1 /* 2 * GRE over IPv6 protocol decoder. 3 * 4 * Authors: Dmitry Kozlov (xeb@mail.ru) 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License 8 * as published by the Free Software Foundation; either version 9 * 2 of the License, or (at your option) any later version. 10 * 11 */ 12 13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 14 15 #include <linux/capability.h> 16 #include <linux/module.h> 17 #include <linux/types.h> 18 #include <linux/kernel.h> 19 #include <linux/slab.h> 20 #include <linux/uaccess.h> 21 #include <linux/skbuff.h> 22 #include <linux/netdevice.h> 23 #include <linux/in.h> 24 #include <linux/tcp.h> 25 #include <linux/udp.h> 26 #include <linux/if_arp.h> 27 #include <linux/init.h> 28 #include <linux/in6.h> 29 #include <linux/inetdevice.h> 30 #include <linux/igmp.h> 31 #include <linux/netfilter_ipv4.h> 32 #include <linux/etherdevice.h> 33 #include <linux/if_ether.h> 34 #include <linux/hash.h> 35 #include <linux/if_tunnel.h> 36 #include <linux/ip6_tunnel.h> 37 38 #include <net/sock.h> 39 #include <net/ip.h> 40 #include <net/ip_tunnels.h> 41 #include <net/icmp.h> 42 #include <net/protocol.h> 43 #include <net/addrconf.h> 44 #include <net/arp.h> 45 #include <net/checksum.h> 46 #include <net/dsfield.h> 47 #include <net/inet_ecn.h> 48 #include <net/xfrm.h> 49 #include <net/net_namespace.h> 50 #include <net/netns/generic.h> 51 #include <net/rtnetlink.h> 52 53 #include <net/ipv6.h> 54 #include <net/ip6_fib.h> 55 #include <net/ip6_route.h> 56 #include <net/ip6_tunnel.h> 57 #include <net/gre.h> 58 #include <net/erspan.h> 59 #include <net/dst_metadata.h> 60 61 62 static bool log_ecn_error = true; 63 module_param(log_ecn_error, bool, 0644); 64 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 65 66 #define IP6_GRE_HASH_SIZE_SHIFT 5 67 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT) 68 69 static unsigned int ip6gre_net_id __read_mostly; 70 struct ip6gre_net { 71 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE]; 72 73 struct ip6_tnl __rcu *collect_md_tun; 74 struct ip6_tnl __rcu *collect_md_tun_erspan; 75 struct net_device *fb_tunnel_dev; 76 }; 77 78 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 79 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 80 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly; 81 static int ip6gre_tunnel_init(struct net_device *dev); 82 static void ip6gre_tunnel_setup(struct net_device *dev); 83 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t); 84 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu); 85 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu); 86 87 /* Tunnel hash table */ 88 89 /* 90 4 hash tables: 91 92 3: (remote,local) 93 2: (remote,*) 94 1: (*,local) 95 0: (*,*) 96 97 We require exact key match i.e. if a key is present in packet 98 it will match only tunnel with the same key; if it is not present, 99 it will match only keyless tunnel. 100 101 All keysless packets, if not matched configured keyless tunnels 102 will match fallback tunnel. 103 */ 104 105 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 106 static u32 HASH_ADDR(const struct in6_addr *addr) 107 { 108 u32 hash = ipv6_addr_hash(addr); 109 110 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 111 } 112 113 #define tunnels_r_l tunnels[3] 114 #define tunnels_r tunnels[2] 115 #define tunnels_l tunnels[1] 116 #define tunnels_wc tunnels[0] 117 118 /* Given src, dst and key, find appropriate for input tunnel. */ 119 120 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 121 const struct in6_addr *remote, const struct in6_addr *local, 122 __be32 key, __be16 gre_proto) 123 { 124 struct net *net = dev_net(dev); 125 int link = dev->ifindex; 126 unsigned int h0 = HASH_ADDR(remote); 127 unsigned int h1 = HASH_KEY(key); 128 struct ip6_tnl *t, *cand = NULL; 129 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 130 int dev_type = (gre_proto == htons(ETH_P_TEB) || 131 gre_proto == htons(ETH_P_ERSPAN) || 132 gre_proto == htons(ETH_P_ERSPAN2)) ? 133 ARPHRD_ETHER : ARPHRD_IP6GRE; 134 int score, cand_score = 4; 135 136 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 137 if (!ipv6_addr_equal(local, &t->parms.laddr) || 138 !ipv6_addr_equal(remote, &t->parms.raddr) || 139 key != t->parms.i_key || 140 !(t->dev->flags & IFF_UP)) 141 continue; 142 143 if (t->dev->type != ARPHRD_IP6GRE && 144 t->dev->type != dev_type) 145 continue; 146 147 score = 0; 148 if (t->parms.link != link) 149 score |= 1; 150 if (t->dev->type != dev_type) 151 score |= 2; 152 if (score == 0) 153 return t; 154 155 if (score < cand_score) { 156 cand = t; 157 cand_score = score; 158 } 159 } 160 161 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 162 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 163 key != t->parms.i_key || 164 !(t->dev->flags & IFF_UP)) 165 continue; 166 167 if (t->dev->type != ARPHRD_IP6GRE && 168 t->dev->type != dev_type) 169 continue; 170 171 score = 0; 172 if (t->parms.link != link) 173 score |= 1; 174 if (t->dev->type != dev_type) 175 score |= 2; 176 if (score == 0) 177 return t; 178 179 if (score < cand_score) { 180 cand = t; 181 cand_score = score; 182 } 183 } 184 185 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 186 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 187 (!ipv6_addr_equal(local, &t->parms.raddr) || 188 !ipv6_addr_is_multicast(local))) || 189 key != t->parms.i_key || 190 !(t->dev->flags & IFF_UP)) 191 continue; 192 193 if (t->dev->type != ARPHRD_IP6GRE && 194 t->dev->type != dev_type) 195 continue; 196 197 score = 0; 198 if (t->parms.link != link) 199 score |= 1; 200 if (t->dev->type != dev_type) 201 score |= 2; 202 if (score == 0) 203 return t; 204 205 if (score < cand_score) { 206 cand = t; 207 cand_score = score; 208 } 209 } 210 211 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 212 if (t->parms.i_key != key || 213 !(t->dev->flags & IFF_UP)) 214 continue; 215 216 if (t->dev->type != ARPHRD_IP6GRE && 217 t->dev->type != dev_type) 218 continue; 219 220 score = 0; 221 if (t->parms.link != link) 222 score |= 1; 223 if (t->dev->type != dev_type) 224 score |= 2; 225 if (score == 0) 226 return t; 227 228 if (score < cand_score) { 229 cand = t; 230 cand_score = score; 231 } 232 } 233 234 if (cand) 235 return cand; 236 237 if (gre_proto == htons(ETH_P_ERSPAN) || 238 gre_proto == htons(ETH_P_ERSPAN2)) 239 t = rcu_dereference(ign->collect_md_tun_erspan); 240 else 241 t = rcu_dereference(ign->collect_md_tun); 242 243 if (t && t->dev->flags & IFF_UP) 244 return t; 245 246 dev = ign->fb_tunnel_dev; 247 if (dev && dev->flags & IFF_UP) 248 return netdev_priv(dev); 249 250 return NULL; 251 } 252 253 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 254 const struct __ip6_tnl_parm *p) 255 { 256 const struct in6_addr *remote = &p->raddr; 257 const struct in6_addr *local = &p->laddr; 258 unsigned int h = HASH_KEY(p->i_key); 259 int prio = 0; 260 261 if (!ipv6_addr_any(local)) 262 prio |= 1; 263 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 264 prio |= 2; 265 h ^= HASH_ADDR(remote); 266 } 267 268 return &ign->tunnels[prio][h]; 269 } 270 271 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 272 { 273 if (t->parms.collect_md) 274 rcu_assign_pointer(ign->collect_md_tun, t); 275 } 276 277 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 278 { 279 if (t->parms.collect_md) 280 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 281 } 282 283 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 284 { 285 if (t->parms.collect_md) 286 rcu_assign_pointer(ign->collect_md_tun, NULL); 287 } 288 289 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 290 struct ip6_tnl *t) 291 { 292 if (t->parms.collect_md) 293 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 294 } 295 296 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 297 const struct ip6_tnl *t) 298 { 299 return __ip6gre_bucket(ign, &t->parms); 300 } 301 302 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 303 { 304 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 