1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 virtual tunneling interface 4 * 5 * Copyright (C) 2013 secunet Security Networks AG 6 * 7 * Author: 8 * Steffen Klassert <steffen.klassert@secunet.com> 9 * 10 * Based on: 11 * net/ipv6/ip6_tunnel.c 12 */ 13 14 #include <linux/module.h> 15 #include <linux/capability.h> 16 #include <linux/errno.h> 17 #include <linux/types.h> 18 #include <linux/sockios.h> 19 #include <linux/icmp.h> 20 #include <linux/if.h> 21 #include <linux/in.h> 22 #include <linux/ip.h> 23 #include <linux/net.h> 24 #include <linux/in6.h> 25 #include <linux/netdevice.h> 26 #include <linux/if_arp.h> 27 #include <linux/icmpv6.h> 28 #include <linux/init.h> 29 #include <linux/route.h> 30 #include <linux/rtnetlink.h> 31 #include <linux/netfilter_ipv6.h> 32 #include <linux/slab.h> 33 #include <linux/hash.h> 34 35 #include <linux/uaccess.h> 36 #include <linux/atomic.h> 37 38 #include <net/icmp.h> 39 #include <net/ip.h> 40 #include <net/ip_tunnels.h> 41 #include <net/ipv6.h> 42 #include <net/ip6_route.h> 43 #include <net/addrconf.h> 44 #include <net/ip6_tunnel.h> 45 #include <net/xfrm.h> 46 #include <net/net_namespace.h> 47 #include <net/netns/generic.h> 48 #include <linux/etherdevice.h> 49 50 #define IP6_VTI_HASH_SIZE_SHIFT 5 51 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT) 52 53 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 54 { 55 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 56 57 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT); 58 } 59 60 static int vti6_dev_init(struct net_device *dev); 61 static void vti6_dev_setup(struct net_device *dev); 62 static struct rtnl_link_ops vti6_link_ops __read_mostly; 63 64 static unsigned int vti6_net_id __read_mostly; 65 struct vti6_net { 66 /* the vti6 tunnel fallback device */ 67 struct net_device *fb_tnl_dev; 68 /* lists for storing tunnels in use */ 69 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE]; 70 struct ip6_tnl __rcu *tnls_wc[1]; 71 struct ip6_tnl __rcu **tnls[2]; 72 }; 73 74 #define for_each_vti6_tunnel_rcu(start) \ 75 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 76 77 /** 78 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses 79 * @net: network namespace 80 * @remote: the address of the tunnel exit-point 81 * @local: the address of the tunnel entry-point 82 * 83 * Return: 84 * tunnel matching given end-points if found, 85 * else fallback tunnel if its device is up, 86 * else %NULL 87 **/ 88 static struct ip6_tnl * 89 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote, 90 const struct in6_addr *local) 91 { 92 unsigned int hash = HASH(remote, local); 93 struct ip6_tnl *t; 94 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 95 struct in6_addr any; 96 97 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 98 if (ipv6_addr_equal(local, &t->parms.laddr) && 99 ipv6_addr_equal(remote, &t->parms.raddr) && 100 (t->dev->flags & IFF_UP)) 101 return t; 102 } 103 104 memset(&any, 0, sizeof(any)); 105 hash = HASH(&any, local); 106 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 107 if (ipv6_addr_equal(local, &t->parms.laddr) && 108 (t->dev->flags & IFF_UP)) 109 return t; 110 } 111 112 hash = HASH(remote, &any); 113 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 114 if (ipv6_addr_equal(remote, &t->parms.raddr) && 115 (t->dev->flags & IFF_UP)) 116 return t; 117 } 118 119 t = rcu_dereference(ip6n->tnls_wc[0]); 120 if (t && (t->dev->flags & IFF_UP)) 121 return t; 122 123 return NULL; 124 } 125 126 /** 127 * vti6_tnl_bucket - get head of list matching given tunnel parameters 128 * @p: parameters containing tunnel end-points 129 * 130 * Description: 131 * vti6_tnl_bucket() returns the head of the list matching the 132 * &struct in6_addr