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 <net/netdev_lock.h> 49 #include <linux/etherdevice.h> 50 51 #define IP6_VTI_HASH_SIZE_SHIFT 5 52 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT) 53 54 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2) 55 { 56 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2); 57 58 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT); 59 } 60 61 static int vti6_dev_init(struct net_device *dev); 62 static void vti6_dev_setup(struct net_device *dev); 63 static struct rtnl_link_ops vti6_link_ops __read_mostly; 64 65 static unsigned int vti6_net_id __read_mostly; 66 struct vti6_net { 67 /* the vti6 tunnel fallback device */ 68 struct net_device *fb_tnl_dev; 69 /* lists for storing tunnels in use */ 70 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE]; 71 struct ip6_tnl __rcu *tnls_wc[1]; 72 struct ip6_tnl __rcu **tnls[2]; 73 }; 74 75 #define for_each_vti6_tunnel_rcu(start) \ 76 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) 77 78 /** 79 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses 80 * @net: network namespace 81 * @remote: the address of the tunnel exit-point 82 * @local: the address of the tunnel entry-point 83 * 84 * Return: 85 * tunnel matching given end-points if found, 86 * else fallback tunnel if its device is up, 87 * else %NULL 88 **/ 89 static struct ip6_tnl * 90 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote, 91 const struct in6_addr *local) 92 { 93 unsigned int hash = HASH(remote, local); 94 struct ip6_tnl *t; 95 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 96 struct in6_addr any; 97 98 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 99 if (ipv6_addr_equal(local, &t->parms.laddr) && 100 ipv6_addr_equal(remote, &t->parms.raddr) && 101 (t->dev->flags & IFF_UP)) 102 return t; 103 } 104 105 memset(&any, 0, sizeof(any)); 106 hash = HASH(&any, local); 107 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 108 if (ipv6_addr_equal(local, &t->parms.laddr) && 109 (t->dev->flags & IFF_UP)) 110 return t; 111 } 112 113 hash = HASH(remote, &any); 114 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) { 115 if (ipv6_addr_equal(remote, &t->parms.raddr) && 116 (t->dev->flags & IFF_UP)) 117 return t; 118 } 119 120 t = rcu_dereference(ip6n->tnls_wc[0]); 121 if (t && (t->dev->flags & IFF_UP)) 122 return t; 123 124 return NULL; 125 } 126 127 /** 128 * vti6_tnl_bucket - get head of list matching given tunnel parameters 129 * @ip6n: the private data for ip6_vti in the netns 130 * @p: parameters containing tunnel end-points 131 * 132 * Description: 133 * vti6_tnl_bucket() returns the head of the list matching the 134 * &struct in6_addr entries laddr and raddr in @p. 135 * 136 * Return: head of IPv6 tunnel list 137 **/ 138 static struct ip6_tnl __rcu ** 139 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p) 140 { 141 const struct in6_addr *remote = &p->raddr; 142 const struct in6_addr *local = &p->laddr; 143 unsigned int h = 0; 144 int prio = 0; 145 146 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) { 147 prio = 1; 148 h = HASH(remote, local); 149 } 150 return &ip6n->tnls[prio][h]; 151 } 152 153 static void 154 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t) 155 { 156 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms); 157 158 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 159 rcu_assign_pointer(*tp, t); 160 } 161 162 static void 163 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t) 164 { 165 struct ip6_tnl __rcu **tp; 166 struct ip6_tnl *iter; 167 168 for (tp = vti6_tnl_bucket(ip6n, &t->parms); 169 (iter = rtnl_dereference(*tp)) != NULL; 170 tp = &iter->next) { 171 if (t == iter) { 172 rcu_assign_pointer(*tp, t->next); 173 break; 174 } 175 } 176 } 177 178 static int vti6_tnl_create2(struct net_device *dev) 179 { 180 struct ip6_tnl *t = netdev_priv(dev); 181 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 182 int err; 183 184 dev->rtnl_link_ops = &vti6_link_ops; 185 err = register_netdevice(dev); 186 if (err < 0) 187 goto out; 188 189 strcpy(t->parms.name, dev->name); 190 191 vti6_tnl_link(ip6n, t); 192 193 return 0; 194 195 out: 196 return err; 197 } 198 199 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p) 200 { 201 struct net_device *dev; 202 struct ip6_tnl *t; 203 