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 vti6_tnl_link(ip6n, t); 197 198 return 0; 199 200 out: 201 return err; 202 } 203 204 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 205 { 206 struct net_device *dev; 207 struct ip6_tnl *t; 208 char name[IFNAMSIZ]; 209 int err; 210 211 if (p->name[0]) { 212 if (!dev_valid_name(p->name)) 213 goto failed; 214 strlcpy(name, p->name, IFNAMSIZ); 215 } else { 216 sprintf(name, "ip6_vti%%d"); 217 } 218 219 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 220 if (!dev) 221 goto failed; 222 223 dev_net_set(dev, net); 224 225 t = netdev_priv(dev); 226 t->parms = *p; 227 t->net = dev_net(dev); 228 229 err = vti6_tnl_create2(dev); 230 if (err < 0) 231 goto failed_free; 232 233 return t; 234 235 failed_free: 236 free_netdev(dev); 237 failed: 238 return NULL; 239 } 240 241 /** 242 * vti6_locate - find or create tunnel matching given parameters 243 * @net: network namespace 244 * @p: tunnel parameters 245 * @create: != 0 if allowed to create new tunnel if no match found 246 * 247 * Description: 248 * vti6_locate() first tries to locate an existing tunnel 249 * based on @parms. If this is unsuccessful, but @create is set a new 250 * tunnel device is created and registered for use. 251 * 252 * Return: 253 * matching tunnel or NULL 254 **/ 255 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 256 int create) 257 { 258 const struct in6_addr *remote = &p->raddr; 259 const struct in6_addr *local = &p->laddr; 260 struct ip6_tnl __rcu **tp; 261 struct ip6_tnl *t; 262 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 263 264 for (tp = vti6_tnl_bucket(ip6n, p); 265 (t = rtnl_dereference(*tp)) != NULL; 266 tp = &t->next) { 267 if (ipv6_addr_equal(local, &t->parms.laddr) && 268 ipv6_addr_equal(remote, &t->parms.raddr)) { 269 if (create) 270 return NULL; 271 272 return t; 273 } 274 } 275 if (!create) 276 return NULL; 277 return vti6_tnl_create(net, p); 278 } 279 280 /** 281 * vti6_dev_uninit - tunnel device uninitializer 282 * @dev: the device to be destroyed 283 * 284 * Description: 285 * vti6_dev_uninit() removes tunnel from its list 286 **/ 287 static void vti6_dev_uninit(struct net_device *dev) 288 { 289 struct ip6_tnl *t = netdev_priv(dev); 290 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 291 292 if (dev == ip6n->fb_tnl_dev) 293 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 294 else 295 vti6_tnl_unlink(ip6n, t); 296 dev_put(dev); 297 } 298 299 static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 300 int encap_type) 301 { 302 struct ip6_tnl *t; 303 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 304 305 rcu_read_lock(); 306 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 307 if (t) { 308 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 309 rcu_read_unlock(); 310 goto discard; 311 } 312 313 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 314 rcu_read_unlock(); 315 goto discard; 316 } 317 318 ipv6h = ipv6_hdr(skb); 319 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 320 t->dev->stats.rx_dropped++; 321 rcu_read_unlock(); 322 goto discard; 323 } 324 325 rcu_read_unlock(); 326 327 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 328 XFRM_SPI_SKB_CB(skb)->family = AF_INET6; 329 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr); 330 return xfrm_input(skb, nexthdr, spi, encap_type); 331 } 332 rcu_read_unlock(); 333 return -EINVAL; 334 discard: 335 kfree_skb(skb); 336 return 0; 337 } 338 339 static int vti6_rcv(struct sk_buff *skb) 340 { 341 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff]; 342 343 return vti6_input_proto(skb, nexthdr, 0, 0); 344 } 345 346 static int vti6_rcv_cb(struct sk_buff *skb, int err) 347 { 348 unsigned short family; 349 struct net_device *dev; 350 struct xfrm_state *x; 351 