entries laddr and raddr in @p. 133 * 134 * Return: head of IPv6 tunnel list 135 **/ 136 static struct ip6_tnl __rcu ** 137 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 138 { 139 const struct in6_addr *remote = &p->raddr; 140 const struct in6_addr *local = &p->laddr; 141 unsigned int h = 0; 142 int prio = 0; 143 144 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 145 prio = 1; 146 h = HASH(remote, local); 147 } 148 return &ip6n->tnls[prio][h]; 149 } 150 151 static void 152 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 153 { 154 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 155 156 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 157 rcu_assign_pointer(*tp, t); 158 } 159 160 static void 161 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 162 { 163 struct ip6_tnl __rcu **tp; 164 struct ip6_tnl *iter; 165 166 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 167 (iter = rtnl_dereference(*tp)) != NULL; 168 tp = &iter->next) { 169 if (t == iter) { 170 rcu_assign_pointer(*tp, t->next); 171 break; 172 } 173 } 174 } 175 176 static void vti6_dev_free(struct net_device *dev) 177 { 178 free_percpu(dev->tstats); 179 } 180 181 static int vti6_tnl_create2(struct net_device *dev) 182 { 183 struct ip6_tnl *t = netdev_priv(dev); 184 struct net *net = dev_net(dev); 185 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 186 int err; 187 188 dev->rtnl_link_ops = &vti6_link_ops; 189 err = register_netdevice(dev); 190 if (err < 0) 191 goto out; 192 193 strcpy(t->parms.name, dev->name); 194 195 dev_hold(dev); 196 vti6_tnl_link(ip6n, t); 197 198 return 0; 199 200 out: 201 return err; 202 } 203 204 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 205 { 206 struct net_device *dev; 207 struct ip6_tnl *t; 208 char name[IFNAMSIZ]; 209 int err; 210 211 if (p->name[0]) { 212 if (!dev_valid_name(p->name)) 213 goto failed; 214 strlcpy(name, p->name, IFNAMSIZ); 215 } else { 216 sprintf(name, "ip6_vti%%d"); 217 } 218 219 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 220 if (!dev) 221 goto failed; 222 223 dev_net_set(dev, net); 224 225 t = netdev_priv(dev); 226 t->parms = *p; 227 t->net = dev_net(dev); 228 229 err = vti6_tnl_create2(dev); 230 if (err < 0) 231 goto failed_free; 232 233 return t; 234 235 failed_free: 236 free_netdev(dev); 237 failed: 238 return NULL; 239 } 240 241 /** 242 * vti6_locate - find or create tunnel matching given parameters 243 * @net: network namespace 244 * @p: tunnel parameters 245 * @create: != 0 if allowed to create new tunnel if no match found 246 * 247 * Description: 248 * vti6_locate() first tries to locate an existing tunnel 249 * based on @parms. If this is unsuccessful, but @create is set a new 250 * tunnel device is created and registered for use. 251 * 252 * Return: 253 * matching tunnel or NULL 254 **/ 255 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 256 int create) 257 { 258 const struct in6_addr *remote = &p->raddr; 259 const struct in6_addr *local = &p->laddr; 260 struct ip6_tnl __rcu **tp; 261 struct ip6_tnl *t; 262 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 263 264 for (tp = vti6_tnl_bucket(ip6n, p); 265 (t = rtnl_dereference(*tp)) != NULL; 266 tp = &t->next) { 267 if (ipv6_addr_equal(local, &t->parms.laddr) && 268 ipv6_addr_equal(remote, &t->parms.raddr)) { 269 if (create) 270 return NULL; 271 272 return t; 273 } 274 } 275 if (!create) 276 return NULL; 277 return vti6_tnl_create(net, p); 278 } 279 280 /** 281 * vti6_dev_uninit - tunnel device uninitializer 282 * @dev: the device to be destroyed 283 * 284 * Description: 285 * vti6_dev_uninit() removes tunnel from its list 286 **/ 287 static void vti6_dev_uninit(struct net_device *dev) 288 { 289 struct ip6_tnl *t = netdev_priv(dev); 