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 int vti6_tnl_create2(struct net_device *dev) 178 { 179 struct ip6_tnl *t = netdev_priv(dev); 180 struct net *net = dev_net(dev); 181 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 182 int err; 183 184 dev->rtnl_link_ops = &vti6_link_ops; 185 err = register_netdevice(dev); 186 if (err < 0) 187 goto out; 188 189 strcpy(t->parms.name, dev->name); 190 191 vti6_tnl_link(ip6n, t); 192 193 return 0; 194 195 out: 196 return err; 197 } 198 199 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 200 { 201 struct net_device *dev; 202 struct ip6_tnl *t; 203 char name[IFNAMSIZ]; 204 int err; 205 206 if (p->name[0]) { 207 if (!dev_valid_name(p->name)) 208 goto failed; 209 strscpy(name, p->name, IFNAMSIZ); 210 } else { 211 sprintf(name, "ip6_vti%%d"); 212 } 213 214 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 215 if (!dev) 216 goto failed; 217 218 dev_net_set(dev, net); 219 220 t = netdev_priv(dev); 221 t->parms = *p; 222 t->net = dev_net(dev); 223 224 err = vti6_tnl_create2(dev); 225 if (err < 0) 226 goto failed_free; 227 228 return t; 229 230 failed_free: 231 free_netdev(dev); 232 failed: 233 return NULL; 234 } 235 236 /** 237 * vti6_locate - find or create tunnel matching given parameters 238 * @net: network namespace 239 * @p: tunnel parameters 240 * @create: != 0 if allowed to create new tunnel if no match found 241 * 242 * Description: 243 * vti6_locate() first tries to locate an existing tunnel 244 * based on @parms. If this is unsuccessful, but @create is set a new 245 * tunnel device is created and registered for use. 246 * 247 * Return: 248 * matching tunnel or NULL 249 **/ 250 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 251 int create) 252 { 253 const struct in6_addr *remote = &p->raddr; 254 const struct in6_addr *local = &p->laddr; 255 struct ip6_tnl __rcu **tp; 256 struct ip6_tnl *t; 257 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 258 259 for (tp = vti6_tnl_bucket(ip6n, p); 260 (t = rtnl_dereference(*tp)) != NULL; 261 tp = &t->next) { 262 if (ipv6_addr_equal(local, &t->parms.laddr) && 263 ipv6_addr_equal(remote, &t->parms.raddr)) { 264 if (create) 265 return NULL; 266 267 return t; 268 } 269 } 270 if (!create) 271 return NULL; 272 return vti6_tnl_create(net, p); 273 } 274 275 /** 276 * vti6_dev_uninit - tunnel device uninitializer 277 * @dev: the device to be destroyed 278 * 279 * Description: 280 * vti6_dev_uninit() removes tunnel from its list 281 **/ 282 static void vti6_dev_uninit(struct net_device *dev) 283 { 284 struct ip6_tnl *t = netdev_priv(dev); 285 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 286 287 if (dev == ip6n->fb_tnl_dev) 288 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 289 else 290 vti6_tnl_unlink(ip6n, t); 291 netdev_put(dev, &t->dev_tracker); 292 } 293 294 static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 295 int encap_type) 296 { 297 struct ip6_tnl *t; 298 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 299 300 rcu_read_lock(); 301 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 302 if (t) { 303 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 304 rcu_read_unlock(); 305 goto discard; 306 } 307 308 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 309 rcu_read_unlock(); 310 goto discard; 311 } 312 313 ipv6h = ipv6_hdr(skb); 314 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 315 DEV_STATS_INC(t->dev, rx_dropped); 316 rcu_read_unlock(); 317 goto discard; 318 } 319 320 rcu_read_unlock(); 321 322 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 323 XFRM_SPI_SKB_CB(skb)->family = AF_INET6; 324 