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 void 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 gre_base_hdr *greh; 431 const struct ipv6hdr *ipv6h; 432 int grehlen = sizeof(*greh); 433 struct ip6_tnl *t; 434 int key_off = 0; 435 __be16 flags; 436 __be32 key; 437 438 if (!pskb_may_pull(skb, offset + grehlen)) 439 return; 440 greh = (const struct gre_base_hdr *)(skb->data + offset); 441 flags = greh->flags; 442 if (flags & (GRE_VERSION | GRE_ROUTING)) 443 return; 444 if (flags & GRE_CSUM) 445 grehlen += 4; 446 if (flags & GRE_KEY) { 447 key_off = grehlen + offset; 448 grehlen += 4; 449 } 450 451 if (!pskb_may_pull(skb, offset + grehlen)) 452 return; 453 ipv6h = (const struct ipv6hdr *)skb->data; 454 greh = (const struct gre_base_hdr *)(skb->data + offset); 455 key = key_off ? *(__be32 *)(skb->data + key_off) : 0; 456 457 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 458 key, greh->protocol); 459 if (!t) 460 return; 461 462 switch (type) { 463 struct ipv6_tlv_tnl_enc_lim *tel; 464 __u32 teli; 465 case ICMPV6_DEST_UNREACH: 466 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 467 t->parms.name); 468 if (code != ICMPV6_PORT_UNREACH) 469 break; 470 return; 471 case ICMPV6_TIME_EXCEED: 472 if (code == ICMPV6_EXC_HOPLIMIT) { 473 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 474 t->parms.name); 475 break; 476 } 477 return; 478 case ICMPV6_PARAMPROB: 479 teli = 0; 480 if (code == ICMPV6_HDR_FIELD) 481 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 482 483 if (teli && teli == be32_to_cpu(info) - 2) { 484 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 485 if (tel->encap_limit == 0) { 486 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 487 t->parms.name); 488 } 489 } else { 490 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 491 t->parms.name); 492 } 493 return; 494 case ICMPV6_PKT_TOOBIG: 495 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 496 return; 497 case NDISC_REDIRECT: 498 ip6_redirect(skb, net, skb->dev->ifindex, 0, 499 sock_net_uid(net, NULL)); 500 return; 501 } 502 503 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 504 t->err_count++; 505 else 506 t->err_count = 1; 507 t->err_time = jiffies; 508 } 509 510 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 511 { 512 const struct ipv6hdr *ipv6h; 513 struct ip6_tnl *tunnel; 514 515 ipv6h = ipv6_hdr(skb); 516 tunnel = ip6gre_tunnel_lookup(skb->dev, 517 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 518 tpi->proto); 519 if (tunnel) { 520 if (tunnel->parms.collect_md) { 521 struct metadata_dst *tun_dst; 522 __be64 tun_id; 523 __be16 flags; 524 525 flags = tpi->flags; 526 tun_id = key32_to_tunnel_id(tpi->key); 527 528 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 529 if (!tun_dst) 530 return PACKET_REJECT; 531 532 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 533 } else { 534 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 535 } 536 537 return PACKET_RCVD; 538 } 539 540 return PACKET_REJECT; 541 } 542 543 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len, 544 struct tnl_ptk_info *tpi) 545 { 546 struct erspan_base_hdr *ershdr; 547 struct erspan_metadata *pkt_md; 548 const struct ipv6hdr *ipv6h; 549 struct erspan_md2 *md2; 550 struct ip6_tnl *tunnel; 551 u8 ver; 552 553 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr)))) 554 return PACKET_REJECT; 555 556 ipv6h = ipv6_hdr(skb); 557 ershdr = (struct erspan_base_hdr *)skb->data; 558 ver = ershdr->ver; 559 tpi->key = cpu_to_be32(get_session_id(ershdr)); 560 561 tunnel = ip6gre_tunnel_lookup(skb->dev, 562 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 563 tpi->proto); 564 if (tunnel) { 565 int len = erspan_hdr_len(ver); 566 567 if (unlikely(!pskb_may_pull(skb, len))) 568 return PACKET_REJECT; 569 570 ershdr = (struct erspan_base_hdr *)skb->data; 571 pkt_md = (struct erspan_metadata *)(ershdr + 1); 572 573 if (__iptunnel_pull_header(skb, len, 574 htons(ETH_P_TEB), 575 false, false) < 0) 576 return PACKET_REJECT; 577 578 if (tunnel->parms.collect_md) { 579 struct metadata_dst *tun_dst; 580 struct ip_tunnel_info *info; 581 struct erspan_metadata *md; 582 __be64 tun_id; 583 __be16 flags; 584 585 tpi->flags |= TUNNEL_KEY; 586 flags = tpi->flags; 587 tun_id = key32_to_tunnel_id(tpi->key); 588 589 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 590 sizeof(*md)); 591 if (!tun_dst) 592 return PACKET_REJECT; 593 594 info = &tun_dst->u.tun_info; 595 md = ip_tunnel_info_opts(info); 596 md->version = ver; 597 md2 = &md->u.md2; 598 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 599 ERSPAN_V2_MDSIZE); 600 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 601 info->options_len = sizeof(*md); 602 603 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 604 605 } else { 606 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 607 } 608 609 return PACKET_RCVD; 610 } 611 612 return PACKET_REJECT; 613 } 614 615 static int gre_rcv(struct sk_buff *skb) 616 { 617 struct tnl_ptk_info tpi; 618 bool csum_err = false; 619 int hdr_len; 620 621 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 622 if (hdr_len < 0) 623 goto drop; 624 625 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 626 goto drop; 627 628 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 629 tpi.proto == htons(ETH_P_ERSPAN2))) { 630 if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD) 631 return 0; 632 goto out; 633 } 634 635 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 636 return 0; 637 638 out: 639 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 640 drop: 641 kfree_skb(skb); 642 return 0; 643 } 644 645 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 646 { 647 return iptunnel_handle_offloads(skb, 648 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 649 } 650 651 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 652 struct net_device *dev, 653 struct flowi6 *fl6, __u8 *dsfield, 654 int *encap_limit) 655 { 656 const struct iphdr *iph = ip_hdr(skb); 657 struct ip6_tnl *t = netdev_priv(dev); 658 659 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 660 *encap_limit = t->parms.encap_limit; 661 662 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 663 664 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 665 *dsfield = ipv4_get_dsfield(iph); 666 else 667 *dsfield = ip6_tclass(t->parms.flowinfo); 668 669 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 670 fl6->flowi6_mark = skb->mark; 671 else 672 fl6->flowi6_mark = t->parms.fwmark; 673 674 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 675 } 676 677 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 678 struct net_device *dev, 679 struct flowi6 *fl6, __u8 *dsfield, 680 int *encap_limit) 681 { 682 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 683 struct ip6_tnl *t = netdev_priv(dev); 684 __u16 offset; 685 686 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 687 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 688 689 if (offset > 0) { 690 struct ipv6_tlv_tnl_enc_lim *tel; 691 692 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 693 if (tel->encap_limit == 0) { 694 icmpv6_send(skb, ICMPV6_PARAMPROB, 695 ICMPV6_HDR_FIELD, offset + 2); 696 return -1; 697 } 698 *encap_limit = tel->encap_limit - 1; 699 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 700 *encap_limit = t->parms.encap_limit; 701 } 702 703 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 704 705 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 706 *dsfield = ipv6_get_dsfield(ipv6h); 707 else 708 *dsfield = ip6_tclass(t->parms.flowinfo); 709 710 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 711 fl6->flowlabel |= ip6_flowlabel(ipv6h); 712 713 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 714 fl6->flowi6_mark = skb->mark; 715 else 716 fl6->flowi6_mark = t->parms.fwmark; 717 718 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 719 720 return 0; 721 } 722 723 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 724 struct net_device *dev, __u8 dsfield, 725 struct flowi6 *fl6, int encap_limit, 726 __u32 *pmtu, __be16 proto) 727 { 728 struct ip6_tnl *tunnel = netdev_priv(dev); 729 __be16 protocol; 730 731 if (dev->type == ARPHRD_ETHER) 732 IPCB(skb)->flags = 0; 733 734 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 735 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 736 else 737 fl6->daddr = tunnel->parms.raddr; 738 739 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 740 return -ENOMEM; 741 742 /* Push GRE header. */ 743 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 744 745 if (tunnel->parms.collect_md) { 746 struct ip_tunnel_info *tun_info; 747 const struct ip_tunnel_key *key; 748 __be16 flags; 749 750 tun_info = skb_tunnel_info(skb); 751 if (unlikely(!tun_info || 752 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 753 ip_tunnel_info_af(tun_info) != AF_INET6)) 754 return -EINVAL; 755 756 key = &tun_info->key; 757 memset(fl6, 0, sizeof(*fl6)); 758 fl6->flowi6_proto = IPPROTO_GRE; 759 fl6->daddr = key->u.ipv6.dst; 760 fl6->flowlabel = key->label; 761 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 762 763 dsfield = key->tos; 764 flags = key->tun_flags & 765 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 766 tunnel->tun_hlen = gre_calc_hlen(flags); 767 768 gre_build_header(skb, tunnel->tun_hlen, 769 flags, protocol, 770 tunnel_id_to_key32(tun_info->key.tun_id), 771 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 772 : 0); 773 774 } else { 775 if (tunnel->parms.o_flags & TUNNEL_SEQ) 776 tunnel->o_seqno++; 777 778 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 779 protocol, tunnel->parms.o_key, 780 htonl(tunnel->o_seqno)); 781 } 782 783 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 784 NEXTHDR_GRE); 785 } 786 787 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 788 { 789 struct ip6_tnl *t = netdev_priv(dev); 790 int encap_limit = -1; 791 struct flowi6 fl6; 792 __u8 dsfield = 0; 793 __u32 mtu; 794 int err; 795 796 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 797 798 if (!t->parms.collect_md) 799 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 800 &dsfield, &encap_limit); 801 802 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 803 if (err) 804 return -1; 805 806 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 807 skb->protocol); 808 if (err != 0) { 809 /* XXX: send ICMP error even if DF is not set. */ 810 if (err == -EMSGSIZE) 811 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 812 htonl(mtu)); 813 return -1; 814 } 815 816 return 0; 817 } 818 819 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 820 { 821 struct ip6_tnl *t = netdev_priv(dev); 822 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 823 int encap_limit = -1; 824 struct flowi6 fl6; 825 __u8 dsfield = 0; 826 __u32 mtu; 827 int err; 828 829 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 830 return -1; 831 832 if (!t->parms.collect_md && 833 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 834 return -1; 835 836 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 837 return -1; 838 839 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 840 &mtu, skb->protocol); 841 if (err != 0) { 842 if (err == -EMSGSIZE) 843 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 844 return -1; 845 } 846 847 return 0; 848 } 849 850 /** 851 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 852 * @t: the outgoing tunnel device 853 * @hdr: IPv6 header from the incoming packet 854 * 855 * Description: 856 * Avoid trivial tunneling loop by checking that tunnel exit-point 857 * doesn't match source of incoming packet. 858 * 859 * Return: 860 * 1 if conflict, 861 * 0 else 862 **/ 863 864 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 865 const struct ipv6hdr *hdr) 866 { 867 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 868 } 869 870 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 871 { 872 struct ip6_tnl *t = netdev_priv(dev); 873 int encap_limit = -1; 874 struct flowi6 fl6; 875 __u32 mtu; 876 int err; 877 878 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 879 encap_limit = t->parms.encap_limit; 880 881 if (!t->parms.collect_md) 882 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 883 884 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 885 if (err) 886 return err; 887 888 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 889 890 return err; 891 } 892 893 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 894 struct net_device *dev) 895 { 896 struct ip6_tnl *t = netdev_priv(dev); 897 struct net_device_stats *stats = &t->dev->stats; 898 int ret; 899 900 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 901 goto tx_err; 902 903 switch (skb->protocol) { 904 case htons(ETH_P_IP): 905 ret = ip6gre_xmit_ipv4(skb, dev); 906 break; 907 case htons(ETH_P_IPV6): 908 ret = ip6gre_xmit_ipv6(skb, dev); 909 break; 910 default: 911 ret = ip6gre_xmit_other(skb, dev); 912 break; 913 } 914 915 if (ret < 0) 916 goto tx_err; 917 918 return NETDEV_TX_OK; 919 920 tx_err: 921 stats->tx_errors++; 922 stats->tx_dropped++; 923 kfree_skb(skb); 924 return NETDEV_TX_OK; 925 } 926 927 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 928 struct net_device *dev) 929 { 930 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 931 struct ip6_tnl *t = netdev_priv(dev); 932 struct dst_entry *dst = skb_dst(skb); 933 struct net_device_stats *stats; 934 bool truncate = false; 935 int encap_limit = -1; 936 __u8 dsfield = false; 937 struct flowi6 fl6; 938 int err = -EINVAL; 939 __u32 mtu; 940 int nhoff; 941 int thoff; 942 943 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 