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_track(dev, &t->dev_tracker); 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_track(dev, &t->dev_tracker); 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 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id); 759 760 dsfield = key->tos; 761 flags = key->tun_flags & 762 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 763 tunnel->tun_hlen = gre_calc_hlen(flags); 764 765 gre_build_header(skb, tunnel->tun_hlen, 766 flags, protocol, 767 tunnel_id_to_key32(tun_info->key.tun_id), 768 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++) 769 : 0); 770 771 } else { 772 if (tunnel->parms.o_flags & TUNNEL_SEQ) 773 tunnel->o_seqno++; 774 775 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 776 protocol, tunnel->parms.o_key, 777 htonl(tunnel->o_seqno)); 778 } 779 780 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 781 NEXTHDR_GRE); 782 } 783 784 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 785 { 786 struct ip6_tnl *t = netdev_priv(dev); 787 int encap_limit = -1; 788 struct flowi6 fl6; 789 __u8 dsfield = 0; 790 __u32 mtu; 791 int err; 792 793 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 794 795 if (!t->parms.collect_md) 796 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 797 &dsfield, &encap_limit); 798 799 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 800 if (err) 801 return -1; 802 803 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 804 skb->protocol); 805 if (err != 0) { 806 /* XXX: send ICMP error even if DF is not set. */ 807 if (err == -EMSGSIZE) 808 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 809 htonl(mtu)); 810 return -1; 811 } 812 813 return 0; 814 } 815 816 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 817 { 818 struct ip6_tnl *t = netdev_priv(dev); 819 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 820 int encap_limit = -1; 821 struct flowi6 fl6; 822 __u8 dsfield = 0; 823 __u32 mtu; 824 int err; 825 826 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 827 return -1; 828 829 if (!t->parms.collect_md && 830 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 831 return -1; 832 833 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 834 return -1; 835 836 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 837 &mtu, skb->protocol); 838 if (err != 0) { 839 if (err == -EMSGSIZE) 840 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 841 return -1; 842 } 843 844 return 0; 845 } 846 847 /** 848 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 849 * @t: the outgoing tunnel device 850 * @hdr: IPv6 header from the incoming packet 851 * 852 * Description: 853 * Avoid trivial tunneling loop by checking that tunnel exit-point 854 * doesn't match source of incoming packet. 855 * 856 * Return: 857 * 1 if conflict, 858 * 0 else 859 **/ 860 861 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 862 const struct ipv6hdr *hdr) 863 { 864 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 865 } 866 867 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 868 { 869 struct ip6_tnl *t = netdev_priv(dev); 870 int encap_limit = -1; 871 struct flowi6 fl6; 872 __u32 mtu; 873 int err; 874 875 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 876 encap_limit = t->parms.encap_limit; 877 878 if (!t->parms.collect_md) 879 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 880 881 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 882 if (err) 883 return err; 884 885 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 886 887 return err; 888 } 889 890 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 891 struct net_device *dev) 892 { 893 struct ip6_tnl *t = netdev_priv(dev); 894 struct net_device_stats *stats = &t->dev->stats; 895 int ret; 896 897 if (!pskb_inet_may_pull(skb)) 898 goto tx_err; 899 900 