290 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 291 292 if (dev == ip6n->fb_tnl_dev) 293 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 294 else 295 vti6_tnl_unlink(ip6n, t); 296 dev_put(dev); 297 } 298 299 static int vti6_rcv(struct sk_buff *skb) 300 { 301 struct ip6_tnl *t; 302 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 303 304 rcu_read_lock(); 305 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 306 if (t) { 307 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 308 rcu_read_unlock(); 309 goto discard; 310 } 311 312 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 313 rcu_read_unlock(); 314 return 0; 315 } 316 317 ipv6h = ipv6_hdr(skb); 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 const 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->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 fl->u.ip6.flowi6_oif = dev->ifindex; 454 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 455 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 456 if (dst->error) { 457 dst_release(dst); 458 dst = NULL; 459 goto tx_err_link_failure; 460 } 461 skb_dst_set(skb, dst); 462 } 463 464 dst_hold(dst); 465 dst = xfrm_lookup(t->net, dst, fl, NULL, 0); 466 if (IS_ERR(dst)) { 467 err = PTR_ERR(dst); 468 dst = NULL; 469 goto tx_err_link_failure; 470 } 471 472 x = dst->xfrm; 473 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 474 goto tx_err_link_failure; 475 476 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 477 (const struct in6_addr *)&x->id.daddr)) 478 goto tx_err_link_failure; 479 480 tdev = dst->dev; 481 482 if (tdev == dev) { 483 stats->collisions++; 484 net_warn_ratelimited("%s: Local routing loop detected!\n", 485 t->parms.name); 486 goto tx_err_dst_release; 487 } 488 489 mtu = dst_mtu(dst); 490 if (skb->len > mtu) { 491 skb_dst_update_pmtu_no_confirm(skb, mtu); 492 493 if (skb->protocol == htons(ETH_P_IPV6)) { 494 if (mtu < IPV6_MIN_MTU) 495 mtu = IPV6_MIN_MTU; 496 497 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 498 } else { 499 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 500 htonl(mtu)); 501 } 502 503 err = -EMSGSIZE; 504 goto tx_err_dst_release; 505 } 506 507 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 508 skb_dst_set(skb, dst); 509 skb->dev = skb_dst(skb)->dev; 510 511 err = dst_output(t->net, skb->sk, skb); 512 if (net_xmit_eval(err) == 0) 513 err = pkt_len; 514 iptunnel_xmit_stats(dev, err); 515 516 return 0; 517 tx_err_link_failure: 518 stats->tx_carrier_errors++; 519 dst_link_failure(skb); 520 tx_err_dst_release: 521 dst_release(dst); 522 return err; 523 } 524 525 static netdev_tx_t 526 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 527 { 528 struct ip6_tnl *t = netdev_priv(dev); 529 struct net_device_stats *stats = &t->dev->stats; 530 struct flowi fl; 531 int ret; 532 533 if (!pskb_inet_may_pull(skb)) 534 goto tx_err; 535 536 memset(&fl, 0, sizeof(fl)); 537 538 switch (skb->protocol) { 539 case htons(ETH_P_IPV6): 540 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 541 vti6_addr_conflict(t, ipv6_hdr(skb))) 542 goto tx_err; 543 544 xfrm_decode_session(skb, &fl, AF_INET6); 545 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 546 break; 547 case htons(ETH_P_IP): 548 xfrm_decode_session(skb, &fl, AF_INET); 549 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 550 break; 551 default: 552 goto tx_err; 553 } 554 555 /* override mark with tunnel output key */ 556 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 557 558 ret = vti6_xmit(skb, dev, &fl); 559 if (ret < 0) 560 goto tx_err; 561 562 return NETDEV_TX_OK; 563 564 tx_err: 565 stats->tx_errors++; 566 stats->tx_dropped++; 567 kfree_skb(skb); 568 return NETDEV_TX_OK; 569 } 570 571 