944 goto tx_err; 945 946 if (gre_handle_offloads(skb, false)) 947 goto tx_err; 948 949 if (skb->len > dev->mtu + dev->hard_header_len) { 950 pskb_trim(skb, dev->mtu + dev->hard_header_len); 951 truncate = true; 952 } 953 954 nhoff = skb_network_header(skb) - skb_mac_header(skb); 955 if (skb->protocol == htons(ETH_P_IP) && 956 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 957 truncate = true; 958 959 thoff = skb_transport_header(skb) - skb_mac_header(skb); 960 if (skb->protocol == htons(ETH_P_IPV6) && 961 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 962 truncate = true; 963 964 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 965 goto tx_err; 966 967 t->parms.o_flags &= ~TUNNEL_KEY; 968 IPCB(skb)->flags = 0; 969 970 /* For collect_md mode, derive fl6 from the tunnel key, 971 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 972 */ 973 if (t->parms.collect_md) { 974 struct ip_tunnel_info *tun_info; 975 const struct ip_tunnel_key *key; 976 struct erspan_metadata *md; 977 __be32 tun_id; 978 979 tun_info = skb_tunnel_info(skb); 980 if (unlikely(!tun_info || 981 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 982 ip_tunnel_info_af(tun_info) != AF_INET6)) 983 return -EINVAL; 984 985 key = &tun_info->key; 986 memset(&fl6, 0, sizeof(fl6)); 987 fl6.flowi6_proto = IPPROTO_GRE; 988 fl6.daddr = key->u.ipv6.dst; 989 fl6.flowlabel = key->label; 990 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 991 992 dsfield = key->tos; 993 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 994 goto tx_err; 995 md = ip_tunnel_info_opts(tun_info); 996 if (!md) 997 goto tx_err; 998 999 tun_id = tunnel_id_to_key32(key->tun_id); 1000 if (md->version == 1) { 1001 erspan_build_header(skb, 1002 ntohl(tun_id), 1003 ntohl(md->u.index), truncate, 1004 false); 1005 } else if (md->version == 2) { 1006 erspan_build_header_v2(skb, 1007 ntohl(tun_id), 1008 md->u.md2.dir, 1009 get_hwid(&md->u.md2), 1010 truncate, false); 1011 } else { 1012 goto tx_err; 1013 } 1014 } else { 1015 switch (skb->protocol) { 1016 case htons(ETH_P_IP): 1017 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1018 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1019 &dsfield, &encap_limit); 1020 break; 1021 case htons(ETH_P_IPV6): 1022 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 1023 goto tx_err; 1024 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1025 &dsfield, &encap_limit)) 1026 goto tx_err; 1027 break; 1028 default: 1029 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1030 break; 1031 } 1032 1033 if (t->parms.erspan_ver == 1) 1034 erspan_build_header(skb, ntohl(t->parms.o_key), 1035 t->parms.index, 1036 truncate, false); 1037 else if (t->parms.erspan_ver == 2) 1038 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1039 t->parms.dir, 1040 t->parms.hwid, 1041 truncate, false); 1042 else 1043 goto tx_err; 1044 1045 fl6.daddr = t->parms.raddr; 1046 } 1047 1048 /* Push GRE header. */ 1049 gre_build_header(skb, 8, TUNNEL_SEQ, 1050 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++)); 1051 1052 /* TooBig packet may have updated dst->dev's mtu */ 1053 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1054 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu); 1055 1056 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1057 NEXTHDR_GRE); 1058 if (err != 0) { 1059 /* XXX: send ICMP error even if DF is not set. */ 1060 if (err == -EMSGSIZE) { 1061 if (skb->protocol == htons(ETH_P_IP)) 1062 icmp_send(skb, ICMP_DEST_UNREACH, 1063 ICMP_FRAG_NEEDED, htonl(mtu)); 1064 else 1065 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1066 } 1067 1068 goto tx_err; 1069 } 1070 return NETDEV_TX_OK; 1071 1072 tx_err: 1073 stats = &t->dev->stats; 1074 stats->tx_errors++; 1075 stats->tx_dropped++; 1076 kfree_skb(skb); 1077 return NETDEV_TX_OK; 1078 } 1079 1080 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1081 { 1082 struct net_device *dev = t->dev; 1083 struct __ip6_tnl_parm *p = &t->parms; 1084 struct flowi6 *fl6 = &t->fl.u.ip6; 1085 1086 if (dev->type != ARPHRD_ETHER) { 1087 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1088 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1089 } 1090 1091 /* Set up flowi template */ 1092 fl6->saddr = p->laddr; 1093 fl6->daddr = p->raddr; 1094 fl6->flowi6_oif = p->link; 1095 fl6->flowlabel = 0; 1096 fl6->flowi6_proto = IPPROTO_GRE; 1097 1098 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1099 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1100 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1101 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1102 1103 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1104 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1105 1106 if (p->flags&IP6_TNL_F_CAP_XMIT && 1107 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1108 dev->flags |= IFF_POINTOPOINT; 1109 else 1110 dev->flags &= ~IFF_POINTOPOINT; 1111 } 1112 1113 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1114 int t_hlen) 1115 { 1116 const struct __ip6_tnl_parm *p = &t->parms; 1117 struct net_device *dev = t->dev; 1118 1119 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1120 int strict = (ipv6_addr_type(&p->raddr) & 1121 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1122 1123 struct rt6_info *rt = rt6_lookup(t->net, 1124 &p->raddr, &p->laddr, 1125 p->link, NULL, strict); 1126 1127 if (!rt) 1128 return; 1129 1130 if (rt->dst.dev) { 1131 dev->hard_header_len = rt->dst.dev->hard_header_len + 1132 t_hlen; 1133 1134 if (set_mtu) { 1135 dev->mtu = rt->dst.dev->mtu - t_hlen; 1136 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1137 dev->mtu -= 8; 1138 if (dev->type == ARPHRD_ETHER) 1139 dev->mtu -= ETH_HLEN; 1140 1141 if (dev->mtu < IPV6_MIN_MTU) 1142 dev->mtu = IPV6_MIN_MTU; 1143 } 1144 } 1145 ip6_rt_put(rt); 1146 } 1147 } 1148 1149 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1150 { 1151 int t_hlen; 1152 1153 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1154 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1155 1156 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1157 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1158 return t_hlen; 1159 } 1160 1161 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1162 { 1163 ip6gre_tnl_link_config_common(t); 1164 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1165 } 1166 1167 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1168 const struct __ip6_tnl_parm *p) 1169 { 1170 t->parms.laddr = p->laddr; 1171 t->parms.raddr = p->raddr; 1172 t->parms.flags = p->flags; 1173 t->parms.hop_limit = p->hop_limit; 1174 t->parms.encap_limit = p->encap_limit; 1175 t->parms.flowinfo = p->flowinfo; 1176 t->parms.link = p->link; 1177 t->parms.proto = p->proto; 1178 t->parms.i_key = p->i_key; 1179 