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 53static 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 60static int vti6_dev_init(struct net_device *dev); 61static void vti6_dev_setup(struct net_device *dev); 62static struct rtnl_link_ops vti6_link_ops __read_mostly; 63 64static unsigned int vti6_net_id __read_mostly; 65struct 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 **/ 88static struct ip6_tnl * 89vti6_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 **/ 136static struct ip6_tnl __rcu ** 137vti6_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 151static void 152vti6_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 160static void 161vti6_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 176static void vti6_dev_free(struct net_device *dev) 177{ 178 free_percpu(dev->tstats); 179} 180 181static 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 vti6_tnl_link(ip6n, t); 196 197 return 0; 198 199out: 200 return err; 201} 202 203static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 204{ 205 struct net_device *dev; 206 struct ip6_tnl *t; 207 char name[IFNAMSIZ]; 208 int err; 209 210 if (p->name[0]) { 211 if (!dev_valid_name(p->name)) 212 goto failed; 213 strlcpy(name, p->name, IFNAMSIZ); 214 } else { 215 sprintf(name, "ip6_vti%%d"); 216 } 217 218 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 219 if (!dev) 220 goto failed; 221 222 dev_net_set(dev, net); 223 224 t = netdev_priv(dev); 225 t->parms = *p; 226 t->net = dev_net(dev); 227 228 err = vti6_tnl_create2(dev); 229 if (err < 0) 230 goto failed_free; 231 232 return t; 233 234failed_free: 235 free_netdev(dev); 236failed: 237 return NULL; 238} 239 240/** 241 * vti6_locate - find or create tunnel matching given parameters 242 * @net: network namespace 243 * @p: tunnel parameters 244 * @create: != 0 if allowed to create new tunnel if no match found 245 * 246 * Description: 247 * vti6_locate() first tries to locate an existing tunnel 248 * based on @parms. If this is unsuccessful, but @create is set a new 249 * tunnel device is created and registered for use. 250 * 251 * Return: 252 * matching tunnel or NULL 253 **/ 254static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 255 int create) 256{ 257 const struct in6_addr *remote = &p->raddr; 258 const struct in6_addr *local = &p->laddr; 259 struct ip6_tnl __rcu **tp; 260 struct ip6_tnl *t; 261 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 262 263 for (tp = vti6_tnl_bucket(ip6n, p); 264 (t = rtnl_dereference(*tp)) != NULL; 265 tp = &t->next) { 266 if (ipv6_addr_equal(local, &t->parms.laddr) && 267 ipv6_addr_equal(remote, &t->parms.raddr)) { 268 if (create) 269 return NULL; 270 271 return t; 272 } 273 } 274 if (!create) 275 return NULL; 276 return vti6_tnl_create(net, p); 277} 278 279/** 280 * vti6_dev_uninit - tunnel device uninitializer 281 * @dev: the device to be destroyed 282 * 283 * Description: 284 * vti6_dev_uninit() removes tunnel from its list 285 **/ 286static void vti6_dev_uninit(struct net_device *dev) 287{ 288 struct ip6_tnl *t = netdev_priv(dev); 289 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 290 291 if (dev == ip6n->fb_tnl_dev) 292 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 293 else 294 vti6_tnl_unlink(ip6n, t); 295 dev_put(dev); 296} 297 298static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 299 int encap_type) 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 goto discard; 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 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 327 XFRM_SPI_SKB_CB(skb)->family = AF_INET6; 328 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr); 329 return xfrm_input(skb, nexthdr, spi, encap_type); 330 } 331 rcu_read_unlock(); 332 return -EINVAL; 333discard: 334 kfree_skb(skb); 335 return 0; 336} 