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 * @ip6n: the private data for ip6_vti in the netns 129 * @p: parameters containing tunnel end-points 130 * 131 * Description: 132 * vti6_tnl_bucket() returns the head of the list matching the 133 * &struct in6_addr entries laddr and raddr in @p. 134 * 135 * Return: head of IPv6 tunnel list 136 **/ 137 static struct ip6_tnl __rcu ** 138 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 139 { 140 const struct in6_addr *remote = &p->raddr; 141 const struct in6_addr *local = &p->laddr; 142 unsigned int h = 0; 143 int prio = 0; 144 145 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 146 prio = 1; 147 h = HASH(remote, local); 148 } 149 return &ip6n->tnls[prio][h]; 150 } 151 152 static void 153 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 154 { 155 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 156 157 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 158 rcu_assign_pointer(*tp, t); 159 } 160 161 static void 162 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 163 { 164 struct ip6_tnl __rcu **tp; 165 struct ip6_tnl *iter; 166 167 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 168 (iter = rtnl_dereference(*tp)) != NULL; 169 tp = &iter->next) { 170 if (t == iter) { 171 rcu_assign_pointer(*tp, t->next); 172 break; 173 } 174 } 175 } 176 177 static void vti6_dev_free(struct net_device *dev) 178 { 179 free_percpu(dev->tstats); 180 } 181 182 static int vti6_tnl_create2(struct net_device *dev) 183 { 184 struct ip6_tnl *t = netdev_priv(dev); 185 struct net *net = dev_net(dev); 186 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 187 int err; 188 189 dev->rtnl_link_ops = &vti6_link_ops; 190 err = register_netdevice(dev); 191 if (err < 0) 192 goto out; 193 194 strcpy(t->parms.name, dev->name); 195 196 dev_hold(dev); 197 vti6_tnl_link(ip6n, t); 198 199 return 0; 200 201 out: 202 return err; 203 } 204 205 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 206 { 207 struct net_device *dev; 208 struct ip6_tnl *t; 209 char name[IFNAMSIZ]; 210 int err; 211 212 if (p->name[0]) { 213 if (!dev_valid_name(p->name)) 214 goto failed; 215 strlcpy(name, p->name, IFNAMSIZ); 216 } else { 217 sprintf(name, "ip6_vti%%d"); 218 } 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_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 301 int encap_type) 302 { 303 struct ip6_tnl *t; 304 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 305 306 rcu_read_lock(); 307 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 308 if (t) { 309 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 310 rcu_read_unlock(); 311 goto discard; 312 } 313 314 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 315 rcu_read_unlock(); 316 goto discard; 317 } 318 319 ipv6h = ipv6_hdr(skb); 320 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 321 t->dev->stats.rx_dropped++; 322 rcu_read_unlock(); 323 goto discard; 324 } 325 326 rcu_read_unlock(); 327 328 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 329 XFRM_SPI_SKB_CB(skb)->family = AF_INET6; 330 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr); 331 return xfrm_input(skb, nexthdr, spi, encap_type); 332 } 333 rcu_read_unlock(); 334 return -EINVAL; 335 discard: 336 kfree_skb(skb); 337 return 0; 338 } 339 340 static int vti6_rcv(struct sk_buff *skb) 341 { 342 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff]; 343 344 return vti6_input_proto(skb, nexthdr, 0, 0); 345 } 346 347 static int vti6_rcv_cb(struct sk_buff *skb, int err) 348 { 349 unsigned short family; 350 struct net_device *dev; 351 struct xfrm_state *x; 352 const struct xfrm_mode *inner_mode; 353 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 354 u32 orig_mark = skb->mark; 355 int ret; 356 357 if (!t) 358 return 1; 359 360 dev = t->dev; 361 362 if (err) { 363 dev->stats.rx_errors++; 364 dev->stats.rx_dropped++; 365 366 return 0; 367 } 368 369 x = xfrm_input_state(skb); 370 371 inner_mode = &x->inner_mode; 372 373 if (x->sel.family == AF_UNSPEC) { 374 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 375 if (inner_mode == NULL) { 376 XFRM_INC_STATS(dev_net(skb->dev), 377 LINUX_MIB_XFRMINSTATEMODEERROR); 378 return -EINVAL; 379 } 380 } 381 382 family = inner_mode->family; 383 384 skb->mark = be32_to_cpu(t->parms.i_key); 385 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 386 skb->mark = orig_mark; 387 388 if (!ret) 389 return -EPERM; 390 391 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 392 skb->dev = dev; 393 dev_sw_netstats_rx_add(dev, skb->len); 394 395 return 0; 396 } 397 398 /** 399 * vti6_addr_conflict - compare packet addresses to tunnel's own 400 * @t: the outgoing tunnel device 401 * @hdr: IPv6 header from the incoming packet 402 * 403 * Description: 404 * Avoid trivial tunneling loop by checking that tunnel exit-point 405 * doesn't match source of incoming packet. 406 * 407 * Return: 408 * 1 if conflict, 409 * 0 else 410 **/ 411 static inline bool 412 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 413 { 414 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 415 } 416 417 static bool vti6_state_check(const struct xfrm_state *x, 418 const struct in6_addr *dst, 419 const struct in6_addr *src) 420 { 421 xfrm_address_t *daddr = (xfrm_address_t *)dst; 422 xfrm_address_t *saddr = (xfrm_address_t *)src; 423 424 /* if there is no transform then this tunnel is not functional. 425 * Or if the xfrm is not mode tunnel. 426 */ 427 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 428 x->props.family != AF_INET6) 429 return false; 430 431 if (ipv6_addr_any(dst)) 432 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 433 434 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 435 return false; 436 437 return true; 438 } 439 440 /** 441 * vti6_xmit - send a packet 442 * @skb: the outgoing socket buffer 443 * @dev: the outgoing tunnel device 444 * @fl: the flow informations for the xfrm_lookup 445 **/ 446 static int 447 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 448 { 449 struct ip6_tnl *t = netdev_priv(dev); 450 struct net_device_stats *stats = &t->dev->stats; 451 struct dst_entry *dst = skb_dst(skb); 452 struct net_device *tdev; 453 struct xfrm_state *x; 454 int pkt_len = skb->len; 455 int err = -1; 456 int mtu; 457 458 if (!dst) { 459 switch (skb->protocol) { 460 case htons(ETH_P_IP): { 461 struct rtable *rt; 462 463 fl->u.ip4.flowi4_oif = dev->ifindex; 464 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 465 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 466 if (IS_ERR(rt)) 467 goto tx_err_link_failure; 468 dst = &rt->dst; 469 skb_dst_set(skb, dst); 470 break; 471 } 472 case htons(ETH_P_IPV6): 473 fl->u.ip6.flowi6_oif = dev->ifindex; 474 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 475 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 476 if (dst->error) { 477 dst_release(dst); 478 dst = NULL; 479 goto tx_err_link_failure; 480 } 481 skb_dst_set(skb, dst); 482 break; 483 default: 484 goto tx_err_link_failure; 485 } 486 } 487 488 dst_hold(dst); 489 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 490 if (IS_ERR(dst)) { 491 err = PTR_ERR(dst); 492 dst = NULL; 493 goto tx_err_link_failure; 494 } 495 496 if (dst->flags & DST_XFRM_QUEUE) 497 goto queued; 498 499 x = dst->xfrm; 500 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 501 goto tx_err_link_failure; 502 503 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 504 (const struct in6_addr *)&x->id.daddr)) 505 goto tx_err_link_failure; 506 507 tdev = dst->dev; 508 509 if (tdev == dev) { 510 stats->collisions++; 511 net_warn_ratelimited("%s: Local routing loop detected!