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, bool keep_mtu) 626 { 627 struct net_device *dev = t->dev; 628 struct __ip6_tnl_parm *p = &t->parms; 629 struct net_device *tdev = NULL; 630 int mtu; 631 632 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 633 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 634 635 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 636 IP6_TNL_F_CAP_PER_PACKET); 637 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 638 639 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 640 dev->flags |= IFF_POINTOPOINT; 641 else 642 dev->flags &= ~IFF_POINTOPOINT; 643 644 if (keep_mtu && dev->mtu) { 645 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu); 646 return; 647 } 648 649 if (p->flags & IP6_TNL_F_CAP_XMIT) { 650 int strict = (ipv6_addr_type(&p->raddr) & 651 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 652 struct rt6_info *rt = rt6_lookup(t->net, 653 &p->raddr, &p->laddr, 654 p->link, NULL, strict); 655 656 if (rt) 657 tdev = rt->dst.dev; 658 ip6_rt_put(rt); 659 } 660 661 if (!tdev && p->link) 662 tdev = __dev_get_by_index(t->net, p->link); 663 664 if (tdev) 665 mtu = tdev->mtu - sizeof(struct ipv6hdr); 666 else 667 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 668 669 dev->mtu = max_t(int, mtu, IPV6_MIN_MTU); 670 } 671 672 /** 673 * vti6_tnl_change - update the tunnel parameters 674 * @t: tunnel to be changed 675 * @p: tunnel configuration parameters 676 * @keep_mtu: MTU was set from userspace, don't re-compute it 677 * 678 * Description: 679 * vti6_tnl_change() updates the tunnel parameters 680 **/ 681 static int 682 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 683 bool keep_mtu) 684 { 685 t->parms.laddr = p->laddr; 686 t->parms.raddr = p->raddr; 687 t->parms.link = p->link; 688 t->parms.i_key = p->i_key; 689 t->parms.o_key = p->o_key; 690 t->parms.proto = p->proto; 691 t->parms.fwmark = p->fwmark; 692 dst_cache_reset(&t->dst_cache); 693 vti6_link_config(t, keep_mtu); 694 return 0; 695 } 696 697 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 698 bool keep_mtu) 699 { 700 struct net *net = dev_net(t->dev); 701 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 702 int err; 703 704 vti6_tnl_unlink(ip6n, t); 705 synchronize_net(); 706 err = vti6_tnl_change(t, p, keep_mtu); 707 vti6_tnl_link(ip6n, t); 708 netdev_state_change(t->dev); 709 return err; 710 } 711 712 static void 713 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 714 { 715 p->laddr = u->laddr; 716 p->raddr = u->raddr; 717 p->link = u->link; 718 p->i_key = u->i_key; 719 p->o_key = u->o_key; 720 p->proto = u->proto; 721 722 memcpy(p->name, u->name, sizeof(u->name)); 723 } 724 725 static void 726 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 727 { 728 u->laddr = p->laddr; 729 u->raddr = p->raddr; 730 u->link = p->link; 731 u->i_key = p->i_key; 732 u->o_key = p->o_key; 733 if (u->i_key) 734 u->i_flags |= GRE_KEY; 735 if (u->o_key) 736 u->o_flags |= GRE_KEY; 737 u->proto = p->proto; 738 739 memcpy(u->name, p->name, sizeof(u->name)); 740 } 741 742 /** 743 * vti6_tnl_ioctl - configure vti6 tunnels from userspace 744 * @dev: virtual device associated with tunnel 745 * @ifr: parameters passed from userspace 746 * @cmd: command to be performed 747 * 748 * Description: 749 * vti6_ioctl() is used for managing vti6 tunnels 750 * from userspace. 751 * 752 * The possible commands are the following: 753 * %SIOCGETTUNNEL: get tunnel parameters for device 754 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 755 * %SIOCCHGTUNNEL: change tunnel parameters to those given 756 * %SIOCDELTUNNEL: delete tunnel 757 * 758 * The fallback device "ip6_vti0", created during module 759 * initialization, can be used for creating other tunnel devices. 