1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * GRE over IPv6 protocol decoder. 4 * 5 * Authors: Dmitry Kozlov (xeb@mail.ru) 6 */ 7 8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 10 #include <linux/capability.h> 11 #include <linux/module.h> 12 #include <linux/types.h> 13 #include <linux/kernel.h> 14 #include <linux/slab.h> 15 #include <linux/uaccess.h> 16 #include <linux/skbuff.h> 17 #include <linux/netdevice.h> 18 #include <linux/in.h> 19 #include <linux/tcp.h> 20 #include <linux/udp.h> 21 #include <linux/if_arp.h> 22 #include <linux/init.h> 23 #include <linux/in6.h> 24 #include <linux/inetdevice.h> 25 #include <linux/igmp.h> 26 #include <linux/netfilter_ipv4.h> 27 #include <linux/etherdevice.h> 28 #include <linux/if_ether.h> 29 #include <linux/hash.h> 30 #include <linux/if_tunnel.h> 31 #include <linux/ip6_tunnel.h> 32 33 #include <net/sock.h> 34 #include <net/ip.h> 35 #include <net/ip_tunnels.h> 36 #include <net/icmp.h> 37 #include <net/protocol.h> 38 #include <net/addrconf.h> 39 #include <net/arp.h> 40 #include <net/checksum.h> 41 #include <net/dsfield.h> 42 #include <net/inet_ecn.h> 43 #include <net/xfrm.h> 44 #include <net/net_namespace.h> 45 #include <net/netns/generic.h> 46 #include <net/rtnetlink.h> 47 48 #include <net/ipv6.h> 49 #include <net/ip6_fib.h> 50 #include <net/ip6_route.h> 51 #include <net/ip6_tunnel.h> 52 #include <net/gre.h> 53 #include <net/erspan.h> 54 #include <net/dst_metadata.h> 55 56 57 static bool log_ecn_error = true; 58 module_param(log_ecn_error, bool, 0644); 59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 60 61 #define IP6_GRE_HASH_SIZE_SHIFT 5 62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT) 63 64 static unsigned int ip6gre_net_id __read_mostly; 65 struct ip6gre_net { 66 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE]; 67 68 struct ip6_tnl __rcu *collect_md_tun; 69 struct ip6_tnl __rcu *collect_md_tun_erspan; 70 struct net_device *fb_tunnel_dev; 71 }; 72 73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly; 76 static int ip6gre_tunnel_init(struct net_device *dev); 77 static void ip6gre_tunnel_setup(struct net_device *dev); 78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t); 79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu); 80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu); 81 82 /* Tunnel hash table */ 83 84 /* 85 4 hash tables: 86 87 3: (remote,local) 88 2: (remote,*) 89 1: (*,local) 90 0: (*,*) 91 92 We require exact key match i.e. if a key is present in packet 93 it will match only tunnel with the same key; if it is not present, 94 it will match only keyless tunnel. 95 96 All keysless packets, if not matched configured keyless tunnels 97 will match fallback tunnel. 98 */ 99 100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 101 static u32 HASH_ADDR(const struct in6_addr *addr) 102 { 103 u32 hash = ipv6_addr_hash(addr); 104 105 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 106 } 107 108 #define tunnels_r_l tunnels[3] 109 #define tunnels_r tunnels[2] 110 #define tunnels_l tunnels[1] 111 #define tunnels_wc tunnels[0] 112 113 /* Given src, dst and key, find appropriate for input tunnel. */ 114 115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 116 const struct in6_addr *remote, const struct in6_addr *local, 117 __be32 key, __be16 gre_proto) 118 { 119 struct net *net = dev_net(dev); 120 int link = dev->ifindex; 121 unsigned int h0 = HASH_ADDR(remote); 122 unsigned int h1 = HASH_KEY(key); 123 struct ip6_tnl *t, *cand = NULL; 124 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 125 int dev_type = (gre_proto == htons(ETH_P_TEB) || 126 gre_proto == htons(ETH_P_ERSPAN) || 127 gre_proto == htons(ETH_P_ERSPAN2)) ? 128 ARPHRD_ETHER : ARPHRD_IP6GRE; 129 int score, cand_score = 4; 130 struct net_device *ndev; 131 132 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 133 if (!ipv6_addr_equal(local, &t->parms.laddr) || 134 !ipv6_addr_equal(remote, &t->parms.raddr) || 135 key != t->parms.i_key || 136 !(t->dev->flags & IFF_UP)) 137 continue; 138 139 if (t->dev->type != ARPHRD_IP6GRE && 140 t->dev->type != dev_type) 141 continue; 142 143 score = 0; 144 if (t->parms.link != link) 145 score |= 1; 146 if (t->dev->type != dev_type) 147 score |= 2; 148 if (score == 0) 149 return t; 150 151 if (score < cand_score) { 152 cand = t; 153 cand_score = score; 154 } 155 } 156 157 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 158 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 159 key != t->parms.i_key || 160 !(t->dev->flags & IFF_UP)) 161 continue; 162 163 if (t->dev->type != ARPHRD_IP6GRE && 164 t->dev->type != dev_type) 165 continue; 166 167 score = 0; 168 if (t->parms.link != link) 169 score |= 1; 170 if (t->dev->type != dev_type) 171 score |= 2; 172 if (score == 0) 173 return t; 174 175 if (score < cand_score) { 176 cand = t; 177 cand_score = score; 178 } 179 } 180 181 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 182 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 183 (!ipv6_addr_equal(local, &t->parms.raddr) || 184 !ipv6_addr_is_multicast(local))) || 185 key != t->parms.i_key || 186 !(t->dev->flags & IFF_UP)) 187 continue; 188 189 if (t->dev->type != ARPHRD_IP6GRE && 190 t->dev->type != dev_type) 191 continue; 192 193 score = 0; 194 if (t->parms.link != link) 195 score |= 1; 196 if (t->dev->type != dev_type) 197 score |= 2; 198 if (score == 0) 199 return t; 200 201 if (score < cand_score) { 202 cand = t; 203 cand_score = score; 204 } 205 } 206 207 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 208 if (t->parms.i_key != key || 209 !(t->dev->flags & IFF_UP)) 210 continue; 211 212 if (t->dev->type != ARPHRD_IP6GRE && 213 t->dev->type != dev_type) 214 continue; 215 216 score = 0; 217 if (t->parms.link != link) 218 score |= 1; 219 if (t->dev->type != dev_type) 220 score |= 2; 221 if (score == 0) 222 return t; 223 224 if (score < cand_score) { 225 cand = t; 226 cand_score = score; 227 } 228 } 229 230 if (cand) 231 return cand; 232 233 if (gre_proto == htons(ETH_P_ERSPAN) || 234 gre_proto == htons(ETH_P_ERSPAN2)) 235 t = rcu_dereference(ign->collect_md_tun_erspan); 236 else 237 t = rcu_dereference(ign->collect_md_tun); 238 239 if (t && t->dev->flags & IFF_UP) 240 return t; 241 242 ndev = READ_ONCE(ign->fb_tunnel_dev); 243 if (ndev && ndev->flags & IFF_UP) 244 return netdev_priv(ndev); 245 246 return NULL; 247 } 248 249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 250 const struct __ip6_tnl_parm *p) 251 { 252 const struct in6_addr *remote = &p->raddr; 253 const struct in6_addr *local = &p->laddr; 254 unsigned int h = HASH_KEY(p->i_key); 255 int prio = 0; 256 257 if (!ipv6_addr_any(local)) 258 prio |= 1; 259 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 260 prio |= 2; 261 h ^= HASH_ADDR(remote); 262 } 263 264 return &ign->tunnels[prio][h]; 265 } 266 267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 268 { 269 if (t->parms.collect_md) 270 rcu_assign_pointer(ign->collect_md_tun, t); 271 } 272 273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 274 { 275 if (t->parms.collect_md) 276 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 277 } 278 279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 280 { 281 if (t->parms.collect_md) 282 rcu_assign_pointer(ign->collect_md_tun, NULL); 283 } 284 285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 286 struct ip6_tnl *t) 287 { 288 if (t->parms.collect_md) 289 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 290 } 291 292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 293 const struct ip6_tnl *t) 294 { 295 return __ip6gre_bucket(ign, &t->parms); 296 } 297 298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 299 { 300 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 301 302 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 303 rcu_assign_pointer(*tp, t); 304 } 305 306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 307 { 308 struct ip6_tnl __rcu **tp; 309 struct ip6_tnl *iter; 310 311 for (tp = ip6gre_bucket(ign, t); 312 (iter = rtnl_dereference(*tp)) != NULL; 313 tp = &iter->next) { 314 if (t == iter) { 315 rcu_assign_pointer(*tp, t->next); 316 break; 317 } 318 } 319 } 320 