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 901 goto tx_err; 902 903 switch (skb->protocol) { 904 case htons(ETH_P_IP): 905 ret = ip6gre_xmit_ipv4(skb, dev); 906 break; 907 case htons(ETH_P_IPV6): 908 ret = ip6gre_xmit_ipv6(skb, dev); 909 break; 910 default: 911 ret = ip6gre_xmit_other(skb, dev); 912 break; 913 } 914 915 if (ret < 0) 916 goto tx_err; 917 918 return NETDEV_TX_OK; 919 920 tx_err: 921 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 922 stats->tx_errors++; 923 stats->tx_dropped++; 924 kfree_skb(skb); 925 return NETDEV_TX_OK; 926 } 927 928 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 929 struct net_device *dev) 930 { 931 struct ip_tunnel_info *tun_info = NULL; 932 struct ip6_tnl *t = netdev_priv(dev); 933 struct dst_entry *dst = skb_dst(skb); 934 struct net_device_stats *stats; 935 bool truncate = false; 936 int encap_limit = -1; 937 __u8 dsfield = false; 938 struct flowi6 fl6; 939 int err = -EINVAL; 940 __be16 proto; 941 __u32 mtu; 942 int nhoff; 943 int thoff; 944 945 if (!pskb_inet_may_pull(skb)) 946 goto tx_err; 947 948 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 949 goto tx_err; 950 951 if (gre_handle_offloads(skb, false)) 952 goto tx_err; 953 954 if (skb->len > dev->mtu + dev->hard_header_len) { 955 pskb_trim(skb, dev->mtu + dev->hard_header_len); 956 truncate = true; 957 } 958 959 nhoff = skb_network_header(skb) - skb_mac_header(skb); 960 if (skb->protocol == htons(ETH_P_IP) && 961 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 962 truncate = true; 963 964 thoff = skb_transport_header(skb) - skb_mac_header(skb); 965 if (skb->protocol == htons(ETH_P_IPV6) && 966 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 967 truncate = true; 968 969 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 970 goto tx_err; 971 972 t->parms.o_flags &= ~TUNNEL_KEY; 973 IPCB(skb)->flags = 0; 974 975 /* For collect_md mode, derive fl6 from the tunnel key, 976 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 977 */ 978 if (t->parms.collect_md) { 979 const struct ip_tunnel_key *key; 980 struct erspan_metadata *md; 981 __be32 tun_id; 982 983 tun_info = skb_tunnel_info_txcheck(skb); 984 if (IS_ERR(tun_info) || 985 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 986 goto tx_err; 987 988 key = &tun_info->key; 989 memset(&fl6, 0, sizeof(fl6)); 990 fl6.flowi6_proto = IPPROTO_GRE; 991 fl6.daddr = key->u.ipv6.dst; 992 fl6.flowlabel = key->label; 993 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 994 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 995 996 dsfield = key->tos; 997 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 998 goto tx_err; 999 if (tun_info->options_len < sizeof(*md)) 1000 goto tx_err; 1001 md = ip_tunnel_info_opts(tun_info); 1002 1003 tun_id = tunnel_id_to_key32(key->tun_id); 1004 if (md->version == 1) { 1005 erspan_build_header(skb, 1006 ntohl(tun_id), 1007 ntohl(md->u.index), truncate, 1008 false); 1009 } else if (md->version == 2) { 1010 erspan_build_header_v2(skb, 1011 ntohl(tun_id), 1012 md->u.md2.dir, 1013 get_hwid(&md->u.md2), 1014 truncate, false); 1015 } else { 1016 goto tx_err; 1017 } 1018 } else { 1019 switch (skb->protocol) { 1020 case htons(ETH_P_IP): 1021 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1022 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1023 &dsfield, &encap_limit); 1024 break; 1025 case htons(ETH_P_IPV6): 1026 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1027 goto tx_err; 1028 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1029 &dsfield, &encap_limit)) 1030 goto tx_err; 1031 break; 1032 default: 1033 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1034 break; 1035 } 1036 1037 if (t->parms.erspan_ver == 1) 1038 erspan_build_header(skb, ntohl(t->parms.o_key), 1039 t->parms.index, 1040 truncate, false); 1041 else