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 53static 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 60static int vti6_dev_init(struct net_device *dev); 61static void vti6_dev_setup(struct net_device *dev); 62static struct rtnl_link_ops vti6_link_ops __read_mostly; 63 64static unsigned int vti6_net_id __read_mostly; 65struct 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 **/ 88static struct ip6_tnl * 89vti6_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 **/ 137static struct ip6_tnl __rcu ** 138vti6_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 152static void 153vti6_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 161static void 162vti6_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 177static void vti6_dev_free(struct net_device *dev) 178{ 179 free_percpu(dev->tstats); 180} 181 182static 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 200out: 201 return err; 202} 203 204static 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 strscpy(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 235failed_free: 236 free_netdev(dev); 237failed: 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 **/ 255static 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 **/ 287static 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 netdev_put(dev, &t->dev_tracker); 297} 298 299static 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 DEV_STATS_INC(t->dev, 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; 334discard: 335 kfree_skb(skb); 336 return 0; 337} 338 339static 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 346static 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_INC(dev, rx_errors); 363 DEV_STATS_INC(dev, 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 **/ 410static inline bool 411vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr) 412{ 413 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 414} 415 416static 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 **/ 445static int 446vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl) 447{ 448 struct ip6_tnl *t = netdev_priv(dev); 449 struct dst_entry *dst = skb_dst(skb); 450 struct net_device *tdev; 451 struct xfrm_state *x; 452 int pkt_len = skb->len; 453 int err = -1; 454 int mtu; 455 456 if (!dst) { 457 switch (skb->protocol) { 458 case htons(ETH_P_IP): { 459 struct rtable *rt; 460 461 fl->u.ip4.flowi4_oif = dev->ifindex; 462 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC; 463 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4); 464 if (IS_ERR(rt)) 465 goto tx_err_link_failure; 466 dst = &rt->dst; 467 skb_dst_set(skb, dst); 468 break; 469 } 470 case htons(ETH_P_IPV6): 471 fl->u.ip6.flowi6_oif = dev->ifindex; 472 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC; 473 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6); 474 if (dst->error) { 475 dst_release(dst); 476 dst = NULL; 477 goto tx_err_link_failure; 478 } 479 skb_dst_set(skb, dst); 480 break; 481 default: 482 goto tx_err_link_failure; 483 } 484 } 485 486 dst_hold(dst); 487 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0); 488 if (IS_ERR(dst)) { 489 err = PTR_ERR(dst); 490 dst = NULL; 491 goto tx_err_link_failure; 492 } 493 494 if (dst->flags & DST_XFRM_QUEUE) 495 goto xmit; 496 497 x = dst->xfrm; 498 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr)) 499 goto tx_err_link_failure; 500 501 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr, 502 (const struct in6_addr *)&x->id.daddr)) 503 goto tx_err_link_failure; 504 505 tdev = dst->dev; 506 507 if (tdev == dev) { 508 DEV_STATS_INC(dev, collisions); 509 net_warn_ratelimited("%s: Local routing loop detected!\n", 510 t->parms.name); 511 goto tx_err_dst_release; 512 } 513 514 mtu = dst_mtu(dst); 515 if (skb->len > mtu) { 516 skb_dst_update_pmtu_no_confirm(skb, mtu); 517 518 if (skb->protocol == htons(ETH_P_IPV6)) { 519 if (mtu < IPV6_MIN_MTU) 520 mtu = IPV6_MIN_MTU; 521 522 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 523 } else { 524 if (!(ip_hdr(skb)->frag_off & htons(IP_DF))) 525 goto xmit; 526 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 527 htonl(mtu)); 528 } 529 530 err = -EMSGSIZE; 531 goto tx_err_dst_release; 532 } 533 534xmit: 535 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev))); 536 skb_dst_set(skb, dst); 537 skb->dev = skb_dst(skb)->dev; 538 539 err = dst_output(t->net, skb->sk, skb); 540 if (net_xmit_eval(err) == 0) 541 err = pkt_len; 542 iptunnel_xmit_stats(dev, err); 543 544 return 0; 545tx_err_link_failure: 546 DEV_STATS_INC(dev, tx_carrier_errors); 547 dst_link_failure(skb); 548tx_err_dst_release: 549 dst_release(dst); 550 return err; 551} 552 553static netdev_tx_t 554vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) 