1 /* 2 * IPv6 virtual tunneling interface 3 * 4 * Copyright (C) 2013 secunet Security Networks AG 5 * 6 * Author: 7 * Steffen Klassert <steffen.klassert@secunet.com> 8 * 9 * Based on: 10 * net/ipv6/ip6_tunnel.c 11 * 12 * This program is free software; you can redistribute it and/or 13 * modify it under the terms of the GNU General Public License 14 * as published by the Free Software Foundation; either version 15 * 2 of the License, or (at your option) any later version. 16 */ 17 18 #include <linux/module.h> 19 #include <linux/capability.h> 20 #include <linux/errno.h> 21 #include <linux/types.h> 22 #include <linux/sockios.h> 23 #include <linux/icmp.h> 24 #include <linux/if.h> 25 #include <linux/in.h> 26 #include <linux/ip.h> 27 #include <linux/net.h> 28 #include <linux/in6.h> 29 #include <linux/netdevice.h> 30 #include <linux/if_arp.h> 31 #include <linux/icmpv6.h> 32 #include <linux/init.h> 33 #include <linux/route.h> 34 #include <linux/rtnetlink.h> 35 #include <linux/netfilter_ipv6.h> 36 #include <linux/slab.h> 37 #include <linux/hash.h> 38 39 #include <linux/uaccess.h> 40 #include <linux/atomic.h> 41 42 #include <net/icmp.h> 43 #include <net/ip.h> 44 #include <net/ip_tunnels.h> 45 #include <net/ipv6.h> 46 #include <net/ip6_route.h> 47 #include <net/addrconf.h> 48 #include <net/ip6_tunnel.h> 49 #include <net/xfrm.h> 50 #include <net/net_namespace.h> 51 #include <net/netns/generic.h> 52 #include <linux/etherdevice.h> 53 54 #define IP6_VTI_HASH_SIZE_SHIFT 5 55 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT) 56 57 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 58 { 59 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 60 61 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT); 62 } 63 64 static int vti6_dev_init(struct net_device *dev); 65 static void vti6_dev_setup(struct net_device *dev); 66 static struct rtnl_link_ops vti6_link_ops __read_mostly; 67 68 static unsigned int vti6_net_id __read_mostly; 69 struct vti6_net { 70 /* the vti6 tunnel fallback device */ 71 struct net_device *fb_tnl_dev; 72 /* lists for storing tunnels in use */ 73 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE]; 74 struct ip6_tnl __rcu *tnls_wc[1]; 75 struct ip6_tnl __rcu **tnls[2]; 76 }; 77 78 #define for_each_vti6_tunnel_rcu(start) \ 79 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 80 81 /** 82 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses 83 * @net: network namespace 84 * @remote: the address of the tunnel exit-point 85 * @local: the address of the tunnel entry-point 86 * 87 * Return: 88 * tunnel matching given end-points if found, 89 * else fallback tunnel if its device is up, 90 * else %NULL 91 **/ 92 static struct ip6_tnl * 93 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote, 94 const struct in6_addr *local) 95 { 96 unsigned int hash = HASH(remote, local); 97 struct ip6_tnl *t; 98 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 99 struct in6_addr any; 100 101 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 102 if (ipv6_addr_equal(local, &t->parms.laddr) && 103 ipv6_addr_equal(remote, &t->parms.raddr) && 104 (t->dev->flags & IFF_UP)) 105 return t; 106 } 107 108 memset(&any, 0, sizeof(any)); 109 hash = HASH(&any, local); 110 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 111 if (ipv6_addr_equal(local, &t->parms.laddr) && 112 (t->dev->flags & IFF_UP)) 113 return t; 114 } 115 116 hash = HASH(remote, &any); 117 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 118 if (ipv6_addr_equal(remote, &t->parms.raddr) && 119 (t->dev->flags & IFF_UP)) 120 return t; 121 } 122 123 t = rcu_dereference(ip6n->tnls_wc[0]); 124 if (t && (t->dev->flags & IFF_UP)) 125 return t; 126 127 return NULL; 128 } 129 130 /** 131 * vti6_tnl_bucket - get head of list matching given tunnel parameters 132 * @p: parameters containing tunnel end-points 133 * 134 * Description: 135 * vti6_tnl_bucket() returns the head of the list matching the 136 * &struct in6_addr entries laddr and raddr in @p. 137 * 138 * Return: head of IPv6 tunnel list 139 **/ 140 static struct ip6_tnl __rcu ** 141 