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, 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 = ipv6_hdr(skb); 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 */ 674 675 if (offset > 0) { 676 struct ipv6_tlv_tnl_enc_lim *tel; 677 678 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 679 if (tel->encap_limit == 0) { 680 icmpv6_send(skb, ICMPV6_PARAMPROB, 681 ICMPV6_HDR_FIELD, offset + 2); 682 return -1; 683 } 684 *encap_limit = tel->encap_limit - 1; 685 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 686 *encap_limit = t->parms.encap_limit; 687 } 688 689 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 690 691 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 692 *dsfield = ipv6_get_dsfield(ipv6h); 693 else 694 *dsfield = ip6_tclass(t->parms.flowinfo); 695 696 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 697 fl6->flowlabel |= ip6_flowlabel(ipv6h); 698 699 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 700 fl6->flowi6_mark = skb->mark; 701 else 702 fl6->flowi6_mark = t->parms.fwmark; 703 704 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 705 706 return 0; 707 } 708 709 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 710 struct net_device *dev, __u8 dsfield, 711 struct flowi6 *fl6, int encap_limit, 712 __u32 *pmtu, __be16 proto) 713 { 714 struct ip6_tnl *tunnel = netdev_priv(dev); 715 __be16 protocol; 716 717 if (dev->type == ARPHRD_ETHER) 718 IPCB(skb)->flags = 0; 719 720 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 721 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 722 else 723 fl6->daddr = tunnel->parms.raddr; 724 725 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 726 return -ENOMEM; 727 728 /* Push GRE header. */ 729 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 730 731 if (tunnel->parms.collect_md) { 732 struct ip_tunnel_info *tun_info; 733 const struct ip_tunnel_key *key; 734 __be16 flags; 735 736 tun_info = skb_tunnel_info(skb); 737 if (unlikely(!tun_info || 738 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 739 ip_tunnel_info_af(tun_info) != AF_INET6)) 740 return -EINVAL; 741 742 key = &tun_info->key; 743 memset(fl6, 0, sizeof(*fl6)); 744 fl6->flowi6_proto = IPPROTO_GRE; 745 fl6->daddr = key->u.ipv6.dst; 746 fl6->flowlabel = key->label; 747 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 748 749 dsfield = key->tos; 750 flags = key->tun_flags & 751 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 752 tunnel->tun_hlen = gre_calc_hlen(flags); 753 754 gre_build_header(skb, tunnel->tun_hlen, 755 flags, protocol, 756 tunnel_id_to_key32(tun_info->key.tun_id), 757 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 758 : 0); 759 760 } else { 761 if (tunnel->parms.o_flags & TUNNEL_SEQ) 762 tunnel->o_seqno++; 763 764 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 765 protocol, tunnel->parms.o_key, 766 htonl(tunnel->o_seqno)); 767 } 768 769 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 770 NEXTHDR_GRE); 771 } 772 773 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 774 { 775 struct ip6_tnl *t = netdev_priv(dev); 776 int encap_limit = -1; 777 struct flowi6 fl6; 778 __u8 dsfield = 0; 779 __u32 mtu; 780 int err; 781 782 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 783 784 if (!t->parms.collect_md) 785 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 786 &dsfield, &encap_limit); 787 788 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 789 if (err) 790 return -1; 791 792 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 793 skb->protocol); 794 if (err != 0) { 795 /* XXX: send ICMP error even if DF is not set. */ 796 if (err == -EMSGSIZE) 797 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 798 htonl(mtu)); 799 return -1; 800 } 801 802 return 0; 803 } 804 805 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 806 { 807 struct ip6_tnl *t = netdev_priv(dev); 808 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 809 int encap_limit = -1; 810 struct flowi6 fl6; 811 __u8 dsfield = 0; 812 __u32 mtu; 813 int err; 814 815 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 816 return -1; 817 818 if (!t->parms.collect_md && 819 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 820 return -1; 821 822 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 823 return -1; 824 825 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 826 &mtu, skb->protocol); 827 if (err != 0) { 828 if (err == -EMSGSIZE) 829 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 830 return -1; 831 } 832 833 return 0; 834 } 835 836 /** 837 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 838 * @t: the outgoing tunnel device 839 * @hdr: IPv6 header from the incoming packet 840 * 841 * Description: 842 * Avoid trivial tunneling loop by checking that tunnel exit-point 843 * doesn't match source of incoming packet. 844 * 845 * Return: 846 * 1 if conflict, 847 * 0 else 848 **/ 849 850 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 851 const struct ipv6hdr *hdr) 852 { 853 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 854 } 855 856 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 857 { 858 struct ip6_tnl *t = netdev_priv(dev); 859 int encap_limit = -1; 860 struct flowi6 fl6; 861 __u32 mtu; 862 int err; 863 864 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 865 encap_limit = t->parms.encap_limit; 866 867 if (!t->parms.collect_md) 868 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 869 870 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 871 if (err) 872 return err; 873 874 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 875 876 return err; 877 } 878 879 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 880 struct net_device *dev) 881 { 882 struct ip6_tnl *t = netdev_priv(dev); 883 struct net_device_stats *stats = &t->dev->stats; 884 int ret; 885 886 if (!pskb_inet_may_pull(skb)) 887 goto tx_err; 888 889 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 890 goto tx_err; 891 892 switch (skb->protocol) { 893 case