555{ 556 struct ip6_tnl *t = netdev_priv(dev); 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 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 572 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6); 573 break; 574 case htons(ETH_P_IP): 575 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 576 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET); 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 591tx_err: 592 DEV_STATS_INC(dev, tx_errors); 593 DEV_STATS_INC(dev, tx_dropped); 594 kfree_skb(skb); 595 return NETDEV_TX_OK; 596} 597 598static 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 654static 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 __dev_addr_set(dev, &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 **/ 710static int 711vti6_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 726static 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 741static void 742vti6_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 754static void 755vti6_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_siocdevprivate - configure vti6 tunnels from userspace 773 * @dev: virtual device associated with tunnel 774 * @ifr: unused 775 * @data: parameters passed from userspace 776 * @cmd: command to be performed 777 * 778 * Description: 779 * vti6_siocdevprivate() is used for managing vti6 tunnels 780 * from userspace. 781 * 782 * The possible commands are the following: 783 * %SIOCGETTUNNEL: get tunnel parameters for device 784 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters 785 * %SIOCCHGTUNNEL: change tunnel parameters to those given 786 * %SIOCDELTUNNEL: delete tunnel 787 * 788 * The fallback device "ip6_vti0", created during module 789 * initialization, can be used for creating other tunnel devices. 790 * 791 * Return: 792 * 0 on success, 793 * %-EFAULT if unable to copy data to or from userspace, 794 * %-EPERM if current process hasn't %CAP_NET_ADMIN set 795 * %-EINVAL if passed tunnel parameters are invalid, 796 * %-EEXIST if changing a tunnel's parameters would cause a conflict 797 * %-ENODEV if attempting to change or delete a nonexisting device 798 **/ 799static int 800vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd) 801{ 802 int err = 0; 803 struct ip6_tnl_parm2 p; 804 struct __ip6_tnl_parm p1; 805 struct ip6_tnl *t = NULL; 806 struct net *net = dev_net(dev); 807 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 808 809 memset(&p1, 0, sizeof(p1)); 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 890static 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 **/ 906static 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 **/ 928static 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 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL); 938 netdev_lockdep_set_classes(dev); 939 return 0; 940} 941 942/** 943 * vti6_dev_init - initializer for all non fallback tunnel devices 944 * @dev: virtual device associated with tunnel 945 **/ 946static int vti6_dev_init(struct net_device *dev) 947{ 948 struct ip6_tnl *t = netdev_priv(dev); 949 int err = vti6_dev_init_gen(dev); 950 951 if (err) 952 return err; 953 vti6_link_config(t, true); 954 return 0; 955} 956 957/** 958 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device 959 * @dev: fallback device 960 * 961 * Return: 0 962 **/ 963static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev) 964{ 965 struct ip6_tnl *t = netdev_priv(dev); 966 struct net *net = dev_net(dev); 967 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 968 969 t->parms.proto = IPPROTO_IPV6; 970 971 rcu_assign_pointer(ip6n->tnls_wc[0], t); 972 return 0; 973} 974 975static int vti6_validate(struct nlattr *tb[], struct nlattr *data[], 976 struct netlink_ext_ack *extack) 977{ 978 return 0; 979} 980 981static void vti6_netlink_parms(struct nlattr *data[], 982 struct __ip6_tnl_parm *parms) 983{ 984 memset(parms, 0, sizeof(*parms)); 985 986 if (!data) 987 return; 988 989 if (data[IFLA_VTI_LINK]) 990 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 991 992 if (data[IFLA_VTI_LOCAL]) 993 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]); 994 995 if (data[IFLA_VTI_REMOTE]) 996 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]); 997 998 if (data[IFLA_VTI_IKEY]) 999 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 1000 1001 if (data[IFLA_VTI_OKEY]) 1002 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 1003 1004 if (data[IFLA_VTI_FWMARK]) 1005 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 1006} 1007 1008static int vti6_newlink(struct net *src_net, struct net_device *dev, 1009 struct nlattr *tb[], struct nlattr *data[], 1010 struct netlink_ext_ack *extack) 1011{ 1012 struct net *net = dev_net(dev); 1013 struct ip6_tnl *nt; 1014 1015 nt = netdev_priv(dev); 1016 vti6_netlink_parms(data, &nt->parms); 1017 1018 nt->parms.proto = IPPROTO_IPV6; 1019 1020 if (vti6_locate(net, &nt->parms, 0)) 1021 return -EEXIST; 1022 1023 return vti6_tnl_create2(dev); 1024} 1025 1026static void vti6_dellink(struct net_device *dev, struct list_head *head) 1027{ 1028 struct net *net = dev_net(dev); 1029 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1030 1031 if (dev != ip6n->fb_tnl_dev) 1032 unregister_netdevice_queue(dev, head); 1033} 1034 1035static int vti6_changelink(struct net_device *dev, struct nlattr *tb[], 1036 struct nlattr *data[], 1037 struct netlink_ext_ack *extack) 1038{ 1039 struct ip6_tnl *t; 1040 struct __ip6_tnl_parm p; 1041 struct net *net = dev_net(dev); 1042 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1043 1044 if (dev == ip6n->fb_tnl_dev) 1045 return -EINVAL; 1046 1047 vti6_netlink_parms(data, &p); 1048 1049 t = vti6_locate(net, &p, 0); 1050 1051 if (t) { 1052 if (t->dev != dev) 1053 return -EEXIST; 1054 } else 1055 t = netdev_priv(dev); 1056 1057 return vti6_update(t, &p, tb && tb[IFLA_MTU]); 1058} 1059 1060static size_t vti6_get_size(const struct net_device *dev) 1061{ 1062 return 1063 /* IFLA_VTI_LINK */ 1064 nla_total_size(4) + 1065 /* IFLA_VTI_LOCAL */ 1066 nla_total_size(sizeof(struct in6_addr)) + 1067 /* IFLA_VTI_REMOTE */ 1068 nla_total_size(sizeof(struct in6_addr)) + 1069 /* IFLA_VTI_IKEY */ 1070 nla_total_size(4) + 1071 /* IFLA_VTI_OKEY */ 1072 nla_total_size(4) + 1073 /* IFLA_VTI_FWMARK */ 1074 nla_total_size(4) + 1075 0; 1076} 1077 1078static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev) 1079{ 1080 struct ip6_tnl *tunnel = netdev_priv(dev); 1081 struct __ip6_tnl_parm *parm = &tunnel->parms; 1082 1083 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) || 1084 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) || 1085 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) || 1086 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) || 1087 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) || 1088 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark)) 1089 goto nla_put_failure; 1090 return 0; 1091 1092nla_put_failure: 1093 return -EMSGSIZE; 1094} 1095 1096static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = { 1097 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 1098 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) }, 1099 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) }, 1100 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 1101 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 1102 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 1103}; 1104 1105static struct rtnl_link_ops vti6_link_ops __read_mostly = { 1106 .kind = "vti6", 1107 .maxtype = IFLA_VTI_MAX, 1108 .policy = vti6_policy, 1109 .priv_size = sizeof(struct ip6_tnl), 1110 .setup = vti6_dev_setup, 1111 .validate = vti6_validate, 1112 .newlink = vti6_newlink, 1113 .dellink = vti6_dellink, 1114 .changelink = vti6_changelink, 1115 .get_size = vti6_get_size, 1116 .fill_info = vti6_fill_info, 1117 .get_link_net = ip6_tnl_get_link_net, 1118}; 1119 1120static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n, 1121 struct list_head *list) 1122{ 1123 int h; 1124 struct ip6_tnl *t; 1125 1126 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) { 1127 t = rtnl_dereference(ip6n->tnls_r_l[h]); 1128 while (t) { 1129 unregister_netdevice_queue(t->dev, list); 1130 t = rtnl_dereference(t->next); 1131 } 1132 } 1133 1134 t = rtnl_dereference(ip6n->tnls_wc[0]); 1135 if (t) 1136 unregister_netdevice_queue(t->dev, list); 1137} 1138 1139static int __net_init vti6_init_net(struct net *net) 1140{ 1141 struct vti6_net *ip6n = net_generic(net, vti6_net_id); 1142 struct ip6_tnl *t = NULL; 1143 int err; 1144 1145 ip6n->tnls[0] = ip6n->tnls_wc; 1146 ip6n->tnls[1] = ip6n->tnls_r_l; 1147 1148 if (!net_has_fallback_tunnels(net)) 1149 return 0; 1150 err = -ENOMEM; 1151 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0", 1152 NET_NAME_UNKNOWN, vti6_dev_setup); 1153 1154 if (!ip6n->fb_tnl_dev) 1155 goto err_alloc_dev; 1156 dev_net_set(ip6n->fb_tnl_dev, net); 1157 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops; 1158 1159 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev); 1160 if (err < 0) 1161 goto err_register; 1162 1163 err = register_netdev(ip6n->fb_tnl_dev); 1164 if (err < 0) 1165 goto err_register; 1166 1167 t = netdev_priv(ip6n->fb_tnl_dev); 1168 1169 strcpy(t->parms.name, ip6n->fb_tnl_dev->name); 1170 return 0; 1171 1172err_register: 1173 free_netdev(ip6n->fb_tnl_dev); 1174err_alloc_dev: 1175 return err; 1176} 1177 1178static void __net_exit vti6_exit_batch_rtnl(struct