321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 322 const struct __ip6_tnl_parm *parms, 323 int type) 324 { 325 const struct in6_addr *remote = &parms->raddr; 326 const struct in6_addr *local = &parms->laddr; 327 __be32 key = parms->i_key; 328 int link = parms->link; 329 struct ip6_tnl *t; 330 struct ip6_tnl __rcu **tp; 331 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 332 333 for (tp = __ip6gre_bucket(ign, parms); 334 (t = rtnl_dereference(*tp)) != NULL; 335 tp = &t->next) 336 if (ipv6_addr_equal(local, &t->parms.laddr) && 337 ipv6_addr_equal(remote, &t->parms.raddr) && 338 key == t->parms.i_key && 339 link == t->parms.link && 340 type == t->dev->type) 341 break; 342 343 return t; 344 } 345 346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 347 const struct __ip6_tnl_parm *parms, int create) 348 { 349 struct ip6_tnl *t, *nt; 350 struct net_device *dev; 351 char name[IFNAMSIZ]; 352 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 353 354 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 355 if (t && create) 356 return NULL; 357 if (t || !create) 358 return t; 359 360 if (parms->name[0]) { 361 if (!dev_valid_name(parms->name)) 362 return NULL; 363 strlcpy(name, parms->name, IFNAMSIZ); 364 } else { 365 strcpy(name, "ip6gre%d"); 366 } 367 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 368 ip6gre_tunnel_setup); 369 if (!dev) 370 return NULL; 371 372 dev_net_set(dev, net); 373 374 nt = netdev_priv(dev); 375 nt->parms = *parms; 376 dev->rtnl_link_ops = &ip6gre_link_ops; 377 378 nt->dev = dev; 379 nt->net = dev_net(dev); 380 381 if (register_netdevice(dev) < 0) 382 goto failed_free; 383 384 ip6gre_tnl_link_config(nt, 1); 385 386 /* Can use a lockless transmit, unless we generate output sequences */ 387 if (!(nt->parms.o_flags & TUNNEL_SEQ)) 388 dev->features |= NETIF_F_LLTX; 389 390 ip6gre_tunnel_link(ign, nt); 391 return nt; 392 393 failed_free: 394 free_netdev(dev); 395 return NULL; 396 } 397 398 static void ip6erspan_tunnel_uninit(struct net_device *dev) 399 { 400 struct ip6_tnl *t = netdev_priv(dev); 401 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 402 403 ip6erspan_tunnel_unlink_md(ign, t); 404 ip6gre_tunnel_unlink(ign, t); 405 dst_cache_reset(&t->dst_cache); 406 dev_put(dev); 407 } 408 409 static void ip6gre_tunnel_uninit(struct net_device *dev) 410 { 411 struct ip6_tnl *t = netdev_priv(dev); 412 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 413 414 ip6gre_tunnel_unlink_md(ign, t); 415 ip6gre_tunnel_unlink(ign, t); 416 if (ign->fb_tunnel_dev == dev) 417 WRITE_ONCE(ign->fb_tunnel_dev, NULL); 418 dst_cache_reset(&t->dst_cache); 419 dev_put(dev); 420 } 421 422 423 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 424 u8 type, u8 code, int offset, __be32 info) 425 { 426 struct net *net = dev_net(skb->dev); 427 const struct ipv6hdr *ipv6h; 428 struct tnl_ptk_info tpi; 429 struct ip6_tnl *t; 430 431 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 432 offset) < 0) 433 return -EINVAL; 434 435 ipv6h = (const struct ipv6hdr *)skb->data; 436 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 437 tpi.key, tpi.proto); 438 if (!t) 439 return -ENOENT; 440 441 switch (type) { 442 case ICMPV6_DEST_UNREACH: 443 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 444 t->parms.name); 445 if (code != ICMPV6_PORT_UNREACH) 446 break; 447 return 0; 448 case ICMPV6_TIME_EXCEED: 449 if (code == ICMPV6_EXC_HOPLIMIT) { 450 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 451 t->parms.name); 452 break; 453 } 454 return 0; 455 case ICMPV6_PARAMPROB: { 456 struct ipv6_tlv_tnl_enc_lim *tel; 457 __u32 teli; 458 459 teli = 0; 460 if (code == ICMPV6_HDR_FIELD) 461 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 462 463 if (teli && teli == be32_to_cpu(info) - 2) { 464 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 465 if (tel->encap_limit == 0) { 466 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 467 t->parms.name); 468 } 469 } else { 470 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 471 t->parms.name); 472 } 473 return 0; 474 } 475 case ICMPV6_PKT_TOOBIG: 476 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 477 return 0; 478 case NDISC_REDIRECT: 479 ip6_redirect(skb, net, skb->dev->ifindex, 0, 480 sock_net_uid(net, NULL)); 481 return 0; 482 } 483 484 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 485 t->err_count++; 486 else 487 t->err_count = 1; 488 t->err_time = jiffies; 489 490 return 0; 491 } 492 493 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 494 { 495 const struct ipv6hdr *ipv6h; 496 struct ip6_tnl *tunnel; 497 498 ipv6h = ipv6_hdr(skb); 499 tunnel = ip6gre_tunnel_lookup(skb->dev, 500 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 501 tpi->proto); 502 if (tunnel) { 503 if (tunnel->parms.collect_md) { 504 struct metadata_dst *tun_dst; 505 __be64 tun_id; 506 __be16 flags; 507 508 flags = tpi->flags; 509 tun_id = key32_to_tunnel_id(tpi->key); 510 511 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 512 if (!tun_dst) 513 return PACKET_REJECT; 514 515 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 516 } else { 517 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 518 } 519 520 return PACKET_RCVD; 521 } 522 523 return PACKET_REJECT; 524 } 525 526 static int ip6erspan_rcv(struct sk_buff *skb, 527 struct tnl_ptk_info *tpi, 528 int gre_hdr_len) 529 { 530 struct erspan_base_hdr *ershdr; 531 const struct ipv6hdr *ipv6h; 532 struct erspan_md2 *md2; 533 struct ip6_tnl *tunnel; 534 u8 ver; 535 536 ipv6h = ipv6_hdr(skb); 537 ershdr = (struct erspan_base_hdr *)skb->data; 538 ver = ershdr->ver; 539 540 tunnel = ip6gre_tunnel_lookup(skb->dev, 541 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 542 tpi->proto); 543 if (tunnel) { 544 int len = erspan_hdr_len(ver); 545 546 if (unlikely(!pskb_may_pull(skb, len))) 547 return PACKET_REJECT; 548 549 if (__iptunnel_pull_header(skb, len, 550 htons(ETH_P_TEB), 551 false, false) < 0) 552 return PACKET_REJECT; 553 554 if (tunnel->parms.collect_md) { 555 struct erspan_metadata *pkt_md, *md; 556 struct metadata_dst *tun_dst; 557 struct ip_tunnel_info *info; 558 unsigned char *gh; 559 __be64 tun_id; 560 __be16 flags; 561 562 tpi->flags |= TUNNEL_KEY; 563 flags = tpi->flags; 564 tun_id = key32_to_tunnel_id(tpi->key); 565 566 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 567 sizeof(*md)); 568 if (!tun_dst) 569 return PACKET_REJECT; 570 571 /* skb can be uncloned in __iptunnel_pull_header, so 572 * old pkt_md is no longer valid and we need to reset 573 * it 574 */ 575 gh = skb_network_header(skb) + 576 skb_network_header_len(skb); 577 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 578 sizeof(*ershdr)); 579 info = &tun_dst->u.tun_info; 580 md = ip_tunnel_info_opts(info); 581 md->version = ver; 582 md2 = &md->u.md2; 583 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 584 ERSPAN_V2_MDSIZE); 585 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 586 info->options_len = sizeof(*md); 587 588 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 589 590 } else { 591 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 592 } 593 594 return PACKET_RCVD; 595 } 596 597 return PACKET_REJECT; 598 } 599 600 static int gre_rcv(struct sk_buff *skb) 601 { 602 struct tnl_ptk_info tpi; 603 bool csum_err = false; 604 int hdr_len; 605 606 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 607 if (hdr_len < 0) 608 goto drop; 609 610 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 611 goto drop; 612 613 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 614 tpi.proto == htons(ETH_P_ERSPAN2))) { 615 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) 616 return 0; 617 goto out; 618 } 619 620 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 621 return 0; 622 623 out: 624 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 625 drop: 626 kfree_skb(skb); 627 return 0; 628 } 629 630 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 631 { 632 return iptunnel_handle_offloads(skb, 633 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 634 } 635 636 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 637 struct net_device *dev, 638 struct flowi6 *fl6, __u8 *dsfield, 