if (t->parms.erspan_ver == 2) 1042 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1043 t->parms.dir, 1044 t->parms.hwid, 1045 truncate, false); 1046 else 1047 goto tx_err; 1048 1049 fl6.daddr = t->parms.raddr; 1050 } 1051 1052 /* Push GRE header. */ 1053 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1054 : htons(ETH_P_ERSPAN2); 1055 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(t->o_seqno++)); 1056 1057 /* TooBig packet may have updated dst->dev's mtu */ 1058 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1059 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1060 1061 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1062 NEXTHDR_GRE); 1063 if (err != 0) { 1064 /* XXX: send ICMP error even if DF is not set. */ 1065 if (err == -EMSGSIZE) { 1066 if (skb->protocol == htons(ETH_P_IP)) 1067 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1068 ICMP_FRAG_NEEDED, htonl(mtu)); 1069 else 1070 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1071 } 1072 1073 goto tx_err; 1074 } 1075 return NETDEV_TX_OK; 1076 1077 tx_err: 1078 stats = &t->dev->stats; 1079 if (!IS_ERR(tun_info)) 1080 stats->tx_errors++; 1081 stats->tx_dropped++; 1082 kfree_skb(skb); 1083 return NETDEV_TX_OK; 1084 } 1085 1086 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1087 { 1088 struct net_device *dev = t->dev; 1089 struct __ip6_tnl_parm *p = &t->parms; 1090 struct flowi6 *fl6 = &t->fl.u.ip6; 1091 1092 if (dev->type != ARPHRD_ETHER) { 1093 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1094 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1095 } 1096 1097 /* Set up flowi template */ 1098 fl6->saddr = p->laddr; 1099 fl6->daddr = p->raddr; 1100 fl6->flowi6_oif = p->link; 1101 fl6->flowlabel = 0; 1102 fl6->flowi6_proto = IPPROTO_GRE; 1103 fl6->fl6_gre_key = t->parms.o_key; 1104 1105 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1106 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1107 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1108 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1109 1110 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1111 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1112 1113 if (p->flags&IP6_TNL_F_CAP_XMIT && 1114 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1115 dev->flags |= IFF_POINTOPOINT; 1116 else 1117 dev->flags &= ~IFF_POINTOPOINT; 1118 } 1119 1120 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1121 int t_hlen) 1122 { 1123 const struct __ip6_tnl_parm *p = &t->parms; 1124 struct net_device *dev = t->dev; 1125 1126 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1127 int strict = (ipv6_addr_type(&p->raddr) & 1128 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1129 1130 struct rt6_info *rt = rt6_lookup(t->net, 1131 &p->raddr, &p->laddr, 1132 p->link, NULL, strict); 1133 1134 if (!rt) 1135 return; 1136 1137 if (rt->dst.dev) { 1138 unsigned short dst_len = rt->dst.dev->hard_header_len + 1139 t_hlen; 1140 1141 if (t->dev->header_ops) 1142 dev->hard_header_len = dst_len; 1143 else 1144 dev->needed_headroom = dst_len; 1145 1146 if (set_mtu) { 1147 dev->mtu = rt->dst.dev->mtu - t_hlen; 1148 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1149 dev->mtu -= 8; 1150 if (dev->type == ARPHRD_ETHER) 1151 dev->mtu -= ETH_HLEN; 1152 1153 if (dev->mtu < IPV6_MIN_MTU) 1154 dev->mtu = IPV6_MIN_MTU; 1155 } 1156 } 1157 ip6_rt_put(rt); 1158 } 1159 } 1160 1161 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1162 { 1163 int t_hlen; 1164 1165 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1166 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1167 1168 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1169 1170 if (tunnel->dev->header_ops) 1171 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1172 else 1173 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1174 