list_head *net_list, 1179 struct list_head *dev_to_kill) 1180{ 1181 struct vti6_net *ip6n; 1182 struct net *net; 1183 1184 ASSERT_RTNL(); 1185 list_for_each_entry(net, net_list, exit_list) { 1186 ip6n = net_generic(net, vti6_net_id); 1187 vti6_destroy_tunnels(ip6n, dev_to_kill); 1188 } 1189} 1190 1191static struct pernet_operations vti6_net_ops = { 1192 .init = vti6_init_net, 1193 .exit_batch_rtnl = vti6_exit_batch_rtnl, 1194 .id = &vti6_net_id, 1195 .size = sizeof(struct vti6_net), 1196}; 1197 1198static struct xfrm6_protocol vti_esp6_protocol __read_mostly = { 1199 .handler = vti6_rcv, 1200 .input_handler = vti6_input_proto, 1201 .cb_handler = vti6_rcv_cb, 1202 .err_handler = vti6_err, 1203 .priority = 100, 1204}; 1205 1206static struct xfrm6_protocol vti_ah6_protocol __read_mostly = { 1207 .handler = vti6_rcv, 1208 .input_handler = vti6_input_proto, 1209 .cb_handler = vti6_rcv_cb, 1210 .err_handler = vti6_err, 1211 .priority = 100, 1212}; 1213 1214static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = { 1215 .handler = vti6_rcv, 1216 .input_handler = vti6_input_proto, 1217 .cb_handler = vti6_rcv_cb, 1218 .err_handler = vti6_err, 1219 .priority = 100, 1220}; 1221 1222#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1223static int vti6_rcv_tunnel(struct sk_buff *skb) 1224{ 1225 const xfrm_address_t *saddr; 1226 __be32 spi; 1227 1228 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr; 1229 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr); 1230 1231 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0); 1232} 1233 1234static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = { 1235 .handler = vti6_rcv_tunnel, 1236 .cb_handler = vti6_rcv_cb, 1237 .err_handler = vti6_err, 1238 .priority = 0, 1239}; 1240 1241static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = { 1242 .handler = vti6_rcv_tunnel, 1243 .cb_handler = vti6_rcv_cb, 1244 .err_handler = vti6_err, 1245 .priority = 0, 1246}; 1247#endif 1248 1249/** 1250 * vti6_tunnel_init - register protocol and reserve needed resources 1251 * 1252 * Return: 0 on success 1253 **/ 1254static int __init vti6_tunnel_init(void) 1255{ 1256 const char *msg; 1257 int err; 1258 1259 msg = "tunnel device"; 1260 err = register_pernet_device(&vti6_net_ops); 1261 if (err < 0) 1262 goto pernet_dev_failed; 1263 1264 msg = "tunnel protocols"; 1265 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP); 1266 if (err < 0) 1267 goto xfrm_proto_esp_failed; 1268 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH); 1269 if (err < 0) 1270 goto xfrm_proto_ah_failed; 1271 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP); 1272 if (err < 0) 1273 goto xfrm_proto_comp_failed; 1274#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1275 msg = "ipv6 tunnel"; 1276 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6); 1277 if (err < 0) 1278 goto vti_tunnel_ipv6_failed; 1279 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET); 1280 if (err < 0) 1281 goto vti_tunnel_ip6ip_failed; 1282#endif 1283 1284 msg = "netlink interface"; 1285 err = rtnl_link_register(&vti6_link_ops); 1286 if (err < 0) 1287 goto rtnl_link_failed; 1288 1289 return 0; 1290 1291rtnl_link_failed: 1292#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1293 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1294vti_tunnel_ip6ip_failed: 1295 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1296vti_tunnel_ipv6_failed: 1297#endif 1298 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1299xfrm_proto_comp_failed: 1300 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1301xfrm_proto_ah_failed: 1302 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1303xfrm_proto_esp_failed: 1304 unregister_pernet_device(&vti6_net_ops); 1305pernet_dev_failed: 1306 pr_err("vti6 init: failed to register %s\n", msg); 1307 return err; 1308} 1309 1310/** 1311 * vti6_tunnel_cleanup - free resources and unregister protocol 1312 **/ 1313static void __exit vti6_tunnel_cleanup(void) 1314{ 1315 rtnl_link_unregister(&vti6_link_ops); 1316#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL) 1317 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET); 1318 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6); 1319#endif 1320 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP); 1321 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH); 1322 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP); 1323 unregister_pernet_device(&vti6_net_ops); 1324} 1325 1326module_init(vti6_tunnel_init); 1327module_exit(vti6_tunnel_cleanup); 1328MODULE_LICENSE("GPL"); 1329MODULE_ALIAS_RTNL_LINK("vti6"); 1330MODULE_ALIAS_NETDEV("ip6_vti0"); 1331MODULE_AUTHOR("Steffen Klassert"); 1332MODULE_DESCRIPTION("IPv6 virtual tunnel interface"); 1333