639 int *encap_limit) 640 { 641 const struct iphdr *iph = ip_hdr(skb); 642 struct ip6_tnl *t = netdev_priv(dev); 643 644 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 645 *encap_limit = t->parms.encap_limit; 646 647 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 648 649 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 650 *dsfield = ipv4_get_dsfield(iph); 651 else 652 *dsfield = ip6_tclass(t->parms.flowinfo); 653 654 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 655 fl6->flowi6_mark = skb->mark; 656 else 657 fl6->flowi6_mark = t->parms.fwmark; 658 659 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 660 } 661 662 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 663 struct net_device *dev, 664 struct flowi6 *fl6, __u8 *dsfield, 665 int *encap_limit) 666 { 667 struct ipv6hdr *ipv6h; 668 struct ip6_tnl *t = netdev_priv(dev); 669 __u16 offset; 670 671 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 672 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 673 ipv6h = ipv6_hdr(skb); 674 675 if (offset > 0) { 676 struct ipv6_tlv_tnl_enc_lim *tel; 677 678 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 679 if (tel->encap_limit == 0) { 680 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 681 ICMPV6_HDR_FIELD, offset + 2); 682 return -1; 683 } 684 *encap_limit = tel->encap_limit - 1; 685 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 686 *encap_limit = t->parms.encap_limit; 687 } 688 689 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 690 691 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 692 *dsfield = ipv6_get_dsfield(ipv6h); 693 else 694 *dsfield = ip6_tclass(t->parms.flowinfo); 695 696 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 697 fl6->flowlabel |= ip6_flowlabel(ipv6h); 698 699 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 700 fl6->flowi6_mark = skb->mark; 701 else 702 fl6->flowi6_mark = t->parms.fwmark; 703 704 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 705 706 return 0; 707 } 708 709 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb) 710 { 711 struct ip_tunnel_info *tun_info; 712 713 tun_info = skb_tunnel_info(skb); 714 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))) 715 return ERR_PTR(-EINVAL); 716 717 return tun_info; 718 } 719 720 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 721 struct net_device *dev, __u8 dsfield, 722 struct flowi6 *fl6, int encap_limit, 723 __u32 *pmtu, __be16 proto) 724 { 725 struct ip6_tnl *tunnel = netdev_priv(dev); 726 __be16 protocol; 727 728 if (dev->type == ARPHRD_ETHER) 729 IPCB(skb)->flags = 0; 730 731 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 732 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 733 else 734 fl6->daddr = tunnel->parms.raddr; 735 736 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 737 return -ENOMEM; 738 739 /* Push GRE header. */ 740 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 741 742 if (tunnel->parms.collect_md) { 743 struct ip_tunnel_info *tun_info; 744 const struct ip_tunnel_key *key; 745 __be16 flags; 746 747 tun_info = skb_tunnel_info_txcheck(skb); 748 if (IS_ERR(tun_info) || 749 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 750 return -EINVAL; 751 752 key = &tun_info->key; 753 memset(fl6, 0, sizeof(*fl6)); 754 fl6->flowi6_proto = IPPROTO_GRE; 755 fl6->daddr = key->u.ipv6.dst; 756 fl6->flowlabel = key->label; 757 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 758 759 dsfield = key->tos; 760 flags = key->tun_flags & 761 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 762 tunnel->tun_hlen = gre_calc_hlen(flags); 763 764 gre_build_header(skb, tunnel->tun_hlen, 765 flags, protocol, 766 tunnel_id_to_key32(tun_info->key.tun_id), 767 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 768 : 0); 769 770 } else { 771 if (tunnel->parms.o_flags & TUNNEL_SEQ) 772 tunnel->o_seqno++; 773 774 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 775 protocol, tunnel->parms.o_key, 776 htonl(tunnel->o_seqno)); 777 } 778 779 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 780 NEXTHDR_GRE); 781 } 782 783 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 784 { 785 struct ip6_tnl *t = netdev_priv(dev); 786 int encap_limit = -1; 787 struct flowi6 fl6; 788 __u8 dsfield = 0; 789 __u32 mtu; 790 int err; 791 792 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 793 794 if (!t->parms.collect_md) 795 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 796 &dsfield, &encap_limit); 797 798 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 799 if (err) 800 return -1; 801 802 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 803 skb->protocol); 804 if (err != 0) { 805 /* XXX: send ICMP error even if DF is not set. */ 806 if (err == -EMSGSIZE) 807 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 808 htonl(mtu)); 809 return -1; 810 } 811 812 return 0; 813 } 814 815 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 816 { 817 struct ip6_tnl *t = netdev_priv(dev); 818 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 819 int encap_limit = -1; 820 struct flowi6 fl6; 821 __u8 dsfield = 0; 822 __u32 mtu; 823 int err; 824 825 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 826 return -1; 827 828 if (!t->parms.collect_md && 829 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 830 return -1; 831 832 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 833 return -1; 834 835 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 836 &mtu, skb->protocol); 837 if (err != 0) { 838 if (err == -EMSGSIZE) 839 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 840 return -1; 841 } 842 843 return 0; 844 } 845 846 /** 847 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 848 * @t: the outgoing tunnel device 849 * @hdr: IPv6 header from the incoming packet 850 * 851 * Description: 852 * Avoid trivial tunneling loop by checking that tunnel exit-point 853 * doesn't match source of incoming packet. 854 * 855 * Return: 856 * 1 if conflict, 857 * 0 else 858 **/ 859 860 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 861 const struct ipv6hdr *hdr) 862 { 863 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 864 } 865 866 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 867 { 868 struct ip6_tnl *t = netdev_priv(dev); 869 int encap_limit = -1; 870 struct flowi6 fl6; 871 __u32 mtu; 872 int err; 873 874 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 875 encap_limit = t->parms.encap_limit; 876 877 if (!t->parms.collect_md) 878 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 879 880 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 881 if (err) 882 return err; 883 884 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 885 886 return err; 887 } 888 889 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 890 struct net_device *dev) 891 { 892 struct ip6_tnl *t = netdev_priv(dev); 893 struct net_device_stats *stats = &t->dev->stats; 894 int ret; 895 896 if (!pskb_inet_may_pull(skb)) 897 goto tx_err; 898 899 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 900 goto tx_err; 901 902 switch (skb->protocol) { 903 case htons(ETH_P_IP): 904 ret = ip6gre_xmit_ipv4(skb, dev); 905 break; 906 case htons(ETH_P_IPV6): 907 ret = ip6gre_xmit_ipv6(skb, dev); 908 break; 909 default: 910 ret = ip6gre_xmit_other(skb, dev); 911 break; 912 } 913 914 if (ret < 0) 915 goto tx_err; 916 917 return NETDEV_TX_OK; 918 919 tx_err: 920 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 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 ip_tunnel_info *tun_info = NULL; 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 __be16 proto; 940 __u32 mtu; 941 int nhoff; 942 int thoff; 943 944 if (!pskb_inet_may_pull(skb)) 945 goto tx_err; 946 947 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 948 goto tx_err; 949 950 if (gre_handle_offloads(skb, false)) 951 goto tx_err; 952 953 if (skb->len > dev->mtu + dev->hard_header_len) { 954 pskb_trim(skb, dev->mtu + dev->hard_header_len); 955 truncate = true; 956 } 957 958 nhoff = skb_network_header(skb) - skb_mac_header(skb); 959 if (skb->protocol == htons(ETH_P_IP) && 960 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 961 truncate = true; 962 963 thoff = skb_transport_header(skb) - skb_mac_header(skb); 964 if (skb->protocol == htons(ETH_P_IPV6) && 965 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 966 truncate = true; 967 968 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 969 goto tx_err; 970 971 t->parms.o_flags &= ~TUNNEL_KEY; 972 IPCB(skb)->flags = 0; 973 974 /* For collect_md mode, derive fl6 from the tunnel key, 975 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 976 */ 977 if (t->parms.collect_md) { 978 const struct ip_tunnel_key *key; 979 struct erspan_metadata *md; 980 __be32 tun_id; 981 982 tun_info = skb_tunnel_info_txcheck(skb); 983 if (IS_ERR(tun_info) || 984 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 985 goto tx_err; 986 987 key = &tun_info->key; 988 memset(&fl6, 0, sizeof(fl6)); 989 fl6.flowi6_proto = IPPROTO_GRE; 990 fl6.daddr = key->u.ipv6.dst; 991 fl6.flowlabel = key->label; 992 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 993 994 dsfield = key->tos; 995 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 996 goto tx_err; 997 if (tun_info->options_len < sizeof(*md)) 998 goto tx_err; 999 md = ip_tunnel_info_opts(tun_info); 1000 1001 tun_id = tunnel_id_to_key32(key->tun_id); 1002 if (md->version == 1) { 1003 erspan_build_header(skb, 1004 ntohl(tun_id), 1005 ntohl(md->u.index), truncate, 1006 false); 1007 } else if (md->version == 2) { 1008 erspan_build_header_v2(skb, 1009 ntohl(tun_id), 1010 md->u.md2.dir, 1011 get_hwid(&md->u.md2), 1012 truncate, false); 1013 } else { 1014 goto tx_err; 1015 } 1016 } else { 1017 switch (skb->protocol) { 1018 case htons(ETH_P_IP): 1019 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1020 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1021 &dsfield, &encap_limit); 1022 break; 1023 case htons(ETH_P_IPV6): 1024 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1025 goto tx_err; 1026 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1027 &dsfield, &encap_limit)) 1028 goto tx_err; 1029 break; 1030 default: 1031 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1032 break; 1033 } 1034 1035 if (t->parms.erspan_ver == 1) 1036 erspan_build_header(skb, ntohl(t->parms.o_key), 1037 t->parms.index, 1038 truncate, false); 1039 else if (t->parms.erspan_ver == 2) 1040 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1041 t->parms.dir, 1042 t->parms.hwid, 1043 truncate, false); 1044 else 1045 goto tx_err; 1046 1047 fl6.daddr = t->parms.raddr; 1048 } 1049 1050 /* Push GRE header. */ 1051 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1052 : htons(ETH_P_ERSPAN2); 1053 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++)); 1054 1055 /* TooBig packet may have updated dst->dev's mtu */ 1056 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1057 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1058 1059 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1060 NEXTHDR_GRE); 1061 if (err != 0) { 1062 /* XXX: send ICMP error even if DF is not set. */ 1063 if (err == -EMSGSIZE) { 1064 if (skb->protocol == htons(ETH_P_IP)) 1065 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1066 ICMP_FRAG_NEEDED, htonl(mtu)); 1067 else 1068 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1069 } 1070 1071 goto tx_err; 1072 } 1073 return NETDEV_TX_OK; 1074 1075 tx_err: 1076 stats = &t->dev->stats; 1077 if (!IS_ERR(tun_info)) 1078 stats->tx_errors++; 1079 stats->tx_dropped++; 1080 kfree_skb(skb); 1081 return NETDEV_TX_OK; 1082 } 1083 1084 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1085 { 1086 struct net_device *dev = t->dev; 1087 struct __ip6_tnl_parm *p = &t->parms; 1088 struct flowi6 *fl6 = &t->fl.u.ip6; 1089 1090 if (dev->type != ARPHRD_ETHER) { 1091 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1092 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1093 } 1094 1095 /* Set up flowi template */ 1096 fl6->saddr = p->laddr; 1097 fl6->daddr = p->raddr; 1098 fl6->flowi6_oif = p->link; 1099 fl6->flowlabel = 0; 1100 fl6->flowi6_proto = IPPROTO_GRE; 1101 1102 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1103 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1104 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1105 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1106 1107 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1108 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1109 1110 if (p->flags&IP6_TNL_F_CAP_XMIT && 1111 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1112 dev->flags |= IFF_POINTOPOINT; 1113 else 1114 dev->flags &= ~IFF_POINTOPOINT; 1115 } 1116 1117 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1118 int t_hlen) 1119 { 1120 const struct __ip6_tnl_parm *p = &t->parms; 1121 struct net_device *dev = t->dev; 1122 1123 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1124 int strict = (ipv6_addr_type(&p->raddr) & 1125 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1126 1127 struct rt6_info *rt = rt6_lookup(t->net, 1128 &p->raddr, &p->laddr, 1129 p->link, NULL, strict); 1130 1131 if (!rt) 1132 return; 1133 1134 if (rt->dst.dev) { 1135 unsigned short dst_len = rt->dst.dev->hard_header_len + 1136 t_hlen; 1137 1138 if (t->dev->header_ops) 1139 dev->hard_header_len = dst_len; 1140 else 1141 dev->needed_headroom = dst_len; 1142 1143 if (set_mtu) { 1144 dev->mtu = rt->dst.dev->mtu - t_hlen; 1145 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1146 dev->mtu -= 8; 1147 if (dev->type == ARPHRD_ETHER) 1148 dev->mtu -= ETH_HLEN; 1149 1150 if (dev->mtu < IPV6_MIN_MTU) 1151 dev->mtu = IPV6_MIN_MTU; 1152 } 1153 } 1154 ip6_rt_put(rt); 1155 } 1156 } 1157 1158 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1159 { 1160 int t_hlen; 1161 1162 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1163 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1164 1165 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1166 1167 if (tunnel->dev->header_ops) 1168 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1169 else 1170 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1171 1172 return t_hlen; 1173 } 1174 1175 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1176 { 1177 ip6gre_tnl_link_config_common(t); 1178 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1179 } 1180 1181 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1182 const struct __ip6_tnl_parm *p) 1183 { 1184 t->parms.laddr = p->laddr; 1185 t->parms.raddr = p->raddr; 1186 t->parms.flags = p->flags; 1187 t->parms.hop_limit = p->hop_limit; 1188 t->parms.encap_limit = p->encap_limit; 1189 t->parms.flowinfo = p->flowinfo; 1190 t->parms.link = p->link; 1191 t->parms.proto = p->proto; 1192 t->parms.i_key = p->i_key; 1193 t->parms.o_key = p->o_key; 1194 t->parms.i_flags = p->i_flags; 1195 t->parms.o_flags = p->o_flags; 1196 t->parms.fwmark = p->fwmark; 1197 t->parms.erspan_ver = p->erspan_ver; 1198 t->parms.index = p->index; 1199 t->parms.dir = p->dir; 1200 t->parms.hwid = p->hwid; 1201 dst_cache_reset(&t->dst_cache); 1202 } 1203 1204 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1205 int set_mtu) 1206 { 1207 ip6gre_tnl_copy_tnl_parm(t, p); 1208 ip6gre_tnl_link_config(t, set_mtu); 1209 return 0; 1210 } 1211 1212 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1213 const struct ip6_tnl_parm2 *u) 1214 { 1215 p->laddr = u->laddr; 1216 p->raddr = u->raddr; 1217 p->flags = u->flags; 1218 p->hop_limit = u->hop_limit; 1219 p->encap_limit = u->encap_limit; 1220 p->flowinfo = u->flowinfo; 1221 p->link = u->link; 1222 p->i_key = u->i_key; 1223 p->o_key = u->o_key; 1224 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1225 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1226 memcpy(p->name, u->name, sizeof(u->name)); 1227 } 1228 1229 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1230 const struct __ip6_tnl_parm *p) 1231 { 1232 u->proto = IPPROTO_GRE; 1233 u->laddr = p->laddr; 1234 u->raddr = p->raddr; 1235 u->flags = p->flags; 1236 u->hop_limit = p->hop_limit; 1237 u->encap_limit = p->encap_limit; 1238 u->flowinfo = p->flowinfo; 1239 u->link = p->link; 1240 u->i_key = p->i_key; 1241 u->o_key = p->o_key; 1242 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1243 