htons(ETH_P_IP): 894 ret = ip6gre_xmit_ipv4(skb, dev); 895 break; 896 case htons(ETH_P_IPV6): 897 ret = ip6gre_xmit_ipv6(skb, dev); 898 break; 899 default: 900 ret = ip6gre_xmit_other(skb, dev); 901 break; 902 } 903 904 if (ret < 0) 905 goto tx_err; 906 907 return NETDEV_TX_OK; 908 909 tx_err: 910 stats->tx_errors++; 911 stats->tx_dropped++; 912 kfree_skb(skb); 913 return NETDEV_TX_OK; 914 } 915 916 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 917 struct net_device *dev) 918 { 919 struct ip6_tnl *t = netdev_priv(dev); 920 struct dst_entry *dst = skb_dst(skb); 921 struct net_device_stats *stats; 922 bool truncate = false; 923 int encap_limit = -1; 924 __u8 dsfield = false; 925 struct flowi6 fl6; 926 int err = -EINVAL; 927 __be16 proto; 928 __u32 mtu; 929 int nhoff; 930 int thoff; 931 932 if (!pskb_inet_may_pull(skb)) 933 goto tx_err; 934 935 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 936 goto tx_err; 937 938 if (gre_handle_offloads(skb, false)) 939 goto tx_err; 940 941 if (skb->len > dev->mtu + dev->hard_header_len) { 942 pskb_trim(skb, dev->mtu + dev->hard_header_len); 943 truncate = true; 944 } 945 946 nhoff = skb_network_header(skb) - skb_mac_header(skb); 947 if (skb->protocol == htons(ETH_P_IP) && 948 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 949 truncate = true; 950 951 thoff = skb_transport_header(skb) - skb_mac_header(skb); 952 if (skb->protocol == htons(ETH_P_IPV6) && 953 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 954 truncate = true; 955 956 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 957 goto tx_err; 958 959 t->parms.o_flags &= ~TUNNEL_KEY; 960 IPCB(skb)->flags = 0; 961 962 /* For collect_md mode, derive fl6 from the tunnel key, 963 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 964 */ 965 if (t->parms.collect_md) { 966 struct ip_tunnel_info *tun_info; 967 const struct ip_tunnel_key *key; 968 struct erspan_metadata *md; 969 __be32 tun_id; 970 971 tun_info = skb_tunnel_info(skb); 972 if (unlikely(!tun_info || 973 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 974 ip_tunnel_info_af(tun_info) != AF_INET6)) 975 return -EINVAL; 976 977 key = &tun_info->key; 978 memset(&fl6, 0, sizeof(fl6)); 979 fl6.flowi6_proto = IPPROTO_GRE; 980 fl6.daddr = key->u.ipv6.dst; 981 fl6.flowlabel = key->label; 982 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 983 984 dsfield = key->tos; 985 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 986 goto tx_err; 987 md = ip_tunnel_info_opts(tun_info); 988 if (!md) 989 goto tx_err; 990 991 tun_id = tunnel_id_to_key32(key->tun_id); 992 if (md->version == 1) { 993 erspan_build_header(skb, 994 ntohl(tun_id), 995 ntohl(md->u.index), truncate, 996 false); 997 } else if (md->version == 2) { 998 erspan_build_header_v2(skb, 999 ntohl(tun_id), 1000 md->u.md2.dir, 1001 get_hwid(&md->u.md2), 1002 truncate, false); 1003 } else { 1004 goto tx_err; 1005 } 1006 } else { 1007 switch (skb->protocol) { 1008 case htons(ETH_P_IP): 1009 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1010 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1011 &dsfield, &encap_limit); 1012 break; 1013 case htons(ETH_P_IPV6): 1014 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1015 goto tx_err; 1016 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1017 &dsfield, &encap_limit)) 1018 goto tx_err; 1019 break; 1020 default: 1021 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1022 break; 1023 } 1024 1025 if (t->parms.erspan_ver == 1) 1026 erspan_build_header(skb, ntohl(t->parms.o_key), 1027 t->parms.index, 1028 truncate, false); 1029 else if (t->parms.erspan_ver == 2) 1030 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1031 t->parms.dir, 1032 t->parms.hwid, 1033 truncate, false); 1034 else 1035 goto tx_err; 1036 1037 fl6.daddr = t->parms.raddr; 1038 } 1039 1040 /* Push GRE header. */ 1041 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1042 : htons(ETH_P_ERSPAN2); 1043 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++)); 1044 1045 /* TooBig packet may have updated dst->dev's mtu */ 1046 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1047 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu); 1048 1049 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1050 NEXTHDR_GRE); 1051 if (err != 0) { 1052 /* XXX: send ICMP error even if DF is not set. */ 1053 if (err == -EMSGSIZE) { 1054 if (skb->protocol == htons(ETH_P_IP)) 1055 icmp_send(skb, ICMP_DEST_UNREACH, 1056 ICMP_FRAG_NEEDED, htonl(mtu)); 1057 else 1058 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1059 } 1060 1061 goto tx_err; 1062 } 1063 return NETDEV_TX_OK; 1064 1065 tx_err: 1066 stats = &t->dev->stats; 1067 stats->tx_errors++; 1068 stats->tx_dropped++; 1069 kfree_skb(skb); 1070 return NETDEV_TX_OK; 1071 } 1072 1073 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1074 { 1075 struct net_device *dev = t->dev; 1076 struct __ip6_tnl_parm *p = &t->parms; 1077 struct flowi6 *fl6 = &t->fl.u.ip6; 1078 1079 if (dev->type != ARPHRD_ETHER) { 1080 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1081 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1082 } 1083 1084 /* Set up flowi template */ 1085 fl6->saddr = p->laddr; 1086 fl6->daddr = p->raddr; 1087 fl6->flowi6_oif = p->link; 1088 fl6->flowlabel = 0; 1089 fl6->flowi6_proto = IPPROTO_GRE; 1090 1091 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1092 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1093 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1094 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1095 1096 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1097 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1098 1099 if (p->flags&IP6_TNL_F_CAP_XMIT && 1100 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1101 dev->flags |= IFF_POINTOPOINT; 1102 else 1103 dev->flags &= ~IFF_POINTOPOINT; 1104 } 1105 1106 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1107 int t_hlen) 1108 { 1109 const struct __ip6_tnl_parm *p = &t->parms; 1110 