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 net_device *fb_tunnel_dev; 75 }; 76 77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 78 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 79 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly; 80 static int ip6gre_tunnel_init(struct net_device *dev); 81 static void ip6gre_tunnel_setup(struct net_device *dev); 82 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t); 83 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu); 84 85 /* Tunnel hash table */ 86 87 /* 88 4 hash tables: 89 90 3: (remote,local) 91 2: (remote,*) 92 1: (*,local) 93 0: (*,*) 94 95 We require exact key match i.e. if a key is present in packet 96 it will match only tunnel with the same key; if it is not present, 97 it will match only keyless tunnel. 98 99 All keysless packets, if not matched configured keyless tunnels 100 will match fallback tunnel. 101 */ 102 103 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 104 static u32 HASH_ADDR(const struct in6_addr *addr) 105 { 106 u32 hash = ipv6_addr_hash(addr); 107 108 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 109 } 110 111 #define tunnels_r_l tunnels[3] 112 #define tunnels_r tunnels[2] 113 #define tunnels_l tunnels[1] 114 #define tunnels_wc tunnels[0] 115 116 /* Given src, dst and key, find appropriate for input tunnel. */ 117 118 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 119 const struct in6_addr *remote, const struct in6_addr *local, 120 __be32 key, __be16 gre_proto) 121 { 122 struct net *net = dev_net(dev); 123 int link = dev->ifindex; 124 unsigned int h0 = HASH_ADDR(remote); 125 unsigned int h1 = HASH_KEY(key); 126 struct ip6_tnl *t, *cand = NULL; 127 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 128 int dev_type = (gre_proto == htons(ETH_P_TEB) || 129 gre_proto == htons(ETH_P_ERSPAN)) ? 130 ARPHRD_ETHER : ARPHRD_IP6GRE; 131 int score, cand_score = 4; 132 133 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 134 if (!ipv6_addr_equal(local, &t->parms.laddr) || 135 !ipv6_addr_equal(remote, &t->parms.raddr) || 136 key != t->parms.i_key || 137 !(t->dev->flags & IFF_UP)) 138 continue; 139 140 if (t->dev->type != ARPHRD_IP6GRE && 141 t->dev->type != dev_type) 142 continue; 143 144 score = 0; 145 if (t->parms.link != link) 146 score |= 1; 147 if (t->dev->type != dev_type) 148 score |= 2; 149 if (score == 0) 150 return t; 151 152 if (score < cand_score) { 153 cand = t; 154 cand_score = score; 155 } 156 } 157 158 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 159 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 160 key != t->parms.i_key || 161 !(t->dev->flags & IFF_UP)) 162 continue; 163 164 if (t->dev->type != ARPHRD_IP6GRE && 165 t->dev->type != dev_type) 166 continue; 167 168 score = 0; 169 if (t->parms.link != link) 170 score |= 1; 171 if (t->dev->type != dev_type) 172 score |= 2; 173 if (score == 0) 174 return t; 175 176 if (score < cand_score) { 177 cand = t; 178 cand_score = score; 179 } 180 } 181 182 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 183 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 184 (!ipv6_addr_equal(local, &t->parms.raddr) || 185 !ipv6_addr_is_multicast(local))) || 186 key != t->parms.i_key || 187 !(t->dev->flags & IFF_UP)) 188 continue; 189 190 if (t->dev->type != ARPHRD_IP6GRE && 191 t->dev->type != dev_type) 192 continue; 193 194 score = 0; 195 if (t->parms.link != link) 196 score |= 1; 197 if (t->dev->type != dev_type) 198 score |= 2; 199 if (score == 0) 200 return t; 201 202 if (score < cand_score) { 203 cand = t; 204 cand_score = score; 205 } 206 } 207 208 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 209 if (t->parms.i_key != key || 210 !(t->dev->flags & IFF_UP)) 211 continue; 212 213 if (t->dev->type != ARPHRD_IP6GRE && 214 t->dev->type != dev_type) 215 continue; 216 217 score = 0; 218 if (t->parms.link != link) 219 score |= 1; 220 if (t->dev->type != dev_type) 221 score |= 2; 222 if (score == 0) 223 return t; 224 225 if (score < cand_score) { 226 cand = t; 227 cand_score = score; 228 } 229 } 230 231 if (cand) 232 return cand; 233 234 t = rcu_dereference(ign->collect_md_tun); 235 if (t && t->dev->flags & IFF_UP) 236 return t; 237 238 dev = ign->fb_tunnel_dev; 239 if (dev->flags & IFF_UP) 240 return netdev_priv(dev); 241 242 return NULL; 243 } 244 245 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 246 const struct __ip6_tnl_parm *p) 247 { 248 const struct in6_addr *remote = &p->raddr; 249 const struct in6_addr *local = &p->laddr; 250 unsigned int h = HASH_KEY(p->i_key); 251 int prio = 0; 252 253 if (!ipv6_addr_any(local)) 254 prio |= 1; 255 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 256 prio |= 2; 257 h ^= HASH_ADDR(remote); 258 } 259 260 return &ign->tunnels[prio][h]; 261 } 262 263 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 264 const struct ip6_tnl *t) 265 { 266 return __ip6gre_bucket(ign, &t->parms); 267 } 268 269 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 270 { 271 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 272 273 if (t->parms.collect_md) 274 rcu_assign_pointer(ign->collect_md_tun, t); 275 276 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 277 rcu_assign_pointer(*tp, t); 278 } 279 280 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 281 { 282 struct ip6_tnl __rcu **tp; 283 struct ip6_tnl *iter; 284 285 if (t->parms.collect_md) 286 rcu_assign_pointer(ign->collect_md_tun, NULL); 287 288 for (tp = ip6gre_bucket(ign, t); 289 (iter = rtnl_dereference(*tp)) != NULL; 290 tp = &iter->next) { 291 if (t == iter) { 292 rcu_assign_pointer(*tp, t->next); 293 break; 294 } 295 } 296 } 297 298 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 299 const struct __ip6_tnl_parm *parms, 300 int type) 301 { 302 const struct in6_addr *remote = &parms->raddr; 303 const struct in6_addr *local = &parms->laddr; 304 __be32 key = parms->i_key; 305 int link = parms->link; 306 struct ip6_tnl *t; 307 struct ip6_tnl __rcu **tp; 308 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 309 310 for (tp = __ip6gre_bucket(ign, parms); 311 (t = rtnl_dereference(*tp)) != NULL; 312 tp = &t->next) 313 if (ipv6_addr_equal(local, &t->parms.laddr) && 314 ipv6_addr_equal(remote, &t->parms.raddr) && 315 key == t->parms.i_key && 316 link == t->parms.link && 317 type == t->dev->type) 318 break; 319 320 return t; 321 } 322 323 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 324 const struct __ip6_tnl_parm *parms, int create) 325 { 326 struct ip6_tnl *t, *nt; 327 struct net_device *dev; 328 char name[IFNAMSIZ]; 329 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 330 331 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 332 if (t && create) 333 return NULL; 334 if (t || !create) 335 return t; 336 337 if (parms->name[0]) 338 strlcpy(name, parms->name, IFNAMSIZ); 339 else 340 strcpy(name, "ip6gre%d"); 341 342 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 343 ip6gre_tunnel_setup); 344 if (!dev) 345 return NULL; 346 347 dev_net_set(dev, net); 348 349 nt = netdev_priv(dev); 350 nt->parms = *parms; 351 dev->rtnl_link_ops = &ip6gre_link_ops; 352 353 nt->dev = dev; 354 nt->net = dev_net(dev); 355 356 if (register_netdevice(dev) < 0) 357 goto failed_free; 358 359 ip6gre_tnl_link_config(nt, 1); 360 361 /* Can use a lockless transmit, unless we generate output sequences */ 362 if (!