const struct xfrm_mode *inner_mode; 352 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 353 u32 orig_mark = skb->mark; 354 int ret; 355 356 if (!t) 357 return 1; 358 359 dev = t->dev; 360 361 if (err) { 362 dev->stats.rx_errors++; 363 dev->stats.rx_dropped++; 364 365 return 0; 366 } 367 368 x = xfrm_input_state(skb); 369 370 inner_mode = &x->inner_mode; 371 372 if (x->sel.family == AF_UNSPEC) { 373 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 374 if (inner_mode == NULL) { 375 XFRM_INC_STATS(dev_net(skb->dev), 376 LINUX_MIB_XFRMINSTATEMODEERROR); 377 return -EINVAL; 378 } 379 } 380 381 family = inner_mode->family; 382 383 skb->mark = be32_to_cpu(t->parms.i_key); 384 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 385 skb->mark = orig_mark; 386 387 if (!ret) 388 return -EPERM; 389 390 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 391 skb->dev = dev; 392 dev_sw_netstats_rx_add(dev, skb->len); 393 394 return 0; 395 } 396 397 /** 398 * vti6_addr_conflict - compare packet addresses to tunnel's own 399 * @t: the outgoing tunnel device 400 * @hdr: IPv6 header from the incoming packet 401 * 402 * Description: 403 * Avoid trivial tunneling loop by checking that tunnel exit-point 404 * doesn't match source of incoming packet. 405 * 406 * Return: 407 * 1 if conflict, 408 * 0 else 409 **/ 410 static inline bool 411 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 412 { 413 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 414 } 415 416 static bool vti6_state_check(const struct xfrm_state *x, 417 const struct in6_addr *dst, 418 const struct in6_addr *src) 419 { 420 xfrm_address_t *daddr = (xfrm_address_t *)dst; 421 xfrm_address_t *saddr = (xfrm_address_t *)src; 422 423 /* if there is no transform then this tunnel is not functional. 424 * Or if the xfrm is not mode tunnel. 425 */ 426 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 427 x->props.family != AF_INET6) 428 return false; 429 430 if (ipv6_addr_any(dst)) 431 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 432 433 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 434 return false; 435 436 return true; 437 } 438 439 /** 440 * vti6_xmit - send a packet 441 * @skb: the outgoing socket buffer 442 * @dev: the outgoing tunnel device 443 * @fl: the flow informations for the xfrm_lookup 444 **/ 445 static int 446 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 447 { 448 struct ip6_tnl *t = netdev_priv(dev); 449 struct net_device_stats *stats = &t->dev->stats; 450 struct dst_entry *dst = skb_dst(skb); 451 struct net_device *tdev; 452 struct xfrm_state *x; 453 int pkt_len = skb->len; 454 int err = -1; 455 int mtu; 456 457 if (!dst) { 458 switch (skb->protocol) { 459 case htons(ETH_P_IP): { 460 struct rtable *rt; 461 462 fl->u.ip4.flowi4_oif = dev->ifindex; 463 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 464 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 465 if (IS_ERR(rt)) 466 goto tx_err_link_failure; 467 dst = &rt->dst; 468 skb_dst_set(skb, dst); 469 break; 470 } 471 case htons(ETH_P_IPV6): 472 fl->u.ip6.flowi6_oif = dev->ifindex; 473 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 474 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 475 if (dst->error) { 476 dst_release(dst); 477 dst = NULL; 478 goto tx_err_link_failure; 479 } 480 skb_dst_set(skb, dst); 481 break; 482 default: 483 goto tx_err_link_failure; 484 } 485 } 486 487 dst_hold(dst); 488 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 489 if (IS_ERR(dst)) { 490 err = PTR_ERR(dst); 491 dst = NULL; 492 goto tx_err_link_failure; 493 } 494 495 if (dst->flags & DST_XFRM_QUEUE) 496 goto xmit; 497 498 x = dst->xfrm; 499 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 500 goto tx_err_link_failure; 501 502 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 503 (const struct in6_addr *)&x->id.daddr)) 504 goto tx_err_link_failure; 505 506 tdev = dst->dev; 507 508 if (tdev == dev) { 509 stats->collisions++; 510 net_warn_ratelimited("%s: Local routing loop detected!