337 338static int vti6_rcv(struct sk_buff *skb) 339{ 340 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff]; 341 342 return vti6_input_proto(skb, nexthdr, 0, 0); 343} 344 345static int vti6_rcv_cb(struct sk_buff *skb, int err) 346{ 347 unsigned short family; 348 struct net_device *dev; 349 struct xfrm_state *x; 350 const struct xfrm_mode *inner_mode; 351 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 352 u32 orig_mark = skb->mark; 353 int ret; 354 355 if (!t) 356 return 1; 357 358 dev = t->dev; 359 360 if (err) { 361 dev->stats.rx_errors++; 362 dev->stats.rx_dropped++; 363 364 return 0; 365 } 366 367 x = xfrm_input_state(skb); 368 369 inner_mode = &x->inner_mode; 370 371 if (x->sel.family == AF_UNSPEC) { 372 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 373 if (inner_mode == NULL) { 374 XFRM_INC_STATS(dev_net(skb->dev), 375 LINUX_MIB_XFRMINSTATEMODEERROR); 376 return -EINVAL; 377 } 378 } 379 380 family = inner_mode->family; 381 382 skb->mark = be32_to_cpu(t->parms.i_key); 383 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 384 skb->mark = orig_mark; 385 386 if (!ret) 387 return -EPERM; 388 389 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 390 skb->dev = dev; 391 dev_sw_netstats_rx_add(dev, skb->len); 392 393 return 0; 394} 395 396/** 397 * vti6_addr_conflict - compare packet addresses to tunnel's own 398 * @t: the outgoing tunnel device 399 * @hdr: IPv6 header from the incoming packet 400 * 401 * Description: 402 * Avoid trivial tunneling loop by checking that tunnel exit-point 403 * doesn't match source of incoming packet. 404 * 405 * Return: 406 * 1 if conflict, 407 * 0 else 408 **/ 409static inline bool 410vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 411{ 412 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 413} 414 415static bool vti6_state_check(const struct xfrm_state *x, 416 const struct in6_addr *dst, 417 const struct in6_addr *src) 418{ 419 xfrm_address_t *daddr = (xfrm_address_t *)dst; 420 xfrm_address_t *saddr = (xfrm_address_t *)src; 421 422 /* if there is no transform then this tunnel is not functional. 423 * Or if the xfrm is not mode tunnel. 424 */ 425 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 426 x->props.family != AF_INET6) 427 return false; 428 429 if (ipv6_addr_any(dst)) 430 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 431 432 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 433 return false; 434 435 return true; 436} 437 438/** 439 * vti6_xmit - send a packet 440 * @skb: the outgoing socket buffer 441 * @dev: the outgoing tunnel device 442 * @fl: the flow informations for the xfrm_lookup 443 **/ 444static int 445vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 446{ 447 struct ip6_tnl *t = netdev_priv(dev); 448 struct net_device_stats *stats = &t->dev->stats; 449 struct dst_entry *dst = skb_dst(skb); 450 struct net_device *tdev; 451 struct xfrm_state *x; 452 int pkt_len = skb->len; 453 int err = -1; 454 int mtu; 455 456 if (!dst) { 457 switch (skb->protocol) { 458 case htons(ETH_P_IP): { 459 struct rtable *rt; 460 461 fl->u.ip4.flowi4_oif = dev->ifindex; 462 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 463 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 464 if (IS_ERR(rt)) 465 goto tx_err_link_failure; 466 dst = &rt->dst; 467 skb_dst_set(skb, dst); 468 break; 469 } 470 case htons(ETH_P_IPV6): 471 fl->u.ip6.flowi6_oif = dev->ifindex; 472 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 473 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 474 if (dst->error) { 475 dst_release(dst); 476 dst = NULL; 477 goto tx_err_link_failure; 478 } 479 skb_dst_set(skb, dst); 480 break; 481 default: 482 goto tx_err_link_failure; 483 } 484 } 485 486 dst_hold(dst); 487 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 488 if (IS_ERR(dst)) { 489 err = PTR_ERR(dst); 490 dst = NULL; 491 goto tx_err_link_failure; 492 } 493 494 if (dst->flags & DST_XFRM_QUEUE) 495 goto xmit; 496 497 x = dst->xfrm; 498 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 499 goto tx_err_link_failure; 500 501 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 502 (const struct in6_addr *)&x->id.daddr)) 503 goto tx_err_link_failure; 504 505 tdev = dst->dev; 506 507 if (tdev == dev) { 508 stats->collisions++; 509 net_warn_ratelimited("%s: Local routing loop detected!