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 vti6_net *ip6n = net_generic(t->net, vti6_net_id); 181 int err; 182 183 dev->rtnl_link_ops = &vti6_link_ops; 184 err = register_netdevice(dev); 185 if (err < 0) 186 goto out; 187 188 strcpy(t->parms.name, dev->name); 189 190 vti6_tnl_link(ip6n, t); 191 192 return 0; 193 194 out: 195 return err; 196 } 197 198 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 199 { 200 struct net_device *dev; 201 struct ip6_tnl *t; 202 char name[IFNAMSIZ]; 203 int err; 204 205 if (p->name[0]) { 206 if (!dev_valid_name(p->name)) 207 goto failed; 208 strscpy(name, p->name, IFNAMSIZ); 209 } else { 210 sprintf(name, "ip6_vti%%d"); 211 } 212 213 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 214 if (!dev) 215 goto failed; 216 217 dev_net_set(dev, net); 218 219 t = netdev_priv(dev); 220 t->parms = *p; 221 t->net = dev_net(dev); 222 223 err = vti6_tnl_create2(dev); 224 if (err < 0) 225 goto failed_free; 226 227 return t; 228 229 failed_free: 230 free_netdev(dev); 231 failed: 232 return NULL; 233 } 234 235 /** 236 * vti6_locate - find or create tunnel matching given parameters 237 * @net: network namespace 238 * @p: tunnel parameters 239 * @create: != 0 if allowed to create new tunnel if no match found 240 * 241 * Description: 242 * vti6_locate() first tries to locate an existing tunnel 243 * based on @parms. If this is unsuccessful, but @create is set a new 244 * tunnel device is created and registered for use. 245 * 246 * Return: 247 * matching tunnel or NULL 248 **/ 249 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 250 int create) 251 { 252 const struct in6_addr *remote = &p->raddr; 253 const struct in6_addr *local = &p->laddr; 254 struct ip6_tnl __rcu **tp; 255 struct ip6_tnl *t; 256 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 257 258 for (tp = vti6_tnl_bucket(ip6n, p); 259 (t = rtnl_dereference(*tp)) != NULL; 260 tp = &t->next) { 261 if (ipv6_addr_equal(local, &t->parms.laddr) && 262 ipv6_addr_equal(remote, &t->parms.raddr)) { 263 if (create) 264 return NULL; 265 266 return t; 267 } 268 } 269 if (!create) 270 return NULL; 271 return vti6_tnl_create(net, p); 272 } 273 274 /** 275 * vti6_dev_uninit - tunnel device uninitializer 276 * @dev: the device to be destroyed 277 * 278 * Description: 279 * vti6_dev_uninit() removes tunnel from its list 280 **/ 281 static void vti6_dev_uninit(struct net_device *dev) 282 { 283 struct ip6_tnl *t = netdev_priv(dev); 284 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 285 286 if (dev == ip6n->fb_tnl_dev) 287 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 288 else 289 vti6_tnl_unlink(ip6n, t); 290 netdev_put(dev, &t->dev_tracker); 291 } 292 293 static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 294 int encap_type) 295 { 296 struct ip6_tnl *t; 297 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 298 299 rcu_read_lock(); 300 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 301 if (t) { 302 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 303 rcu_read_unlock(); 304 goto discard; 305 } 306 307 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 308 rcu_read_unlock(); 309 goto discard; 310 } 311 312 ipv6h = ipv6_hdr(skb); 313 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 314 DEV_STATS_INC(t->dev, rx_dropped); 315 rcu_read_unlock(); 316 goto discard; 317 } 318 319 rcu_read_unlock(); 320 321 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 322 XFRM_SPI_SKB_CB(skb)->family = AF_INET6; 323 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr); 324 return xfrm_input(skb, nexthdr, spi, encap_type); 325 } 326 rcu_read_unlock(); 327 return -EINVAL; 328 discard: 329 kfree_skb(skb); 330 return 0; 331 } 332 333 static int vti6_rcv(struct sk_buff *skb) 334 { 335 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff]; 336 337 return vti6_input_proto(skb, nexthdr, 0, 0); 338 } 339 340 static int vti6_rcv_cb(struct sk_buff *skb, int err) 341 { 342 unsigned short family; 343 struct net_device *dev; 344 struct xfrm_state *x; 345 const struct xfrm_mode *inner_mode; 346 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 347 u32 orig_mark = skb->mark; 348 int ret; 349 350 if (!t) 351 return 1; 352 353 dev = t->dev; 354 355 if (err) { 356 DEV_STATS_INC(dev, rx_errors); 357 DEV_STATS_INC(dev, rx_dropped); 358 359 return 0; 360 } 361 362 x = xfrm_input_state(skb); 363 364 inner_mode = &x->inner_mode; 365 366 if (x->sel.family == AF_UNSPEC) { 367 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 368 if (inner_mode == NULL) { 369 XFRM_INC_STATS(dev_net(skb->dev), 370 LINUX_MIB_XFRMINSTATEMODEERROR); 371 return -EINVAL; 372 } 373 } 374 375 family = inner_mode->family; 376 377 skb->mark = be32_to_cpu(t->parms.i_key); 378 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 379 skb->mark = orig_mark; 380 381 if (!ret) 382 return -EPERM; 383 384 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 385 skb->dev = dev; 386 dev_sw_netstats_rx_add(dev, skb->len); 387 388 return 0; 389 } 390 391 /** 392 * vti6_addr_conflict - compare packet addresses to tunnel's own 393 * @t: the outgoing tunnel device 394 * @hdr: IPv6 header from the incoming packet 395 * 396 * Description: 397 * Avoid trivial tunneling loop by checking that tunnel exit-point 398 * doesn't match source of incoming packet. 399 * 400 * Return: 401 * 1 if conflict, 402 * 0 else 403 **/ 404 static inline bool 405 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 406 { 407 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 408 } 409 410 static bool vti6_state_check(const struct xfrm_state *x, 411 const struct in6_addr *dst, 412 const struct in6_addr *src) 413 { 414 xfrm_address_t *daddr = (xfrm_address_t *)dst; 415 xfrm_address_t *saddr = (xfrm_address_t *)src; 416 417 /* if there is no transform then this tunnel is not functional. 418 * Or if the xfrm is not mode tunnel. 419 */ 420 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 421 x->props.family != AF_INET6) 422 return false; 423 424 if (ipv6_addr_any(dst)) 425 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 426 427 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 428 return false; 429 430 return true; 431 } 432 433 /** 434 * vti6_xmit - send a packet 435 * @skb: the outgoing socket buffer 436 * @dev: the outgoing tunnel device 437 * @fl: the flow informations for the xfrm_lookup 438 **/ 439 static int 440 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 441 { 442 struct ip6_tnl *t = netdev_priv(dev); 443 struct dst_entry *dst = skb_dst(skb); 444 struct net_device *tdev; 445 struct xfrm_state *x; 446 int pkt_len = skb->len; 447 int err = -1; 448 int mtu; 449 450 if (!dst) { 451 switch (skb->protocol) { 452 case htons(ETH_P_IP): { 453 struct rtable *rt; 454 455 fl->u.ip4.flowi4_oif = dev->ifindex; 456 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 457 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 458 if (IS_ERR(rt)) 459 goto tx_err_link_failure; 460 dst = &rt->dst; 461 skb_dst_set(skb, dst); 462 break; 463 } 464 case htons(ETH_P_IPV6): 465 fl->u.ip6.flowi6_oif = dev->ifindex; 466 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 467 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 468 if (dst->error) { 469 dst_release(dst); 470 dst = NULL; 471 goto tx_err_link_failure; 472 } 473 skb_dst_set(skb, dst); 474 break; 475 default: 476 goto tx_err_link_failure; 477 } 478 } 479 480 dst_hold(dst); 481 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 482 if (IS_ERR(dst)) { 483 err = PTR_ERR(dst); 484 dst = NULL; 485 goto tx_err_link_failure; 486 } 487 488 if (dst->flags & DST_XFRM_QUEUE) 489 goto xmit; 490 491 x = dst->xfrm; 492 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 493 goto tx_err_link_failure; 494 495 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 496 (const struct in6_addr *)&x->id.daddr)) 497 goto tx_err_link_failure; 498 499 tdev = dst->dev; 500 501 if (tdev == dev) { 502 DEV_STATS_INC(dev, collisions); 503 net_warn_ratelimited("%s: Local routing loop detected!