305 306 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 307 rcu_assign_pointer(*tp, t); 308 } 309 310 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 311 { 312 struct ip6_tnl __rcu **tp; 313 struct ip6_tnl *iter; 314 315 for (tp = ip6gre_bucket(ign, t); 316 (iter = rtnl_dereference(*tp)) != NULL; 317 tp = &iter->next) { 318 if (t == iter) { 319 rcu_assign_pointer(*tp, t->next); 320 break; 321 } 322 } 323 } 324 325 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 326 const struct __ip6_tnl_parm *parms, 327 int type) 328 { 329 const struct in6_addr *remote = &parms->raddr; 330 const struct in6_addr *local = &parms->laddr; 331 __be32 key = parms->i_key; 332 int link = parms->link; 333 struct ip6_tnl *t; 334 struct ip6_tnl __rcu **tp; 335 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 336 337 for (tp = __ip6gre_bucket(ign, parms); 338 (t = rtnl_dereference(*tp)) != NULL; 339 tp = &t->next) 340 if (ipv6_addr_equal(local, &t->parms.laddr) && 341 ipv6_addr_equal(remote, &t->parms.raddr) && 342 key == t->parms.i_key && 343 link == t->parms.link && 344 type == t->dev->type) 345 break; 346 347 return t; 348 } 349 350 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 351 const struct __ip6_tnl_parm *parms, int create) 352 { 353 struct ip6_tnl *t, *nt; 354 struct net_device *dev; 355 char name[IFNAMSIZ]; 356 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 357 358 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 359 if (t && create) 360 return NULL; 361 if (t || !create) 362 return t; 363 364 if (parms->name[0]) { 365 if (!dev_valid_name(parms->name)) 366 return NULL; 367 strlcpy(name, parms->name, IFNAMSIZ); 368 } else { 369 strcpy(name, "ip6gre%d"); 370 } 371 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 372 ip6gre_tunnel_setup); 373 if (!dev) 374 return NULL; 375 376 dev_net_set(dev, net); 377 378 nt = netdev_priv(dev); 379 nt->parms = *parms; 380 dev->rtnl_link_ops = &ip6gre_link_ops; 381 382 nt->dev = dev; 383 nt->net = dev_net(dev); 384 385 if (register_netdevice(dev) < 0) 386 goto failed_free; 387 388 ip6gre_tnl_link_config(nt, 1); 389 390 /* Can use a lockless transmit, unless we generate output sequences */ 391 if (!(nt->parms.o_flags & TUNNEL_SEQ)) 392 dev->features |= NETIF_F_LLTX; 393 394 dev_hold(dev); 395 ip6gre_tunnel_link(ign, nt); 396 return nt; 397 398 failed_free: 399 free_netdev(dev); 400 return NULL; 401 } 402 403 static void ip6erspan_tunnel_uninit(struct net_device *dev) 404 { 405 struct ip6_tnl *t = netdev_priv(dev); 406 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 407 408 ip6erspan_tunnel_unlink_md(ign, t); 409 ip6gre_tunnel_unlink(ign, t); 410 dst_cache_reset(&t->dst_cache); 411 dev_put(dev); 412 } 413 414 static void ip6gre_tunnel_uninit(struct net_device *dev) 415 { 416 struct ip6_tnl *t = netdev_priv(dev); 417 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 418 419 ip6gre_tunnel_unlink_md(ign, t); 420 ip6gre_tunnel_unlink(ign, t); 421 dst_cache_reset(&t->dst_cache); 422 dev_put(dev); 423 } 424 425 426 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 427 u8 type, u8 code, int offset, __be32 info) 428 { 429 struct net *net = dev_net(skb->dev); 430 const struct ipv6hdr *ipv6h; 431 struct tnl_ptk_info tpi; 432 struct ip6_tnl *t; 433 434 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 435 offset) < 0) 436 return -EINVAL; 437 438 ipv6h = (const struct ipv6hdr *)skb->data; 439 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 440 tpi.key, tpi.proto); 441 if (!t) 442 return -ENOENT; 443 444 switch (type) { 445 struct ipv6_tlv_tnl_enc_lim *tel; 446 __u32 teli; 447 case ICMPV6_DEST_UNREACH: 448 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 449 t->parms.name); 450 if (code != ICMPV6_PORT_UNREACH) 451 break; 452 return 0; 453 case ICMPV6_TIME_EXCEED: 454 if (code == ICMPV6_EXC_HOPLIMIT) { 455 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 456 t->parms.name); 457 break; 458 } 459 return 0; 460 case ICMPV6_PARAMPROB: 461 teli = 0; 462 if (code == ICMPV6_HDR_FIELD) 463 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 464 465 if (teli && teli == be32_to_cpu(info) - 2) { 466 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 467 if (tel->encap_limit == 0) { 468 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 469 t->parms.name); 470 } 471 } else { 472 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 473 t->parms.name); 474 } 475 return 0; 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, int gre_hdr_len, 528 struct tnl_ptk_info *tpi) 529 { 530 struct erspan_base_hdr *ershdr; 531 struct erspan_metadata *pkt_md; 532 const struct ipv6hdr *ipv6h; 533 struct erspan_md2 *md2; 534 struct ip6_tnl *tunnel; 535 u8 ver; 536 537 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr)))) 538 return PACKET_REJECT; 539 540 ipv6h = ipv6_hdr(skb); 541 ershdr = (struct erspan_base_hdr *)skb->data; 542 ver = ershdr->ver; 543 tpi->key = cpu_to_be32(get_session_id(ershdr)); 544 545 tunnel = ip6gre_tunnel_lookup(skb->dev, 546 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 547 tpi->proto); 548 if (tunnel) { 549 int len = erspan_hdr_len(ver); 550 551 if (unlikely(!pskb_may_pull(skb, len))) 552 return PACKET_REJECT; 553 554 ershdr = (struct erspan_base_hdr *)skb->data; 555 pkt_md = (struct erspan_metadata *)(ershdr + 1); 556 557 if (__iptunnel_pull_header(skb, len, 558 htons(ETH_P_TEB), 559 false, false) < 0) 560 return PACKET_REJECT; 561 562 if (tunnel->parms.collect_md) { 563 struct metadata_dst *tun_dst; 564 struct ip_tunnel_info *info; 565 struct erspan_metadata *md; 566 __be64 tun_id; 567 __be16 flags; 568 569 tpi->flags |= TUNNEL_KEY; 570 flags = tpi->flags; 571 tun_id = key32_to_tunnel_id(tpi->key); 572 573 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 574 sizeof(*md)); 575 if (!tun_dst) 576 return PACKET_REJECT; 577 578 info = &tun_dst->u.tun_info; 579 md = ip_tunnel_info_opts(info); 580 md->version = ver; 581 md2 = &md->u.md2; 582 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 583 ERSPAN_V2_MDSIZE); 584 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 585 info->options_len = sizeof(*md); 586 587 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 588 589 } else { 590 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 591 } 592 593 return PACKET_RCVD; 594 } 595 596 return PACKET_REJECT; 597 } 598 599 static int gre_rcv(struct sk_buff *skb) 600 { 601 struct tnl_ptk_info tpi; 602 bool csum_err = false; 603 int hdr_len; 604 605 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 606 if (hdr_len < 0) 607 goto drop; 608 609 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 610 goto drop; 611 612 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 613 tpi.proto == htons(ETH_P_ERSPAN2))) { 614 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD) 615 return 0; 616 goto out; 617 } 618 619 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 620 return 0; 621 622 out: 623 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 624 drop: 625 kfree_skb(skb); 626 return 0; 627 } 628 629 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 630 { 631 return iptunnel_handle_offloads(skb, 632 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 633 } 634 635 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 636 struct net_device *dev, 637 struct flowi6 *fl6, __u8 *dsfield, 638 int *encap_limit) 639 { 640 const struct iphdr *iph = ip_hdr(skb); 641 struct ip6_tnl *t = netdev_priv(dev); 642 643 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 644 *encap_limit = t->parms.encap_limit; 645 646 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 647 648 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 649 *dsfield = ipv4_get_dsfield(iph); 650 else 651 *dsfield = ip6_tclass(t->parms.flowinfo); 652 653 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 654 fl6->flowi6_mark = skb->mark; 655 else 656 fl6->flowi6_mark = t->parms.fwmark; 657 658 