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; 130 131 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 132 if (!ipv6_addr_equal(local, &t->parms.laddr) || 133 !ipv6_addr_equal(remote, &t->parms.raddr) || 134 key != t->parms.i_key || 135 !(t->dev->flags & IFF_UP)) 136 continue; 137 138 if (t->dev->type != ARPHRD_IP6GRE && 139 t->dev->type != dev_type) 140 continue; 141 142 score = 0; 143 if (t->parms.link != link) 144 score |= 1; 145 if (t->dev->type != dev_type) 146 score |= 2; 147 if (score == 0) 148 return t; 149 150 if (score < cand_score) { 151 cand = t; 152 cand_score = score; 153 } 154 } 155 156 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 157 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 158 key != t->parms.i_key || 159 !(t->dev->flags & IFF_UP)) 160 continue; 161 162 if (t->dev->type != ARPHRD_IP6GRE && 163 t->dev->type != dev_type) 164 continue; 165 166 score = 0; 167 if (t->parms.link != link) 168 score |= 1; 169 if (t->dev->type != dev_type) 170 score |= 2; 171 if (score == 0) 172 return t; 173 174 if (score < cand_score) { 175 cand = t; 176 cand_score = score; 177 } 178 } 179 180 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 181 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 182 (!ipv6_addr_equal(local, &t->parms.raddr) || 183 !ipv6_addr_is_multicast(local))) || 184 key != t->parms.i_key || 185 !(t->dev->flags & IFF_UP)) 186 continue; 187 188 if (t->dev->type != ARPHRD_IP6GRE && 189 t->dev->type != dev_type) 190 continue; 191 192 score = 0; 193 if (t->parms.link != link) 194 score |= 1; 195 if (t->dev->type != dev_type) 196 score |= 2; 197 if (score == 0) 198 return t; 199 200 if (score < cand_score) { 201 cand = t; 202 cand_score = score; 203 } 204 } 205 206 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 207 if (t->parms.i_key != key || 208 !(t->dev->flags & IFF_UP)) 209 continue; 210 211 if (t->dev->type != ARPHRD_IP6GRE && 212 t->dev->type != dev_type) 213 continue; 214 215 score = 0; 216 if (t->parms.link != link) 217 score |= 1; 218 if (t->dev->type != dev_type) 219 score |= 2; 220 if (score == 0) 221 return t; 222 223 if (score < cand_score) { 224 cand = t; 225 cand_score = score; 226 } 227 } 228 229 if (cand) 230 return cand; 231 232 if (gre_proto == htons(ETH_P_ERSPAN) || 233 gre_proto == htons(ETH_P_ERSPAN2)) 234 t = rcu_dereference(ign->collect_md_tun_erspan); 235 else 236 t = rcu_dereference(ign->collect_md_tun); 237 238 if (t && t->dev->flags & IFF_UP) 239 return t; 240 241 dev = ign->fb_tunnel_dev; 242 if (dev && dev->flags & IFF_UP) 243 return netdev_priv(dev); 244 245 return NULL; 246 } 247 248 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 249 const struct __ip6_tnl_parm *p) 250 { 251 const struct in6_addr *remote = &p->raddr; 252 const struct in6_addr *local = &p->laddr; 253 unsigned int h = HASH_KEY(p->i_key); 254 int prio = 0; 255 256 if (!ipv6_addr_any(local)) 257 prio |= 1; 258 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 259 prio |= 2; 260 h ^= HASH_ADDR(remote); 261 } 262 263 return &ign->tunnels[prio][h]; 264 } 265 266 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 267 { 268 if (t->parms.collect_md) 269 rcu_assign_pointer(ign->collect_md_tun, t); 270 } 271 272 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 273 { 274 if (t->parms.collect_md) 275 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 276 } 277 278 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 279 { 280 if (t->parms.collect_md) 281 rcu_assign_pointer(ign->collect_md_tun, NULL); 282 } 283 284 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 285 struct ip6_tnl *t) 286 { 287 if (t->parms.collect_md) 288 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 289 } 290 291 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 292 const struct ip6_tnl *t) 293 { 294 return __ip6gre_bucket(ign, &t->parms); 295 } 296 297 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 298 { 299 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 300 301 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 302 rcu_assign_pointer(*tp, t); 303 } 304 305 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 306 { 307 struct ip6_tnl __rcu **tp; 308 struct ip6_tnl *iter; 309 310 for (tp = ip6gre_bucket(ign, t); 311 (iter = rtnl_dereference(*tp)) != NULL; 312 tp = &iter->next) { 313 if (t == iter) { 314 rcu_assign_pointer(*tp, t->next); 315 break; 316 } 317 } 318 } 319 320 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 321 const struct __ip6_tnl_parm *parms, 322 int type) 323 { 324 const struct in6_addr *remote = &parms->raddr; 325 const struct in6_addr *local = &parms->laddr; 326 __be32 key = parms->i_key; 327 int link = parms->link; 328 struct ip6_tnl *t; 329 struct ip6_tnl __rcu **tp; 330 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 331 332 for (tp = __ip6gre_bucket(ign, parms); 333 (t = rtnl_dereference(*tp)) != NULL; 334 tp = &t->next) 335 if (ipv6_addr_equal(local, &t->parms.laddr) && 336 ipv6_addr_equal(remote, &t->parms.raddr) && 337 key == t->parms.i_key && 338 link == t->parms.link && 339 type == t->dev->type) 340 break; 341 342 return t; 343 } 344 345 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 346 const struct __ip6_tnl_parm *parms, int create) 347 { 348 struct ip6_tnl *t, *nt; 349 struct net_device *dev; 350 char name[IFNAMSIZ]; 351 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 352 353 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 354 if (t && create) 355 return NULL; 356 if (t || !create) 357 return t; 358 359 if (parms->name[0]) { 360 if (!dev_valid_name(parms->name)) 361 return NULL; 362 strlcpy(name, parms->name, IFNAMSIZ); 363 } else { 364 strcpy(name, "ip6gre%d"); 365 } 366 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 367 ip6gre_tunnel_setup); 368 if (!dev) 369 return NULL; 370 371 dev_net_set(dev, net); 372 373 nt = netdev_priv(dev); 374 nt->parms = *parms; 375 dev->rtnl_link_ops = &ip6gre_link_ops; 376 377 nt->dev = dev; 378 nt->net = dev_net(dev); 379 380 if (register_netdevice(dev) < 0) 381 goto failed_free; 382 383 ip6gre_tnl_link_config(nt, 1); 384 385 /* Can use a lockless transmit, unless we generate output sequences */ 386 if (!