1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * GRE over IPv6 protocol decoder. 4 * 5 * Authors: Dmitry Kozlov (xeb@mail.ru) 6 */ 7 8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 10 #include <linux/capability.h> 11 #include <linux/module.h> 12 #include <linux/types.h> 13 #include <linux/kernel.h> 14 #include <linux/slab.h> 15 #include <linux/uaccess.h> 16 #include <linux/skbuff.h> 17 #include <linux/netdevice.h> 18 #include <linux/in.h> 19 #include <linux/tcp.h> 20 #include <linux/udp.h> 21 #include <linux/if_arp.h> 22 #include <linux/init.h> 23 #include <linux/in6.h> 24 #include <linux/inetdevice.h> 25 #include <linux/igmp.h> 26 #include <linux/netfilter_ipv4.h> 27 #include <linux/etherdevice.h> 28 #include <linux/if_ether.h> 29 #include <linux/hash.h> 30 #include <linux/if_tunnel.h> 31 #include <linux/ip6_tunnel.h> 32 33 #include <net/sock.h> 34 #include <net/ip.h> 35 #include <net/ip_tunnels.h> 36 #include <net/icmp.h> 37 #include <net/protocol.h> 38 #include <net/addrconf.h> 39 #include <net/arp.h> 40 #include <net/checksum.h> 41 #include <net/dsfield.h> 42 #include <net/inet_ecn.h> 43 #include <net/xfrm.h> 44 #include <net/net_namespace.h> 45 #include <net/netns/generic.h> 46 #include <net/rtnetlink.h> 47 48 #include <net/ipv6.h> 49 #include <net/ip6_fib.h> 50 #include <net/ip6_route.h> 51 #include <net/ip6_tunnel.h> 52 #include <net/gre.h> 53 #include <net/erspan.h> 54 #include <net/dst_metadata.h> 55 56 57 static bool log_ecn_error = true; 58 module_param(log_ecn_error, bool, 0644); 59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 60 61 #define IP6_GRE_HASH_SIZE_SHIFT 5 62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT) 63 64 static unsigned int ip6gre_net_id __read_mostly; 65 struct ip6gre_net { 66 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE]; 67 68 struct ip6_tnl __rcu *collect_md_tun; 69 struct ip6_tnl __rcu *collect_md_tun_erspan; 70 struct net_device *fb_tunnel_dev; 71 }; 72 73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly; 76 static int ip6gre_tunnel_init(struct net_device *dev); 77 static void ip6gre_tunnel_setup(struct net_device *dev); 78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t); 79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu); 80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu); 81 82 /* Tunnel hash table */ 83 84 /* 85 4 hash tables: 86 87 3: (remote,local) 88 2: (remote,*) 89 1: (*,local) 90 0: (*,*) 91 92 We require exact key match i.e. if a key is present in packet 93 it will match only tunnel with the same key; if it is not present, 94 it will match only keyless tunnel. 95 96 All keysless packets, if not matched configured keyless tunnels 97 will match fallback tunnel. 98 */ 99 100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 101 static u32 HASH_ADDR(const struct in6_addr *addr) 102 { 103 u32 hash = ipv6_addr_hash(addr); 104 105 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 106 } 107 108 #define tunnels_r_l tunnels[3] 109 #define tunnels_r tunnels[2] 110 #define tunnels_l tunnels[1] 111 #define tunnels_wc tunnels[0] 112 113 /* Given src, dst and key, find appropriate for input tunnel. */ 114 115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 116 const struct in6_addr *remote, const struct in6_addr *local, 117 __be32 key, __be16 gre_proto) 118 { 119 struct net *net = dev_net(dev); 120 int link = dev->ifindex; 121 unsigned int h0 = HASH_ADDR(remote); 122 unsigned int h1 = HASH_KEY(key); 123 struct ip6_tnl *t, *cand = NULL; 124 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 125 int dev_type = (gre_proto == htons(ETH_P_TEB) || 126 gre_proto == htons(ETH_P_ERSPAN) || 127 gre_proto == htons(ETH_P_ERSPAN2)) ? 128 ARPHRD_ETHER : ARPHRD_IP6GRE; 129 int score, cand_score = 4; 130 struct net_device *ndev; 131 132 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 133 if (!ipv6_addr_equal(local, &t->parms.laddr) || 134 !ipv6_addr_equal(remote, &t->parms.raddr) || 135 key != t->parms.i_key || 136 !(t->dev->flags & IFF_UP)) 137 continue; 138 139 if (t->dev->type != ARPHRD_IP6GRE && 140 t->dev->type != dev_type) 141 continue; 142 143 score = 0; 144 if (t->parms.link != link) 145 score |= 1; 146 if (t->dev->type != dev_type) 147 score |= 2; 148 if (score == 0) 149 return t; 150 151 if (score < cand_score) { 152 cand = t; 153 cand_score = score; 154 } 155 } 156 157 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 158 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 159 key != t->parms.i_key || 160 !(t->dev->flags & IFF_UP)) 161 continue; 162 163 if (t->dev->type != ARPHRD_IP6GRE && 164 t->dev->type != dev_type) 165 continue; 166 167 score = 0; 168 if (t->parms.link != link) 169 score |= 1; 170 if (t->dev->type != dev_type) 171 score |= 2; 172 if (score == 0) 173 return t; 174 175 if (score < cand_score) { 176 cand = t; 177 cand_score = score; 178 } 179 } 180 181 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 182 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 183 (!ipv6_addr_equal(local, &t->parms.raddr) || 184 !ipv6_addr_is_multicast(local))) || 185 key != t->parms.i_key || 186 !(t->dev->flags & IFF_UP)) 187 continue; 188 189 if (t->dev->type != ARPHRD_IP6GRE && 190 t->dev->type != dev_type) 191 continue; 192 193 score = 0; 194 if (t->parms.link != link) 195 score |= 1; 196 if (t->dev->type != dev_type) 197 score |= 2; 198 if (score == 0) 199 return t; 200 201 if (score < cand_score) { 202 cand = t; 203 cand_score = score; 204 } 205 } 206 207 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 208 if (t->parms.i_key != key || 209 !(t->dev->flags & IFF_UP)) 210 continue; 211 212 if (t->dev->type != ARPHRD_IP6GRE && 213 t->dev->type != dev_type) 214 continue; 215 216 score = 0; 217 if (t->parms.link != link) 218 score |= 1; 219 if (t->dev->type != dev_type) 220 score |= 2; 221 if (score == 0) 222 return t; 223 224 if (score < cand_score) { 225 cand = t; 226 cand_score = score; 227 } 228 } 229 230 if (cand) 231 return cand; 232 233 if (gre_proto == htons(ETH_P_ERSPAN) || 234 gre_proto == htons(ETH_P_ERSPAN2)) 235 t = rcu_dereference(ign->collect_md_tun_erspan); 236 else 237 t = rcu_dereference(ign->collect_md_tun); 238 239 if (t && t->dev->flags & IFF_UP) 240 return t; 241 242 ndev = READ_ONCE(ign->fb_tunnel_dev); 243 if (ndev && ndev->flags & IFF_UP) 244 return netdev_priv(ndev); 245 246 return NULL; 247 } 248 249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 250 const struct __ip6_tnl_parm *p) 251 { 252 const struct in6_addr *remote = &p->raddr; 253 const struct in6_addr *local = &p->laddr; 254 unsigned int h = HASH_KEY(p->i_key); 255 int prio = 0; 256 257 if (!ipv6_addr_any(local)) 258 prio |= 1; 259 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 260 prio |= 2; 261 h ^= HASH_ADDR(remote); 262 } 263 264 return &ign->tunnels[prio][h]; 265 } 266 267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 268 { 269 if (t->parms.collect_md) 270 rcu_assign_pointer(ign->collect_md_tun, t); 271 } 272 273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 274 { 275 if (t->parms.collect_md) 276 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 277 } 278 279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 280 { 281 if (t->parms.collect_md) 282 rcu_assign_pointer(ign->collect_md_tun, NULL); 283 } 284 285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 286 struct ip6_tnl *t) 287 { 288 if (t->parms.collect_md) 289 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 290 } 291 292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 293 const struct ip6_tnl *t) 294 { 295 return __ip6gre_bucket(ign, &t->parms); 296 } 297 298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 299 { 300 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 301 302 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 303 rcu_assign_pointer(*tp, t); 304 } 305 306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 307 { 308 struct ip6_tnl __rcu **tp; 309 struct ip6_tnl *iter; 310 311 for (tp = ip6gre_bucket(ign, t); 312 (iter = rtnl_dereference(*tp)) != NULL; 313 tp = &iter->next) { 314 if (t == iter) { 315 rcu_assign_pointer(*tp, t->next); 316 break; 317 } 318 } 319 } 320 