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 142 { 143 const struct in6_addr *remote = &p->raddr; 144 const struct in6_addr *local = &p->laddr; 145 unsigned int h = 0; 146 int prio = 0; 147 148 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 149 prio = 1; 150 h = HASH(remote, local); 151 } 152 return &ip6n->tnls[prio][h]; 153 } 154 155 static void 156 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 157 { 158 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 159 160 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 161 rcu_assign_pointer(*tp, t); 162 } 163 164 static void 165 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 166 { 167 struct ip6_tnl __rcu **tp; 168 struct ip6_tnl *iter; 169 170 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 171 (iter = rtnl_dereference(*tp)) != NULL; 172 tp = &iter->next) { 173 if (t == iter) { 174 rcu_assign_pointer(*tp, t->next); 175 break; 176 } 177 } 178 } 179 180 static void vti6_dev_free(struct net_device *dev) 181 { 182 free_percpu(dev->tstats); 183 } 184 185 static int vti6_tnl_create2(struct net_device *dev) 186 { 187 struct ip6_tnl *t = netdev_priv(dev); 188 struct net *net = dev_net(dev); 189 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 190 int err; 191 192 dev->rtnl_link_ops = &vti6_link_ops; 193 err = register_netdevice(dev); 194 if (err < 0) 195 goto out; 196 197 strcpy(t->parms.name, dev->name); 198 199 dev_hold(dev); 200 vti6_tnl_link(ip6n, t); 201 202 return 0; 203 204 out: 205 return err; 206 } 207 208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 209 { 210 struct net_device *dev; 211 struct ip6_tnl *t; 212 char name[IFNAMSIZ]; 213 int err; 214 215 if (p->name[0]) 216 strlcpy(name, p->name, IFNAMSIZ); 217 else 218 sprintf(name, "ip6_vti%%d"); 219 220 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 221 if (!dev) 222 goto failed; 223 224 dev_net_set(dev, net); 225 226 t = netdev_priv(dev); 227 t->parms = *p; 228 t->net = dev_net(dev); 229 230 err = vti6_tnl_create2(dev); 231 if (err < 0) 232 goto failed_free; 233 234 return t; 235 236 failed_free: 237 free_netdev(dev); 238 failed: 239 return NULL; 240 } 241 242 /** 243 * vti6_locate - find or create tunnel matching given parameters 244 * @net: network namespace 245 * @p: tunnel parameters 246 * @create: != 0 if allowed to create new tunnel if no match found 247 * 248 * Description: 249 * vti6_locate() first tries to locate an existing tunnel 250 * based on @parms. If this is unsuccessful, but @create is set a new 251 * tunnel device is created and registered for use. 252 * 253 * Return: 254 * matching tunnel or NULL 255 **/ 256 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 257 int create) 258 { 259 const struct in6_addr *remote = &p->raddr; 260 const struct in6_addr *local = &p->laddr; 261 struct ip6_tnl __rcu **tp; 262 struct ip6_tnl *t; 263 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 264 265 for (tp = vti6_tnl_bucket(ip6n, p); 266 (t = rtnl_dereference(*tp)) != NULL; 267 tp = &t->next) { 268 if (ipv6_addr_equal(local, &t->parms.laddr) && 269 ipv6_addr_equal(remote, &t->parms.raddr)) { 270 if (create) 271 return NULL; 272 273 return t; 274 } 275 } 276 if (!create) 277 return NULL; 278 return vti6_tnl_create(net, p); 279 } 280 281 /** 282 * vti6_dev_uninit - tunnel device uninitializer 283 * @dev: the device to be destroyed 284 * 285 * Description: 286 * vti6_dev_uninit() removes tunnel from its list 287 **/ 288 static void vti6_dev_uninit(struct net_device *dev) 289 { 290 struct ip6_tnl *t = netdev_priv(dev); 291 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 292 293 if (dev == ip6n->fb_tnl_dev) 294 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 295 else 296 vti6_tnl_unlink(ip6n, t); 297 dev_put(dev); 298 } 299 300 static int vti6_rcv(struct sk_buff *skb) 301 { 302 struct ip6_tnl *t; 303 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 304 305 rcu_read_lock(); 306 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 307 if (t) { 308 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 309 rcu_read_unlock(); 310 goto