(nt->parms.o_flags & TUNNEL_SEQ)) 387 dev->features |= NETIF_F_LLTX; 388 389 dev_hold(dev); 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 dst_cache_reset(&t->dst_cache); 417 dev_put(dev); 418 } 419 420 421 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 422 u8 type, u8 code, int offset, __be32 info) 423 { 424 struct net *net = dev_net(skb->dev); 425 const struct ipv6hdr *ipv6h; 426 struct tnl_ptk_info tpi; 427 struct ip6_tnl *t; 428 429 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 430 offset) < 0) 431 return -EINVAL; 432 433 ipv6h = (const struct ipv6hdr *)skb->data; 434 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 435 tpi.key, tpi.proto); 436 if (!t) 437 return -ENOENT; 438 439 switch (type) { 440 struct ipv6_tlv_tnl_enc_lim *tel; 441 __u32 teli; 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 teli = 0; 457 if (code == ICMPV6_HDR_FIELD) 458 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 459 460 if (teli && teli == be32_to_cpu(info) - 2) { 461 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 462 if (tel->encap_limit == 0) { 463 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 464 t->parms.name); 465 } 466 } else { 467 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 468 t->parms.name); 469 } 470 return 0; 471 case ICMPV6_PKT_TOOBIG: 472 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 473 return 0; 474 case NDISC_REDIRECT: 475 ip6_redirect(skb, net, skb->dev->ifindex, 0, 476 sock_net_uid(net, NULL)); 477 return 0; 478 } 479 480 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 481 t->err_count++; 482 else 483 t->err_count = 1; 484 t->err_time = jiffies; 485 486 return 0; 487 } 488 489 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 490 { 491 const struct ipv6hdr *ipv6h; 492 struct ip6_tnl *tunnel; 493 494 ipv6h = ipv6_hdr(skb); 495 tunnel = ip6gre_tunnel_lookup(skb->dev, 496 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 497 tpi->proto); 498 if (tunnel) { 499 if (tunnel->parms.collect_md) { 500 struct metadata_dst *tun_dst; 501 __be64 tun_id; 502 __be16 flags; 503 504 flags = tpi->flags; 505 tun_id = key32_to_tunnel_id(tpi->key); 506 507 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 508 if (!tun_dst) 509 return PACKET_REJECT; 510 511 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 512 } else { 513 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 514 } 515 516 return PACKET_RCVD; 517 } 518 519 return PACKET_REJECT; 520 } 521 522 static int ip6erspan_rcv(struct sk_buff *skb, 523 struct tnl_ptk_info *tpi, 524 int gre_hdr_len) 525 { 526 struct erspan_base_hdr *ershdr; 527 const struct ipv6hdr *ipv6h; 528 struct erspan_md2 *md2; 529 struct ip6_tnl *tunnel; 530 u8 ver; 531 532 ipv6h = ipv6_hdr(skb); 533 ershdr = (struct erspan_base_hdr *)skb->data; 534 ver = ershdr->ver; 535 536 tunnel = ip6gre_tunnel_lookup(skb->dev, 537 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 538 tpi->proto); 539 if (tunnel) { 540 int len = erspan_hdr_len(ver); 541 542 if (unlikely(!pskb_may_pull(skb, len))) 543 return PACKET_REJECT; 544 545 if (__iptunnel_pull_header(skb, len, 546 htons(ETH_P_TEB), 547 false, false) < 0) 548 return PACKET_REJECT; 549 550 if (tunnel->parms.collect_md) { 551 struct erspan_metadata *pkt_md, *md; 552 struct metadata_dst *tun_dst; 553 struct ip_tunnel_info *info; 554 unsigned char *gh; 555 __be64 tun_id; 556 __be16 flags; 557 558 tpi->flags |= TUNNEL_KEY; 559 flags = tpi->flags; 560 tun_id = key32_to_tunnel_id(tpi->key); 561 562 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 563 sizeof(*md)); 564 if (!tun_dst) 565 return PACKET_REJECT; 566 567 /* skb can be uncloned in __iptunnel_pull_header, so 568 * old pkt_md is no longer valid and we need to reset 569 * it 570 */ 571 gh = skb_network_header(skb) + 572 skb_network_header_len(skb); 573 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 574 sizeof(*ershdr)); 575 info = &tun_dst->u.tun_info; 576 md = ip_tunnel_info_opts(info); 577 md->version = ver; 578 md2 = &md->u.md2; 579 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 580 ERSPAN_V2_MDSIZE); 581 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 582 info->options_len = sizeof(*md); 583 584 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 585 586 } else { 587 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 588 } 589 590 return PACKET_RCVD; 591 } 592 593 return PACKET_REJECT; 594 } 595 596 static int gre_rcv(struct sk_buff *skb) 597 { 598 struct tnl_ptk_info tpi; 599 bool csum_err = false; 600 int hdr_len; 601 602 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 603 if (hdr_len < 0) 604 goto drop; 605 606 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 607 goto drop; 608 609 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 610 tpi.proto == htons(ETH_P_ERSPAN2))) { 611 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) 612 return 0; 613 goto out; 614 } 615 616 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 617 return 0; 618 619 out: 620 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 621 drop: 622 kfree_skb(skb); 623 return 0; 624 } 625 626 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 627 { 628 return iptunnel_handle_offloads(skb, 629 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 630 } 631 632 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 633 struct net_device *dev, 634 struct flowi6 *fl6, __u8 *dsfield, 635 int *encap_limit) 636 { 637 const struct iphdr *iph = ip_hdr(skb); 638 struct ip6_tnl *t = netdev_priv(dev); 639 640 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 641 *encap_limit = t->parms.encap_limit; 642 643 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 644 645 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 646 *dsfield = ipv4_get_dsfield(iph); 647 else 648 *dsfield = ip6_tclass(t->parms.flowinfo); 649 650 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 651 fl6->flowi6_mark = skb->mark; 652 else 653 fl6->flowi6_mark = t->parms.fwmark; 654 655 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 656 } 657 658 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 659 struct net_device *dev, 660 struct flowi6 *fl6, __u8 *dsfield, 661 int *encap_limit) 662 { 663 struct ipv6hdr *ipv6h; 664 struct ip6_tnl *t = netdev_priv(dev); 665 __u16 offset; 666 667 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 668 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 669 ipv6h = ipv6_hdr(skb); 670 671 if (offset > 0) { 672 struct ipv6_tlv_tnl_enc_lim *tel; 673 674 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 675 if (tel->encap_limit == 0) { 676 icmpv6_send(skb, ICMPV6_PARAMPROB, 677 ICMPV6_HDR_FIELD, offset + 2); 678 return -1; 679 } 680 *encap_limit = tel->encap_limit - 1; 681 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 682 *encap_limit = t->parms.encap_limit; 683 } 684 685 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 686 687 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 688 *dsfield = ipv6_get_dsfield(ipv6h); 689 else 690 *dsfield = ip6_tclass(t->parms.flowinfo); 691 692 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 693 fl6->flowlabel |= ip6_flowlabel(ipv6h); 694 695 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 696 fl6->flowi6_mark = skb->mark; 697 else 698 fl6->flowi6_mark = t->parms.fwmark; 699 700 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 