t->parms.o_key = p->o_key; 1180 t->parms.i_flags = p->i_flags; 1181 t->parms.o_flags = p->o_flags; 1182 t->parms.fwmark = p->fwmark; 1183 dst_cache_reset(&t->dst_cache); 1184 } 1185 1186 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1187 int set_mtu) 1188 { 1189 ip6gre_tnl_copy_tnl_parm(t, p); 1190 ip6gre_tnl_link_config(t, set_mtu); 1191 return 0; 1192 } 1193 1194 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1195 const struct ip6_tnl_parm2 *u) 1196 { 1197 p->laddr = u->laddr; 1198 p->raddr = u->raddr; 1199 p->flags = u->flags; 1200 p->hop_limit = u->hop_limit; 1201 p->encap_limit = u->encap_limit; 1202 p->flowinfo = u->flowinfo; 1203 p->link = u->link; 1204 p->i_key = u->i_key; 1205 p->o_key = u->o_key; 1206 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1207 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1208 memcpy(p->name, u->name, sizeof(u->name)); 1209 } 1210 1211 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1212 const struct __ip6_tnl_parm *p) 1213 { 1214 u->proto = IPPROTO_GRE; 1215 u->laddr = p->laddr; 1216 u->raddr = p->raddr; 1217 u->flags = p->flags; 1218 u->hop_limit = p->hop_limit; 1219 u->encap_limit = p->encap_limit; 1220 u->flowinfo = p->flowinfo; 1221 u->link = p->link; 1222 u->i_key = p->i_key; 1223 u->o_key = p->o_key; 1224 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1225 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1226 memcpy(u->name, p->name, sizeof(u->name)); 1227 } 1228 1229 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1230 struct ifreq *ifr, int cmd) 1231 { 1232 int err = 0; 1233 struct ip6_tnl_parm2 p; 1234 struct __ip6_tnl_parm p1; 1235 struct ip6_tnl *t = netdev_priv(dev); 1236 struct net *net = t->net; 1237 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1238 1239 memset(&p1, 0, sizeof(p1)); 1240 1241 switch (cmd) { 1242 case SIOCGETTUNNEL: 1243 if (dev == ign->fb_tunnel_dev) { 1244 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1245 err = -EFAULT; 1246 break; 1247 } 1248 ip6gre_tnl_parm_from_user(&p1, &p); 1249 t = ip6gre_tunnel_locate(net, &p1, 0); 1250 if (!t) 1251 t = netdev_priv(dev); 1252 } 1253 memset(&p, 0, sizeof(p)); 1254 ip6gre_tnl_parm_to_user(&p, &t->parms); 1255 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1256 err = -EFAULT; 1257 break; 1258 1259 case SIOCADDTUNNEL: 1260 case SIOCCHGTUNNEL: 1261 err = -EPERM; 1262 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1263 goto done; 1264 1265 err = -EFAULT; 1266 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1267 goto done; 1268 1269 err = -EINVAL; 1270 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1271 goto done; 1272 1273 if (!(p.i_flags&GRE_KEY)) 1274 p.i_key = 0; 1275 if (!(p.o_flags&GRE_KEY)) 1276 p.o_key = 0; 1277 1278 ip6gre_tnl_parm_from_user(&p1, &p); 1279 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1280 1281 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1282 if (t) { 1283 if (t->dev != dev) { 1284 err = -EEXIST; 1285 break; 1286 } 1287 } else { 1288 t = netdev_priv(dev); 1289 1290 ip6gre_tunnel_unlink(ign, t); 1291 synchronize_net(); 1292 ip6gre_tnl_change(t, &p1, 1); 1293 ip6gre_tunnel_link(ign, t); 1294 netdev_state_change(dev); 1295 } 1296 } 1297 1298 if (t) { 1299 err = 0; 1300 1301 memset(&p, 0, sizeof(p)); 1302 ip6gre_tnl_parm_to_user(&p, &t->parms); 1303 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1304 err = -EFAULT; 1305 } else 1306 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1307 break; 1308 1309 case SIOCDELTUNNEL: 1310 err = -EPERM; 1311 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1312 goto done; 1313 1314 if (dev == ign->fb_tunnel_dev) { 1315 err = -EFAULT; 1316 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1317 goto done; 1318 err = -ENOENT; 1319 ip6gre_tnl_parm_from_user(&p1, &p); 1320 t = ip6gre_tunnel_locate(net, &p1, 0); 1321 if (!t) 1322 goto done; 1323 err = -EPERM; 1324 if (t == netdev_priv(ign->fb_tunnel_dev)) 1325 goto done; 1326 dev = t->dev; 1327 } 1328 unregister_netdevice(dev); 1329 err = 0; 1330 break; 1331 1332 default: 1333 err = -EINVAL; 1334 } 1335 1336 done: 1337 return err; 1338 } 1339 1340 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1341 unsigned short type, const void *daddr, 1342 const void *saddr, unsigned int len) 1343 { 1344 struct ip6_tnl *t = netdev_priv(dev); 1345 struct ipv6hdr *ipv6h; 1346 __be16 *p; 1347 1348 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1349 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1350 t->fl.u.ip6.flowlabel, 1351 true, &t->fl.u.ip6)); 1352 ipv6h->hop_limit = t->parms.hop_limit; 1353 ipv6h->nexthdr = NEXTHDR_GRE; 1354 ipv6h->saddr = t->parms.laddr; 1355 ipv6h->daddr = t->parms.raddr; 1356 1357 p = (__be16 *)(ipv6h + 1); 1358 p[0] = t->parms.o_flags; 1359 p[1] = htons(type); 1360 1361 /* 1362 * Set the source hardware address. 1363 */ 1364 1365 if (saddr) 1366 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1367 if (daddr) 1368 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1369 if (!ipv6_addr_any(&ipv6h->daddr)) 1370 return t->hlen; 1371 1372 return -t->hlen; 1373 } 1374 1375 static const struct header_ops ip6gre_header_ops = { 1376 .create = ip6gre_header, 1377 }; 1378 1379 static const struct net_device_ops ip6gre_netdev_ops = { 1380 .ndo_init = ip6gre_tunnel_init, 1381 .ndo_uninit = ip6gre_tunnel_uninit, 1382 .ndo_start_xmit = ip6gre_tunnel_xmit, 1383 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1384 .ndo_change_mtu = ip6_tnl_change_mtu, 1385 .ndo_get_stats64 = ip_tunnel_get_stats64, 1386 .ndo_get_iflink = ip6_tnl_get_iflink, 1387 }; 1388 1389 static void ip6gre_dev_free(struct net_device *dev) 1390 { 1391 struct ip6_tnl *t = netdev_priv(dev); 1392 1393 gro_cells_destroy(&t->gro_cells); 1394 dst_cache_destroy(&t->dst_cache); 1395 free_percpu(dev->tstats); 1396 } 1397 1398 static void ip6gre_tunnel_setup(struct net_device *dev) 1399 { 1400 dev->netdev_ops = &ip6gre_netdev_ops; 1401 dev->needs_free_netdev = true; 1402 dev->priv_destructor = ip6gre_dev_free; 1403 1404 dev->type = ARPHRD_IP6GRE; 1405 1406 dev->flags |= IFF_NOARP; 1407 dev->addr_len = sizeof(struct in6_addr); 1408 netif_keep_dst(dev); 1409 /* This perm addr will be used as interface identifier by IPv6 */ 1410 dev->addr_assign_type = NET_ADDR_RANDOM; 1411 eth_random_addr(dev->perm_addr); 1412 } 1413 1414 #define GRE6_FEATURES (NETIF_F_SG | \ 1415 NETIF_F_FRAGLIST | \ 1416 NETIF_F_HIGHDMA | \ 1417 NETIF_F_HW_CSUM) 1418 1419 static void ip6gre_tnl_init_features(struct net_device *dev) 1420 { 1421 struct ip6_tnl *nt = netdev_priv(dev); 1422 1423 dev->features |= GRE6_FEATURES; 1424 dev->hw_features |= GRE6_FEATURES; 1425 1426 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1427 /* TCP offload with GRE SEQ is not supported, nor 1428 * can we support 2 levels of outer headers requiring 1429 * an update. 