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 572 u8 type, u8 code, int offset, __be32 info) 573 { 574 __be32 spi; 575 __u32 mark; 576 struct xfrm_state *x; 577 struct ip6_tnl *t; 578 struct ip_esp_hdr *esph; 579 struct ip_auth_hdr *ah; 580 struct ip_comp_hdr *ipch; 581 struct net *net = dev_net(skb->dev); 582 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 583 int protocol = iph->nexthdr; 584 585 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 586 if (!t) 587 return -1; 588 589 mark = be32_to_cpu(t->parms.o_key); 590 591 switch (protocol) { 592 case IPPROTO_ESP: 593 esph = (struct ip_esp_hdr *)(skb->data + offset); 594 spi = esph->spi; 595 break; 596 case IPPROTO_AH: 597 ah = (struct ip_auth_hdr *)(skb->data + offset); 598 spi = ah->spi; 599 break; 600 case IPPROTO_COMP: 601 ipch = (struct ip_comp_hdr *)(skb->data + offset); 602 spi = htonl(ntohs(ipch->cpi)); 603 break; 604 default: 605 return 0; 606 } 607 608 if (type != ICMPV6_PKT_TOOBIG && 609 type != NDISC_REDIRECT) 610 return 0; 611 612 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 613 spi, protocol, AF_INET6); 614 if (!x) 615 return 0; 616 617 if (type == NDISC_REDIRECT) 618 ip6_redirect(skb, net, skb->dev->ifindex, 0, 619 sock_net_uid(net, NULL)); 620 else 621 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 622 xfrm_state_put(x); 623 624 return 0; 625 } 626 627 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 628 { 629 struct net_device *dev = t->dev; 630 struct __ip6_tnl_parm *p = &t->parms; 631 struct net_device *tdev = NULL; 632 int mtu; 633 634 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 635 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 636 637 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 638 IP6_TNL_F_CAP_PER_PACKET); 639 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 640 641 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 642 dev->flags |= IFF_POINTOPOINT; 643 else 644 dev->flags &= ~IFF_POINTOPOINT; 645 646 if (keep_mtu && dev->mtu) { 647 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu); 648 return; 649 } 650 651 if (p->flags & IP6_TNL_F_CAP_XMIT) { 652 int strict = (ipv6_addr_type(&p->raddr) & 653 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 654 struct rt6_info *rt = rt6_lookup(t->net, 655 &p->raddr, &p->laddr, 656 p->link, NULL, strict); 657 658 if (rt) 659 tdev = rt->dst.dev; 660 ip6_rt_put(rt); 661 } 662 663 if (!tdev && p->link) 664 tdev = __dev_get_by_index(t->net, p->link); 665 666 if (tdev) 667 mtu = tdev->mtu - sizeof(struct ipv6hdr); 668 else 669 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 670 671 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU); 672 } 673 674 /** 675 * vti6_tnl_change - update the tunnel parameters 676 * @t: tunnel to be changed 677 * @p: tunnel configuration parameters 678 * @keep_mtu: MTU was set from userspace, don't re-compute it 679 * 680 * Description: 681 * vti6_tnl_change() updates the tunnel parameters 682 **/ 683 static int 684 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 685 bool keep_mtu) 686 { 687 t->parms.laddr = p->laddr; 688 t->parms.raddr = p->raddr; 689 t->parms.link = p->link; 690 t->parms.i_key = p->i_key; 691 t->parms.o_key = p->o_key; 692 t->parms.proto = p->proto; 693 t->parms.fwmark = p->fwmark; 694 dst_cache_reset(&t->dst_cache); 695 vti6_link_config(t, keep_mtu); 696 return 0; 697 } 698 699 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 700 bool keep_mtu) 701 { 702 struct net *net = dev_net(t->dev); 703 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 704 int err; 705 706 vti6_tnl_unlink(ip6n, t); 707 synchronize_net(); 708 err = vti6_tnl_change(t, p, keep_mtu); 709 vti6_tnl_link(ip6n, t); 710 netdev_state_change(t->dev); 