760 * 761 * Return: 762 * 0 on success, 763 * %-EFAULT if unable to copy data to or from userspace, 764 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 765 * %-EINVAL if passed tunnel parameters are invalid, 766 * %-EEXIST if changing a tunnel's parameters would cause a conflict 767 * %-ENODEV if attempting to change or delete a nonexisting device 768 **/ 769 static int 770 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 771 { 772 int err = 0; 773 struct ip6_tnl_parm2 p; 774 struct __ip6_tnl_parm p1; 775 struct ip6_tnl *t = NULL; 776 struct net *net = dev_net(dev); 777 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 778 779 switch (cmd) { 780 case SIOCGETTUNNEL: 781 if (dev == ip6n->fb_tnl_dev) { 782 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 783 err = -EFAULT; 784 break; 785 } 786 vti6_parm_from_user(&p1, &p); 787 t = vti6_locate(net, &p1, 0); 788 } else { 789 memset(&p, 0, sizeof(p)); 790 } 791 if (!t) 792 t = netdev_priv(dev); 793 vti6_parm_to_user(&p, &t->parms); 794 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 795 err = -EFAULT; 796 break; 797 case SIOCADDTUNNEL: 798 case SIOCCHGTUNNEL: 799 err = -EPERM; 800 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 801 break; 802 err = -EFAULT; 803 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 804 break; 805 err = -EINVAL; 806 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 807 break; 808 vti6_parm_from_user(&p1, &p); 809 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 810 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 811 if (t) { 812 if (t->dev != dev) { 813 err = -EEXIST; 814 break; 815 } 816 } else 817 t = netdev_priv(dev); 818 819 err = vti6_update(t, &p1, false); 820 } 821 if (t) { 822 err = 0; 823 vti6_parm_to_user(&p, &t->parms); 824 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 825 err = -EFAULT; 826 827 } else 828 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 829 break; 830 case SIOCDELTUNNEL: 831 err = -EPERM; 832 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 833 break; 834 835 if (dev == ip6n->fb_tnl_dev) { 836 err = -EFAULT; 837 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 838 break; 839 err = -ENOENT; 840 vti6_parm_from_user(&p1, &p); 841 t = vti6_locate(net, &p1, 0); 842 if (!t) 843 break; 844 err = -EPERM; 845 if (t->dev == ip6n->fb_tnl_dev) 846 break; 847 dev = t->dev; 848 } 849 err = 0; 850 unregister_netdevice(dev); 851 break; 852 default: 853 err = -EINVAL; 854 } 855 return err; 856 } 857 858 static const struct net_device_ops vti6_netdev_ops = { 859 .ndo_init = vti6_dev_init, 860 .ndo_uninit = vti6_dev_uninit, 861 .ndo_start_xmit = vti6_tnl_xmit, 862 .ndo_do_ioctl = vti6_ioctl, 863 .ndo_get_stats64 = ip_tunnel_get_stats64, 864 .ndo_get_iflink = ip6_tnl_get_iflink, 865 }; 866 867 /** 868 * vti6_dev_setup - setup virtual tunnel device 869 * @dev: virtual device associated with tunnel 870 * 871 * Description: 872 * Initialize function pointers and device parameters 873 **/ 874 static void vti6_dev_setup(struct net_device *dev) 875 { 876 dev->netdev_ops = &vti6_netdev_ops; 877 dev->needs_free_netdev = true; 878 dev->priv_destructor = vti6_dev_free; 879 880 dev->type = ARPHRD_TUNNEL6; 881 dev->min_mtu = IPV6_MIN_MTU; 882 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 883 dev->flags |= IFF_NOARP; 884 dev->addr_len = sizeof(struct in6_addr); 885 netif_keep_dst(dev); 886 /* This perm addr will be used as interface identifier by IPv6 */ 887 dev->addr_assign_type = NET_ADDR_RANDOM; 888 eth_random_addr(dev->perm_addr); 889 } 890 891 /** 892 * vti6_dev_init_gen - general initializer for all tunnel devices 893 * @dev: virtual device associated with tunnel 894 **/ 895 static inline int vti6_dev_init_gen(struct net_device *dev) 896 { 897 struct ip6_tnl *t = netdev_priv(dev); 898 899 t->dev = dev; 900 t->net = dev_net(dev); 901 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 902 if (!dev->tstats) 903 return -ENOMEM; 904 return 0; 905 } 906 907 /** 908 * vti6_dev_init - initializer for all non fallback tunnel devices 909 * @dev: virtual device associated with tunnel 910 **/ 911 static int vti6_dev_init(struct net_device *dev) 912 { 913 struct ip6_tnl *t = netdev_priv(dev); 914 int err = vti6_dev_init_gen(dev); 915 916 if (err) 917 return err; 918 vti6_link_config(t, true); 919 return 0; 920 } 921 922 /** 923 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 924 * @dev: fallback device 925 * 926 * Return: 0 927 **/ 928 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 929 { 930 struct ip6_tnl *t = netdev_priv(dev); 931 struct net *net = dev_net(dev); 932 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 933 934 t->parms.proto = IPPROTO_IPV6; 935 dev_hold(dev); 936 937 rcu_assign_pointer(ip6n->tnls_wc[0], t); 938 return 0; 939 } 940 941 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 