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 659 } 660 661 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 662 struct net_device *dev, 663 struct flowi6 *fl6, __u8 *dsfield, 664 int *encap_limit) 665 { 666 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 667 struct ip6_tnl *t = netdev_priv(dev); 668 __u16 offset; 669 670 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 671 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 672 673 if (offset > 0) { 674 struct ipv6_tlv_tnl_enc_lim *tel; 675 676 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 677 if (tel->encap_limit == 0) { 678 icmpv6_send(skb, ICMPV6_PARAMPROB, 679 ICMPV6_HDR_FIELD, offset + 2); 680 return -1; 681 } 682 *encap_limit = tel->encap_limit - 1; 683 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 684 *encap_limit = t->parms.encap_limit; 685 } 686 687 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 688 689 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 690 *dsfield = ipv6_get_dsfield(ipv6h); 691 else 692 *dsfield = ip6_tclass(t->parms.flowinfo); 693 694 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 695 fl6->flowlabel |= ip6_flowlabel(ipv6h); 696 697 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 698 fl6->flowi6_mark = skb->mark; 699 else 700 fl6->flowi6_mark = t->parms.fwmark; 701 702 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 703 704 return 0; 705 } 706 707 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 708 struct net_device *dev, __u8 dsfield, 709 struct flowi6 *fl6, int encap_limit, 710 __u32 *pmtu, __be16 proto) 711 { 712 struct ip6_tnl *tunnel = netdev_priv(dev); 713 __be16 protocol; 714 715 if (dev->type == ARPHRD_ETHER) 716 IPCB(skb)->flags = 0; 717 718 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 719 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 720 else 721 fl6->daddr = tunnel->parms.raddr; 722 723 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 724 return -ENOMEM; 725 726 /* Push GRE header. */ 727 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 728 729 if (tunnel->parms.collect_md) { 730 struct ip_tunnel_info *tun_info; 731 const struct ip_tunnel_key *key; 732 __be16 flags; 733 734 tun_info = skb_tunnel_info(skb); 735 if (unlikely(!tun_info || 736 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 737 ip_tunnel_info_af(tun_info) != AF_INET6)) 738 return -EINVAL; 739 740 key = &tun_info->key; 741 memset(fl6, 0, sizeof(*fl6)); 742 fl6->flowi6_proto = IPPROTO_GRE; 743 fl6->daddr = key->u.ipv6.dst; 744 fl6->flowlabel = key->label; 745 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 746 747 dsfield = key->tos; 748 flags = key->tun_flags & 749 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 750 tunnel->tun_hlen = gre_calc_hlen(flags); 751 752 gre_build_header(skb, tunnel->tun_hlen, 753 flags, protocol, 754 tunnel_id_to_key32(tun_info->key.tun_id), 755 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 756 : 0); 757 758 } else { 759 if (tunnel->parms.o_flags & TUNNEL_SEQ) 760 tunnel->o_seqno++; 761 762 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 763 protocol, tunnel->parms.o_key, 764 htonl(tunnel->o_seqno)); 765 } 766 767 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 768 NEXTHDR_GRE); 769 } 770 771 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 772 { 773 struct ip6_tnl *t = netdev_priv(dev); 774 int encap_limit = -1; 775 struct flowi6 fl6; 776 __u8 dsfield = 0; 777 __u32 mtu; 778 int err; 779 780 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 781 782 if (!t->parms.collect_md) 783 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 784 &dsfield, &encap_limit); 785 786 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 787 if (err) 788 return -1; 789 790 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 791 skb->protocol); 792 if (err != 0) { 793 /* XXX: send ICMP error even if DF is not set. */ 794 if (err == -EMSGSIZE) 795 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 796 htonl(mtu)); 797 return -1; 798 } 799 800 return 0; 801 } 802 803 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 804 { 805 struct ip6_tnl *t = netdev_priv(dev); 806 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 807 int encap_limit = -1; 808 struct flowi6 fl6; 809 __u8 dsfield = 0; 810 __u32 mtu; 811 int err; 812 813 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 814 return -1; 815 816 if (!t->parms.collect_md && 817 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 818 return -1; 819 820 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 821 return -1; 822 823 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 824 &mtu, skb->protocol); 825 if (err != 0) { 826 if (err == -EMSGSIZE) 827 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 828 return -1; 829 } 830 831 return 0; 832 } 833 834 /** 835 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 836 * @t: the outgoing tunnel device 837 * @hdr: IPv6 header from the incoming packet 838 * 839 * Description: 840 * Avoid trivial tunneling loop by checking that tunnel exit-point 841 * doesn't match source of incoming packet. 842 * 843 * Return: 844 * 1 if conflict, 845 * 0 else 846 **/ 847 848 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 849 const struct ipv6hdr *hdr) 850 { 851 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 852 } 853 854 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 855 { 856 struct ip6_tnl *t = netdev_priv(dev); 857 int encap_limit = -1; 858 struct flowi6 fl6; 859 __u32 mtu; 860 int err; 861 862 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 863 encap_limit = t->parms.encap_limit; 864 865 if (!t->parms.collect_md) 866 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 867 868 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 869 if (err) 870 return err; 871 872 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 873 874 return err; 875 } 876 877 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 878 struct net_device *dev) 879 { 880 struct ip6_tnl *t = netdev_priv(dev); 881 struct net_device_stats *stats = &t->dev->stats; 882 int ret; 883 884 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 885 goto tx_err; 886 887 switch (skb->protocol) { 888 case htons(ETH_P_IP): 889 ret = ip6gre_xmit_ipv4(skb, dev); 890 break; 891 case htons(ETH_P_IPV6): 892 ret = ip6gre_xmit_ipv6(skb, dev); 893 break; 894 default: 895 ret = ip6gre_xmit_other(skb, dev); 896 break; 897 } 898 899 if (ret < 0) 900 goto tx_err; 901 902 return NETDEV_TX_OK; 903 904 tx_err: 905 stats->tx_errors++; 906 stats->tx_dropped++; 907 kfree_skb(skb); 908 return NETDEV_TX_OK; 909 } 910 911 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 912 struct net_device *dev) 913 { 914 struct ip6_tnl *t = netdev_priv(dev); 915 struct dst_entry *dst = skb_dst(skb); 916 struct net_device_stats *stats; 917 bool truncate = false; 918 int encap_limit = -1; 919 __u8 dsfield = false; 920 struct flowi6 fl6; 921 int err = -EINVAL; 922 __u32 mtu; 923 int nhoff; 924 int thoff; 925 926 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 927 goto tx_err; 928 929 if (gre_handle_offloads(skb, false)) 930 goto tx_err; 931 932 if (skb->len > dev->mtu + dev->hard_header_len) { 933 pskb_trim(skb, dev->mtu + dev->hard_header_len); 934 truncate = true; 935 } 936 937 nhoff = skb_network_header(skb) - skb_mac_header(skb); 938 if (skb->protocol == htons(ETH_P_IP) && 939 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 940 truncate = true; 941 942 thoff = skb_transport_header(skb) - skb_mac_header(skb); 943 if (skb->protocol == htons(ETH_P_IPV6) && 944 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 945 truncate = true; 946 947 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 948 goto tx_err; 949 950 t->parms.o_flags &= ~TUNNEL_KEY; 951 IPCB(skb)->flags = 0; 952 953 /* For collect_md mode, derive fl6 from the tunnel key, 954 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 955 */ 956 if (t->parms.collect_md) { 