(dev_net(t->dev), net)) 1881 unregister_netdevice_queue(t->dev, 1882 head); 1883 t = rtnl_dereference(t->next); 1884 } 1885 } 1886 } 1887 } 1888 1889 static int __net_init sit_init_net(struct net *net) 1890 { 1891 struct sit_net *sitn = net_generic(net, sit_net_id); 1892 struct ip_tunnel *t; 1893 int err; 1894 1895 sitn->tunnels[0] = sitn->tunnels_wc; 1896 sitn->tunnels[1] = sitn->tunnels_l; 1897 sitn->tunnels[2] = sitn->tunnels_r; 1898 sitn->tunnels[3] = sitn->tunnels_r_l; 1899 1900 if (!net_has_fallback_tunnels(net)) 1901 return 0; 1902 1903 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", 1904 NET_NAME_UNKNOWN, 1905 ipip6_tunnel_setup); 1906 if (!sitn->fb_tunnel_dev) { 1907 err = -ENOMEM; 1908 goto err_alloc_dev; 1909 } 1910 dev_net_set(sitn->fb_tunnel_dev, net); 1911 sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops; 1912 /* FB netdevice is special: we have one, and only one per netns. 1913 * Allowing to move it to another netns is clearly unsafe. 1914 */ 1915 sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1916 1917 err = register_netdev(sitn->fb_tunnel_dev); 1918 if (err) 1919 goto err_reg_dev; 1920 1921 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); 1922 ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); 1923 1924 t = netdev_priv(sitn->fb_tunnel_dev); 1925 1926 strcpy(t->parms.name, sitn->fb_tunnel_dev->name); 1927 return 0; 1928 1929 err_reg_dev: 1930 free_netdev(sitn->fb_tunnel_dev); 1931 err_alloc_dev: 1932 return err; 1933 } 1934 1935 static void __net_exit sit_exit_batch_net(struct list_head *net_list) 1936 { 1937 LIST_HEAD(list); 1938 struct net *net; 1939 1940 rtnl_lock(); 1941 list_for_each_entry(net, net_list, exit_list) 1942 sit_destroy_tunnels(net, &list); 1943 1944 unregister_netdevice_many(&list); 1945 rtnl_unlock(); 1946 } 1947 1948 static struct pernet_operations sit_net_ops = { 1949 .init = sit_init_net, 1950 .exit_batch = sit_exit_batch_net, 1951 .id = &sit_net_id, 1952 .size = sizeof(struct sit_net), 1953 }; 1954 1955 static void __exit sit_cleanup(void) 1956 { 1957 rtnl_link_unregister(&sit_link_ops); 1958 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 1959 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 1960 #if IS_ENABLED(CONFIG_MPLS) 1961 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS); 1962 #endif 1963 1964 unregister_pernet_device(&sit_net_ops); 1965 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1966 } 1967 1968 static int __init sit_init(void) 1969 { 1970 int err; 1971 1972 pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n"); 1973 1974 err = register_pernet_device(&sit_net_ops); 1975 if (err < 0) 1976 return err; 1977 err = xfrm4_tunnel_register(&sit_handler, AF_INET6); 1978 if (err < 0) { 1979 pr_info("%s: can't register ip6ip4\n", __func__); 1980 goto xfrm_tunnel_failed; 1981 } 1982 err = xfrm4_tunnel_register(&ipip_handler, AF_INET); 1983 if (err < 0) { 1984 pr_info("%s: can't register ip4ip4\n", __func__); 1985 goto xfrm_tunnel4_failed; 1986 } 1987 #if IS_ENABLED(CONFIG_MPLS) 1988 err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS); 1989 if (err < 0) { 1990 pr_info("%s: can't register mplsip\n", __func__); 1991 goto xfrm_tunnel_mpls_failed; 1992 } 1993 #endif 1994 err = rtnl_link_register(&sit_link_ops); 1995 if (err < 0) 1996 goto rtnl_link_failed; 1997 1998 out: 1999 return err; 2000 2001 rtnl_link_failed: 2002 #if IS_ENABLED(CONFIG_MPLS) 2003 xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS); 2004 xfrm_tunnel_mpls_failed: 2005 #endif 2006 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 2007 xfrm_tunnel4_failed: 2008 xfrm4_tunnel_deregister(&sit_handler, AF_INET6); 2009 xfrm_tunnel_failed: 2010 unregister_pernet_device(&sit_net_ops); 2011 goto out; 2012 } 2013 2014 module_init(sit_init); 2015 module_exit(sit_cleanup); 2016 MODULE_LICENSE("GPL"); 2017 MODULE_ALIAS_RTNL_LINK("sit"); 2018 MODULE_ALIAS_NETDEV("sit0"); 2019