321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 322 const struct __ip6_tnl_parm *parms, 323 int type) 324 { 325 const struct in6_addr *remote = &parms->raddr; 326 const struct in6_addr *local = &parms->laddr; 327 __be32 key = parms->i_key; 328 int link = parms->link; 329 struct ip6_tnl *t; 330 struct ip6_tnl __rcu **tp; 331 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 332 333 for (tp = __ip6gre_bucket(ign, parms); 334 (t = rtnl_dereference(*tp)) != NULL; 335 tp = &t->next) 336 if (ipv6_addr_equal(local, &t->parms.laddr) && 337 ipv6_addr_equal(remote, &t->parms.raddr) && 338 key == t->parms.i_key && 339 link == t->parms.link && 340 type == t->dev->type) 341 break; 342 343 return t; 344 } 345 346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 347 const struct __ip6_tnl_parm *parms, int create) 348 { 349 struct ip6_tnl *t, *nt; 350 struct net_device *dev; 351 char name[IFNAMSIZ]; 352 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 353 354 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 355 if (t && create) 356 return NULL; 357 if (t || !create) 358 return t; 359 360 if (parms->name[0]) { 361 if (!dev_valid_name(parms->name)) 362 return NULL; 363 strlcpy(name, parms->name, IFNAMSIZ); 364 } else { 365 strcpy(name, "ip6gre%d"); 366 } 367 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 368 ip6gre_tunnel_setup); 369 if (!dev) 370 return NULL; 371 372 dev_net_set(dev, net); 373 374 nt = netdev_priv(dev); 375 nt->parms = *parms; 376 dev->rtnl_link_ops = &ip6gre_link_ops; 377 378 nt->dev = dev; 379 nt->net = dev_net(dev); 380 381 if (register_netdevice(dev) < 0) 382 goto failed_free; 383 384 ip6gre_tnl_link_config(nt, 1); 385 386 /* Can use a lockless transmit, unless we generate output sequences */ 387 if (!(nt->parms.o_flags & TUNNEL_SEQ)) 388 dev->features |= NETIF_F_LLTX; 389 390 dev_hold(dev); 391 ip6gre_tunnel_link(ign, nt); 392 return nt; 393 394 failed_free: 395 free_netdev(dev); 396 return NULL; 397 } 398 399 static void ip6erspan_tunnel_uninit(struct net_device *dev) 400 { 401 struct ip6_tnl *t = netdev_priv(dev); 402 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 403 404 ip6erspan_tunnel_unlink_md(ign, t); 405 ip6gre_tunnel_unlink(ign, t); 406 dst_cache_reset(&t->dst_cache); 407 dev_put(dev); 408 } 409 410 static void ip6gre_tunnel_uninit(struct net_device *dev) 411 { 412 struct ip6_tnl *t = netdev_priv(dev); 413 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 414 415 ip6gre_tunnel_unlink_md(ign, t); 416 ip6gre_tunnel_unlink(ign, t); 417 if (ign->fb_tunnel_dev == dev) 418 WRITE_ONCE(ign->fb_tunnel_dev, NULL); 419 dst_cache_reset(&t->dst_cache); 420 dev_put(dev); 421 } 422 423 424 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 425 u8 type, u8 code, int offset, __be32 info) 426 { 427 struct net *net = dev_net(skb->dev); 428 const struct ipv6hdr *ipv6h; 429 struct tnl_ptk_info tpi; 430 struct ip6_tnl *t; 431 432 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 433 offset) < 0) 434 return -EINVAL; 435 436 ipv6h = (const struct ipv6hdr *)skb->data; 437 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 438 tpi.key, tpi.proto); 439 if (!t) 440 return -ENOENT; 441 442 switch (type) { 443 case ICMPV6_DEST_UNREACH: 444 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 445 t->parms.name); 446 if (code != ICMPV6_PORT_UNREACH) 447 break; 448 return 0; 449 case ICMPV6_TIME_EXCEED: 450 if (code == ICMPV6_EXC_HOPLIMIT) { 451 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 452 t->parms.name); 453 break; 454 } 455 return 0; 456 case ICMPV6_PARAMPROB: { 457 struct ipv6_tlv_tnl_enc_lim *tel; 458 __u32 teli; 459 460 teli = 0; 461 if (code == ICMPV6_HDR_FIELD) 462 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 463 464 if (teli && teli == be32_to_cpu(info) - 2) { 465 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 466 if (tel->encap_limit == 0) { 467 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 468 t->parms.name); 469 } 470 } else { 471 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 472 t->parms.name); 473 } 474 return 0; 475 } 476 case ICMPV6_PKT_TOOBIG: 477 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 478 return 0; 479 case NDISC_REDIRECT: 480 ip6_redirect(skb, net, skb->dev->ifindex, 0, 481 sock_net_uid(net, NULL)); 482 return 0; 483 } 484 485 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 486 t->err_count++; 487 else 488 t->err_count = 1; 489 t->err_time = jiffies; 490 491 return 0; 492 } 493 494 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 495 { 496 const struct ipv6hdr *ipv6h; 497 struct ip6_tnl *tunnel; 498 499 ipv6h = ipv6_hdr(skb); 500 tunnel = ip6gre_tunnel_lookup(skb->dev, 501 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 502 tpi->proto); 503 if (tunnel) { 504 if (tunnel->parms.collect_md) { 505 struct metadata_dst *tun_dst; 506 __be64 tun_id; 507 __be16 flags; 508 509 flags = tpi->flags; 510 tun_id = key32_to_tunnel_id(tpi->key); 511 512 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 513 if (!tun_dst) 514 return PACKET_REJECT; 515 516 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 517 } else { 518 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 519 } 520 521 return PACKET_RCVD; 522 } 523 524 return PACKET_REJECT; 525 } 526 527 static int ip6erspan_rcv(struct sk_buff *skb, 528 struct tnl_ptk_info *tpi, 529 int gre_hdr_len) 530 { 531 struct erspan_base_hdr *ershdr; 532 const struct ipv6hdr *ipv6h; 533 struct erspan_md2 *md2; 534 struct ip6_tnl *tunnel; 535 u8 ver; 536 537 ipv6h = ipv6_hdr(skb); 538 ershdr = (struct erspan_base_hdr *)skb->data; 539 ver = ershdr->ver; 540 541 tunnel = ip6gre_tunnel_lookup(skb->dev, 542 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 543 tpi->proto); 544 if (tunnel) { 545 int len = erspan_hdr_len(ver); 546 547 if (unlikely(!pskb_may_pull(skb, len))) 548 return PACKET_REJECT; 549 550 if (__iptunnel_pull_header(skb, len, 551 htons(ETH_P_TEB), 552 false, false) < 0) 553 return PACKET_REJECT; 554 555 if (tunnel->parms.collect_md) { 556 struct erspan_metadata *pkt_md, *md; 557 struct metadata_dst *tun_dst; 558 struct ip_tunnel_info *info; 559 unsigned char *gh; 560 __be64 tun_id; 561 __be16 flags; 562 563 tpi->flags |= TUNNEL_KEY; 564 flags = tpi->flags; 565 tun_id = key32_to_tunnel_id(tpi->key); 566 567 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 568 sizeof(*md)); 569 if (!tun_dst) 570 return PACKET_REJECT; 571 572 /* skb can be uncloned in __iptunnel_pull_header, so 573 * old pkt_md is no longer valid and we need to reset 574 * it 575 */ 576 gh = skb_network_header(skb) + 577 skb_network_header_len(skb); 578 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 579 sizeof(*ershdr)); 580 info = &tun_dst->u.tun_info; 581 md = ip_tunnel_info_opts(info); 582 md->version = ver; 583 md2 = &md->u.md2; 584 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 585 ERSPAN_V2_MDSIZE); 586 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 587 info->options_len = sizeof(*md); 588 589 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 590 591 } else { 592 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 593 } 594 595 return PACKET_RCVD; 596 } 597 598 return PACKET_REJECT; 599 } 600 601 static int gre_rcv(struct sk_buff *skb) 602 { 603 struct tnl_ptk_info tpi; 604 bool csum_err = false; 605 int hdr_len; 606 607 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 608 if (hdr_len < 0) 609 goto drop; 610 611 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 612 goto drop; 613 614 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 615 tpi.proto == htons(ETH_P_ERSPAN2))) { 616 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) 617 return 0; 618 goto out; 619 } 620 621 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 622 return 0; 623 624 out: 625 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 626 drop: 627 kfree_skb(skb); 628 return 0; 629 } 630 631 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 632 { 633 return iptunnel_handle_offloads(skb, 634 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 635 } 636 637 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 638 struct net_device *dev, 639 struct flowi6 *fl6, __u8 *dsfield, 640 int *encap_limit) 641 { 642 const struct iphdr *iph = ip_hdr(skb); 643 struct ip6_tnl *t = netdev_priv(dev); 644 645 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 646 *encap_limit = t->parms.encap_limit; 647 648 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 649 650 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 651 *dsfield = ipv4_get_dsfield(iph); 652 else 653 *dsfield = ip6_tclass(t->parms.flowinfo); 654 655 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 656 fl6->flowi6_mark = skb->mark; 657 else 658 fl6->flowi6_mark = t->parms.fwmark; 659 660 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 661 } 662 663 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 664 struct net_device *dev, 665 struct flowi6 *fl6, __u8 *dsfield, 666 int *encap_limit) 667 { 668 struct ipv6hdr *ipv6h; 669 struct ip6_tnl *t = netdev_priv(dev); 670 __u16 offset; 671 672 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 673 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 674 ipv6h = ipv6_hdr(skb); 675 676 if (offset > 0) { 677 struct ipv6_tlv_tnl_enc_lim *tel; 678 679 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 680 if (tel->encap_limit == 0) { 681 icmpv6_send(skb, ICMPV6_PARAMPROB, 682 ICMPV6_HDR_FIELD, offset + 2); 683 return -1; 684 } 685 *encap_limit = tel->encap_limit - 1; 686 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 687 *encap_limit = t->parms.encap_limit; 688 } 689 690 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 691 692 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 693 *dsfield = ipv6_get_dsfield(ipv6h); 694 else 695 *dsfield = ip6_tclass(t->parms.flowinfo); 696 697 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 698 fl6->flowlabel |= ip6_flowlabel(ipv6h); 699 700 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 701 fl6->flowi6_mark = skb->mark; 702 else 703 fl6->flowi6_mark = t->parms.fwmark; 704 705 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 706 707 return 0; 708 } 709 710 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb) 711 { 712 struct ip_tunnel_info *tun_info; 713 714 tun_info = skb_tunnel_info(skb); 715 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))) 716 return ERR_PTR(-EINVAL); 717 718 return tun_info; 719 } 720 721 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 722 struct net_device *dev, __u8 dsfield, 723 struct flowi6 *fl6, int encap_limit, 724 __u32 *pmtu, __be16 proto) 725 { 726 struct ip6_tnl *tunnel = netdev_priv(dev); 727 __be16 protocol; 728 729 if (dev->type == ARPHRD_ETHER) 730 IPCB(skb)->flags = 0; 731 732 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 733 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 734 else 735 fl6->daddr = tunnel->parms.raddr; 736 737 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 738 return -ENOMEM; 739 740 /* Push GRE header. */ 741 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 742 743 if (tunnel->parms.collect_md) { 744 struct ip_tunnel_info *tun_info; 745 const struct ip_tunnel_key *key; 746 __be16 flags; 747 748 tun_info = skb_tunnel_info_txcheck(skb); 749 if (IS_ERR(tun_info) || 750 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 751 return -EINVAL; 752 753 key = &tun_info->key; 754 memset(fl6, 0, sizeof(*fl6)); 755 fl6->flowi6_proto = IPPROTO_GRE; 756 fl6->daddr = key->u.ipv6.dst; 757 fl6->flowlabel = key->label; 758 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 759 760 dsfield = key->tos; 761 flags = key->tun_flags & 762 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 763 tunnel->tun_hlen = gre_calc_hlen(flags); 764 765 gre_build_header(skb, tunnel->tun_hlen, 766 flags, protocol, 767 tunnel_id_to_key32(tun_info->key.tun_id), 768 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 769 : 0); 770 771 } else { 772 if (tunnel->parms.o_flags & TUNNEL_SEQ) 773 tunnel->o_seqno++; 774 775 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 776 protocol, tunnel->parms.o_key, 777 htonl(tunnel->o_seqno)); 778 } 779 780 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 781 NEXTHDR_GRE); 782 } 783 784 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 785 { 786 struct ip6_tnl *t = netdev_priv(dev); 787 int encap_limit = -1; 788 struct flowi6 fl6; 789 __u8 dsfield = 0; 790 __u32 mtu; 791 int err; 792 793 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 794 795 if (!t->parms.collect_md) 796 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 797 &dsfield, &encap_limit); 798 799 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 800 if (err) 801 return -1; 802 803 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 804 skb->protocol); 805 if (err != 0) { 806 /* XXX: send ICMP error even if DF is not set. */ 807 if (err == -EMSGSIZE) 808 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 809 htonl(mtu)); 810 return -1; 811 } 812 813 return 0; 814 } 815 816 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 817 { 818 struct ip6_tnl *t = netdev_priv(dev); 819 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 820 int encap_limit = -1; 821 struct flowi6 fl6; 822 __u8 dsfield = 0; 823 __u32 mtu; 824 int err; 825 826 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 827 return -1; 828 829 if (!t->parms.collect_md && 830 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 831 return -1; 832 833 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 834 return -1; 835 836 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 837 &mtu, skb->protocol); 838 if (err != 0) { 839 if (err == -EMSGSIZE) 840 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 841 return -1; 842 } 843 844 return 0; 845 } 846 847 /** 848 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 849 * @t: the outgoing tunnel device 850 * @hdr: IPv6 header from the incoming packet 851 * 852 * Description: 853 * Avoid trivial tunneling loop by checking that tunnel exit-point 854 * doesn't match source of incoming packet. 855 * 856 * Return: 857 * 1 if conflict, 858 * 0 else 859 **/ 860 861 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 862 const struct ipv6hdr *hdr) 863 { 864 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 865 } 866 867 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 868 { 869 struct ip6_tnl *t = netdev_priv(dev); 870 int encap_limit = -1; 871 struct flowi6 fl6; 872 __u32 mtu; 873 int err; 874 875 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 876 encap_limit = t->parms.encap_limit; 877 878 if (!t->parms.collect_md) 879 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 880 881 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 882 if (err) 883 return err; 884 885 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 886 887 return err; 888 } 889 890 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 891 struct net_device *dev) 892 { 893 struct ip6_tnl *t = netdev_priv(dev); 894 struct net_device_stats *stats = &t->dev->stats; 895 int ret; 896 897 if (!pskb_inet_may_pull(skb)) 898 goto tx_err; 899 900 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 901 goto tx_err; 902 903 switch (skb->protocol) { 904 case htons(ETH_P_IP): 905 ret = ip6gre_xmit_ipv4(skb, dev); 906 break; 907 case htons(ETH_P_IPV6): 908 ret = ip6gre_xmit_ipv6(skb, dev); 909 break; 910 default: 911 ret = ip6gre_xmit_other(skb, dev); 912 break; 913 } 914 915 if (ret < 0) 916 goto tx_err; 917 918 return NETDEV_TX_OK; 919 920 tx_err: 921 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 922 stats->tx_errors++; 923 stats->tx_dropped++; 924 kfree_skb(skb); 925 return NETDEV_TX_OK; 926 } 927 928 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 929 struct net_device *dev) 930 { 931 struct ip_tunnel_info *tun_info = NULL; 932 struct ip6_tnl *t = netdev_priv(dev); 933 struct dst_entry *dst = skb_dst(skb); 934 struct net_device_stats *stats; 935 bool truncate = false; 936 int encap_limit = -1; 937 __u8 dsfield = false; 938 struct flowi6 fl6; 939 int err = -EINVAL; 940 __be16 proto; 941 __u32 mtu; 942 int nhoff; 943 int thoff; 944 945 if (!pskb_inet_may_pull(skb)) 946 goto tx_err; 947 948 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 949 goto tx_err; 950 951 if (gre_handle_offloads(skb, false)) 952 goto tx_err; 953 954 if (skb->len > dev->mtu + dev->hard_header_len) { 955 pskb_trim(skb, dev->mtu + dev->hard_header_len); 956 truncate = true; 957 } 958 959 nhoff = skb_network_header(skb) - skb_mac_header(skb); 960 if (skb->protocol == htons(ETH_P_IP) && 961 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 962 truncate = true; 963 964 thoff = skb_transport_header(skb) - skb_mac_header(skb); 965 if (skb->protocol == htons(ETH_P_IPV6) && 966 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 967 truncate = true; 968 969 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 970 goto tx_err; 971 972 t->parms.o_flags &= ~TUNNEL_KEY; 973 IPCB(skb)->flags = 0; 974 975 /* For collect_md mode, derive fl6 from the tunnel key, 976 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 977 */ 978 if (t->parms.collect_md) { 979 const struct ip_tunnel_key *key; 980 struct erspan_metadata *md; 981 __be32 tun_id; 982 983 tun_info = skb_tunnel_info_txcheck(skb); 984 if (IS_ERR(tun_info) || 985 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 986 goto tx_err; 987 988 key = &tun_info->key; 989 memset(&fl6, 0, sizeof(fl6)); 990 fl6.flowi6_proto = IPPROTO_GRE; 991 fl6.daddr = key->u.ipv6.dst; 992 fl6.flowlabel = key->label; 993 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 994 995 dsfield = key->tos; 996 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 997 goto tx_err; 998 if (tun_info->options_len < sizeof(*md)) 999 goto tx_err; 1000 md = ip_tunnel_info_opts(tun_info); 1001 1002 tun_id = tunnel_id_to_key32(key->tun_id); 1003 if (md->version == 1) { 1004 erspan_build_header(skb, 1005 ntohl(tun_id), 1006 ntohl(md->u.index), truncate, 1007 false); 1008 } else if (md->version == 2) { 1009 erspan_build_header_v2(skb, 1010 ntohl(tun_id), 1011 md->u.md2.dir, 1012 get_hwid(&md->u.md2), 1013 truncate, false); 1014 } else { 1015 goto tx_err; 1016 } 1017 } else { 1018 switch (skb->protocol) { 1019 case htons(ETH_P_IP): 1020 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1021 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1022 &dsfield, &encap_limit); 1023 break; 1024 case htons(ETH_P_IPV6): 1025 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1026 goto tx_err; 1027 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1028 &dsfield, &encap_limit)) 1029 goto tx_err; 1030 break; 1031 default: 1032 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1033 break; 1034 } 1035 1036 if (t->parms.erspan_ver == 1) 1037 erspan_build_header(skb, ntohl(t->parms.o_key), 1038 t->parms.index, 1039 truncate, false); 1040 else if (t->parms.erspan_ver == 2) 1041 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1042 t->parms.dir, 1043 t->parms.hwid, 1044 truncate, false); 1045 else 1046 goto tx_err; 1047 1048 fl6.daddr = t->parms.raddr; 1049 } 1050 1051 /* Push GRE header. */ 1052 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1053 : htons(ETH_P_ERSPAN2); 1054 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++)); 1055 1056 /* TooBig packet may have updated dst->dev's mtu */ 1057 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1058 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1059 1060 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1061 NEXTHDR_GRE); 1062 if (err != 0) { 1063 /* XXX: send ICMP error even if DF is not set. */ 1064 if (err == -EMSGSIZE) { 1065 if (skb->protocol == htons(ETH_P_IP)) 1066 icmp_send(skb, ICMP_DEST_UNREACH, 1067 ICMP_FRAG_NEEDED, htonl(mtu)); 1068 else 1069 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1070 } 1071 1072 goto tx_err; 1073 } 1074 return NETDEV_TX_OK; 1075 1076 tx_err: 1077 stats = &t->dev->stats; 1078 if (!IS_ERR(tun_info)) 1079 stats->tx_errors++; 1080 stats->tx_dropped++; 1081 kfree_skb(skb); 1082 return NETDEV_TX_OK; 1083 } 1084 1085 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1086 { 1087 struct net_device *dev = t->dev; 1088 struct __ip6_tnl_parm *p = &t->parms; 1089 struct flowi6 *fl6 = &t->fl.u.ip6; 1090 1091 if (dev->type != ARPHRD_ETHER) { 1092 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1093 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1094 } 1095 1096 /* Set up flowi template */ 1097 fl6->saddr = p->laddr; 1098 fl6->daddr = p->raddr; 1099 fl6->flowi6_oif = p->link; 1100 fl6->flowlabel = 0; 1101 fl6->flowi6_proto = IPPROTO_GRE; 1102 1103 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1104 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1105 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1106 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1107 1108 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1109 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1110 1111 if (p->flags&IP6_TNL_F_CAP_XMIT && 1112 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1113 dev->flags |= IFF_POINTOPOINT; 1114 else 1115 dev->flags &= ~IFF_POINTOPOINT; 1116 } 1117 1118 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1119 int t_hlen) 1120 { 1121 const struct __ip6_tnl_parm *p = &t->parms; 1122 struct net_device *dev = t->dev; 1123 1124 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1125 int strict = (ipv6_addr_type(&p->raddr) & 1126 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1127 1128 struct rt6_info *rt = rt6_lookup(t->net, 1129 &p->raddr, &p->laddr, 1130 p->link, NULL, strict); 1131 1132 if (!rt) 1133 return; 1134 1135 if (rt->dst.dev) { 1136 dev->needed_headroom = rt->dst.dev->hard_header_len + 1137 t_hlen; 1138 1139 if (set_mtu) { 1140 dev->mtu = rt->dst.dev->mtu - t_hlen; 1141 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1142 dev->mtu -= 8; 1143 if (dev->type == ARPHRD_ETHER) 1144 dev->mtu -= ETH_HLEN; 1145 1146 if (dev->mtu < IPV6_MIN_MTU) 1147 dev->mtu = IPV6_MIN_MTU; 1148 } 1149 } 1150 ip6_rt_put(rt); 1151 } 1152 } 1153 1154 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1155 { 1156 int t_hlen; 1157 1158 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1159 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1160 1161 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1162 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1163 return t_hlen; 1164 } 1165 1166 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1167 { 1168 ip6gre_tnl_link_config_common(t); 1169 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1170 } 1171 1172 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1173 const struct __ip6_tnl_parm *p) 1174 { 1175 t->parms.laddr = p->laddr; 1176 t->parms.raddr = p->raddr; 1177 t->parms.flags = p->flags; 1178 t->parms.hop_limit = p->hop_limit; 1179 t->parms.encap_limit = p->encap_limit; 1180 t->parms.flowinfo = p->flowinfo; 1181 t->parms.link = p->link; 1182 t->parms.proto = p->proto; 1183 t->parms.i_key = p->i_key; 1184 t->parms.o_key = p->o_key; 1185 t->parms.i_flags = p->i_flags; 1186 t->parms.o_flags = p->o_flags; 1187 t->parms.fwmark = p->fwmark; 1188 t->parms.erspan_ver = p->erspan_ver; 1189 t->parms.index = p->index; 1190 t->parms.dir = p->dir; 1191 t->parms.hwid = p->hwid; 1192 dst_cache_reset(&t->dst_cache); 1193 } 1194 1195 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1196 int set_mtu) 1197 { 1198 ip6gre_tnl_copy_tnl_parm(t, p); 1199 ip6gre_tnl_link_config(t, set_mtu); 1200 return 0; 1201 } 1202 1203 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1204 const struct ip6_tnl_parm2 *u) 1205 { 1206 p->laddr = u->laddr; 1207 p->raddr = u->raddr; 1208 p->flags = u->flags; 1209 p->hop_limit = u->hop_limit; 1210 p->encap_limit = u->encap_limit; 1211 p->flowinfo = u->flowinfo; 1212 p->link = u->link; 1213 p->i_key = u->i_key; 1214 p->o_key = u->o_key; 1215 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1216 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1217 memcpy(p->name, u->name, sizeof(u->name)); 1218 } 1219 1220 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1221 const struct __ip6_tnl_parm *p) 1222 { 1223 u->proto = IPPROTO_GRE; 1224 u->laddr = p->laddr; 1225 u->raddr = p->raddr; 1226 u->flags = p->flags; 1227 u->hop_limit = p->hop_limit; 1228 u->encap_limit = p->encap_limit; 1229 u->flowinfo = p->flowinfo; 1230 u->link = p->link; 1231 u->i_key = p->i_key; 1232 u->o_key = p->o_key; 1233 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1234 