\n", 510 t->parms.name); 511 goto tx_err_dst_release; 512 } 513 514 mtu = dst_mtu(dst); 515 if (skb->len > mtu) { 516 skb_dst_update_pmtu_no_confirm(skb, mtu); 517 518 if (skb->protocol == htons(ETH_P_IPV6)) { 519 if (mtu < IPV6_MIN_MTU) 520 mtu = IPV6_MIN_MTU; 521 522 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 523 } else { 524 if (!(ip_hdr(skb)->frag_off & htons(IP_DF))) 525 goto xmit; 526 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 527 htonl(mtu)); 528 } 529 530 err = -EMSGSIZE; 531 goto tx_err_dst_release; 532 } 533 534xmit: 535 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 536 skb_dst_set(skb, dst); 537 skb->dev = skb_dst(skb)->dev; 538 539 err = dst_output(t->net, skb->sk, skb); 540 if (net_xmit_eval(err) == 0) 541 err = pkt_len; 542 iptunnel_xmit_stats(dev, err); 543 544 return 0; 545tx_err_link_failure: 546 stats->tx_carrier_errors++; 547 dst_link_failure(skb); 548tx_err_dst_release: 549 dst_release(dst); 550 return err; 551} 552 553static netdev_tx_t 554vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 555{ 556 struct ip6_tnl *t = netdev_priv(dev); 557 struct net_device_stats *stats = &t->dev->stats; 558 struct flowi fl; 559 int ret; 560 561 if (!pskb_inet_may_pull(skb)) 562 goto tx_err; 563 564 memset(&fl, 0, sizeof(fl)); 565 566 switch (skb->protocol) { 567 case htons(ETH_P_IPV6): 568 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 569 vti6_addr_conflict(t, ipv6_hdr(skb))) 570 goto tx_err; 571 572 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 573 xfrm_decode_session(skb, &fl, AF_INET6); 574 break; 575 case htons(ETH_P_IP): 576 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 577 xfrm_decode_session(skb, &fl, AF_INET); 578 break; 579 default: 580 goto tx_err; 581 } 582 583 /* override mark with tunnel output key */ 584 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 585 586 ret = vti6_xmit(skb, dev, &fl); 587 if (ret < 0) 588 goto tx_err; 589 590 return NETDEV_TX_OK; 591 592tx_err: 593 stats->tx_errors++; 594 stats->tx_dropped++; 595 kfree_skb(skb); 596 return NETDEV_TX_OK; 597} 598 599static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 600 u8 type, u8 code, int offset, __be32 info) 601{ 602 __be32 spi; 603 __u32 mark; 604 struct xfrm_state *x; 605 struct ip6_tnl *t; 606 struct ip_esp_hdr *esph; 607 struct ip_auth_hdr *ah; 608 struct ip_comp_hdr *ipch; 609 struct net *net = dev_net(skb->dev); 610 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 611 int protocol = iph->nexthdr; 612 613 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 614 if (!t) 615 return -1; 616 617 mark = be32_to_cpu(t->parms.o_key); 618 619 switch (protocol) { 620 case IPPROTO_ESP: 621 esph = (struct ip_esp_hdr *)(skb->data + offset); 622 spi = esph->spi; 623 break; 624 case IPPROTO_AH: 625 ah = (struct ip_auth_hdr *)(skb->data + offset); 626 spi = ah->spi; 627 break; 628 case IPPROTO_COMP: 629 ipch = (struct ip_comp_hdr *)(skb->data + offset); 630 spi = htonl(ntohs(ipch->cpi)); 631 break; 632 default: 633 return 0; 634 } 635 636 if (type != ICMPV6_PKT_TOOBIG && 637 type != NDISC_REDIRECT) 638 return 0; 639 640 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 641 spi, protocol, AF_INET6); 642 if (!x) 643 return 0; 644 645 if (type == NDISC_REDIRECT) 646 ip6_redirect(skb, net, skb->dev->ifindex, 