957 struct ip_tunnel_info *tun_info; 958 const struct ip_tunnel_key *key; 959 struct erspan_metadata *md; 960 __be32 tun_id; 961 962 tun_info = skb_tunnel_info(skb); 963 if (unlikely(!tun_info || 964 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 965 ip_tunnel_info_af(tun_info) != AF_INET6)) 966 return -EINVAL; 967 968 key = &tun_info->key; 969 memset(&fl6, 0, sizeof(fl6)); 970 fl6.flowi6_proto = IPPROTO_GRE; 971 fl6.daddr = key->u.ipv6.dst; 972 fl6.flowlabel = key->label; 973 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 974 975 dsfield = key->tos; 976 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 977 goto tx_err; 978 md = ip_tunnel_info_opts(tun_info); 979 if (!md) 980 goto tx_err; 981 982 tun_id = tunnel_id_to_key32(key->tun_id); 983 if (md->version == 1) { 984 erspan_build_header(skb, 985 ntohl(tun_id), 986 ntohl(md->u.index), truncate, 987 false); 988 } else if (md->version == 2) { 989 erspan_build_header_v2(skb, 990 ntohl(tun_id), 991 md->u.md2.dir, 992 get_hwid(&md->u.md2), 993 truncate, false); 994 } else { 995 goto tx_err; 996 } 997 } else { 998 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 999 1000 switch (skb->protocol) { 1001 case htons(ETH_P_IP): 1002 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1003 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1004 &dsfield, &encap_limit); 1005 break; 1006 case htons(ETH_P_IPV6): 1007 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 1008 goto tx_err; 1009 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1010 &dsfield, &encap_limit)) 1011 goto tx_err; 1012 break; 1013 default: 1014 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1015 break; 1016 } 1017 1018 if (t->parms.erspan_ver == 1) 1019 erspan_build_header(skb, ntohl(t->parms.o_key), 1020 t->parms.index, 1021 truncate, false); 1022 else if (t->parms.erspan_ver == 2) 1023 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1024 t->parms.dir, 1025 t->parms.hwid, 1026 truncate, false); 1027 else 1028 goto tx_err; 1029 1030 fl6.daddr = t->parms.raddr; 1031 } 1032 1033 /* Push GRE header. */ 1034 gre_build_header(skb, 8, TUNNEL_SEQ, 1035 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++)); 1036 1037 /* TooBig packet may have updated dst->dev's mtu */ 1038 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1039 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu); 1040 1041 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1042 NEXTHDR_GRE); 1043 if (err != 0) { 1044 /* XXX: send ICMP error even if DF is not set. */ 1045 if (err == -EMSGSIZE) { 1046 if (skb->protocol == htons(ETH_P_IP)) 1047 icmp_send(skb, ICMP_DEST_UNREACH, 1048 ICMP_FRAG_NEEDED, htonl(mtu)); 1049 else 1050 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1051 } 1052 1053 goto tx_err; 1054 } 1055 return NETDEV_TX_OK; 1056 1057 tx_err: 1058 stats = &t->dev->stats; 1059 stats->tx_errors++; 1060 stats->tx_dropped++; 1061 kfree_skb(skb); 1062 return NETDEV_TX_OK; 1063 } 1064 1065 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1066 { 1067 struct net_device *dev = t->dev; 1068 struct __ip6_tnl_parm *p = &t->parms; 1069 struct flowi6 *fl6 = &t->fl.u.ip6; 1070 1071 if (dev->type != ARPHRD_ETHER) { 1072 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1073 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1074 } 1075 1076 /* Set up flowi template */ 1077 fl6->saddr = p->laddr; 1078 fl6->daddr = p->raddr; 1079 fl6->flowi6_oif = p->link; 1080 fl6->flowlabel = 0; 1081 fl6->flowi6_proto = IPPROTO_GRE; 1082 1083 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1084 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1085 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1086 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1087 1088 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1089 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1090 1091 if (p->flags&IP6_TNL_F_CAP_XMIT && 1092 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1093 dev->flags |= IFF_POINTOPOINT; 1094 else 1095 dev->flags &= ~IFF_POINTOPOINT; 1096 } 1097 1098 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1099 int t_hlen) 1100 { 1101 const struct __ip6_tnl_parm *p = &t->parms; 1102 struct net_device *dev = t->dev; 1103 1104 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1105 int strict = (ipv6_addr_type(&p->raddr) & 1106 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1107 1108 struct rt6_info *rt = rt6_lookup(t->net, 1109 &p->raddr, &p->laddr, 1110 p->link, NULL, strict); 1111 1112 if (!rt) 1113 return; 1114 1115 if (rt->dst.dev) { 1116 dev->needed_headroom = rt->dst.dev->hard_header_len + 1117 t_hlen; 1118 1119 if (set_mtu) { 1120 dev->mtu = rt->dst.dev->mtu - t_hlen; 1121 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1122 dev->mtu -= 8; 1123 if (dev->type == ARPHRD_ETHER) 1124 dev->mtu -= ETH_HLEN; 1125 1126 if (dev->mtu < IPV6_MIN_MTU) 1127 dev->mtu = IPV6_MIN_MTU; 1128 } 1129 } 1130 ip6_rt_put(rt); 1131 } 1132 } 1133 1134 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1135 { 1136 int t_hlen; 1137 1138 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1139 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1140 1141 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1142 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1143 return t_hlen; 1144 } 1145 1146 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1147 { 1148 ip6gre_tnl_link_config_common(t); 1149 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1150 } 1151 1152 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1153 const struct __ip6_tnl_parm *p) 1154 { 1155 t->parms.laddr = p->laddr; 1156 t->parms.raddr = p->raddr; 1157 t->parms.flags = p->flags; 1158 t->parms.hop_limit = p->hop_limit; 1159 t->parms.encap_limit = p->encap_limit; 1160 t->parms.flowinfo = p->flowinfo; 1161 t->parms.link = p->link; 1162 t->parms.proto = p->proto; 1163 t->parms.i_key = p->i_key; 1164 t->parms.o_key = p->o_key; 1165 t->parms.i_flags = p->i_flags; 1166 t->parms.o_flags = p->o_flags; 1167 t->parms.fwmark = p->fwmark; 1168 dst_cache_reset(&t->dst_cache); 1169 } 1170 1171 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1172 int set_mtu) 1173 { 1174 ip6gre_tnl_copy_tnl_parm(t, p); 1175 ip6gre_tnl_link_config(t, set_mtu); 1176 return 0; 1177 } 1178 1179 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1180 const struct ip6_tnl_parm2 *u) 1181 { 1182 p->laddr = u->laddr; 1183 p->raddr = u->raddr; 1184 p->flags = u->flags; 1185 p->hop_limit = u->hop_limit; 1186 p->encap_limit = u->encap_limit; 1187 p->flowinfo = u->flowinfo; 1188 p->link = u->link; 1189 p->i_key = u->i_key; 1190 p->o_key = u->o_key; 1191 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1192 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1193 memcpy(p->name, u->name, sizeof(u->name)); 1194 } 1195 1196 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1197 const struct __ip6_tnl_parm *p) 1198 { 1199 u->proto = IPPROTO_GRE; 1200 u->laddr = p->laddr; 1201 u->raddr = p->raddr; 1202 u->flags = p->flags; 1203 u->hop_limit = p->hop_limit; 1204 u->encap_limit = p->encap_limit; 1205 u->flowinfo = p->flowinfo; 1206 u->link = p->link; 1207 u->i_key = p->i_key; 1208 u->o_key = p->o_key; 1209 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1210 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1211 memcpy(u->name, p->name, sizeof(u->name)); 1212 } 1213 1214 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1215 struct ifreq *ifr, int cmd) 1216 { 1217 int err = 0; 1218 struct ip6_tnl_parm2 p; 1219 struct __ip6_tnl_parm p1; 1220 struct ip6_tnl *t = netdev_priv(dev); 1221 struct net *net = t->net; 1222 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1223 1224 memset(&p1, 0, sizeof(p1)); 1225 1226 switch (cmd) { 1227 case SIOCGETTUNNEL: 1228 if (dev == ign->fb_tunnel_dev) { 1229 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1230 err = -EFAULT; 1231 break; 1232 } 1233 ip6gre_tnl_parm_from_user(&p1, &p); 1234 t = ip6gre_tunnel_locate(net, &p1, 0); 1235 if (!t) 1236 t = netdev_priv(dev); 1237 } 1238 memset(&p, 0, sizeof(p)); 1239 ip6gre_tnl_parm_to_user(&p, &t->parms); 1240 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1241 err = -EFAULT; 1242 break; 1243 1244 case SIOCADDTUNNEL: 1245 case SIOCCHGTUNNEL: 1246 err = -EPERM; 1247 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1248 goto done; 1249 1250 err = -EFAULT; 1251 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1252 goto done; 1253 1254 err = -EINVAL; 1255 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1256 goto done; 1257 1258 if (!