(nt->parms.o_flags & TUNNEL_SEQ)) 363 dev->features |= NETIF_F_LLTX; 364 365 dev_hold(dev); 366 ip6gre_tunnel_link(ign, nt); 367 return nt; 368 369 failed_free: 370 free_netdev(dev); 371 return NULL; 372 } 373 374 static void ip6gre_tunnel_uninit(struct net_device *dev) 375 { 376 struct ip6_tnl *t = netdev_priv(dev); 377 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 378 379 ip6gre_tunnel_unlink(ign, t); 380 dst_cache_reset(&t->dst_cache); 381 dev_put(dev); 382 } 383 384 385 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 386 u8 type, u8 code, int offset, __be32 info) 387 { 388 struct net *net = dev_net(skb->dev); 389 const struct gre_base_hdr *greh; 390 const struct ipv6hdr *ipv6h; 391 int grehlen = sizeof(*greh); 392 struct ip6_tnl *t; 393 int key_off = 0; 394 __be16 flags; 395 __be32 key; 396 397 if (!pskb_may_pull(skb, offset + grehlen)) 398 return; 399 greh = (const struct gre_base_hdr *)(skb->data + offset); 400 flags = greh->flags; 401 if (flags & (GRE_VERSION | GRE_ROUTING)) 402 return; 403 if (flags & GRE_CSUM) 404 grehlen += 4; 405 if (flags & GRE_KEY) { 406 key_off = grehlen + offset; 407 grehlen += 4; 408 } 409 410 if (!pskb_may_pull(skb, offset + grehlen)) 411 return; 412 ipv6h = (const struct ipv6hdr *)skb->data; 413 greh = (const struct gre_base_hdr *)(skb->data + offset); 414 key = key_off ? *(__be32 *)(skb->data + key_off) : 0; 415 416 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 417 key, greh->protocol); 418 if (!t) 419 return; 420 421 switch (type) { 422 struct ipv6_tlv_tnl_enc_lim *tel; 423 __u32 teli; 424 case ICMPV6_DEST_UNREACH: 425 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 426 t->parms.name); 427 if (code != ICMPV6_PORT_UNREACH) 428 break; 429 return; 430 case ICMPV6_TIME_EXCEED: 431 if (code == ICMPV6_EXC_HOPLIMIT) { 432 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 433 t->parms.name); 434 break; 435 } 436 return; 437 case ICMPV6_PARAMPROB: 438 teli = 0; 439 if (code == ICMPV6_HDR_FIELD) 440 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 441 442 if (teli && teli == be32_to_cpu(info) - 2) { 443 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 444 if (tel->encap_limit == 0) { 445 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 446 t->parms.name); 447 } 448 } else { 449 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 450 t->parms.name); 451 } 452 return; 453 case ICMPV6_PKT_TOOBIG: 454 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 455 return; 456 case NDISC_REDIRECT: 457 ip6_redirect(skb, net, skb->dev->ifindex, 0, 458 sock_net_uid(net, NULL)); 459 return; 460 } 461 462 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 463 t->err_count++; 464 else 465 t->err_count = 1; 466 t->err_time = jiffies; 467 } 468 469 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 470 { 471 const struct ipv6hdr *ipv6h; 472 struct ip6_tnl *tunnel; 473 474 ipv6h = ipv6_hdr(skb); 475 tunnel = ip6gre_tunnel_lookup(skb->dev, 476 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 477 tpi->proto); 478 if (tunnel) { 479 if (tunnel->parms.collect_md) { 480 struct metadata_dst *tun_dst; 481 __be64 tun_id; 482 __be16 flags; 483 484 flags = tpi->flags; 485 tun_id = key32_to_tunnel_id(tpi->key); 486 487 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 488 if (!tun_dst) 489 return PACKET_REJECT; 490 491 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 492 } else { 493 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 494 } 495 496 return PACKET_RCVD; 497 } 498 499 return PACKET_REJECT; 500 } 501 502 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len, 503 struct tnl_ptk_info *tpi) 504 { 505 struct erspan_base_hdr *ershdr; 506 struct erspan_metadata *pkt_md; 507 const struct ipv6hdr *ipv6h; 508 struct ip6_tnl *tunnel; 509 u8 ver; 510 511 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr)))) 512 return PACKET_REJECT; 513 514 ipv6h = ipv6_hdr(skb); 515 ershdr = (struct erspan_base_hdr *)skb->data; 516 ver = ershdr->ver; 517 tpi->key = cpu_to_be32(get_session_id(ershdr)); 518 519 tunnel = ip6gre_tunnel_lookup(skb->dev, 520 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 521 tpi->proto); 522 if (tunnel) { 523 int len = erspan_hdr_len(ver); 524 525 if (unlikely(!pskb_may_pull(skb, len))) 526 return PACKET_REJECT; 527 528 ershdr = (struct erspan_base_hdr *)skb->data; 529 pkt_md = (struct erspan_metadata *)(ershdr + 1); 530 531 if (__iptunnel_pull_header(skb, len, 532 htons(ETH_P_TEB), 533 false, false) < 0) 534 return PACKET_REJECT; 535 536 if (tunnel->parms.collect_md) { 537 struct metadata_dst *tun_dst; 538 struct ip_tunnel_info *info; 539 struct erspan_metadata *md; 540 __be64 tun_id; 541 __be16 flags; 542 543 tpi->flags |= TUNNEL_KEY; 544 flags = tpi->flags; 545 tun_id = key32_to_tunnel_id(tpi->key); 546 547 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 548 sizeof(*md)); 549 if (!tun_dst) 550 return PACKET_REJECT; 551 552 info = &tun_dst->u.tun_info; 553 md = ip_tunnel_info_opts(info); 554 555 memcpy(md, pkt_md, sizeof(*md)); 556 md->version = ver; 557 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 558 info->options_len = sizeof(*md); 559 560 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 561 562 } else { 563 tunnel->parms.erspan_ver = ver; 564 565 if (ver == 1) { 566 tunnel->parms.index = ntohl(pkt_md->u.index); 567 } else { 568 tunnel->parms.dir = pkt_md->u.md2.dir; 569 tunnel->parms.hwid = get_hwid(&pkt_md->u.md2); 570 } 571 572 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 573 } 574 575 return PACKET_RCVD; 576 } 577 578 return PACKET_REJECT; 579 } 580 581 static int gre_rcv(struct sk_buff *skb) 582 { 583 struct tnl_ptk_info tpi; 584 bool csum_err = false; 585 int hdr_len; 586 587 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 588 if (hdr_len < 0) 589 goto drop; 590 591 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 592 goto drop; 593 594 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 595 tpi.proto == htons(ETH_P_ERSPAN2))) { 596 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD) 597 return 0; 598 goto out; 599 } 600 601 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 602 return 0; 603 604 out: 605 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 606 drop: 607 kfree_skb(skb); 608 return 0; 609 } 610 611 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 612 { 613 return iptunnel_handle_offloads(skb, 614 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 615 } 616 617 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 618 struct net_device *dev, 619 struct flowi6 *fl6, __u8 *dsfield, 620 int *encap_limit) 621 { 622 const struct iphdr *iph = ip_hdr(skb); 623 struct ip6_tnl *t = netdev_priv(dev); 624 625 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 626 *encap_limit = t->parms.encap_limit; 627 628 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 629 630 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 631 *dsfield = ipv4_get_dsfield(iph); 632 else 633 *dsfield = ip6_tclass(t->parms.flowinfo); 