0, 647 sock_net_uid(net, NULL)); 648 else 649 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 650 xfrm_state_put(x); 651 652 return 0; 653} 654 655static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 656{ 657 struct net_device *dev = t->dev; 658 struct __ip6_tnl_parm *p = &t->parms; 659 struct net_device *tdev = NULL; 660 int mtu; 661 662 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 663 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 664 665 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 666 IP6_TNL_F_CAP_PER_PACKET); 667 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 668 669 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 670 dev->flags |= IFF_POINTOPOINT; 671 else 672 dev->flags &= ~IFF_POINTOPOINT; 673 674 if (keep_mtu && dev->mtu) { 675 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu); 676 return; 677 } 678 679 if (p->flags & IP6_TNL_F_CAP_XMIT) { 680 int strict = (ipv6_addr_type(&p->raddr) & 681 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 682 struct rt6_info *rt = rt6_lookup(t->net, 683 &p->raddr, &p->laddr, 684 p->link, NULL, strict); 685 686 if (rt) 687 tdev = rt->dst.dev; 688 ip6_rt_put(rt); 689 } 690 691 if (!tdev && p->link) 692 tdev = __dev_get_by_index(t->net, p->link); 693 694 if (tdev) 695 mtu = tdev->mtu - sizeof(struct ipv6hdr); 696 else 697 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 698 699 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU); 700} 701 702/** 703 * vti6_tnl_change - update the tunnel parameters 704 * @t: tunnel to be changed 705 * @p: tunnel configuration parameters 706 * @keep_mtu: MTU was set from userspace, don't re-compute it 707 * 708 * Description: 709 * vti6_tnl_change() updates the tunnel parameters 710 **/ 711static int 712vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 713 bool keep_mtu) 714{ 715 t->parms.laddr = p->laddr; 716 t->parms.raddr = p->raddr; 717 t->parms.link = p->link; 718 t->parms.i_key = p->i_key; 719 t->parms.o_key = p->o_key; 720 t->parms.proto = p->proto; 721 t->parms.fwmark = p->fwmark; 722 dst_cache_reset(&t->dst_cache); 723 vti6_link_config(t, keep_mtu); 724 return 0; 725} 726 727static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 728 bool keep_mtu) 729{ 730 struct net *net = dev_net(t->dev); 731 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 732 int err; 733 734 vti6_tnl_unlink(ip6n, t); 735 synchronize_net(); 736 err = vti6_tnl_change(t, p, keep_mtu); 737 vti6_tnl_link(ip6n, t); 738 netdev_state_change(t->dev); 739 return err; 740} 741 742static void 743vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 744{ 745 p->laddr = u->laddr; 746 p->raddr = u->raddr; 747 p->link = u->link; 748 p->i_key = u->i_key; 749 p->o_key = u->o_key; 750 p->proto = u->proto; 751 752 memcpy(p->name, u->name, sizeof(u->name)); 753} 754 755static void 756vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 757{ 758 u->laddr = p->laddr; 759 u->raddr = p->raddr; 760 u->link = p->link; 761 u->i_key = p->i_key; 762 u->o_key = p->o_key; 763 if (u->i_key) 764 u->i_flags |= GRE_KEY; 765 if (u->o_key) 766 u->o_flags |= GRE_KEY; 767 u->proto = p->proto; 768 769 memcpy(u->name, p->name, sizeof(u->name)); 770} 771 772/** 773 * vti6_ioctl - configure vti6 tunnels from userspace 774 * @dev: virtual device associated with tunnel 775 * @ifr: parameters passed from userspace 776 * @cmd: command to be performed 777 * 778 * Description: 779 * vti6_ioctl() is used for managing vti6 tunnels 780 * from userspace. 781 * 782 * The possible commands are the following: 783 * %SIOCGETTUNNEL: get tunnel parameters for device 784 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 785 * %SIOCCHGTUNNEL: change tunnel parameters to those given 786 * %SIOCDELTUNNEL: delete tunnel 787 * 788 * The fallback device "ip6_vti0", created during module 789 * initialization, can be used for creating other tunnel devices. 