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 * @p: parameters containing tunnel end-points 129 * 130 * Description: 131 * vti6_tnl_bucket() returns the head of the list matching the 132 * &struct in6_addr entries laddr and raddr in @p. 133 * 134 * Return: head of IPv6 tunnel list 135 **/ 136 static struct ip6_tnl __rcu ** 137 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 138 { 139 const struct in6_addr *remote = &p->raddr; 140 const struct in6_addr *local = &p->laddr; 141 unsigned int h = 0; 142 int prio = 0; 143 144 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 145 prio = 1; 146 h = HASH(remote, local); 147 } 148 return &ip6n->tnls[prio][h]; 149 } 150 151 static void 152 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 153 { 154 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 155 156 rcu_assign_pointer(t->next , rtnl_dereference(*tp)); 157 rcu_assign_pointer(*tp, t); 158 } 159 160 static void 161 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 162 { 163 struct ip6_tnl __rcu **tp; 164 struct ip6_tnl *iter; 165 166 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 167 (iter = rtnl_dereference(*tp)) != NULL; 168 tp = &iter->next) { 169 if (t == iter) { 170 rcu_assign_pointer(*tp, t->next); 171 break; 172 } 173 } 174 } 175 176 static void vti6_dev_free(struct net_device *dev) 177 { 178 free_percpu(dev->tstats); 179 } 180 181 static int vti6_tnl_create2(struct net_device *dev) 182 { 183 struct ip6_tnl *t = netdev_priv(dev); 184 struct net *net = dev_net(dev); 185 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 186 int err; 187 188 dev->rtnl_link_ops = &vti6_link_ops; 189 err = register_netdevice(dev); 190 if (err < 0) 191 goto out; 192 193 strcpy(t->parms.name, dev->name); 194 195 dev_hold(dev); 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 queued; 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_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 524 } else { 525 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 526 htonl(mtu)); 527 } 528 529 err = -EMSGSIZE; 530 goto tx_err_dst_release; 531 } 532 533 queued: 534 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 535 skb_dst_set(skb, dst); 536 skb->dev = skb_dst(skb)->dev; 537 538 err = dst_output(t->net, skb->sk, skb); 539 if (net_xmit_eval(err) == 0) 540 err = pkt_len; 541 iptunnel_xmit_stats(dev, err); 542 543 return 0; 544 tx_err_link_failure: 545 stats->tx_carrier_errors++; 546 dst_link_failure(skb); 547 tx_err_dst_release: 548 dst_release(dst); 549 return err; 550 } 551 552 static netdev_tx_t 553 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 554 { 555 struct ip6_tnl *t = netdev_priv(dev); 556 struct net_device_stats *stats = &t->dev->stats; 557 struct flowi fl; 558 int ret; 559 560 if (!pskb_inet_may_pull(skb)) 561 goto tx_err; 562 563 memset(&fl, 0, sizeof(fl)); 564 565 switch (skb->protocol) { 566 case htons(ETH_P_IPV6): 567 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 568 vti6_addr_conflict(t, ipv6_hdr(skb))) 569 goto tx_err; 570 571 xfrm_decode_session(skb, &fl, AF_INET6); 572 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 573 break; 574 case htons(ETH_P_IP): 575 xfrm_decode_session(skb, &fl, AF_INET); 576 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 577 break; 578 default: 579 goto tx_err; 580 } 581 582 /* override mark with tunnel output key */ 583 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 584 585 ret = vti6_xmit(skb, dev, &fl); 586 if (ret < 0) 587 goto tx_err; 588 589 return NETDEV_TX_OK; 590 591 tx_err: 592 stats->tx_errors++; 593 stats->tx_dropped++; 594 kfree_skb(skb); 595 return NETDEV_TX_OK; 596 } 597 598 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 599 u8 type, u8 code, int offset, __be32 info) 600 { 601 __be32 spi; 602 __u32 mark; 603 struct xfrm_state *x; 604 struct ip6_tnl *t; 605 struct ip_esp_hdr *esph; 606 struct ip_auth_hdr *ah; 607 struct ip_comp_hdr *ipch; 608 struct net *net = dev_net(skb->dev); 609 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 610 int protocol = iph->nexthdr; 611 612 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 613 if (!t) 614 return -1; 615 616 mark = be32_to_cpu(t->parms.o_key); 617 618 switch (protocol) { 619 case IPPROTO_ESP: 620 esph = (struct ip_esp_hdr *)(skb->data + offset); 621 spi = esph->spi; 622 break; 623 case IPPROTO_AH: 624 ah = (struct ip_auth_hdr *)(skb->data + offset); 625 spi = ah->spi; 626 break; 627 case IPPROTO_COMP: 628 ipch = (struct ip_comp_hdr *)(skb->data + offset); 629 spi = htonl(ntohs(ipch->cpi)); 630 break; 631 default: 632 return 0; 633 } 634 635 if (type != ICMPV6_PKT_TOOBIG && 636 type != NDISC_REDIRECT) 637 return 0; 638 639 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 640 spi, protocol, AF_INET6); 641 if (!x) 642 return 0; 643 644 if (type == NDISC_REDIRECT) 645 ip6_redirect(skb, net, skb->dev->ifindex, 0, 646 sock_net_uid(net, NULL)); 647 else 648 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 649 xfrm_state_put(x); 650 651 return 0; 652 } 653 654 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 655 { 656 struct net_device *dev = t->dev; 657 struct __ip6_tnl_parm *p = &t->parms; 658 struct net_device *tdev = NULL; 659 int mtu; 660 661 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 662 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 663 664 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 665 IP6_TNL_F_CAP_PER_PACKET); 666 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 667 668 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 669 dev->flags |= IFF_POINTOPOINT; 670 else 671 dev->flags &= ~IFF_POINTOPOINT; 672 673 if (keep_mtu && dev->mtu) { 674 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu); 675 return; 676 } 677 678 if (p->flags & IP6_TNL_F_CAP_XMIT) { 679 int strict = (ipv6_addr_type(&p->raddr) & 680 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 681 struct rt6_info *rt = rt6_lookup(t->net, 682 &p->raddr, &p->laddr, 683 p->link, NULL, strict); 684 685 if (rt) 686 tdev = rt->dst.dev; 687 ip6_rt_put(rt); 688 } 689 690 if (!tdev && p->link) 691 tdev = __dev_get_by_index(t->net, p->link); 692 693 if (tdev) 694 mtu = tdev->mtu - sizeof(struct ipv6hdr); 695 else 696 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 697 698 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU); 699 } 700 701 /** 702 * vti6_tnl_change - update the tunnel parameters 703 * @t: tunnel to be changed 704 * @p: tunnel configuration parameters 705 * @keep_mtu: MTU was set from userspace, don't re-compute it 706 * 707 * Description: 708 * vti6_tnl_change() updates the tunnel parameters 709 **/ 710 static int 711 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 712 bool keep_mtu) 713 { 714 t->parms.laddr = p->laddr; 715 t->parms.raddr = p->raddr; 716 t->parms.link = p->link; 717 t->parms.i_key = p->i_key; 718 t->parms.o_key = p->o_key; 719 t->parms.proto = p->proto; 720 t->parms.fwmark = p->fwmark; 721 dst_cache_reset(&t->dst_cache); 722 vti6_link_config(t, keep_mtu); 723 return 0; 724 } 725 726 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 727 bool keep_mtu) 728 { 729 struct net *net = dev_net(t->dev); 730 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 731 int err; 732 733 vti6_tnl_unlink(ip6n, t); 734 synchronize_net(); 735 err = vti6_tnl_change(t, p, keep_mtu); 736 vti6_tnl_link(ip6n, t); 737 netdev_state_change(t->dev); 738 return err; 739 } 740 741 static void 742 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 743 { 744 p->laddr = u->laddr; 745 p->raddr = u->raddr; 746 p->link = u->link; 747 p->i_key = u->i_key; 748 p->o_key = u->o_key; 749 p->proto = u->proto; 750 751 memcpy(p->name, u->name, sizeof(u->name)); 752 } 753 754 static void 755 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 756 { 757 u->laddr = p->laddr; 758 u->raddr = p->raddr; 759 u->link = p->link; 760 u->i_key = p->i_key; 761 u->o_key = p->o_key; 762 if (u->i_key) 763 u->i_flags |= GRE_KEY; 764 if (u->o_key) 765 u->o_flags |= GRE_KEY; 766 u->proto = p->proto; 767 768 memcpy(u->name, p->name, sizeof(u->name)); 769 } 770 771 /** 772 * vti6_ioctl - configure vti6 tunnels from userspace 773 * @dev: virtual device associated with tunnel 774 * @ifr: parameters passed from userspace 775 * @cmd: command to be performed 776 * 777 * Description: 778 * vti6_ioctl() is used for managing vti6 tunnels 779 * from userspace. 