634 635 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 636 fl6->flowi6_mark = skb->mark; 637 else 638 fl6->flowi6_mark = t->parms.fwmark; 639 640 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 641 } 642 643 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 644 struct net_device *dev, 645 struct flowi6 *fl6, __u8 *dsfield, 646 int *encap_limit) 647 { 648 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 649 struct ip6_tnl *t = netdev_priv(dev); 650 __u16 offset; 651 652 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 653 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 654 655 if (offset > 0) { 656 struct ipv6_tlv_tnl_enc_lim *tel; 657 658 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 659 if (tel->encap_limit == 0) { 660 icmpv6_send(skb, ICMPV6_PARAMPROB, 661 ICMPV6_HDR_FIELD, offset + 2); 662 return -1; 663 } 664 *encap_limit = tel->encap_limit - 1; 665 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 666 *encap_limit = t->parms.encap_limit; 667 } 668 669 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 670 671 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 672 *dsfield = ipv6_get_dsfield(ipv6h); 673 else 674 *dsfield = ip6_tclass(t->parms.flowinfo); 675 676 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 677 fl6->flowlabel |= ip6_flowlabel(ipv6h); 678 679 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 680 fl6->flowi6_mark = skb->mark; 681 else 682 fl6->flowi6_mark = t->parms.fwmark; 683 684 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 685 686 return 0; 687 } 688 689 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 690 struct net_device *dev, __u8 dsfield, 691 struct flowi6 *fl6, int encap_limit, 692 __u32 *pmtu, __be16 proto) 693 { 694 struct ip6_tnl *tunnel = netdev_priv(dev); 695 __be16 protocol; 696 697 if (dev->type == ARPHRD_ETHER) 698 IPCB(skb)->flags = 0; 699 700 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 701 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 702 else 703 fl6->daddr = tunnel->parms.raddr; 704 705 if (tunnel->parms.o_flags & TUNNEL_SEQ) 706 tunnel->o_seqno++; 707 708 /* Push GRE header. */ 709 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 710 711 if (tunnel->parms.collect_md) { 712 struct ip_tunnel_info *tun_info; 713 const struct ip_tunnel_key *key; 714 __be16 flags; 715 716 tun_info = skb_tunnel_info(skb); 717 if (unlikely(!tun_info || 718 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 719 ip_tunnel_info_af(tun_info) != AF_INET6)) 720 return -EINVAL; 721 722 key = &tun_info->key; 723 memset(fl6, 0, sizeof(*fl6)); 724 fl6->flowi6_proto = IPPROTO_GRE; 725 fl6->daddr = key->u.ipv6.dst; 726 fl6->flowlabel = key->label; 727 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 728 729 dsfield = key->tos; 730 flags = key->tun_flags & (TUNNEL_CSUM | TUNNEL_KEY); 731 tunnel->tun_hlen = gre_calc_hlen(flags); 732 733 gre_build_header(skb, tunnel->tun_hlen, 734 flags, protocol, 735 tunnel_id_to_key32(tun_info->key.tun_id), 0); 736 737 } else { 738 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 739 protocol, tunnel->parms.o_key, 740 htonl(tunnel->o_seqno)); 741 } 742 743 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 744 NEXTHDR_GRE); 745 } 746 747 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 748 { 749 struct ip6_tnl *t = netdev_priv(dev); 750 int encap_limit = -1; 751 struct flowi6 fl6; 752 __u8 dsfield = 0; 753 __u32 mtu; 754 int err; 755 756 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 757 758 if (!t->parms.collect_md) 759 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 760 &dsfield, &encap_limit); 761 762 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 763 if (err) 764 return -1; 765 766 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 767 skb->protocol); 768 if (err != 0) { 769 /* XXX: send ICMP error even if DF is not set. */ 770 if (err == -EMSGSIZE) 771 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 772 htonl(mtu)); 773 return -1; 774 } 775 776 return 0; 777 } 778 779 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 780 { 781 struct ip6_tnl *t = netdev_priv(dev); 782 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 783 int encap_limit = -1; 784 struct flowi6 fl6; 785 __u8 dsfield = 0; 786 __u32 mtu; 787 int err; 788 789 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 790 return -1; 791 792 if (!t->parms.collect_md && 793 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 794 return -1; 795 796 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 797 return -1; 798 799 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 800 &mtu, skb->protocol); 801 if (err != 0) { 802 if (err == -EMSGSIZE) 803 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 804 return -1; 805 } 806 807 return 0; 808 } 809 810 /** 811 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 812 * @t: the outgoing tunnel device 813 * @hdr: IPv6 header from the incoming packet 814 * 815 * Description: 816 * Avoid trivial tunneling loop by checking that tunnel exit-point 817 * doesn't match source of incoming packet. 818 * 819 * Return: 820 * 1 if conflict, 821 * 0 else 822 **/ 823 824 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 825 const struct ipv6hdr *hdr) 826 { 827 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 828 } 829 830 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 831 { 832 struct ip6_tnl *t = netdev_priv(dev); 833 int encap_limit = -1; 834 struct flowi6 fl6; 835 __u32 mtu; 836 int err; 837 838 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 839 encap_limit = t->parms.encap_limit; 840 841 if (!t->parms.collect_md) 842 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 843 844 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 845 if (err) 846 return err; 847 848 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 849 850 return err; 851 } 852 853 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 854 struct net_device *dev) 855 { 856 struct ip6_tnl *t = netdev_priv(dev); 857 struct net_device_stats *stats = &t->dev->stats; 858 int ret; 859 860 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 861 goto tx_err; 862 863 switch (skb->protocol) { 864 case htons(ETH_P_IP): 865 ret = ip6gre_xmit_ipv4(skb, dev); 866 break; 867 case htons(ETH_P_IPV6): 868 ret = ip6gre_xmit_ipv6(skb, dev); 869 break; 870 default: 871 ret = ip6gre_xmit_other(skb, dev); 872 break; 873 } 874 875 if (ret < 0) 876 goto tx_err; 877 878 return NETDEV_TX_OK; 879 880 tx_err: 881 stats->tx_errors++; 882 stats->tx_dropped++; 883 kfree_skb(skb); 884 return NETDEV_TX_OK; 885 } 886 887 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 888 struct net_device *dev) 889 { 890 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 891 struct ip6_tnl *t = netdev_priv(dev); 892 struct dst_entry *dst = skb_dst(skb); 893 struct net_device_stats *stats; 894 bool truncate = false; 895 int encap_limit = -1; 896 __u8 dsfield = false; 897 struct flowi6 fl6; 898 int err = -EINVAL; 899 __u32 mtu; 900 901 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 902 goto tx_err; 903 904 if (gre_handle_offloads(skb, false)) 905 goto tx_err; 906 907 if (skb->len > dev->mtu + dev->hard_header_len) { 908 pskb_trim(skb, dev->mtu + dev->hard_header_len); 909 truncate = true; 910 } 911 912 t->parms.o_flags &= ~TUNNEL_KEY; 913 IPCB(skb)->flags = 0; 914 915 /* For collect_md mode, derive fl6 from the tunnel key, 916 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 917 */ 918 if (t->parms.collect_md) { 919 struct ip_tunnel_info *tun_info; 920 const struct ip_tunnel_key *key; 921 struct erspan_metadata *md; 922 __be32 tun_id; 923 924 tun_info = skb_tunnel_info(skb); 925 if (unlikely(!tun_info || 926 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 927 ip_tunnel_info_af(tun_info) != AF_INET6)) 928 return -EINVAL; 929 930 key = &tun_info->key; 931 memset(&fl6, 0, sizeof(fl6)); 932 fl6.flowi6_proto = IPPROTO_GRE; 933 fl6.daddr = key->u.ipv6.dst; 934 fl6.flowlabel = key->label; 935 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 936 937 dsfield = key->tos; 938 md = ip_tunnel_info_opts(tun_info); 939 if (!md) 940 goto tx_err; 941 942 tun_id = tunnel_id_to_key32(key->tun_id); 943 if (md->version == 1) { 944 erspan_build_header(skb, 945 ntohl(tun_id), 946 ntohl(md->u.index), truncate, 947 false); 948 } else if (md->version == 2) { 949 erspan_build_header_v2(skb, 950 ntohl(tun_id), 951 md->u.md2.dir, 952 get_hwid(&md->u.md2), 953 truncate, false); 954 } 955 } else { 956 switch (skb->protocol) { 957 case htons(ETH_P_IP): 958 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 959 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 960 &dsfield, &encap_limit); 961 break; 962 case htons(ETH_P_IPV6): 963 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 964 goto tx_err; 965 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 966 &dsfield, &encap_limit)) 967 goto tx_err; 968 break; 969 default: 970 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 971 break; 972 } 973 974 if (t->parms.erspan_ver == 1) 975 erspan_build_header(skb, ntohl(t->parms.o_key), 976 t->parms.index, 977 truncate, false); 978 else 979 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 980 t->parms.dir, 981 t->parms.hwid, 982 truncate, false); 983 fl6.daddr = t->parms.raddr; 984 } 985 986 /* Push GRE header. */ 987 gre_build_header(skb, 8, TUNNEL_SEQ, 988 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++)); 989 990 /* TooBig packet may have updated dst->dev's mtu */ 991 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 992 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu); 993 994 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 995 NEXTHDR_GRE); 996 if (err != 0) { 997 /* XXX: send ICMP error even if DF is not set. */ 998 if (err == -EMSGSIZE) { 999 if (skb->protocol == htons(ETH_P_IP)) 1000 icmp_send(skb, ICMP_DEST_UNREACH, 1001 ICMP_FRAG_NEEDED, htonl(mtu)); 1002 else 1003 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1004 } 1005 1006 goto tx_err; 1007 } 1008 return NETDEV_TX_OK; 1009 1010 tx_err: 1011 stats = &t->dev->stats; 1012 stats->tx_errors++; 1013 stats->tx_dropped++; 1014 kfree_skb(skb); 1015 return NETDEV_TX_OK; 1016 } 1017 1018 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1019 { 1020 struct net_device *dev = t->dev; 1021 struct __ip6_tnl_parm *p = &t->parms; 1022 struct flowi6 *fl6 = &t->fl.u.ip6; 1023 int t_hlen; 1024 1025 if (dev->type != ARPHRD_ETHER) { 1026 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1027 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1028 } 1029 1030 /* Set up flowi template */ 1031 fl6->saddr = p->laddr; 1032 fl6->daddr = p->raddr; 1033 fl6->flowi6_oif = p->link; 1034 fl6->flowlabel = 0; 1035 fl6->flowi6_proto = IPPROTO_GRE; 1036 1037 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1038 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1039 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1040 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1041 1042 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1043 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1044 1045 if (p->flags&IP6_TNL_F_CAP_XMIT && 1046 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1047 dev->flags |= IFF_POINTOPOINT; 1048 else 1049 dev->flags &= ~IFF_POINTOPOINT; 1050 1051 t->tun_hlen = gre_calc_hlen(t->parms.o_flags); 1052 1053 t->hlen = t->encap_hlen + t->tun_hlen; 1054 1055 t_hlen = t->hlen + sizeof(struct ipv6hdr); 1056 1057 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1058 int strict = (ipv6_addr_type(&p->raddr) & 1059 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1060 1061 struct rt6_info *rt = rt6_lookup(t->net, 1062 &p->raddr, &p->laddr, 1063 p->link, strict); 1064 1065 if (!rt) 1066 return; 1067 1068 if (rt->dst.dev) { 1069 dev->hard_header_len = rt->dst.dev->hard_header_len + 1070 t_hlen; 1071 1072 if (set_mtu) { 1073 dev->mtu = rt->dst.dev->mtu - t_hlen; 1074 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1075 dev->mtu -= 8; 1076 if (dev->type == ARPHRD_ETHER) 1077 dev->mtu -= ETH_HLEN; 1078 1079 if (dev->mtu < IPV6_MIN_MTU) 1080 dev->mtu = IPV6_MIN_MTU; 1081 } 1082 } 1083 ip6_rt_put(rt); 1084 } 1085 } 1086 1087 static int ip6gre_tnl_change(struct ip6_tnl *t, 1088 const struct __ip6_tnl_parm *p, int set_mtu) 1089 { 1090 t->parms.laddr = p->laddr; 1091 t->parms.raddr = p->raddr; 1092 t->parms.flags = p->flags; 1093 t->parms.hop_limit = p->hop_limit; 1094 t->parms.encap_limit = p->encap_limit; 1095 t->parms.flowinfo = p->flowinfo; 1096 t->parms.link = p->link; 1097 t->parms.proto = p->proto; 1098 t->parms.i_key = p->i_key; 1099 t->parms.o_key = p->o_key; 1100 t->parms.i_flags = p->i_flags; 1101 t->parms.o_flags = p->o_flags; 1102 t->parms.fwmark = p->fwmark; 1103 dst_cache_reset(&t->dst_cache); 1104 ip6gre_tnl_link_config(t, set_mtu); 1105 return 0; 1106 } 1107 1108 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1109 const struct ip6_tnl_parm2 *u) 1110 { 1111 p->laddr = u->laddr; 1112 p->raddr = u->raddr; 1113 p->flags = u->flags; 1114 p->hop_limit = u->hop_limit; 1115 p->encap_limit = u->encap_limit; 1116 p->flowinfo = u->flowinfo; 1117 p->link = u->link; 1118 p->i_key = u->i_key; 1119 p->o_key = u->o_key; 1120 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1121 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1122 memcpy(p->name, u->name, sizeof(u->name)); 1123 } 1124 1125 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1126 const struct __ip6_tnl_parm *p) 1127 { 1128 u->proto = IPPROTO_GRE; 1129 u->laddr = p->laddr; 1130 u->raddr = p->raddr; 1131 u->flags = p->flags; 1132 u->hop_limit = p->hop_limit; 1133 u->encap_limit = p->encap_limit; 1134 u->flowinfo = p->flowinfo; 1135 u->link = p->link; 1136 u->i_key = p->i_key; 1137 u->o_key = p->o_key; 1138 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1139 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1140 memcpy(u->name, p->name, sizeof(u->name)); 1141 } 1142 1143 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1144 struct ifreq *ifr, int cmd) 1145 { 1146 int err = 0; 1147 struct ip6_tnl_parm2 p; 1148 struct __ip6_tnl_parm p1; 1149 struct ip6_tnl *t = netdev_priv(dev); 1150 struct net *net = t->net; 1151 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1152 1153 memset(&p1, 0, sizeof(p1)); 1154 1155 switch (cmd) { 1156 case SIOCGETTUNNEL: 1157 if (dev == ign->fb_tunnel_dev) { 1158 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1159 err = -EFAULT; 1160 break; 1161 } 1162 ip6gre_tnl_parm_from_user(&p1, &p); 1163 t = ip6gre_tunnel_locate(net, &p1, 0); 1164 if (!t) 1165 t = netdev_priv(dev); 1166 } 1167 memset(&p, 0, sizeof(p)); 1168 ip6gre_tnl_parm_to_user(&p, &t->parms); 1169 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1170 err = -EFAULT; 1171 break; 1172 1173 case SIOCADDTUNNEL: 1174 case SIOCCHGTUNNEL: 1175 err = -EPERM; 1176 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1177 goto done; 1178 1179 err = -EFAULT; 1180 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1181 goto done; 1182 1183 err = -EINVAL; 1184 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1185 goto done; 1186 1187 if (!