1175 return t_hlen; 1176 } 1177 1178 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1179 { 1180 ip6gre_tnl_link_config_common(t); 1181 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1182 } 1183 1184 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1185 const struct __ip6_tnl_parm *p) 1186 { 1187 t->parms.laddr = p->laddr; 1188 t->parms.raddr = p->raddr; 1189 t->parms.flags = p->flags; 1190 t->parms.hop_limit = p->hop_limit; 1191 t->parms.encap_limit = p->encap_limit; 1192 t->parms.flowinfo = p->flowinfo; 1193 t->parms.link = p->link; 1194 t->parms.proto = p->proto; 1195 t->parms.i_key = p->i_key; 1196 t->parms.o_key = p->o_key; 1197 t->parms.i_flags = p->i_flags; 1198 t->parms.o_flags = p->o_flags; 1199 t->parms.fwmark = p->fwmark; 1200 t->parms.erspan_ver = p->erspan_ver; 1201 t->parms.index = p->index; 1202 t->parms.dir = p->dir; 1203 t->parms.hwid = p->hwid; 1204 dst_cache_reset(&t->dst_cache); 1205 } 1206 1207 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1208 int set_mtu) 1209 { 1210 ip6gre_tnl_copy_tnl_parm(t, p); 1211 ip6gre_tnl_link_config(t, set_mtu); 1212 return 0; 1213 } 1214 1215 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1216 const struct ip6_tnl_parm2 *u) 1217 { 1218 p->laddr = u->laddr; 1219 p->raddr = u->raddr; 1220 p->flags = u->flags; 1221 p->hop_limit = u->hop_limit; 1222 p->encap_limit = u->encap_limit; 1223 p->flowinfo = u->flowinfo; 1224 p->link = u->link; 1225 p->i_key = u->i_key; 1226 p->o_key = u->o_key; 1227 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1228 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1229 memcpy(p->name, u->name, sizeof(u->name)); 1230 } 1231 1232 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1233 const struct __ip6_tnl_parm *p) 1234 { 1235 u->proto = IPPROTO_GRE; 1236 u->laddr = p->laddr; 1237 u->raddr = p->raddr; 1238 u->flags = p->flags; 1239 u->hop_limit = p->hop_limit; 1240 u->encap_limit = p->encap_limit; 1241 u->flowinfo = p->flowinfo; 1242 u->link = p->link; 1243 u->i_key = p->i_key; 1244 u->o_key = p->o_key; 1245 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1246 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1247 memcpy(u->name, p->name, sizeof(u->name)); 1248 } 1249 1250 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1251 struct ifreq *ifr, void __user *data, 1252 int cmd) 1253 { 1254 int err = 0; 1255 struct ip6_tnl_parm2 p; 1256 struct __ip6_tnl_parm p1; 1257 struct ip6_tnl *t = netdev_priv(dev); 1258 struct net *net = t->net; 1259 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1260 1261 memset(&p1, 0, sizeof(p1)); 1262 1263 switch (cmd) { 1264 case SIOCGETTUNNEL: 1265 if (dev == ign->fb_tunnel_dev) { 1266 if (copy_from_user(&p, data, sizeof(p))) { 1267 err = -EFAULT; 1268 break; 1269 } 1270 ip6gre_tnl_parm_from_user(&p1, &p); 1271 t = ip6gre_tunnel_locate(net, &p1, 0); 1272 if (!t) 1273 t = netdev_priv(dev); 1274 } 1275 memset(&p, 0, sizeof(p)); 1276 ip6gre_tnl_parm_to_user(&p, &t->parms); 1277 if (copy_to_user(data, &p, sizeof(p))) 1278 err = -EFAULT; 1279 break; 1280 1281 case SIOCADDTUNNEL: 1282 case SIOCCHGTUNNEL: 1283 err = -EPERM; 1284 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1285 goto done; 1286 1287 err = -EFAULT; 1288 if (copy_from_user(&p, data, sizeof(p))) 1289 goto done; 1290 1291 err = -EINVAL; 1292 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1293 goto done; 1294 1295 if (!(p.i_flags&GRE_KEY)) 1296 p.i_key = 0; 1297 if (!