1430 */ 1431 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1432 nt->encap.type == TUNNEL_ENCAP_NONE) { 1433 dev->features |= NETIF_F_GSO_SOFTWARE; 1434 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1435 } 1436 1437 /* Can use a lockless transmit, unless we generate 1438 * output sequences 1439 */ 1440 dev->features |= NETIF_F_LLTX; 1441 } 1442 } 1443 1444 static int ip6gre_tunnel_init_common(struct net_device *dev) 1445 { 1446 struct ip6_tnl *tunnel; 1447 int ret; 1448 int t_hlen; 1449 1450 tunnel = netdev_priv(dev); 1451 1452 tunnel->dev = dev; 1453 tunnel->net = dev_net(dev); 1454 strcpy(tunnel->parms.name, dev->name); 1455 1456 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1457 if (!dev->tstats) 1458 return -ENOMEM; 1459 1460 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1461 if (ret) 1462 goto cleanup_alloc_pcpu_stats; 1463 1464 ret = gro_cells_init(&tunnel->gro_cells, dev); 1465 if (ret) 1466 goto cleanup_dst_cache_init; 1467 1468 t_hlen = ip6gre_calc_hlen(tunnel); 1469 dev->mtu = ETH_DATA_LEN - t_hlen; 1470 if (dev->type == ARPHRD_ETHER) 1471 dev->mtu -= ETH_HLEN; 1472 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1473 dev->mtu -= 8; 1474 1475 if (tunnel->parms.collect_md) { 1476 dev->features |= NETIF_F_NETNS_LOCAL; 1477 netif_keep_dst(dev); 1478 } 1479 ip6gre_tnl_init_features(dev); 1480 1481 return 0; 1482 1483 cleanup_dst_cache_init: 1484 dst_cache_destroy(&tunnel->dst_cache); 1485 cleanup_alloc_pcpu_stats: 1486 free_percpu(dev->tstats); 1487 dev->tstats = NULL; 1488 return ret; 1489 } 1490 1491 static int ip6gre_tunnel_init(struct net_device *dev) 1492 { 1493 struct ip6_tnl *tunnel; 1494 int ret; 1495 1496 ret = ip6gre_tunnel_init_common(dev); 1497 if (ret) 1498 return ret; 1499 1500 tunnel = netdev_priv(dev); 1501 1502 if (tunnel->parms.collect_md) 1503 return 0; 1504 1505 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1506 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1507 1508 if (ipv6_addr_any(&tunnel->parms.raddr)) 1509 dev->header_ops = &ip6gre_header_ops; 1510 1511 return 0; 1512 } 1513 1514 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1515 { 1516 struct ip6_tnl *tunnel = netdev_priv(dev); 1517 1518 tunnel->dev = dev; 1519 tunnel->net = dev_net(dev); 1520 strcpy(tunnel->parms.name, dev->name); 1521 1522 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1523 1524 dev_hold(dev); 1525 } 1526 1527 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1528 .handler = gre_rcv, 1529 .err_handler = ip6gre_err, 1530 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1531 }; 1532 1533 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1534 { 1535 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1536 struct net_device *dev, *aux; 1537 int prio; 1538 1539 for_each_netdev_safe(net, dev, aux) 1540 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1541 dev->rtnl_link_ops == &ip6gre_tap_ops || 1542 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1543 unregister_netdevice_queue(dev, head); 1544 1545 for (prio = 0; prio < 4; prio++) { 1546 int h; 1547 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1548 struct ip6_tnl *t; 1549 1550 t = rtnl_dereference(ign->tunnels[prio][h]); 1551 1552 while (t) { 1553 /* If dev is in the same netns, it has already 1554 * been added to the list by the previous loop. 1555 */ 1556 if (!net_eq(dev_net(t->dev), net)) 1557 unregister_netdevice_queue(t->dev, 1558 head); 1559 t = rtnl_dereference(t->next); 1560 } 1561 } 1562 } 1563 } 1564 1565 static int __net_init ip6gre_init_net(struct net *net) 1566 { 1567 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1568 int err; 1569 1570 if (!net_has_fallback_tunnels(net)) 1571 return 0; 1572 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1573 NET_NAME_UNKNOWN, 1574 ip6gre_tunnel_setup); 1575 if (!ign->fb_tunnel_dev) { 1576 err = -ENOMEM; 1577 goto err_alloc_dev; 1578 } 1579 dev_net_set(ign->fb_tunnel_dev, net); 1580 /* FB netdevice is special: we have one, and only one per netns. 1581 * Allowing to move it to another netns is clearly unsafe. 1582 */ 1583 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1584 1585 1586 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1587 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1588 1589 err = register_netdev(ign->fb_tunnel_dev); 1590 if (err) 1591 goto err_reg_dev; 1592 1593 rcu_assign_pointer(ign->tunnels_wc[0], 1594 netdev_priv(ign->fb_tunnel_dev)); 1595 return 0; 1596 1597 err_reg_dev: 1598 free_netdev(ign->fb_tunnel_dev); 1599 err_alloc_dev: 1600 return err; 1601 } 1602 1603 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1604 { 1605 struct net *net; 1606 LIST_HEAD(list); 1607 1608 rtnl_lock(); 1609 list_for_each_entry(net, net_list, exit_list) 1610 ip6gre_destroy_tunnels(net, &list); 1611 unregister_netdevice_many(&list); 1612 rtnl_unlock(); 1613 } 1614 1615 static struct pernet_operations ip6gre_net_ops = { 1616 .init = ip6gre_init_net, 1617 .exit_batch = ip6gre_exit_batch_net, 1618 .id = &ip6gre_net_id, 1619 .size = sizeof(struct ip6gre_net), 1620 }; 1621 1622 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1623 struct netlink_ext_ack *extack) 1624 { 1625 __be16 flags; 1626 1627 if (!data) 1628 return 0; 1629 1630 flags = 0; 1631 if (data[IFLA_GRE_IFLAGS]) 1632 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1633 if (data[IFLA_GRE_OFLAGS]) 1634 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1635 if (flags & (GRE_VERSION|GRE_ROUTING)) 1636 return -EINVAL; 1637 1638 return 0; 1639 } 1640 1641 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1642 struct netlink_ext_ack *extack) 1643 { 1644 struct in6_addr daddr; 1645 1646 if (tb[IFLA_ADDRESS]) { 1647 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1648 return -EINVAL; 1649 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1650 return -EADDRNOTAVAIL; 1651 } 1652 1653 if (!data) 1654 goto out; 1655 1656 if (data[IFLA_GRE_REMOTE]) { 1657 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1658 if (ipv6_addr_any(&daddr)) 1659 return -EINVAL; 1660 } 1661 1662 out: 1663 return ip6gre_tunnel_validate(tb, data, extack); 1664 } 1665 1666 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1667 struct netlink_ext_ack *extack) 1668 { 1669 __be16 flags = 0; 1670 int ret, ver = 0; 1671 1672 if (!data) 1673 return 0; 1674 1675 ret = ip6gre_tap_validate(tb, data, extack); 1676 if (ret) 1677 return ret; 1678 1679 /* ERSPAN should only have GRE sequence and key flag */ 1680 if (data[IFLA_GRE_OFLAGS]) 1681 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1682 if (data[IFLA_GRE_IFLAGS]) 1683 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1684 if (!data[IFLA_GRE_COLLECT_METADATA] && 1685 flags != (GRE_SEQ | GRE_KEY)) 1686 return -EINVAL; 1687 1688 /* ERSPAN Session ID only has 10-bit. Since we reuse 1689 * 32-bit key field as ID, check it's range. 