struct net_device *dev = t->dev; 1111 1112 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1113 int strict = (ipv6_addr_type(&p->raddr) & 1114 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1115 1116 struct rt6_info *rt = rt6_lookup(t->net, 1117 &p->raddr, &p->laddr, 1118 p->link, NULL, strict); 1119 1120 if (!rt) 1121 return; 1122 1123 if (rt->dst.dev) { 1124 dev->needed_headroom = rt->dst.dev->hard_header_len + 1125 t_hlen; 1126 1127 if (set_mtu) { 1128 dev->mtu = rt->dst.dev->mtu - t_hlen; 1129 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1130 dev->mtu -= 8; 1131 if (dev->type == ARPHRD_ETHER) 1132 dev->mtu -= ETH_HLEN; 1133 1134 if (dev->mtu < IPV6_MIN_MTU) 1135 dev->mtu = IPV6_MIN_MTU; 1136 } 1137 } 1138 ip6_rt_put(rt); 1139 } 1140 } 1141 1142 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1143 { 1144 int t_hlen; 1145 1146 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1147 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1148 1149 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1150 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1151 return t_hlen; 1152 } 1153 1154 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1155 { 1156 ip6gre_tnl_link_config_common(t); 1157 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1158 } 1159 1160 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1161 const struct __ip6_tnl_parm *p) 1162 { 1163 t->parms.laddr = p->laddr; 1164 t->parms.raddr = p->raddr; 1165 t->parms.flags = p->flags; 1166 t->parms.hop_limit = p->hop_limit; 1167 t->parms.encap_limit = p->encap_limit; 1168 t->parms.flowinfo = p->flowinfo; 1169 t->parms.link = p->link; 1170 t->parms.proto = p->proto; 1171 t->parms.i_key = p->i_key; 1172 t->parms.o_key = p->o_key; 1173 t->parms.i_flags = p->i_flags; 1174 t->parms.o_flags = p->o_flags; 1175 t->parms.fwmark = p->fwmark; 1176 t->parms.erspan_ver = p->erspan_ver; 1177 t->parms.index = p->index; 1178 t->parms.dir = p->dir; 1179 t->parms.hwid = p->hwid; 1180 dst_cache_reset(&t->dst_cache); 1181 } 1182 1183 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1184 int set_mtu) 1185 { 1186 ip6gre_tnl_copy_tnl_parm(t, p); 1187 ip6gre_tnl_link_config(t, set_mtu); 1188 return 0; 1189 } 1190 1191 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1192 const struct ip6_tnl_parm2 *u) 1193 { 1194 p->laddr = u->laddr; 1195 p->raddr = u->raddr; 1196 p->flags = u->flags; 1197 p->hop_limit = u->hop_limit; 1198 p->encap_limit = u->encap_limit; 1199 p->flowinfo = u->flowinfo; 1200 p->link = u->link; 1201 p->i_key = u->i_key; 1202 p->o_key = u->o_key; 1203 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1204 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1205 memcpy(p->name, u->name, sizeof(u->name)); 1206 } 1207 1208 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1209 const struct __ip6_tnl_parm *p) 1210 { 1211 u->proto = IPPROTO_GRE; 1212 u->laddr = p->laddr; 1213 u->raddr = p->raddr; 1214 u->flags = p->flags; 1215 u->hop_limit = p->hop_limit; 1216 u->encap_limit = p->encap_limit; 1217 u->flowinfo = p->flowinfo; 1218 u->link = p->link; 1219 u->i_key = p->i_key; 1220 u->o_key = p->o_key; 1221 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1222 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1223 memcpy(u->name, p->name, sizeof(u->name)); 1224 } 1225 1226 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1227 struct ifreq *ifr, int cmd) 1228 { 1229 int err = 0; 1230 struct ip6_tnl_parm2 p; 1231 struct __ip6_tnl_parm p1; 1232 struct ip6_tnl *t = netdev_priv(dev); 1233 struct net *net = t->net; 1234 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1235 1236 memset(&p1, 0, sizeof(p1)); 1237 1238 switch (cmd) { 1239 case SIOCGETTUNNEL: 1240 if (dev == ign->fb_tunnel_dev) { 1241 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1242 err = -EFAULT; 1243 break; 1244 } 1245 ip6gre_tnl_parm_from_user(&p1, &p); 1246 t = ip6gre_tunnel_locate(net, &p1, 0); 1247 if (!t) 1248 t = netdev_priv(dev); 1249 } 1250 memset(&p, 0, sizeof(p)); 1251 ip6gre_tnl_parm_to_user(&p, &t->parms); 1252 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1253 err = -EFAULT; 1254 break; 1255 1256 case SIOCADDTUNNEL: 1257 case SIOCCHGTUNNEL: 1258 err = -EPERM; 1259 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1260 goto done; 1261 1262 err = -EFAULT; 1263 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1264 goto done; 1265 1266 err = -EINVAL; 1267 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1268 goto done; 1269 1270 if (!(p.i_flags&GRE_KEY)) 1271 p.i_key = 0; 1272 if (!(p.o_flags&GRE_KEY)) 1273 p.o_key = 0; 1274 1275 ip6gre_tnl_parm_from_user(&p1, &p); 1276 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1277 1278 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1279 if (t) { 1280 if (t->dev != dev) { 1281 err = -EEXIST; 1282 break; 1283 } 1284 } else { 1285 t = netdev_priv(dev); 1286 1287 ip6gre_tunnel_unlink(ign, t); 1288 synchronize_net(); 1289 ip6gre_tnl_change(t, &p1, 1); 1290 ip6gre_tunnel_link(ign, t); 1291 netdev_state_change(dev); 1292 } 1293 } 1294 1295 if (t) { 1296 err = 0; 1297 1298 memset(&p, 0, sizeof(p)); 1299 ip6gre_tnl_parm_to_user(&p, &t->parms); 1300 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1301 err = -EFAULT; 1302 } else 1303 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1304 break; 1305 1306 case SIOCDELTUNNEL: 1307 err = -EPERM; 1308 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1309 goto done; 1310 1311 if (dev == ign->fb_tunnel_dev) { 1312 err = -EFAULT; 1313 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1314 goto done; 1315 err = -ENOENT; 1316 ip6gre_tnl_parm_from_user(&p1, &p); 1317 t = ip6gre_tunnel_locate(net, &p1, 0); 1318 if (!t) 1319 goto done; 1320 err = -EPERM; 1321 if (t == netdev_priv(ign->fb_tunnel_dev)) 1322 goto done; 1323 dev = t->dev; 1324 } 1325 unregister_netdevice(dev); 1326 err = 0; 1327 break; 1328 1329 default: 1330 err = -EINVAL; 1331 } 1332 1333 done: 1334 return err; 1335 } 1336 1337 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1338 unsigned short type, const void *daddr, 1339 const void *saddr, unsigned int len) 1340 { 1341 struct ip6_tnl *t = netdev_priv(dev); 1342 struct ipv6hdr *ipv6h; 1343 __be16 *p; 1344 1345 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1346 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1347 t->fl.u.ip6.flowlabel, 1348 true, &t->fl.u.ip6)); 1349 ipv6h->hop_limit = t->parms.hop_limit; 1350 ipv6h->nexthdr = NEXTHDR_GRE; 1351 ipv6h->saddr = t->parms.laddr; 1352 ipv6h->daddr = t->parms.raddr; 1353 1354 p = (__be16 *)(ipv6h + 1); 1355 p[0] = t->parms.o_flags; 1356 p[1] = htons(type); 1357 1358 /* 1359 * Set the source hardware address. 