\n", 511 t->parms.name); 512 goto tx_err_dst_release; 513 } 514 515 mtu = dst_mtu(dst); 516 if (skb->len > mtu) { 517 skb_dst_update_pmtu_no_confirm(skb, mtu); 518 519 if (skb->protocol == htons(ETH_P_IPV6)) { 520 if (mtu < IPV6_MIN_MTU) 521 mtu = IPV6_MIN_MTU; 522 523 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 524 } else { 525 if (!(ip_hdr(skb)->frag_off & htons(IP_DF))) 526 goto xmit; 527 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 528 htonl(mtu)); 529 } 530 531 err = -EMSGSIZE; 532 goto tx_err_dst_release; 533 } 534 535 xmit: 536 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 537 skb_dst_set(skb, dst); 538 skb->dev = skb_dst(skb)->dev; 539 540 err = dst_output(t->net, skb->sk, skb); 541 if (net_xmit_eval(err) == 0) 542 err = pkt_len; 543 iptunnel_xmit_stats(dev, err); 544 545 return 0; 546 tx_err_link_failure: 547 stats->tx_carrier_errors++; 548 dst_link_failure(skb); 549 tx_err_dst_release: 550 dst_release(dst); 551 return err; 552 } 553 554 static netdev_tx_t 555 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 556 { 557 struct ip6_tnl *t = netdev_priv(dev); 558 struct net_device_stats *stats = &t->dev->stats; 559 struct flowi fl; 560 int ret; 561 562 if (!pskb_inet_may_pull(skb)) 563 goto tx_err; 564 565 memset(&fl, 0, sizeof(fl)); 566 567 switch (skb->protocol) { 568 case htons(ETH_P_IPV6): 569 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 570 vti6_addr_conflict(t, ipv6_hdr(skb))) 571 goto tx_err; 572 573 xfrm_decode_session(skb, &fl, AF_INET6); 574 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 575 break; 576 case htons(ETH_P_IP): 577 xfrm_decode_session(skb, &fl, AF_INET); 578 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 579 break; 580 default: 581 goto tx_err; 582 } 583 584 /* override mark with tunnel output key */ 585 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 586 587 ret = vti6_xmit(skb, dev, &fl); 588 if (ret < 0) 589 goto tx_err; 590 591 return NETDEV_TX_OK; 592 593 tx_err: 594 stats->tx_errors++; 595 stats->tx_dropped++; 596 kfree_skb(skb); 597 return NETDEV_TX_OK; 598 } 599 600 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 601 u8 type, u8 code, int offset, __be32 info) 602 { 603 __be32 spi; 604 __u32 mark; 605 struct xfrm_state *x; 606 struct ip6_tnl *t; 607 struct ip_esp_hdr *esph; 608 struct ip_auth_hdr *ah; 609 struct ip_comp_hdr *ipch; 610 struct net *net = dev_net(skb->dev); 611 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 612 int protocol = iph->nexthdr; 613 614 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 615 if (!t) 616 return -1; 617 618 mark = be32_to_cpu(t->parms.o_key); 619 620 switch (protocol) { 621 case IPPROTO_ESP: 622 esph = (struct ip_esp_hdr *)(skb->data + offset); 623 spi = esph->spi; 624 break; 625 case IPPROTO_AH: 626 ah = (struct ip_auth_hdr *)(skb->data + offset); 627 spi = ah->spi; 628 break; 629 case IPPROTO_COMP: 630 ipch = (struct ip_comp_hdr *)(skb->data + offset); 631 spi = htonl(ntohs(ipch->cpi)); 632 break; 633 default: 634 return 0; 635 } 636 637 if (type != ICMPV6_PKT_TOOBIG && 638 type != NDISC_REDIRECT) 639 return 0; 640 641 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 642 spi, protocol, AF_INET6); 643 if (!x) 644 return 0; 645 646 if (type == NDISC_REDIRECT) 647 ip6_redirect(skb, net, skb->dev->ifindex, 0, 648 sock_net_uid(net, NULL)); 649 else 650 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 651 xfrm_state_put(x); 652 653 return 0; 654 } 655 656 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 657 { 658 