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1244 memcpy(u->name, p->name, sizeof(u->name)); 1245 } 1246 1247 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1248 struct ifreq *ifr, int cmd) 1249 { 1250 int err = 0; 1251 struct ip6_tnl_parm2 p; 1252 struct __ip6_tnl_parm p1; 1253 struct ip6_tnl *t = netdev_priv(dev); 1254 struct net *net = t->net; 1255 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1256 1257 memset(&p1, 0, sizeof(p1)); 1258 1259 switch (cmd) { 1260 case SIOCGETTUNNEL: 1261 if (dev == ign->fb_tunnel_dev) { 1262 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1263 err = -EFAULT; 1264 break; 1265 } 1266 ip6gre_tnl_parm_from_user(&p1, &p); 1267 t = ip6gre_tunnel_locate(net, &p1, 0); 1268 if (!t) 1269 t = netdev_priv(dev); 1270 } 1271 memset(&p, 0, sizeof(p)); 1272 ip6gre_tnl_parm_to_user(&p, &t->parms); 1273 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1274 err = -EFAULT; 1275 break; 1276 1277 case SIOCADDTUNNEL: 1278 case SIOCCHGTUNNEL: 1279 err = -EPERM; 1280 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1281 goto done; 1282 1283 err = -EFAULT; 1284 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1285 goto done; 1286 1287 err = -EINVAL; 1288 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1289 goto done; 1290 1291 if (!(p.i_flags&GRE_KEY)) 1292 p.i_key = 0; 1293 if (!(p.o_flags&GRE_KEY)) 1294 p.o_key = 0; 1295 1296 ip6gre_tnl_parm_from_user(&p1, &p); 1297 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1298 1299 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1300 if (t) { 1301 if (t->dev != dev) { 1302 err = -EEXIST; 1303 break; 1304 } 1305 } else { 1306 t = netdev_priv(dev); 1307 1308 ip6gre_tunnel_unlink(ign, t); 1309 synchronize_net(); 1310 ip6gre_tnl_change(t, &p1, 1); 1311 ip6gre_tunnel_link(ign, t); 1312 netdev_state_change(dev); 1313 } 1314 } 1315 1316 if (t) { 1317 err = 0; 1318 1319 memset(&p, 0, sizeof(p)); 1320 ip6gre_tnl_parm_to_user(&p, &t->parms); 1321 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1322 err = -EFAULT; 1323 } else 1324 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1325 break; 1326 1327 case SIOCDELTUNNEL: 1328 err = -EPERM; 1329 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1330 goto done; 1331 1332 if (dev == ign->fb_tunnel_dev) { 1333 err = -EFAULT; 1334 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1335 goto done; 1336 err = -ENOENT; 1337 ip6gre_tnl_parm_from_user(&p1, &p); 1338 t = ip6gre_tunnel_locate(net, &p1, 0); 1339 if (!t) 1340 goto done; 1341 err = -EPERM; 1342 if (t == netdev_priv(ign->fb_tunnel_dev)) 1343 goto done; 1344 dev = t->dev; 1345 } 1346 unregister_netdevice(dev); 1347 err = 0; 1348 break; 1349 1350 default: 1351 err = -EINVAL; 1352 } 1353 1354 done: 1355 return err; 1356 } 1357 1358 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1359 unsigned short type, const void *daddr, 1360 const void *saddr, unsigned int len) 1361 { 1362 struct ip6_tnl *t = netdev_priv(dev); 1363 struct ipv6hdr *ipv6h; 1364 __be16 *p; 1365 1366 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1367 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1368 t->fl.u.ip6.flowlabel, 1369 true, &t->fl.u.ip6)); 1370 ipv6h->hop_limit = t->parms.hop_limit; 1371 ipv6h->nexthdr = NEXTHDR_GRE; 1372 ipv6h->saddr = t->parms.laddr; 1373 ipv6h->daddr = t->parms.raddr; 1374 1375 p = (__be16 *)(ipv6h + 1); 1376 p[0] = t->parms.o_flags; 1377 p[1] = htons(type); 1378 1379 /* 1380 * Set the source hardware address. 1381 */ 1382 1383 if (saddr) 1384 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1385 if (daddr) 1386 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1387 if (!ipv6_addr_any(&ipv6h->daddr)) 1388 return t->hlen; 1389 1390 return -t->hlen; 1391 } 1392 1393 static const struct header_ops ip6gre_header_ops = { 1394 .create = ip6gre_header, 1395 }; 1396 1397 static const struct net_device_ops ip6gre_netdev_ops = { 1398 .ndo_init = ip6gre_tunnel_init, 1399 .ndo_uninit = ip6gre_tunnel_uninit, 1400 .ndo_start_xmit = ip6gre_tunnel_xmit, 1401 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1402 .ndo_change_mtu = ip6_tnl_change_mtu, 1403 .ndo_get_stats64 = dev_get_tstats64, 1404 .ndo_get_iflink = ip6_tnl_get_iflink, 1405 }; 1406 1407 static void ip6gre_dev_free(struct net_device *dev) 1408 { 1409 struct ip6_tnl *t = netdev_priv(dev); 1410 1411 gro_cells_destroy(&t->gro_cells); 1412 dst_cache_destroy(&t->dst_cache); 1413 free_percpu(dev->tstats); 1414 } 1415 1416 static void ip6gre_tunnel_setup(struct net_device *dev) 1417 { 1418 dev->netdev_ops = &ip6gre_netdev_ops; 1419 dev->needs_free_netdev = true; 1420 dev->priv_destructor = ip6gre_dev_free; 1421 1422 dev->type = ARPHRD_IP6GRE; 1423 1424 dev->flags |= IFF_NOARP; 1425 dev->addr_len = sizeof(struct in6_addr); 1426 netif_keep_dst(dev); 1427 /* This perm addr will be used as interface identifier by IPv6 */ 1428 dev->addr_assign_type = NET_ADDR_RANDOM; 1429 eth_random_addr(dev->perm_addr); 1430 } 1431 1432 #define GRE6_FEATURES (NETIF_F_SG | \ 1433 NETIF_F_FRAGLIST | \ 1434 NETIF_F_HIGHDMA | \ 1435 NETIF_F_HW_CSUM) 1436 1437 static void ip6gre_tnl_init_features(struct net_device *dev) 1438 { 1439 struct ip6_tnl *nt = netdev_priv(dev); 1440 1441 dev->features |= GRE6_FEATURES; 1442 dev->hw_features |= GRE6_FEATURES; 1443 1444 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1445 /* TCP offload with GRE SEQ is not supported, nor 1446 * can we support 2 levels of outer headers requiring 1447 * an update. 1448 */ 1449 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1450 nt->encap.type == TUNNEL_ENCAP_NONE) { 1451 dev->features |= NETIF_F_GSO_SOFTWARE; 1452 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1453 } 1454 1455 /* Can use a lockless transmit, unless we generate 1456 * output sequences 1457 */ 1458 dev->features |= NETIF_F_LLTX; 1459 } 1460 } 1461 1462 static int ip6gre_tunnel_init_common(struct net_device *dev) 1463 { 1464 struct ip6_tnl *tunnel; 1465 int ret; 1466 int t_hlen; 1467 1468 tunnel = netdev_priv(dev); 1469 1470 tunnel->dev = dev; 1471 tunnel->net = dev_net(dev); 1472 strcpy(tunnel->parms.name, dev->name); 1473 1474 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1475 if (!dev->tstats) 1476 return -ENOMEM; 1477 1478 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1479 if (ret) 1480 goto cleanup_alloc_pcpu_stats; 1481 1482 ret = gro_cells_init(&tunnel->gro_cells, dev); 1483 if (ret) 1484 goto cleanup_dst_cache_init; 1485 1486 t_hlen = ip6gre_calc_hlen(tunnel); 1487 dev->mtu = ETH_DATA_LEN - t_hlen; 1488 if (dev->type == ARPHRD_ETHER) 1489 dev->mtu -= ETH_HLEN; 1490 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1491 dev->mtu -= 8; 1492 1493 if (tunnel->parms.collect_md) { 1494 netif_keep_dst(dev); 1495 } 1496 ip6gre_tnl_init_features(dev); 1497 1498 dev_hold(dev); 1499 return 0; 1500 1501 cleanup_dst_cache_init: 1502 dst_cache_destroy(&tunnel->dst_cache); 1503 cleanup_alloc_pcpu_stats: 1504 free_percpu(dev->tstats); 1505 dev->tstats = NULL; 1506 return ret; 1507 } 1508 1509 static int ip6gre_tunnel_init(struct net_device *dev) 1510 { 1511 struct ip6_tnl *tunnel; 1512 int ret; 1513 1514 ret = ip6gre_tunnel_init_common(dev); 1515 if (ret) 1516 return ret; 1517 1518 tunnel = netdev_priv(dev); 1519 1520 if (tunnel->parms.collect_md) 1521 return 0; 1522 1523 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1524 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1525 1526 if (ipv6_addr_any(&tunnel->parms.raddr)) 1527 dev->header_ops = &ip6gre_header_ops; 1528 1529 return 0; 1530 } 1531 1532 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1533 { 1534 struct ip6_tnl *tunnel = netdev_priv(dev); 1535 1536 tunnel->dev = dev; 1537 tunnel->net = dev_net(dev); 1538 strcpy(tunnel->parms.name, dev->name); 1539 1540 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1541 } 1542 1543 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1544 .handler = gre_rcv, 1545 .err_handler = ip6gre_err, 1546 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1547 }; 1548 1549 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1550 { 1551 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1552 struct net_device *dev, *aux; 1553 int