1360 */ 1361 1362 if (saddr) 1363 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1364 if (daddr) 1365 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1366 if (!ipv6_addr_any(&ipv6h->daddr)) 1367 return t->hlen; 1368 1369 return -t->hlen; 1370 } 1371 1372 static const struct header_ops ip6gre_header_ops = { 1373 .create = ip6gre_header, 1374 }; 1375 1376 static const struct net_device_ops ip6gre_netdev_ops = { 1377 .ndo_init = ip6gre_tunnel_init, 1378 .ndo_uninit = ip6gre_tunnel_uninit, 1379 .ndo_start_xmit = ip6gre_tunnel_xmit, 1380 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1381 .ndo_change_mtu = ip6_tnl_change_mtu, 1382 .ndo_get_stats64 = ip_tunnel_get_stats64, 1383 .ndo_get_iflink = ip6_tnl_get_iflink, 1384 }; 1385 1386 static void ip6gre_dev_free(struct net_device *dev) 1387 { 1388 struct ip6_tnl *t = netdev_priv(dev); 1389 1390 gro_cells_destroy(&t->gro_cells); 1391 dst_cache_destroy(&t->dst_cache); 1392 free_percpu(dev->tstats); 1393 } 1394 1395 static void ip6gre_tunnel_setup(struct net_device *dev) 1396 { 1397 dev->netdev_ops = &ip6gre_netdev_ops; 1398 dev->needs_free_netdev = true; 1399 dev->priv_destructor = ip6gre_dev_free; 1400 1401 dev->type = ARPHRD_IP6GRE; 1402 1403 dev->flags |= IFF_NOARP; 1404 dev->addr_len = sizeof(struct in6_addr); 1405 netif_keep_dst(dev); 1406 /* This perm addr will be used as interface identifier by IPv6 */ 1407 dev->addr_assign_type = NET_ADDR_RANDOM; 1408 eth_random_addr(dev->perm_addr); 1409 } 1410 1411 #define GRE6_FEATURES (NETIF_F_SG | \ 1412 NETIF_F_FRAGLIST | \ 1413 NETIF_F_HIGHDMA | \ 1414 NETIF_F_HW_CSUM) 1415 1416 static void ip6gre_tnl_init_features(struct net_device *dev) 1417 { 1418 struct ip6_tnl *nt = netdev_priv(dev); 1419 1420 dev->features |= GRE6_FEATURES; 1421 dev->hw_features |= GRE6_FEATURES; 1422 1423 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1424 /* TCP offload with GRE SEQ is not supported, nor 1425 * can we support 2 levels of outer headers requiring 1426 * an update. 1427 */ 1428 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1429 nt->encap.type == TUNNEL_ENCAP_NONE) { 1430 dev->features |= NETIF_F_GSO_SOFTWARE; 1431 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1432 } 1433 1434 /* Can use a lockless transmit, unless we generate 1435 * output sequences 1436 */ 1437 dev->features |= NETIF_F_LLTX; 1438 } 1439 } 1440 1441 static int ip6gre_tunnel_init_common(struct net_device *dev) 1442 { 1443 struct ip6_tnl *tunnel; 1444 int ret; 1445 int t_hlen; 1446 1447 tunnel = netdev_priv(dev); 1448 1449 tunnel->dev = dev; 1450 tunnel->net = dev_net(dev); 1451 strcpy(tunnel->parms.name, dev->name); 1452 1453 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1454 if (!dev->tstats) 1455 return -ENOMEM; 1456 1457 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1458 if (ret) 1459 goto cleanup_alloc_pcpu_stats; 1460 1461 ret = gro_cells_init(&tunnel->gro_cells, dev); 1462 if (ret) 1463 goto cleanup_dst_cache_init; 1464 1465 t_hlen = ip6gre_calc_hlen(tunnel); 1466 dev->mtu = ETH_DATA_LEN - t_hlen; 1467 if (dev->type == ARPHRD_ETHER) 1468 dev->mtu -= ETH_HLEN; 1469 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1470 dev->mtu -= 8; 1471 1472 if (tunnel->parms.collect_md) { 1473 dev->features |= NETIF_F_NETNS_LOCAL; 1474 netif_keep_dst(dev); 1475 } 1476 ip6gre_tnl_init_features(dev); 1477 1478 return 0; 1479 1480 cleanup_dst_cache_init: 1481 dst_cache_destroy(&tunnel->dst_cache); 1482 cleanup_alloc_pcpu_stats: 1483 free_percpu(dev->tstats); 1484 dev->tstats = NULL; 1485 return ret; 1486 } 1487 1488 static int ip6gre_tunnel_init(struct net_device *dev) 1489 { 1490 struct ip6_tnl *tunnel; 1491 int ret; 1492 1493 ret = ip6gre_tunnel_init_common(dev); 1494 if (ret) 1495 return ret; 1496 1497 tunnel = netdev_priv(dev); 1498 1499 if (tunnel->parms.collect_md) 1500 return 0; 1501 1502 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1503 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1504 1505 if (ipv6_addr_any(&tunnel->parms.raddr)) 1506 dev->header_ops = &ip6gre_header_ops; 1507 1508 return 0; 1509 } 1510 1511 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1512 { 1513 struct ip6_tnl *tunnel = netdev_priv(dev); 1514 1515 tunnel->dev = dev; 1516 tunnel->net = dev_net(dev); 1517 strcpy(tunnel->parms.name, dev->name); 1518 1519 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1520 1521 dev_hold(dev); 1522 } 1523 1524 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1525 .handler = gre_rcv, 1526 .err_handler = ip6gre_err, 1527 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1528 }; 1529 1530 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1531 { 1532 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1533 struct net_device *dev, *aux; 1534 int prio; 1535 1536 for_each_netdev_safe(net, dev, aux) 1537 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1538 dev->rtnl_link_ops == &ip6gre_tap_ops || 1539 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1540 unregister_netdevice_queue(dev, head); 1541 1542 for (prio = 0; prio < 4; prio++) { 1543 int h; 1544 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1545 struct ip6_tnl *t; 1546 1547 t = rtnl_dereference(ign->tunnels[prio][h]); 1548 1549 while (t) { 1550 /* If dev is in the same netns, it has already 1551 * been added to the list by the previous loop. 1552 */ 1553 if (!net_eq(dev_net(t->dev), net)) 1554 unregister_netdevice_queue(t->dev, 1555 head); 1556 t = rtnl_dereference(t->next); 1557 } 1558 } 1559 } 1560 } 1561 1562 static int __net_init ip6gre_init_net(struct net *net) 1563 { 1564 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1565 int err; 1566 1567 if (!net_has_fallback_tunnels(net)) 1568 return 0; 1569 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1570 NET_NAME_UNKNOWN, 1571 ip6gre_tunnel_setup); 1572 if (!ign->fb_tunnel_dev) { 1573 err = -ENOMEM; 1574 goto err_alloc_dev; 1575 } 1576 dev_net_set(ign->fb_tunnel_dev, net); 1577 /* FB netdevice is special: we have one, and only one per netns. 