790 * 791 * Return: 792 * 0 on success, 793 * %-EFAULT if unable to copy data to or from userspace, 794 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 795 * %-EINVAL if passed tunnel parameters are invalid, 796 * %-EEXIST if changing a tunnel's parameters would cause a conflict 797 * %-ENODEV if attempting to change or delete a nonexisting device 798 **/ 799static int 800vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 801{ 802 int err = 0; 803 struct ip6_tnl_parm2 p; 804 struct __ip6_tnl_parm p1; 805 struct ip6_tnl *t = NULL; 806 struct net *net = dev_net(dev); 807 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 808 809 memset(&p1, 0, sizeof(p1)); 810 811 switch (cmd) { 812 case SIOCGETTUNNEL: 813 if (dev == ip6n->fb_tnl_dev) { 814 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 815 err = -EFAULT; 816 break; 817 } 818 vti6_parm_from_user(&p1, &p); 819 t = vti6_locate(net, &p1, 0); 820 } else { 821 memset(&p, 0, sizeof(p)); 822 } 823 if (!t) 824 t = netdev_priv(dev); 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 break; 829 case SIOCADDTUNNEL: 830 case SIOCCHGTUNNEL: 831 err = -EPERM; 832 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 833 break; 834 err = -EFAULT; 835 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 836 break; 837 err = -EINVAL; 838 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 839 break; 840 vti6_parm_from_user(&p1, &p); 841 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 842 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 843 if (t) { 844 if (t->dev != dev) { 845 err = -EEXIST; 846 break; 847 } 848 } else 849 t = netdev_priv(dev); 850 851 err = vti6_update(t, &p1, false); 852 } 853 if (t) { 854 err = 0; 855 vti6_parm_to_user(&p, &t->parms); 856 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 857 err = -EFAULT; 858 859 } else 860 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 861 break; 862 case SIOCDELTUNNEL: 863 err = -EPERM; 864 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 865 break; 866 867 if (dev == ip6n->fb_tnl_dev) { 868 err = -EFAULT; 869 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 870 break; 871 err = -ENOENT; 872 vti6_parm_from_user(&p1, &p); 873 t = vti6_locate(net, &p1, 0); 874 if (!t) 875 break; 876 err = -EPERM; 877 if (t->dev == ip6n->fb_tnl_dev) 878 break; 879 dev = t->dev; 880 } 881 err = 0; 882 unregister_netdevice(dev); 883 break; 884 default: 885 err = -EINVAL; 886 } 887 return err; 888} 889 890static const struct net_device_ops vti6_netdev_ops = { 891 .ndo_init = vti6_dev_init, 892 .ndo_uninit = vti6_dev_uninit, 893 .ndo_start_xmit = vti6_tnl_xmit, 894 .ndo_do_ioctl = vti6_ioctl, 895 .ndo_get_stats64 = ip_tunnel_get_stats64, 896 .ndo_get_iflink = ip6_tnl_get_iflink, 897}; 898 899/** 900 * vti6_dev_setup - setup virtual tunnel device 901 * @dev: virtual device associated with tunnel 902 * 903 * Description: 904 * Initialize function pointers and device parameters 905 **/ 906static void vti6_dev_setup(struct net_device *dev) 907{ 908 dev->netdev_ops = &vti6_netdev_ops; 909 dev->header_ops = &ip_tunnel_header_ops; 910 dev->needs_free_netdev = true; 911 dev->priv_destructor = vti6_dev_free; 912 913 dev->type = ARPHRD_TUNNEL6; 914 dev->min_mtu = IPV4_MIN_MTU; 915 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 916 dev->flags |= IFF_NOARP; 917 dev->addr_len = sizeof(struct in6_addr); 918 netif_keep_dst(dev); 919 /* This perm addr will be used as interface identifier by IPv6 */ 920 dev->addr_assign_type = NET_ADDR_RANDOM; 921 eth_random_addr(dev->perm_addr); 922} 923 924/** 925 * vti6_dev_init_gen - general initializer for all tunnel devices 926 * @dev: virtual device associated with tunnel 927 **/ 928static inline