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1235 memcpy(u->name, p->name, sizeof(u->name)); 1236 } 1237 1238 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1239 struct ifreq *ifr, int cmd) 1240 { 1241 int err = 0; 1242 struct ip6_tnl_parm2 p; 1243 struct __ip6_tnl_parm p1; 1244 struct ip6_tnl *t = netdev_priv(dev); 1245 struct net *net = t->net; 1246 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1247 1248 memset(&p1, 0, sizeof(p1)); 1249 1250 switch (cmd) { 1251 case SIOCGETTUNNEL: 1252 if (dev == ign->fb_tunnel_dev) { 1253 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1254 err = -EFAULT; 1255 break; 1256 } 1257 ip6gre_tnl_parm_from_user(&p1, &p); 1258 t = ip6gre_tunnel_locate(net, &p1, 0); 1259 if (!t) 1260 t = netdev_priv(dev); 1261 } 1262 memset(&p, 0, sizeof(p)); 1263 ip6gre_tnl_parm_to_user(&p, &t->parms); 1264 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1265 err = -EFAULT; 1266 break; 1267 1268 case SIOCADDTUNNEL: 1269 case SIOCCHGTUNNEL: 1270 err = -EPERM; 1271 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1272 goto done; 1273 1274 err = -EFAULT; 1275 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1276 goto done; 1277 1278 err = -EINVAL; 1279 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1280 goto done; 1281 1282 if (!(p.i_flags&GRE_KEY)) 1283 p.i_key = 0; 1284 if (!(p.o_flags&GRE_KEY)) 1285 p.o_key = 0; 1286 1287 ip6gre_tnl_parm_from_user(&p1, &p); 1288 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1289 1290 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1291 if (t) { 1292 if (t->dev != dev) { 1293 err = -EEXIST; 1294 break; 1295 } 1296 } else { 1297 t = netdev_priv(dev); 1298 1299 ip6gre_tunnel_unlink(ign, t); 1300 synchronize_net(); 1301 ip6gre_tnl_change(t, &p1, 1); 1302 ip6gre_tunnel_link(ign, t); 1303 netdev_state_change(dev); 1304 } 1305 } 1306 1307 if (t) { 1308 err = 0; 1309 1310 memset(&p, 0, sizeof(p)); 1311 ip6gre_tnl_parm_to_user(&p, &t->parms); 1312 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1313 err = -EFAULT; 1314 } else 1315 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1316 break; 1317 1318 case SIOCDELTUNNEL: 1319 err = -EPERM; 1320 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1321 goto done; 1322 1323 if (dev == ign->fb_tunnel_dev) { 1324 err = -EFAULT; 1325 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1326 goto done; 1327 err = -ENOENT; 1328 ip6gre_tnl_parm_from_user(&p1, &p); 1329 t = ip6gre_tunnel_locate(net, &p1, 0); 1330 if (!t) 1331 goto done; 1332 err = -EPERM; 1333 if (t == netdev_priv(ign->fb_tunnel_dev)) 1334 goto done; 1335 dev = t->dev; 1336 } 1337 unregister_netdevice(dev); 1338 err = 0; 1339 break; 1340 1341 default: 1342 err = -EINVAL; 1343 } 1344 1345 done: 1346 return err; 1347 } 1348 1349 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1350 unsigned short type, const void *daddr, 1351 const void *saddr, unsigned int len) 1352 { 1353 struct ip6_tnl *t = netdev_priv(dev); 1354 struct ipv6hdr *ipv6h; 1355 __be16 *p; 1356 1357 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1358 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1359 t->fl.u.ip6.flowlabel, 1360 true, &t->fl.u.ip6)); 1361 ipv6h->hop_limit = t->parms.hop_limit; 1362 ipv6h->nexthdr = NEXTHDR_GRE; 1363 ipv6h->saddr = t->parms.laddr; 1364 ipv6h->daddr = t->parms.raddr; 1365 1366 p = (__be16 *)(ipv6h + 1); 1367 p[0] = t->parms.o_flags; 1368 p[1] = htons(type); 1369 1370 /* 1371 * Set the source hardware address. 1372 */ 1373 1374 if (saddr) 1375 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1376 if (daddr) 1377 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1378 if (!ipv6_addr_any(&ipv6h->daddr)) 1379 return t->hlen; 1380 1381 return -t->hlen; 1382 } 1383 1384 static const struct header_ops ip6gre_header_ops = { 1385 .create = ip6gre_header, 1386 }; 1387 1388 static const struct net_device_ops ip6gre_netdev_ops = { 1389 .ndo_init = ip6gre_tunnel_init, 1390 .ndo_uninit = ip6gre_tunnel_uninit, 1391 .ndo_start_xmit = ip6gre_tunnel_xmit, 1392 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1393 .ndo_change_mtu = ip6_tnl_change_mtu, 1394 .ndo_get_stats64 = ip_tunnel_get_stats64, 1395 .ndo_get_iflink = ip6_tnl_get_iflink, 1396 }; 1397 1398 static void ip6gre_dev_free(struct net_device *dev) 1399 { 1400 struct ip6_tnl *t = netdev_priv(dev); 1401 1402 gro_cells_destroy(&t->gro_cells); 1403 dst_cache_destroy(&t->dst_cache); 1404 free_percpu(dev->tstats); 1405 } 1406 1407 static void ip6gre_tunnel_setup(struct net_device *dev) 1408 { 1409 dev->netdev_ops = &ip6gre_netdev_ops; 1410 dev->needs_free_netdev = true; 1411 dev->priv_destructor = ip6gre_dev_free; 1412 1413 dev->type = ARPHRD_IP6GRE; 1414 1415 dev->flags |= IFF_NOARP; 1416 dev->addr_len = sizeof(struct in6_addr); 1417 netif_keep_dst(dev); 1418 /* This perm addr will be used as interface identifier by IPv6 */ 1419 dev->addr_assign_type = NET_ADDR_RANDOM; 1420 eth_random_addr(dev->perm_addr); 1421 } 1422 1423 #define GRE6_FEATURES (NETIF_F_SG | \ 1424 NETIF_F_FRAGLIST | \ 1425 NETIF_F_HIGHDMA | \ 1426 NETIF_F_HW_CSUM) 1427 1428 static void ip6gre_tnl_init_features(struct net_device *dev) 1429 { 1430 struct ip6_tnl *nt = netdev_priv(dev); 1431 1432 dev->features |= GRE6_FEATURES; 1433 dev->hw_features |= GRE6_FEATURES; 1434 1435 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1436 /* TCP offload with GRE SEQ is not supported, nor 1437 * can we support 2 levels of outer headers requiring 1438 * an update. 1439 */ 1440 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1441 nt->encap.type == TUNNEL_ENCAP_NONE) { 1442 dev->features |= NETIF_F_GSO_SOFTWARE; 1443 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1444 } 1445 1446 /* Can use a lockless transmit, unless we generate 1447 * output sequences 1448 */ 1449 dev->features |= NETIF_F_LLTX; 1450 } 1451 } 1452 1453 static int ip6gre_tunnel_init_common(struct net_device *dev) 1454 { 1455 struct ip6_tnl *tunnel; 1456 int ret; 1457 int t_hlen; 1458 1459 tunnel = netdev_priv(dev); 1460 1461 tunnel->dev = dev; 1462 tunnel->net = dev_net(dev); 1463 strcpy(tunnel->parms.name, dev->name); 1464 1465 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1466 if (!dev->tstats) 1467 return -ENOMEM; 1468 1469 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1470 if (ret) 1471 goto cleanup_alloc_pcpu_stats; 1472 1473 ret = gro_cells_init(&tunnel->gro_cells, dev); 1474 if (ret) 1475 goto cleanup_dst_cache_init; 1476 1477 t_hlen = ip6gre_calc_hlen(tunnel); 1478 dev->mtu = ETH_DATA_LEN - t_hlen; 1479 if (dev->type == ARPHRD_ETHER) 1480 dev->mtu -= ETH_HLEN; 1481 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1482 dev->mtu -= 8; 1483 1484 if (tunnel->parms.collect_md) { 1485 netif_keep_dst(dev); 1486 } 1487 ip6gre_tnl_init_features(dev); 1488 1489 return 0; 1490 1491 cleanup_dst_cache_init: 1492 dst_cache_destroy(&tunnel->dst_cache); 1493 cleanup_alloc_pcpu_stats: 1494 free_percpu(dev->tstats); 1495 dev->tstats = NULL; 1496 return ret; 1497 } 1498 1499 static int ip6gre_tunnel_init(struct net_device *dev) 1500 { 1501 struct ip6_tnl *tunnel; 1502 int ret; 1503 1504 ret = ip6gre_tunnel_init_common(dev); 1505 if (ret) 1506 return ret; 1507 1508 tunnel = netdev_priv(dev); 1509 1510 if (tunnel->parms.collect_md) 1511 return 0; 1512 1513 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1514 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1515 1516 if (ipv6_addr_any(&tunnel->parms.raddr)) 1517 dev->header_ops = &ip6gre_header_ops; 1518 1519 return 0; 1520 } 1521 1522 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1523 { 1524 struct ip6_tnl *tunnel = netdev_priv(dev); 1525 1526 tunnel->dev = dev; 1527 tunnel->net = dev_net(dev); 1528 strcpy(tunnel->parms.name, dev->name); 1529 1530 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1531 1532 dev_hold(dev); 1533 } 1534 1535 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1536 .handler = gre_rcv, 1537 .err_handler = ip6gre_err, 1538 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1539 }; 1540 1541 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1542 { 1543 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1544 struct