780 * 781 * The possible commands are the following: 782 * %SIOCGETTUNNEL: get tunnel parameters for device 783 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 784 * %SIOCCHGTUNNEL: change tunnel parameters to those given 785 * %SIOCDELTUNNEL: delete tunnel 786 * 787 * The fallback device "ip6_vti0", created during module 788 * initialization, can be used for creating other tunnel devices. 789 * 790 * Return: 791 * 0 on success, 792 * %-EFAULT if unable to copy data to or from userspace, 793 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 794 * %-EINVAL if passed tunnel parameters are invalid, 795 * %-EEXIST if changing a tunnel's parameters would cause a conflict 796 * %-ENODEV if attempting to change or delete a nonexisting device 797 **/ 798 static int 799 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 800 { 801 int err = 0; 802 struct ip6_tnl_parm2 p; 803 struct __ip6_tnl_parm p1; 804 struct ip6_tnl *t = NULL; 805 struct net *net = dev_net(dev); 806 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 807 808 switch (cmd) { 809 case SIOCGETTUNNEL: 810 if (dev == ip6n->fb_tnl_dev) { 811 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 812 err = -EFAULT; 813 break; 814 } 815 vti6_parm_from_user(&p1, &p); 816 t = vti6_locate(net, &p1, 0); 817 } else { 818 memset(&p, 0, sizeof(p)); 819 } 820 if (!t) 821 t = netdev_priv(dev); 822 vti6_parm_to_user(&p, &t->parms); 823 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 824 err = -EFAULT; 825 break; 826 case SIOCADDTUNNEL: 827 case SIOCCHGTUNNEL: 828 err = -EPERM; 829 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 830 break; 831 err = -EFAULT; 832 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 833 break; 834 err = -EINVAL; 835 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 836 break; 837 vti6_parm_from_user(&p1, &p); 838 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 839 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 840 if (t) { 841 if (t->dev != dev) { 842 err = -EEXIST; 843 break; 844 } 845 } else 846 t = netdev_priv(dev); 847 848 err = vti6_update(t, &p1, false); 849 } 850 if (t) { 851 err = 0; 852 vti6_parm_to_user(&p, &t->parms); 853 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 854 err = -EFAULT; 855 856 } else 857 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 858 break; 859 case SIOCDELTUNNEL: 860 err = -EPERM; 861 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 862 break; 863 864 if (dev == ip6n->fb_tnl_dev) { 865 err = -EFAULT; 866 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 867 break; 868 err = -ENOENT; 869 vti6_parm_from_user(&p1, &p); 870 t = vti6_locate(net, &p1, 0); 871 if (!t) 872 break; 873 err = -EPERM; 874 if (t->dev == ip6n->fb_tnl_dev) 875 break; 876 dev = t->dev; 877 } 878 err = 0; 879 unregister_netdevice(dev); 880 break; 881 default: 882 err = -EINVAL; 883 } 884 return err; 885 } 886 887 static const struct net_device_ops vti6_netdev_ops = { 888 .ndo_init = vti6_dev_init, 889 .ndo_uninit = vti6_dev_uninit, 890 .ndo_start_xmit = vti6_tnl_xmit, 891 .ndo_do_ioctl = vti6_ioctl, 892 .ndo_get_stats64 = ip_tunnel_get_stats64, 893 .ndo_get_iflink = ip6_tnl_get_iflink, 894 }; 895 896 /** 897 * vti6_dev_setup - setup virtual tunnel device 898 * @dev: virtual device associated with tunnel 899 * 900 * Description: 901 * Initialize function pointers and device parameters 902 **/ 903 static void vti6_dev_setup(struct net_device *dev) 904 { 905 dev->netdev_ops = &vti6_netdev_ops; 906 dev->header_ops = &ip_tunnel_header_ops; 907 dev->needs_free_netdev = true; 908 dev->priv_destructor = vti6_dev_free; 909 910 dev->type = ARPHRD_TUNNEL6; 911 dev->min_mtu = IPV4_MIN_MTU; 912 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 913 dev->flags |= IFF_NOARP; 914 dev->addr_len = sizeof(struct in6_addr); 915 netif_keep_dst(dev); 916 /* This perm addr will be used as interface identifier by IPv6 */ 917 dev->addr_assign_type = NET_ADDR_RANDOM; 918 eth_random_addr(dev->perm_addr); 919 } 920 921 /** 922 * vti6_dev_init_gen - general initializer for all tunnel devices 923 * @dev: virtual device associated with tunnel 924 **/ 925 static inline int vti6_dev_init_gen(struct net_device *dev) 926 { 927 struct ip6_tnl *t = netdev_priv(dev); 928 929 t->dev = dev; 930 t->net = dev_net(dev); 931 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 932 if (!dev->tstats) 933 return -ENOMEM; 934 return 0; 935 } 936 937 /** 938 * vti6_dev_init - initializer for all non fallback tunnel devices 939 * @dev: virtual device associated with tunnel 940 **/ 941 static int vti6_dev_init(struct net_device *dev) 942 { 943 struct ip6_tnl *t = netdev_priv(dev); 944 int err = vti6_dev_init_gen(dev); 945 946 if (err) 947 return err; 948 vti6_link_config(t, true); 949 return 0; 950 } 951 952 /** 953 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 954 * @dev: fallback device 955 * 956 * Return: 0 957 **/ 958 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 959 { 960 struct ip6_tnl *t = netdev_priv(dev); 961 struct net *net = dev_net(dev); 962 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 963 964 t->parms.proto = IPPROTO_IPV6; 965 dev_hold(dev); 966 967 rcu_assign_pointer(ip6n->tnls_wc[0], t); 968 return 0; 969 } 970 971 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 972 struct netlink_ext_ack *extack) 973 { 974 return 0; 975 } 976 977 static void vti6_netlink_parms(struct nlattr *data[], 978 struct __ip6_tnl_parm *parms) 979 { 980 memset(parms, 0, sizeof(*parms)); 981 982 if (!data) 983 return; 984 985 if (data[IFLA_VTI_LINK]) 986 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 987 988 if (data[IFLA_VTI_LOCAL]) 989 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 990 991 if (data[IFLA_VTI_REMOTE]) 992 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 993 994 if (data[IFLA_VTI_IKEY]) 995 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 996 997 if (data[IFLA_VTI_OKEY]) 998 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 999 1000 if (data[IFLA_VTI_FWMARK]) 1001 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 1002 } 1003 1004 static int vti6_newlink(struct net *src_net, struct net_device *dev, 1005 struct nlattr *tb[], struct nlattr *data[], 1006 struct netlink_ext_ack *extack) 1007 { 1008 struct net *net = dev_net(dev); 1009 struct ip6_tnl *nt; 1010 1011 nt = netdev_priv(dev); 1012 vti6_netlink_parms(data, &nt->parms); 1013 1014 nt->parms.proto = IPPROTO_IPV6; 1015 1016 if (vti6_locate(net, &nt->parms, 0)) 1017 return -EEXIST; 1018 1019 return vti6_tnl_create2(dev); 1020 } 1021 1022 static void vti6_dellink(struct net_device *dev, struct list_head *head) 1023 { 1024 struct net *net = dev_net(dev); 1025 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1026 1027 if (dev != ip6n->fb_tnl_dev) 1028 unregister_netdevice_queue(dev, head); 1029 } 1030 1031 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1032 struct nlattr *data[], 1033 struct netlink_ext_ack *extack) 1034 { 1035 struct ip6_tnl *t; 1036 struct __ip6_tnl_parm p; 1037 struct net *net = dev_net(dev); 1038 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1039 1040 if (dev == ip6n->fb_tnl_dev) 1041 return -EINVAL; 1042 1043 vti6_netlink_parms(data, &p); 1044 1045 t = vti6_locate(net, &p, 0); 1046 1047 if (t) { 1048 if (t->dev != dev) 1049 return -EEXIST; 1050 } else 1051 t = netdev_priv(dev); 1052 1053 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1054 } 