prio; 1554 1555 for_each_netdev_safe(net, dev, aux) 1556 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1557 dev->rtnl_link_ops == &ip6gre_tap_ops || 1558 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1559 unregister_netdevice_queue(dev, head); 1560 1561 for (prio = 0; prio < 4; prio++) { 1562 int h; 1563 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1564 struct ip6_tnl *t; 1565 1566 t = rtnl_dereference(ign->tunnels[prio][h]); 1567 1568 while (t) { 1569 /* If dev is in the same netns, it has already 1570 * been added to the list by the previous loop. 1571 */ 1572 if (!net_eq(dev_net(t->dev), net)) 1573 unregister_netdevice_queue(t->dev, 1574 head); 1575 t = rtnl_dereference(t->next); 1576 } 1577 } 1578 } 1579 } 1580 1581 static int __net_init ip6gre_init_net(struct net *net) 1582 { 1583 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1584 struct net_device *ndev; 1585 int err; 1586 1587 if (!net_has_fallback_tunnels(net)) 1588 return 0; 1589 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1590 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1591 if (!ndev) { 1592 err = -ENOMEM; 1593 goto err_alloc_dev; 1594 } 1595 ign->fb_tunnel_dev = ndev; 1596 dev_net_set(ign->fb_tunnel_dev, net); 1597 /* FB netdevice is special: we have one, and only one per netns. 1598 * Allowing to move it to another netns is clearly unsafe. 1599 */ 1600 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1601 1602 1603 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1604 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1605 1606 err = register_netdev(ign->fb_tunnel_dev); 1607 if (err) 1608 goto err_reg_dev; 1609 1610 rcu_assign_pointer(ign->tunnels_wc[0], 1611 netdev_priv(ign->fb_tunnel_dev)); 1612 return 0; 1613 1614 err_reg_dev: 1615 free_netdev(ndev); 1616 err_alloc_dev: 1617 return err; 1618 } 1619 1620 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1621 { 1622 struct net *net; 1623 LIST_HEAD(list); 1624 1625 rtnl_lock(); 1626 list_for_each_entry(net, net_list, exit_list) 1627 ip6gre_destroy_tunnels(net, &list); 1628 unregister_netdevice_many(&list); 1629 rtnl_unlock(); 1630 } 1631 1632 static struct pernet_operations ip6gre_net_ops = { 1633 .init = ip6gre_init_net, 1634 .exit_batch = ip6gre_exit_batch_net, 1635 .id = &ip6gre_net_id, 1636 .size = sizeof(struct ip6gre_net), 1637 }; 1638 1639 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1640 struct netlink_ext_ack *extack) 1641 { 1642 __be16 flags; 1643 1644 if (!data) 1645 return 0; 1646 1647 flags = 0; 1648 if (data[IFLA_GRE_IFLAGS]) 1649 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1650 if (data[IFLA_GRE_OFLAGS]) 1651 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1652 if (flags & (GRE_VERSION|GRE_ROUTING)) 1653 return -EINVAL; 1654 1655 return 0; 1656 } 1657 1658 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1659 struct netlink_ext_ack *extack) 1660 { 1661 struct in6_addr daddr; 1662 1663 if (tb[IFLA_ADDRESS]) { 1664 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1665 return -EINVAL; 1666 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1667 return -EADDRNOTAVAIL; 1668 } 1669 1670 if (!data) 1671 goto out; 1672 1673 if (data[IFLA_GRE_REMOTE]) { 1674 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1675 if (ipv6_addr_any(&daddr)) 1676 return -EINVAL; 1677 } 1678 1679 out: 1680 return ip6gre_tunnel_validate(tb, data, extack); 1681 } 1682 1683 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1684 struct netlink_ext_ack *extack) 1685 { 1686 __be16 flags = 0; 1687 int ret, ver = 0; 1688 1689 if (!data) 1690 return 0; 1691 1692 ret = ip6gre_tap_validate(tb, data, extack); 1693 if (ret) 1694 return ret; 1695 1696 /* ERSPAN should only have GRE sequence and key flag */ 1697 if (data[IFLA_GRE_OFLAGS]) 1698 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1699 if (data[IFLA_GRE_IFLAGS]) 1700 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1701 if (!data[IFLA_GRE_COLLECT_METADATA] && 1702 flags != (GRE_SEQ | GRE_KEY)) 1703 return -EINVAL; 1704 1705 /* ERSPAN Session ID only has 10-bit. Since we reuse 1706 * 32-bit key field as ID, check it's range. 1707 */ 1708 if (data[IFLA_GRE_IKEY] && 1709 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1710 return -EINVAL; 1711 1712 if (data[IFLA_GRE_OKEY] && 1713 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1714 return -EINVAL; 1715 1716 if (data[IFLA_GRE_ERSPAN_VER]) { 1717 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1718 if (ver != 1 && ver != 2) 1719 return -EINVAL; 1720 } 1721 1722 if (ver == 1) { 1723 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1724 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1725 1726 if (index & ~INDEX_MASK) 1727 return -EINVAL; 1728 } 1729 } else if (ver == 2) { 1730 if (data[IFLA_GRE_ERSPAN_DIR]) { 1731 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1732 1733 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1734 return -EINVAL; 1735 } 1736 1737 if (data[IFLA_GRE_ERSPAN_HWID]) { 1738 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1739 1740 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1741 return -EINVAL; 1742 } 1743 } 1744 1745 return 0; 1746 } 1747 1748 static void ip6erspan_set_version(struct nlattr *data[], 1749 struct __ip6_tnl_parm *parms) 1750 { 1751 if (!data) 1752 return; 1753 1754 parms->erspan_ver = 1; 1755 if (data[IFLA_GRE_ERSPAN_VER]) 1756 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1757 1758 if (parms->erspan_ver == 1) { 1759 if (data[IFLA_GRE_ERSPAN_INDEX]) 1760 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1761 } else if (parms->erspan_ver == 2) { 1762 if (data[IFLA_GRE_ERSPAN_DIR]) 1763 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1764 if (data[IFLA_GRE_ERSPAN_HWID]) 1765 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1766 } 1767 } 1768 1769 static void ip6gre_netlink_parms(struct nlattr *data[], 1770 struct __ip6_tnl_parm *parms) 1771 { 1772 memset(parms, 0, sizeof(*parms)); 1773 1774 if (!data) 1775 return; 1776 1777 if (data[IFLA_GRE_LINK]) 1778 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1779 1780 if (data[IFLA_GRE_IFLAGS]) 1781 parms->i_flags = gre_flags_to_tnl_flags( 1782 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1783 1784 if (data[IFLA_GRE_OFLAGS]) 1785 parms->o_flags = gre_flags_to_tnl_flags( 1786 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1787 1788 if (data[IFLA_GRE_IKEY]) 1789 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1790 1791 if (data[IFLA_GRE_OKEY]) 1792 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1793 1794 if (data[IFLA_GRE_LOCAL]) 1795 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1796 1797 if (data[IFLA_GRE_REMOTE]) 1798 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1799 1800 if (data[IFLA_GRE_TTL]) 1801 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1802 1803 if (data[IFLA_GRE_ENCAP_LIMIT]) 1804 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1805 1806 if (data[IFLA_GRE_FLOWINFO]) 1807 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1808 1809 if (data[IFLA_GRE_FLAGS]) 1810 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1811 1812 if (data[IFLA_GRE_FWMARK]) 1813 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1814 1815 if (data[IFLA_GRE_COLLECT_METADATA]) 1816 parms->collect_md = true; 1817 } 1818 1819 static int ip6gre_tap_init(struct net_device *dev) 1820 { 1821 int ret; 1822 1823 ret = ip6gre_tunnel_init_common(dev); 1824 if (ret) 1825 return ret; 1826 1827 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1828 1829 return 0; 1830 } 1831 1832 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1833 .ndo_init = ip6gre_tap_init, 1834 .ndo_uninit = ip6gre_tunnel_uninit, 1835 .ndo_start_xmit = ip6gre_tunnel_xmit, 1836 .ndo_set_mac_address = eth_mac_addr, 1837 .ndo_validate_addr = eth_validate_addr, 1838 .ndo_change_mtu = ip6_tnl_change_mtu, 1839 .ndo_get_stats64 = dev_get_tstats64, 1840 .ndo_get_iflink = ip6_tnl_get_iflink, 1841 }; 1842 1843 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1844 { 1845 int t_hlen; 1846 1847 tunnel->tun_hlen = 8; 1848 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1849 