1578 * Allowing to move it to another netns is clearly unsafe. 1579 */ 1580 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1581 1582 1583 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1584 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1585 1586 err = register_netdev(ign->fb_tunnel_dev); 1587 if (err) 1588 goto err_reg_dev; 1589 1590 rcu_assign_pointer(ign->tunnels_wc[0], 1591 netdev_priv(ign->fb_tunnel_dev)); 1592 return 0; 1593 1594 err_reg_dev: 1595 free_netdev(ign->fb_tunnel_dev); 1596 err_alloc_dev: 1597 return err; 1598 } 1599 1600 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1601 { 1602 struct net *net; 1603 LIST_HEAD(list); 1604 1605 rtnl_lock(); 1606 list_for_each_entry(net, net_list, exit_list) 1607 ip6gre_destroy_tunnels(net, &list); 1608 unregister_netdevice_many(&list); 1609 rtnl_unlock(); 1610 } 1611 1612 static struct pernet_operations ip6gre_net_ops = { 1613 .init = ip6gre_init_net, 1614 .exit_batch = ip6gre_exit_batch_net, 1615 .id = &ip6gre_net_id, 1616 .size = sizeof(struct ip6gre_net), 1617 }; 1618 1619 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1620 struct netlink_ext_ack *extack) 1621 { 1622 __be16 flags; 1623 1624 if (!data) 1625 return 0; 1626 1627 flags = 0; 1628 if (data[IFLA_GRE_IFLAGS]) 1629 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1630 if (data[IFLA_GRE_OFLAGS]) 1631 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1632 if (flags & (GRE_VERSION|GRE_ROUTING)) 1633 return -EINVAL; 1634 1635 return 0; 1636 } 1637 1638 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1639 struct netlink_ext_ack *extack) 1640 { 1641 struct in6_addr daddr; 1642 1643 if (tb[IFLA_ADDRESS]) { 1644 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1645 return -EINVAL; 1646 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1647 return -EADDRNOTAVAIL; 1648 } 1649 1650 if (!data) 1651 goto out; 1652 1653 if (data[IFLA_GRE_REMOTE]) { 1654 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1655 if (ipv6_addr_any(&daddr)) 1656 return -EINVAL; 1657 } 1658 1659 out: 1660 return ip6gre_tunnel_validate(tb, data, extack); 1661 } 1662 1663 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1664 struct netlink_ext_ack *extack) 1665 { 1666 __be16 flags = 0; 1667 int ret, ver = 0; 1668 1669 if (!data) 1670 return 0; 1671 1672 ret = ip6gre_tap_validate(tb, data, extack); 1673 if (ret) 1674 return ret; 1675 1676 /* ERSPAN should only have GRE sequence and key flag */ 1677 if (data[IFLA_GRE_OFLAGS]) 1678 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1679 if (data[IFLA_GRE_IFLAGS]) 1680 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1681 if (!data[IFLA_GRE_COLLECT_METADATA] && 1682 flags != (GRE_SEQ | GRE_KEY)) 1683 return -EINVAL; 1684 1685 /* ERSPAN Session ID only has 10-bit. Since we reuse 1686 * 32-bit key field as ID, check it's range. 1687 */ 1688 if (data[IFLA_GRE_IKEY] && 1689 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1690 return -EINVAL; 1691 1692 if (data[IFLA_GRE_OKEY] && 1693 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1694 return -EINVAL; 1695 1696 if (data[IFLA_GRE_ERSPAN_VER]) { 1697 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1698 if (ver != 1 && ver != 2) 1699 return -EINVAL; 1700 } 1701 1702 if (ver == 1) { 1703 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1704 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1705 1706 if (index & ~INDEX_MASK) 1707 return -EINVAL; 1708 } 1709 } else if (ver == 2) { 1710 if (data[IFLA_GRE_ERSPAN_DIR]) { 1711 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1712 1713 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1714 return -EINVAL; 1715 } 1716 1717 if (data[IFLA_GRE_ERSPAN_HWID]) { 1718 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1719 1720 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1721 return -EINVAL; 1722 } 1723 } 1724 1725 return 0; 1726 } 1727 1728 static void ip6erspan_set_version(struct nlattr *data[], 1729 struct __ip6_tnl_parm *parms) 1730 { 1731 if (!data) 1732 return; 1733 1734 parms->erspan_ver = 1; 1735 if (data[IFLA_GRE_ERSPAN_VER]) 1736 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1737 1738 if (parms->erspan_ver == 1) { 1739 if (data[IFLA_GRE_ERSPAN_INDEX]) 1740 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1741 } else if (parms->erspan_ver == 2) { 1742 if (data[IFLA_GRE_ERSPAN_DIR]) 1743 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1744 if (data[IFLA_GRE_ERSPAN_HWID]) 1745 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1746 } 1747 } 1748 1749 static void ip6gre_netlink_parms(struct nlattr *data[], 1750 struct __ip6_tnl_parm *parms) 1751 { 1752 memset(parms, 0, sizeof(*parms)); 1753 1754 if (!data) 1755 return; 1756 1757 if (data[IFLA_GRE_LINK]) 1758 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1759 1760 if (data[IFLA_GRE_IFLAGS]) 1761 parms->i_flags = gre_flags_to_tnl_flags( 1762 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1763 1764 if (data[IFLA_GRE_OFLAGS]) 1765 parms->o_flags = gre_flags_to_tnl_flags( 1766 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1767 1768 if (data[IFLA_GRE_IKEY]) 1769 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1770 1771 if (data[IFLA_GRE_OKEY]) 1772 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1773 1774 