\n", 512 t->parms.name); 513 goto tx_err_dst_release; 514 } 515 516 mtu = dst_mtu(dst); 517 if (skb->len > mtu) { 518 skb_dst_update_pmtu_no_confirm(skb, mtu); 519 520 if (skb->protocol == htons(ETH_P_IPV6)) { 521 if (mtu < IPV6_MIN_MTU) 522 mtu = IPV6_MIN_MTU; 523 524 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 525 } else { 526 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 527 htonl(mtu)); 528 } 529 530 err = -EMSGSIZE; 531 goto tx_err_dst_release; 532 } 533 534 queued: 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; 545 tx_err_link_failure: 546 stats->tx_carrier_errors++; 547 dst_link_failure(skb); 548 tx_err_dst_release: 549 dst_release(dst); 550 return err; 551 } 552 553 static netdev_tx_t 554 vti6_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 xfrm_decode_session(skb, &fl, AF_INET6); 573 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 574 break; 575 case htons(ETH_P_IP): 576 xfrm_decode_session(skb, &fl, AF_INET); 577 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 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; 591 592 tx_err: 593 stats->tx_errors++; 594 stats->tx_dropped++; 595 kfree_skb(skb); 596 return NETDEV_TX_OK; 597 } 598 599 static 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 } 654 655 static 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 **/ 711 static int 712 vti6_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 } 726 727 static 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 } 741 742 static void 743 vti6_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 } 754 755 static void 756 vti6_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 **/ 799 static int 800 vti6_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 switch (cmd) { 810 case SIOCGETTUNNEL: 811 if (dev == ip6n->fb_tnl_dev) { 812 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 813 err = -EFAULT; 814 break; 815 } 816 vti6_parm_from_user(&p1, &p); 817 t = vti6_locate(net, &p1, 0); 818 } else { 819 memset(&p, 0, sizeof(p)); 820 } 821 if (!t) 822 t = netdev_priv(dev); 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 break; 827 case SIOCADDTUNNEL: 828 case SIOCCHGTUNNEL: 829 err = -EPERM; 830 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 831 break; 832 err = -EFAULT; 833 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 834 break; 835 err = -EINVAL; 836 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 837 break; 838 vti6_parm_from_user(&p1, &p); 839 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 840 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 841 if (t) { 842 if (t->dev != dev) { 843 err = -EEXIST; 844 break; 845 } 846 } else 847 t = netdev_priv(dev); 848 849 err = vti6_update(t, &p1, false); 850 } 851 if (t) { 852 err = 0; 853 vti6_parm_to_user(&p, &t->parms); 854 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 855 err = -EFAULT; 856 857 } else 858 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 859 break; 860 case SIOCDELTUNNEL: 861 err = -EPERM; 862 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 863 break; 864 865 if (dev == ip6n->fb_tnl_dev) { 866 err = -EFAULT; 867 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 868 break; 869 err = -ENOENT; 870 vti6_parm_from_user(&p1, &p); 871 t = vti6_locate(net, &p1, 0); 872 if (!t) 873 break; 874 err = -EPERM; 875 if (t->dev == ip6n->fb_tnl_dev) 876 break; 877 dev = t->dev; 878 } 879 err = 0; 880 unregister_netdevice(dev); 881 break; 882 default: 883 err = -EINVAL; 884 } 885 return err; 886 } 887 888 static const struct net_device_ops vti6_netdev_ops = { 889 .ndo_init = vti6_dev_init, 890 .ndo_uninit = vti6_dev_uninit, 891 .ndo_start_xmit = vti6_tnl_xmit, 892 .ndo_do_ioctl = vti6_ioctl, 893 .ndo_get_stats64 = dev_get_tstats64, 894 .ndo_get_iflink = ip6_tnl_get_iflink, 895 }; 896 897 /** 898 * vti6_dev_setup - setup virtual tunnel device 899 * @dev: virtual device associated with tunnel 900 * 901 * Description: 902 * Initialize function pointers and device parameters 903 **/ 904 static void vti6_dev_setup(struct net_device *dev) 905 { 906 dev->netdev_ops = &vti6_netdev_ops; 907 dev->header_ops = &ip_tunnel_header_ops; 908 dev->needs_free_netdev = true; 909 dev->priv_destructor = vti6_dev_free; 910 911 dev->type = ARPHRD_TUNNEL6; 912 dev->min_mtu = IPV4_MIN_MTU; 