discard; 311 } 312 313 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 314 rcu_read_unlock(); 315 return 0; 316 } 317 318 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 319 t->dev->stats.rx_dropped++; 320 rcu_read_unlock(); 321 goto discard; 322 } 323 324 rcu_read_unlock(); 325 326 return xfrm6_rcv_tnl(skb, t); 327 } 328 rcu_read_unlock(); 329 return -EINVAL; 330 discard: 331 kfree_skb(skb); 332 return 0; 333 } 334 335 static int vti6_rcv_cb(struct sk_buff *skb, int err) 336 { 337 unsigned short family; 338 struct net_device *dev; 339 struct pcpu_sw_netstats *tstats; 340 struct xfrm_state *x; 341 struct xfrm_mode *inner_mode; 342 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 343 u32 orig_mark = skb->mark; 344 int ret; 345 346 if (!t) 347 return 1; 348 349 dev = t->dev; 350 351 if (err) { 352 dev->stats.rx_errors++; 353 dev->stats.rx_dropped++; 354 355 return 0; 356 } 357 358 x = xfrm_input_state(skb); 359 360 inner_mode = x->inner_mode; 361 362 if (x->sel.family == AF_UNSPEC) { 363 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 364 if (inner_mode == NULL) { 365 XFRM_INC_STATS(dev_net(skb->dev), 366 LINUX_MIB_XFRMINSTATEMODEERROR); 367 return -EINVAL; 368 } 369 } 370 371 family = inner_mode->afinfo->family; 372 373 skb->mark = be32_to_cpu(t->parms.i_key); 374 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 375 skb->mark = orig_mark; 376 377 if (!ret) 378 return -EPERM; 379 380 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 381 skb->dev = dev; 382 383 tstats = this_cpu_ptr(dev->tstats); 384 u64_stats_update_begin(&tstats->syncp); 385 tstats->rx_packets++; 386 tstats->rx_bytes += skb->len; 387 u64_stats_update_end(&tstats->syncp); 388 389 return 0; 390 } 391 392 /** 393 * vti6_addr_conflict - compare packet addresses to tunnel's own 394 * @t: the outgoing tunnel device 395 * @hdr: IPv6 header from the incoming packet 396 * 397 * Description: 398 * Avoid trivial tunneling loop by checking that tunnel exit-point 399 * doesn't match source of incoming packet. 400 * 401 * Return: 402 * 1 if conflict, 403 * 0 else 404 **/ 405 static inline bool 406 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 407 { 408 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 409 } 410 411 static bool vti6_state_check(const struct xfrm_state *x, 412 const struct in6_addr *dst, 413 const struct in6_addr *src) 414 { 415 xfrm_address_t *daddr = (xfrm_address_t *)dst; 416 xfrm_address_t *saddr = (xfrm_address_t *)src; 417 418 /* if there is no transform then this tunnel is not functional. 419 * Or if the xfrm is not mode tunnel. 420 */ 421 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 422 x->props.family != AF_INET6) 423 return false; 424 425 if (ipv6_addr_any(dst)) 426 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 427 428 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 429 return false; 430 431 return true; 432 } 433 434 /** 435 * vti6_xmit - send a packet 436 * @skb: the outgoing socket buffer 437 * @dev: the outgoing tunnel device 438 * @fl: the flow informations for the xfrm_lookup 439 **/ 440 static int 441 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 442 { 443 struct ip6_tnl *t = netdev_priv(dev); 444 struct net_device_stats *stats = &t->dev->stats; 445 struct dst_entry *dst = skb_dst(skb); 446 struct net_device *tdev; 447 struct xfrm_state *x; 448 int pkt_len = skb->len; 449 int err = -1; 450 int mtu; 451 452 if (!dst) 453 goto tx_err_link_failure; 454 455 dst_hold(dst); 456 dst = xfrm_lookup(t->net, dst, fl, NULL, 0); 457 if (IS_ERR(dst)) { 458 err = PTR_ERR(dst); 459 dst = NULL; 460 goto tx_err_link_failure; 461 } 462 463 x = dst->xfrm; 464 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 465 goto tx_err_link_failure; 466 467 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 468 (const struct in6_addr *)&x->id.daddr)) 469 goto tx_err_link_failure; 470 471 tdev = dst->dev; 472 473 if (tdev == dev) { 474 stats->collisions++; 475 net_warn_ratelimited("%s: Local routing loop detected!