(p.i_flags&GRE_KEY)) 1259 p.i_key = 0; 1260 if (!(p.o_flags&GRE_KEY)) 1261 p.o_key = 0; 1262 1263 ip6gre_tnl_parm_from_user(&p1, &p); 1264 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1265 1266 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1267 if (t) { 1268 if (t->dev != dev) { 1269 err = -EEXIST; 1270 break; 1271 } 1272 } else { 1273 t = netdev_priv(dev); 1274 1275 ip6gre_tunnel_unlink(ign, t); 1276 synchronize_net(); 1277 ip6gre_tnl_change(t, &p1, 1); 1278 ip6gre_tunnel_link(ign, t); 1279 netdev_state_change(dev); 1280 } 1281 } 1282 1283 if (t) { 1284 err = 0; 1285 1286 memset(&p, 0, sizeof(p)); 1287 ip6gre_tnl_parm_to_user(&p, &t->parms); 1288 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1289 err = -EFAULT; 1290 } else 1291 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1292 break; 1293 1294 case SIOCDELTUNNEL: 1295 err = -EPERM; 1296 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1297 goto done; 1298 1299 if (dev == ign->fb_tunnel_dev) { 1300 err = -EFAULT; 1301 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1302 goto done; 1303 err = -ENOENT; 1304 ip6gre_tnl_parm_from_user(&p1, &p); 1305 t = ip6gre_tunnel_locate(net, &p1, 0); 1306 if (!t) 1307 goto done; 1308 err = -EPERM; 1309 if (t == netdev_priv(ign->fb_tunnel_dev)) 1310 goto done; 1311 dev = t->dev; 1312 } 1313 unregister_netdevice(dev); 1314 err = 0; 1315 break; 1316 1317 default: 1318 err = -EINVAL; 1319 } 1320 1321 done: 1322 return err; 1323 } 1324 1325 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1326 unsigned short type, const void *daddr, 1327 const void *saddr, unsigned int len) 1328 { 1329 struct ip6_tnl *t = netdev_priv(dev); 1330 struct ipv6hdr *ipv6h; 1331 __be16 *p; 1332 1333 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1334 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1335 t->fl.u.ip6.flowlabel, 1336 true, &t->fl.u.ip6)); 1337 ipv6h->hop_limit = t->parms.hop_limit; 1338 ipv6h->nexthdr = NEXTHDR_GRE; 1339 ipv6h->saddr = t->parms.laddr; 1340 ipv6h->daddr = t->parms.raddr; 1341 1342 p = (__be16 *)(ipv6h + 1); 1343 p[0] = t->parms.o_flags; 1344 p[1] = htons(type); 1345 1346 /* 1347 * Set the source hardware address. 1348 */ 1349 1350 if (saddr) 1351 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1352 if (daddr) 1353 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1354 if (!ipv6_addr_any(&ipv6h->daddr)) 1355 return t->hlen; 1356 1357 return -t->hlen; 1358 } 1359 1360 static const struct header_ops ip6gre_header_ops = { 1361 .create = ip6gre_header, 1362 }; 1363 1364 static const struct net_device_ops ip6gre_netdev_ops = { 1365 .ndo_init = ip6gre_tunnel_init, 1366 .ndo_uninit = ip6gre_tunnel_uninit, 1367 .ndo_start_xmit = ip6gre_tunnel_xmit, 1368 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1369 .ndo_change_mtu = ip6_tnl_change_mtu, 1370 .ndo_get_stats64 = ip_tunnel_get_stats64, 1371 .ndo_get_iflink = ip6_tnl_get_iflink, 1372 }; 1373 1374 static void ip6gre_dev_free(struct net_device *dev) 1375 { 1376 struct ip6_tnl *t = netdev_priv(dev); 1377 1378 gro_cells_destroy(&t->gro_cells); 1379 dst_cache_destroy(&t->dst_cache); 1380 free_percpu(dev->tstats); 1381 } 1382 1383 static void ip6gre_tunnel_setup(struct net_device *dev) 1384 { 1385 dev->netdev_ops = &ip6gre_netdev_ops; 1386 dev->needs_free_netdev = true; 1387 dev->priv_destructor = ip6gre_dev_free; 1388 1389 dev->type = ARPHRD_IP6GRE; 1390 1391 dev->flags |= IFF_NOARP; 1392 dev->addr_len = sizeof(struct in6_addr); 1393 netif_keep_dst(dev); 1394 /* This perm addr will be used as interface identifier by IPv6 */ 1395 dev->addr_assign_type = NET_ADDR_RANDOM; 1396 eth_random_addr(dev->perm_addr); 1397 } 1398 1399 #define GRE6_FEATURES (NETIF_F_SG | \ 1400 NETIF_F_FRAGLIST | \ 1401 NETIF_F_HIGHDMA | \ 1402 NETIF_F_HW_CSUM) 1403 1404 static void ip6gre_tnl_init_features(struct net_device *dev) 1405 { 1406 struct ip6_tnl *nt = netdev_priv(dev); 1407 1408 dev->features |= GRE6_FEATURES; 1409 dev->hw_features |= GRE6_FEATURES; 1410 1411 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1412 /* TCP offload with GRE SEQ is not supported, nor 1413 * can we support 2 levels of outer headers requiring 1414 * an update. 1415 */ 1416 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1417 nt->encap.type == TUNNEL_ENCAP_NONE) { 1418 dev->features |= NETIF_F_GSO_SOFTWARE; 1419 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1420 } 1421 1422 /* Can use a lockless transmit, unless we generate 1423 * output sequences 1424 */ 1425 dev->features |= NETIF_F_LLTX; 1426 } 1427 } 1428 1429 static int ip6gre_tunnel_init_common(struct net_device *dev) 1430 { 1431 struct ip6_tnl *tunnel; 1432 int ret; 1433 int t_hlen; 1434 1435 tunnel = netdev_priv(dev); 1436 1437 tunnel->dev = dev; 1438 tunnel->net = dev_net(dev); 1439 strcpy(tunnel->parms.name, dev->name); 1440 1441 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1442 if (!dev->tstats) 1443 return -ENOMEM; 1444 1445 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1446 if (ret) 1447 goto cleanup_alloc_pcpu_stats; 1448 1449 ret = gro_cells_init(&tunnel->gro_cells, dev); 1450 if (ret) 1451 goto cleanup_dst_cache_init; 1452 1453 t_hlen = ip6gre_calc_hlen(tunnel); 1454 dev->mtu = ETH_DATA_LEN - t_hlen; 1455 if (dev->type == ARPHRD_ETHER) 1456 dev->mtu -= ETH_HLEN; 1457 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1458 dev->mtu -= 8; 1459 1460 if (tunnel->parms.collect_md) { 1461 dev->features |= NETIF_F_NETNS_LOCAL; 1462 netif_keep_dst(dev); 1463 } 1464 ip6gre_tnl_init_features(dev); 1465 1466 return 0; 1467 1468 cleanup_dst_cache_init: 1469 dst_cache_destroy(&tunnel->dst_cache); 1470 cleanup_alloc_pcpu_stats: 1471 free_percpu(dev->tstats); 1472 dev->tstats = NULL; 1473 return ret; 1474 } 1475 1476 static int ip6gre_tunnel_init(struct net_device *dev) 1477 { 1478 struct ip6_tnl *tunnel; 1479 int ret; 1480 1481 ret = ip6gre_tunnel_init_common(dev); 1482 if (ret) 1483 return ret; 1484 1485 tunnel = netdev_priv(dev); 1486 1487 if (tunnel->parms.collect_md) 1488 return 0; 1489 1490 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1491 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1492 1493 if (ipv6_addr_any(&tunnel->parms.raddr)) 1494 dev->header_ops = &ip6gre_header_ops; 1495 1496 return 0; 1497 } 1498 1499 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1500 { 1501 struct ip6_tnl *tunnel = netdev_priv(dev); 1502 1503 tunnel->dev = dev; 1504 tunnel->net = dev_net(dev); 1505 strcpy(tunnel->parms.name, dev->name); 1506 1507 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1508 1509 dev_hold(dev); 1510 } 1511 1512 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1513 .handler = gre_rcv, 1514 .err_handler = ip6gre_err, 1515 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1516 }; 1517 1518 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1519 { 1520 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1521 struct net_device *dev, *aux; 1522 int prio; 1523 1524 for_each_netdev_safe(net, dev, aux) 1525 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1526 dev->rtnl_link_ops == &ip6gre_tap_ops || 1527 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1528 unregister_netdevice_queue(dev, head); 1529 1530 for (prio = 0; prio < 4; prio++) { 1531 int h; 1532 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1533 struct ip6_tnl *t; 1534 1535 t = rtnl_dereference(ign->tunnels[prio][h]); 1536 1537 while (t) { 1538 /* If dev is in the same netns, it has already 1539 * been added to the list by the previous loop. 1540 */ 1541 if (!net_eq(dev_net(t->dev), net)) 1542 unregister_netdevice_queue(t->dev, 1543 head); 1544 t = rtnl_dereference(t->next); 1545 } 1546 } 1547 } 1548 } 1549 1550 static int __net_init ip6gre_init_net(struct net *net) 1551 { 1552 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1553 int err; 1554 1555 if (!net_has_fallback_tunnels(net)) 1556 return 0; 1557 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1558 NET_NAME_UNKNOWN, 1559 ip6gre_tunnel_setup); 1560 if (!ign->fb_tunnel_dev) { 1561 err = -ENOMEM; 1562 goto err_alloc_dev; 1563 } 1564 dev_net_set(ign->fb_tunnel_dev, net); 1565 /* FB netdevice is special: we have one, and only one per netns. 