erspan_hdr_len(tunnel->parms.erspan_ver); 1850 1851 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1852 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1853 return t_hlen; 1854 } 1855 1856 static int ip6erspan_tap_init(struct net_device *dev) 1857 { 1858 struct ip6_tnl *tunnel; 1859 int t_hlen; 1860 int ret; 1861 1862 tunnel = netdev_priv(dev); 1863 1864 tunnel->dev = dev; 1865 tunnel->net = dev_net(dev); 1866 strcpy(tunnel->parms.name, dev->name); 1867 1868 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1869 if (!dev->tstats) 1870 return -ENOMEM; 1871 1872 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1873 if (ret) 1874 goto cleanup_alloc_pcpu_stats; 1875 1876 ret = gro_cells_init(&tunnel->gro_cells, dev); 1877 if (ret) 1878 goto cleanup_dst_cache_init; 1879 1880 t_hlen = ip6erspan_calc_hlen(tunnel); 1881 dev->mtu = ETH_DATA_LEN - t_hlen; 1882 if (dev->type == ARPHRD_ETHER) 1883 dev->mtu -= ETH_HLEN; 1884 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1885 dev->mtu -= 8; 1886 1887 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1888 ip6erspan_tnl_link_config(tunnel, 1); 1889 1890 dev_hold(dev); 1891 return 0; 1892 1893 cleanup_dst_cache_init: 1894 dst_cache_destroy(&tunnel->dst_cache); 1895 cleanup_alloc_pcpu_stats: 1896 free_percpu(dev->tstats); 1897 dev->tstats = NULL; 1898 return ret; 1899 } 1900 1901 static const struct net_device_ops ip6erspan_netdev_ops = { 1902 .ndo_init = ip6erspan_tap_init, 1903 .ndo_uninit = ip6erspan_tunnel_uninit, 1904 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1905 .ndo_set_mac_address = eth_mac_addr, 1906 .ndo_validate_addr = eth_validate_addr, 1907 .ndo_change_mtu = ip6_tnl_change_mtu, 1908 .ndo_get_stats64 = dev_get_tstats64, 1909 .ndo_get_iflink = ip6_tnl_get_iflink, 1910 }; 1911 1912 static void ip6gre_tap_setup(struct net_device *dev) 1913 { 1914 1915 ether_setup(dev); 1916 1917 dev->max_mtu = 0; 1918 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1919 dev->needs_free_netdev = true; 1920 dev->priv_destructor = ip6gre_dev_free; 1921 1922 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1923 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1924 netif_keep_dst(dev); 1925 } 1926 1927 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1928 struct ip_tunnel_encap *ipencap) 1929 { 1930 bool ret = false; 1931 1932 memset(ipencap, 0, sizeof(*ipencap)); 1933 1934 if (!data) 1935 return ret; 1936 1937 if (data[IFLA_GRE_ENCAP_TYPE]) { 1938 ret = true; 1939 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1940 } 1941 1942 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1943 ret = true; 1944 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1945 } 1946 1947 if (data[IFLA_GRE_ENCAP_SPORT]) { 1948 ret = true; 1949 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1950 } 1951 1952 if (data[IFLA_GRE_ENCAP_DPORT]) { 1953 ret = true; 1954 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1955 } 1956 1957 return ret; 1958 } 1959 1960 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1961 struct nlattr *tb[], struct nlattr *data[], 1962 struct netlink_ext_ack *extack) 1963 { 1964 struct ip6_tnl *nt; 1965 struct ip_tunnel_encap ipencap; 1966 int err; 1967 1968 nt = netdev_priv(dev); 1969 1970 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1971 int err = ip6_tnl_encap_setup(nt, &ipencap); 1972 1973 if (err < 0) 1974 return err; 1975 } 1976 1977 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1978 eth_hw_addr_random(dev); 1979 1980 nt->dev = dev; 1981 nt->net = dev_net(dev); 1982 1983 err = register_netdevice(dev); 1984 if (err) 1985 goto out; 1986 1987 if (tb[IFLA_MTU]) 1988 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1989 1990 out: 1991 return err; 1992 } 1993 1994 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1995 struct nlattr *tb[], struct nlattr *data[], 1996 struct netlink_ext_ack *extack) 1997 { 1998 struct ip6_tnl *nt = netdev_priv(dev); 1999 struct net *net = dev_net(dev); 2000 struct ip6gre_net *ign; 2001 int err; 2002 2003 ip6gre_netlink_parms(data, &nt->parms); 2004 ign = net_generic(net, ip6gre_net_id); 2005 2006 if (nt->parms.collect_md) { 2007 if (rtnl_dereference(ign->collect_md_tun)) 2008 return -EEXIST; 2009 } else { 2010 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2011 return -EEXIST; 2012 } 2013 2014 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2015 if (!err) { 2016 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2017 ip6gre_tunnel_link_md(ign, nt); 2018 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2019 } 2020 return err; 2021 } 2022 2023 static struct ip6_tnl * 2024 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2025 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2026 struct netlink_ext_ack *extack) 2027 { 2028 struct ip6_tnl *t, *nt = netdev_priv(dev); 2029 struct net *net = nt->net; 2030 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2031 struct ip_tunnel_encap ipencap; 2032 2033 if (dev == ign->fb_tunnel_dev) 2034 return ERR_PTR(-EINVAL); 2035 2036 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2037 int err = ip6_tnl_encap_setup(nt, &ipencap); 2038 2039 if (err < 0) 2040 return ERR_PTR(err); 2041 } 2042 2043 ip6gre_netlink_parms(data, p_p); 2044 2045 t = ip6gre_tunnel_locate(net, p_p, 0); 2046 2047 if (t) { 2048 if (t->dev != dev) 2049 return ERR_PTR(-EEXIST); 2050 } else { 2051 t = nt; 2052 } 2053 2054 return t; 2055 } 2056 2057 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2058 struct nlattr *data[], 2059 struct netlink_ext_ack *extack) 2060 { 2061 struct ip6_tnl *t = netdev_priv(dev); 2062 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2063 struct __ip6_tnl_parm p; 2064 2065 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2066 if (IS_ERR(t)) 2067 return PTR_ERR(t); 2068 2069 ip6gre_tunnel_unlink_md(ign, t); 2070 ip6gre_tunnel_unlink(ign, t); 2071 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2072 ip6gre_tunnel_link_md(ign, t); 2073 ip6gre_tunnel_link(ign, t); 2074 return 0; 2075 } 2076 2077 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2078 { 2079 struct net *net = dev_net(dev); 2080 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2081 2082 if (dev != ign->fb_tunnel_dev) 2083 unregister_netdevice_queue(dev, head); 2084 } 2085 2086 static size_t ip6gre_get_size(const struct net_device *dev) 2087 { 2088 return 2089 /* IFLA_GRE_LINK */ 2090 nla_total_size(4) + 2091 /* IFLA_GRE_IFLAGS */ 2092 nla_total_size(2) + 2093 /* IFLA_GRE_OFLAGS */ 2094 nla_total_size(2) + 2095 /* IFLA_GRE_IKEY */ 2096 nla_total_size(4) + 2097 /* IFLA_GRE_OKEY */ 2098 nla_total_size(4) + 2099 /* IFLA_GRE_LOCAL */ 2100 nla_total_size(sizeof(struct in6_addr)) + 2101 /* IFLA_GRE_REMOTE */ 2102 nla_total_size(sizeof(struct in6_addr)) + 2103 /* IFLA_GRE_TTL */ 2104 nla_total_size(1) + 2105 /* IFLA_GRE_ENCAP_LIMIT */ 2106 nla_total_size(1) + 2107 /* IFLA_GRE_FLOWINFO */ 2108 nla_total_size(4) + 2109 /* IFLA_GRE_FLAGS */ 2110 nla_total_size(4) + 2111 /* IFLA_GRE_ENCAP_TYPE */ 2112 nla_total_size(2) + 2113 /* IFLA_GRE_ENCAP_FLAGS */ 2114 nla_total_size(2) + 2115 /* IFLA_GRE_ENCAP_SPORT */ 2116 nla_total_size(2) + 2117 /* IFLA_GRE_ENCAP_DPORT */ 2118 nla_total_size(2) + 2119 /* IFLA_GRE_COLLECT_METADATA */ 2120 nla_total_size(0) + 2121 /* IFLA_GRE_FWMARK */ 2122 nla_total_size(4) + 2123 /* IFLA_GRE_ERSPAN_INDEX */ 2124 nla_total_size(4) + 2125 0; 2126 } 2127 2128 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2129 { 2130 struct ip6_tnl *t = netdev_priv(dev); 2131 struct __ip6_tnl_parm *p = &t->parms; 2132 __be16 o_flags = p->o_flags; 2133 2134 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2135 if (!p->collect_md) 2136 o_flags |= TUNNEL_KEY; 2137 2138 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2139 goto nla_put_failure; 2140 2141 if (p->erspan_ver == 1) { 2142 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2143 goto nla_put_failure; 2144 } else { 2145 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2146 goto nla_put_failure; 2147 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2148 goto nla_put_failure; 2149 } 2150 } 2151 2152 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2153 