char name[IFNAMSIZ]; 204 int err; 205 206 if (p->name[0]) { 207 if (!dev_valid_name(p->name)) 208 goto failed; 209 strscpy(name, p->name, IFNAMSIZ); 210 } else { 211 sprintf(name, "ip6_vti%%d"); 212 } 213 214 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup); 215 if (!dev) 216 goto failed; 217 218 dev_net_set(dev, net); 219 220 t = netdev_priv(dev); 221 t->parms = *p; 222 t->net = dev_net(dev); 223 224 err = vti6_tnl_create2(dev); 225 if (err < 0) 226 goto failed_free; 227 228 return t; 229 230 failed_free: 231 free_netdev(dev); 232 failed: 233 return NULL; 234 } 235 236 /** 237 * vti6_locate - find or create tunnel matching given parameters 238 * @net: network namespace 239 * @p: tunnel parameters 240 * @create: != 0 if allowed to create new tunnel if no match found 241 * 242 * Description: 243 * vti6_locate() first tries to locate an existing tunnel 244 * based on @parms. If this is unsuccessful, but @create is set a new 245 * tunnel device is created and registered for use. 246 * 247 * Return: 248 * matching tunnel or NULL 249 **/ 250 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p, 251 int create) 252 { 253 const struct in6_addr *remote = &p->raddr; 254 const struct in6_addr *local = &p->laddr; 255 struct ip6_tnl __rcu **tp; 256 struct ip6_tnl *t; 257 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 258 259 for (tp = vti6_tnl_bucket(ip6n, p); 260 (t = rtnl_dereference(*tp)) != NULL; 261 tp = &t->next) { 262 if (ipv6_addr_equal(local, &t->parms.laddr) && 263 ipv6_addr_equal(remote, &t->parms.raddr)) { 264 if (create) 265 return NULL; 266 267 return t; 268 } 269 } 270 if (!create) 271 return NULL; 272 return vti6_tnl_create(net, p); 273 } 274 275 /** 276 * vti6_dev_uninit - tunnel device uninitializer 277 * @dev: the device to be destroyed 278 * 279 * Description: 280 * vti6_dev_uninit() removes tunnel from its list 281 **/ 282 static void vti6_dev_uninit(struct net_device *dev) 283 { 284 struct ip6_tnl *t = netdev_priv(dev); 285 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id); 286 287 if (dev == ip6n->fb_tnl_dev) 288 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL); 289 else 290 vti6_tnl_unlink(ip6n, t); 291 netdev_put(dev, &t->dev_tracker); 292 } 293 294 static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 295 int encap_type) 296 { 297 struct ip6_tnl *t; 298 const struct ipv6hdr *ipv6h = ipv6_hdr(skb); 299 300 rcu_read_lock(); 301 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr); 302 if (t) { 303 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) { 304 rcu_read_unlock(); 305 goto discard; 306 } 307 308 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 309 rcu_read_unlock(); 310 goto discard; 311 } 312 313 ipv6h = ipv6_hdr(skb); 314 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) { 315 DEV_STATS_INC(t->dev, rx_dropped); 316 rcu_read_unlock(); 317 goto discard; 318 } 319 320 rcu_read_unlock(); 321 322 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t; 323 XFRM_SPI_SKB_CB(skb)->family = AF_INET6; 324 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr); 325 return xfrm_input(skb, nexthdr, spi, encap_type); 326 } 327 rcu_read_unlock(); 328 return -EINVAL; 329 discard: 330 kfree_skb(skb); 331 return 0; 332 } 333 334 static int vti6_rcv(struct sk_buff *skb) 335 { 336 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff]; 337 338 return vti6_input_proto(skb, nexthdr, 0, 0); 339 } 340 341 static int vti6_rcv_cb(struct sk_buff *skb, int err) 342 { 343 unsigned short family; 344 struct net_device *dev; 345 struct xfrm_state *x; 346 const struct xfrm_mode *inner_mode; 347 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6; 348 u32 orig_mark = skb->mark; 349 int ret; 350 351 if (!t) 352 return 1; 353 354 dev = t->dev; 355 356 if (err) { 357 DEV_STATS_INC(dev, rx_errors); 358 DEV_STATS_INC(dev, rx_dropped); 359 360 return 0; 361 } 362 363 x = xfrm_input_state(skb); 364 365 inner_mode = &x->inner_mode; 366 367 if (x->sel.family == AF_UNSPEC) { 368 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 369 if (inner_mode == NULL) { 370 XFRM_INC_STATS(dev_net(skb->dev), 371 LINUX_MIB_XFRMINSTATEMODEERROR); 372 return -EINVAL; 373 } 374 } 375 376 family = inner_mode->family; 377 378 skb->mark = be32_to_cpu(t->parms.i_key); 379 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 380 skb->mark = orig_mark; 381 382 if (!ret) 383 return -EPERM; 384 385 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev))); 386 skb->dev = dev; 387 dev_sw_netstats_rx_add(dev, skb->len); 388 389 return 0; 390 } 391 392 /** 393 * vti6_addr_conflict - compare packet addresses to tunnel's own 394 * @t: the outgoing tunnel device 395 * @hdr: IPv6 header from the incoming packet 396 * 397 * Description: 398 * Avoid trivial tunneling loop by checking that tunnel exit-point 399 * doesn't match source of incoming packet. 