711 return err; 712 } 713 714 static void 715 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 716 { 717 p->laddr = u->laddr; 718 p->raddr = u->raddr; 719 p->link = u->link; 720 p->i_key = u->i_key; 721 p->o_key = u->o_key; 722 p->proto = u->proto; 723 724 memcpy(p->name, u->name, sizeof(u->name)); 725 } 726 727 static void 728 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 729 { 730 u->laddr = p->laddr; 731 u->raddr = p->raddr; 732 u->link = p->link; 733 u->i_key = p->i_key; 734 u->o_key = p->o_key; 735 if (u->i_key) 736 u->i_flags |= GRE_KEY; 737 if (u->o_key) 738 u->o_flags |= GRE_KEY; 739 u->proto = p->proto; 740 741 memcpy(u->name, p->name, sizeof(u->name)); 742 } 743 744 /** 745 * vti6_ioctl - configure vti6 tunnels from userspace 746 * @dev: virtual device associated with tunnel 747 * @ifr: parameters passed from userspace 748 * @cmd: command to be performed 749 * 750 * Description: 751 * vti6_ioctl() is used for managing vti6 tunnels 752 * from userspace. 753 * 754 * The possible commands are the following: 755 * %SIOCGETTUNNEL: get tunnel parameters for device 756 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 757 * %SIOCCHGTUNNEL: change tunnel parameters to those given 758 * %SIOCDELTUNNEL: delete tunnel 759 * 760 * The fallback device "ip6_vti0", created during module 761 * initialization, can be used for creating other tunnel devices. 762 * 763 * Return: 764 * 0 on success, 765 * %-EFAULT if unable to copy data to or from userspace, 766 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 767 * %-EINVAL if passed tunnel parameters are invalid, 768 * %-EEXIST if changing a tunnel's parameters would cause a conflict 769 * %-ENODEV if attempting to change or delete a nonexisting device 770 **/ 771 static int 772 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 773 { 774 int err = 0; 775 struct ip6_tnl_parm2 p; 776 struct __ip6_tnl_parm p1; 777 struct ip6_tnl *t = NULL; 778 struct net *net = dev_net(dev); 779 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 780 781 switch (cmd) { 782 case SIOCGETTUNNEL: 783 if (dev == ip6n->fb_tnl_dev) { 784 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 785 err = -EFAULT; 786 break; 787 } 788 vti6_parm_from_user(&p1, &p); 789 t = vti6_locate(net, &p1, 0); 790 } else { 791 memset(&p, 0, sizeof(p)); 792 } 793 if (!t) 794 t = netdev_priv(dev); 795 vti6_parm_to_user(&p, &t->parms); 796 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 797 err = -EFAULT; 798 break; 799 case SIOCADDTUNNEL: 800 case SIOCCHGTUNNEL: 801 err = -EPERM; 802 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 803 break; 804 err = -EFAULT; 805 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 806 break; 807 err = -EINVAL; 808 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 809 break; 810 vti6_parm_from_user(&p1, &p); 811 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 812 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 813 if (t) { 814 if (t->dev != dev) { 815 err = -EEXIST; 816 break; 817 } 818 } else 819 t = netdev_priv(dev); 820 821 err = vti6_update(t, &p1, false); 822 } 823 if (t) { 824 err = 0; 825 vti6_parm_to_user(&p, &t->parms); 826 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 827 err = -EFAULT; 828 829 } else 830 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 831 break; 832 case SIOCDELTUNNEL: 833 err = -EPERM; 834 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 835 break; 836 837 if (dev == ip6n->fb_tnl_dev) { 838 err = -EFAULT; 839 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 840 break; 841 err = -ENOENT; 842 vti6_parm_from_user(&p1, &p); 843 t = vti6_locate(net, &p1, 0); 844 if (!t) 845 break; 846 err = -EPERM; 847 if (t->dev == ip6n->fb_tnl_dev) 