if (data[IFLA_GRE_LOCAL]) 1775 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1776 1777 if (data[IFLA_GRE_REMOTE]) 1778 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1779 1780 if (data[IFLA_GRE_TTL]) 1781 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1782 1783 if (data[IFLA_GRE_ENCAP_LIMIT]) 1784 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1785 1786 if (data[IFLA_GRE_FLOWINFO]) 1787 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1788 1789 if (data[IFLA_GRE_FLAGS]) 1790 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1791 1792 if (data[IFLA_GRE_FWMARK]) 1793 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1794 1795 if (data[IFLA_GRE_COLLECT_METADATA]) 1796 parms->collect_md = true; 1797 } 1798 1799 static int ip6gre_tap_init(struct net_device *dev) 1800 { 1801 int ret; 1802 1803 ret = ip6gre_tunnel_init_common(dev); 1804 if (ret) 1805 return ret; 1806 1807 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1808 1809 return 0; 1810 } 1811 1812 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1813 .ndo_init = ip6gre_tap_init, 1814 .ndo_uninit = ip6gre_tunnel_uninit, 1815 .ndo_start_xmit = ip6gre_tunnel_xmit, 1816 .ndo_set_mac_address = eth_mac_addr, 1817 .ndo_validate_addr = eth_validate_addr, 1818 .ndo_change_mtu = ip6_tnl_change_mtu, 1819 .ndo_get_stats64 = ip_tunnel_get_stats64, 1820 .ndo_get_iflink = ip6_tnl_get_iflink, 1821 }; 1822 1823 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1824 { 1825 int t_hlen; 1826 1827 tunnel->tun_hlen = 8; 1828 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1829 erspan_hdr_len(tunnel->parms.erspan_ver); 1830 1831 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1832 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1833 return t_hlen; 1834 } 1835 1836 static int ip6erspan_tap_init(struct net_device *dev) 1837 { 1838 struct ip6_tnl *tunnel; 1839 int t_hlen; 1840 int ret; 1841 1842 tunnel = netdev_priv(dev); 1843 1844 tunnel->dev = dev; 1845 tunnel->net = dev_net(dev); 1846 strcpy(tunnel->parms.name, dev->name); 1847 1848 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1849 if (!dev->tstats) 1850 return -ENOMEM; 1851 1852 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1853 if (ret) 1854 goto cleanup_alloc_pcpu_stats; 1855 1856 ret = gro_cells_init(&tunnel->gro_cells, dev); 1857 if (ret) 1858 goto cleanup_dst_cache_init; 1859 1860 t_hlen = ip6erspan_calc_hlen(tunnel); 1861 dev->mtu = ETH_DATA_LEN - t_hlen; 1862 if (dev->type == ARPHRD_ETHER) 1863 dev->mtu -= ETH_HLEN; 1864 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1865 dev->mtu -= 8; 1866 1867 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1868 ip6erspan_tnl_link_config(tunnel, 1); 1869 1870 return 0; 1871 1872 cleanup_dst_cache_init: 1873 dst_cache_destroy(&tunnel->dst_cache); 1874 cleanup_alloc_pcpu_stats: 1875 free_percpu(dev->tstats); 1876 dev->tstats = NULL; 1877 return ret; 1878 } 1879 1880 static const struct net_device_ops ip6erspan_netdev_ops = { 1881 .ndo_init = ip6erspan_tap_init, 1882 .ndo_uninit = ip6erspan_tunnel_uninit, 1883 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1884 .ndo_set_mac_address = eth_mac_addr, 1885 .ndo_validate_addr = eth_validate_addr, 1886 .ndo_change_mtu = ip6_tnl_change_mtu, 1887 .ndo_get_stats64 = ip_tunnel_get_stats64, 1888 .ndo_get_iflink = ip6_tnl_get_iflink, 1889 }; 1890 1891 static void ip6gre_tap_setup(struct net_device *dev) 1892 { 1893 1894 ether_setup(dev); 1895 1896 dev->max_mtu = 0; 1897 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1898 dev->needs_free_netdev = true; 1899 dev->priv_destructor = ip6gre_dev_free; 1900 1901 dev->features |= NETIF_F_NETNS_LOCAL; 1902 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1903 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1904 netif_keep_dst(dev); 1905 } 1906 1907 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1908 struct ip_tunnel_encap *ipencap) 1909 { 1910 bool ret = false; 1911 1912 memset(ipencap, 0, sizeof(*ipencap)); 1913 1914 if (!data) 1915 return ret; 1916 1917 if (data[IFLA_GRE_ENCAP_TYPE]) { 1918 ret = true; 1919 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1920 } 1921 1922 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1923 ret = true; 1924 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1925 } 1926 1927 if (data[IFLA_GRE_ENCAP_SPORT]) { 1928 ret = true; 1929 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1930 } 1931 1932 if (data[IFLA_GRE_ENCAP_DPORT]) { 1933 ret = true; 1934 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1935 } 1936 1937 return ret; 1938 } 1939 1940 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1941 struct nlattr *tb[], struct nlattr *data[], 1942 struct netlink_ext_ack *extack) 1943 { 1944 struct ip6_tnl *nt; 1945 struct ip_tunnel_encap ipencap; 1946 int err; 1947 1948 nt = netdev_priv(dev); 1949 1950 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1951 int err = ip6_tnl_encap_setup(nt, &ipencap); 1952 1953 if (err < 0) 1954 return err; 1955 } 1956 1957 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1958 eth_hw_addr_random(dev); 1959 1960 nt->dev = dev; 1961 nt->net = dev_net(dev); 1962 1963 err = register_netdevice(dev); 1964 if (err) 1965 goto out; 1966 1967 if (tb[IFLA_MTU]) 1968 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1969 1970 dev_hold(dev); 1971 1972 out: 1973 return err; 1974 } 1975 1976 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1977 struct nlattr *tb[], struct nlattr *data[], 1978 struct netlink_ext_ack *extack) 1979 { 1980 struct ip6_tnl *nt = netdev_priv(dev); 1981 struct net *net = dev_net(dev); 1982 struct ip6gre_net *ign; 1983 int err; 1984 1985 ip6gre_netlink_parms(data, &nt->parms); 1986 ign = net_generic(net, ip6gre_net_id); 1987 1988 if (nt->parms.collect_md) { 1989 if (rtnl_dereference(ign->collect_md_tun)) 1990 return -EEXIST; 1991 } else { 1992 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1993 return -EEXIST; 1994 } 1995 1996 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 