942 struct netlink_ext_ack *extack) 943 { 944 return 0; 945 } 946 947 static void vti6_netlink_parms(struct nlattr *data[], 948 struct __ip6_tnl_parm *parms) 949 { 950 memset(parms, 0, sizeof(*parms)); 951 952 if (!data) 953 return; 954 955 if (data[IFLA_VTI_LINK]) 956 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 957 958 if (data[IFLA_VTI_LOCAL]) 959 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 960 961 if (data[IFLA_VTI_REMOTE]) 962 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 963 964 if (data[IFLA_VTI_IKEY]) 965 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 966 967 if (data[IFLA_VTI_OKEY]) 968 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 969 970 if (data[IFLA_VTI_FWMARK]) 971 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 972 } 973 974 static int vti6_newlink(struct net *src_net, struct net_device *dev, 975 struct nlattr *tb[], struct nlattr *data[], 976 struct netlink_ext_ack *extack) 977 { 978 struct net *net = dev_net(dev); 979 struct ip6_tnl *nt; 980 981 nt = netdev_priv(dev); 982 vti6_netlink_parms(data, &nt->parms); 983 984 nt->parms.proto = IPPROTO_IPV6; 985 986 if (vti6_locate(net, &nt->parms, 0)) 987 return -EEXIST; 988 989 return vti6_tnl_create2(dev); 990 } 991 992 static void vti6_dellink(struct net_device *dev, struct list_head *head) 993 { 994 struct net *net = dev_net(dev); 995 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 996 997 if (dev != ip6n->fb_tnl_dev) 998 unregister_netdevice_queue(dev, head); 999 } 1000 1001 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1002 struct nlattr *data[], 1003 struct netlink_ext_ack *extack) 1004 { 1005 struct ip6_tnl *t; 1006 struct __ip6_tnl_parm p; 1007 struct net *net = dev_net(dev); 1008 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1009 1010 if (dev == ip6n->fb_tnl_dev) 1011 return -EINVAL; 1012 1013 vti6_netlink_parms(data, &p); 1014 1015 t = vti6_locate(net, &p, 0); 1016 1017 if (t) { 1018 if (t->dev != dev) 1019 return -EEXIST; 1020 } else 1021 t = netdev_priv(dev); 1022 1023 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1024 } 1025 1026 static size_t vti6_get_size(const struct net_device *dev) 1027 { 1028 return 1029 /* IFLA_VTI_LINK */ 1030 nla_total_size(4) + 1031 /* IFLA_VTI_LOCAL */ 1032 nla_total_size(sizeof(struct in6_addr)) + 1033 /* IFLA_VTI_REMOTE */ 1034 nla_total_size(sizeof(struct in6_addr)) + 1035 /* IFLA_VTI_IKEY */ 1036 nla_total_size(4) + 1037 /* IFLA_VTI_OKEY */ 1038 nla_total_size(4) + 1039 /* IFLA_VTI_FWMARK */ 1040 nla_total_size(4) + 1041 0; 1042 } 1043 1044 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1045 { 1046 struct ip6_tnl *tunnel = netdev_priv(dev); 1047 struct __ip6_tnl_parm *parm = &tunnel->parms; 1048 1049 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1050 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1051 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1052 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1053 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1054 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1055 goto nla_put_failure; 1056 return 0; 1057 1058 nla_put_failure: 1059 return -EMSGSIZE; 1060 } 1061 1062 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1063 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1064 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1065 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1066 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1067 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1068 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1069 }; 1070 1071 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1072 .kind = "vti6", 1073 .maxtype = IFLA_VTI_MAX, 1074 .policy = vti6_policy, 1075 .priv_size = sizeof(struct ip6_tnl), 1076 .setup = vti6_dev_setup, 1077 .validate = vti6_validate, 1078 .newlink = vti6_newlink, 1079 .dellink = vti6_dellink, 1080 .changelink = vti6_changelink, 1081 .get_size = vti6_get_size, 1082 .fill_info = vti6_fill_info, 1083 .get_link_net = ip6_tnl_get_link_net, 1084 }; 1085 1086 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1087 struct list_head *list) 1088 { 1089 int h; 1090 struct ip6_tnl *t; 1091 1092 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1093 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1094 