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2154 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2155 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2156 gre_tnl_flags_to_gre_flags(o_flags)) || 2157 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2158 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2159 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2160 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2161 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2162 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2163 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2164 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2165 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2166 goto nla_put_failure; 2167 2168 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2169 t->encap.type) || 2170 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2171 t->encap.sport) || 2172 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2173 t->encap.dport) || 2174 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2175 t->encap.flags)) 2176 goto nla_put_failure; 2177 2178 if (p->collect_md) { 2179 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2180 goto nla_put_failure; 2181 } 2182 2183 return 0; 2184 2185 nla_put_failure: 2186 return -EMSGSIZE; 2187 } 2188 2189 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2190 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2191 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2192 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2193 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2194 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2195 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2196 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2197 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2198 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2199 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2200 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2201 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2202 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2203 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2204 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2205 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2206 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2207 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2208 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2209 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2210 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2211 }; 2212 2213 static void ip6erspan_tap_setup(struct net_device *dev) 2214 { 2215 ether_setup(dev); 2216 2217 dev->max_mtu = 0; 2218 dev->netdev_ops = &ip6erspan_netdev_ops; 2219 dev->needs_free_netdev = true; 2220 dev->priv_destructor = ip6gre_dev_free; 2221 2222 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2223 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2224 netif_keep_dst(dev); 2225 } 2226 2227 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2228 struct nlattr *tb[], struct nlattr *data[], 2229 struct netlink_ext_ack *extack) 2230 { 2231 struct ip6_tnl *nt = netdev_priv(dev); 2232 struct net *net = dev_net(dev); 2233 struct ip6gre_net *ign; 2234 int err; 2235 2236 ip6gre_netlink_parms(data, &nt->parms); 2237 ip6erspan_set_version(data, &nt->parms); 2238 ign = net_generic(net, ip6gre_net_id); 2239 2240 if (nt->parms.collect_md) { 2241 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2242 return -EEXIST; 2243 } else { 2244 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2245 return -EEXIST; 2246 } 2247 2248 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2249 if (!err) { 2250 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2251 ip6erspan_tunnel_link_md(ign, nt); 2252 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2253 } 2254 return err; 2255 } 2256 2257 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2258 { 2259 ip6gre_tnl_link_config_common(t); 2260 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2261 } 2262 2263 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2264 const struct __ip6_tnl_parm *p, int set_mtu) 2265 { 2266 ip6gre_tnl_copy_tnl_parm(t, p); 2267 ip6erspan_tnl_link_config(t, set_mtu); 2268 return 0; 2269 } 2270 2271 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2272 struct nlattr *data[], 2273 struct netlink_ext_ack *extack) 2274 { 2275 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2276 struct __ip6_tnl_parm p; 2277 struct ip6_tnl *t; 2278 2279 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2280 if (IS_ERR(t)) 2281 return PTR_ERR(t); 2282 2283 ip6erspan_set_version(data, &p); 2284 ip6gre_tunnel_unlink_md(ign, t); 2285 ip6gre_tunnel_unlink(ign, t); 2286 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2287 ip6erspan_tunnel_link_md(ign, t); 2288 ip6gre_tunnel_link(ign, t); 2289 return 0; 2290 } 2291 2292 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2293 .kind = "ip6gre", 2294 .maxtype = IFLA_GRE_MAX, 2295 .policy = ip6gre_policy, 2296 .priv_size = sizeof(struct ip6_tnl), 2297 .setup = ip6gre_tunnel_setup, 2298 .validate = ip6gre_tunnel_validate, 2299 .newlink = ip6gre_newlink, 2300 .changelink = ip6gre_changelink, 2301 .dellink = ip6gre_dellink, 2302 .get_size = ip6gre_get_size, 2303 .fill_info = ip6gre_fill_info, 2304 .get_link_net = ip6_tnl_get_link_net, 2305 }; 2306 2307 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2308 .kind = "ip6gretap", 2309 .maxtype = IFLA_GRE_MAX, 2310 .policy = ip6gre_policy, 2311 .priv_size = sizeof(struct ip6_tnl), 2312 .setup = ip6gre_tap_setup, 2313 .validate = ip6gre_tap_validate, 2314 .newlink = ip6gre_newlink, 2315 .changelink = ip6gre_changelink, 2316 .get_size = ip6gre_get_size, 2317 .fill_info = ip6gre_fill_info, 2318 .get_link_net = ip6_tnl_get_link_net, 2319 }; 2320 2321 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2322 .kind = "ip6erspan", 2323 .maxtype = IFLA_GRE_MAX, 2324 .policy = ip6gre_policy, 2325 .priv_size = sizeof(struct ip6_tnl), 2326 .setup = ip6erspan_tap_setup, 2327 .validate = ip6erspan_tap_validate, 2328 .newlink = ip6erspan_newlink, 2329 .changelink = ip6erspan_changelink, 2330 .get_size = ip6gre_get_size, 2331 .fill_info = ip6gre_fill_info, 2332 .get_link_net = ip6_tnl_get_link_net, 2333 }; 2334 2335 /* 2336 * And now the modules code and kernel interface. 2337 */ 2338 2339 static int __init ip6gre_init(void) 2340 { 2341 int err; 2342 2343 pr_info("GRE over IPv6 tunneling driver\n"); 2344 2345 err = register_pernet_device(&ip6gre_net_ops); 2346 if (err < 0) 2347 return err; 2348 2349 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2350 if (err < 0) { 2351 pr_info("%s: can't add protocol\n", __func__); 2352 goto add_proto_failed; 2353 } 2354 2355 err = rtnl_link_register(&ip6gre_link_ops); 2356 if (err < 0) 2357 goto rtnl_link_failed; 2358 2359 err = rtnl_link_register(&ip6gre_tap_ops); 2360 if (err < 0) 2361 goto tap_ops_failed; 2362 2363 err = rtnl_link_register(&ip6erspan_tap_ops); 2364 if (err < 0) 2365 goto erspan_link_failed; 2366 2367 out: 2368 return err; 2369 2370 erspan_link_failed: 2371 rtnl_link_unregister(&ip6gre_tap_ops); 2372 tap_ops_failed: 2373 rtnl_link_unregister(&ip6gre_link_ops); 2374 rtnl_link_failed: 2375 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2376 add_proto_failed: 2377 unregister_pernet_device(&ip6gre_net_ops); 2378 goto out; 2379 } 2380 2381 static void __exit ip6gre_fini(void) 2382 { 2383 rtnl_link_unregister(&ip6gre_tap_ops); 2384 rtnl_link_unregister(&ip6gre_link_ops); 2385 rtnl_link_unregister(&ip6erspan_tap_ops); 2386 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2387 unregister_pernet_device(&ip6gre_net_ops); 2388 } 2389 2390 module_init(ip6gre_init); 2391 module_exit(ip6gre_fini); 2392 MODULE_LICENSE("GPL"); 2393 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2394 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2395 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2396 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2397 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2398 MODULE_ALIAS_NETDEV("ip6gre0"); 2399