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr); 325 return xfrm_input(skb, nexthdr, spi, encap_type); 326 } 327 rcu_read_unlock(); 328 return -EINVAL; 329 discard: 330 kfree_skb(skb); 331 return 0; 332 } 333 334 static int vti6_rcv(struct sk_buff *skb) 335 { 336 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff]; 337 338 return vti6_input_proto(skb, nexthdr, 0, 0); 339 } 340 341 static int vti6_rcv_cb(struct sk_buff *skb, int err) 342 { 343 unsigned short family; 344 struct net_device *dev; 345 struct xfrm_state *x; 346 const struct xfrm_mode *inner_mode; 347 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 348 u32 orig_mark = skb->mark; 349 int ret; 350 351 if (!t) 352 return 1; 353 354 dev = t->dev; 355 356 if (err) { 357 DEV_STATS_INC(dev, rx_errors); 358 DEV_STATS_INC(dev, rx_dropped); 359 360 return 0; 361 } 362 363 x = xfrm_input_state(skb); 364 365 inner_mode = &x->inner_mode; 366 367 if (x->sel.family == AF_UNSPEC) { 368 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 369 if (inner_mode == NULL) { 370 XFRM_INC_STATS(dev_net(skb->dev), 371 LINUX_MIB_XFRMINSTATEMODEERROR); 372 return -EINVAL; 373 } 374 } 375 376 family = inner_mode->family; 377 378 skb->mark = be32_to_cpu(t->parms.i_key); 379 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 380 skb->mark = orig_mark; 381 382 if (!ret) 383 return -EPERM; 384 385 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 386 skb->dev = dev; 387 dev_sw_netstats_rx_add(dev, skb->len); 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 dst_entry *dst = skb_dst(skb); 445 struct net_device *tdev; 446 struct xfrm_state *x; 447 int pkt_len = skb->len; 448 int err = -1; 449 int mtu; 450 451 if (!dst) { 452 switch (skb->protocol) { 453 case htons(ETH_P_IP): { 454 struct rtable *rt; 455 456 fl->u.ip4.flowi4_oif = dev->ifindex; 457 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 458 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 459 if (IS_ERR(rt)) 460 goto tx_err_link_failure; 461 dst = &rt->dst; 462 skb_dst_set(skb, dst); 463 break; 464 } 465 case htons(ETH_P_IPV6): 466 fl->u.ip6.flowi6_oif = dev->ifindex; 467 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 468 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 469 if (dst->error) { 470 dst_release(dst); 471 dst = NULL; 472 goto tx_err_link_failure; 473 } 474 skb_dst_set(skb, dst); 475 break; 476 default: 477 goto tx_err_link_failure; 478 } 479 } 480 481 dst_hold(dst); 482 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 483 if (IS_ERR(dst)) { 484 err = PTR_ERR(dst); 485 dst = NULL; 486 goto tx_err_link_failure; 487 } 488 489 if (dst->flags & DST_XFRM_QUEUE) 490 goto xmit; 491 492 x = dst->xfrm; 493 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 494 goto tx_err_link_failure; 495 496 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 497 (const struct in6_addr *)&x->id.daddr)) 498 goto tx_err_link_failure; 499 500 tdev = dst->dev; 501 502 if (tdev == dev) { 503 DEV_STATS_INC(dev, collisions); 504 net_warn_ratelimited("%s: Local routing loop detected!\n", 505 t->parms.name); 506 goto tx_err_dst_release; 507 } 508 509 mtu = dst_mtu(dst); 510 if (skb->len > mtu) { 511 skb_dst_update_pmtu_no_confirm(skb, mtu); 512 513 if (skb->protocol == htons(ETH_P_IPV6)) { 514 if (mtu < IPV6_MIN_MTU) 515 mtu = IPV6_MIN_MTU; 516 517 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 518 } else { 519 if (!