\n", 476 t->parms.name); 477 goto tx_err_dst_release; 478 } 479 480 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 481 skb_dst_set(skb, dst); 482 skb->dev = skb_dst(skb)->dev; 483 484 mtu = dst_mtu(dst); 485 if (!skb->ignore_df && skb->len > mtu) { 486 skb_dst_update_pmtu(skb, mtu); 487 488 if (skb->protocol == htons(ETH_P_IPV6)) { 489 if (mtu < IPV6_MIN_MTU) 490 mtu = IPV6_MIN_MTU; 491 492 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 493 } else { 494 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 495 htonl(mtu)); 496 } 497 498 return -EMSGSIZE; 499 } 500 501 err = dst_output(t->net, skb->sk, skb); 502 if (net_xmit_eval(err) == 0) { 503 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats); 504 505 u64_stats_update_begin(&tstats->syncp); 506 tstats->tx_bytes += pkt_len; 507 tstats->tx_packets++; 508 u64_stats_update_end(&tstats->syncp); 509 } else { 510 stats->tx_errors++; 511 stats->tx_aborted_errors++; 512 } 513 514 return 0; 515 tx_err_link_failure: 516 stats->tx_carrier_errors++; 517 dst_link_failure(skb); 518 tx_err_dst_release: 519 dst_release(dst); 520 return err; 521 } 522 523 static netdev_tx_t 524 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 525 { 526 struct ip6_tnl *t = netdev_priv(dev); 527 struct net_device_stats *stats = &t->dev->stats; 528 struct ipv6hdr *ipv6h; 529 struct flowi fl; 530 int ret; 531 532 memset(&fl, 0, sizeof(fl)); 533 534 switch (skb->protocol) { 535 case htons(ETH_P_IPV6): 536 ipv6h = ipv6_hdr(skb); 537 538 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 539 vti6_addr_conflict(t, ipv6h)) 540 goto tx_err; 541 542 xfrm_decode_session(skb, &fl, AF_INET6); 543 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 544 break; 545 case htons(ETH_P_IP): 546 xfrm_decode_session(skb, &fl, AF_INET); 547 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 548 break; 549 default: 550 goto tx_err; 551 } 552 553 /* override mark with tunnel output key */ 554 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 555 556 ret = vti6_xmit(skb, dev, &fl); 557 if (ret < 0) 558 goto tx_err; 559 560 return NETDEV_TX_OK; 561 562 tx_err: 563 stats->tx_errors++; 564 stats->tx_dropped++; 565 kfree_skb(skb); 566 return NETDEV_TX_OK; 567 } 568 569 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 570 u8 type, u8 code, int offset, __be32 info) 571 { 572 __be32 spi; 573 __u32 mark; 574 struct xfrm_state *x; 575 struct ip6_tnl *t; 576 struct ip_esp_hdr *esph; 577 struct ip_auth_hdr *ah; 578 struct ip_comp_hdr *ipch; 579 struct net *net = dev_net(skb->dev); 580 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 581 int protocol = iph->nexthdr; 582 583 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 584 if (!t) 585 return -1; 586 587 mark = be32_to_cpu(t->parms.o_key); 588 589 switch (protocol) { 590 case IPPROTO_ESP: 591 esph = (struct ip_esp_hdr *)(skb->data + offset); 592 spi = esph->spi; 593 break; 594 case IPPROTO_AH: 595 ah = (struct ip_auth_hdr *)(skb->data + offset); 596 spi = ah->spi; 597 break; 598 case IPPROTO_COMP: 599 ipch = (struct ip_comp_hdr *)(skb->data + offset); 600 spi = htonl(ntohs(ipch->cpi)); 601 break; 602 default: 603 return 0; 604 } 605 606 if (type != ICMPV6_PKT_TOOBIG && 607 type != NDISC_REDIRECT) 608 return 0; 609 610 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 611 spi, protocol, AF_INET6); 612 if (!x) 613 return 0; 614 615 if (type == NDISC_REDIRECT) 616 ip6_redirect(skb, net, skb->dev->ifindex, 0, 617 sock_net_uid(net, NULL)); 618 else 619 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 620 xfrm_state_put(x); 621 622 return 0; 623 } 624 625 static void vti6_link_config(struct ip6_tnl *t) 626 { 627 struct net_device *dev = t->dev; 628 struct __ip6_tnl_parm *p = &t->parms; 629 struct net_device *tdev = NULL; 630 631 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 632 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 633 634 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 635 IP6_TNL_F_CAP_PER_PACKET); 636 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 637 638 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 639 dev->flags |= IFF_POINTOPOINT; 