1690 */ 1691 if (data[IFLA_GRE_IKEY] && 1692 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1693 return -EINVAL; 1694 1695 if (data[IFLA_GRE_OKEY] && 1696 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1697 return -EINVAL; 1698 1699 if (data[IFLA_GRE_ERSPAN_VER]) { 1700 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1701 if (ver != 1 && ver != 2) 1702 return -EINVAL; 1703 } 1704 1705 if (ver == 1) { 1706 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1707 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1708 1709 if (index & ~INDEX_MASK) 1710 return -EINVAL; 1711 } 1712 } else if (ver == 2) { 1713 if (data[IFLA_GRE_ERSPAN_DIR]) { 1714 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1715 1716 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1717 return -EINVAL; 1718 } 1719 1720 if (data[IFLA_GRE_ERSPAN_HWID]) { 1721 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1722 1723 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1724 return -EINVAL; 1725 } 1726 } 1727 1728 return 0; 1729 } 1730 1731 static void ip6gre_netlink_parms(struct nlattr *data[], 1732 struct __ip6_tnl_parm *parms) 1733 { 1734 memset(parms, 0, sizeof(*parms)); 1735 1736 if (!data) 1737 return; 1738 1739 if (data[IFLA_GRE_LINK]) 1740 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1741 1742 if (data[IFLA_GRE_IFLAGS]) 1743 parms->i_flags = gre_flags_to_tnl_flags( 1744 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1745 1746 if (data[IFLA_GRE_OFLAGS]) 1747 parms->o_flags = gre_flags_to_tnl_flags( 1748 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1749 1750 if (data[IFLA_GRE_IKEY]) 1751 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1752 1753 if (data[IFLA_GRE_OKEY]) 1754 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1755 1756 if (data[IFLA_GRE_LOCAL]) 1757 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1758 1759 if (data[IFLA_GRE_REMOTE]) 1760 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1761 1762 if (data[IFLA_GRE_TTL]) 1763 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1764 1765 if (data[IFLA_GRE_ENCAP_LIMIT]) 1766 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1767 1768 if (data[IFLA_GRE_FLOWINFO]) 1769 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1770 1771 if (data[IFLA_GRE_FLAGS]) 1772 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1773 1774 if (data[IFLA_GRE_FWMARK]) 1775 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1776 1777 if (data[IFLA_GRE_COLLECT_METADATA]) 1778 parms->collect_md = true; 1779 1780 if (data[IFLA_GRE_ERSPAN_VER]) 1781 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1782 1783 if (parms->erspan_ver == 1) { 1784 if (data[IFLA_GRE_ERSPAN_INDEX]) 1785 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1786 } else if (parms->erspan_ver == 2) { 1787 if (data[IFLA_GRE_ERSPAN_DIR]) 1788 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1789 if (data[IFLA_GRE_ERSPAN_HWID]) 1790 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1791 } 1792 } 1793 1794 static int ip6gre_tap_init(struct net_device *dev) 1795 { 1796 int ret; 1797 1798 ret = ip6gre_tunnel_init_common(dev); 1799 if (ret) 1800 return ret; 1801 1802 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1803 1804 return 0; 1805 } 1806 1807 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1808 .ndo_init = ip6gre_tap_init, 1809 .ndo_uninit = ip6gre_tunnel_uninit, 1810 .ndo_start_xmit = ip6gre_tunnel_xmit, 1811 .ndo_set_mac_address = eth_mac_addr, 1812 .ndo_validate_addr = eth_validate_addr, 1813 .ndo_change_mtu = ip6_tnl_change_mtu, 1814 .ndo_get_stats64 = ip_tunnel_get_stats64, 1815 .ndo_get_iflink = ip6_tnl_get_iflink, 1816 }; 1817 1818 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1819 { 1820 int t_hlen; 1821 1822 tunnel->tun_hlen = 8; 1823 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1824 erspan_hdr_len(tunnel->parms.erspan_ver); 1825 1826 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1827 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1828 return t_hlen; 1829 } 1830 1831 static int ip6erspan_tap_init(struct net_device *dev) 1832 { 1833 struct ip6_tnl *tunnel; 1834 int t_hlen; 1835 int ret; 1836 1837 tunnel = netdev_priv(dev); 1838 1839 tunnel->dev = dev; 1840 tunnel->net = dev_net(dev); 1841 strcpy(tunnel->parms.name, dev->name); 1842 1843 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1844 if (!dev->tstats) 1845 return -ENOMEM; 1846 1847 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1848 if (ret) 1849 goto cleanup_alloc_pcpu_stats; 1850 1851 ret = gro_cells_init(&tunnel->gro_cells, dev); 1852 if (ret) 1853 goto cleanup_dst_cache_init; 1854 1855 t_hlen = ip6erspan_calc_hlen(tunnel); 1856 dev->mtu = ETH_DATA_LEN - t_hlen; 1857 if (dev->type == ARPHRD_ETHER) 1858 dev->mtu -= ETH_HLEN; 1859 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1860 dev->mtu -= 8; 1861 1862 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1863 ip6erspan_tnl_link_config(tunnel, 1); 1864 1865 return 0; 1866 1867 cleanup_dst_cache_init: 1868 dst_cache_destroy(&tunnel->dst_cache); 1869 cleanup_alloc_pcpu_stats: 1870 free_percpu(dev->tstats); 1871 dev->tstats = NULL; 1872 return ret; 1873 } 1874 1875 static const struct net_device_ops ip6erspan_netdev_ops = { 1876 .ndo_init = ip6erspan_tap_init, 1877 .ndo_uninit = ip6erspan_tunnel_uninit, 1878 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1879 .ndo_set_mac_address = eth_mac_addr, 1880 .ndo_validate_addr = eth_validate_addr, 1881 .ndo_change_mtu = ip6_tnl_change_mtu, 1882 .ndo_get_stats64 = ip_tunnel_get_stats64, 1883 .ndo_get_iflink = ip6_tnl_get_iflink, 1884 }; 1885 1886 static void ip6gre_tap_setup(struct net_device *dev) 1887 { 1888 1889 ether_setup(dev); 1890 1891 dev->max_mtu = 0; 1892 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1893 dev->needs_free_netdev = true; 1894 dev->priv_destructor = ip6gre_dev_free; 1895 1896 dev->features |= NETIF_F_NETNS_LOCAL; 1897 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1898 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1899 netif_keep_dst(dev); 1900 } 1901 1902 bool is_ip6gretap_dev(const struct net_device *dev) 1903 { 1904 return dev->netdev_ops == &ip6gre_tap_netdev_ops; 1905 } 1906 EXPORT_SYMBOL_GPL(is_ip6gretap_dev); 1907 1908 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1909 struct ip_tunnel_encap *ipencap) 1910 { 1911 bool ret = false; 1912 1913 memset(ipencap, 0, sizeof(*ipencap)); 1914 1915 if (!data) 1916 return ret; 1917 1918 if (data[IFLA_GRE_ENCAP_TYPE]) { 1919 ret = true; 1920 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1921 } 1922 1923 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1924 ret = true; 1925 