701 702 return 0; 703 } 704 705 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 706 struct net_device *dev, __u8 dsfield, 707 struct flowi6 *fl6, int encap_limit, 708 __u32 *pmtu, __be16 proto) 709 { 710 struct ip6_tnl *tunnel = netdev_priv(dev); 711 __be16 protocol; 712 713 if (dev->type == ARPHRD_ETHER) 714 IPCB(skb)->flags = 0; 715 716 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 717 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 718 else 719 fl6->daddr = tunnel->parms.raddr; 720 721 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 722 return -ENOMEM; 723 724 /* Push GRE header. */ 725 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 726 727 if (tunnel->parms.collect_md) { 728 struct ip_tunnel_info *tun_info; 729 const struct ip_tunnel_key *key; 730 __be16 flags; 731 732 tun_info = skb_tunnel_info(skb); 733 if (unlikely(!tun_info || 734 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 735 ip_tunnel_info_af(tun_info) != AF_INET6)) 736 return -EINVAL; 737 738 key = &tun_info->key; 739 memset(fl6, 0, sizeof(*fl6)); 740 fl6->flowi6_proto = IPPROTO_GRE; 741 fl6->daddr = key->u.ipv6.dst; 742 fl6->flowlabel = key->label; 743 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 744 745 dsfield = key->tos; 746 flags = key->tun_flags & 747 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 748 tunnel->tun_hlen = gre_calc_hlen(flags); 749 750 gre_build_header(skb, tunnel->tun_hlen, 751 flags, protocol, 752 tunnel_id_to_key32(tun_info->key.tun_id), 753 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 754 : 0); 755 756 } else { 757 if (tunnel->parms.o_flags & TUNNEL_SEQ) 758 tunnel->o_seqno++; 759 760 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 761 protocol, tunnel->parms.o_key, 762 htonl(tunnel->o_seqno)); 763 } 764 765 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 766 NEXTHDR_GRE); 767 } 768 769 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 770 { 771 struct ip6_tnl *t = netdev_priv(dev); 772 int encap_limit = -1; 773 struct flowi6 fl6; 774 __u8 dsfield = 0; 775 __u32 mtu; 776 int err; 777 778 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 779 780 if (!t->parms.collect_md) 781 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 782 &dsfield, &encap_limit); 783 784 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 785 if (err) 786 return -1; 787 788 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 789 skb->protocol); 790 if (err != 0) { 791 /* XXX: send ICMP error even if DF is not set. */ 792 if (err == -EMSGSIZE) 793 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 794 htonl(mtu)); 795 return -1; 796 } 797 798 return 0; 799 } 800 801 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 802 { 803 struct ip6_tnl *t = netdev_priv(dev); 804 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 805 int encap_limit = -1; 806 struct flowi6 fl6; 807 __u8 dsfield = 0; 808 __u32 mtu; 809 int err; 810 811 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 812 return -1; 813 814 if (!t->parms.collect_md && 815 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 816 return -1; 817 818 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 819 return -1; 820 821 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 822 &mtu, skb->protocol); 823 if (err != 0) { 824 if (err == -EMSGSIZE) 825 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 826 return -1; 827 } 828 829 return 0; 830 } 831 832 /** 833 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 834 * @t: the outgoing tunnel device 835 * @hdr: IPv6 header from the incoming packet 836 * 837 * Description: 838 * Avoid trivial tunneling loop by checking that tunnel exit-point 839 * doesn't match source of incoming packet. 840 * 841 * Return: 842 * 1 if conflict, 843 * 0 else 844 **/ 845 846 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 847 const struct ipv6hdr *hdr) 848 { 849 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 850 } 851 852 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 853 { 854 struct ip6_tnl *t = netdev_priv(dev); 855 int encap_limit = -1; 856 struct flowi6 fl6; 857 __u32 mtu; 858 int err; 859 860 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 861 encap_limit = t->parms.encap_limit; 862 863 if (!t->parms.collect_md) 864 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 865 866 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 867 if (err) 868 return err; 869 870 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 871 872 return err; 873 } 874 875 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 876 struct net_device *dev) 877 { 878 struct ip6_tnl *t = netdev_priv(dev); 879 struct net_device_stats *stats = &t->dev->stats; 880 int ret; 881 882 if (!pskb_inet_may_pull(skb)) 883 goto tx_err; 884 885 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 886 goto tx_err; 887 888 switch (skb->protocol) { 889 case htons(ETH_P_IP): 890 ret = ip6gre_xmit_ipv4(skb, dev); 891 break; 892 case htons(ETH_P_IPV6): 893 ret = ip6gre_xmit_ipv6(skb, dev); 894 break; 895 default: 896 ret = ip6gre_xmit_other(skb, dev); 897 break; 898 } 899 900 if (ret < 0) 901 goto tx_err; 902 903 return NETDEV_TX_OK; 904 905 tx_err: 906 stats->tx_errors++; 907 stats->tx_dropped++; 908 kfree_skb(skb); 909 return NETDEV_TX_OK; 910 } 911 912 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 913 struct net_device *dev) 914 { 915 struct ip6_tnl *t = netdev_priv(dev); 916 struct dst_entry *dst = skb_dst(skb); 917 struct net_device_stats *stats; 918 bool truncate = false; 919 int encap_limit = -1; 920 __u8 dsfield = false; 921 struct flowi6 fl6; 922 int err = -EINVAL; 923 __be16 proto; 924 __u32 mtu; 925 int nhoff; 926 int thoff; 927 928 if (!pskb_inet_may_pull(skb)) 929 goto tx_err; 930 931 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 932 goto tx_err; 933 934 if (gre_handle_offloads(skb, false)) 935 goto tx_err; 936 937 if (skb->len > dev->mtu + dev->hard_header_len) { 938 pskb_trim(skb, dev->mtu + dev->hard_header_len); 939 truncate = true; 940 } 941 942 nhoff = skb_network_header(skb) - skb_mac_header(skb); 943 if (skb->protocol == htons(ETH_P_IP) && 944 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 945 truncate = true; 946 947 thoff = skb_transport_header(skb) - skb_mac_header(skb); 948 if (skb->protocol == htons(ETH_P_IPV6) && 949 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 950 truncate = true; 951 952 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 953 goto tx_err; 954 955 t->parms.o_flags &= ~TUNNEL_KEY; 956 IPCB(skb)->flags = 0; 957 958 /* For collect_md mode, derive fl6 from the tunnel key, 959 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 960 */ 961 if (t->parms.collect_md) { 962 struct ip_tunnel_info *tun_info; 963 const struct ip_tunnel_key *key; 964 struct erspan_metadata *md; 965 __be32 tun_id; 966 967 tun_info = skb_tunnel_info(skb); 968 if (unlikely(!tun_info || 969 !