(p.o_flags&GRE_KEY)) 1298 p.o_key = 0; 1299 1300 ip6gre_tnl_parm_from_user(&p1, &p); 1301 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1302 1303 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1304 if (t) { 1305 if (t->dev != dev) { 1306 err = -EEXIST; 1307 break; 1308 } 1309 } else { 1310 t = netdev_priv(dev); 1311 1312 ip6gre_tunnel_unlink(ign, t); 1313 synchronize_net(); 1314 ip6gre_tnl_change(t, &p1, 1); 1315 ip6gre_tunnel_link(ign, t); 1316 netdev_state_change(dev); 1317 } 1318 } 1319 1320 if (t) { 1321 err = 0; 1322 1323 memset(&p, 0, sizeof(p)); 1324 ip6gre_tnl_parm_to_user(&p, &t->parms); 1325 if (copy_to_user(data, &p, sizeof(p))) 1326 err = -EFAULT; 1327 } else 1328 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1329 break; 1330 1331 case SIOCDELTUNNEL: 1332 err = -EPERM; 1333 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1334 goto done; 1335 1336 if (dev == ign->fb_tunnel_dev) { 1337 err = -EFAULT; 1338 if (copy_from_user(&p, data, sizeof(p))) 1339 goto done; 1340 err = -ENOENT; 1341 ip6gre_tnl_parm_from_user(&p1, &p); 1342 t = ip6gre_tunnel_locate(net, &p1, 0); 1343 if (!t) 1344 goto done; 1345 err = -EPERM; 1346 if (t == netdev_priv(ign->fb_tunnel_dev)) 1347 goto done; 1348 dev = t->dev; 1349 } 1350 unregister_netdevice(dev); 1351 err = 0; 1352 break; 1353 1354 default: 1355 err = -EINVAL; 1356 } 1357 1358 done: 1359 return err; 1360 } 1361 1362 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1363 unsigned short type, const void *daddr, 1364 const void *saddr, unsigned int len) 1365 { 1366 struct ip6_tnl *t = netdev_priv(dev); 1367 struct ipv6hdr *ipv6h; 1368 __be16 *p; 1369 1370 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1371 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1372 t->fl.u.ip6.flowlabel, 1373 true, &t->fl.u.ip6)); 1374 ipv6h->hop_limit = t->parms.hop_limit; 1375 ipv6h->nexthdr = NEXTHDR_GRE; 1376 ipv6h->saddr = t->parms.laddr; 1377 ipv6h->daddr = t->parms.raddr; 1378 1379 p = (__be16 *)(ipv6h + 1); 1380 p[0] = t->parms.o_flags; 1381 p[1] = htons(type); 1382 1383 /* 1384 * Set the source hardware address. 1385 */ 1386 1387 if (saddr) 1388 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1389 if (daddr) 1390 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1391 if (!ipv6_addr_any(&ipv6h->daddr)) 1392 return t->hlen; 1393 1394 return -t->hlen; 1395 } 1396 1397 static const struct header_ops ip6gre_header_ops = { 1398 .create = ip6gre_header, 1399 }; 1400 1401 static const struct net_device_ops ip6gre_netdev_ops = { 1402 .ndo_init = ip6gre_tunnel_init, 1403 .ndo_uninit = ip6gre_tunnel_uninit, 1404 .ndo_start_xmit = ip6gre_tunnel_xmit, 1405 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1406 .ndo_change_mtu = ip6_tnl_change_mtu, 1407 .ndo_get_stats64 = dev_get_tstats64, 1408 .ndo_get_iflink = ip6_tnl_get_iflink, 1409 }; 1410 1411 static void ip6gre_dev_free(struct net_device *dev) 1412 { 1413 struct ip6_tnl *t = netdev_priv(dev); 1414 1415 gro_cells_destroy(&t->gro_cells); 1416 dst_cache_destroy(&t->dst_cache); 1417 free_percpu(dev->tstats); 1418 } 1419 1420 static void ip6gre_tunnel_setup(struct net_device *dev) 1421 { 1422 dev->netdev_ops = &ip6gre_netdev_ops; 1423 dev->needs_free_netdev = true; 1424 dev->priv_destructor = ip6gre_dev_free; 1425 1426 dev->type = ARPHRD_IP6GRE; 1427 1428 dev->flags |= IFF_NOARP; 1429 dev->addr_len = sizeof(struct in6_addr); 1430 netif_keep_dst(dev); 1431 /* This perm addr will be used as interface identifier by IPv6 */ 1432 dev->addr_assign_type = NET_ADDR_RANDOM; 1433 eth_random_addr(dev->perm_addr); 1434 } 1435 1436 #define GRE6_FEATURES (NETIF_F_SG | \ 1437 NETIF_F_FRAGLIST | \ 1438 NETIF_F_HIGHDMA | \ 1439 NETIF_F_HW_CSUM) 1440 1441 static void ip6gre_tnl_init_features(struct net_device *dev) 1442 { 1443 struct ip6_tnl *nt = netdev_priv(dev); 1444 1445 dev->features |= GRE6_FEATURES; 1446 dev->hw_features |= GRE6_FEATURES; 1447 1448 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1449 /* TCP offload with GRE SEQ is not supported, nor 1450 * can we support 2 levels of outer headers requiring 1451 * an update. 