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1926 } 1927 1928 if (data[IFLA_GRE_ENCAP_SPORT]) { 1929 ret = true; 1930 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1931 } 1932 1933 if (data[IFLA_GRE_ENCAP_DPORT]) { 1934 ret = true; 1935 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1936 } 1937 1938 return ret; 1939 } 1940 1941 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1942 struct nlattr *tb[], struct nlattr *data[], 1943 struct netlink_ext_ack *extack) 1944 { 1945 struct ip6_tnl *nt; 1946 struct ip_tunnel_encap ipencap; 1947 int err; 1948 1949 nt = netdev_priv(dev); 1950 1951 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1952 int err = ip6_tnl_encap_setup(nt, &ipencap); 1953 1954 if (err < 0) 1955 return err; 1956 } 1957 1958 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1959 eth_hw_addr_random(dev); 1960 1961 nt->dev = dev; 1962 nt->net = dev_net(dev); 1963 1964 err = register_netdevice(dev); 1965 if (err) 1966 goto out; 1967 1968 if (tb[IFLA_MTU]) 1969 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1970 1971 dev_hold(dev); 1972 1973 out: 1974 return err; 1975 } 1976 1977 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1978 struct nlattr *tb[], struct nlattr *data[], 1979 struct netlink_ext_ack *extack) 1980 { 1981 struct ip6_tnl *nt = netdev_priv(dev); 1982 struct net *net = dev_net(dev); 1983 struct ip6gre_net *ign; 1984 int err; 1985 1986 ip6gre_netlink_parms(data, &nt->parms); 1987 ign = net_generic(net, ip6gre_net_id); 1988 1989 if (nt->parms.collect_md) { 1990 if (rtnl_dereference(ign->collect_md_tun)) 1991 return -EEXIST; 1992 } else { 1993 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1994 return -EEXIST; 1995 } 1996 1997 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 1998 if (!err) { 1999 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2000 ip6gre_tunnel_link_md(ign, nt); 2001 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2002 } 2003 return err; 2004 } 2005 2006 static struct ip6_tnl * 2007 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2008 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2009 struct netlink_ext_ack *extack) 2010 { 2011 struct ip6_tnl *t, *nt = netdev_priv(dev); 2012 struct net *net = nt->net; 2013 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2014 struct ip_tunnel_encap ipencap; 2015 2016 if (dev == ign->fb_tunnel_dev) 2017 return ERR_PTR(-EINVAL); 2018 2019 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2020 int err = ip6_tnl_encap_setup(nt, &ipencap); 2021 2022 if (err < 0) 2023 return ERR_PTR(err); 2024 } 2025 2026 ip6gre_netlink_parms(data, p_p); 2027 2028 t = ip6gre_tunnel_locate(net, p_p, 0); 2029 2030 if (t) { 2031 if (t->dev != dev) 2032 return ERR_PTR(-EEXIST); 2033 } else { 2034 t = nt; 2035 } 2036 2037 return t; 2038 } 2039 2040 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2041 struct nlattr *data[], 2042 struct netlink_ext_ack *extack) 2043 { 2044 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2045 struct __ip6_tnl_parm p; 2046 struct ip6_tnl *t; 2047 2048 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2049 if (IS_ERR(t)) 2050 return PTR_ERR(t); 2051 2052 ip6gre_tunnel_unlink_md(ign, t); 2053 ip6gre_tunnel_unlink(ign, t); 2054 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2055 ip6gre_tunnel_link_md(ign, t); 2056 ip6gre_tunnel_link(ign, t); 2057 return 0; 2058 } 2059 2060 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2061 { 2062 struct net *net = dev_net(dev); 2063 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2064 2065 if (dev != ign->fb_tunnel_dev) 2066 unregister_netdevice_queue(dev, head); 2067 } 2068 2069 static size_t ip6gre_get_size(const struct net_device *dev) 2070 { 2071 return 2072 /* IFLA_GRE_LINK */ 2073 nla_total_size(4) + 2074 /* IFLA_GRE_IFLAGS */ 2075 nla_total_size(2) + 2076 /* IFLA_GRE_OFLAGS */ 2077 nla_total_size(2) + 2078 /* IFLA_GRE_IKEY */ 2079 nla_total_size(4) + 2080 /* IFLA_GRE_OKEY */ 2081 nla_total_size(4) + 2082 /* IFLA_GRE_LOCAL */ 2083 nla_total_size(sizeof(struct in6_addr)) + 2084 /* IFLA_GRE_REMOTE */ 2085 nla_total_size(sizeof(struct in6_addr)) + 2086 /* IFLA_GRE_TTL */ 2087 nla_total_size(1) + 2088 /* IFLA_GRE_ENCAP_LIMIT */ 2089 nla_total_size(1) + 2090 /* IFLA_GRE_FLOWINFO */ 2091 nla_total_size(4) + 2092 /* IFLA_GRE_FLAGS */ 2093 nla_total_size(4) + 2094 /* IFLA_GRE_ENCAP_TYPE */ 2095 nla_total_size(2) + 2096 /* IFLA_GRE_ENCAP_FLAGS */ 2097 nla_total_size(2) + 2098 /* IFLA_GRE_ENCAP_SPORT */ 2099 nla_total_size(2) + 2100 /* IFLA_GRE_ENCAP_DPORT */ 2101 nla_total_size(2) + 2102 /* IFLA_GRE_COLLECT_METADATA */ 2103 nla_total_size(0) + 2104 /* IFLA_GRE_FWMARK */ 2105 nla_total_size(4) + 2106 /* IFLA_GRE_ERSPAN_INDEX */ 2107 nla_total_size(4) + 2108 0; 2109 } 2110 2111 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2112 { 2113 struct ip6_tnl *t = netdev_priv(dev); 2114 struct __ip6_tnl_parm *p = &t->parms; 2115 2116 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2117 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2118 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2119 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2120 gre_tnl_flags_to_gre_flags(p->o_flags)) || 2121 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2122 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2123 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2124 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2125 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2126 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2127 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2128 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2129 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) || 2130 nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2131 goto nla_put_failure; 2132 2133 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2134 t->encap.type) || 2135 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2136 t->encap.sport) || 2137 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2138 t->encap.dport) || 2139 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2140 t->encap.flags)) 2141 goto nla_put_failure; 2142 2143 if (p->collect_md) { 2144 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2145 goto nla_put_failure; 2146 } 2147 2148 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2149 goto nla_put_failure; 2150 2151 if (p->erspan_ver == 1) { 2152 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2153 goto nla_put_failure; 2154 } else if (p->erspan_ver == 2) { 2155 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2156 goto nla_put_failure; 2157 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2158 goto nla_put_failure; 2159 } 2160 2161 return 0; 2162 2163 nla_put_failure: 2164 return -EMSGSIZE; 