net_device *dev, *aux; 1545 int prio; 1546 1547 for_each_netdev_safe(net, dev, aux) 1548 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1549 dev->rtnl_link_ops == &ip6gre_tap_ops || 1550 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1551 unregister_netdevice_queue(dev, head); 1552 1553 for (prio = 0; prio < 4; prio++) { 1554 int h; 1555 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1556 struct ip6_tnl *t; 1557 1558 t = rtnl_dereference(ign->tunnels[prio][h]); 1559 1560 while (t) { 1561 /* If dev is in the same netns, it has already 1562 * been added to the list by the previous loop. 1563 */ 1564 if (!net_eq(dev_net(t->dev), net)) 1565 unregister_netdevice_queue(t->dev, 1566 head); 1567 t = rtnl_dereference(t->next); 1568 } 1569 } 1570 } 1571 } 1572 1573 static int __net_init ip6gre_init_net(struct net *net) 1574 { 1575 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1576 struct net_device *ndev; 1577 int err; 1578 1579 if (!net_has_fallback_tunnels(net)) 1580 return 0; 1581 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1582 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1583 if (!ndev) { 1584 err = -ENOMEM; 1585 goto err_alloc_dev; 1586 } 1587 ign->fb_tunnel_dev = ndev; 1588 dev_net_set(ign->fb_tunnel_dev, net); 1589 /* FB netdevice is special: we have one, and only one per netns. 1590 * Allowing to move it to another netns is clearly unsafe. 1591 */ 1592 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1593 1594 1595 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1596 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1597 1598 err = register_netdev(ign->fb_tunnel_dev); 1599 if (err) 1600 goto err_reg_dev; 1601 1602 rcu_assign_pointer(ign->tunnels_wc[0], 1603 netdev_priv(ign->fb_tunnel_dev)); 1604 return 0; 1605 1606 err_reg_dev: 1607 free_netdev(ndev); 1608 err_alloc_dev: 1609 return err; 1610 } 1611 1612 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1613 { 1614 struct net *net; 1615 LIST_HEAD(list); 1616 1617 rtnl_lock(); 1618 list_for_each_entry(net, net_list, exit_list) 1619 ip6gre_destroy_tunnels(net, &list); 1620 unregister_netdevice_many(&list); 1621 rtnl_unlock(); 1622 } 1623 1624 static struct pernet_operations ip6gre_net_ops = { 1625 .init = ip6gre_init_net, 1626 .exit_batch = ip6gre_exit_batch_net, 1627 .id = &ip6gre_net_id, 1628 .size = sizeof(struct ip6gre_net), 1629 }; 1630 1631 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1632 struct netlink_ext_ack *extack) 1633 { 1634 __be16 flags; 1635 1636 if (!data) 1637 return 0; 1638 1639 flags = 0; 1640 if (data[IFLA_GRE_IFLAGS]) 1641 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1642 if (data[IFLA_GRE_OFLAGS]) 1643 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1644 if (flags & (GRE_VERSION|GRE_ROUTING)) 1645 return -EINVAL; 1646 1647 return 0; 1648 } 1649 1650 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1651 struct netlink_ext_ack *extack) 1652 { 1653 struct in6_addr daddr; 1654 1655 if (tb[IFLA_ADDRESS]) { 1656 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1657 return -EINVAL; 1658 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1659 return -EADDRNOTAVAIL; 1660 } 1661 1662 if (!data) 1663 goto out; 1664 1665 if (data[IFLA_GRE_REMOTE]) { 1666 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1667 if (ipv6_addr_any(&daddr)) 1668 return -EINVAL; 1669 } 1670 1671 out: 1672 return ip6gre_tunnel_validate(tb, data, extack); 1673 } 1674 1675 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1676 struct netlink_ext_ack *extack) 1677 { 1678 __be16 flags = 0; 1679 int ret, ver = 0; 1680 1681 if (!data) 1682 return 0; 1683 1684 ret = ip6gre_tap_validate(tb, data, extack); 1685 if (ret) 1686 return ret; 1687 1688 /* ERSPAN should only have GRE sequence and key flag */ 1689 if (data[IFLA_GRE_OFLAGS]) 1690 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1691 if (data[IFLA_GRE_IFLAGS]) 1692 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1693 if (!data[IFLA_GRE_COLLECT_METADATA] && 1694 flags != (GRE_SEQ | GRE_KEY)) 1695 return -EINVAL; 1696 1697 /* ERSPAN Session ID only has 10-bit. Since we reuse 1698 * 32-bit key field as ID, check it's range. 1699 */ 1700 if (data[IFLA_GRE_IKEY] && 1701 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1702 return -EINVAL; 1703 1704 if (data[IFLA_GRE_OKEY] && 1705 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1706 return -EINVAL; 1707 1708 if (data[IFLA_GRE_ERSPAN_VER]) { 1709 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1710 if (ver != 1 && ver != 2) 1711 return -EINVAL; 1712 } 1713 1714 if (ver == 1) { 1715 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1716 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1717 1718 if (index & ~INDEX_MASK) 1719 return -EINVAL; 1720 } 1721 } else if (ver == 2) { 1722 if (data[IFLA_GRE_ERSPAN_DIR]) { 1723 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1724 1725 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1726 return -EINVAL; 1727 } 1728 1729 if (data[IFLA_GRE_ERSPAN_HWID]) { 1730 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1731 1732 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1733 return -EINVAL; 1734 } 1735 } 1736 1737 return 0; 1738 } 1739 1740 static void ip6erspan_set_version(struct nlattr *data[], 1741 struct __ip6_tnl_parm *parms) 1742 { 1743 if (!data) 1744 return; 1745 1746 parms->erspan_ver = 1; 1747 if (data[IFLA_GRE_ERSPAN_VER]) 1748 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1749 1750 if (parms->erspan_ver == 1) { 1751 if (data[IFLA_GRE_ERSPAN_INDEX]) 1752 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1753 } else if (parms->erspan_ver == 2) { 1754 if (data[IFLA_GRE_ERSPAN_DIR]) 1755 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1756 if (data[IFLA_GRE_ERSPAN_HWID]) 1757 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1758 } 1759 } 1760 1761 static void ip6gre_netlink_parms(struct nlattr *data[], 1762 struct __ip6_tnl_parm *parms) 1763 { 1764 memset(parms, 0, sizeof(*parms)); 1765 1766 if (!data) 1767 return; 1768 1769 if (data[IFLA_GRE_LINK]) 1770 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1771 1772 if (data[IFLA_GRE_IFLAGS]) 1773 parms->i_flags = gre_flags_to_tnl_flags( 1774 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1775 1776 if (data[IFLA_GRE_OFLAGS]) 1777 parms->o_flags = gre_flags_to_tnl_flags( 1778 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1779 1780 if (data[IFLA_GRE_IKEY]) 1781 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1782 1783 if (data[IFLA_GRE_OKEY]) 1784 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1785 1786 if (data[IFLA_GRE_LOCAL]) 1787 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1788 1789 if (data[IFLA_GRE_REMOTE]) 1790 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1791 1792 if (data[IFLA_GRE_TTL]) 1793 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1794 1795 if (data[IFLA_GRE_ENCAP_LIMIT]) 1796 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1797 1798 if (data[IFLA_GRE_FLOWINFO]) 1799 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1800 1801 if (data[IFLA_GRE_FLAGS]) 1802 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1803 1804 if (data[IFLA_GRE_FWMARK]) 1805 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1806 1807 if (data[IFLA_GRE_COLLECT_METADATA]) 1808 parms->collect_md = true; 1809 } 1810 1811 static int ip6gre_tap_init(struct net_device *dev) 1812 { 1813 int ret; 1814 1815 ret = ip6gre_tunnel_init_common(dev); 1816 if (ret) 1817 return ret; 1818 1819 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1820 1821 return 0; 1822 } 1823 1824 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1825 .ndo_init = ip6gre_tap_init, 1826 .ndo_uninit = ip6gre_tunnel_uninit, 1827 .ndo_start_xmit = ip6gre_tunnel_xmit, 1828 .ndo_set_mac_address = eth_mac_addr, 1829 .ndo_validate_addr = eth_validate_addr, 1830 .ndo_change_mtu = ip6_tnl_change_mtu, 1831 .ndo_get_stats64 = ip_tunnel_get_stats64, 1832 .ndo_get_iflink = ip6_tnl_get_iflink, 1833 }; 1834 1835 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1836 { 1837 int t_hlen; 1838 1839 tunnel->tun_hlen = 8; 1840 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1841 