(tun_info->mode & IP_TUNNEL_INFO_TX) || 970 ip_tunnel_info_af(tun_info) != AF_INET6)) 971 goto tx_err; 972 973 key = &tun_info->key; 974 memset(&fl6, 0, sizeof(fl6)); 975 fl6.flowi6_proto = IPPROTO_GRE; 976 fl6.daddr = key->u.ipv6.dst; 977 fl6.flowlabel = key->label; 978 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 979 980 dsfield = key->tos; 981 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 982 goto tx_err; 983 if (tun_info->options_len < sizeof(*md)) 984 goto tx_err; 985 md = ip_tunnel_info_opts(tun_info); 986 987 tun_id = tunnel_id_to_key32(key->tun_id); 988 if (md->version == 1) { 989 erspan_build_header(skb, 990 ntohl(tun_id), 991 ntohl(md->u.index), truncate, 992 false); 993 } else if (md->version == 2) { 994 erspan_build_header_v2(skb, 995 ntohl(tun_id), 996 md->u.md2.dir, 997 get_hwid(&md->u.md2), 998 truncate, false); 999 } else { 1000 goto tx_err; 1001 } 1002 } else { 1003 switch (skb->protocol) { 1004 case htons(ETH_P_IP): 1005 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1006 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1007 &dsfield, &encap_limit); 1008 break; 1009 case htons(ETH_P_IPV6): 1010 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1011 goto tx_err; 1012 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1013 &dsfield, &encap_limit)) 1014 goto tx_err; 1015 break; 1016 default: 1017 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1018 break; 1019 } 1020 1021 if (t->parms.erspan_ver == 1) 1022 erspan_build_header(skb, ntohl(t->parms.o_key), 1023 t->parms.index, 1024 truncate, false); 1025 else if (t->parms.erspan_ver == 2) 1026 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1027 t->parms.dir, 1028 t->parms.hwid, 1029 truncate, false); 1030 else 1031 goto tx_err; 1032 1033 fl6.daddr = t->parms.raddr; 1034 } 1035 1036 /* Push GRE header. */ 1037 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1038 : htons(ETH_P_ERSPAN2); 1039 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++)); 1040 1041 /* TooBig packet may have updated dst->dev's mtu */ 1042 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1043 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1044 1045 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1046 NEXTHDR_GRE); 1047 if (err != 0) { 1048 /* XXX: send ICMP error even if DF is not set. */ 1049 if (err == -EMSGSIZE) { 1050 if (skb->protocol == htons(ETH_P_IP)) 1051 icmp_send(skb, ICMP_DEST_UNREACH, 1052 ICMP_FRAG_NEEDED, htonl(mtu)); 1053 else 1054 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1055 } 1056 1057 goto tx_err; 1058 } 1059 return NETDEV_TX_OK; 1060 1061 tx_err: 1062 stats = &t->dev->stats; 1063 stats->tx_errors++; 1064 stats->tx_dropped++; 1065 kfree_skb(skb); 1066 return NETDEV_TX_OK; 1067 } 1068 1069 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1070 { 1071 struct net_device *dev = t->dev; 1072 struct __ip6_tnl_parm *p = &t->parms; 1073 struct flowi6 *fl6 = &t->fl.u.ip6; 1074 1075 if (dev->type != ARPHRD_ETHER) { 1076 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 1077 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1078 } 1079 1080 /* Set up flowi template */ 1081 fl6->saddr = p->laddr; 1082 fl6->daddr = p->raddr; 1083 fl6->flowi6_oif = p->link; 1084 fl6->flowlabel = 0; 1085 fl6->flowi6_proto = IPPROTO_GRE; 1086 1087 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1088 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1089 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1090 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1091 1092 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1093 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1094 1095 if (p->flags&IP6_TNL_F_CAP_XMIT && 1096 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1097 dev->flags |= IFF_POINTOPOINT; 1098 else 1099 dev->flags &= ~IFF_POINTOPOINT; 1100 } 1101 1102 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1103 int t_hlen) 1104 { 1105 const struct __ip6_tnl_parm *p = &t->parms; 1106 struct net_device *dev = t->dev; 1107 1108 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1109 int strict = (ipv6_addr_type(&p->raddr) & 1110 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1111 1112 struct rt6_info *rt = rt6_lookup(t->net, 1113 &p->raddr, &p->laddr, 1114 p->link, NULL, strict); 1115 1116 if (!rt) 1117 return; 1118 1119 if (rt->dst.dev) { 1120 dev->needed_headroom = rt->dst.dev->hard_header_len + 1121 t_hlen; 1122 1123 if (set_mtu) { 1124 dev->mtu = rt->dst.dev->mtu - t_hlen; 1125 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1126 dev->mtu -= 8; 1127 if (dev->type == ARPHRD_ETHER) 1128 dev->mtu -= ETH_HLEN; 1129 1130 if (dev->mtu < IPV6_MIN_MTU) 1131 dev->mtu = IPV6_MIN_MTU; 1132 } 1133 } 1134 ip6_rt_put(rt); 1135 } 1136 } 1137 1138 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1139 { 1140 int t_hlen; 1141 1142 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1143 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1144 1145 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1146 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1147 return t_hlen; 1148 } 1149 1150 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1151 { 1152 ip6gre_tnl_link_config_common(t); 1153 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1154 } 1155 1156 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1157 const struct __ip6_tnl_parm *p) 1158 { 1159 t->parms.laddr = p->laddr; 1160 t->parms.raddr = p->raddr; 1161 t->parms.flags = p->flags; 1162 t->parms.hop_limit = p->hop_limit; 1163 t->parms.encap_limit = p->encap_limit; 1164 t->parms.flowinfo = p->flowinfo; 1165 t->parms.link = p->link; 1166 t->parms.proto = p->proto; 1167 t->parms.i_key = p->i_key; 1168 t->parms.o_key = p->o_key; 1169 t->parms.i_flags = p->i_flags; 1170 t->parms.o_flags = p->o_flags; 1171 t->parms.fwmark = p->fwmark; 1172 t->parms.erspan_ver = p->erspan_ver; 1173 t->parms.index = p->index; 1174 t->parms.dir = p->dir; 1175 t->parms.hwid = p->hwid; 1176 dst_cache_reset(&t->dst_cache); 1177 } 1178 1179 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1180 int set_mtu) 1181 { 1182 ip6gre_tnl_copy_tnl_parm(t, p); 1183 ip6gre_tnl_link_config(t, set_mtu); 1184 return 0; 1185 } 1186 1187 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1188 const struct ip6_tnl_parm2 *u) 1189 { 1190 p->laddr = u->laddr; 1191 p->raddr = u->raddr; 1192 p->flags = u->flags; 1193 p->hop_limit = u->hop_limit; 1194 p->encap_limit = u->encap_limit; 1195 p->flowinfo = u->flowinfo; 1196 p->link = u->link; 1197 p->i_key = u->i_key; 1198 p->o_key = u->o_key; 1199 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1200 