(ip_hdr(skb)->frag_off & htons(IP_DF))) 520 goto xmit; 521 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 522 htonl(mtu)); 523 } 524 525 err = -EMSGSIZE; 526 goto tx_err_dst_release; 527 } 528 529 xmit: 530 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 531 skb_dst_set(skb, dst); 532 skb->dev = skb_dst(skb)->dev; 533 534 err = dst_output(t->net, skb->sk, skb); 535 if (net_xmit_eval(err) == 0) 536 err = pkt_len; 537 iptunnel_xmit_stats(dev, err); 538 539 return 0; 540 tx_err_link_failure: 541 DEV_STATS_INC(dev, tx_carrier_errors); 542 dst_link_failure(skb); 543 tx_err_dst_release: 544 dst_release(dst); 545 return err; 546 } 547 548 static netdev_tx_t 549 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 550 { 551 struct ip6_tnl *t = netdev_priv(dev); 552 struct flowi fl; 553 int ret; 554 555 if (!pskb_inet_may_pull(skb)) 556 goto tx_err; 557 558 memset(&fl, 0, sizeof(fl)); 559 560 switch (skb->protocol) { 561 case htons(ETH_P_IPV6): 562 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 563 vti6_addr_conflict(t, ipv6_hdr(skb))) 564 goto tx_err; 565 566 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 567 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6); 568 break; 569 case htons(ETH_P_IP): 570 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 571 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET); 572 break; 573 default: 574 goto tx_err; 575 } 576 577 /* override mark with tunnel output key */ 578 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 579 580 ret = vti6_xmit(skb, dev, &fl); 581 if (ret < 0) 582 goto tx_err; 583 584 return NETDEV_TX_OK; 585 586 tx_err: 587 DEV_STATS_INC(dev, tx_errors); 588 DEV_STATS_INC(dev, tx_dropped); 589 kfree_skb(skb); 590 return NETDEV_TX_OK; 591 } 592 593 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 594 u8 type, u8 code, int offset, __be32 info) 595 { 596 __be32 spi; 597 __u32 mark; 598 struct xfrm_state *x; 599 struct ip6_tnl *t; 600 struct ip_esp_hdr *esph; 601 struct ip_auth_hdr *ah; 602 struct ip_comp_hdr *ipch; 603 struct net *net = dev_net(skb->dev); 604 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 605 int protocol = iph->nexthdr; 606 607 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 608 if (!t) 609 return -1; 610 611 mark = be32_to_cpu(t->parms.o_key); 612 613 switch (protocol) { 614 case IPPROTO_ESP: 615 esph = (struct ip_esp_hdr *)(skb->data + offset); 616 spi = esph->spi; 617 break; 618 case IPPROTO_AH: 619 ah = (struct ip_auth_hdr *)(skb->data + offset); 620 spi = ah->spi; 621 break; 622 case IPPROTO_COMP: 623 ipch = (struct ip_comp_hdr *)(skb->data + offset); 624 spi = htonl(ntohs(ipch->cpi)); 625 break; 626 default: 627 return 0; 628 } 629 630 if (type != ICMPV6_PKT_TOOBIG && 631 type != NDISC_REDIRECT) 632 return 0; 633 634 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 635 spi, protocol, AF_INET6); 636 if (!x) 637 return 0; 638 639 if (type == NDISC_REDIRECT) 640 ip6_redirect(skb, net, skb->dev->ifindex, 0, 641 sock_net_uid(net, NULL)); 642 else 643 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 644 xfrm_state_put(x); 645 646 return 0; 647 } 648 649 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 650 { 651 struct net_device *dev = t->dev; 652 struct __ip6_tnl_parm *p = &t->parms; 653 struct net_device *tdev = NULL; 654 int mtu; 655 656 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 657 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 658 659 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 660 IP6_TNL_F_CAP_PER_PACKET); 661 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 662 663 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 664 dev->flags |= IFF_POINTOPOINT; 665 else 666 dev->flags &= ~IFF_POINTOPOINT; 667 668 if (keep_mtu && dev->mtu) { 669 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu); 670 return; 671 } 672 673 