1452 */ 1453 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1454 nt->encap.type == TUNNEL_ENCAP_NONE) { 1455 dev->features |= NETIF_F_GSO_SOFTWARE; 1456 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1457 } 1458 1459 /* Can use a lockless transmit, unless we generate 1460 * output sequences 1461 */ 1462 dev->features |= NETIF_F_LLTX; 1463 } 1464 } 1465 1466 static int ip6gre_tunnel_init_common(struct net_device *dev) 1467 { 1468 struct ip6_tnl *tunnel; 1469 int ret; 1470 int t_hlen; 1471 1472 tunnel = netdev_priv(dev); 1473 1474 tunnel->dev = dev; 1475 tunnel->net = dev_net(dev); 1476 strcpy(tunnel->parms.name, dev->name); 1477 1478 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1479 if (!dev->tstats) 1480 return -ENOMEM; 1481 1482 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1483 if (ret) 1484 goto cleanup_alloc_pcpu_stats; 1485 1486 ret = gro_cells_init(&tunnel->gro_cells, dev); 1487 if (ret) 1488 goto cleanup_dst_cache_init; 1489 1490 t_hlen = ip6gre_calc_hlen(tunnel); 1491 dev->mtu = ETH_DATA_LEN - t_hlen; 1492 if (dev->type == ARPHRD_ETHER) 1493 dev->mtu -= ETH_HLEN; 1494 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1495 dev->mtu -= 8; 1496 1497 if (tunnel->parms.collect_md) { 1498 netif_keep_dst(dev); 1499 } 1500 ip6gre_tnl_init_features(dev); 1501 1502 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL); 1503 return 0; 1504 1505 cleanup_dst_cache_init: 1506 dst_cache_destroy(&tunnel->dst_cache); 1507 cleanup_alloc_pcpu_stats: 1508 free_percpu(dev->tstats); 1509 dev->tstats = NULL; 1510 return ret; 1511 } 1512 1513 static int ip6gre_tunnel_init(struct net_device *dev) 1514 { 1515 struct ip6_tnl *tunnel; 1516 int ret; 1517 1518 ret = ip6gre_tunnel_init_common(dev); 1519 if (ret) 1520 return ret; 1521 1522 tunnel = netdev_priv(dev); 1523 1524 if (tunnel->parms.collect_md) 1525 return 0; 1526 1527 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1528 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1529 1530 if (ipv6_addr_any(&tunnel->parms.raddr)) 1531 dev->header_ops = &ip6gre_header_ops; 1532 1533 return 0; 1534 } 1535 1536 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1537 { 1538 struct ip6_tnl *tunnel = netdev_priv(dev); 1539 1540 tunnel->dev = dev; 1541 tunnel->net = dev_net(dev); 1542 strcpy(tunnel->parms.name, dev->name); 1543 1544 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1545 } 1546 1547 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1548 .handler = gre_rcv, 1549 .err_handler = ip6gre_err, 1550 .flags = INET6_PROTO_FINAL, 1551 }; 1552 1553 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1554 { 1555 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1556 struct net_device *dev, *aux; 1557 int prio; 1558 1559 for_each_netdev_safe(net, dev, aux) 1560 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1561 dev->rtnl_link_ops == &ip6gre_tap_ops || 1562 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1563 unregister_netdevice_queue(dev, head); 1564 1565 for (prio = 0; prio < 4; prio++) { 1566 int h; 1567 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1568 struct ip6_tnl *t; 1569 1570 t = rtnl_dereference(ign->tunnels[prio][h]); 1571 1572 while (t) { 1573 /* If dev is in the same netns, it has already 1574 * been added to the list by the previous loop. 1575 */ 1576 if (!net_eq(dev_net(t->dev), net)) 1577 unregister_netdevice_queue(t->dev, 1578 head); 1579 t = rtnl_dereference(t->next); 1580 } 1581 } 1582 } 1583 } 1584 1585 static int __net_init ip6gre_init_net(struct net *net) 1586 { 1587 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1588 struct net_device *ndev; 1589 int err; 1590 1591 if (!net_has_fallback_tunnels(net)) 1592 return 0; 1593 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1594 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1595 if (!ndev) { 1596 err = -ENOMEM; 1597 goto err_alloc_dev; 1598 } 1599 ign->fb_tunnel_dev = ndev; 1600 dev_net_set(ign->fb_tunnel_dev, net); 1601 /* FB netdevice is special: we have one, and only one per netns. 