(p.i_flags&GRE_KEY)) 1188 p.i_key = 0; 1189 if (!(p.o_flags&GRE_KEY)) 1190 p.o_key = 0; 1191 1192 ip6gre_tnl_parm_from_user(&p1, &p); 1193 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1194 1195 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1196 if (t) { 1197 if (t->dev != dev) { 1198 err = -EEXIST; 1199 break; 1200 } 1201 } else { 1202 t = netdev_priv(dev); 1203 1204 ip6gre_tunnel_unlink(ign, t); 1205 synchronize_net(); 1206 ip6gre_tnl_change(t, &p1, 1); 1207 ip6gre_tunnel_link(ign, t); 1208 netdev_state_change(dev); 1209 } 1210 } 1211 1212 if (t) { 1213 err = 0; 1214 1215 memset(&p, 0, sizeof(p)); 1216 ip6gre_tnl_parm_to_user(&p, &t->parms); 1217 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1218 err = -EFAULT; 1219 } else 1220 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1221 break; 1222 1223 case SIOCDELTUNNEL: 1224 err = -EPERM; 1225 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1226 goto done; 1227 1228 if (dev == ign->fb_tunnel_dev) { 1229 err = -EFAULT; 1230 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1231 goto done; 1232 err = -ENOENT; 1233 ip6gre_tnl_parm_from_user(&p1, &p); 1234 t = ip6gre_tunnel_locate(net, &p1, 0); 1235 if (!t) 1236 goto done; 1237 err = -EPERM; 1238 if (t == netdev_priv(ign->fb_tunnel_dev)) 1239 goto done; 1240 dev = t->dev; 1241 } 1242 unregister_netdevice(dev); 1243 err = 0; 1244 break; 1245 1246 default: 1247 err = -EINVAL; 1248 } 1249 1250 done: 1251 return err; 1252 } 1253 1254 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1255 unsigned short type, const void *daddr, 1256 const void *saddr, unsigned int len) 1257 { 1258 struct ip6_tnl *t = netdev_priv(dev); 1259 struct ipv6hdr *ipv6h; 1260 __be16 *p; 1261 1262 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1263 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1264 t->fl.u.ip6.flowlabel, 1265 true, &t->fl.u.ip6)); 1266 ipv6h->hop_limit = t->parms.hop_limit; 1267 ipv6h->nexthdr = NEXTHDR_GRE; 1268 ipv6h->saddr = t->parms.laddr; 1269 ipv6h->daddr = t->parms.raddr; 1270 1271 p = (__be16 *)(ipv6h + 1); 1272 p[0] = t->parms.o_flags; 1273 p[1] = htons(type); 1274 1275 /* 1276 * Set the source hardware address. 1277 */ 1278 1279 if (saddr) 1280 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1281 if (daddr) 1282 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1283 if (!ipv6_addr_any(&ipv6h->daddr)) 1284 return t->hlen; 1285 1286 return -t->hlen; 1287 } 1288 1289 static const struct header_ops ip6gre_header_ops = { 1290 .create = ip6gre_header, 1291 }; 1292 1293 static const struct net_device_ops ip6gre_netdev_ops = { 1294 .ndo_init = ip6gre_tunnel_init, 1295 .ndo_uninit = ip6gre_tunnel_uninit, 1296 .ndo_start_xmit = ip6gre_tunnel_xmit, 1297 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1298 .ndo_change_mtu = ip6_tnl_change_mtu, 1299 .ndo_get_stats64 = ip_tunnel_get_stats64, 1300 .ndo_get_iflink = ip6_tnl_get_iflink, 1301 }; 1302 1303 static void ip6gre_dev_free(struct net_device *dev) 1304 { 1305 struct ip6_tnl *t = netdev_priv(dev); 1306 1307 dst_cache_destroy(&t->dst_cache); 1308 free_percpu(dev->tstats); 1309 } 1310 1311 static void ip6gre_tunnel_setup(struct net_device *dev) 1312 { 1313 dev->netdev_ops = &ip6gre_netdev_ops; 1314 dev->needs_free_netdev = true; 1315 dev->priv_destructor = ip6gre_dev_free; 1316 1317 dev->type = ARPHRD_IP6GRE; 1318 1319 dev->flags |= IFF_NOARP; 1320 dev->addr_len = sizeof(struct in6_addr); 1321 netif_keep_dst(dev); 1322 /* This perm addr will be used as interface identifier by IPv6 */ 1323 dev->addr_assign_type = NET_ADDR_RANDOM; 1324 eth_random_addr(dev->perm_addr); 1325 } 1326 1327 #define GRE6_FEATURES (NETIF_F_SG | \ 1328 NETIF_F_FRAGLIST | \ 1329 NETIF_F_HIGHDMA | \ 1330 NETIF_F_HW_CSUM) 1331 1332 static void ip6gre_tnl_init_features(struct net_device *dev) 1333 { 1334 struct ip6_tnl *nt = netdev_priv(dev); 1335 1336 dev->features |= GRE6_FEATURES; 1337 dev->hw_features |= GRE6_FEATURES; 1338 1339 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1340 /* TCP offload with GRE SEQ is not supported, nor 1341 * can we support 2 levels of outer headers requiring 1342 * an update. 1343 */ 1344 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1345 nt->encap.type == TUNNEL_ENCAP_NONE) { 1346 dev->features |= NETIF_F_GSO_SOFTWARE; 1347 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1348 } 1349 1350 /* Can use a lockless transmit, unless we generate 1351 * output sequences 1352 */ 1353 dev->features |= NETIF_F_LLTX; 1354 } 1355 } 1356 1357 static int ip6gre_tunnel_init_common(struct net_device *dev) 1358 { 1359 struct ip6_tnl *tunnel; 1360 int ret; 1361 int t_hlen; 1362 1363 tunnel = netdev_priv(dev); 1364 1365 tunnel->dev = dev; 1366 tunnel->net = dev_net(dev); 1367 strcpy(tunnel->parms.name, dev->name); 1368 1369 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1370 if (!dev->tstats) 1371 return -ENOMEM; 1372 1373 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1374 if (ret) { 1375 free_percpu(dev->tstats); 1376 dev->tstats = NULL; 1377 return ret; 1378 } 1379 1380 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1381 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1382 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1383 1384 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1385 dev->mtu = ETH_DATA_LEN - t_hlen; 1386 if (dev->type == ARPHRD_ETHER) 1387 dev->mtu -= ETH_HLEN; 1388 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1389 dev->mtu -= 8; 1390 1391 if (tunnel->parms.collect_md) { 1392 dev->features |= NETIF_F_NETNS_LOCAL; 1393 netif_keep_dst(dev); 1394 } 1395 ip6gre_tnl_init_features(dev); 1396 1397 return 0; 1398 } 1399 1400 static int ip6gre_tunnel_init(struct net_device *dev) 1401 { 1402 struct ip6_tnl *tunnel; 1403 int ret; 1404 1405 ret = ip6gre_tunnel_init_common(dev); 1406 if (ret) 1407 return ret; 1408 1409 tunnel = netdev_priv(dev); 1410 1411 if (tunnel->parms.collect_md) 1412 return 0; 1413 1414 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1415 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1416 1417 if (ipv6_addr_any(&tunnel->parms.raddr)) 1418 dev->header_ops = &ip6gre_header_ops; 1419 1420 return 0; 1421 } 1422 1423 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1424 { 1425 struct ip6_tnl *tunnel = netdev_priv(dev); 1426 1427 tunnel->dev = dev; 1428 