struct net_device *dev = t->dev; 659 struct __ip6_tnl_parm *p = &t->parms; 660 struct net_device *tdev = NULL; 661 int mtu; 662 663 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 664 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 665 666 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 667 IP6_TNL_F_CAP_PER_PACKET); 668 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 669 670 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 671 dev->flags |= IFF_POINTOPOINT; 672 else 673 dev->flags &= ~IFF_POINTOPOINT; 674 675 if (keep_mtu && dev->mtu) { 676 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu); 677 return; 678 } 679 680 if (p->flags & IP6_TNL_F_CAP_XMIT) { 681 int strict = (ipv6_addr_type(&p->raddr) & 682 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 683 struct rt6_info *rt = rt6_lookup(t->net, 684 &p->raddr, &p->laddr, 685 p->link, NULL, strict); 686 687 if (rt) 688 tdev = rt->dst.dev; 689 ip6_rt_put(rt); 690 } 691 692 if (!tdev && p->link) 693 tdev = __dev_get_by_index(t->net, p->link); 694 695 if (tdev) 696 mtu = tdev->mtu - sizeof(struct ipv6hdr); 697 else 698 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 699 700 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU); 701 } 702 703 /** 704 * vti6_tnl_change - update the tunnel parameters 705 * @t: tunnel to be changed 706 * @p: tunnel configuration parameters 707 * @keep_mtu: MTU was set from userspace, don't re-compute it 708 * 709 * Description: 710 * vti6_tnl_change() updates the tunnel parameters 711 **/ 712 static int 713 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 714 bool keep_mtu) 715 { 716 t->parms.laddr = p->laddr; 717 t->parms.raddr = p->raddr; 718 t->parms.link = p->link; 719 t->parms.i_key = p->i_key; 720 t->parms.o_key = p->o_key; 721 t->parms.proto = p->proto; 722 t->parms.fwmark = p->fwmark; 723 dst_cache_reset(&t->dst_cache); 724 vti6_link_config(t, keep_mtu); 725 return 0; 726 } 727 728 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 729 bool keep_mtu) 730 { 731 struct net *net = dev_net(t->dev); 732 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 733 int err; 734 735 vti6_tnl_unlink(ip6n, t); 736 synchronize_net(); 737 err = vti6_tnl_change(t, p, keep_mtu); 738 vti6_tnl_link(ip6n, t); 739 netdev_state_change(t->dev); 740 return err; 741 } 742 743 static void 744 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 745 { 746 p->laddr = u->laddr; 747 p->raddr = u->raddr; 748 p->link = u->link; 749 p->i_key = u->i_key; 750 p->o_key = u->o_key; 751 p->proto = u->proto; 752 753 memcpy(p->name, u->name, sizeof(u->name)); 754 } 755 756 static void 757 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 758 { 759 u->laddr = p->laddr; 760 u->raddr = p->raddr; 761 u->link = p->link; 762 u->i_key = p->i_key; 763 u->o_key = p->o_key; 764 if (u->i_key) 765 u->i_flags |= GRE_KEY; 766 if (u->o_key) 767 u->o_flags |= GRE_KEY; 768 u->proto = p->proto; 769 770 memcpy(u->name, p->name, sizeof(u->name)); 771 } 772 773 /** 774 * vti6_siocdevprivate - configure vti6 tunnels from userspace 775 * @dev: virtual device associated with tunnel 776 * @ifr: unused 777 * @data: parameters passed from userspace 778 * @cmd: command to be performed 779 * 780 * Description: 781 * vti6_siocdevprivate() is used for managing vti6 tunnels 782 * from userspace. 783 * 784 * The possible commands are the following: 785 * %SIOCGETTUNNEL: get tunnel parameters for device 786 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 787 * %SIOCCHGTUNNEL: change tunnel parameters to those given 788 * %SIOCDELTUNNEL: delete tunnel 789 * 790 * The fallback device "ip6_vti0", created during module 791 * initialization, can be used for creating other tunnel devices. 