if (p->flags & IP6_TNL_F_CAP_XMIT) { 674 int strict = (ipv6_addr_type(&p->raddr) & 675 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 676 struct rt6_info *rt = rt6_lookup(t->net, 677 &p->raddr, &p->laddr, 678 p->link, NULL, strict); 679 680 if (rt) 681 tdev = rt->dst.dev; 682 ip6_rt_put(rt); 683 } 684 685 if (!tdev && p->link) 686 tdev = __dev_get_by_index(t->net, p->link); 687 688 if (tdev) 689 mtu = tdev->mtu - sizeof(struct ipv6hdr); 690 else 691 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 692 693 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU); 694 } 695 696 /** 697 * vti6_tnl_change - update the tunnel parameters 698 * @t: tunnel to be changed 699 * @p: tunnel configuration parameters 700 * @keep_mtu: MTU was set from userspace, don't re-compute it 701 * 702 * Description: 703 * vti6_tnl_change() updates the tunnel parameters 704 **/ 705 static int 706 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 707 bool keep_mtu) 708 { 709 t->parms.laddr = p->laddr; 710 t->parms.raddr = p->raddr; 711 t->parms.link = p->link; 712 t->parms.i_key = p->i_key; 713 t->parms.o_key = p->o_key; 714 t->parms.proto = p->proto; 715 t->parms.fwmark = p->fwmark; 716 dst_cache_reset(&t->dst_cache); 717 vti6_link_config(t, keep_mtu); 718 return 0; 719 } 720 721 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 722 bool keep_mtu) 723 { 724 struct net *net = dev_net(t->dev); 725 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 726 int err; 727 728 vti6_tnl_unlink(ip6n, t); 729 synchronize_net(); 730 err = vti6_tnl_change(t, p, keep_mtu); 731 vti6_tnl_link(ip6n, t); 732 netdev_state_change(t->dev); 733 return err; 734 } 735 736 static void 737 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 738 { 739 p->laddr = u->laddr; 740 p->raddr = u->raddr; 741 p->link = u->link; 742 p->i_key = u->i_key; 743 p->o_key = u->o_key; 744 p->proto = u->proto; 745 746 memcpy(p->name, u->name, sizeof(u->name)); 747 } 748 749 static void 750 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 751 { 752 u->laddr = p->laddr; 753 u->raddr = p->raddr; 754 u->link = p->link; 755 u->i_key = p->i_key; 756 u->o_key = p->o_key; 757 if (u->i_key) 758 u->i_flags |= GRE_KEY; 759 if (u->o_key) 760 u->o_flags |= GRE_KEY; 761 u->proto = p->proto; 762 763 memcpy(u->name, p->name, sizeof(u->name)); 764 } 765 766 /** 767 * vti6_siocdevprivate - configure vti6 tunnels from userspace 768 * @dev: virtual device associated with tunnel 769 * @ifr: unused 770 * @data: parameters passed from userspace 771 * @cmd: command to be performed 772 * 773 * Description: 774 * vti6_siocdevprivate() is used for managing vti6 tunnels 775 * from userspace. 776 * 777 * The possible commands are the following: 778 * %SIOCGETTUNNEL: get tunnel parameters for device 779 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 780 * %SIOCCHGTUNNEL: change tunnel parameters to those given 781 * %SIOCDELTUNNEL: delete tunnel 782 * 783 * The fallback device "ip6_vti0", created during module 784 * initialization, can be used for creating other tunnel devices. 785 * 786 * Return: 787 * 0 on success, 788 * %-EFAULT if unable to copy data to or from userspace, 789 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 790 * %-EINVAL if passed tunnel parameters are invalid, 791 * %-EEXIST if changing a tunnel's parameters would cause a conflict 792 * %-ENODEV if attempting to change or delete a nonexisting device 793 **/ 794 static int 795 vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd) 796 { 797 int err = 0; 798 struct ip6_tnl_parm2 p; 799 struct __ip6_tnl_parm p1; 800 struct ip6_tnl *t = NULL; 801 struct net *net = dev_net(dev); 802 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 803 804 memset(&p1, 0, sizeof(p1)); 805 806 switch (cmd) { 807 case SIOCGETTUNNEL: 808 if (dev == ip6n->fb_tnl_dev) { 809 if (copy_from_user(&p, data, sizeof(p))) { 810 err = -EFAULT; 811 break; 812 } 813 vti6_parm_from_user(&p1, &p); 814 t = vti6_locate(net, &p1, 0); 815 } else { 816 memset(&p, 0, sizeof(p)); 817 } 818 if (!t) 819 t = netdev_priv(dev); 820 vti6_parm_to_user(&p, &t->parms); 821 if (copy_to_user(data, &p, sizeof(p))) 822 err = -EFAULT; 823 break; 824 case SIOCADDTUNNEL: 825 case SIOCCHGTUNNEL: 826 err = -EPERM; 827 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 828 break; 829 err = -EFAULT; 830 if (copy_from_user(&p, data, sizeof(p))) 831 break; 832 err = -EINVAL; 833 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 834 break; 835 vti6_parm_from_user(&p1, &p); 836 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 837 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 838 if (t) { 839 if (t->dev != dev) { 840 err = -EEXIST; 841 break; 842 } 843 } else 844 t = netdev_priv(dev); 845 846 err = vti6_update(t, &p1, false); 847 } 848 if (t) { 849 err = 0; 850 vti6_parm_to_user(&p, &t->parms); 851 if (copy_to_user(data, &p, sizeof(p))) 852 err = -EFAULT; 853 854 } else 855 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 856 break; 857 case SIOCDELTUNNEL: 858 err = -EPERM; 859 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 860 break; 861 862 if (dev == ip6n->fb_tnl_dev) { 863 err = -EFAULT; 864 if (copy_from_user(&p, data, sizeof(p))) 865 break; 866 err = -ENOENT; 867 vti6_parm_from_user(&p1, &p); 868 t = vti6_locate(net, &p1, 0); 869 if (!t) 870 break; 871 err = -EPERM; 872 if (t->dev == ip6n->fb_tnl_dev) 873 break; 874 dev = t->dev; 875 } 876 err = 0; 877 unregister_netdevice(dev); 878 break; 879 default: 880 err = -EINVAL; 881 } 882 return err; 883 } 884 885 static const struct net_device_ops vti6_netdev_ops = { 886 .ndo_init = vti6_dev_init, 887 .ndo_uninit = vti6_dev_uninit, 888 .ndo_start_xmit = vti6_tnl_xmit, 889 .ndo_siocdevprivate = vti6_siocdevprivate, 890 .ndo_get_iflink = ip6_tnl_get_iflink, 891 }; 892 893 /** 894 * vti6_dev_setup - setup virtual tunnel device 895 * @dev: virtual device associated with tunnel 896 * 897 * Description: 898 * Initialize function pointers and device parameters 899 **/ 900 static void vti6_dev_setup(struct net_device *dev) 901 { 902 dev->netdev_ops = &vti6_netdev_ops; 903 dev->header_ops = &ip_tunnel_header_ops; 904 905 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 906 dev->type = ARPHRD_TUNNEL6; 907 dev->min_mtu = IPV4_MIN_MTU; 908 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 909 dev->flags |= IFF_NOARP; 910 dev->addr_len = sizeof(struct in6_addr); 911 netif_keep_dst(dev); 912 /* This perm addr will be used as interface identifier by IPv6 */ 913 dev->addr_assign_type = NET_ADDR_RANDOM; 914 eth_random_addr(dev->perm_addr); 915 } 916 917 /** 918 * vti6_dev_init_gen - general initializer for all tunnel devices 919 * @dev: virtual device associated with tunnel 920 **/ 921 static inline int vti6_dev_init_gen(struct net_device *dev) 922 { 923 struct ip6_tnl *t = netdev_priv(dev); 924 925 t->dev = dev; 926 t->net = dev_net(dev); 927 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL); 928 netdev_lockdep_set_classes(dev); 929 return 0; 930 } 931 932 /** 933 * vti6_dev_init - initializer for all non fallback tunnel devices 934 * @dev: virtual device associated with tunnel 935 **/ 936 static int vti6_dev_init(struct net_device *dev) 937 { 938 struct ip6_tnl *t = netdev_priv(dev); 939 int err = vti6_dev_init_gen(dev); 940 941 if (err) 942 return err; 943 vti6_link_config(t, true); 944 return 0; 945 } 946 947 /** 948 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 949 * @dev: fallback device 950 * 951 * Return: 0 952 **/ 953 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 954 { 955 struct ip6_tnl *t = netdev_priv(dev); 956 struct net *net = dev_net(dev); 957 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 958 959 t->parms.proto = IPPROTO_IPV6; 960 961 rcu_assign_pointer(ip6n->tnls_wc[0], t); 962 return 0; 963 } 964 965 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 966 struct netlink_ext_ack *extack) 967 { 968 return 0; 969 } 970 971 static void vti6_netlink_parms(struct nlattr *data[], 972 struct __ip6_tnl_parm *parms) 973 { 974 memset(parms, 0, sizeof(*parms)); 975 976 if (!data) 977 return; 978 979 if (data[IFLA_VTI_LINK]) 980 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 981 982 if (data[IFLA_VTI_LOCAL]) 983 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 984 985 if (data[IFLA_VTI_REMOTE]) 986 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 987 988 if (data[IFLA_VTI_IKEY]) 989 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 990 991 if (data[IFLA_VTI_OKEY]) 992 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 993 994 if (data[IFLA_VTI_FWMARK]) 995 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 996 } 997 998 static int vti6_newlink(struct net *src_net, struct net_device *dev, 999 struct nlattr *tb[], struct nlattr *data[], 1000 struct netlink_ext_ack *extack) 1001 { 1002 struct net *net = dev_net(dev); 1003 struct ip6_tnl *nt; 1004 1005 nt = netdev_priv(dev); 1006 vti6_netlink_parms(data, &nt->parms); 1007 1008 nt->parms.proto = IPPROTO_IPV6; 1009 1010 if (vti6_locate(net, &nt->parms, 0)) 1011 return -EEXIST; 1012 1013 return vti6_tnl_create2(dev); 1014 } 1015 1016 static void vti6_dellink(struct net_device *dev, struct list_head *head) 1017 { 1018 struct net *net = dev_net(dev); 1019 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1020 1021 if (dev != ip6n->fb_tnl_dev) 1022 unregister_netdevice_queue(dev, head); 1023 } 1024 1025 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1026 struct nlattr *data[], 1027 struct netlink_ext_ack *extack) 1028 { 1029 struct ip6_tnl *t; 1030 struct __ip6_tnl_parm p; 1031 struct net *net = dev_net(dev); 1032 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1033 1034 if (dev == ip6n->fb_tnl_dev) 1035 return -EINVAL; 1036 1037 vti6_netlink_parms(data, &p); 1038 1039 t = vti6_locate(net, &p, 0); 1040 1041 if (t) { 1042 if (t->dev != dev) 1043 return -EEXIST; 1044 } else 1045 t = netdev_priv(dev); 1046 1047 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1048 } 1049 1050 static size_t vti6_get_size(const struct net_device *dev) 1051 { 1052 return 1053 /* IFLA_VTI_LINK */ 1054 nla_total_size(4) + 1055 /* IFLA_VTI_LOCAL */ 1056 nla_total_size(sizeof(struct in6_addr)) + 1057 /* IFLA_VTI_REMOTE */ 1058 nla_total_size(sizeof(struct in6_addr)) + 1059 /* IFLA_VTI_IKEY */ 1060 nla_total_size(4) + 1061 /* IFLA_VTI_OKEY */ 1062 nla_total_size(4) + 1063 /* IFLA_VTI_FWMARK */ 1064 nla_total_size(4) + 1065 0; 1066 } 1067 1068 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1069 { 1070 struct ip6_tnl *tunnel = netdev_priv(dev); 1071 struct __ip6_tnl_parm *parm = &tunnel->parms; 1072 1073 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1074 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1075 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1076 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1077 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1078 