while (t) { 1095 unregister_netdevice_queue(t->dev, list); 1096 t = rtnl_dereference(t->next); 1097 } 1098 } 1099 1100 t = rtnl_dereference(ip6n->tnls_wc[0]); 1101 unregister_netdevice_queue(t->dev, list); 1102 } 1103 1104 static int __net_init vti6_init_net(struct net *net) 1105 { 1106 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1107 struct ip6_tnl *t = NULL; 1108 int err; 1109 1110 ip6n->tnls[0] = ip6n->tnls_wc; 1111 ip6n->tnls[1] = ip6n->tnls_r_l; 1112 1113 err = -ENOMEM; 1114 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1115 NET_NAME_UNKNOWN, vti6_dev_setup); 1116 1117 if (!ip6n->fb_tnl_dev) 1118 goto err_alloc_dev; 1119 dev_net_set(ip6n->fb_tnl_dev, net); 1120 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1121 1122 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1123 if (err < 0) 1124 goto err_register; 1125 1126 err = register_netdev(ip6n->fb_tnl_dev); 1127 if (err < 0) 1128 goto err_register; 1129 1130 t = netdev_priv(ip6n->fb_tnl_dev); 1131 1132 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1133 return 0; 1134 1135 err_register: 1136 free_netdev(ip6n->fb_tnl_dev); 1137 err_alloc_dev: 1138 return err; 1139 } 1140 1141 static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1142 { 1143 struct vti6_net *ip6n; 1144 struct net *net; 1145 LIST_HEAD(list); 1146 1147 rtnl_lock(); 1148 list_for_each_entry(net, net_list, exit_list) { 1149 ip6n = net_generic(net, vti6_net_id); 1150 vti6_destroy_tunnels(ip6n, &list); 1151 } 1152 unregister_netdevice_many(&list); 1153 rtnl_unlock(); 1154 } 1155 1156 static struct pernet_operations vti6_net_ops = { 1157 .init = vti6_init_net, 1158 .exit_batch = vti6_exit_batch_net, 1159 .id = &vti6_net_id, 1160 .size = sizeof(struct vti6_net), 1161 }; 1162 1163 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1164 .handler = vti6_rcv, 1165 .cb_handler = vti6_rcv_cb, 1166 .err_handler = vti6_err, 1167 .priority = 100, 1168 }; 1169 1170 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1171 .handler = vti6_rcv, 1172 .cb_handler = vti6_rcv_cb, 1173 .err_handler = vti6_err, 1174 .priority = 100, 1175 }; 1176 1177 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1178 .handler = vti6_rcv, 1179 .cb_handler = vti6_rcv_cb, 1180 .err_handler = vti6_err, 1181 .priority = 100, 1182 }; 1183 1184 /** 1185 * vti6_tunnel_init - register protocol and reserve needed resources 1186 * 1187 * Return: 0 on success 1188 **/ 1189 static int __init vti6_tunnel_init(void) 1190 { 1191 const char *msg; 1192 int err; 1193 1194 msg = "tunnel device"; 1195 err = register_pernet_device(&vti6_net_ops); 1196 if (err < 0) 1197 goto pernet_dev_failed; 1198 1199 msg = "tunnel protocols"; 1200 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1201 if (err < 0) 1202 goto xfrm_proto_esp_failed; 1203 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1204 if (err < 0) 1205 goto xfrm_proto_ah_failed; 1206 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1207 if (err < 0) 1208 goto xfrm_proto_comp_failed; 1209 1210 msg = "netlink interface"; 1211 err = rtnl_link_register(&vti6_link_ops); 1212 if (err < 0) 1213 goto rtnl_link_failed; 1214 1215 return 0; 1216 1217 rtnl_link_failed: 1218 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1219 xfrm_proto_comp_failed: 1220 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1221 xfrm_proto_ah_failed: 1222 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1223 xfrm_proto_esp_failed: 1224 unregister_pernet_device(&vti6_net_ops); 1225 pernet_dev_failed: 1226 pr_err("vti6 init: failed to register %s\n", msg); 1227 return err; 1228 } 1229 1230 /** 1231 * vti6_tunnel_cleanup - free resources and unregister protocol 1232 **/ 1233 static void __exit vti6_tunnel_cleanup(void) 1234 { 1235 rtnl_link_unregister(&vti6_link_ops); 1236 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1237 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1238 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1239 unregister_pernet_device(&vti6_net_ops); 1240 } 1241 1242 module_init(vti6_tunnel_init); 1243 module_exit(vti6_tunnel_cleanup); 1244 MODULE_LICENSE("GPL"); 1245 MODULE_ALIAS_RTNL_LINK("vti6"); 1246 MODULE_ALIAS_NETDEV("ip6_vti0"); 1247 MODULE_AUTHOR("Steffen Klassert"); 1248 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1249