400 * 401 * Return: 402 * 1 if conflict, 403 * 0 else 404 **/ 405 static inline bool 406 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 407 { 408 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 409 } 410 411 static bool vti6_state_check(const struct xfrm_state *x, 412 const struct in6_addr *dst, 413 const struct in6_addr *src) 414 { 415 xfrm_address_t *daddr = (xfrm_address_t *)dst; 416 xfrm_address_t *saddr = (xfrm_address_t *)src; 417 418 /* if there is no transform then this tunnel is not functional. 419 * Or if the xfrm is not mode tunnel. 420 */ 421 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 422 x->props.family != AF_INET6) 423 return false; 424 425 if (ipv6_addr_any(dst)) 426 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6); 427 428 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6)) 429 return false; 430 431 return true; 432 } 433 434 /** 435 * vti6_xmit - send a packet 436 * @skb: the outgoing socket buffer 437 * @dev: the outgoing tunnel device 438 * @fl: the flow informations for the xfrm_lookup 439 **/ 440 static int 441 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 442 { 443 struct ip6_tnl *t = netdev_priv(dev); 444 struct dst_entry *dst = skb_dst(skb); 445 struct net_device *tdev; 446 struct xfrm_state *x; 447 int pkt_len = skb->len; 448 int err = -1; 449 int mtu; 450 451 if (!dst) { 452 switch (skb->protocol) { 453 case htons(ETH_P_IP): { 454 struct rtable *rt; 455 456 fl->u.ip4.flowi4_oif = dev->ifindex; 457 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 458 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 459 if (IS_ERR(rt)) 460 goto tx_err_link_failure; 461 dst = &rt->dst; 462 skb_dst_set(skb, dst); 463 break; 464 } 465 case htons(ETH_P_IPV6): 466 fl->u.ip6.flowi6_oif = dev->ifindex; 467 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 468 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 469 if (dst->error) { 470 dst_release(dst); 471 dst = NULL; 472 goto tx_err_link_failure; 473 } 474 skb_dst_set(skb, dst); 475 break; 476 default: 477 goto tx_err_link_failure; 478 } 479 } 480 481 dst_hold(dst); 482 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 483 if (IS_ERR(dst)) { 484 err = PTR_ERR(dst); 485 dst = NULL; 486 goto tx_err_link_failure; 487 } 488 489 if (dst->flags & DST_XFRM_QUEUE) 490 goto xmit; 491 492 x = dst->xfrm; 493 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 494 goto tx_err_link_failure; 495 496 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 497 (const struct in6_addr *)&x->id.daddr)) 498 goto tx_err_link_failure; 499 500 tdev = dst->dev; 501 502 if (tdev == dev) { 503 DEV_STATS_INC(dev, collisions); 504 net_warn_ratelimited("%s: Local routing loop detected!\n", 505 t->parms.name); 506 goto tx_err_dst_release; 507 } 508 509 mtu = dst_mtu(dst); 510 if (skb->len > mtu) { 511 skb_dst_update_pmtu_no_confirm(skb, mtu); 512 513 if (skb->protocol == htons(ETH_P_IPV6)) { 514 if (mtu < IPV6_MIN_MTU) 515 mtu = IPV6_MIN_MTU; 516 517 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 518 } else { 519 if (!(ip_hdr(skb)->frag_off & htons(IP_DF))) 520 goto xmit; 521 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 522 htonl(mtu)); 523 } 524 525 err = -EMSGSIZE; 526 goto tx_err_dst_release; 527 } 528 529 xmit: 530 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 531 skb_dst_set(skb, dst); 532 skb->dev = skb_dst(skb)->dev; 533 534 err = dst_output(t->net, skb->sk, skb); 535 if (net_xmit_eval(err) == 0) 536 err = pkt_len; 537 iptunnel_xmit_stats(dev, err); 538 539 return 0; 540 tx_err_link_failure: 541 DEV_STATS_INC(dev, tx_carrier_errors); 542 dst_link_failure(skb); 543 tx_err_dst_release: 544 dst_release(dst); 545 return err; 546 } 547 548 static netdev_tx_t 549 