\n", 504 t->parms.name); 505 goto tx_err_dst_release; 506 } 507 508 mtu = dst_mtu(dst); 509 if (skb->len > mtu) { 510 skb_dst_update_pmtu_no_confirm(skb, mtu); 511 512 if (skb->protocol == htons(ETH_P_IPV6)) { 513 if (mtu < IPV6_MIN_MTU) 514 mtu = IPV6_MIN_MTU; 515 516 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 517 } else { 518 if (!(ip_hdr(skb)->frag_off & htons(IP_DF))) 519 goto xmit; 520 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 521 htonl(mtu)); 522 } 523 524 err = -EMSGSIZE; 525 goto tx_err_dst_release; 526 } 527 528 xmit: 529 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 530 skb_dst_set(skb, dst); 531 skb->dev = skb_dst(skb)->dev; 532 533 err = dst_output(t->net, skb->sk, skb); 534 if (net_xmit_eval(err) == 0) 535 err = pkt_len; 536 iptunnel_xmit_stats(dev, err); 537 538 return 0; 539 tx_err_link_failure: 540 DEV_STATS_INC(dev, tx_carrier_errors); 541 dst_link_failure(skb); 542 tx_err_dst_release: 543 dst_release(dst); 544 return err; 545 } 546 547 static netdev_tx_t 548 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 549 { 550 struct ip6_tnl *t = netdev_priv(dev); 551 struct flowi fl; 552 int ret; 553 554 if (!pskb_inet_may_pull(skb)) 555 goto tx_err; 556 557 memset(&fl, 0, sizeof(fl)); 558 559 switch (skb->protocol) { 560 case htons(ETH_P_IPV6): 561 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 562 vti6_addr_conflict(t, ipv6_hdr(skb))) 563 goto tx_err; 564 565 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 566 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6); 567 break; 568 case htons(ETH_P_IP): 569 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 570 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET); 571 break; 572 default: 573 goto tx_err; 574 } 575 576 /* override mark with tunnel output key */ 577 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 578 579 ret = vti6_xmit(skb, dev, &fl); 580 if (ret < 0) 581 goto tx_err; 582 583 return NETDEV_TX_OK; 584 585 tx_err: 586 DEV_STATS_INC(dev, tx_errors); 587 DEV_STATS_INC(dev, tx_dropped); 588 kfree_skb(skb); 589 return NETDEV_TX_OK; 590 } 591 592 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 593 u8 type, u8 code, int offset, __be32 info) 594 { 595 __be32 spi; 596 __u32 mark; 597 struct xfrm_state *x; 598 struct ip6_tnl *t; 599 struct ip_esp_hdr *esph; 600 struct ip_auth_hdr *ah; 601 struct ip_comp_hdr *ipch; 602 struct net *net = dev_net(skb->dev); 603 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 604 int protocol = iph->nexthdr; 605 606 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 607 if (!t) 608 return -1; 609 610 mark = be32_to_cpu(t->parms.o_key); 611 612 switch (protocol) { 613 case IPPROTO_ESP: 614 esph = (struct ip_esp_hdr *)(skb->data + offset); 615 spi = esph->spi; 616 break; 617 case IPPROTO_AH: 618 ah = (struct ip_auth_hdr *)(skb->data + offset); 619 spi = ah->spi; 620 break; 621 case IPPROTO_COMP: 622 ipch = (struct ip_comp_hdr *)(skb->data + offset); 623 spi = htonl(ntohs(ipch->cpi)); 624 break; 625 default: 626 return 0; 627 } 628 629 if (type != ICMPV6_PKT_TOOBIG && 630 type != NDISC_REDIRECT) 631 return 0; 632 633 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 634 spi, protocol, AF_INET6); 635 if (!x) 636 return 0; 637 638 if (type == NDISC_REDIRECT) 639 ip6_redirect(skb, net, skb->dev->ifindex, 0, 640 sock_net_uid(net, NULL)); 641 else 642 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 643 xfrm_state_put(x); 644 645 return 0; 646 } 647 648 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 649 { 650 struct net_device *dev = t->dev; 651 struct __ip6_tnl_parm *p = &t->parms; 652 struct net_device *tdev = NULL; 653 int mtu; 654 655 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 656 