792 * 793 * Return: 794 * 0 on success, 795 * %-EFAULT if unable to copy data to or from userspace, 796 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 797 * %-EINVAL if passed tunnel parameters are invalid, 798 * %-EEXIST if changing a tunnel's parameters would cause a conflict 799 * %-ENODEV if attempting to change or delete a nonexisting device 800 **/ 801 static int 802 vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd) 803 { 804 int err = 0; 805 struct ip6_tnl_parm2 p; 806 struct __ip6_tnl_parm p1; 807 struct ip6_tnl *t = NULL; 808 struct net *net = dev_net(dev); 809 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 810 811 switch (cmd) { 812 case SIOCGETTUNNEL: 813 if (dev == ip6n->fb_tnl_dev) { 814 if (copy_from_user(&p, data, sizeof(p))) { 815 err = -EFAULT; 816 break; 817 } 818 vti6_parm_from_user(&p1, &p); 819 t = vti6_locate(net, &p1, 0); 820 } else { 821 memset(&p, 0, sizeof(p)); 822 } 823 if (!t) 824 t = netdev_priv(dev); 825 vti6_parm_to_user(&p, &t->parms); 826 if (copy_to_user(data, &p, sizeof(p))) 827 err = -EFAULT; 828 break; 829 case SIOCADDTUNNEL: 830 case SIOCCHGTUNNEL: 831 err = -EPERM; 832 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 833 break; 834 err = -EFAULT; 835 if (copy_from_user(&p, data, sizeof(p))) 836 break; 837 err = -EINVAL; 838 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 839 break; 840 vti6_parm_from_user(&p1, &p); 841 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 842 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 843 if (t) { 844 if (t->dev != dev) { 845 err = -EEXIST; 846 break; 847 } 848 } else 849 t = netdev_priv(dev); 850 851 err = vti6_update(t, &p1, false); 852 } 853 if (t) { 854 err = 0; 855 vti6_parm_to_user(&p, &t->parms); 856 if (copy_to_user(data, &p, sizeof(p))) 857 err = -EFAULT; 858 859 } else 860 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 861 break; 862 case SIOCDELTUNNEL: 863 err = -EPERM; 864 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 865 break; 866 867 if (dev == ip6n->fb_tnl_dev) { 868 err = -EFAULT; 869 if (copy_from_user(&p, data, sizeof(p))) 870 break; 871 err = -ENOENT; 872 vti6_parm_from_user(&p1, &p); 873 t = vti6_locate(net, &p1, 0); 874 if (!t) 875 break; 876 err = -EPERM; 877 if (t->dev == ip6n->fb_tnl_dev) 878 break; 879 dev = t->dev; 880 } 881 err = 0; 882 unregister_netdevice(dev); 883 break; 884 default: 885 err = -EINVAL; 886 } 887 return err; 888 } 889 890 static const struct net_device_ops vti6_netdev_ops = { 891 .ndo_init = vti6_dev_init, 892 .ndo_uninit = vti6_dev_uninit, 893 .ndo_start_xmit = vti6_tnl_xmit, 894 .ndo_siocdevprivate = vti6_siocdevprivate, 895 .ndo_get_stats64 = dev_get_tstats64, 896 .ndo_get_iflink = ip6_tnl_get_iflink, 897 }; 898 899 /** 900 * vti6_dev_setup - setup virtual tunnel device 901 * @dev: virtual device associated with tunnel 902 * 903 * Description: 904 * Initialize function pointers and device parameters 905 **/ 906 static void vti6_dev_setup(struct net_device *dev) 907 { 908 dev->netdev_ops = &vti6_netdev_ops; 909 dev->header_ops = &ip_tunnel_header_ops; 910 dev->needs_free_netdev = true; 911 dev->priv_destructor = vti6_dev_free; 912 913 dev->type = ARPHRD_TUNNEL6; 914 dev->min_mtu = IPV4_MIN_MTU; 915 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 916 dev->flags |= IFF_NOARP; 917 dev->addr_len = sizeof(struct in6_addr); 918 netif_keep_dst(dev); 919 /* This perm addr will be used as interface identifier by IPv6 */ 920 dev->addr_assign_type = NET_ADDR_RANDOM; 921 eth_random_addr(dev->perm_addr); 922 } 923 924 /** 925 * vti6_dev_init_gen - general initializer for all tunnel devices 926 * @dev: virtual device associated with tunnel 927 **/ 928 static inline int vti6_dev_init_gen(struct net_device *dev) 929 { 930 struct ip6_tnl *t = netdev_priv(dev); 931 932 t->dev = dev; 933 t->net = dev_net(dev); 934 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 935 if (!dev->tstats) 936 return -ENOMEM; 937 dev_hold(dev); 938 return 0; 939 } 940 941 /** 942 * vti6_dev_init - initializer for all non fallback tunnel devices 943 * @dev: virtual device associated with tunnel 944 **/ 945 static int vti6_dev_init(struct net_device *dev) 946 { 947 struct ip6_tnl *t = netdev_priv(dev); 948 int err = vti6_dev_init_gen(dev); 949 950 if (err) 951 return err; 952 vti6_link_config(t, true); 953 return 0; 954 } 955 956 /** 957 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 958 * @dev: fallback device 959 * 960 * Return: 0 961 **/ 962 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 963 { 964 struct ip6_tnl *t = netdev_priv(dev); 965 struct net *net = dev_net(dev); 966 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 967 968 t->parms.proto = IPPROTO_IPV6; 969 970 rcu_assign_pointer(ip6n->tnls_wc[0], t); 971 return 0; 972 } 973 974 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 975 struct netlink_ext_ack *extack) 976 { 977 return 0; 978 } 979 980 static void vti6_netlink_parms(struct nlattr *data[], 981 struct __ip6_tnl_parm *parms) 982 { 983 memset(parms, 0, sizeof(*parms)); 984 985 if (!data) 986 return; 987 988 if (data[IFLA_VTI_LINK]) 989 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 990 991 if (data[IFLA_VTI_LOCAL]) 992 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 993 994 if (data[IFLA_VTI_REMOTE]) 995 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 996 997 if (data[IFLA_VTI_IKEY]) 998 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 999 1000 if (data[IFLA_VTI_OKEY]) 1001 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 1002 1003 if (data[IFLA_VTI_FWMARK]) 1004 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 1005 } 1006 1007 static int vti6_newlink(struct net *src_net, struct net_device *dev, 1008 struct nlattr *tb[], struct nlattr *data[], 1009 struct netlink_ext_ack *extack) 1010 { 1011 struct net *net = dev_net(dev); 1012 struct ip6_tnl *nt; 1013 1014 nt = netdev_priv(dev); 1015 vti6_netlink_parms(data, &nt->parms); 1016 1017 nt->parms.proto = IPPROTO_IPV6; 1018 1019 if (vti6_locate(net, &nt->parms, 0)) 1020 return -EEXIST; 1021 1022 return vti6_tnl_create2(dev); 1023 } 1024 1025 static void vti6_dellink(struct net_device *dev, struct list_head *head) 1026 { 1027 struct net *net = dev_net(dev); 1028 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1029 1030 if (dev != ip6n->fb_tnl_dev) 1031 unregister_netdevice_queue(dev, head); 1032 } 1033 1034 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1035 struct nlattr *data[], 1036 struct netlink_ext_ack *extack) 1037 { 1038 struct ip6_tnl *t; 1039 struct __ip6_tnl_parm p; 1040 struct net *net = dev_net(dev); 1041 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1042 1043 if (dev == ip6n->fb_tnl_dev) 1044 return -EINVAL; 1045 1046 vti6_netlink_parms(data, &p); 1047 1048 t = vti6_locate(net, &p, 0); 1049 1050 if (t) { 1051 if (t->dev != dev) 1052 return -EEXIST; 1053 } else 1054 t = netdev_priv(dev); 1055 1056 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1057 } 1058 1059 static size_t vti6_get_size(const struct net_device *dev) 1060 { 1061 return 1062 /* IFLA_VTI_LINK */ 1063 nla_total_size(4) + 1064 /* IFLA_VTI_LOCAL */ 1065 nla_total_size(sizeof(struct in6_addr)) + 1066 /* IFLA_VTI_REMOTE */ 1067 nla_total_size(sizeof(struct in6_addr)) + 1068 /* IFLA_VTI_IKEY */ 1069 nla_total_size(4) + 1070 /* IFLA_VTI_OKEY */ 1071 nla_total_size(4) + 1072 /* IFLA_VTI_FWMARK */ 1073 nla_total_size(4) + 1074 0; 1075 } 1076 1077 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1078 { 1079 struct ip6_tnl *tunnel = netdev_priv(dev); 1080 struct __ip6_tnl_parm *parm = &tunnel->parms; 1081 1082 