tunnel->net = dev_net(dev); 1429 strcpy(tunnel->parms.name, dev->name); 1430 1431 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1432 1433 dev_hold(dev); 1434 } 1435 1436 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1437 .handler = gre_rcv, 1438 .err_handler = ip6gre_err, 1439 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1440 }; 1441 1442 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1443 { 1444 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1445 struct net_device *dev, *aux; 1446 int prio; 1447 1448 for_each_netdev_safe(net, dev, aux) 1449 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1450 dev->rtnl_link_ops == &ip6gre_tap_ops || 1451 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1452 unregister_netdevice_queue(dev, head); 1453 1454 for (prio = 0; prio < 4; prio++) { 1455 int h; 1456 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1457 struct ip6_tnl *t; 1458 1459 t = rtnl_dereference(ign->tunnels[prio][h]); 1460 1461 while (t) { 1462 /* If dev is in the same netns, it has already 1463 * been added to the list by the previous loop. 1464 */ 1465 if (!net_eq(dev_net(t->dev), net)) 1466 unregister_netdevice_queue(t->dev, 1467 head); 1468 t = rtnl_dereference(t->next); 1469 } 1470 } 1471 } 1472 } 1473 1474 static int __net_init ip6gre_init_net(struct net *net) 1475 { 1476 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1477 int err; 1478 1479 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1480 NET_NAME_UNKNOWN, 1481 ip6gre_tunnel_setup); 1482 if (!ign->fb_tunnel_dev) { 1483 err = -ENOMEM; 1484 goto err_alloc_dev; 1485 } 1486 dev_net_set(ign->fb_tunnel_dev, net); 1487 /* FB netdevice is special: we have one, and only one per netns. 1488 * Allowing to move it to another netns is clearly unsafe. 1489 */ 1490 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1491 1492 1493 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1494 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1495 1496 err = register_netdev(ign->fb_tunnel_dev); 1497 if (err) 1498 goto err_reg_dev; 1499 1500 rcu_assign_pointer(ign->tunnels_wc[0], 1501 netdev_priv(ign->fb_tunnel_dev)); 1502 return 0; 1503 1504 err_reg_dev: 1505 free_netdev(ign->fb_tunnel_dev); 1506 err_alloc_dev: 1507 return err; 1508 } 1509 1510 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1511 { 1512 struct net *net; 1513 LIST_HEAD(list); 1514 1515 rtnl_lock(); 1516 list_for_each_entry(net, net_list, exit_list) 1517 ip6gre_destroy_tunnels(net, &list); 1518 unregister_netdevice_many(&list); 1519 rtnl_unlock(); 1520 } 1521 1522 static struct pernet_operations ip6gre_net_ops = { 1523 .init = ip6gre_init_net, 1524 .exit_batch = ip6gre_exit_batch_net, 1525 .id = &ip6gre_net_id, 1526 .size = sizeof(struct ip6gre_net), 1527 }; 1528 1529 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1530 struct netlink_ext_ack *extack) 1531 { 1532 __be16 flags; 1533 1534 if (!data) 1535 return 0; 1536 1537 flags = 0; 1538 if (data[IFLA_GRE_IFLAGS]) 1539 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1540 if (data[IFLA_GRE_OFLAGS]) 1541 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1542 if (flags & (GRE_VERSION|GRE_ROUTING)) 1543 return -EINVAL; 1544 1545 return 0; 1546 } 1547 1548 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1549 struct netlink_ext_ack *extack) 1550 { 1551 struct in6_addr daddr; 1552 1553 if (tb[IFLA_ADDRESS]) { 1554 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1555 return -EINVAL; 1556 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1557 return -EADDRNOTAVAIL; 1558 } 1559 1560 if (!data) 1561 goto out; 1562 1563 if (data[IFLA_GRE_REMOTE]) { 1564 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1565 if (ipv6_addr_any(&daddr)) 1566 return -EINVAL; 1567 } 1568 1569 out: 1570 return ip6gre_tunnel_validate(tb, data, extack); 1571 } 1572 1573 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1574 struct netlink_ext_ack *extack) 1575 { 1576 __be16 flags = 0; 1577 int ret, ver = 0; 1578 1579 if (!data) 1580 return 0; 1581 1582 ret = ip6gre_tap_validate(tb, data, extack); 1583 if (ret) 1584 return ret; 1585 1586 /* ERSPAN should only have GRE sequence and key flag */ 1587 if (data[IFLA_GRE_OFLAGS]) 1588 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1589 if (data[IFLA_GRE_IFLAGS]) 1590 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1591 if (!data[IFLA_GRE_COLLECT_METADATA] && 1592 flags != (GRE_SEQ | GRE_KEY)) 1593 return -EINVAL; 1594 1595 /* ERSPAN Session ID only has 10-bit. Since we reuse 1596 * 32-bit key field as ID, check it's range. 1597 */ 1598 if (data[IFLA_GRE_IKEY] && 1599 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1600 return -EINVAL; 1601 1602 if (data[IFLA_GRE_OKEY] && 1603 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1604 return -EINVAL; 1605 1606 if (data[IFLA_GRE_ERSPAN_VER]) { 1607 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1608 if (ver != 1 && ver != 2) 1609 return -EINVAL; 1610 } 1611 1612 if (ver == 1) { 1613 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1614 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1615 1616 if (index & ~INDEX_MASK) 1617 return -EINVAL; 1618 } 1619 } else if (ver == 2) { 1620 if (data[IFLA_GRE_ERSPAN_DIR]) { 1621 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1622 1623 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1624 return -EINVAL; 1625 } 1626 1627 if (data[IFLA_GRE_ERSPAN_HWID]) { 1628 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1629 1630 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1631 return -EINVAL; 1632 } 1633 } 1634 1635 return 0; 1636 } 1637 1638 static void ip6gre_netlink_parms(struct nlattr *data[], 1639 struct __ip6_tnl_parm *parms) 1640 { 1641 memset(parms, 0, sizeof(*parms)); 1642 1643 if (!data) 1644 return; 1645 1646 if (data[IFLA_GRE_LINK]) 1647 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1648 1649 if (data[IFLA_GRE_IFLAGS]) 1650 parms->i_flags = gre_flags_to_tnl_flags( 1651 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1652 1653 if (data[IFLA_GRE_OFLAGS]) 1654 parms->o_flags = gre_flags_to_tnl_flags( 1655 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1656 1657 if (data[IFLA_GRE_IKEY]) 1658 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1659 1660 if (data[IFLA_GRE_OKEY]) 1661 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1662 1663 if (data[IFLA_GRE_LOCAL]) 1664 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1665 1666 if (data[IFLA_GRE_REMOTE]) 1667 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1668 1669 if (data[IFLA_GRE_TTL]) 1670 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1671 1672 if (data[IFLA_GRE_ENCAP_LIMIT]) 1673 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1674 1675 if (data[IFLA_GRE_FLOWINFO]) 1676 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1677 1678 if (data[IFLA_GRE_FLAGS]) 1679 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1680 1681 if (data[IFLA_GRE_FWMARK]) 1682 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1683 1684 if (data[IFLA_GRE_COLLECT_METADATA]) 1685 parms->collect_md = true; 1686 1687 if (data[IFLA_GRE_ERSPAN_VER]) 1688 