2165 } 2166 2167 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2168 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2169 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2170 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2171 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2172 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2173 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 2174 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 2175 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2176 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2177 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2178 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2179 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2180 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2181 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2182 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2183 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2184 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2185 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2186 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2187 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2188 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2189 }; 2190 2191 static void ip6erspan_tap_setup(struct net_device *dev) 2192 { 2193 ether_setup(dev); 2194 2195 dev->netdev_ops = &ip6erspan_netdev_ops; 2196 dev->needs_free_netdev = true; 2197 dev->priv_destructor = ip6gre_dev_free; 2198 2199 dev->features |= NETIF_F_NETNS_LOCAL; 2200 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2201 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2202 netif_keep_dst(dev); 2203 } 2204 2205 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2206 struct nlattr *tb[], struct nlattr *data[], 2207 struct netlink_ext_ack *extack) 2208 { 2209 struct ip6_tnl *nt = netdev_priv(dev); 2210 struct net *net = dev_net(dev); 2211 struct ip6gre_net *ign; 2212 int err; 2213 2214 ip6gre_netlink_parms(data, &nt->parms); 2215 ign = net_generic(net, ip6gre_net_id); 2216 2217 if (nt->parms.collect_md) { 2218 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2219 return -EEXIST; 2220 } else { 2221 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2222 return -EEXIST; 2223 } 2224 2225 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2226 if (!err) { 2227 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2228 ip6erspan_tunnel_link_md(ign, nt); 2229 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2230 } 2231 return err; 2232 } 2233 2234 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2235 { 2236 ip6gre_tnl_link_config_common(t); 2237 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2238 } 2239 2240 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2241 const struct __ip6_tnl_parm *p, int set_mtu) 2242 { 2243 ip6gre_tnl_copy_tnl_parm(t, p); 2244 ip6erspan_tnl_link_config(t, set_mtu); 2245 return 0; 2246 } 2247 2248 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2249 struct nlattr *data[], 2250 struct netlink_ext_ack *extack) 2251 { 2252 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2253 struct __ip6_tnl_parm p; 2254 struct ip6_tnl *t; 2255 2256 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2257 if (IS_ERR(t)) 2258 return PTR_ERR(t); 2259 2260 ip6gre_tunnel_unlink_md(ign, t); 2261 ip6gre_tunnel_unlink(ign, t); 2262 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2263 ip6erspan_tunnel_link_md(ign, t); 2264 ip6gre_tunnel_link(ign, t); 2265 return 0; 2266 } 2267 2268 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2269 .kind = "ip6gre", 2270 .maxtype = IFLA_GRE_MAX, 2271 .policy = ip6gre_policy, 2272 .priv_size = sizeof(struct ip6_tnl), 2273 .setup = ip6gre_tunnel_setup, 2274 .validate = ip6gre_tunnel_validate, 2275 .newlink = ip6gre_newlink, 2276 .changelink = ip6gre_changelink, 2277 .dellink = ip6gre_dellink, 2278 .get_size = ip6gre_get_size, 2279 .fill_info = ip6gre_fill_info, 2280 .get_link_net = ip6_tnl_get_link_net, 2281 }; 2282 2283 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2284 .kind = "ip6gretap", 2285 .maxtype = IFLA_GRE_MAX, 2286 .policy = ip6gre_policy, 2287 .priv_size = sizeof(struct ip6_tnl), 2288 .setup = ip6gre_tap_setup, 2289 .validate = ip6gre_tap_validate, 2290 .newlink = ip6gre_newlink, 2291 .changelink = ip6gre_changelink, 2292 .get_size = ip6gre_get_size, 2293 .fill_info = ip6gre_fill_info, 2294 .get_link_net = ip6_tnl_get_link_net, 2295 }; 2296 2297 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2298 .kind = "ip6erspan", 2299 .maxtype = IFLA_GRE_MAX, 2300 .policy = ip6gre_policy, 2301 .priv_size = sizeof(struct ip6_tnl), 2302 .setup = ip6erspan_tap_setup, 2303 .validate = ip6erspan_tap_validate, 2304 .newlink = ip6erspan_newlink, 2305 .changelink = ip6erspan_changelink, 2306 .get_size = ip6gre_get_size, 2307 .fill_info = ip6gre_fill_info, 2308 .get_link_net = ip6_tnl_get_link_net, 2309 }; 2310 2311 /* 2312 * And now the modules code and kernel interface. 2313 */ 2314 2315 static int __init ip6gre_init(void) 2316 { 2317 int err; 2318 2319 pr_info("GRE over IPv6 tunneling driver\n"); 2320 2321 err = register_pernet_device(&ip6gre_net_ops); 2322 if (err < 0) 2323 return err; 2324 2325 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2326 if (err < 0) { 2327 pr_info("%s: can't add protocol\n", __func__); 2328 goto add_proto_failed; 2329 } 2330 2331 err = rtnl_link_register(&ip6gre_link_ops); 2332 if (err < 0) 2333 goto rtnl_link_failed; 2334 2335 err = rtnl_link_register(&ip6gre_tap_ops); 2336 if (err < 0) 2337 goto tap_ops_failed; 2338 2339 err = rtnl_link_register(&ip6erspan_tap_ops); 2340 if (err < 0) 2341 goto erspan_link_failed; 2342 2343 out: 2344 return err; 2345 2346 erspan_link_failed: 2347 rtnl_link_unregister(&ip6gre_tap_ops); 2348 tap_ops_failed: 2349 rtnl_link_unregister(&ip6gre_link_ops); 2350 rtnl_link_failed: 2351 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2352 add_proto_failed: 2353 unregister_pernet_device(&ip6gre_net_ops); 2354 goto out; 2355 } 2356 2357 static void __exit ip6gre_fini(void) 2358 { 2359 rtnl_link_unregister(&ip6gre_tap_ops); 2360 rtnl_link_unregister(&ip6gre_link_ops); 2361 rtnl_link_unregister(&ip6erspan_tap_ops); 2362 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2363 unregister_pernet_device(&ip6gre_net_ops); 2364 } 2365 2366 module_init(ip6gre_init); 2367 module_exit(ip6gre_fini); 2368 MODULE_LICENSE("GPL"); 2369 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2370 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2371 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2372 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2373 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2374 MODULE_ALIAS_NETDEV("ip6gre0"); 2375