1997 if (!err) { 1998 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1999 ip6gre_tunnel_link_md(ign, nt); 2000 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2001 } 2002 return err; 2003 } 2004 2005 static struct ip6_tnl * 2006 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2007 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2008 struct netlink_ext_ack *extack) 2009 { 2010 struct ip6_tnl *t, *nt = netdev_priv(dev); 2011 struct net *net = nt->net; 2012 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2013 struct ip_tunnel_encap ipencap; 2014 2015 if (dev == ign->fb_tunnel_dev) 2016 return ERR_PTR(-EINVAL); 2017 2018 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2019 int err = ip6_tnl_encap_setup(nt, &ipencap); 2020 2021 if (err < 0) 2022 return ERR_PTR(err); 2023 } 2024 2025 ip6gre_netlink_parms(data, p_p); 2026 2027 t = ip6gre_tunnel_locate(net, p_p, 0); 2028 2029 if (t) { 2030 if (t->dev != dev) 2031 return ERR_PTR(-EEXIST); 2032 } else { 2033 t = nt; 2034 } 2035 2036 return t; 2037 } 2038 2039 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2040 struct nlattr *data[], 2041 struct netlink_ext_ack *extack) 2042 { 2043 struct ip6_tnl *t = netdev_priv(dev); 2044 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2045 struct __ip6_tnl_parm p; 2046 2047 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2048 if (IS_ERR(t)) 2049 return PTR_ERR(t); 2050 2051 ip6gre_tunnel_unlink_md(ign, t); 2052 ip6gre_tunnel_unlink(ign, t); 2053 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2054 ip6gre_tunnel_link_md(ign, t); 2055 ip6gre_tunnel_link(ign, t); 2056 return 0; 2057 } 2058 2059 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2060 { 2061 struct net *net = dev_net(dev); 2062 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2063 2064 if (dev != ign->fb_tunnel_dev) 2065 unregister_netdevice_queue(dev, head); 2066 } 2067 2068 static size_t ip6gre_get_size(const struct net_device *dev) 2069 { 2070 return 2071 /* IFLA_GRE_LINK */ 2072 nla_total_size(4) + 2073 /* IFLA_GRE_IFLAGS */ 2074 nla_total_size(2) + 2075 /* IFLA_GRE_OFLAGS */ 2076 nla_total_size(2) + 2077 /* IFLA_GRE_IKEY */ 2078 nla_total_size(4) + 2079 /* IFLA_GRE_OKEY */ 2080 nla_total_size(4) + 2081 /* IFLA_GRE_LOCAL */ 2082 nla_total_size(sizeof(struct in6_addr)) + 2083 /* IFLA_GRE_REMOTE */ 2084 nla_total_size(sizeof(struct in6_addr)) + 2085 /* IFLA_GRE_TTL */ 2086 nla_total_size(1) + 2087 /* IFLA_GRE_ENCAP_LIMIT */ 2088 nla_total_size(1) + 2089 /* IFLA_GRE_FLOWINFO */ 2090 nla_total_size(4) + 2091 /* IFLA_GRE_FLAGS */ 2092 nla_total_size(4) + 2093 /* IFLA_GRE_ENCAP_TYPE */ 2094 nla_total_size(2) + 2095 /* IFLA_GRE_ENCAP_FLAGS */ 2096 nla_total_size(2) + 2097 /* IFLA_GRE_ENCAP_SPORT */ 2098 nla_total_size(2) + 2099 /* IFLA_GRE_ENCAP_DPORT */ 2100 nla_total_size(2) + 2101 /* IFLA_GRE_COLLECT_METADATA */ 2102 nla_total_size(0) + 2103 /* IFLA_GRE_FWMARK */ 2104 nla_total_size(4) + 2105 /* IFLA_GRE_ERSPAN_INDEX */ 2106 nla_total_size(4) + 2107 0; 2108 } 2109 2110 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2111 { 2112 struct ip6_tnl *t = netdev_priv(dev); 2113 struct __ip6_tnl_parm *p = &t->parms; 2114 __be16 o_flags = p->o_flags; 2115 2116 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2117 if (!p->collect_md) 2118 o_flags |= TUNNEL_KEY; 2119 2120 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2121 goto nla_put_failure; 2122 2123 if (p->erspan_ver == 1) { 2124 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2125 goto nla_put_failure; 2126 } else { 2127 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2128 goto nla_put_failure; 2129 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2130 goto nla_put_failure; 2131 } 2132 } 2133 2134 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2135 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2136 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2137 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2138 gre_tnl_flags_to_gre_flags(o_flags)) || 2139 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2140 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2141 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2142 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2143 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2144 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2145 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2146 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2147 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2148 goto nla_put_failure; 2149 2150 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2151 t->encap.type) || 2152 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2153 t->encap.sport) || 2154 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2155 t->encap.dport) || 2156 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2157 t->encap.flags)) 2158 goto nla_put_failure; 2159 2160 if (p->collect_md) { 2161 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2162 goto nla_put_failure; 2163 } 2164 2165 return 0; 2166 2167 nla_put_failure: 2168 return -EMSGSIZE; 2169 } 2170 2171 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2172 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2173 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2174 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2175 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2176 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2177 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 2178 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 2179 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2180 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2181 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2182 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2183 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2184 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2185 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2186 