erspan_hdr_len(tunnel->parms.erspan_ver); 1842 1843 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1844 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1845 return t_hlen; 1846 } 1847 1848 static int ip6erspan_tap_init(struct net_device *dev) 1849 { 1850 struct ip6_tnl *tunnel; 1851 int t_hlen; 1852 int ret; 1853 1854 tunnel = netdev_priv(dev); 1855 1856 tunnel->dev = dev; 1857 tunnel->net = dev_net(dev); 1858 strcpy(tunnel->parms.name, dev->name); 1859 1860 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1861 if (!dev->tstats) 1862 return -ENOMEM; 1863 1864 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1865 if (ret) 1866 goto cleanup_alloc_pcpu_stats; 1867 1868 ret = gro_cells_init(&tunnel->gro_cells, dev); 1869 if (ret) 1870 goto cleanup_dst_cache_init; 1871 1872 t_hlen = ip6erspan_calc_hlen(tunnel); 1873 dev->mtu = ETH_DATA_LEN - t_hlen; 1874 if (dev->type == ARPHRD_ETHER) 1875 dev->mtu -= ETH_HLEN; 1876 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1877 dev->mtu -= 8; 1878 1879 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1880 ip6erspan_tnl_link_config(tunnel, 1); 1881 1882 return 0; 1883 1884 cleanup_dst_cache_init: 1885 dst_cache_destroy(&tunnel->dst_cache); 1886 cleanup_alloc_pcpu_stats: 1887 free_percpu(dev->tstats); 1888 dev->tstats = NULL; 1889 return ret; 1890 } 1891 1892 static const struct net_device_ops ip6erspan_netdev_ops = { 1893 .ndo_init = ip6erspan_tap_init, 1894 .ndo_uninit = ip6erspan_tunnel_uninit, 1895 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1896 .ndo_set_mac_address = eth_mac_addr, 1897 .ndo_validate_addr = eth_validate_addr, 1898 .ndo_change_mtu = ip6_tnl_change_mtu, 1899 .ndo_get_stats64 = ip_tunnel_get_stats64, 1900 .ndo_get_iflink = ip6_tnl_get_iflink, 1901 }; 1902 1903 static void ip6gre_tap_setup(struct net_device *dev) 1904 { 1905 1906 ether_setup(dev); 1907 1908 dev->max_mtu = 0; 1909 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1910 dev->needs_free_netdev = true; 1911 dev->priv_destructor = ip6gre_dev_free; 1912 1913 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1914 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1915 netif_keep_dst(dev); 1916 } 1917 1918 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1919 struct ip_tunnel_encap *ipencap) 1920 { 1921 bool ret = false; 1922 1923 memset(ipencap, 0, sizeof(*ipencap)); 1924 1925 if (!data) 1926 return ret; 1927 1928 if (data[IFLA_GRE_ENCAP_TYPE]) { 1929 ret = true; 1930 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1931 } 1932 1933 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1934 ret = true; 1935 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1936 } 1937 1938 if (data[IFLA_GRE_ENCAP_SPORT]) { 1939 ret = true; 1940 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1941 } 1942 1943 if (data[IFLA_GRE_ENCAP_DPORT]) { 1944 ret = true; 1945 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1946 } 1947 1948 return ret; 1949 } 1950 1951 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1952 struct nlattr *tb[], struct nlattr *data[], 1953 struct netlink_ext_ack *extack) 1954 { 1955 struct ip6_tnl *nt; 1956 struct ip_tunnel_encap ipencap; 1957 int err; 1958 1959 nt = netdev_priv(dev); 1960 1961 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1962 int err = ip6_tnl_encap_setup(nt, &ipencap); 1963 1964 if (err < 0) 1965 return err; 1966 } 1967 1968 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1969 eth_hw_addr_random(dev); 1970 1971 nt->dev = dev; 1972 nt->net = dev_net(dev); 1973 1974 err = register_netdevice(dev); 1975 if (err) 1976 goto out; 1977 1978 if (tb[IFLA_MTU]) 1979 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1980 1981 dev_hold(dev); 1982 1983 out: 1984 return err; 1985 } 1986 1987 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1988 struct nlattr *tb[], struct nlattr *data[], 1989 struct netlink_ext_ack *extack) 1990 { 1991 struct ip6_tnl *nt = netdev_priv(dev); 1992 struct net *net = dev_net(dev); 1993 struct ip6gre_net *ign; 1994 int err; 1995 1996 ip6gre_netlink_parms(data, &nt->parms); 1997 ign = net_generic(net, ip6gre_net_id); 1998 1999 if (nt->parms.collect_md) { 2000 if (rtnl_dereference(ign->collect_md_tun)) 2001 return -EEXIST; 2002 } else { 2003 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2004 return -EEXIST; 2005 } 2006 2007 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2008 if (!err) { 2009 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2010 ip6gre_tunnel_link_md(ign, nt); 2011 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2012 } 2013 return err; 2014 } 2015 2016 static struct ip6_tnl * 2017 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2018 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2019 struct netlink_ext_ack *extack) 2020 { 2021 struct ip6_tnl *t, *nt = netdev_priv(dev); 2022 struct net *net = nt->net; 2023 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2024 struct ip_tunnel_encap ipencap; 2025 2026 if (dev == ign->fb_tunnel_dev) 2027 return ERR_PTR(-EINVAL); 2028 2029 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2030 int err = ip6_tnl_encap_setup(nt, &ipencap); 2031 2032 if (err < 0) 2033 return ERR_PTR(err); 2034 } 2035 2036 ip6gre_netlink_parms(data, p_p); 2037 2038 t = ip6gre_tunnel_locate(net, p_p, 0); 2039 2040 if (t) { 2041 if (t->dev != dev) 2042 return ERR_PTR(-EEXIST); 2043 } else { 2044 t = nt; 2045 } 2046 2047 return t; 2048 } 2049 2050 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2051 struct nlattr *data[], 2052 struct netlink_ext_ack *extack) 2053 { 2054 struct ip6_tnl *t = netdev_priv(dev); 2055 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2056 struct __ip6_tnl_parm p; 2057 2058 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2059 if (IS_ERR(t)) 2060 return PTR_ERR(t); 2061 2062 ip6gre_tunnel_unlink_md(ign, t); 2063 ip6gre_tunnel_unlink(ign, t); 2064 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2065 ip6gre_tunnel_link_md(ign, t); 2066 ip6gre_tunnel_link(ign, t); 2067 return 0; 2068 } 2069 2070 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2071 { 2072 struct net *net = dev_net(dev); 2073 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2074 2075 if (dev != ign->fb_tunnel_dev) 2076 unregister_netdevice_queue(dev, head); 2077 } 2078 2079 static size_t ip6gre_get_size(const struct net_device *dev) 2080 { 2081 return 2082 /* IFLA_GRE_LINK */ 2083 nla_total_size(4) + 2084 /* IFLA_GRE_IFLAGS */ 2085 nla_total_size(2) + 2086 /* IFLA_GRE_OFLAGS */ 2087 nla_total_size(2) + 2088 /* IFLA_GRE_IKEY */ 2089 nla_total_size(4) + 2090 /* IFLA_GRE_OKEY */ 2091 nla_total_size(4) + 2092 /* IFLA_GRE_LOCAL */ 2093 nla_total_size(sizeof(struct in6_addr)) + 2094 /* IFLA_GRE_REMOTE */ 2095 nla_total_size(sizeof(struct in6_addr)) + 2096 /* IFLA_GRE_TTL */ 2097 nla_total_size(1) + 2098 /* IFLA_GRE_ENCAP_LIMIT */ 2099 nla_total_size(1) + 2100 /* IFLA_GRE_FLOWINFO */ 2101 nla_total_size(4) + 2102 /* IFLA_GRE_FLAGS */ 2103 nla_total_size(4) + 2104 /* IFLA_GRE_ENCAP_TYPE */ 2105 nla_total_size(2) + 2106 /* IFLA_GRE_ENCAP_FLAGS */ 2107 nla_total_size(2) + 2108 /* IFLA_GRE_ENCAP_SPORT */ 2109 nla_total_size(2) + 2110 /* IFLA_GRE_ENCAP_DPORT */ 2111 nla_total_size(2) + 2112 /* IFLA_GRE_COLLECT_METADATA */ 2113 nla_total_size(0) + 2114 /* IFLA_GRE_FWMARK */ 2115 nla_total_size(4) + 2116 /* IFLA_GRE_ERSPAN_INDEX */ 2117 nla_total_size(4) + 2118 0; 2119 } 2120 2121 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2122 { 2123 struct ip6_tnl *t = netdev_priv(dev); 2124 struct __ip6_tnl_parm *p = &t->parms; 2125 __be16 o_flags = p->o_flags; 2126 2127 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2128 if (!p->collect_md) 2129 o_flags |= TUNNEL_KEY; 2130 2131 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2132 goto nla_put_failure; 2133 2134 if (p->erspan_ver == 1) { 2135 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2136 goto nla_put_failure; 2137 } else { 2138 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2139 goto nla_put_failure; 2140 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2141 goto nla_put_failure; 2142 } 2143 } 2144 2145 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2146 