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 657 658 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 659 IP6_TNL_F_CAP_PER_PACKET); 660 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 661 662 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 663 dev->flags |= IFF_POINTOPOINT; 664 else 665 dev->flags &= ~IFF_POINTOPOINT; 666 667 if (keep_mtu && dev->mtu) { 668 WRITE_ONCE(dev->mtu, 669 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 dev->needs_free_netdev = true; 905 906 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 907 dev->type = ARPHRD_TUNNEL6; 908 dev->min_mtu = IPV4_MIN_MTU; 909 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 910 dev->flags |= IFF_NOARP; 911 dev->addr_len = sizeof(struct in6_addr); 912 netif_keep_dst(dev); 913 /* This perm addr will be used as interface identifier by IPv6 */ 914 dev->addr_assign_type = NET_ADDR_RANDOM; 915 eth_random_addr(dev->perm_addr); 916 } 917 918 /** 919 * vti6_dev_init_gen - general initializer for all tunnel devices 920 * @dev: virtual device associated with tunnel 921 **/ 922 static inline int vti6_dev_init_gen(struct net_device *dev) 923 { 924 struct ip6_tnl *t = netdev_priv(dev); 925 926 t->dev = 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->net = net; 960 t->parms.proto = IPPROTO_IPV6; 961 962 rcu_assign_pointer(ip6n->tnls_wc[0], t); 963 return 0; 964 } 965 966 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 967 struct netlink_ext_ack *extack) 968 { 969 return 0; 970 } 971 972 static void vti6_netlink_parms(struct nlattr *data[], 973 struct __ip6_tnl_parm *parms) 974 { 975 memset(parms, 0, sizeof(*parms)); 976 977 if (!data) 978 return; 979 980 if (data[IFLA_VTI_LINK]) 981 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 982 983 if (data[IFLA_VTI_LOCAL]) 984 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 985 986 if (data[IFLA_VTI_REMOTE]) 987 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 988 989 if (data[IFLA_VTI_IKEY]) 990 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 991 992 if (data[IFLA_VTI_OKEY]) 993 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 994 995 if (data[IFLA_VTI_FWMARK]) 996 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 997 } 998 999 static int vti6_newlink(struct net_device *dev, 1000 struct rtnl_newlink_params *params, 1001 struct netlink_ext_ack *extack) 1002 { 1003 struct nlattr **data = params->data; 1004 struct ip6_tnl *nt; 1005 struct net *net; 1006 1007 net = params->link_net ? : dev_net(dev); 1008 nt = netdev_priv(dev); 1009 vti6_netlink_parms(data, &nt->parms); 1010 1011 nt->parms.proto = IPPROTO_IPV6; 1012 nt->net = net; 1013 1014 if (vti6_locate(net, &nt->parms, 0)) 1015 return -EEXIST; 1016 1017 return vti6_tnl_create2(dev); 1018 } 1019 1020 static void vti6_dellink(struct net_device *dev, struct list_head *head) 1021 { 1022 struct net *net = dev_net(dev); 1023 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1024 1025 if (dev != ip6n->fb_tnl_dev) 1026 unregister_netdevice_queue(dev, head); 1027 } 1028 1029 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1030 struct nlattr *data[], 1031 struct netlink_ext_ack *extack) 1032 { 1033 struct ip6_tnl *t; 1034 struct __ip6_tnl_parm p; 1035 struct net *net = dev_net(dev); 1036 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1037 1038 if (dev == ip6n->fb_tnl_dev) 1039 return -EINVAL; 1040 1041 vti6_netlink_parms(data, &p); 1042 1043 t = vti6_locate(net, &p, 0); 1044 1045 if (t) { 1046 if (t->dev != dev) 1047 return -EEXIST; 1048 } else 1049 t = netdev_priv(dev); 1050 1051 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1052 } 1053 1054 static size_t vti6_get_size(const struct net_device *dev) 1055 { 1056 return 1057 /* IFLA_VTI_LINK */ 1058 nla_total_size(4) + 1059 /* IFLA_VTI_LOCAL */ 1060 nla_total_size(sizeof(struct in6_addr)) + 1061 /* IFLA_VTI_REMOTE */ 1062 nla_total_size(sizeof(struct in6_addr)) + 1063 /* IFLA_VTI_IKEY */ 1064 