1566 * Allowing to move it to another netns is clearly unsafe. 1567 */ 1568 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1569 1570 1571 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1572 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1573 1574 err = register_netdev(ign->fb_tunnel_dev); 1575 if (err) 1576 goto err_reg_dev; 1577 1578 rcu_assign_pointer(ign->tunnels_wc[0], 1579 netdev_priv(ign->fb_tunnel_dev)); 1580 return 0; 1581 1582 err_reg_dev: 1583 free_netdev(ign->fb_tunnel_dev); 1584 err_alloc_dev: 1585 return err; 1586 } 1587 1588 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1589 { 1590 struct net *net; 1591 LIST_HEAD(list); 1592 1593 rtnl_lock(); 1594 list_for_each_entry(net, net_list, exit_list) 1595 ip6gre_destroy_tunnels(net, &list); 1596 unregister_netdevice_many(&list); 1597 rtnl_unlock(); 1598 } 1599 1600 static struct pernet_operations ip6gre_net_ops = { 1601 .init = ip6gre_init_net, 1602 .exit_batch = ip6gre_exit_batch_net, 1603 .id = &ip6gre_net_id, 1604 .size = sizeof(struct ip6gre_net), 1605 }; 1606 1607 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1608 struct netlink_ext_ack *extack) 1609 { 1610 __be16 flags; 1611 1612 if (!data) 1613 return 0; 1614 1615 flags = 0; 1616 if (data[IFLA_GRE_IFLAGS]) 1617 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1618 if (data[IFLA_GRE_OFLAGS]) 1619 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1620 if (flags & (GRE_VERSION|GRE_ROUTING)) 1621 return -EINVAL; 1622 1623 return 0; 1624 } 1625 1626 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1627 struct netlink_ext_ack *extack) 1628 { 1629 struct in6_addr daddr; 1630 1631 if (tb[IFLA_ADDRESS]) { 1632 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1633 return -EINVAL; 1634 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1635 return -EADDRNOTAVAIL; 1636 } 1637 1638 if (!data) 1639 goto out; 1640 1641 if (data[IFLA_GRE_REMOTE]) { 1642 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1643 if (ipv6_addr_any(&daddr)) 1644 return -EINVAL; 1645 } 1646 1647 out: 1648 return ip6gre_tunnel_validate(tb, data, extack); 1649 } 1650 1651 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1652 struct netlink_ext_ack *extack) 1653 { 1654 __be16 flags = 0; 1655 int ret, ver = 0; 1656 1657 if (!data) 1658 return 0; 1659 1660 ret = ip6gre_tap_validate(tb, data, extack); 1661 if (ret) 1662 return ret; 1663 1664 /* ERSPAN should only have GRE sequence and key flag */ 1665 if (data[IFLA_GRE_OFLAGS]) 1666 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1667 if (data[IFLA_GRE_IFLAGS]) 1668 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1669 if (!data[IFLA_GRE_COLLECT_METADATA] && 1670 flags != (GRE_SEQ | GRE_KEY)) 1671 return -EINVAL; 1672 1673 /* ERSPAN Session ID only has 10-bit. Since we reuse 1674 * 32-bit key field as ID, check it's range. 1675 */ 1676 if (data[IFLA_GRE_IKEY] && 1677 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1678 return -EINVAL; 1679 1680 if (data[IFLA_GRE_OKEY] && 1681 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1682 return -EINVAL; 1683 1684 if (data[IFLA_GRE_ERSPAN_VER]) { 1685 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1686 if (ver != 1 && ver != 2) 1687 return -EINVAL; 1688 } 1689 1690 if (ver == 1) { 1691 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1692 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1693 1694 if (index & ~INDEX_MASK) 1695 return -EINVAL; 1696 } 1697 } else if (ver == 2) { 1698 if (data[IFLA_GRE_ERSPAN_DIR]) { 1699 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1700 1701 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1702 return -EINVAL; 1703 } 1704 1705 if (data[IFLA_GRE_ERSPAN_HWID]) { 1706 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1707 1708 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1709 return -EINVAL; 1710 } 1711 } 1712 1713 return 0; 1714 } 1715 1716 static void ip6gre_netlink_parms(struct nlattr *data[], 1717 struct __ip6_tnl_parm *parms) 1718 { 1719 memset(parms, 0, sizeof(*parms)); 1720 1721 if (!data) 1722 return; 1723 1724 if (data[IFLA_GRE_LINK]) 1725 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1726 1727 if (data[IFLA_GRE_IFLAGS]) 1728 parms->i_flags = gre_flags_to_tnl_flags( 1729 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1730 1731 if (data[IFLA_GRE_OFLAGS]) 1732 parms->o_flags = gre_flags_to_tnl_flags( 1733 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1734 1735 if (data[IFLA_GRE_IKEY]) 1736 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1737 1738 if (data[IFLA_GRE_OKEY]) 1739 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1740 1741 if (data[IFLA_GRE_LOCAL]) 1742 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1743 1744 if (data[IFLA_GRE_REMOTE]) 1745 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1746 1747 if (data[IFLA_GRE_TTL]) 1748 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1749 1750 if (data[IFLA_GRE_ENCAP_LIMIT]) 1751 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1752 1753 if (data[IFLA_GRE_FLOWINFO]) 1754 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1755 1756 if (data[IFLA_GRE_FLAGS]) 1757 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1758 1759 if (data[IFLA_GRE_FWMARK]) 1760 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1761 1762 if (data[IFLA_GRE_COLLECT_METADATA]) 1763 parms->collect_md = true; 1764 1765 parms->erspan_ver = 1; 1766 if (data[IFLA_GRE_ERSPAN_VER]) 1767 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1768 1769 if (parms->erspan_ver == 1) { 1770 if (data[IFLA_GRE_ERSPAN_INDEX]) 1771 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1772 } else if (parms->erspan_ver == 2) { 1773 if (data[IFLA_GRE_ERSPAN_DIR]) 1774 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1775 if (data[IFLA_GRE_ERSPAN_HWID]) 1776 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1777 } 1778 } 1779 1780 static int ip6gre_tap_init(struct net_device *dev) 1781 { 1782 int ret; 1783 1784 ret = ip6gre_tunnel_init_common(dev); 1785 if (ret) 1786 return ret; 1787 1788 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1789 1790 return 0; 1791 } 1792 1793 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1794 .ndo_init = ip6gre_tap_init, 1795 .ndo_uninit = ip6gre_tunnel_uninit, 1796 .ndo_start_xmit = ip6gre_tunnel_xmit, 1797 .ndo_set_mac_address = eth_mac_addr, 1798 .ndo_validate_addr = eth_validate_addr, 1799 .ndo_change_mtu = ip6_tnl_change_mtu, 1800 .ndo_get_stats64 = ip_tunnel_get_stats64, 1801 .ndo_get_iflink = ip6_tnl_get_iflink, 1802 }; 1803 1804 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1805 { 1806 int t_hlen; 1807 1808 tunnel->tun_hlen = 8; 1809 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1810 erspan_hdr_len(tunnel->parms.erspan_ver); 1811 1812 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1813 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1814 return t_hlen; 1815 } 1816 1817 static int ip6erspan_tap_init(struct net_device *dev) 1818 { 1819 struct ip6_tnl *tunnel; 1820 int t_hlen; 1821 int ret; 1822 1823 tunnel = netdev_priv(dev); 1824 1825 tunnel->dev = dev; 1826 tunnel->net = dev_net(dev); 1827 strcpy(tunnel->parms.name, dev->name); 1828 1829 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1830 if (!dev->tstats) 1831 return -ENOMEM; 1832 1833 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1834 if (ret) 1835 goto cleanup_alloc_pcpu_stats; 1836 1837 ret = gro_cells_init(&tunnel->gro_cells, dev); 1838 if (ret) 1839 goto cleanup_dst_cache_init; 1840 1841 t_hlen = ip6erspan_calc_hlen(tunnel); 1842 dev->mtu = ETH_DATA_LEN - t_hlen; 1843 if (dev->type == ARPHRD_ETHER) 1844 dev->mtu -= ETH_HLEN; 1845 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1846 dev->mtu -= 8; 1847 1848 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1849 ip6erspan_tnl_link_config(tunnel, 1); 1850 1851 return 0; 1852 1853 cleanup_dst_cache_init: 1854 dst_cache_destroy(&tunnel->dst_cache); 1855 cleanup_alloc_pcpu_stats: 1856 free_percpu(dev->tstats); 1857 dev->tstats = NULL; 1858 return ret; 1859 } 1860 1861 static const struct net_device_ops ip6erspan_netdev_ops = { 1862 .ndo_init = ip6erspan_tap_init, 1863 .ndo_uninit = ip6erspan_tunnel_uninit, 1864 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1865 .ndo_set_mac_address = eth_mac_addr, 1866 .ndo_validate_addr = eth_validate_addr, 1867 .ndo_change_mtu = ip6_tnl_change_mtu, 1868 .ndo_get_stats64 = ip_tunnel_get_stats64, 1869 .ndo_get_iflink = ip6_tnl_get_iflink, 1870 }; 1871 1872 static void ip6gre_tap_setup(struct net_device *dev) 1873 { 1874 1875 ether_setup(dev); 1876 1877 dev->max_mtu = 0; 1878 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1879 dev->needs_free_netdev = true; 1880 dev->priv_destructor = ip6gre_dev_free; 1881 1882 dev->features |= NETIF_F_NETNS_LOCAL; 1883 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1884 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1885 netif_keep_dst(dev); 1886 } 1887 1888 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1889 struct ip_tunnel_encap *ipencap) 1890 { 1891 bool ret = false; 1892 1893 memset(ipencap, 0, sizeof(*ipencap)); 1894 1895 if (!data) 1896 return ret; 1897 1898 if (data[IFLA_GRE_ENCAP_TYPE]) { 1899 ret = true; 1900 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1901 } 1902 1903 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1904 ret = true; 1905 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1906 } 1907 1908 if (data[IFLA_GRE_ENCAP_SPORT]) { 1909 ret = true; 1910 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1911 } 1912 1913 if (data[IFLA_GRE_ENCAP_DPORT]) { 1914 ret = true; 1915 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1916 } 1917 1918 return ret; 1919 } 1920 1921 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1922 struct nlattr *tb[], struct nlattr *data[], 1923 struct netlink_ext_ack *extack) 1924 { 1925 struct ip6_tnl *nt; 1926 struct ip_tunnel_encap ipencap; 1927 int err; 1928 1929 nt = netdev_priv(dev); 1930 1931 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1932 int err = ip6_tnl_encap_setup(nt, &ipencap); 1933 1934 if (err < 0) 1935 return err; 1936 } 1937 1938 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1939 eth_hw_addr_random(dev); 1940 1941 nt->dev = dev; 1942 nt->net = dev_net(dev); 1943 1944 err = register_netdevice(dev); 1945 if (err) 1946 goto out; 1947 1948 if (tb[IFLA_MTU]) 1949 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1950 1951 dev_hold(dev); 1952 1953 out: 1954 return err; 1955 } 1956 1957 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1958 struct nlattr *tb[], struct nlattr *data[], 1959 struct netlink_ext_ack *extack) 1960 { 1961 struct ip6_tnl *nt = netdev_priv(dev); 1962 struct net *net = dev_net(dev); 1963 struct ip6gre_net *ign; 1964 int err; 1965 1966 ip6gre_netlink_parms(data, &nt->parms); 1967 ign = net_generic(net, ip6gre_net_id); 1968 1969 if (nt->parms.collect_md) { 1970 if (rtnl_dereference(ign->collect_md_tun)) 1971 return -EEXIST; 1972 } else { 1973 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1974 return -EEXIST; 1975 } 1976 1977 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 1978 if (!err) { 1979 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1980 ip6gre_tunnel_link_md(ign, nt); 1981 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 1982 } 1983 return err; 1984 } 1985 1986 static struct ip6_tnl * 1987 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 1988 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 1989 struct netlink_ext_ack *extack) 1990 { 1991 struct ip6_tnl *t, *nt = netdev_priv(dev); 1992 struct net *net = nt->net; 1993 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1994 struct ip_tunnel_encap ipencap; 1995 1996 if (dev == ign->fb_tunnel_dev) 1997 return ERR_PTR(-EINVAL); 1998 1999 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2000 int err = ip6_tnl_encap_setup(nt, &ipencap); 2001 2002 if (err < 0) 2003 return ERR_PTR(err); 2004 } 2005 2006 ip6gre_netlink_parms(data, p_p); 2007 2008 t = ip6gre_tunnel_locate(net, p_p, 0); 2009 2010 if (t) { 2011 if (t->dev != dev) 2012 return ERR_PTR(-EEXIST); 2013 } else { 2014 t = nt; 2015 } 2016 2017 return t; 2018 } 2019 2020 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2021 struct nlattr *data[], 2022 struct netlink_ext_ack *extack) 2023 { 2024 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2025 struct __ip6_tnl_parm p; 2026 struct ip6_tnl *t; 2027 2028 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2029 if (IS_ERR(t)) 2030 return PTR_ERR(t); 2031 2032 ip6gre_tunnel_unlink_md(ign, t); 2033 ip6gre_tunnel_unlink(ign, t); 2034 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2035 ip6gre_tunnel_link_md(ign, t); 2036 ip6gre_tunnel_link(ign, t); 2037 return 0; 2038 } 2039 2040 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2041 { 2042 struct net *net = dev_net(dev); 2043 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2044 2045 if (dev != ign->fb_tunnel_dev) 2046 unregister_netdevice_queue(dev, head); 2047 } 2048 2049 static size_t ip6gre_get_size(const struct net_device *dev) 2050 { 2051 return 2052 /* IFLA_GRE_LINK */ 2053 nla_total_size(4) + 2054 /* IFLA_GRE_IFLAGS */ 2055 nla_total_size(2) + 2056 /* IFLA_GRE_OFLAGS */ 2057 nla_total_size(2) + 2058 /* IFLA_GRE_IKEY */ 2059 nla_total_size(4) + 2060 /* IFLA_GRE_OKEY */ 2061 nla_total_size(4) + 2062 /* IFLA_GRE_LOCAL */ 2063 nla_total_size(sizeof(struct in6_addr)) + 2064 /* IFLA_GRE_REMOTE */ 2065 nla_total_size(sizeof(struct in6_addr)) + 2066 /* IFLA_GRE_TTL */ 2067 nla_total_size(1) + 2068 /* IFLA_GRE_ENCAP_LIMIT */ 2069 nla_total_size(1) + 2070 /* IFLA_GRE_FLOWINFO */ 2071 nla_total_size(4) + 2072 /* IFLA_GRE_FLAGS */ 2073 nla_total_size(4) + 2074 /* IFLA_GRE_ENCAP_TYPE */ 2075 nla_total_size(2) + 2076 /* IFLA_GRE_ENCAP_FLAGS */ 2077 nla_total_size(2) + 2078 /* IFLA_GRE_ENCAP_SPORT */ 2079 nla_total_size(2) + 2080 /* IFLA_GRE_ENCAP_DPORT */ 2081 nla_total_size(2) + 2082 /* IFLA_GRE_COLLECT_METADATA */ 2083 nla_total_size(0) + 2084 /* IFLA_GRE_FWMARK */ 2085 nla_total_size(4) + 2086 /* IFLA_GRE_ERSPAN_INDEX */ 2087 nla_total_size(4) + 2088 0; 2089 } 2090 2091 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2092 { 2093 struct ip6_tnl *t = netdev_priv(dev); 2094 struct __ip6_tnl_parm *p = &t->parms; 2095 2096 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2097 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2098 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2099 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2100 gre_tnl_flags_to_gre_flags(p->o_flags)) || 2101 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2102 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2103 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2104 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2105 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2106 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2107 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2108 