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1201 memcpy(p->name, u->name, sizeof(u->name)); 1202 } 1203 1204 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1205 const struct __ip6_tnl_parm *p) 1206 { 1207 u->proto = IPPROTO_GRE; 1208 u->laddr = p->laddr; 1209 u->raddr = p->raddr; 1210 u->flags = p->flags; 1211 u->hop_limit = p->hop_limit; 1212 u->encap_limit = p->encap_limit; 1213 u->flowinfo = p->flowinfo; 1214 u->link = p->link; 1215 u->i_key = p->i_key; 1216 u->o_key = p->o_key; 1217 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1218 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1219 memcpy(u->name, p->name, sizeof(u->name)); 1220 } 1221 1222 static int ip6gre_tunnel_ioctl(struct net_device *dev, 1223 struct ifreq *ifr, int cmd) 1224 { 1225 int err = 0; 1226 struct ip6_tnl_parm2 p; 1227 struct __ip6_tnl_parm p1; 1228 struct ip6_tnl *t = netdev_priv(dev); 1229 struct net *net = t->net; 1230 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1231 1232 memset(&p1, 0, sizeof(p1)); 1233 1234 switch (cmd) { 1235 case SIOCGETTUNNEL: 1236 if (dev == ign->fb_tunnel_dev) { 1237 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 1238 err = -EFAULT; 1239 break; 1240 } 1241 ip6gre_tnl_parm_from_user(&p1, &p); 1242 t = ip6gre_tunnel_locate(net, &p1, 0); 1243 if (!t) 1244 t = netdev_priv(dev); 1245 } 1246 memset(&p, 0, sizeof(p)); 1247 ip6gre_tnl_parm_to_user(&p, &t->parms); 1248 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1249 err = -EFAULT; 1250 break; 1251 1252 case SIOCADDTUNNEL: 1253 case SIOCCHGTUNNEL: 1254 err = -EPERM; 1255 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1256 goto done; 1257 1258 err = -EFAULT; 1259 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1260 goto done; 1261 1262 err = -EINVAL; 1263 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1264 goto done; 1265 1266 if (!(p.i_flags&GRE_KEY)) 1267 p.i_key = 0; 1268 if (!(p.o_flags&GRE_KEY)) 1269 p.o_key = 0; 1270 1271 ip6gre_tnl_parm_from_user(&p1, &p); 1272 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1273 1274 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1275 if (t) { 1276 if (t->dev != dev) { 1277 err = -EEXIST; 1278 break; 1279 } 1280 } else { 1281 t = netdev_priv(dev); 1282 1283 ip6gre_tunnel_unlink(ign, t); 1284 synchronize_net(); 1285 ip6gre_tnl_change(t, &p1, 1); 1286 ip6gre_tunnel_link(ign, t); 1287 netdev_state_change(dev); 1288 } 1289 } 1290 1291 if (t) { 1292 err = 0; 1293 1294 memset(&p, 0, sizeof(p)); 1295 ip6gre_tnl_parm_to_user(&p, &t->parms); 1296 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 1297 err = -EFAULT; 1298 } else 1299 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1300 break; 1301 1302 case SIOCDELTUNNEL: 1303 err = -EPERM; 1304 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1305 goto done; 1306 1307 if (dev == ign->fb_tunnel_dev) { 1308 err = -EFAULT; 1309 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 1310 goto done; 1311 err = -ENOENT; 1312 ip6gre_tnl_parm_from_user(&p1, &p); 1313 t = ip6gre_tunnel_locate(net, &p1, 0); 1314 if (!t) 1315 goto done; 1316 err = -EPERM; 1317 if (t == netdev_priv(ign->fb_tunnel_dev)) 1318 goto done; 1319 dev = t->dev; 1320 } 1321 unregister_netdevice(dev); 1322 err = 0; 1323 break; 1324 1325 default: 1326 err = -EINVAL; 1327 } 1328 1329 done: 1330 return err; 1331 } 1332 1333 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1334 unsigned short type, const void *daddr, 1335 const void *saddr, unsigned int len) 1336 { 1337 struct ip6_tnl *t = netdev_priv(dev); 1338 struct ipv6hdr *ipv6h; 1339 __be16 *p; 1340 1341 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1342 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1343 t->fl.u.ip6.flowlabel, 1344 true, &t->fl.u.ip6)); 1345 ipv6h->hop_limit = t->parms.hop_limit; 1346 ipv6h->nexthdr = NEXTHDR_GRE; 1347 ipv6h->saddr = t->parms.laddr; 1348 ipv6h->daddr = t->parms.raddr; 1349 1350 p = (__be16 *)(ipv6h + 1); 1351 p[0] = t->parms.o_flags; 1352 p[1] = htons(type); 1353 1354 /* 1355 * Set the source hardware address. 1356 */ 1357 1358 if (saddr) 1359 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1360 if (daddr) 1361 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1362 if (!ipv6_addr_any(&ipv6h->daddr)) 1363 return t->hlen; 1364 1365 return -t->hlen; 1366 } 1367 1368 static const struct header_ops ip6gre_header_ops = { 1369 .create = ip6gre_header, 1370 }; 1371 1372 static const struct net_device_ops ip6gre_netdev_ops = { 1373 .ndo_init = ip6gre_tunnel_init, 1374 .ndo_uninit = ip6gre_tunnel_uninit, 1375 .ndo_start_xmit = ip6gre_tunnel_xmit, 1376 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 1377 .ndo_change_mtu = ip6_tnl_change_mtu, 1378 .ndo_get_stats64 = ip_tunnel_get_stats64, 1379 .ndo_get_iflink = ip6_tnl_get_iflink, 1380 }; 1381 1382 static void ip6gre_dev_free(struct net_device *dev) 1383 { 1384 struct ip6_tnl *t = netdev_priv(dev); 1385 1386 gro_cells_destroy(&t->gro_cells); 1387 dst_cache_destroy(&t->dst_cache); 1388 free_percpu(dev->tstats); 1389 } 1390 1391 static void ip6gre_tunnel_setup(struct net_device *dev) 1392 { 1393 dev->netdev_ops = &ip6gre_netdev_ops; 1394 dev->needs_free_netdev = true; 1395 dev->priv_destructor = ip6gre_dev_free; 1396 1397 dev->type = ARPHRD_IP6GRE; 1398 1399 dev->flags |= IFF_NOARP; 1400 dev->addr_len = sizeof(struct in6_addr); 1401 netif_keep_dst(dev); 1402 /* This perm addr will be used as interface identifier by IPv6 */ 1403 dev->addr_assign_type = NET_ADDR_RANDOM; 1404 eth_random_addr(dev->perm_addr); 1405 } 1406 1407 #define GRE6_FEATURES (NETIF_F_SG | \ 1408 NETIF_F_FRAGLIST | \ 1409 NETIF_F_HIGHDMA | \ 1410 NETIF_F_HW_CSUM) 1411 1412 static void ip6gre_tnl_init_features(struct net_device *dev) 1413 { 1414 struct ip6_tnl *nt = netdev_priv(dev); 1415 1416 dev->features |= GRE6_FEATURES; 1417 dev->hw_features |= GRE6_FEATURES; 1418 1419 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1420 /* TCP offload with GRE SEQ is not supported, nor 1421 * can we support 2 levels of outer headers requiring 1422 * an update. 