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 550 { 551 struct ip6_tnl *t = netdev_priv(dev); 552 struct flowi fl; 553 int ret; 554 555 if (!pskb_inet_may_pull(skb)) 556 goto tx_err; 557 558 memset(&fl, 0, sizeof(fl)); 559 560 switch (skb->protocol) { 561 case htons(ETH_P_IPV6): 562 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) || 563 vti6_addr_conflict(t, ipv6_hdr(skb))) 564 goto tx_err; 565 566 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 567 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6); 568 break; 569 case htons(ETH_P_IP): 570 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 571 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET); 572 break; 573 default: 574 goto tx_err; 575 } 576 577 /* override mark with tunnel output key */ 578 fl.flowi_mark = be32_to_cpu(t->parms.o_key); 579 580 ret = vti6_xmit(skb, dev, &fl); 581 if (ret < 0) 582 goto tx_err; 583 584 return NETDEV_TX_OK; 585 586 tx_err: 587 DEV_STATS_INC(dev, tx_errors); 588 DEV_STATS_INC(dev, tx_dropped); 589 kfree_skb(skb); 590 return NETDEV_TX_OK; 591 } 592 593 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 594 u8 type, u8 code, int offset, __be32 info) 595 { 596 __be32 spi; 597 __u32 mark; 598 struct xfrm_state *x; 599 struct ip6_tnl *t; 600 struct ip_esp_hdr *esph; 601 struct ip_auth_hdr *ah; 602 struct ip_comp_hdr *ipch; 603 struct net *net = dev_net(skb->dev); 604 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data; 605 int protocol = iph->nexthdr; 606 607 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr); 608 if (!t) 609 return -1; 610 611 mark = be32_to_cpu(t->parms.o_key); 612 613 switch (protocol) { 614 case IPPROTO_ESP: 615 esph = (struct ip_esp_hdr *)(skb->data + offset); 616 spi = esph->spi; 617 break; 618 case IPPROTO_AH: 619 ah = (struct ip_auth_hdr *)(skb->data + offset); 620 spi = ah->spi; 621 break; 622 case IPPROTO_COMP: 623 ipch = (struct ip_comp_hdr *)(skb->data + offset); 624 spi = htonl(ntohs(ipch->cpi)); 625 break; 626 default: 627 return 0; 628 } 629 630 if (type != ICMPV6_PKT_TOOBIG && 631 type != NDISC_REDIRECT) 632 return 0; 633 634 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 635 spi, protocol, AF_INET6); 636 if (!x) 637 return 0; 638 639 if (type == NDISC_REDIRECT) 640 ip6_redirect(skb, net, skb->dev->ifindex, 0, 641 sock_net_uid(net, NULL)); 642 else 643 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 644 xfrm_state_put(x); 645 646 return 0; 647 } 648 649 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu) 650 { 651 struct net_device *dev = t->dev; 652 struct __ip6_tnl_parm *p = &t->parms; 653 struct net_device *tdev = NULL; 654 int mtu; 655 656 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 657 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 658 659 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV | 660 IP6_TNL_F_CAP_PER_PACKET); 661 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 662 663 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV) 664 dev->flags |= IFF_POINTOPOINT; 665 else 666 dev->flags &= ~IFF_POINTOPOINT; 667 668 if (keep_mtu && dev->mtu) { 669 WRITE_ONCE(dev->mtu, 670 clamp(dev->mtu, dev->min_mtu, dev->max_mtu)); 671 return; 672 } 673 674 if (p->flags & IP6_TNL_F_CAP_XMIT) { 675 int strict = (ipv6_addr_type(&p->raddr) & 676 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)); 677 struct rt6_info *rt = rt6_lookup(t->net, 678 &p->raddr, &p->laddr, 679 p->link, NULL, strict); 680 681 if (rt) 682 tdev = rt->dst.dev; 683 ip6_rt_put(rt); 684 } 685 686 if (!tdev && p->link) 687 tdev = __dev_get_by_index(t->net, p->link); 688 689 if (tdev) 690 mtu = tdev->mtu - sizeof(struct ipv6hdr); 691 else 692 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr); 693 694 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU); 695 } 696 697 /** 698 * vti6_tnl_change - update the tunnel parameters 699 * @t: tunnel to be changed 700 * @p: tunnel configuration parameters 701 * @keep_mtu: MTU was set from userspace, don't re-compute it 702 * 703 * Description: 704 * vti6_tnl_change() updates the tunnel parameters 705 **/ 706 static int 707 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 708 bool keep_mtu) 709 { 710 t->parms.laddr = p->laddr; 711 t->parms.raddr = p->raddr; 712 t->parms.link = p->link; 713 t->parms.i_key = p->i_key; 714 t->parms.o_key = p->o_key; 715 t->parms.proto = p->proto; 716 t->parms.fwmark = p->fwmark; 717 dst_cache_reset(&t->dst_cache); 718 vti6_link_config(t, keep_mtu); 719 return 0; 720 } 721 722 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p, 723 bool keep_mtu) 724 { 725 struct net *net = dev_net(t->dev); 726 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 727 int err; 728 729 vti6_tnl_unlink(ip6n, t); 730 synchronize_net(); 731 err = vti6_tnl_change(t, p, keep_mtu); 732 vti6_tnl_link(ip6n, t); 733 netdev_state_change(t->dev); 734 return err; 735 } 736 737 static void 738 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u) 739 { 740 p->laddr = u->laddr; 741 p->raddr = u->raddr; 742 p->link = u->link; 743 p->i_key = u->i_key; 744 p->o_key = u->o_key; 745 p->proto = u->proto; 746 747 memcpy(p->name, u->name, sizeof(u->name)); 748 } 749 750 static void 751 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p) 752 { 753 u->laddr = p->laddr; 754 u->raddr = p->raddr; 755 u->link = p->link; 756 u->i_key = p->i_key; 757 u->o_key = p->o_key; 758 if (u->i_key) 759 u->i_flags |= GRE_KEY; 760 if (u->o_key) 761 u->o_flags |= GRE_KEY; 762 u->proto = p->proto; 763 764 memcpy(u->name, p->name, sizeof(u->name)); 765 } 766 767 /** 768 * vti6_siocdevprivate - configure vti6 tunnels from userspace 769 * @dev: virtual device associated with tunnel 770 * @ifr: unused 771 * @data: parameters passed from userspace 772 * @cmd: command to be performed 773 * 774 * Description: 775 * vti6_siocdevprivate() is used for managing vti6 tunnels 776 * from userspace. 777 * 778 * The possible commands are the following: 779 * %SIOCGETTUNNEL: get tunnel parameters for device 780 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 781 * %SIOCCHGTUNNEL: change tunnel parameters to those given 782 * %SIOCDELTUNNEL: delete tunnel 783 * 784 * The fallback device "ip6_vti0", created during module 785 * initialization, can be used for creating other tunnel devices. 786 * 787 * Return: 788 * 0 on success, 789 * %-EFAULT if unable to copy data to or from userspace, 790 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 791 * %-EINVAL if passed tunnel parameters are invalid, 792 * %-EEXIST if changing a tunnel's parameters would cause a conflict 793 * %-ENODEV if attempting to change or delete a nonexisting device 794 **/ 795 static int 796 vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd) 797 { 798 int err = 0; 799 struct ip6_tnl_parm2 p; 800 struct __ip6_tnl_parm p1; 801 struct ip6_tnl *t = NULL; 802 struct net *net = dev_net(dev); 803 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 804 805 memset(&p1, 0, sizeof(p1)); 806 807 switch (cmd) { 808 case SIOCGETTUNNEL: 809 if (dev == ip6n->fb_tnl_dev) { 810 if (copy_from_user(&p, data, sizeof(p))) { 811 err = -EFAULT; 812 break; 813 } 814 vti6_parm_from_user(&p1, &p); 815 t = vti6_locate(net, &p1, 0); 816 } else { 817 memset(&p, 0, sizeof(p)); 818 } 819 if (!t) 820 t = netdev_priv(dev); 821 vti6_parm_to_user(&p, &t->parms); 822 if (copy_to_user(data, &p, sizeof(p))) 823 err = -EFAULT; 824 break; 825 case SIOCADDTUNNEL: 826 case SIOCCHGTUNNEL: 827 err = -EPERM; 828 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 829 break; 830 err = -EFAULT; 831 if (copy_from_user(&p, data, sizeof(p))) 832 break; 833 err = -EINVAL; 834 if (p.proto != IPPROTO_IPV6 && p.proto != 0) 835 break; 836 vti6_parm_from_user(&p1, &p); 837 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL); 838 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) { 839 if (t) { 840 if (t->dev != dev) { 841 err = -EEXIST; 842 break; 843 } 844 } else 845 t = netdev_priv(dev); 846 847 err = vti6_update(t, &p1, false); 848 } 849 if (t) { 850 err = 0; 851 vti6_parm_to_user(&p, &t->parms); 852 if (copy_to_user(data, &p, sizeof(p))) 853 err = -EFAULT; 854 855 } else 856 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 857 break; 858 case SIOCDELTUNNEL: 859 err = -EPERM; 860 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 861 break; 862 863 if (dev == ip6n->fb_tnl_dev) { 864 err = -EFAULT; 865 if (copy_from_user(&p, data, sizeof(p))) 866 break; 867 err = -ENOENT; 868 vti6_parm_from_user(&p1, &p); 869 t = vti6_locate(net, &p1, 0); 870 if (!t) 871 break; 872 err = -EPERM; 873 if (t->dev == ip6n->fb_tnl_dev) 874 break; 875 dev = t->dev; 876 } 877 err = 0; 878 unregister_netdevice(dev); 879 break; 880 default: 881 err = -EINVAL; 882 } 883 return err; 884 } 885 886 static const struct net_device_ops vti6_netdev_ops = { 887 .ndo_init = vti6_dev_init, 888 .ndo_uninit = vti6_dev_uninit, 889 .ndo_start_xmit = vti6_tnl_xmit, 890 .ndo_siocdevprivate = vti6_siocdevprivate, 891 .ndo_get_iflink = ip6_tnl_get_iflink, 892 }; 893 894 /** 895 * vti6_dev_setup - setup virtual tunnel device 896 * @dev: virtual device associated with tunnel 897 * 898 * Description: 899 * Initialize function pointers and device parameters 900 **/ 901 static void vti6_dev_setup(struct net_device *dev) 902 { 903 dev->netdev_ops = &vti6_netdev_ops; 904 dev->header_ops = &ip_tunnel_header_ops; 905 dev->needs_free_netdev = true; 906 907 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 908 dev->type = ARPHRD_TUNNEL6; 909 dev->min_mtu = IPV4_MIN_MTU; 910 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr); 911 dev->flags |= IFF_NOARP; 912 dev->addr_len = sizeof(struct in6_addr); 913 netif_keep_dst(dev); 914 /* This perm addr will be used as interface identifier by IPv6 */ 915 dev->addr_assign_type = NET_ADDR_RANDOM; 916 eth_random_addr(dev->perm_addr); 917 } 918 919 /** 920 * vti6_dev_init_gen - general initializer for all tunnel devices 921 * @dev: virtual device associated with tunnel 922 **/ 923 static inline int vti6_dev_init_gen(struct net_device *dev) 924 { 925 struct ip6_tnl *t = netdev_priv(dev); 926 927 t->dev = dev; 928 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL); 929 netdev_lockdep_set_classes(dev); 930 return 0; 931 } 932 933 /** 934 * vti6_dev_init - initializer for all non fallback tunnel devices 935 * @dev: virtual device associated with tunnel 936 **/ 937 static int vti6_dev_init(struct net_device *dev) 938 { 939 struct ip6_tnl *t = netdev_priv(dev); 940 int err = vti6_dev_init_gen(dev); 941 942 if (err) 943 return err; 944 vti6_link_config(t, true); 945 return 0; 946 } 947 948 /** 949 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 950 * @dev: fallback device 951 * 952 * Return: 0 953 **/ 954 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 955 { 956 struct ip6_tnl *t = netdev_priv(dev); 957 struct net *net = dev_net(dev); 958 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 959 960 t->net = net; 961 t->parms.proto = IPPROTO_IPV6; 962 963 rcu_assign_pointer(ip6n->tnls_wc[0], t); 964 return 0; 965 } 966 967 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 968 struct netlink_ext_ack *extack) 969 { 970 return 0; 971 } 972 973 static void vti6_netlink_parms(struct nlattr *data[], 974 struct __ip6_tnl_parm *parms) 975 { 976 memset(parms, 0, sizeof(*parms)); 977 978 if (!data) 979 return; 980 981 if (data[IFLA_VTI_LINK]) 982 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 983 984 if (data[IFLA_VTI_LOCAL]) 985 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 986 987 if (data[IFLA_VTI_REMOTE]) 988 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 989 990 if (data[IFLA_VTI_IKEY]) 991 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 992 993 if (data[IFLA_VTI_OKEY]) 994 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 995 996 if (data[IFLA_VTI_FWMARK]) 997 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 998 } 999 1000 static int vti6_newlink(struct net_device *dev, 1001 struct rtnl_newlink_params *params, 1002 struct netlink_ext_ack *extack) 1003 { 1004 struct nlattr **data = params->data; 1005 struct ip6_tnl *nt; 1006 struct net *net; 1007 1008 net = params->link_net ? : dev_net(dev); 1009 nt = netdev_priv(dev); 1010 vti6_netlink_parms(data, &nt->parms); 1011 1012 nt->parms.proto = IPPROTO_IPV6; 1013 nt->net = net; 1014 1015 if (vti6_locate(net, &nt->parms, 0)) 1016 return -EEXIST; 1017 1018 return vti6_tnl_create2(dev); 1019 } 1020 1021 static void vti6_dellink(struct net_device *dev, struct list_head *head) 1022 { 1023 struct net *net = dev_net(dev); 1024 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1025 1026 if (dev != ip6n->fb_tnl_dev) 1027 unregister_netdevice_queue(dev, head); 1028 } 1029 1030 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1031 struct nlattr *data[], 1032 struct netlink_ext_ack *extack) 1033 { 1034 struct ip6_tnl *t; 1035 struct __ip6_tnl_parm