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2187 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2188 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2189 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2190 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2191 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2192 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2193 }; 2194 2195 static void ip6erspan_tap_setup(struct net_device *dev) 2196 { 2197 ether_setup(dev); 2198 2199 dev->netdev_ops = &ip6erspan_netdev_ops; 2200 dev->needs_free_netdev = true; 2201 dev->priv_destructor = ip6gre_dev_free; 2202 2203 dev->features |= NETIF_F_NETNS_LOCAL; 2204 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2205 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2206 netif_keep_dst(dev); 2207 } 2208 2209 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2210 struct nlattr *tb[], struct nlattr *data[], 2211 struct netlink_ext_ack *extack) 2212 { 2213 struct ip6_tnl *nt = netdev_priv(dev); 2214 struct net *net = dev_net(dev); 2215 struct ip6gre_net *ign; 2216 int err; 2217 2218 ip6gre_netlink_parms(data, &nt->parms); 2219 ip6erspan_set_version(data, &nt->parms); 2220 ign = net_generic(net, ip6gre_net_id); 2221 2222 if (nt->parms.collect_md) { 2223 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2224 return -EEXIST; 2225 } else { 2226 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2227 return -EEXIST; 2228 } 2229 2230 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2231 if (!err) { 2232 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2233 ip6erspan_tunnel_link_md(ign, nt); 2234 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2235 } 2236 return err; 2237 } 2238 2239 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2240 { 2241 ip6gre_tnl_link_config_common(t); 2242 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2243 } 2244 2245 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2246 const struct __ip6_tnl_parm *p, int set_mtu) 2247 { 2248 ip6gre_tnl_copy_tnl_parm(t, p); 2249 ip6erspan_tnl_link_config(t, set_mtu); 2250 return 0; 2251 } 2252 2253 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2254 struct nlattr *data[], 2255 struct netlink_ext_ack *extack) 2256 { 2257 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2258 struct __ip6_tnl_parm p; 2259 struct ip6_tnl *t; 2260 2261 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2262 if (IS_ERR(t)) 2263 return PTR_ERR(t); 2264 2265 ip6erspan_set_version(data, &p); 2266 ip6gre_tunnel_unlink_md(ign, t); 2267 ip6gre_tunnel_unlink(ign, t); 2268 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2269 ip6erspan_tunnel_link_md(ign, t); 2270 ip6gre_tunnel_link(ign, t); 2271 return 0; 2272 } 2273 2274 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2275 .kind = "ip6gre", 2276 .maxtype = IFLA_GRE_MAX, 2277 .policy = ip6gre_policy, 2278 .priv_size = sizeof(struct ip6_tnl), 2279 .setup = ip6gre_tunnel_setup, 2280 .validate = ip6gre_tunnel_validate, 2281 .newlink = ip6gre_newlink, 2282 .changelink = ip6gre_changelink, 2283 .dellink = ip6gre_dellink, 2284 .get_size = ip6gre_get_size, 2285 .fill_info = ip6gre_fill_info, 2286 .get_link_net = ip6_tnl_get_link_net, 2287 }; 2288 2289 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2290 .kind = "ip6gretap", 2291 .maxtype = IFLA_GRE_MAX, 2292 .policy = ip6gre_policy, 2293 .priv_size = sizeof(struct ip6_tnl), 2294 .setup = ip6gre_tap_setup, 2295 .validate = ip6gre_tap_validate, 2296 .newlink = ip6gre_newlink, 2297 .changelink = ip6gre_changelink, 2298 .get_size = ip6gre_get_size, 2299 .fill_info = ip6gre_fill_info, 2300 .get_link_net = ip6_tnl_get_link_net, 2301 }; 2302 2303 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2304 .kind = "ip6erspan", 2305 .maxtype = IFLA_GRE_MAX, 2306 .policy = ip6gre_policy, 2307 .priv_size = sizeof(struct ip6_tnl), 2308 .setup = ip6erspan_tap_setup, 2309 .validate = ip6erspan_tap_validate, 2310 .newlink = ip6erspan_newlink, 2311 .changelink = ip6erspan_changelink, 2312 .get_size = ip6gre_get_size, 2313 .fill_info = ip6gre_fill_info, 2314 .get_link_net = ip6_tnl_get_link_net, 2315 }; 2316 2317 /* 2318 * And now the modules code and kernel interface. 2319 */ 2320 2321 static int __init ip6gre_init(void) 2322 { 2323 int err; 2324 2325 pr_info("GRE over IPv6 tunneling driver\n"); 2326 2327 err = register_pernet_device(&ip6gre_net_ops); 2328 if (err < 0) 2329 return err; 2330 2331 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2332 if (err < 0) { 2333 pr_info("%s: can't add protocol\n", __func__); 2334 goto add_proto_failed; 2335 } 2336 2337 err = rtnl_link_register(&ip6gre_link_ops); 2338 if (err < 0) 2339 goto rtnl_link_failed; 2340 2341 err = rtnl_link_register(&ip6gre_tap_ops); 2342 if (err < 0) 2343 goto tap_ops_failed; 2344 2345 err = rtnl_link_register(&ip6erspan_tap_ops); 2346 if (err < 0) 2347 goto erspan_link_failed; 2348 2349 out: 2350 return err; 2351 2352 erspan_link_failed: 2353 rtnl_link_unregister(&ip6gre_tap_ops); 2354 tap_ops_failed: 2355 rtnl_link_unregister(&ip6gre_link_ops); 2356 rtnl_link_failed: 2357 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2358 add_proto_failed: 2359 unregister_pernet_device(&ip6gre_net_ops); 2360 goto out; 2361 } 2362 2363 static void __exit ip6gre_fini(void) 2364 { 2365 rtnl_link_unregister(&ip6gre_tap_ops); 2366 rtnl_link_unregister(&ip6gre_link_ops); 2367 rtnl_link_unregister(&ip6erspan_tap_ops); 2368 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2369 unregister_pernet_device(&ip6gre_net_ops); 2370 } 2371 2372 module_init(ip6gre_init); 2373 module_exit(ip6gre_fini); 2374 MODULE_LICENSE("GPL"); 2375 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2376 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2377 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2378 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2379 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2380 MODULE_ALIAS_NETDEV("ip6gre0"); 2381