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1083 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1084 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1085 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1086 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1087 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1088 goto nla_put_failure; 1089 return 0; 1090 1091 nla_put_failure: 1092 return -EMSGSIZE; 1093 } 1094 1095 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1096 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1097 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1098 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1099 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1100 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1101 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1102 }; 1103 1104 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1105 .kind = "vti6", 1106 .maxtype = IFLA_VTI_MAX, 1107 .policy = vti6_policy, 1108 .priv_size = sizeof(struct ip6_tnl), 1109 .setup = vti6_dev_setup, 1110 .validate = vti6_validate, 1111 .newlink = vti6_newlink, 1112 .dellink = vti6_dellink, 1113 .changelink = vti6_changelink, 1114 .get_size = vti6_get_size, 1115 .fill_info = vti6_fill_info, 1116 .get_link_net = ip6_tnl_get_link_net, 1117 }; 1118 1119 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1120 struct list_head *list) 1121 { 1122 int h; 1123 struct ip6_tnl *t; 1124 1125 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1126 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1127 while (t) { 1128 unregister_netdevice_queue(t->dev, list); 1129 t = rtnl_dereference(t->next); 1130 } 1131 } 1132 1133 t = rtnl_dereference(ip6n->tnls_wc[0]); 1134 if (t) 1135 unregister_netdevice_queue(t->dev, list); 1136 } 1137 1138 static int __net_init vti6_init_net(struct net *net) 1139 { 1140 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1141 struct ip6_tnl *t = NULL; 1142 int err; 1143 1144 ip6n->tnls[0] = ip6n->tnls_wc; 1145 ip6n->tnls[1] = ip6n->tnls_r_l; 1146 1147 if (!net_has_fallback_tunnels(net)) 1148 return 0; 1149 err = -ENOMEM; 1150 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1151 NET_NAME_UNKNOWN, vti6_dev_setup); 1152 1153 if (!ip6n->fb_tnl_dev) 1154 goto err_alloc_dev; 1155 dev_net_set(ip6n->fb_tnl_dev, net); 1156 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1157 1158 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1159 if (err < 0) 1160 goto err_register; 1161 1162 err = register_netdev(ip6n->fb_tnl_dev); 1163 if (err < 0) 1164 goto err_register; 1165 1166 t = netdev_priv(ip6n->fb_tnl_dev); 1167 1168 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1169 return 0; 1170 1171 err_register: 1172 free_netdev(ip6n->fb_tnl_dev); 1173 err_alloc_dev: 1174 return err; 1175 } 1176 1177 static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1178 { 1179 struct vti6_net *ip6n; 1180 struct net *net; 1181 LIST_HEAD(list); 1182 1183 rtnl_lock(); 1184 list_for_each_entry(net, net_list, exit_list) { 1185 ip6n = net_generic(net, vti6_net_id); 1186 vti6_destroy_tunnels(ip6n, &list); 1187 } 1188 unregister_netdevice_many(&list); 1189 rtnl_unlock(); 1190 } 1191 1192 static struct pernet_operations vti6_net_ops = { 1193 .init = vti6_init_net, 1194 .exit_batch = vti6_exit_batch_net, 1195 .id = &vti6_net_id, 1196 .size = sizeof(struct vti6_net), 1197 }; 1198 1199 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1200 .handler = vti6_rcv, 1201 .input_handler = vti6_input_proto, 1202 .cb_handler = vti6_rcv_cb, 1203 .err_handler = vti6_err, 1204 .priority = 100, 1205 }; 1206 1207 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1208 .handler = vti6_rcv, 1209 .input_handler = vti6_input_proto, 1210 .cb_handler = vti6_rcv_cb, 