848 break; 849 dev = t->dev; 850 } 851 err = 0; 852 unregister_netdevice(dev); 853 break; 854 default: 855 err = -EINVAL; 856 } 857 return err; 858 } 859 860 static const struct net_device_ops vti6_netdev_ops = { 861 .ndo_init = vti6_dev_init, 862 .ndo_uninit = vti6_dev_uninit, 863 .ndo_start_xmit = vti6_tnl_xmit, 864 .ndo_do_ioctl = vti6_ioctl, 865 .ndo_get_stats64 = ip_tunnel_get_stats64, 866 .ndo_get_iflink = ip6_tnl_get_iflink, 867 }; 868 869 /** 870 * vti6_dev_setup - setup virtual tunnel device 871 * @dev: virtual device associated with tunnel 872 * 873 * Description: 874 * Initialize function pointers and device parameters 875 **/ 876 static void vti6_dev_setup(struct net_device *dev) 877 { 878 dev->netdev_ops = &vti6_netdev_ops; 879 dev->needs_free_netdev = true; 880 dev->priv_destructor = vti6_dev_free; 881 882 dev->type = ARPHRD_TUNNEL6; 883 dev->min_mtu = IPV4_MIN_MTU; 884 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 885 dev->flags |= IFF_NOARP; 886 dev->addr_len = sizeof(struct in6_addr); 887 netif_keep_dst(dev); 888 /* This perm addr will be used as interface identifier by IPv6 */ 889 dev->addr_assign_type = NET_ADDR_RANDOM; 890 eth_random_addr(dev->perm_addr); 891 } 892 893 /** 894 * vti6_dev_init_gen - general initializer for all tunnel devices 895 * @dev: virtual device associated with tunnel 896 **/ 897 static inline int vti6_dev_init_gen(struct net_device *dev) 898 { 899 struct ip6_tnl *t = netdev_priv(dev); 900 901 t->dev = dev; 902 t->net = dev_net(dev); 903 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 904 if (!dev->tstats) 905 return -ENOMEM; 906 return 0; 907 } 908 909 /** 910 * vti6_dev_init - initializer for all non fallback tunnel devices 911 * @dev: virtual device associated with tunnel 912 **/ 913 static int vti6_dev_init(struct net_device *dev) 914 { 915 struct ip6_tnl *t = netdev_priv(dev); 916 int err = vti6_dev_init_gen(dev); 917 918 if (err) 919 return err; 920 vti6_link_config(t, true); 921 return 0; 922 } 923 924 /** 925 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 926 * @dev: fallback device 927 * 928 * Return: 0 929 **/ 930 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 931 { 932 struct ip6_tnl *t = netdev_priv(dev); 933 struct net *net = dev_net(dev); 934 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 935 936 t->parms.proto = IPPROTO_IPV6; 937 dev_hold(dev); 938 939 rcu_assign_pointer(ip6n->tnls_wc[0], t); 940 return 0; 941 } 942 943 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 944 struct netlink_ext_ack *extack) 945 { 946 return 0; 947 } 948 949 static void vti6_netlink_parms(struct nlattr *data[], 950 struct __ip6_tnl_parm *parms) 951 { 952 memset(parms, 0, sizeof(*parms)); 953 954 if (!data) 955 return; 956 957 if (data[IFLA_VTI_LINK]) 958 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 959 960 if (data[IFLA_VTI_LOCAL]) 961 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 962 963 if (data[IFLA_VTI_REMOTE]) 964 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 965 966 if (data[IFLA_VTI_IKEY]) 967 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 968 969 if (data[IFLA_VTI_OKEY]) 970 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 971 972 if (data[IFLA_VTI_FWMARK]) 973 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 974 } 975 976 static int vti6_newlink(struct net *src_net, struct net_device *dev, 977 struct nlattr *tb[], struct nlattr *data[], 978 struct netlink_ext_ack *extack) 979 { 980 struct net *net = dev_net(dev); 981 struct ip6_tnl *nt; 982 983 nt = netdev_priv(dev); 984 vti6_netlink_parms(data, &nt->parms); 985 986 nt->parms.proto = IPPROTO_IPV6; 987 988 if (vti6_locate(net, &nt->parms, 0)) 989 return -EEXIST; 990 991 return