int vti6_dev_init_gen(struct net_device *dev) 929{ 930 struct ip6_tnl *t = netdev_priv(dev); 931 932 t->dev = dev; 933 t->net = dev_net(dev); 934 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 935 if (!dev->tstats) 936 return -ENOMEM; 937 dev_hold(dev); 938 return 0; 939} 940 941/** 942 * vti6_dev_init - initializer for all non fallback tunnel devices 943 * @dev: virtual device associated with tunnel 944 **/ 945static int vti6_dev_init(struct net_device *dev) 946{ 947 struct ip6_tnl *t = netdev_priv(dev); 948 int err = vti6_dev_init_gen(dev); 949 950 if (err) 951 return err; 952 vti6_link_config(t, true); 953 return 0; 954} 955 956/** 957 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 958 * @dev: fallback device 959 * 960 * Return: 0 961 **/ 962static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 963{ 964 struct ip6_tnl *t = netdev_priv(dev); 965 struct net *net = dev_net(dev); 966 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 967 968 t->parms.proto = IPPROTO_IPV6; 969 970 rcu_assign_pointer(ip6n->tnls_wc[0], t); 971 return 0; 972} 973 974static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 975 struct netlink_ext_ack *extack) 976{ 977 return 0; 978} 979 980static void vti6_netlink_parms(struct nlattr *data[], 981 struct __ip6_tnl_parm *parms) 982{ 983 memset(parms, 0, sizeof(*parms)); 984 985 if (!data) 986 return; 987 988 if (data[IFLA_VTI_LINK]) 989 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 990 991 if (data[IFLA_VTI_LOCAL]) 992 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 993 994 if (data[IFLA_VTI_REMOTE]) 995 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 996 997 if (data[IFLA_VTI_IKEY]) 998 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 999 1000 if (data[IFLA_VTI_OKEY]) 1001 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 1002 1003 if (data[IFLA_VTI_FWMARK]) 1004 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 1005} 1006 1007static int vti6_newlink(struct net *src_net, struct net_device *dev, 1008 struct nlattr *tb[], struct nlattr *data[], 1009 struct netlink_ext_ack *extack) 1010{ 1011 struct net *net = dev_net(dev); 1012 struct ip6_tnl *nt; 1013 1014 nt = netdev_priv(dev); 1015 vti6_netlink_parms(data, &nt->parms); 1016 1017 nt->parms.proto = IPPROTO_IPV6; 1018 1019 if (vti6_locate(net, &nt->parms, 0)) 1020 return -EEXIST; 1021 1022 return vti6_tnl_create2(dev); 1023} 1024 1025static void vti6_dellink(struct net_device *dev, struct list_head *head) 1026{ 1027 struct net *net = dev_net(dev); 1028 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1029 1030 if (dev != ip6n->fb_tnl_dev) 1031 unregister_netdevice_queue(dev, head); 1032} 1033 1034static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1035 struct nlattr *data[], 1036 struct netlink_ext_ack *extack) 1037{ 1038 struct ip6_tnl *t; 1039 struct __ip6_tnl_parm p; 1040 struct net *net = dev_net(dev); 1041 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1042 1043 if (dev == ip6n->fb_tnl_dev) 1044 return -EINVAL; 1045 1046 vti6_netlink_parms(data, &p); 1047 1048 t = vti6_locate(net, &p, 0); 1049 1050 if (t) { 1051 if (t->dev != dev) 1052 return -EEXIST; 1053 } else 1054 t = netdev_priv(dev); 1055 1056 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1057} 1058 1059static size_t vti6_get_size(const struct net_device *dev) 1060{ 1061 return 1062 /* IFLA_VTI_LINK */ 1063 nla_total_size(4) + 1064 /* IFLA_VTI_LOCAL */ 1065 nla_total_size(sizeof(struct in6_addr)) + 1066 /* IFLA_VTI_REMOTE */ 1067 nla_total_size(sizeof(struct in6_addr)) + 1068 /* IFLA_VTI_IKEY */ 1069 nla_total_size(4) + 1070 /* IFLA_VTI_OKEY */ 1071 nla_total_size(4) + 1072 /* IFLA_VTI_FWMARK */ 1073 nla_total_size(4) + 1074 0; 1075} 1076 1077static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1078{ 1079 struct ip6_tnl *tunnel = netdev_priv(dev); 1080 struct __ip6_tnl_parm *parm = &tunnel->parms; 1081 1082 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1083 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1084 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1085 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1086 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1087 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1088 goto nla_put_failure; 1089 return 0; 1090 1091nla_put_failure: 1092 return -EMSGSIZE; 1093} 1094 1095static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1096 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1097 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1098 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1099 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1100 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1101 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1102}; 1103 1104static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1105 .kind = "vti6", 1106 .maxtype = IFLA_VTI_MAX, 1107 .policy = vti6_policy, 1108 .priv_size = sizeof(struct ip6_tnl), 1109 .setup = vti6_dev_setup, 1110 .validate = vti6_validate, 1111 .newlink = vti6_newlink, 1112 .dellink = vti6_dellink, 1113 .changelink = vti6_changelink, 1114 .get_size = vti6_get_size, 1115 .fill_info = vti6_fill_info, 1116 .get_link_net = ip6_tnl_get_link_net, 1117}; 1118 1119static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1120 struct list_head *list) 1121{ 1122 int h; 1123 struct ip6_tnl *t; 1124 1125 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1126 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1127 while (t) { 1128 unregister_netdevice_queue(t->dev, list); 1129 t = rtnl_dereference(t->next); 1130 } 1131 } 1132 1133 t = rtnl_dereference(ip6n->tnls_wc[0]); 1134 if (t) 1135 unregister_netdevice_queue(t->dev, list); 1136} 1137 1138static int __net_init vti6_init_net(struct net *net) 1139{ 1140 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1141 struct ip6_tnl *t = NULL; 1142 int err; 1143 1144 ip6n->tnls[0] = ip6n->tnls_wc; 1145 ip6n->tnls[1] = ip6n->tnls_r_l; 1146 1147 if (!net_has_fallback_tunnels(net)) 1148 return 0; 1149 err = -ENOMEM; 1150 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1151 NET_NAME_UNKNOWN, vti6_dev_setup); 1152 1153 if (!ip6n->fb_tnl_dev) 1154 goto err_alloc_dev; 1155 dev_net_set(ip6n->fb_tnl_dev, net); 1156 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1157 1158 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1159 if (err < 0) 1160 goto err_register; 1161 1162 err = register_netdev(ip6n->fb_tnl_dev); 1163 if (err < 0) 1164 goto err_register; 1165 1166 t = netdev_priv(ip6n->fb_tnl_dev); 1167 1168 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1169 return 0; 1170 1171err_register: 1172 free_netdev(ip6n->fb_tnl_dev); 1173err_alloc_dev: 1174 return err; 1175} 1176 1177static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1178{ 1179 struct vti6_net *ip6n; 1180 struct net *net; 1181 LIST_HEAD(list); 1182 1183 rtnl_lock(); 1184 list_for_each_entry(net, net_list, exit_list) { 1185 ip6n = net_generic(net, vti6_net_id); 1186 vti6_destroy_tunnels(ip6n, &list); 1187 } 1188 unregister_netdevice_many(&list); 1189 rtnl_unlock(); 1190} 1191 1192static struct pernet_operations vti6_net_ops = { 1193 .init = vti6_init_net, 1194 .exit_batch = vti6_exit_batch_net, 1195 .id = &vti6_net_id, 1196 .size = sizeof(struct vti6_net), 1197}; 1198 1199static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1200 .handler = vti6_rcv, 1201 .input_handler = vti6_input_proto, 1202 .cb_handler = vti6_rcv_cb, 1203 .err_handler = vti6_err, 1204 .priority = 100, 1205}; 1206 1207static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1208 .handler = vti6_rcv, 1209 .input_handler = vti6_input_proto, 1210 .cb_handler = vti6_rcv_cb, 1211 .err_handler = vti6_err, 1212 .priority = 100, 1213}; 1214 1215static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1216 .handler = vti6_rcv, 1217 .input_handler = vti6_input_proto, 1218 .cb_handler = vti6_rcv_cb, 1219 .err_handler = vti6_err, 1220 .priority = 100, 1221}; 1222 1223#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1224static int vti6_rcv_tunnel(struct sk_buff *skb) 1225{ 1226 const xfrm_address_t *saddr; 1227 __be32 spi; 1228 1229 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1230 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1231 1232 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1233} 1234 1235static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1236 .handler = vti6_rcv_tunnel, 1237 .cb_handler = vti6_rcv_cb, 1238 .err_handler = vti6_err, 1239 .priority = 0, 1240}; 1241 1242static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1243 .handler = vti6_rcv_tunnel, 1244 .cb_handler = vti6_rcv_cb, 1245 .err_handler = vti6_err, 1246 .priority = 0, 1247}; 1248#endif 1249 1250/** 1251 * vti6_tunnel_init - register protocol and reserve needed resources 1252 * 1253 * Return: 0 on success 1254 **/ 1255static int __init vti6_tunnel_init(void) 1256{ 1257 const char *msg; 1258 int err; 1259 1260 msg = "tunnel device"; 1261 err = register_pernet_device(&vti6_net_ops); 1262 if (err < 0) 1263 goto pernet_dev_failed; 1264 1265 msg = "tunnel protocols"; 1266 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1267 if (err < 0) 1268 goto xfrm_proto_esp_failed; 1269 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1270 if (err < 0) 1271 goto xfrm_proto_ah_failed; 1272 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1273 if (err < 0) 1274 goto xfrm_proto_comp_failed; 1275#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1276 msg = "ipv6 tunnel"; 1277 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1278 if (err < 0) 1279 goto vti_tunnel_ipv6_failed; 1280 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1281 if (err < 0) 1282 goto vti_tunnel_ip6ip_failed; 1283#endif 1284 1285 msg = "netlink interface"; 1286 err = rtnl_link_register(&vti6_link_ops); 1287 if (err < 0) 1288 goto rtnl_link_failed; 1289 1290 return 0; 1291 1292rtnl_link_failed: 1293#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1294 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1295vti_tunnel_ip6ip_failed: 1296 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1297vti_tunnel_ipv6_failed: 1298#endif 1299 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1300xfrm_proto_comp_failed: 1301 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1302xfrm_proto_ah_failed: 1303 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1304xfrm_proto_esp_failed: 1305 unregister_pernet_device(&vti6_net_ops); 1306pernet_dev_failed: 1307 pr_err("vti6 init: failed to register %s\n", msg); 1308 return err; 1309} 1310 1311/** 1312 * vti6_tunnel_cleanup - free resources and unregister protocol 1313 **/ 1314static void __exit vti6_tunnel_cleanup(void) 1315{ 1316 rtnl_link_unregister(&vti6_link_ops); 1317#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1318 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1319 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1320#endif 1321 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1322 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1323 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1324 unregister_pernet_device(&vti6_net_ops); 1325} 1326 1327module_init(vti6_tunnel_init); 1328module_exit(vti6_tunnel_cleanup); 1329MODULE_LICENSE("GPL"); 1330MODULE_ALIAS_RTNL_LINK("vti6"); 1331MODULE_ALIAS_NETDEV("ip6_vti0"); 1332MODULE_AUTHOR("Steffen Klassert"); 1333MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1334