913 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 914 dev->flags |= IFF_NOARP; 915 dev->addr_len = sizeof(struct in6_addr); 916 netif_keep_dst(dev); 917 /* This perm addr will be used as interface identifier by IPv6 */ 918 dev->addr_assign_type = NET_ADDR_RANDOM; 919 eth_random_addr(dev->perm_addr); 920 } 921 922 /** 923 * vti6_dev_init_gen - general initializer for all tunnel devices 924 * @dev: virtual device associated with tunnel 925 **/ 926 static inline int vti6_dev_init_gen(struct net_device *dev) 927 { 928 struct ip6_tnl *t = netdev_priv(dev); 929 930 t->dev = dev; 931 t->net = dev_net(dev); 932 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 933 if (!dev->tstats) 934 return -ENOMEM; 935 return 0; 936 } 937 938 /** 939 * vti6_dev_init - initializer for all non fallback tunnel devices 940 * @dev: virtual device associated with tunnel 941 **/ 942 static int vti6_dev_init(struct net_device *dev) 943 { 944 struct ip6_tnl *t = netdev_priv(dev); 945 int err = vti6_dev_init_gen(dev); 946 947 if (err) 948 return err; 949 vti6_link_config(t, true); 950 return 0; 951 } 952 953 /** 954 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 955 * @dev: fallback device 956 * 957 * Return: 0 958 **/ 959 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 960 { 961 struct ip6_tnl *t = netdev_priv(dev); 962 struct net *net = dev_net(dev); 963 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 964 965 t->parms.proto = IPPROTO_IPV6; 966 dev_hold(dev); 967 968 rcu_assign_pointer(ip6n->tnls_wc[0], t); 969 return 0; 970 } 971 972 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 973 struct netlink_ext_ack *extack) 974 { 975 return 0; 976 } 977 978 static void vti6_netlink_parms(struct nlattr *data[], 979 struct __ip6_tnl_parm *parms) 980 { 981 memset(parms, 0, sizeof(*parms)); 982 983 if (!data) 984 return; 985 986 if (data[IFLA_VTI_LINK]) 987 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 988 989 if (data[IFLA_VTI_LOCAL]) 990 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 991 992 if (data[IFLA_VTI_REMOTE]) 993 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 994 995 if (data[IFLA_VTI_IKEY]) 996 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 997 998 if (data[IFLA_VTI_OKEY]) 999 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 1000 1001 if (data[IFLA_VTI_FWMARK]) 1002 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 1003 } 1004 1005 static int vti6_newlink(struct net *src_net, struct net_device *dev, 1006 struct nlattr *tb[], struct nlattr *data[], 1007 struct netlink_ext_ack *extack) 1008 { 1009 struct net *net = dev_net(dev); 1010 struct ip6_tnl *nt; 1011 1012 nt = netdev_priv(dev); 1013 vti6_netlink_parms(data, &nt->parms); 1014 1015 nt->parms.proto = IPPROTO_IPV6; 1016 1017 if (vti6_locate(net, &nt->parms, 0)) 1018 return -EEXIST; 1019 1020 return vti6_tnl_create2(dev); 1021 } 1022 1023 static void vti6_dellink(struct net_device *dev, struct list_head *head) 1024 { 1025 struct net *net = dev_net(dev); 1026 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1027 1028 if (dev != ip6n->fb_tnl_dev) 1029 unregister_netdevice_queue(dev, head); 1030 } 1031 1032 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1033 struct nlattr *data[], 1034 struct netlink_ext_ack *extack) 1035 { 1036 struct ip6_tnl *t; 1037 struct __ip6_tnl_parm p; 1038 struct net *net = dev_net(dev); 1039 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1040 1041 if (dev == ip6n->fb_tnl_dev) 1042 return -EINVAL; 1043 1044 vti6_netlink_parms(data, &p); 1045 1046 t = vti6_locate(net, &p, 0); 1047 1048 if (t) { 1049 if (t->dev != dev) 1050 return -EEXIST; 1051 } else 1052 t = netdev_priv(dev); 1053 1054 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1055 } 1056 1057 static size_t vti6_get_size(const struct net_device *dev) 1058 { 1059 return 1060 /* IFLA_VTI_LINK */ 1061 nla_total_size(4) + 1062 /* IFLA_VTI_LOCAL */ 1063 nla_total_size(sizeof(struct in6_addr)) + 1064 /* IFLA_VTI_REMOTE */ 1065 nla_total_size(sizeof(struct in6_addr)) + 1066 /* IFLA_VTI_IKEY */ 1067 nla_total_size(4) + 1068 /* IFLA_VTI_OKEY */ 1069 nla_total_size(4) + 1070 /* IFLA_VTI_FWMARK */ 1071 nla_total_size(4) + 1072 0; 1073 } 1074 1075 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1076 { 1077 struct ip6_tnl *tunnel = netdev_priv(dev); 1078 struct __ip6_tnl_parm *parm = &tunnel->parms; 1079 1080 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1081 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1082 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1083 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1084 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1085 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1086 goto nla_put_failure; 1087 return 0; 1088 1089 nla_put_failure: 1090 return -EMSGSIZE; 1091 } 1092 1093 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1094 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1095 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1096 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1097 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1098 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1099 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1100 }; 1101 1102 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1103 .kind = "vti6", 1104 .maxtype = IFLA_VTI_MAX, 1105 .policy = vti6_policy, 1106 .priv_size = sizeof(struct ip6_tnl), 1107 .setup = vti6_dev_setup, 1108 .validate = vti6_validate, 1109 .newlink = vti6_newlink, 1110 .dellink = vti6_dellink, 1111 .changelink = vti6_changelink, 1112 .get_size = vti6_get_size, 1113 .fill_info = vti6_fill_info, 1114 .get_link_net = ip6_tnl_get_link_net, 1115 }; 1116 1117 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1118 struct list_head *list) 1119 { 1120 int h; 1121 struct ip6_tnl *t; 1122 1123 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1124 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1125 while (t) { 1126 unregister_netdevice_queue(t->dev, list); 1127 t = rtnl_dereference(t->next); 1128 } 1129 } 1130 1131 t = rtnl_dereference(ip6n->tnls_wc[0]); 1132 if (t) 1133 unregister_netdevice_queue(t->dev, list); 1134 } 1135 1136 static int __net_init vti6_init_net(struct net *net) 1137 { 1138 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1139 struct ip6_tnl *t = NULL; 1140 int err; 1141 1142 ip6n->tnls[0] = ip6n->tnls_wc; 1143 ip6n->tnls[1] = ip6n->tnls_r_l; 1144 1145 if (!net_has_fallback_tunnels(net)) 1146 return 0; 1147 err = -ENOMEM; 1148 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1149 NET_NAME_UNKNOWN, vti6_dev_setup); 1150 1151 if (!ip6n->fb_tnl_dev) 1152 goto err_alloc_dev; 1153 dev_net_set(ip6n->fb_tnl_dev, net); 1154 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1155 1156 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1157 if (err < 0) 1158 goto err_register; 1159 1160 err = register_netdev(ip6n->fb_tnl_dev); 1161 if (err < 0) 1162 goto err_register; 1163 1164 t = netdev_priv(ip6n->fb_tnl_dev); 1165 1166 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1167 return 0; 1168 1169 err_register: 1170 free_netdev(ip6n->fb_tnl_dev); 1171 err_alloc_dev: 1172 return err; 1173 } 1174 1175 static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1176 { 1177 struct vti6_net *ip6n; 1178 struct net *net; 1179 LIST_HEAD(list); 1180 1181 rtnl_lock(); 1182 list_for_each_entry(net, net_list, exit_list) { 1183 ip6n = net_generic(net, vti6_net_id); 1184 vti6_destroy_tunnels(ip6n, &list); 1185 } 1186 unregister_netdevice_many(&list); 1187 rtnl_unlock(); 1188 } 1189 1190 static struct pernet_operations vti6_net_ops = { 1191 .init = vti6_init_net, 1192 .exit_batch = vti6_exit_batch_net, 1193 .id = &vti6_net_id, 1194 .size = sizeof(struct vti6_net), 1195 }; 1196 1197 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1198 .handler = vti6_rcv, 1199 .input_handler = vti6_input_proto, 1200 .cb_handler = vti6_rcv_cb, 1201 .err_handler = vti6_err, 1202 .priority = 100, 1203 }; 1204 1205 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1206 .handler = vti6_rcv, 1207 .input_handler = vti6_input_proto, 1208 .cb_handler = vti6_rcv_cb, 1209 .err_handler = vti6_err, 1210 .priority = 100, 1211 }; 1212 1213 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1214 .handler = vti6_rcv, 1215 .input_handler = vti6_input_proto, 1216 .cb_handler = vti6_rcv_cb, 1217 .err_handler = vti6_err, 1218 .priority = 100, 1219 }; 1220 1221 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1222 static int vti6_rcv_tunnel(struct sk_buff *skb) 1223 { 1224 const xfrm_address_t *saddr; 1225 __be32 spi; 1226 1227 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1228 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1229 1230 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1231 } 1232 1233 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1234 .handler = vti6_rcv_tunnel, 1235 .cb_handler = vti6_rcv_cb, 1236 .err_handler = vti6_err, 1237 .priority = 0, 1238 }; 1239 1240 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1241 .handler = vti6_rcv_tunnel, 1242 .cb_handler = vti6_rcv_cb, 1243 .err_handler = vti6_err, 1244 .priority = 0, 1245 }; 1246 #endif 1247 1248 /** 1249 * vti6_tunnel_init - register protocol and reserve needed resources 1250 * 1251 * Return: 0 on success 1252 **/ 1253 static int __init vti6_tunnel_init(void) 1254 { 1255 const char *msg; 1256 int err; 1257 1258 msg = "tunnel device"; 1259 err = register_pernet_device(&vti6_net_ops); 1260 if (err < 0) 1261 goto pernet_dev_failed; 1262 1263 msg = "tunnel protocols"; 1264 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1265 if (err < 0) 1266 goto xfrm_proto_esp_failed; 1267 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1268 if (err < 0) 1269 goto xfrm_proto_ah_failed; 1270 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1271 if (err < 0) 1272 goto xfrm_proto_comp_failed; 1273 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1274 msg = "ipv6 tunnel"; 1275 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1276 if (err < 0) 1277 goto vti_tunnel_ipv6_failed; 1278 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1279 if (err < 0) 1280 goto vti_tunnel_ip6ip_failed; 1281 #endif 1282 1283 msg = "netlink interface"; 1284 err = rtnl_link_register(&vti6_link_ops); 1285 if (err < 0) 1286 goto rtnl_link_failed; 1287 1288 return 0; 1289 1290 rtnl_link_failed: 1291 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1292 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1293 vti_tunnel_ip6ip_failed: 1294 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1295 vti_tunnel_ipv6_failed: 1296 #endif 1297 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1298 xfrm_proto_comp_failed: 1299 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1300 xfrm_proto_ah_failed: 1301 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1302 xfrm_proto_esp_failed: 1303 unregister_pernet_device(&vti6_net_ops); 1304 pernet_dev_failed: 1305 pr_err("vti6 init: failed to register %s\n", msg); 1306 return err; 1307 } 1308 1309 /** 1310 * vti6_tunnel_cleanup - free resources and unregister protocol 1311 **/ 1312 static void __exit vti6_tunnel_cleanup(void) 1313 { 1314 rtnl_link_unregister(&vti6_link_ops); 1315 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1316 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1317 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1318 #endif 1319 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1320 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1321 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1322 unregister_pernet_device(&vti6_net_ops); 1323 } 1324 1325 module_init(vti6_tunnel_init); 1326 module_exit(vti6_tunnel_cleanup); 1327 MODULE_LICENSE("GPL"); 1328 MODULE_ALIAS_RTNL_LINK("vti6"); 1329 MODULE_ALIAS_NETDEV("ip6_vti0"); 1330 MODULE_AUTHOR("Steffen Klassert"); 1331 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1332