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1689 1690 if (parms->erspan_ver == 1) { 1691 if (data[IFLA_GRE_ERSPAN_INDEX]) 1692 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1693 } else if (parms->erspan_ver == 2) { 1694 if (data[IFLA_GRE_ERSPAN_DIR]) 1695 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1696 if (data[IFLA_GRE_ERSPAN_HWID]) 1697 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1698 } 1699 } 1700 1701 static int ip6gre_tap_init(struct net_device *dev) 1702 { 1703 int ret; 1704 1705 ret = ip6gre_tunnel_init_common(dev); 1706 if (ret) 1707 return ret; 1708 1709 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1710 1711 return 0; 1712 } 1713 1714 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1715 .ndo_init = ip6gre_tap_init, 1716 .ndo_uninit = ip6gre_tunnel_uninit, 1717 .ndo_start_xmit = ip6gre_tunnel_xmit, 1718 .ndo_set_mac_address = eth_mac_addr, 1719 .ndo_validate_addr = eth_validate_addr, 1720 .ndo_change_mtu = ip6_tnl_change_mtu, 1721 .ndo_get_stats64 = ip_tunnel_get_stats64, 1722 .ndo_get_iflink = ip6_tnl_get_iflink, 1723 }; 1724 1725 static int ip6erspan_tap_init(struct net_device *dev) 1726 { 1727 struct ip6_tnl *tunnel; 1728 int t_hlen; 1729 int ret; 1730 1731 tunnel = netdev_priv(dev); 1732 1733 tunnel->dev = dev; 1734 tunnel->net = dev_net(dev); 1735 strcpy(tunnel->parms.name, dev->name); 1736 1737 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1738 if (!dev->tstats) 1739 return -ENOMEM; 1740 1741 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1742 if (ret) { 1743 free_percpu(dev->tstats); 1744 dev->tstats = NULL; 1745 return ret; 1746 } 1747 1748 tunnel->tun_hlen = 8; 1749 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1750 erspan_hdr_len(tunnel->parms.erspan_ver); 1751 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1752 1753 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1754 dev->mtu = ETH_DATA_LEN - t_hlen; 1755 if (dev->type == ARPHRD_ETHER) 1756 dev->mtu -= ETH_HLEN; 1757 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1758 dev->mtu -= 8; 1759 1760 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1761 tunnel = netdev_priv(dev); 1762 ip6gre_tnl_link_config(tunnel, 1); 1763 1764 return 0; 1765 } 1766 1767 static const struct net_device_ops ip6erspan_netdev_ops = { 1768 .ndo_init = ip6erspan_tap_init, 1769 .ndo_uninit = ip6gre_tunnel_uninit, 1770 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1771 .ndo_set_mac_address = eth_mac_addr, 1772 .ndo_validate_addr = eth_validate_addr, 1773 .ndo_change_mtu = ip6_tnl_change_mtu, 1774 .ndo_get_stats64 = ip_tunnel_get_stats64, 1775 .ndo_get_iflink = ip6_tnl_get_iflink, 1776 }; 1777 1778 static void ip6gre_tap_setup(struct net_device *dev) 1779 { 1780 1781 ether_setup(dev); 1782 1783 dev->max_mtu = 0; 1784 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1785 dev->needs_free_netdev = true; 1786 dev->priv_destructor = ip6gre_dev_free; 1787 1788 dev->features |= NETIF_F_NETNS_LOCAL; 1789 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1790 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1791 netif_keep_dst(dev); 1792 } 1793 1794 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1795 struct ip_tunnel_encap *ipencap) 1796 { 1797 bool ret = false; 1798 1799 memset(ipencap, 0, sizeof(*ipencap)); 1800 1801 if (!data) 1802 return ret; 1803 1804 if (data[IFLA_GRE_ENCAP_TYPE]) { 1805 ret = true; 1806 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1807 } 1808 1809 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1810 ret = true; 1811 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1812 } 1813 1814 if (data[IFLA_GRE_ENCAP_SPORT]) { 1815 ret = true; 1816 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1817 } 1818 1819 if (data[IFLA_GRE_ENCAP_DPORT]) { 1820 ret = true; 1821 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1822 } 1823 1824 return ret; 1825 } 1826 1827 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1828 struct nlattr *tb[], struct nlattr *data[], 1829 struct netlink_ext_ack *extack) 1830 { 1831 struct ip6_tnl *nt; 1832 struct net *net = dev_net(dev); 1833 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1834 struct ip_tunnel_encap ipencap; 1835 int err; 1836 1837 nt = netdev_priv(dev); 1838 1839 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1840 int err = ip6_tnl_encap_setup(nt, &ipencap); 1841 1842 if (err < 0) 1843 return err; 1844 } 1845 1846 ip6gre_netlink_parms(data, &nt->parms); 1847 1848 if (nt->parms.collect_md) { 1849 if (rtnl_dereference(ign->collect_md_tun)) 1850 return -EEXIST; 1851 } else { 1852 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1853 return -EEXIST; 1854 } 1855 1856 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1857 eth_hw_addr_random(dev); 1858 1859 nt->dev = dev; 1860 nt->net = dev_net(dev); 1861 1862 err = register_netdevice(dev); 1863 if (err) 1864 goto out; 1865 1866 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1867 1868 if (tb[IFLA_MTU]) 1869 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1870 1871 dev_hold(dev); 1872 ip6gre_tunnel_link(ign, nt); 1873 1874 out: 1875 return err; 1876 } 1877 1878 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 1879 struct nlattr *data[], 1880 struct netlink_ext_ack *extack) 1881 { 1882 struct ip6_tnl *t, *nt = netdev_priv(dev); 1883 struct net *net = nt->net; 1884 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1885 struct __ip6_tnl_parm p; 1886 struct ip_tunnel_encap ipencap; 1887 1888 if (dev == ign->fb_tunnel_dev) 1889 return -EINVAL; 1890 1891 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1892 int err = ip6_tnl_encap_setup(nt, &ipencap); 1893 1894 if (err < 0) 1895 return err; 1896 } 1897 1898 ip6gre_netlink_parms(data, &p); 1899 1900 t = ip6gre_tunnel_locate(net, &p, 0); 1901 1902 if (t) { 1903 if (t->dev != dev) 1904 return -EEXIST; 1905 } else { 1906 t = nt; 1907 } 1908 1909 ip6gre_tunnel_unlink(ign, t); 1910 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 1911 ip6gre_tunnel_link(ign, t); 1912 return 0; 1913 } 1914 1915 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 1916 { 1917 struct net *net = dev_net(dev); 1918 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1919 1920 if (dev != ign->fb_tunnel_dev) 1921 unregister_netdevice_queue(dev, head); 1922 } 1923 1924 static size_t ip6gre_get_size(const struct net_device *dev) 1925 { 1926 return 1927 /* IFLA_GRE_LINK */ 1928 nla_total_size(4) + 1929 /* IFLA_GRE_IFLAGS */ 1930 nla_total_size(2) + 1931 /* IFLA_GRE_OFLAGS */ 1932 nla_total_size(2) + 1933 /* IFLA_GRE_IKEY */ 1934 nla_total_size(4) + 1935 /* IFLA_GRE_OKEY */ 1936 nla_total_size(4) + 1937 /* IFLA_GRE_LOCAL */ 1938 nla_total_size(sizeof(struct in6_addr)) + 1939 /* IFLA_GRE_REMOTE */ 1940 nla_total_size(sizeof(struct in6_addr)) + 1941 /* IFLA_GRE_TTL */ 1942 nla_total_size(1) + 1943 /* IFLA_GRE_ENCAP_LIMIT */ 1944 nla_total_size(1) + 1945 /* IFLA_GRE_FLOWINFO */ 1946 nla_total_size(4) + 1947 /* IFLA_GRE_FLAGS */ 1948 nla_total_size(4) + 1949 /* IFLA_GRE_ENCAP_TYPE */ 1950 nla_total_size(2) + 1951 /* IFLA_GRE_ENCAP_FLAGS */ 1952 nla_total_size(2) + 1953 /* IFLA_GRE_ENCAP_SPORT */ 1954 nla_total_size(2) + 1955 /* IFLA_GRE_ENCAP_DPORT */ 1956 nla_total_size(2) + 1957 /* IFLA_GRE_COLLECT_METADATA */ 1958 nla_total_size(0) + 1959 /* IFLA_GRE_FWMARK */ 1960 nla_total_size(4) + 1961 /* IFLA_GRE_ERSPAN_INDEX */ 1962 nla_total_size(4) + 1963 0; 1964 } 1965 1966 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 1967 { 1968 struct ip6_tnl *t = netdev_priv(dev); 1969 struct __ip6_tnl_parm *p = &t->parms; 1970 1971 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 1972 nla_put_be16(skb, IFLA_GRE_IFLAGS, 1973 gre_tnl_flags_to_gre_flags(p->i_flags)) || 1974 nla_put_be16(skb, IFLA_GRE_OFLAGS, 1975 gre_tnl_flags_to_gre_flags(p->o_flags)) || 1976 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 1977 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 1978 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 1979 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 1980 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 1981 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 1982 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 1983 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 1984 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) || 1985 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 1986 goto nla_put_failure; 1987 1988 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 1989 t->encap.type) || 1990 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 1991 t->encap.sport) || 1992 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 1993 t->encap.dport) || 1994 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 1995 t->encap.flags)) 1996 goto nla_put_failure; 1997 1998 if (p->collect_md) { 1999 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2000 goto nla_put_failure; 2001 } 2002 2003 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2004 goto nla_put_failure; 2005 2006 if (p->erspan_ver == 1) { 2007 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2008 goto nla_put_failure; 2009 } else if (p->erspan_ver == 2) { 2010 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2011 goto nla_put_failure; 2012 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2013 goto nla_put_failure; 2014 } 2015 2016 return 0; 2017 2018 nla_put_failure: 2019 return -EMSGSIZE; 2020 } 2021 2022 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2023 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2024 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2025 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2026 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2027 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2028 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 2029 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 2030 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2031 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2032 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2033 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2034 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2035 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2036 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2037 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2038 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2039 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2040 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2041 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2042 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2043 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2044 }; 2045 2046 static void ip6erspan_tap_setup(struct net_device *dev) 2047 { 2048 ether_setup(dev); 2049 2050 dev->netdev_ops = &ip6erspan_netdev_ops; 2051 dev->needs_free_netdev = true; 2052 dev->priv_destructor = ip6gre_dev_free; 2053 2054 dev->features |= NETIF_F_NETNS_LOCAL; 2055 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2056 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2057 netif_keep_dst(dev); 2058 } 2059 2060 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2061 .kind = "ip6gre", 2062 .maxtype = IFLA_GRE_MAX, 2063 .policy = ip6gre_policy, 2064 .priv_size = sizeof(struct ip6_tnl), 2065 .setup = ip6gre_tunnel_setup, 2066 .validate = ip6gre_tunnel_validate, 2067 .newlink = ip6gre_newlink, 2068 .changelink = ip6gre_changelink, 2069 .dellink = ip6gre_dellink, 2070 .get_size = ip6gre_get_size, 2071 .fill_info = ip6gre_fill_info, 2072 .get_link_net = ip6_tnl_get_link_net, 2073 }; 2074 2075 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2076 .kind = "ip6gretap", 2077 .maxtype = IFLA_GRE_MAX, 2078 .policy = ip6gre_policy, 2079 .priv_size = sizeof(struct ip6_tnl), 2080 .setup = ip6gre_tap_setup, 2081 .validate = ip6gre_tap_validate, 2082 .newlink = ip6gre_newlink, 2083 .changelink = ip6gre_changelink, 2084 .get_size = ip6gre_get_size, 2085 .fill_info = ip6gre_fill_info, 2086 .get_link_net = ip6_tnl_get_link_net, 2087 }; 2088 2089 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2090 .kind = "ip6erspan", 2091 .maxtype = IFLA_GRE_MAX, 2092 .policy = ip6gre_policy, 2093 .priv_size = sizeof(struct ip6_tnl), 2094 .setup = ip6erspan_tap_setup, 2095 .validate = ip6erspan_tap_validate, 2096 .newlink = ip6gre_newlink, 2097 .changelink = ip6gre_changelink, 2098 .get_size = ip6gre_get_size, 2099 .fill_info = ip6gre_fill_info, 2100 .get_link_net = ip6_tnl_get_link_net, 2101 }; 2102 2103 /* 2104 * And now the modules code and kernel interface. 2105 */ 2106 2107 static int __init ip6gre_init(void) 2108 { 2109 int err; 2110 2111 pr_info("GRE over IPv6 tunneling driver\n"); 2112 2113 err = register_pernet_device(&ip6gre_net_ops); 2114 if (err < 0) 2115 return err; 2116 2117 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2118 if (err < 0) { 2119 pr_info("%s: can't add protocol\n", __func__); 2120 goto add_proto_failed; 2121 } 2122 2123 err = rtnl_link_register(&ip6gre_link_ops); 2124 if (err < 0) 2125 goto rtnl_link_failed; 2126 2127 err = rtnl_link_register(&ip6gre_tap_ops); 2128 if (err < 0) 2129 goto tap_ops_failed; 2130 2131 err = rtnl_link_register(&ip6erspan_tap_ops); 2132 if (err < 0) 2133 goto erspan_link_failed; 2134 2135 out: 2136 return err; 2137 2138 erspan_link_failed: 2139 rtnl_link_unregister(&ip6gre_tap_ops); 2140 tap_ops_failed: 2141 rtnl_link_unregister(&ip6gre_link_ops); 2142 rtnl_link_failed: 2143 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2144 add_proto_failed: 2145 unregister_pernet_device(&ip6gre_net_ops); 2146 goto out; 2147 } 2148 2149 static void __exit ip6gre_fini(void) 2150 { 2151 rtnl_link_unregister(&ip6gre_tap_ops); 2152 rtnl_link_unregister(&ip6gre_link_ops); 2153 rtnl_link_unregister(&ip6erspan_tap_ops); 2154 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2155 unregister_pernet_device(&ip6gre_net_ops); 2156 } 2157 2158 module_init(ip6gre_init); 2159 module_exit(ip6gre_fini); 2160 MODULE_LICENSE("GPL"); 2161 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2162 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2163 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2164 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2165 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2166 MODULE_ALIAS_NETDEV("ip6gre0"); 2167