1055 1056 static size_t vti6_get_size(const struct net_device *dev) 1057 { 1058 return 1059 /* IFLA_VTI_LINK */ 1060 nla_total_size(4) + 1061 /* IFLA_VTI_LOCAL */ 1062 nla_total_size(sizeof(struct in6_addr)) + 1063 /* IFLA_VTI_REMOTE */ 1064 nla_total_size(sizeof(struct in6_addr)) + 1065 /* IFLA_VTI_IKEY */ 1066 nla_total_size(4) + 1067 /* IFLA_VTI_OKEY */ 1068 nla_total_size(4) + 1069 /* IFLA_VTI_FWMARK */ 1070 nla_total_size(4) + 1071 0; 1072 } 1073 1074 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1075 { 1076 struct ip6_tnl *tunnel = netdev_priv(dev); 1077 struct __ip6_tnl_parm *parm = &tunnel->parms; 1078 1079 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1080 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1081 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1082 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1083 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1084 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1085 goto nla_put_failure; 1086 return 0; 1087 1088 nla_put_failure: 1089 return -EMSGSIZE; 1090 } 1091 1092 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1093 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1094 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1095 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1096 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1097 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1098 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1099 }; 1100 1101 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1102 .kind = "vti6", 1103 .maxtype = IFLA_VTI_MAX, 1104 .policy = vti6_policy, 1105 .priv_size = sizeof(struct ip6_tnl), 1106 .setup = vti6_dev_setup, 1107 .validate = vti6_validate, 1108 .newlink = vti6_newlink, 1109 .dellink = vti6_dellink, 1110 .changelink = vti6_changelink, 1111 .get_size = vti6_get_size, 1112 .fill_info = vti6_fill_info, 1113 .get_link_net = ip6_tnl_get_link_net, 1114 }; 1115 1116 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1117 struct list_head *list) 1118 { 1119 int h; 1120 struct ip6_tnl *t; 1121 1122 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1123 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1124 while (t) { 1125 unregister_netdevice_queue(t->dev, list); 1126 t = rtnl_dereference(t->next); 1127 } 1128 } 1129 1130 t = rtnl_dereference(ip6n->tnls_wc[0]); 1131 if (t) 1132 unregister_netdevice_queue(t->dev, list); 1133 } 1134 1135 static int __net_init vti6_init_net(struct net *net) 1136 { 1137 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1138 struct ip6_tnl *t = NULL; 1139 int err; 1140 1141 ip6n->tnls[0] = ip6n->tnls_wc; 1142 ip6n->tnls[1] = ip6n->tnls_r_l; 1143 1144 if (!net_has_fallback_tunnels(net)) 1145 return 0; 1146 err = -ENOMEM; 1147 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1148 NET_NAME_UNKNOWN, vti6_dev_setup); 1149 1150 if (!ip6n->fb_tnl_dev) 1151 goto err_alloc_dev; 1152 dev_net_set(ip6n->fb_tnl_dev, net); 1153 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1154 1155 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1156 if (err < 0) 1157 goto err_register; 1158 1159 err = register_netdev(ip6n->fb_tnl_dev); 1160 if (err < 0) 1161 goto err_register; 1162 1163 t = netdev_priv(ip6n->fb_tnl_dev); 1164 1165 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1166 return 0; 1167 1168 err_register: 1169 free_netdev(ip6n->fb_tnl_dev); 1170 err_alloc_dev: 1171 return err; 1172 } 1173 1174 static void __net_exit vti6_exit_batch_net(struct list_head *net_list) 1175 { 1176 struct vti6_net *ip6n; 1177 struct net *net; 1178 LIST_HEAD(list); 1179 1180 rtnl_lock(); 1181 list_for_each_entry(net, net_list, exit_list) { 1182 ip6n = net_generic(net, vti6_net_id); 1183 vti6_destroy_tunnels(ip6n, &list); 1184 } 1185 unregister_netdevice_many(&list); 1186 rtnl_unlock(); 1187 } 1188 1189 static struct pernet_operations vti6_net_ops = { 1190 .init = vti6_init_net, 1191 .exit_batch = vti6_exit_batch_net, 1192 .id = &vti6_net_id, 1193 .size = sizeof(struct vti6_net), 1194 }; 1195 1196 static struct xfrm6_protocol vti_esp6_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_ah6_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 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1213 .handler = vti6_rcv, 1214 .input_handler = vti6_input_proto, 1215 .cb_handler = vti6_rcv_cb, 1216 .err_handler = vti6_err, 1217 .priority = 100, 1218 }; 1219 1220 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1221 static int vti6_rcv_tunnel(struct sk_buff *skb) 1222 { 1223 const xfrm_address_t *saddr; 1224 __be32 spi; 1225 1226 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1227 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1228 1229 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1230 } 1231 1232 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1233 .handler = vti6_rcv_tunnel, 1234 .cb_handler = vti6_rcv_cb, 1235 .err_handler = vti6_err, 1236 .priority = 0, 1237 }; 1238 1239 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1240 .handler = vti6_rcv_tunnel, 1241 .cb_handler = vti6_rcv_cb, 1242 .err_handler = vti6_err, 1243 .priority = 0, 1244 }; 1245 #endif 1246 1247 /** 1248 * vti6_tunnel_init - register protocol and reserve needed resources 1249 * 1250 * Return: 0 on success 1251 **/ 1252 static int __init vti6_tunnel_init(void) 1253 { 1254 const char *msg; 1255 int err; 1256 1257 msg = "tunnel device"; 1258 err = register_pernet_device(&vti6_net_ops); 1259 if (err < 0) 1260 goto pernet_dev_failed; 1261 1262 msg = "tunnel protocols"; 1263 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1264 if (err < 0) 1265 goto xfrm_proto_esp_failed; 1266 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1267 if (err < 0) 1268 goto xfrm_proto_ah_failed; 1269 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1270 if (err < 0) 1271 goto xfrm_proto_comp_failed; 1272 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1273 msg = "ipv6 tunnel"; 1274 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1275 if (err < 0) 1276 goto vti_tunnel_ipv6_failed; 1277 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1278 if (err < 0) 1279 goto vti_tunnel_ip6ip_failed; 1280 #endif 1281 1282 msg = "netlink interface"; 1283 err = rtnl_link_register(&vti6_link_ops); 1284 if (err < 0) 1285 goto rtnl_link_failed; 1286 1287 return 0; 1288 1289 rtnl_link_failed: 1290 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1291 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1292 vti_tunnel_ip6ip_failed: 1293 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1294 vti_tunnel_ipv6_failed: 1295 #endif 1296 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1297 xfrm_proto_comp_failed: 1298 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1299 xfrm_proto_ah_failed: 1300 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1301 xfrm_proto_esp_failed: 1302 unregister_pernet_device(&vti6_net_ops); 1303 pernet_dev_failed: 1304 pr_err("vti6 init: failed to register %s\n", msg); 1305 return err; 1306 } 1307 1308 /** 1309 * vti6_tunnel_cleanup - free resources and unregister protocol 1310 **/ 1311 static void __exit vti6_tunnel_cleanup(void) 1312 { 1313 rtnl_link_unregister(&vti6_link_ops); 1314 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1315 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1316 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1317 #endif 1318 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1319 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1320 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1321 unregister_pernet_device(&vti6_net_ops); 1322 } 1323 1324 module_init(vti6_tunnel_init); 1325 module_exit(vti6_tunnel_cleanup); 1326 MODULE_LICENSE("GPL"); 1327 MODULE_ALIAS_RTNL_LINK("vti6"); 1328 MODULE_ALIAS_NETDEV("ip6_vti0"); 1329 MODULE_AUTHOR("Steffen Klassert"); 1330 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1331