nla_total_size(4) + 1065 /* IFLA_VTI_OKEY */ 1066 nla_total_size(4) + 1067 /* IFLA_VTI_FWMARK */ 1068 nla_total_size(4) + 1069 0; 1070 } 1071 1072 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1073 { 1074 struct ip6_tnl *tunnel = netdev_priv(dev); 1075 struct __ip6_tnl_parm *parm = &tunnel->parms; 1076 1077 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1078 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1079 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1080 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1081 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1082 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1083 goto nla_put_failure; 1084 return 0; 1085 1086 nla_put_failure: 1087 return -EMSGSIZE; 1088 } 1089 1090 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1091 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1092 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1093 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1094 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1095 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1096 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1097 }; 1098 1099 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1100 .kind = "vti6", 1101 .maxtype = IFLA_VTI_MAX, 1102 .policy = vti6_policy, 1103 .priv_size = sizeof(struct ip6_tnl), 1104 .setup = vti6_dev_setup, 1105 .validate = vti6_validate, 1106 .newlink = vti6_newlink, 1107 .dellink = vti6_dellink, 1108 .changelink = vti6_changelink, 1109 .get_size = vti6_get_size, 1110 .fill_info = vti6_fill_info, 1111 .get_link_net = ip6_tnl_get_link_net, 1112 }; 1113 1114 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1115 struct list_head *list) 1116 { 1117 int h; 1118 struct ip6_tnl *t; 1119 1120 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1121 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1122 while (t) { 1123 unregister_netdevice_queue(t->dev, list); 1124 t = rtnl_dereference(t->next); 1125 } 1126 } 1127 1128 t = rtnl_dereference(ip6n->tnls_wc[0]); 1129 if (t) 1130 unregister_netdevice_queue(t->dev, list); 1131 } 1132 1133 static int __net_init vti6_init_net(struct net *net) 1134 { 1135 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1136 struct ip6_tnl *t = NULL; 1137 int err; 1138 1139 ip6n->tnls[0] = ip6n->tnls_wc; 1140 ip6n->tnls[1] = ip6n->tnls_r_l; 1141 1142 if (!net_has_fallback_tunnels(net)) 1143 return 0; 1144 err = -ENOMEM; 1145 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1146 NET_NAME_UNKNOWN, vti6_dev_setup); 1147 1148 if (!ip6n->fb_tnl_dev) 1149 goto err_alloc_dev; 1150 dev_net_set(ip6n->fb_tnl_dev, net); 1151 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1152 1153 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1154 if (err < 0) 1155 goto err_register; 1156 1157 err = register_netdev(ip6n->fb_tnl_dev); 1158 if (err < 0) 1159 goto err_register; 1160 1161 t = netdev_priv(ip6n->fb_tnl_dev); 1162 1163 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1164 return 0; 1165 1166 err_register: 1167 free_netdev(ip6n->fb_tnl_dev); 1168 err_alloc_dev: 1169 return err; 1170 } 1171 1172 static void __net_exit vti6_exit_batch_rtnl(struct list_head *net_list, 1173 struct list_head *dev_to_kill) 1174 { 1175 struct vti6_net *ip6n; 1176 struct net *net; 1177 1178 ASSERT_RTNL(); 1179 list_for_each_entry(net, net_list, exit_list) { 1180 ip6n = net_generic(net, vti6_net_id); 1181 vti6_destroy_tunnels(ip6n, dev_to_kill); 1182 } 1183 } 1184 1185 static struct pernet_operations vti6_net_ops = { 1186 .init = vti6_init_net, 1187 .exit_batch_rtnl = vti6_exit_batch_rtnl, 1188 .id = &vti6_net_id, 1189 .size = sizeof(struct vti6_net), 1190 }; 1191 1192 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1193 .handler = vti6_rcv, 1194 .input_handler = vti6_input_proto, 1195 .cb_handler = vti6_rcv_cb, 1196 .err_handler = vti6_err, 1197 .priority = 100, 1198 }; 1199 1200 