vti6_tnl_create2(dev); 992 } 993 994 static void vti6_dellink(struct net_device *dev, struct list_head *head) 995 { 996 struct net *net = dev_net(dev); 997 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 998 999 if (dev != ip6n->fb_tnl_dev) 1000 unregister_netdevice_queue(dev, head); 1001 } 1002 1003 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1004 struct nlattr *data[], 1005 struct netlink_ext_ack *extack) 1006 { 1007 struct ip6_tnl *t; 1008 struct __ip6_tnl_parm p; 1009 struct net *net = dev_net(dev); 1010 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1011 1012 if (dev == ip6n->fb_tnl_dev) 1013 return -EINVAL; 1014 1015 vti6_netlink_parms(data, &p); 1016 1017 t = vti6_locate(net, &p, 0); 1018 1019 if (t) { 1020 if (t->dev != dev) 1021 return -EEXIST; 1022 } else 1023 t = netdev_priv(dev); 1024 1025 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1026 } 1027 1028 static size_t vti6_get_size(const struct net_device *dev) 1029 { 1030 return 1031 /* IFLA_VTI_LINK */ 1032 nla_total_size(4) + 1033 /* IFLA_VTI_LOCAL */ 1034 nla_total_size(sizeof(struct in6_addr)) + 1035 /* IFLA_VTI_REMOTE */ 1036 nla_total_size(sizeof(struct in6_addr)) + 1037 /* IFLA_VTI_IKEY */ 1038 nla_total_size(4) + 1039 /* IFLA_VTI_OKEY */ 1040 nla_total_size(4) + 1041 /* IFLA_VTI_FWMARK */ 1042 nla_total_size(4) + 1043 0; 1044 } 1045 1046 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1047 { 1048 struct ip6_tnl *tunnel = netdev_priv(dev); 1049 struct __ip6_tnl_parm *parm = &tunnel->parms; 1050 1051 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1052 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1053 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1054 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1055 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1056 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1057 goto nla_put_failure; 1058 return 0; 1059 1060 nla_put_failure: 1061 return -EMSGSIZE; 1062 } 1063 1064 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1065 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1066 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1067 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1068 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1069 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1070 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1071 }; 1072 1073 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1074 .kind = "vti6", 1075 .maxtype = IFLA_VTI_MAX, 1076 .policy = vti6_policy, 1077 .priv_size = sizeof(struct ip6_tnl), 1078 .setup = vti6_dev_setup, 1079 .validate = vti6_validate, 1080 .newlink = vti6_newlink, 1081 .dellink = vti6_dellink, 1082 .changelink = vti6_changelink, 1083 .get_size = vti6_get_size, 1084 .fill_info = vti6_fill_info, 1085 .get_link_net = ip6_tnl_get_link_net, 1086 }; 1087 1088 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1089 struct list_head *list) 1090 { 1091 int h; 1092 struct ip6_tnl *t; 1093 1094 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1095 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1096 while (t) { 1097 unregister_netdevice_queue(t->dev, list); 1098 t = rtnl_dereference(t->next); 1099 } 1100 } 1101 1102 t = rtnl_dereference(ip6n->tnls_wc[0]); 1103 if (t) 1104 unregister_netdevice_queue(t->dev, list); 1105 } 1106 1107 static int __net_init vti6_init_net(struct net *net) 1108 { 1109 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1110 struct ip6_tnl *t = NULL; 1111 int err; 1112 1113 ip6n->tnls[0] = ip6n->tnls_wc; 1114 ip6n->tnls[1] = ip6n->tnls_r_l; 1115 1116 if (!net_has_fallback_tunnels(net)) 1117 return 0; 1118 err = -ENOMEM; 1119 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1120 