640 else 641 dev->flags &= ~IFF_POINTOPOINT; 642 643 if (p->flags & IP6_TNL_F_CAP_XMIT) { 644 int strict = (ipv6_addr_type(&p->raddr) & 645 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 646 struct rt6_info *rt = rt6_lookup(t->net, 647 &p->raddr, &p->laddr, 648 p->link, strict); 649 650 if (rt) 651 tdev = rt->dst.dev; 652 ip6_rt_put(rt); 653 } 654 655 if (!tdev && p->link) 656 tdev = __dev_get_by_index(t->net, p->link); 657 658 if (tdev) 659 dev->mtu = max_t(int, tdev->mtu - dev->hard_header_len, 660 IPV6_MIN_MTU); 661 } 662 663 /** 664 * vti6_tnl_change - update the tunnel parameters 665 * @t: tunnel to be changed 666 * @p: tunnel configuration parameters 667 * 668 * Description: 669 * vti6_tnl_change() updates the tunnel parameters 670 **/ 671 static int 672 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p) 673 { 674 t->parms.laddr = p->laddr; 675 t->parms.raddr = p->raddr; 676 t->parms.link = p->link; 677 t->parms.i_key = p->i_key; 678 t->parms.o_key = p->o_key; 679 t->parms.proto = p->proto; 680 t->parms.fwmark = p->fwmark; 681 dst_cache_reset(&t->dst_cache); 682 vti6_link_config(t); 683 return 0; 684 } 685 686 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p) 687 { 688 struct net *net = dev_net(t->dev); 689 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 690 int err; 691 692 vti6_tnl_unlink(ip6n, t); 693 synchronize_net(); 694 err = vti6_tnl_change(t, p); 695 vti6_tnl_link(ip6n, t); 696 netdev_state_change(t->dev); 697 return err; 698 } 699 700 static void 701 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 702 { 703 p->laddr = u->laddr; 704 p->raddr = u->raddr; 705 p->link = u->link; 706 p->i_key = u->i_key; 707 p->o_key = u->o_key; 708 p->proto = u->proto; 709 710 memcpy(p->name, u->name, sizeof(u->name)); 711 } 712 713 static void 714 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 715 { 716 u->laddr = p->laddr; 717 u->raddr = p->raddr; 718 u->link = p->link; 719 u->i_key = p->i_key; 720 u->o_key = p->o_key; 721 if (u->i_key) 722 u->i_flags |= GRE_KEY; 723 if (u->o_key) 724 u->o_flags |= GRE_KEY; 725 u->proto = p->proto; 726 727 memcpy(u->name, p->name, sizeof(u->name)); 728 } 729 730 /** 731 * vti6_tnl_ioctl - configure vti6 tunnels from userspace 732 * @dev: virtual device associated with tunnel 733 * @ifr: parameters passed from userspace 734 * @cmd: command to be performed 735 * 736 * Description: 737 * vti6_ioctl() is used for managing vti6 tunnels 738 * from userspace. 739 * 740 * The possible commands are the following: 741 * %SIOCGETTUNNEL: get tunnel parameters for device 742 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 743 * %SIOCCHGTUNNEL: change tunnel parameters to those given 744 * %SIOCDELTUNNEL: delete tunnel 745 * 746 * The fallback device "ip6_vti0", created during module 747 * initialization, can be used for creating other tunnel devices. 748 * 749 * Return: 750 * 0 on success, 751 * %-EFAULT if unable to copy data to or from userspace, 752 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 753 * %-EINVAL if passed tunnel parameters are invalid, 754 * %-EEXIST if changing a tunnel's parameters would cause a conflict 755 * %-ENODEV if attempting to change or delete a nonexisting device 756 **/ 757 static int 758 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 759 { 760 int err = 0; 761 struct ip6_tnl_parm2 p; 762 struct __ip6_tnl_parm p1; 763 struct ip6_tnl *t = NULL; 764 struct net *net = dev_net(dev); 765 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 766 767 switch (cmd) { 768 case SIOCGETTUNNEL: 769 if (dev == ip6n->fb_tnl_dev) { 770 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 771 err = -EFAULT; 772 break; 773 } 774 vti6_parm_from_user(&p1, &p); 775 t = vti6_locate(net, &p1, 0); 776 } else { 777 memset(&p, 0, sizeof(p)); 778 } 779 if (!t) 780 t = netdev_priv(dev); 781 vti6_parm_to_user(&p, &t->parms); 782 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 783 err = -EFAULT; 784 break; 785 case SIOCADDTUNNEL: 786 case SIOCCHGTUNNEL: 787 err = -EPERM; 788 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 