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1201 .handler = vti6_rcv, 1202 .input_handler = vti6_input_proto, 1203 .cb_handler = vti6_rcv_cb, 1204 .err_handler = vti6_err, 1205 .priority = 100, 1206 }; 1207 1208 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1209 .handler = vti6_rcv, 1210 .input_handler = vti6_input_proto, 1211 .cb_handler = vti6_rcv_cb, 1212 .err_handler = vti6_err, 1213 .priority = 100, 1214 }; 1215 1216 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1217 static int vti6_rcv_tunnel(struct sk_buff *skb) 1218 { 1219 const xfrm_address_t *saddr; 1220 __be32 spi; 1221 1222 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1223 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1224 1225 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1226 } 1227 1228 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1229 .handler = vti6_rcv_tunnel, 1230 .cb_handler = vti6_rcv_cb, 1231 .err_handler = vti6_err, 1232 .priority = 0, 1233 }; 1234 1235 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1236 .handler = vti6_rcv_tunnel, 1237 .cb_handler = vti6_rcv_cb, 1238 .err_handler = vti6_err, 1239 .priority = 0, 1240 }; 1241 #endif 1242 1243 /** 1244 * vti6_tunnel_init - register protocol and reserve needed resources 1245 * 1246 * Return: 0 on success 1247 **/ 1248 static int __init vti6_tunnel_init(void) 1249 { 1250 const char *msg; 1251 int err; 1252 1253 msg = "tunnel device"; 1254 err = register_pernet_device(&vti6_net_ops); 1255 if (err < 0) 1256 goto pernet_dev_failed; 1257 1258 msg = "tunnel protocols"; 1259 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1260 if (err < 0) 1261 goto xfrm_proto_esp_failed; 1262 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1263 if (err < 0) 1264 goto xfrm_proto_ah_failed; 1265 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1266 if (err < 0) 1267 goto xfrm_proto_comp_failed; 1268 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1269 msg = "ipv6 tunnel"; 1270 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1271 if (err < 0) 1272 goto vti_tunnel_ipv6_failed; 1273 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1274 if (err < 0) 1275 goto vti_tunnel_ip6ip_failed; 1276 #endif 1277 1278 msg = "netlink interface"; 1279 err = rtnl_link_register(&vti6_link_ops); 1280 if (err < 0) 1281 goto rtnl_link_failed; 1282 1283 return 0; 1284 1285 rtnl_link_failed: 1286 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1287 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1288 vti_tunnel_ip6ip_failed: 1289 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1290 vti_tunnel_ipv6_failed: 1291 #endif 1292 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1293 xfrm_proto_comp_failed: 1294 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1295 xfrm_proto_ah_failed: 1296 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1297 xfrm_proto_esp_failed: 1298 unregister_pernet_device(&vti6_net_ops); 1299 pernet_dev_failed: 1300 pr_err("vti6 init: failed to register %s\n", msg); 1301 return err; 1302 } 1303 1304 /** 1305 * vti6_tunnel_cleanup - free resources and unregister protocol 1306 **/ 1307 static void __exit vti6_tunnel_cleanup(void) 1308 { 1309 rtnl_link_unregister(&vti6_link_ops); 1310 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1311 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1312 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1313 #endif 1314 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1315 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1316 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1317 unregister_pernet_device(&vti6_net_ops); 1318 } 1319 1320 module_init(vti6_tunnel_init); 1321 module_exit(vti6_tunnel_cleanup); 1322 MODULE_LICENSE("GPL"); 1323 MODULE_ALIAS_RTNL_LINK("vti6"); 1324 MODULE_ALIAS_NETDEV("ip6_vti0"); 1325 MODULE_AUTHOR("Steffen Klassert"); 1326 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1327