p; 1036 struct net *net = dev_net(dev); 1037 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1038 1039 if (dev == ip6n->fb_tnl_dev) 1040 return -EINVAL; 1041 1042 vti6_netlink_parms(data, &p); 1043 1044 t = vti6_locate(net, &p, 0); 1045 1046 if (t) { 1047 if (t->dev != dev) 1048 return -EEXIST; 1049 } else 1050 t = netdev_priv(dev); 1051 1052 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1053 } 1054 1055 static size_t vti6_get_size(const struct net_device *dev) 1056 { 1057 return 1058 /* IFLA_VTI_LINK */ 1059 nla_total_size(4) + 1060 /* IFLA_VTI_LOCAL */ 1061 nla_total_size(sizeof(struct in6_addr)) + 1062 /* IFLA_VTI_REMOTE */ 1063 nla_total_size(sizeof(struct in6_addr)) + 1064 /* IFLA_VTI_IKEY */ 1065 nla_total_size(4) + 1066 /* IFLA_VTI_OKEY */ 1067 nla_total_size(4) + 1068 /* IFLA_VTI_FWMARK */ 1069 nla_total_size(4) + 1070 0; 1071 } 1072 1073 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1074 { 1075 struct ip6_tnl *tunnel = netdev_priv(dev); 1076 struct __ip6_tnl_parm *parm = &tunnel->parms; 1077 1078 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1079 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1080 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1081 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1082 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1083 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1084 goto nla_put_failure; 1085 return 0; 1086 1087 nla_put_failure: 1088 return -EMSGSIZE; 1089 } 1090 1091 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1092 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1093 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1094 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1095 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1096 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1097 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1098 }; 1099 1100 static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1101 .kind = "vti6", 1102 .maxtype = IFLA_VTI_MAX, 1103 .policy = vti6_policy, 1104 .priv_size = sizeof(struct ip6_tnl), 1105 .setup = vti6_dev_setup, 1106 .validate = vti6_validate, 1107 .newlink = vti6_newlink, 1108 .dellink = vti6_dellink, 1109 .changelink = vti6_changelink, 1110 .get_size = vti6_get_size, 1111 .fill_info = vti6_fill_info, 1112 .get_link_net = ip6_tnl_get_link_net, 1113 }; 1114 1115 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1116 struct list_head *list) 1117 { 1118 int h; 1119 struct ip6_tnl *t; 1120 1121 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1122 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1123 while (t) { 1124 unregister_netdevice_queue(t->dev, list); 1125 t = rtnl_dereference(t->next); 1126 } 1127 } 1128 1129 t = rtnl_dereference(ip6n->tnls_wc[0]); 1130 if (t) 1131 unregister_netdevice_queue(t->dev, list); 1132 } 1133 1134 static int __net_init vti6_init_net(struct net *net) 1135 { 1136 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1137 struct ip6_tnl *t = NULL; 1138 int err; 1139 1140 ip6n->tnls[0] = ip6n->tnls_wc; 1141 ip6n->tnls[1] = ip6n->tnls_r_l; 1142 1143 if (!net_has_fallback_tunnels(net)) 1144 return 0; 1145 err = -ENOMEM; 1146 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1147 NET_NAME_UNKNOWN, vti6_dev_setup); 1148 1149 if (!ip6n->fb_tnl_dev) 1150 goto err_alloc_dev; 1151 dev_net_set(ip6n->fb_tnl_dev, net); 1152 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1153 1154 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1155 if (err < 0) 1156 goto err_register; 1157 1158 err = register_netdev(ip6n->fb_tnl_dev); 1159 if (err < 0) 1160 goto err_register; 1161 1162 t = netdev_priv(ip6n->fb_tnl_dev); 1163 1164 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1165 return 0; 1166 1167 err_register: 1168 free_netdev(ip6n->fb_tnl_dev); 1169 err_alloc_dev: 1170 return err; 1171 } 1172 1173 static void __net_exit vti6_exit_batch_rtnl(struct list_head *net_list, 1174 struct list_head *dev_to_kill) 1175 { 1176 struct vti6_net *ip6n; 1177 struct net *net; 1178 1179 ASSERT_RTNL(); 1180 list_for_each_entry(net, net_list, exit_list) { 1181 ip6n = net_generic(net, vti6_net_id); 1182 vti6_destroy_tunnels(ip6n, dev_to_kill); 1183 } 1184 } 1185 1186 static struct pernet_operations vti6_net_ops = { 1187 .init = vti6_init_net, 1188 .exit_batch_rtnl = vti6_exit_batch_rtnl, 1189 .id = &vti6_net_id, 1190 .size = sizeof(struct vti6_net), 1191 }; 1192 1193 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1194 .handler = vti6_rcv, 1195 .input_handler = vti6_input_proto, 1196 .cb_handler = vti6_rcv_cb, 1197 .err_handler = vti6_err, 1198 .priority = 100, 1199 }; 1200 1201 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1202 .handler = vti6_rcv, 1203 .input_handler = vti6_input_proto, 1204 .cb_handler = vti6_rcv_cb, 1205 .err_handler = vti6_err, 1206 .priority = 100, 1207 }; 1208 1209 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1210 .handler = vti6_rcv, 1211 .input_handler = vti6_input_proto, 1212 .cb_handler = vti6_rcv_cb, 1213 .err_handler = vti6_err, 1214 .priority = 100, 1215 }; 1216 1217 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1218 static int vti6_rcv_tunnel(struct sk_buff *skb) 1219 { 1220 const xfrm_address_t *saddr; 1221 __be32 spi; 1222 1223 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1224 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1225 1226 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1227 } 1228 1229 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1230 .handler = vti6_rcv_tunnel, 1231 .cb_handler = vti6_rcv_cb, 1232 .err_handler = vti6_err, 1233 .priority = 0, 1234 }; 1235 1236 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1237 .handler = vti6_rcv_tunnel, 1238 .cb_handler = vti6_rcv_cb, 1239 .err_handler = vti6_err, 1240 .priority = 0, 1241 }; 1242 #endif 1243 1244 /** 1245 * vti6_tunnel_init - register protocol and reserve needed resources 1246 * 1247 * Return: 0 on success 1248 **/ 1249 static int __init vti6_tunnel_init(void) 1250 { 1251 const char *msg; 1252 int err; 1253 1254 msg = "tunnel device"; 1255 err = register_pernet_device(&vti6_net_ops); 1256 if (err < 0) 1257 goto pernet_dev_failed; 1258 1259 msg = "tunnel protocols"; 1260 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1261 if (err < 0) 1262 goto xfrm_proto_esp_failed; 1263 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1264 if (err < 0) 1265 goto xfrm_proto_ah_failed; 1266 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1267 if (err < 0) 1268 goto xfrm_proto_comp_failed; 1269 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1270 msg = "ipv6 tunnel"; 1271 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1272 if (err < 0) 1273 goto vti_tunnel_ipv6_failed; 1274 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1275 if (err < 0) 1276 goto vti_tunnel_ip6ip_failed; 1277 #endif 1278 1279 msg = "netlink interface"; 1280 err = rtnl_link_register(&vti6_link_ops); 1281 if (err < 0) 1282 goto rtnl_link_failed; 1283 1284 return 0; 1285 1286 rtnl_link_failed: 1287 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1288 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1289 vti_tunnel_ip6ip_failed: 1290 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1291 vti_tunnel_ipv6_failed: 1292 #endif 1293 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1294 xfrm_proto_comp_failed: 1295 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1296 xfrm_proto_ah_failed: 1297 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1298 xfrm_proto_esp_failed: 1299 unregister_pernet_device(&vti6_net_ops); 1300 pernet_dev_failed: 1301 pr_err("vti6 init: failed to register %s\n", msg); 1302 return err; 1303 } 1304 1305 /** 1306 * vti6_tunnel_cleanup - free resources and unregister protocol 1307 **/ 1308 static void __exit vti6_tunnel_cleanup(void) 1309 { 1310 rtnl_link_unregister(&vti6_link_ops); 1311 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1312 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1313 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1314 #endif 1315 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1316 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1317 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1318 unregister_pernet_device(&vti6_net_ops); 1319 } 1320 1321 module_init(vti6_tunnel_init); 1322 module_exit(vti6_tunnel_cleanup); 1323 MODULE_LICENSE("GPL"); 1324 MODULE_ALIAS_RTNL_LINK("vti6"); 1325 MODULE_ALIAS_NETDEV("ip6_vti0"); 1326 MODULE_AUTHOR("Steffen Klassert"); 1327 MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1328