789 break; 790 err = -EFAULT; 791 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 792 break; 793 err = -EINVAL; 794 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 795 break; 796 vti6_parm_from_user(&p1, &p); 797 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 798 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 799 if (t) { 800 if (t->dev != dev) { 801 err = -EEXIST; 802 break; 803 } 804 } else 805 t = netdev_priv(dev); 806 807 err = vti6_update(t, &p1); 808 } 809 if (t) { 810 err = 0; 811 vti6_parm_to_user(&p, &t->parms); 812 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 813 err = -EFAULT; 814 815 } else 816 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 817 break; 818 case SIOCDELTUNNEL: 819 err = -EPERM; 820 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 821 break; 822 823 if (dev == ip6n->fb_tnl_dev) { 824 err = -EFAULT; 825 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 826 break; 827 err = -ENOENT; 828 vti6_parm_from_user(&p1, &p); 829 t = vti6_locate(net, &p1, 0); 830 if (!t) 831 break; 832 err = -EPERM; 833 if (t->dev == ip6n->fb_tnl_dev) 834 break; 835 dev = t->dev; 836 } 837 err = 0; 838 unregister_netdevice(dev); 839 break; 840 default: 841 err = -EINVAL; 842 } 843 return err; 844 } 845 846 static const struct net_device_ops vti6_netdev_ops = { 847 .ndo_init = vti6_dev_init, 848 .ndo_uninit = vti6_dev_uninit, 849 .ndo_start_xmit = vti6_tnl_xmit, 850 .ndo_do_ioctl = vti6_ioctl, 851 .ndo_get_stats64 = ip_tunnel_get_stats64, 852 .ndo_get_iflink = ip6_tnl_get_iflink, 853 }; 854 855 /** 856 * vti6_dev_setup - setup virtual tunnel device 857 * @dev: virtual device associated with tunnel 858 * 859 * Description: 860 * Initialize function pointers and device parameters 861 **/ 862 static void vti6_dev_setup(struct net_device *dev) 863 { 864 dev->netdev_ops = &vti6_netdev_ops; 865 dev->needs_free_netdev = true; 866 dev->priv_destructor = vti6_dev_free; 867 868 dev->type = ARPHRD_TUNNEL6; 869 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr); 870 dev->mtu = ETH_DATA_LEN; 871 dev->min_mtu = IPV6_MIN_MTU; 872 dev->max_mtu = IP_MAX_MTU; 873 dev->flags |= IFF_NOARP; 874 dev->addr_len = sizeof(struct in6_addr); 875 netif_keep_dst(dev); 876 /* This perm addr will be used as interface identifier by IPv6 */ 877 dev->addr_assign_type = NET_ADDR_RANDOM; 878 eth_random_addr(dev->perm_addr); 879 } 880 881 /** 882 * vti6_dev_init_gen - general initializer for all tunnel devices 883 * @dev: virtual device associated with tunnel 884 **/ 885 static inline int vti6_dev_init_gen(struct net_device *dev) 886 { 887 struct ip6_tnl *t = netdev_priv(dev); 888 889 t->dev = dev; 890 t->net = dev_net(dev); 891 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 892 if (!dev->tstats) 893 return -ENOMEM; 894 return 0; 895 } 896 897 /** 898 * vti6_dev_init - initializer for all non fallback tunnel devices 899 * @dev: virtual device associated with tunnel 900 **/ 901 static int vti6_dev_init(struct net_device *dev) 902 { 903 struct ip6_tnl *t = netdev_priv(dev); 904 int err = vti6_dev_init_gen(dev); 905 906 if (err) 907 return err; 908 vti6_link_config(t); 909 return 0; 910 } 911 912 /** 913 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 914 * @dev: fallback device 915 * 916 * Return: 0 917 **/ 918 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 919 { 920 struct ip6_tnl *t = netdev_priv(dev); 921 struct net *net = dev_net(dev); 922 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 923 924 t->parms.proto = IPPROTO_IPV6; 925 dev_hold(dev); 926 927 rcu_assign_pointer(ip6n->tnls_wc[0], t); 928 return 0; 929 } 930 931 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 932 struct netlink_ext_ack *extack) 933 { 934 return 0; 935 } 936 937 static void vti6_netlink_parms(struct nlattr *data[], 938 struct __ip6_tnl_parm *parms) 939 { 940 memset(parms, 0, sizeof(*parms)); 941 942 if (!data) 943 return; 944 945 if (data[IFLA_VTI_LINK]) 946 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 947 948 if (data[IFLA_VTI_LOCAL]) 949 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 