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1079 goto nla_put_failure; 1080 return 0; 1081 1082 nla_put_failure: 1083 return -EMSGSIZE; 1084 } 1085 1086 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1087 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1088 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1089 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1090 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1091 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1092 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1093 }; 1094 1095 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1096 .kind = "vti6", 1097 .maxtype = IFLA_VTI_MAX, 1098 .policy = vti6_policy, 1099 .priv_size = sizeof(struct ip6_tnl), 1100 .setup = vti6_dev_setup, 1101 .validate = vti6_validate, 1102 .newlink = vti6_newlink, 1103 .dellink = vti6_dellink, 1104 .changelink = vti6_changelink, 1105 .get_size = vti6_get_size, 1106 .fill_info = vti6_fill_info, 1107 .get_link_net = ip6_tnl_get_link_net, 1108 }; 1109 1110 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1111 struct list_head *list) 1112 { 1113 int h; 1114 struct ip6_tnl *t; 1115 1116 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1117 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1118 while (t) { 1119 unregister_netdevice_queue(t->dev, list); 1120 t = rtnl_dereference(t->next); 1121 } 1122 } 1123 1124 t = rtnl_dereference(ip6n->tnls_wc[0]); 1125 if (t) 1126 unregister_netdevice_queue(t->dev, list); 1127 } 1128 1129 static int __net_init vti6_init_net(struct net *net) 1130 { 1131 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1132 struct ip6_tnl *t = NULL; 1133 int err; 1134 1135 ip6n->tnls[0] = ip6n->tnls_wc; 1136 ip6n->tnls[1] = ip6n->tnls_r_l; 1137 1138 if (!net_has_fallback_tunnels(net)) 1139 return 0; 1140 err = -ENOMEM; 1141 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1142 NET_NAME_UNKNOWN, vti6_dev_setup); 1143 1144 if (!ip6n->fb_tnl_dev) 1145 goto err_alloc_dev; 1146 dev_net_set(ip6n->fb_tnl_dev, net); 1147 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1148 1149 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1150 if (err < 0) 1151 goto err_register; 1152 1153 err = register_netdev(ip6n->fb_tnl_dev); 1154 if (err < 0) 1155 goto err_register; 1156 1157 t = netdev_priv(ip6n->fb_tnl_dev); 1158 1159 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1160 return 0; 1161 1162 err_register: 1163 free_netdev(ip6n->fb_tnl_dev); 1164 err_alloc_dev: 1165 return err; 1166 } 1167 1168 static void __net_exit vti6_exit_batch_rtnl(struct list_head *net_list, 1169 struct list_head *dev_to_kill) 1170 { 1171 struct vti6_net *ip6n; 1172 struct net *net; 1173 1174 ASSERT_RTNL(); 1175 list_for_each_entry(net, net_list, exit_list) { 1176 ip6n = net_generic(net, vti6_net_id); 1177 vti6_destroy_tunnels(ip6n, dev_to_kill); 1178 } 1179 } 1180 1181 static struct pernet_operations vti6_net_ops = { 1182 .init = vti6_init_net, 1183 .exit_batch_rtnl = vti6_exit_batch_rtnl, 1184 .id = &vti6_net_id, 1185 .size = sizeof(struct vti6_net), 1186 }; 1187 1188 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1189 .handler = vti6_rcv, 1190 .input_handler = vti6_input_proto, 1191 .cb_handler = vti6_rcv_cb, 1192 .err_handler = vti6_err, 1193 .priority = 100, 1194 }; 1195 1196 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1197 .handler = vti6_rcv, 1198 .input_handler = vti6_input_proto, 1199 .cb_handler = vti6_rcv_cb, 1200 .err_handler = vti6_err, 1201 .priority = 100, 1202 }; 1203 1204 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1205 .handler = vti6_rcv, 