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2147 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2148 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2149 gre_tnl_flags_to_gre_flags(o_flags)) || 2150 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2151 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2152 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2153 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2154 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2155 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2156 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2157 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2158 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2159 goto nla_put_failure; 2160 2161 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2162 t->encap.type) || 2163 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2164 t->encap.sport) || 2165 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2166 t->encap.dport) || 2167 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2168 t->encap.flags)) 2169 goto nla_put_failure; 2170 2171 if (p->collect_md) { 2172 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2173 goto nla_put_failure; 2174 } 2175 2176 return 0; 2177 2178 nla_put_failure: 2179 return -EMSGSIZE; 2180 } 2181 2182 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2183 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2184 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2185 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2186 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2187 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2188 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2189 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2190 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2191 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2192 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2193 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2194 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2195 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2196 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2197 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2198 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2199 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2200 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2201 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2202 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2203 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2204 }; 2205 2206 static void ip6erspan_tap_setup(struct net_device *dev) 2207 { 2208 ether_setup(dev); 2209 2210 dev->max_mtu = 0; 2211 dev->netdev_ops = &ip6erspan_netdev_ops; 2212 dev->needs_free_netdev = true; 2213 dev->priv_destructor = ip6gre_dev_free; 2214 2215 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2216 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2217 netif_keep_dst(dev); 2218 } 2219 2220 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2221 struct nlattr *tb[], struct nlattr *data[], 2222 struct netlink_ext_ack *extack) 2223 { 2224 struct ip6_tnl *nt = netdev_priv(dev); 2225 struct net *net = dev_net(dev); 2226 struct ip6gre_net *ign; 2227 int err; 2228 2229 ip6gre_netlink_parms(data, &nt->parms); 2230 ip6erspan_set_version(data, &nt->parms); 2231 ign = net_generic(net, ip6gre_net_id); 2232 2233 if (nt->parms.collect_md) { 2234 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2235 return -EEXIST; 2236 } else { 2237 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2238 return -EEXIST; 2239 } 2240 2241 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2242 if (!err) { 2243 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2244 ip6erspan_tunnel_link_md(ign, nt); 2245 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2246 } 2247 return err; 2248 } 2249 2250 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2251 { 2252 ip6gre_tnl_link_config_common(t); 2253 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2254 } 2255 2256 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2257 const struct __ip6_tnl_parm *p, int set_mtu) 2258 { 2259 ip6gre_tnl_copy_tnl_parm(t, p); 2260 ip6erspan_tnl_link_config(t, set_mtu); 2261 return 0; 2262 } 2263 2264 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2265 struct nlattr *data[], 2266 struct netlink_ext_ack *extack) 2267 { 2268 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2269 struct __ip6_tnl_parm p; 2270 struct ip6_tnl *t; 2271 2272 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2273 if (IS_ERR(t)) 2274 return PTR_ERR(t); 2275 2276 ip6erspan_set_version(data, &p); 2277 ip6gre_tunnel_unlink_md(ign, t); 2278 ip6gre_tunnel_unlink(ign, t); 2279 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2280 ip6erspan_tunnel_link_md(ign, t); 2281 ip6gre_tunnel_link(ign, t); 2282 return 0; 2283 } 2284 2285 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2286 .kind = "ip6gre", 2287 .maxtype = IFLA_GRE_MAX, 2288 .policy = ip6gre_policy, 2289 .priv_size = sizeof(struct ip6_tnl), 2290 .setup = ip6gre_tunnel_setup, 2291 .validate = ip6gre_tunnel_validate, 2292 .newlink = ip6gre_newlink, 2293 .changelink = ip6gre_changelink, 2294 .dellink = ip6gre_dellink, 2295 .get_size = ip6gre_get_size, 2296 .fill_info = ip6gre_fill_info, 2297 .get_link_net = ip6_tnl_get_link_net, 2298 }; 2299 2300 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2301 .kind = "ip6gretap", 2302 .maxtype = IFLA_GRE_MAX, 2303 .policy = ip6gre_policy, 2304 .priv_size = sizeof(struct ip6_tnl), 2305 .setup = ip6gre_tap_setup, 2306 .validate = ip6gre_tap_validate, 2307 .newlink = ip6gre_newlink, 2308 .changelink = ip6gre_changelink, 2309 .get_size = ip6gre_get_size, 2310 .fill_info = ip6gre_fill_info, 2311 .get_link_net = ip6_tnl_get_link_net, 2312 }; 2313 2314 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2315 .kind = "ip6erspan", 2316 .maxtype = IFLA_GRE_MAX, 2317 .policy = ip6gre_policy, 2318 .priv_size = sizeof(struct ip6_tnl), 2319 .setup = ip6erspan_tap_setup, 2320 .validate = ip6erspan_tap_validate, 2321 .newlink = ip6erspan_newlink, 2322 .changelink = ip6erspan_changelink, 2323 .get_size = ip6gre_get_size, 2324 .fill_info = ip6gre_fill_info, 2325 .get_link_net = ip6_tnl_get_link_net, 2326 }; 2327 2328 /* 2329 * And now the modules code and kernel interface. 2330 */ 2331 2332 static int __init ip6gre_init(void) 2333 { 2334 int err; 2335 2336 pr_info("GRE over IPv6 tunneling driver\n"); 2337 2338 err = register_pernet_device(&ip6gre_net_ops); 2339 if (err < 0) 2340 return err; 2341 2342 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2343 if (err < 0) { 2344 pr_info("%s: can't add protocol\n", __func__); 2345 goto add_proto_failed; 2346 } 2347 2348 err = rtnl_link_register(&ip6gre_link_ops); 2349 if (err < 0) 2350 goto rtnl_link_failed; 2351 2352 err = rtnl_link_register(&ip6gre_tap_ops); 2353 if (err < 0) 2354 goto tap_ops_failed; 2355 2356 err = rtnl_link_register(&ip6erspan_tap_ops); 2357 if (err < 0) 2358 goto erspan_link_failed; 2359 2360 out: 2361 return err; 2362 2363 erspan_link_failed: 2364 rtnl_link_unregister(&ip6gre_tap_ops); 2365 tap_ops_failed: 2366 rtnl_link_unregister(&ip6gre_link_ops); 2367 rtnl_link_failed: 2368 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2369 add_proto_failed: 2370 unregister_pernet_device(&ip6gre_net_ops); 2371 goto out; 2372 } 2373 2374 static void __exit ip6gre_fini(void) 2375 { 2376 rtnl_link_unregister(&ip6gre_tap_ops); 2377 rtnl_link_unregister(&ip6gre_link_ops); 2378 rtnl_link_unregister(&ip6erspan_tap_ops); 2379 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2380 unregister_pernet_device(&ip6gre_net_ops); 2381 } 2382 2383 module_init(ip6gre_init); 2384 module_exit(ip6gre_fini); 2385 MODULE_LICENSE("GPL"); 2386 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2387 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2388 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2389 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2390 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2391 MODULE_ALIAS_NETDEV("ip6gre0"); 2392