1211 .err_handler = vti6_err, 1212 .priority = 100, 1213 }; 1214 1215 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1216 .handler = vti6_rcv, 1217 .input_handler = vti6_input_proto, 1218 .cb_handler = vti6_rcv_cb, 1219 .err_handler = vti6_err, 1220 .priority = 100, 1221 }; 1222 1223 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1224 static int vti6_rcv_tunnel(struct sk_buff *skb) 1225 { 1226 const xfrm_address_t *saddr; 1227 __be32 spi; 1228 1229 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1230 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1231 1232 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1233 } 1234 1235 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1236 .handler = vti6_rcv_tunnel, 1237 .cb_handler = vti6_rcv_cb, 1238 .err_handler = vti6_err, 1239 .priority = 0, 1240 }; 1241 1242 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1243 .handler = vti6_rcv_tunnel, 1244 .cb_handler = vti6_rcv_cb, 1245 .err_handler = vti6_err, 1246 .priority = 0, 1247 }; 1248 #endif 1249 1250 /** 1251 * vti6_tunnel_init - register protocol and reserve needed resources 1252 * 1253 * Return: 0 on success 1254 **/ 1255 static int __init vti6_tunnel_init(void) 1256 { 1257 const char *msg; 1258 int err; 1259 1260 msg = "tunnel device"; 1261 err = register_pernet_device(&vti6_net_ops); 1262 if (err < 0) 1263 goto pernet_dev_failed; 1264 1265 msg = "tunnel protocols"; 1266 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1267 if (err < 0) 1268 goto xfrm_proto_esp_failed; 1269 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1270 if (err < 0) 1271 goto xfrm_proto_ah_failed; 1272 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1273 if (err < 0) 1274 goto xfrm_proto_comp_failed; 1275 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1276 msg = "ipv6 tunnel"; 1277 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1278 if (err < 0) 1279 goto vti_tunnel_ipv6_failed; 1280 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1281 if (err < 0) 1282 goto vti_tunnel_ip6ip_failed; 1283 #endif 1284 1285 msg = "netlink interface"; 1286 err = rtnl_link_register(&vti6_link_ops); 1287 if (err < 0) 1288 goto rtnl_link_failed; 1289 1290 return 0; 1291 1292 rtnl_link_failed: 1293 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1294 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1295 vti_tunnel_ip6ip_failed: 1296 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1297 vti_tunnel_ipv6_failed: 1298 #endif 1299 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1300 xfrm_proto_comp_failed: 1301 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1302 xfrm_proto_ah_failed: 1303 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1304 xfrm_proto_esp_failed: 1305 unregister_pernet_device(&vti6_net_ops); 1306 pernet_dev_failed: 1307 pr_err("vti6 init: failed to register %s\n", msg); 1308 return err; 1309 } 1310 1311 /** 1312 * vti6_tunnel_cleanup - free resources and unregister protocol 1313 **/ 1314 static void __exit vti6_tunnel_cleanup(void) 1315 { 1316 rtnl_link_unregister(&vti6_link_ops); 1317 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1318 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1319 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1320 #endif 1321 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1322 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1323 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1324 unregister_pernet_device(&vti6_net_ops); 1325 } 1326 1327 module_init(vti6_tunnel_init); 1328 module_exit(vti6_tunnel_cleanup); 1329 MODULE_LICENSE("GPL"); 1330 MODULE_ALIAS_RTNL_LINK("vti6"); 1331 MODULE_ALIAS_NETDEV("ip6_vti0"); 1332 MODULE_AUTHOR("Steffen Klassert"); 1333 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1334