1206 .input_handler = vti6_input_proto, 1207 .cb_handler = vti6_rcv_cb, 1208 .err_handler = vti6_err, 1209 .priority = 100, 1210 }; 1211 1212 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1213 static int vti6_rcv_tunnel(struct sk_buff *skb) 1214 { 1215 const xfrm_address_t *saddr; 1216 __be32 spi; 1217 1218 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1219 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1220 1221 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1222 } 1223 1224 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1225 .handler = vti6_rcv_tunnel, 1226 .cb_handler = vti6_rcv_cb, 1227 .err_handler = vti6_err, 1228 .priority = 0, 1229 }; 1230 1231 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1232 .handler = vti6_rcv_tunnel, 1233 .cb_handler = vti6_rcv_cb, 1234 .err_handler = vti6_err, 1235 .priority = 0, 1236 }; 1237 #endif 1238 1239 /** 1240 * vti6_tunnel_init - register protocol and reserve needed resources 1241 * 1242 * Return: 0 on success 1243 **/ 1244 static int __init vti6_tunnel_init(void) 1245 { 1246 const char *msg; 1247 int err; 1248 1249 msg = "tunnel device"; 1250 err = register_pernet_device(&vti6_net_ops); 1251 if (err < 0) 1252 goto pernet_dev_failed; 1253 1254 msg = "tunnel protocols"; 1255 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1256 if (err < 0) 1257 goto xfrm_proto_esp_failed; 1258 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1259 if (err < 0) 1260 goto xfrm_proto_ah_failed; 1261 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1262 if (err < 0) 1263 goto xfrm_proto_comp_failed; 1264 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1265 msg = "ipv6 tunnel"; 1266 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1267 if (err < 0) 1268 goto vti_tunnel_ipv6_failed; 1269 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1270 if (err < 0) 1271 goto vti_tunnel_ip6ip_failed; 1272 #endif 1273 1274 msg = "netlink interface"; 1275 err = rtnl_link_register(&vti6_link_ops); 1276 if (err < 0) 1277 goto rtnl_link_failed; 1278 1279 return 0; 1280 1281 rtnl_link_failed: 1282 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1283 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1284 vti_tunnel_ip6ip_failed: 1285 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1286 vti_tunnel_ipv6_failed: 1287 #endif 1288 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1289 xfrm_proto_comp_failed: 1290 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1291 xfrm_proto_ah_failed: 1292 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1293 xfrm_proto_esp_failed: 1294 unregister_pernet_device(&vti6_net_ops); 1295 pernet_dev_failed: 1296 pr_err("vti6 init: failed to register %s\n", msg); 1297 return err; 1298 } 1299 1300 /** 1301 * vti6_tunnel_cleanup - free resources and unregister protocol 1302 **/ 1303 static void __exit vti6_tunnel_cleanup(void) 1304 { 1305 rtnl_link_unregister(&vti6_link_ops); 1306 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1307 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1308 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1309 #endif 1310 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1311 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1312 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1313 unregister_pernet_device(&vti6_net_ops); 1314 } 1315 1316 module_init(vti6_tunnel_init); 1317 module_exit(vti6_tunnel_cleanup); 1318 MODULE_LICENSE("GPL"); 1319 MODULE_ALIAS_RTNL_LINK("vti6"); 1320 MODULE_ALIAS_NETDEV("ip6_vti0"); 1321 MODULE_AUTHOR("Steffen Klassert"); 1322 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1323