1423 */ 1424 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1425 nt->encap.type == TUNNEL_ENCAP_NONE) { 1426 dev->features |= NETIF_F_GSO_SOFTWARE; 1427 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1428 } 1429 1430 /* Can use a lockless transmit, unless we generate 1431 * output sequences 1432 */ 1433 dev->features |= NETIF_F_LLTX; 1434 } 1435 } 1436 1437 static int ip6gre_tunnel_init_common(struct net_device *dev) 1438 { 1439 struct ip6_tnl *tunnel; 1440 int ret; 1441 int t_hlen; 1442 1443 tunnel = netdev_priv(dev); 1444 1445 tunnel->dev = dev; 1446 tunnel->net = dev_net(dev); 1447 strcpy(tunnel->parms.name, dev->name); 1448 1449 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1450 if (!dev->tstats) 1451 return -ENOMEM; 1452 1453 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1454 if (ret) 1455 goto cleanup_alloc_pcpu_stats; 1456 1457 ret = gro_cells_init(&tunnel->gro_cells, dev); 1458 if (ret) 1459 goto cleanup_dst_cache_init; 1460 1461 t_hlen = ip6gre_calc_hlen(tunnel); 1462 dev->mtu = ETH_DATA_LEN - t_hlen; 1463 if (dev->type == ARPHRD_ETHER) 1464 dev->mtu -= ETH_HLEN; 1465 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1466 dev->mtu -= 8; 1467 1468 if (tunnel->parms.collect_md) { 1469 netif_keep_dst(dev); 1470 } 1471 ip6gre_tnl_init_features(dev); 1472 1473 return 0; 1474 1475 cleanup_dst_cache_init: 1476 dst_cache_destroy(&tunnel->dst_cache); 1477 cleanup_alloc_pcpu_stats: 1478 free_percpu(dev->tstats); 1479 dev->tstats = NULL; 1480 return ret; 1481 } 1482 1483 static int ip6gre_tunnel_init(struct net_device *dev) 1484 { 1485 struct ip6_tnl *tunnel; 1486 int ret; 1487 1488 ret = ip6gre_tunnel_init_common(dev); 1489 if (ret) 1490 return ret; 1491 1492 tunnel = netdev_priv(dev); 1493 1494 if (tunnel->parms.collect_md) 1495 return 0; 1496 1497 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1498 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1499 1500 if (ipv6_addr_any(&tunnel->parms.raddr)) 1501 dev->header_ops = &ip6gre_header_ops; 1502 1503 return 0; 1504 } 1505 1506 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1507 { 1508 struct ip6_tnl *tunnel = netdev_priv(dev); 1509 1510 tunnel->dev = dev; 1511 tunnel->net = dev_net(dev); 1512 strcpy(tunnel->parms.name, dev->name); 1513 1514 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1515 1516 dev_hold(dev); 1517 } 1518 1519 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1520 .handler = gre_rcv, 1521 .err_handler = ip6gre_err, 1522 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1523 }; 1524 1525 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1526 { 1527 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1528 struct net_device *dev, *aux; 1529 int prio; 1530 1531 for_each_netdev_safe(net, dev, aux) 1532 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1533 dev->rtnl_link_ops == &ip6gre_tap_ops || 1534 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1535 unregister_netdevice_queue(dev, head); 1536 1537 for (prio = 0; prio < 4; prio++) { 1538 int h; 1539 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1540 struct ip6_tnl *t; 1541 1542 t = rtnl_dereference(ign->tunnels[prio][h]); 1543 1544 while (t) { 1545 /* If dev is in the same netns, it has already 1546 * been added to the list by the previous loop. 1547 */ 1548 if (!net_eq(dev_net(t->dev), net)) 1549 unregister_netdevice_queue(t->dev, 1550 head); 1551 t = rtnl_dereference(t->next); 1552 } 1553 } 1554 } 1555 } 1556 1557 static int __net_init ip6gre_init_net(struct net *net) 1558 { 1559 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1560 int err; 1561 1562 if (!net_has_fallback_tunnels(net)) 1563 return 0; 1564 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1565 NET_NAME_UNKNOWN, 1566 ip6gre_tunnel_setup); 1567 if (!ign->fb_tunnel_dev) { 1568 err = -ENOMEM; 1569 goto err_alloc_dev; 1570 } 1571 dev_net_set(ign->fb_tunnel_dev, net); 1572 /* FB netdevice is special: we have one, and only one per netns. 1573 * Allowing to move it to another netns is clearly unsafe. 1574 */ 1575 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1576 1577 1578 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1579 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1580 1581 err = register_netdev(ign->fb_tunnel_dev); 1582 if (err) 1583 goto err_reg_dev; 1584 1585 rcu_assign_pointer(ign->tunnels_wc[0], 1586 netdev_priv(ign->fb_tunnel_dev)); 1587 return 0; 1588 1589 err_reg_dev: 1590 free_netdev(ign->fb_tunnel_dev); 1591 err_alloc_dev: 1592 return err; 1593 } 1594 1595 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1596 { 1597 struct net *net; 1598 LIST_HEAD(list); 1599 1600 rtnl_lock(); 1601 list_for_each_entry(net, net_list, exit_list) 1602 ip6gre_destroy_tunnels(net, &list); 1603 unregister_netdevice_many(&list); 1604 rtnl_unlock(); 1605 } 1606 1607 static struct pernet_operations ip6gre_net_ops = { 1608 .init = ip6gre_init_net, 1609 .exit_batch = ip6gre_exit_batch_net, 1610 .id = &ip6gre_net_id, 1611 .size = sizeof(struct ip6gre_net), 1612 }; 1613 1614 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1615 struct netlink_ext_ack *extack) 1616 { 1617 __be16 flags; 1618 1619 if (!data) 1620 return 0; 1621 1622 flags = 0; 1623 if (data[IFLA_GRE_IFLAGS]) 1624 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1625 if (data[IFLA_GRE_OFLAGS]) 1626 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1627 if (flags & (GRE_VERSION|GRE_ROUTING)) 1628 return -EINVAL; 1629 1630 return 0; 1631 } 1632 1633 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1634 struct netlink_ext_ack *extack) 1635 { 1636 struct in6_addr daddr; 1637 1638 if (tb[IFLA_ADDRESS]) { 1639 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1640 return -EINVAL; 1641 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1642 return -EADDRNOTAVAIL; 1643 } 1644 1645 if (!data) 1646 goto out; 1647 1648 if (data[IFLA_GRE_REMOTE]) { 1649 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1650 if (ipv6_addr_any(&daddr)) 1651 return -EINVAL; 1652 } 1653 1654 out: 1655 return ip6gre_tunnel_validate(tb, data, extack); 1656 } 1657 1658 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1659 struct netlink_ext_ack *extack) 1660 { 1661 __be16 flags = 0; 1662 int ret, ver = 0; 1663 1664 if (!data) 1665 return 0; 1666 1667 ret = ip6gre_tap_validate(tb, data, extack); 1668 if (ret) 1669 return ret; 1670 1671 /* ERSPAN should only have GRE sequence and key flag */ 1672 if (data[IFLA_GRE_OFLAGS]) 1673 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1674 if (data[IFLA_GRE_IFLAGS]) 1675 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1676 if (!data[IFLA_GRE_COLLECT_METADATA] && 1677 flags != (GRE_SEQ | GRE_KEY)) 1678 return -EINVAL; 1679 1680 /* ERSPAN Session ID only has 10-bit. Since we reuse 1681 * 32-bit key field as ID, check it's range. 