950 951 if (data[IFLA_VTI_REMOTE]) 952 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 953 954 if (data[IFLA_VTI_IKEY]) 955 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 956 957 if (data[IFLA_VTI_OKEY]) 958 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 959 960 if (data[IFLA_VTI_FWMARK]) 961 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 962 } 963 964 static int vti6_newlink(struct net *src_net, struct net_device *dev, 965 struct nlattr *tb[], struct nlattr *data[], 966 struct netlink_ext_ack *extack) 967 { 968 struct net *net = dev_net(dev); 969 struct ip6_tnl *nt; 970 971 nt = netdev_priv(dev); 972 vti6_netlink_parms(data, &nt->parms); 973 974 nt->parms.proto = IPPROTO_IPV6; 975 976 if (vti6_locate(net, &nt->parms, 0)) 977 return -EEXIST; 978 979 return vti6_tnl_create2(dev); 980 } 981 982 static void vti6_dellink(struct net_device *dev, struct list_head *head) 983 { 984 struct net *net = dev_net(dev); 985 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 986 987 if (dev != ip6n->fb_tnl_dev) 988 unregister_netdevice_queue(dev, head); 989 } 990 991 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 992 struct nlattr *data[], 993 struct netlink_ext_ack *extack) 994 { 995 struct ip6_tnl *t; 996 struct __ip6_tnl_parm p; 997 struct net *net = dev_net(dev); 998 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 999 1000 if (dev == ip6n->fb_tnl_dev) 1001 return -EINVAL; 1002 1003 vti6_netlink_parms(data, &p); 1004 1005 t = vti6_locate(net, &p, 0); 1006 1007 if (t) { 1008 if (t->dev != dev) 1009 return -EEXIST; 1010 } else 1011 t = netdev_priv(dev); 1012 1013 return vti6_update(t, &p); 1014 } 1015 1016 static size_t vti6_get_size(const struct net_device *dev) 1017 { 1018 return 1019 /* IFLA_VTI_LINK */ 1020 nla_total_size(4) + 1021 /* IFLA_VTI_LOCAL */ 1022 nla_total_size(sizeof(struct in6_addr)) + 1023 /* IFLA_VTI_REMOTE */ 1024 nla_total_size(sizeof(struct in6_addr)) + 1025 /* IFLA_VTI_IKEY */ 1026 nla_total_size(4) + 1027 /* IFLA_VTI_OKEY */ 1028 nla_total_size(4) + 1029 /* IFLA_VTI_FWMARK */ 1030 nla_total_size(4) + 1031 0; 1032 } 1033 1034 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1035 { 1036 struct ip6_tnl *tunnel = netdev_priv(dev); 1037 struct __ip6_tnl_parm *parm = &tunnel->parms; 1038 1039 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1040 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1041 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1042 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1043 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1044 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1045 goto nla_put_failure; 1046 return 0; 1047 1048 nla_put_failure: 1049 return -EMSGSIZE; 1050 } 1051 1052 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1053 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1054 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1055 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1056 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1057 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1058 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1059 }; 1060 1061 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1062 .kind = "vti6", 1063 .maxtype = IFLA_VTI_MAX, 1064 .policy = vti6_policy, 1065 .priv_size = sizeof(struct ip6_tnl), 1066 .setup = vti6_dev_setup, 1067 .validate = vti6_validate, 1068 .newlink = vti6_newlink, 1069 .dellink = vti6_dellink, 1070 .changelink = vti6_changelink, 1071 .get_size = vti6_get_size, 1072 .fill_info = vti6_fill_info, 1073 .get_link_net = ip6_tnl_get_link_net, 1074 }; 1075 1076 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1077 struct list_head *list) 1078 { 1079 int h; 1080 struct ip6_tnl *t; 1081 1082 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1083 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1084 while (t) { 1085 unregister_netdevice_queue(t->dev, list); 1086 t = rtnl_dereference(t->next); 1087 } 1088 } 1089 1090 t = rtnl_dereference(ip6n->tnls_wc[0]); 1091 unregister_netdevice_queue(t->dev, list); 