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2109 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) || 2110 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2111 goto nla_put_failure; 2112 2113 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2114 t->encap.type) || 2115 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2116 t->encap.sport) || 2117 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2118 t->encap.dport) || 2119 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2120 t->encap.flags)) 2121 goto nla_put_failure; 2122 2123 if (p->collect_md) { 2124 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2125 goto nla_put_failure; 2126 } 2127 2128 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2129 goto nla_put_failure; 2130 2131 if (p->erspan_ver == 1) { 2132 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2133 goto nla_put_failure; 2134 } else if (p->erspan_ver == 2) { 2135 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2136 goto nla_put_failure; 2137 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2138 goto nla_put_failure; 2139 } 2140 2141 return 0; 2142 2143 nla_put_failure: 2144 return -EMSGSIZE; 2145 } 2146 2147 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2148 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2149 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2150 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2151 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2152 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2153 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 2154 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 2155 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2156 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2157 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2158 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2159 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2160 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2161 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2162 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2163 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2164 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2165 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2166 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2167 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2168 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2169 }; 2170 2171 static void ip6erspan_tap_setup(struct net_device *dev) 2172 { 2173 ether_setup(dev); 2174 2175 dev->netdev_ops = &ip6erspan_netdev_ops; 2176 dev->needs_free_netdev = true; 2177 dev->priv_destructor = ip6gre_dev_free; 2178 2179 dev->features |= NETIF_F_NETNS_LOCAL; 2180 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2181 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2182 netif_keep_dst(dev); 2183 } 2184 2185 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2186 struct nlattr *tb[], struct nlattr *data[], 2187 struct netlink_ext_ack *extack) 2188 { 2189 struct ip6_tnl *nt = netdev_priv(dev); 2190 struct net *net = dev_net(dev); 2191 struct ip6gre_net *ign; 2192 int err; 2193 2194 ip6gre_netlink_parms(data, &nt->parms); 2195 ign = net_generic(net, ip6gre_net_id); 2196 2197 if (nt->parms.collect_md) { 2198 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2199 return -EEXIST; 2200 } else { 2201 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2202 return -EEXIST; 2203 } 2204 2205 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2206 if (!err) { 2207 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2208 ip6erspan_tunnel_link_md(ign, nt); 2209 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2210 } 2211 return err; 2212 } 2213 2214 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2215 { 2216 ip6gre_tnl_link_config_common(t); 2217 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2218 } 2219 2220 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2221 const struct __ip6_tnl_parm *p, int set_mtu) 2222 { 2223 ip6gre_tnl_copy_tnl_parm(t, p); 2224 ip6erspan_tnl_link_config(t, set_mtu); 2225 return 0; 2226 } 2227 2228 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2229 struct nlattr *data[], 2230 struct netlink_ext_ack *extack) 2231 { 2232 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2233 struct __ip6_tnl_parm p; 2234 struct ip6_tnl *t; 2235 2236 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2237 if (IS_ERR(t)) 2238 return PTR_ERR(t); 2239 2240 ip6gre_tunnel_unlink_md(ign, t); 2241 ip6gre_tunnel_unlink(ign, t); 2242 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2243 ip6erspan_tunnel_link_md(ign, t); 2244 ip6gre_tunnel_link(ign, t); 2245 return 0; 2246 } 2247 2248 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2249 .kind = "ip6gre", 2250 .maxtype = IFLA_GRE_MAX, 2251 .policy = ip6gre_policy, 2252 .priv_size = sizeof(struct ip6_tnl), 2253 .setup = ip6gre_tunnel_setup, 2254 .validate = ip6gre_tunnel_validate, 2255 .newlink = ip6gre_newlink, 2256 .changelink = ip6gre_changelink, 2257 .dellink = ip6gre_dellink, 2258 .get_size = ip6gre_get_size, 2259 .fill_info = ip6gre_fill_info, 2260 .get_link_net = ip6_tnl_get_link_net, 2261 }; 2262 2263 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2264 .kind = "ip6gretap", 2265 .maxtype = IFLA_GRE_MAX, 2266 .policy = ip6gre_policy, 2267 .priv_size = sizeof(struct ip6_tnl), 2268 .setup = ip6gre_tap_setup, 2269 .validate = ip6gre_tap_validate, 2270 .newlink = ip6gre_newlink, 2271 .changelink = ip6gre_changelink, 2272 .get_size = ip6gre_get_size, 2273 .fill_info = ip6gre_fill_info, 2274 .get_link_net = ip6_tnl_get_link_net, 2275 }; 2276 2277 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2278 .kind = "ip6erspan", 2279 .maxtype = IFLA_GRE_MAX, 2280 .policy = ip6gre_policy, 2281 .priv_size = sizeof(struct ip6_tnl), 2282 .setup = ip6erspan_tap_setup, 2283 .validate = ip6erspan_tap_validate, 2284 .newlink = ip6erspan_newlink, 2285 .changelink = ip6erspan_changelink, 2286 .get_size = ip6gre_get_size, 2287 .fill_info = ip6gre_fill_info, 2288 .get_link_net = ip6_tnl_get_link_net, 2289 }; 2290 2291 /* 2292 * And now the modules code and kernel interface. 2293 */ 2294 2295 static int __init ip6gre_init(void) 2296 { 2297 int err; 2298 2299 pr_info("GRE over IPv6 tunneling driver\n"); 2300 2301 err = register_pernet_device(&ip6gre_net_ops); 2302 if (err < 0) 2303 return err; 2304 2305 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2306 if (err < 0) { 2307 pr_info("%s: can't add protocol\n", __func__); 2308 goto add_proto_failed; 2309 } 2310 2311 err = rtnl_link_register(&ip6gre_link_ops); 2312 if (err < 0) 2313 goto rtnl_link_failed; 2314 2315 err = rtnl_link_register(&ip6gre_tap_ops); 2316 if (err < 0) 2317 goto tap_ops_failed; 2318 2319 err = rtnl_link_register(&ip6erspan_tap_ops); 2320 if (err < 0) 2321 goto erspan_link_failed; 2322 2323 out: 2324 return err; 2325 2326 erspan_link_failed: 2327 rtnl_link_unregister(&ip6gre_tap_ops); 2328 tap_ops_failed: 2329 rtnl_link_unregister(&ip6gre_link_ops); 2330 rtnl_link_failed: 2331 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2332 add_proto_failed: 2333 unregister_pernet_device(&ip6gre_net_ops); 2334 goto out; 2335 } 2336 2337 static void __exit ip6gre_fini(void) 2338 { 2339 rtnl_link_unregister(&ip6gre_tap_ops); 2340 rtnl_link_unregister(&ip6gre_link_ops); 2341 rtnl_link_unregister(&ip6erspan_tap_ops); 2342 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2343 unregister_pernet_device(&ip6gre_net_ops); 2344 } 2345 2346 module_init(ip6gre_init); 2347 module_exit(ip6gre_fini); 2348 MODULE_LICENSE("GPL"); 2349 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2350 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2351 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2352 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2353 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2354 MODULE_ALIAS_NETDEV("ip6gre0"); 2355