1682 */ 1683 if (data[IFLA_GRE_IKEY] && 1684 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1685 return -EINVAL; 1686 1687 if (data[IFLA_GRE_OKEY] && 1688 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1689 return -EINVAL; 1690 1691 if (data[IFLA_GRE_ERSPAN_VER]) { 1692 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1693 if (ver != 1 && ver != 2) 1694 return -EINVAL; 1695 } 1696 1697 if (ver == 1) { 1698 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1699 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1700 1701 if (index & ~INDEX_MASK) 1702 return -EINVAL; 1703 } 1704 } else if (ver == 2) { 1705 if (data[IFLA_GRE_ERSPAN_DIR]) { 1706 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1707 1708 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1709 return -EINVAL; 1710 } 1711 1712 if (data[IFLA_GRE_ERSPAN_HWID]) { 1713 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1714 1715 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1716 return -EINVAL; 1717 } 1718 } 1719 1720 return 0; 1721 } 1722 1723 static void ip6erspan_set_version(struct nlattr *data[], 1724 struct __ip6_tnl_parm *parms) 1725 { 1726 if (!data) 1727 return; 1728 1729 parms->erspan_ver = 1; 1730 if (data[IFLA_GRE_ERSPAN_VER]) 1731 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1732 1733 if (parms->erspan_ver == 1) { 1734 if (data[IFLA_GRE_ERSPAN_INDEX]) 1735 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1736 } else if (parms->erspan_ver == 2) { 1737 if (data[IFLA_GRE_ERSPAN_DIR]) 1738 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1739 if (data[IFLA_GRE_ERSPAN_HWID]) 1740 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1741 } 1742 } 1743 1744 static void ip6gre_netlink_parms(struct nlattr *data[], 1745 struct __ip6_tnl_parm *parms) 1746 { 1747 memset(parms, 0, sizeof(*parms)); 1748 1749 if (!data) 1750 return; 1751 1752 if (data[IFLA_GRE_LINK]) 1753 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1754 1755 if (data[IFLA_GRE_IFLAGS]) 1756 parms->i_flags = gre_flags_to_tnl_flags( 1757 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1758 1759 if (data[IFLA_GRE_OFLAGS]) 1760 parms->o_flags = gre_flags_to_tnl_flags( 1761 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1762 1763 if (data[IFLA_GRE_IKEY]) 1764 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1765 1766 if (data[IFLA_GRE_OKEY]) 1767 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1768 1769 if (data[IFLA_GRE_LOCAL]) 1770 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1771 1772 if (data[IFLA_GRE_REMOTE]) 1773 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1774 1775 if (data[IFLA_GRE_TTL]) 1776 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1777 1778 if (data[IFLA_GRE_ENCAP_LIMIT]) 1779 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1780 1781 if (data[IFLA_GRE_FLOWINFO]) 1782 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1783 1784 if (data[IFLA_GRE_FLAGS]) 1785 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1786 1787 if (data[IFLA_GRE_FWMARK]) 1788 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1789 1790 if (data[IFLA_GRE_COLLECT_METADATA]) 1791 parms->collect_md = true; 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->needed_headroom = 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->priv_flags &= ~IFF_TX_SKB_SHARING; 1897 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1898 netif_keep_dst(dev); 1899 } 1900 1901 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1902 struct ip_tunnel_encap *ipencap) 1903 { 1904 bool ret = false; 1905 1906 memset(ipencap, 0, sizeof(*ipencap)); 1907 1908 if (!data) 1909 return ret; 1910 1911 if (data[IFLA_GRE_ENCAP_TYPE]) { 1912 ret = true; 1913 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1914 } 1915 1916 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1917 ret = true; 1918 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1919 } 1920 1921 if (data[IFLA_GRE_ENCAP_SPORT]) { 1922 ret = true; 1923 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1924 } 1925 1926 if (data[IFLA_GRE_ENCAP_DPORT]) { 1927 ret = true; 1928 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1929 } 1930 1931 return ret; 1932 } 1933 1934 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1935 struct nlattr *tb[], struct nlattr *data[], 1936 struct netlink_ext_ack *extack) 1937 { 1938 struct ip6_tnl *nt; 1939 struct ip_tunnel_encap ipencap; 1940 int err; 1941 1942 nt = netdev_priv(dev); 1943 1944 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1945 int err = ip6_tnl_encap_setup(nt, &ipencap); 1946 1947 if (err < 0) 1948 return err; 1949 } 1950 1951 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1952 eth_hw_addr_random(dev); 1953 1954 nt->dev = dev; 1955 nt->net = dev_net(dev); 1956 1957 err = register_netdevice(dev); 1958 if (err) 1959 goto out; 1960 1961 if (tb[IFLA_MTU]) 1962 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1963 1964 dev_hold(dev); 1965 1966 out: 1967 return err; 1968 } 1969 1970 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1971 struct nlattr *tb[], struct nlattr *data[], 1972 struct netlink_ext_ack *extack) 1973 { 1974 struct ip6_tnl *nt = netdev_priv(dev); 1975 struct net *net = dev_net(dev); 1976 struct ip6gre_net *ign; 1977 int err; 1978 1979 ip6gre_netlink_parms(data, &nt->parms); 1980 ign = net_generic(net, ip6gre_net_id); 1981 1982 if (nt->parms.collect_md) { 1983 if (rtnl_dereference(ign->collect_md_tun)) 1984 return -EEXIST; 1985 } else { 1986 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1987 return -EEXIST; 1988 } 1989 1990 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 1991 if (!err) { 1992 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1993 ip6gre_tunnel_link_md(ign, nt); 1994 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 1995 } 1996 return err; 1997 } 1998 1999 static struct ip6_tnl * 2000 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2001 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2002 struct netlink_ext_ack *extack) 2003 { 2004 struct ip6_tnl *t, *nt = netdev_priv(dev); 2005 struct net *net = nt->net; 2006 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2007 struct ip_tunnel_encap ipencap; 2008 2009 if (dev == ign->fb_tunnel_dev) 2010 return ERR_PTR(-EINVAL); 2011 2012 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2013 int err = ip6_tnl_encap_setup(nt, &ipencap); 2014 2015 if (err < 0) 2016 return ERR_PTR(err); 2017 } 2018 2019 ip6gre_netlink_parms(data, p_p); 2020 2021 t = ip6gre_tunnel_locate(net, p_p, 0); 2022 2023 if (t) { 2024 if (t->dev != dev) 2025 return ERR_PTR(-EEXIST); 2026 } else { 2027 t = nt; 2028 } 2029 2030 return t; 2031 } 2032 2033 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2034 struct nlattr *data[], 2035 struct netlink_ext_ack *extack) 2036 { 2037 struct ip6_tnl *t = netdev_priv(dev); 2038 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2039 struct __ip6_tnl_parm p; 2040 2041 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2042 if (IS_ERR(t)) 2043 return PTR_ERR(t); 2044 2045 ip6gre_tunnel_unlink_md(ign, t); 2046 ip6gre_tunnel_unlink(ign, t); 2047 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2048 ip6gre_tunnel_link_md(ign, t); 2049 ip6gre_tunnel_link(ign, t); 2050 return 0; 2051 } 2052 2053 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2054 { 2055 struct net *net = dev_net(dev); 2056 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2057 