1092 } 1093 1094 static int __net_init vti6_init_net(struct net *net) 1095 { 1096 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1097 struct ip6_tnl *t = NULL; 1098 int err; 1099 1100 ip6n->tnls[0] = ip6n->tnls_wc; 1101 ip6n->tnls[1] = ip6n->tnls_r_l; 1102 1103 err = -ENOMEM; 1104 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1105 NET_NAME_UNKNOWN, vti6_dev_setup); 1106 1107 if (!ip6n->fb_tnl_dev) 1108 goto err_alloc_dev; 1109 dev_net_set(ip6n->fb_tnl_dev, net); 1110 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1111 1112 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1113 if (err < 0) 1114 goto err_register; 1115 1116 err = register_netdev(ip6n->fb_tnl_dev); 1117 if (err < 0) 1118 goto err_register; 1119 1120 t = netdev_priv(ip6n->fb_tnl_dev); 1121 1122 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1123 return 0; 1124 1125 err_register: 1126 free_netdev(ip6n->fb_tnl_dev); 1127 err_alloc_dev: 1128 return err; 1129 } 1130 1131 static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1132 { 1133 struct vti6_net *ip6n; 1134 struct net *net; 1135 LIST_HEAD(list); 1136 1137 rtnl_lock(); 1138 list_for_each_entry(net, net_list, exit_list) { 1139 ip6n = net_generic(net, vti6_net_id); 1140 vti6_destroy_tunnels(ip6n, &list); 1141 } 1142 unregister_netdevice_many(&list); 1143 rtnl_unlock(); 1144 } 1145 1146 static struct pernet_operations vti6_net_ops = { 1147 .init = vti6_init_net, 1148 .exit_batch = vti6_exit_batch_net, 1149 .id = &vti6_net_id, 1150 .size = sizeof(struct vti6_net), 1151 }; 1152 1153 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1154 .handler = vti6_rcv, 1155 .cb_handler = vti6_rcv_cb, 1156 .err_handler = vti6_err, 1157 .priority = 100, 1158 }; 1159 1160 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1161 .handler = vti6_rcv, 1162 .cb_handler = vti6_rcv_cb, 1163 .err_handler = vti6_err, 1164 .priority = 100, 1165 }; 1166 1167 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1168 .handler = vti6_rcv, 1169 .cb_handler = vti6_rcv_cb, 1170 .err_handler = vti6_err, 1171 .priority = 100, 1172 }; 1173 1174 /** 1175 * vti6_tunnel_init - register protocol and reserve needed resources 1176 * 1177 * Return: 0 on success 1178 **/ 1179 static int __init vti6_tunnel_init(void) 1180 { 1181 const char *msg; 1182 int err; 1183 1184 msg = "tunnel device"; 1185 err = register_pernet_device(&vti6_net_ops); 1186 if (err < 0) 1187 goto pernet_dev_failed; 1188 1189 msg = "tunnel protocols"; 1190 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1191 if (err < 0) 1192 goto xfrm_proto_esp_failed; 1193 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1194 if (err < 0) 1195 goto xfrm_proto_ah_failed; 1196 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1197 if (err < 0) 1198 goto xfrm_proto_comp_failed; 1199 1200 msg = "netlink interface"; 1201 err = rtnl_link_register(&vti6_link_ops); 1202 if (err < 0) 1203 goto rtnl_link_failed; 1204 1205 return 0; 1206 1207 rtnl_link_failed: 1208 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1209 xfrm_proto_comp_failed: 1210 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1211 xfrm_proto_ah_failed: 1212 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1213 xfrm_proto_esp_failed: 1214 unregister_pernet_device(&vti6_net_ops); 1215 pernet_dev_failed: 1216 pr_err("vti6 init: failed to register %s\n", msg); 1217 return err; 1218 } 1219 1220 /** 1221 * vti6_tunnel_cleanup - free resources and unregister protocol 1222 **/ 1223 static void __exit vti6_tunnel_cleanup(void) 1224 { 1225 rtnl_link_unregister(&vti6_link_ops); 1226 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1227 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1228 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1229 unregister_pernet_device(&vti6_net_ops); 1230 } 1231 1232 module_init(vti6_tunnel_init); 1233 module_exit(vti6_tunnel_cleanup); 1234 MODULE_LICENSE("GPL"); 1235 MODULE_ALIAS_RTNL_LINK("vti6"); 1236 MODULE_ALIAS_NETDEV("ip6_vti0"); 1237 MODULE_AUTHOR("Steffen Klassert"); 1238 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1239