NET_NAME_UNKNOWN, vti6_dev_setup); 1121 1122 if (!ip6n->fb_tnl_dev) 1123 goto err_alloc_dev; 1124 dev_net_set(ip6n->fb_tnl_dev, net); 1125 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1126 1127 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1128 if (err < 0) 1129 goto err_register; 1130 1131 err = register_netdev(ip6n->fb_tnl_dev); 1132 if (err < 0) 1133 goto err_register; 1134 1135 t = netdev_priv(ip6n->fb_tnl_dev); 1136 1137 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1138 return 0; 1139 1140 err_register: 1141 free_netdev(ip6n->fb_tnl_dev); 1142 err_alloc_dev: 1143 return err; 1144 } 1145 1146 static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1147 { 1148 struct vti6_net *ip6n; 1149 struct net *net; 1150 LIST_HEAD(list); 1151 1152 rtnl_lock(); 1153 list_for_each_entry(net, net_list, exit_list) { 1154 ip6n = net_generic(net, vti6_net_id); 1155 vti6_destroy_tunnels(ip6n, &list); 1156 } 1157 unregister_netdevice_many(&list); 1158 rtnl_unlock(); 1159 } 1160 1161 static struct pernet_operations vti6_net_ops = { 1162 .init = vti6_init_net, 1163 .exit_batch = vti6_exit_batch_net, 1164 .id = &vti6_net_id, 1165 .size = sizeof(struct vti6_net), 1166 }; 1167 1168 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1169 .handler = vti6_rcv, 1170 .cb_handler = vti6_rcv_cb, 1171 .err_handler = vti6_err, 1172 .priority = 100, 1173 }; 1174 1175 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1176 .handler = vti6_rcv, 1177 .cb_handler = vti6_rcv_cb, 1178 .err_handler = vti6_err, 1179 .priority = 100, 1180 }; 1181 1182 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1183 .handler = vti6_rcv, 1184 .cb_handler = vti6_rcv_cb, 1185 .err_handler = vti6_err, 1186 .priority = 100, 1187 }; 1188 1189 /** 1190 * vti6_tunnel_init - register protocol and reserve needed resources 1191 * 1192 * Return: 0 on success 1193 **/ 1194 static int __init vti6_tunnel_init(void) 1195 { 1196 const char *msg; 1197 int err; 1198 1199 msg = "tunnel device"; 1200 err = register_pernet_device(&vti6_net_ops); 1201 if (err < 0) 1202 goto pernet_dev_failed; 1203 1204 msg = "tunnel protocols"; 1205 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1206 if (err < 0) 1207 goto xfrm_proto_esp_failed; 1208 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1209 if (err < 0) 1210 goto xfrm_proto_ah_failed; 1211 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1212 if (err < 0) 1213 goto xfrm_proto_comp_failed; 1214 1215 msg = "netlink interface"; 1216 err = rtnl_link_register(&vti6_link_ops); 1217 if (err < 0) 1218 goto rtnl_link_failed; 1219 1220 return 0; 1221 1222 rtnl_link_failed: 1223 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1224 xfrm_proto_comp_failed: 1225 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1226 xfrm_proto_ah_failed: 1227 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1228 xfrm_proto_esp_failed: 1229 unregister_pernet_device(&vti6_net_ops); 1230 pernet_dev_failed: 1231 pr_err("vti6 init: failed to register %s\n", msg); 1232 return err; 1233 } 1234 1235 /** 1236 * vti6_tunnel_cleanup - free resources and unregister protocol 1237 **/ 1238 static void __exit vti6_tunnel_cleanup(void) 1239 { 1240 rtnl_link_unregister(&vti6_link_ops); 1241 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1242 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1243 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1244 unregister_pernet_device(&vti6_net_ops); 1245 } 1246 1247 module_init(vti6_tunnel_init); 1248 module_exit(vti6_tunnel_cleanup); 1249 MODULE_LICENSE("GPL"); 1250 MODULE_ALIAS_RTNL_LINK("vti6"); 1251 MODULE_ALIAS_NETDEV("ip6_vti0"); 1252 MODULE_AUTHOR("Steffen Klassert"); 1253 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1254