2058 if (dev != ign->fb_tunnel_dev) 2059 unregister_netdevice_queue(dev, head); 2060 } 2061 2062 static size_t ip6gre_get_size(const struct net_device *dev) 2063 { 2064 return 2065 /* IFLA_GRE_LINK */ 2066 nla_total_size(4) + 2067 /* IFLA_GRE_IFLAGS */ 2068 nla_total_size(2) + 2069 /* IFLA_GRE_OFLAGS */ 2070 nla_total_size(2) + 2071 /* IFLA_GRE_IKEY */ 2072 nla_total_size(4) + 2073 /* IFLA_GRE_OKEY */ 2074 nla_total_size(4) + 2075 /* IFLA_GRE_LOCAL */ 2076 nla_total_size(sizeof(struct in6_addr)) + 2077 /* IFLA_GRE_REMOTE */ 2078 nla_total_size(sizeof(struct in6_addr)) + 2079 /* IFLA_GRE_TTL */ 2080 nla_total_size(1) + 2081 /* IFLA_GRE_ENCAP_LIMIT */ 2082 nla_total_size(1) + 2083 /* IFLA_GRE_FLOWINFO */ 2084 nla_total_size(4) + 2085 /* IFLA_GRE_FLAGS */ 2086 nla_total_size(4) + 2087 /* IFLA_GRE_ENCAP_TYPE */ 2088 nla_total_size(2) + 2089 /* IFLA_GRE_ENCAP_FLAGS */ 2090 nla_total_size(2) + 2091 /* IFLA_GRE_ENCAP_SPORT */ 2092 nla_total_size(2) + 2093 /* IFLA_GRE_ENCAP_DPORT */ 2094 nla_total_size(2) + 2095 /* IFLA_GRE_COLLECT_METADATA */ 2096 nla_total_size(0) + 2097 /* IFLA_GRE_FWMARK */ 2098 nla_total_size(4) + 2099 /* IFLA_GRE_ERSPAN_INDEX */ 2100 nla_total_size(4) + 2101 0; 2102 } 2103 2104 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2105 { 2106 struct ip6_tnl *t = netdev_priv(dev); 2107 struct __ip6_tnl_parm *p = &t->parms; 2108 __be16 o_flags = p->o_flags; 2109 2110 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2111 if (!p->collect_md) 2112 o_flags |= TUNNEL_KEY; 2113 2114 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2115 goto nla_put_failure; 2116 2117 if (p->erspan_ver == 1) { 2118 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2119 goto nla_put_failure; 2120 } else { 2121 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2122 goto nla_put_failure; 2123 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2124 goto nla_put_failure; 2125 } 2126 } 2127 2128 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2129 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2130 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2131 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2132 gre_tnl_flags_to_gre_flags(o_flags)) || 2133 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2134 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2135 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2136 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2137 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2138 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2139 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2140 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2141 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2142 goto nla_put_failure; 2143 2144 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2145 t->encap.type) || 2146 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2147 t->encap.sport) || 2148 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2149 t->encap.dport) || 2150 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2151 t->encap.flags)) 2152 goto nla_put_failure; 2153 2154 if (p->collect_md) { 2155 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2156 goto nla_put_failure; 2157 } 2158 2159 return 0; 2160 2161 nla_put_failure: 2162 return -EMSGSIZE; 2163 } 2164 2165 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2166 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2167 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2168 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2169 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2170 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2171 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2172 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2173 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2174 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2175 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2176 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2177 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2178 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2179 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2180 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2181 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2182 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2183 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2184 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2185 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2186 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2187 }; 2188 2189 static void ip6erspan_tap_setup(struct net_device *dev) 2190 { 2191 ether_setup(dev); 2192 2193 dev->max_mtu = 0; 2194 dev->netdev_ops = &ip6erspan_netdev_ops; 2195 dev->needs_free_netdev = true; 2196 dev->priv_destructor = ip6gre_dev_free; 2197 2198 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2199 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2200 netif_keep_dst(dev); 2201 } 2202 2203 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2204 struct nlattr *tb[], struct nlattr *data[], 2205 struct netlink_ext_ack *extack) 2206 { 2207 struct ip6_tnl *nt = netdev_priv(dev); 2208 struct net *net = dev_net(dev); 2209 struct ip6gre_net *ign; 2210 int err; 2211 2212 ip6gre_netlink_parms(data, &nt->parms); 2213 ip6erspan_set_version(data, &nt->parms); 2214 ign = net_generic(net, ip6gre_net_id); 2215 2216 if (nt->parms.collect_md) { 2217 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2218 return -EEXIST; 2219 } else { 2220 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2221 return -EEXIST; 2222 } 2223 2224 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2225 if (!err) { 2226 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2227 ip6erspan_tunnel_link_md(ign, nt); 2228 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2229 } 2230 return err; 2231 } 2232 2233 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2234 { 2235 ip6gre_tnl_link_config_common(t); 2236 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2237 } 2238 2239 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2240 const struct __ip6_tnl_parm *p, int set_mtu) 2241 { 2242 ip6gre_tnl_copy_tnl_parm(t, p); 2243 ip6erspan_tnl_link_config(t, set_mtu); 2244 return 0; 2245 } 2246 2247 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2248 struct nlattr *data[], 2249 struct netlink_ext_ack *extack) 2250 { 2251 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2252 struct __ip6_tnl_parm p; 2253 struct ip6_tnl *t; 2254 2255 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2256 if (IS_ERR(t)) 2257 return PTR_ERR(t); 2258 2259 ip6erspan_set_version(data, &p); 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