1602 * Allowing to move it to another netns is clearly unsafe. 1603 */ 1604 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1605 1606 1607 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1608 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1609 1610 err = register_netdev(ign->fb_tunnel_dev); 1611 if (err) 1612 goto err_reg_dev; 1613 1614 rcu_assign_pointer(ign->tunnels_wc[0], 1615 netdev_priv(ign->fb_tunnel_dev)); 1616 return 0; 1617 1618 err_reg_dev: 1619 free_netdev(ndev); 1620 err_alloc_dev: 1621 return err; 1622 } 1623 1624 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1625 { 1626 struct net *net; 1627 LIST_HEAD(list); 1628 1629 rtnl_lock(); 1630 list_for_each_entry(net, net_list, exit_list) 1631 ip6gre_destroy_tunnels(net, &list); 1632 unregister_netdevice_many(&list); 1633 rtnl_unlock(); 1634 } 1635 1636 static struct pernet_operations ip6gre_net_ops = { 1637 .init = ip6gre_init_net, 1638 .exit_batch = ip6gre_exit_batch_net, 1639 .id = &ip6gre_net_id, 1640 .size = sizeof(struct ip6gre_net), 1641 }; 1642 1643 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1644 struct netlink_ext_ack *extack) 1645 { 1646 __be16 flags; 1647 1648 if (!data) 1649 return 0; 1650 1651 flags = 0; 1652 if (data[IFLA_GRE_IFLAGS]) 1653 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1654 if (data[IFLA_GRE_OFLAGS]) 1655 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1656 if (flags & (GRE_VERSION|GRE_ROUTING)) 1657 return -EINVAL; 1658 1659 return 0; 1660 } 1661 1662 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1663 struct netlink_ext_ack *extack) 1664 { 1665 struct in6_addr daddr; 1666 1667 if (tb[IFLA_ADDRESS]) { 1668 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1669 return -EINVAL; 1670 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1671 return -EADDRNOTAVAIL; 1672 } 1673 1674 if (!data) 1675 goto out; 1676 1677 if (data[IFLA_GRE_REMOTE]) { 1678 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1679 if (ipv6_addr_any(&daddr)) 1680 return -EINVAL; 1681 } 1682 1683 out: 1684 return ip6gre_tunnel_validate(tb, data, extack); 1685 } 1686 1687 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1688 struct netlink_ext_ack *extack) 1689 { 1690 __be16 flags = 0; 1691 int ret, ver = 0; 1692 1693 if (!data) 1694 return 0; 1695 1696 ret = ip6gre_tap_validate(tb, data, extack); 1697 if (ret) 1698 return ret; 1699 1700 /* ERSPAN should only have GRE sequence and key flag */ 1701 if (data[IFLA_GRE_OFLAGS]) 1702 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1703 if (data[IFLA_GRE_IFLAGS]) 1704 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1705 if (!data[IFLA_GRE_COLLECT_METADATA] && 1706 flags != (GRE_SEQ | GRE_KEY)) 1707 return -EINVAL; 1708 1709 /* ERSPAN Session ID only has 10-bit. Since we reuse 1710 * 32-bit key field as ID, check it's range. 1711 */ 1712 if (data[IFLA_GRE_IKEY] && 1713 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1714 return -EINVAL; 1715 1716 if (data[IFLA_GRE_OKEY] && 1717 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1718 return -EINVAL; 1719 1720 if (data[IFLA_GRE_ERSPAN_VER]) { 1721 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1722 if (ver != 1 && ver != 2) 1723 return -EINVAL; 1724 } 1725 1726 if (ver == 1) { 1727 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1728 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1729 1730 if (index & ~INDEX_MASK) 1731 return -EINVAL; 1732 } 1733 } else if (ver == 2) { 1734 if (data[IFLA_GRE_ERSPAN_DIR]) { 1735 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1736 1737 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1738 return -EINVAL; 1739 } 1740 1741 if (data[IFLA_GRE_ERSPAN_HWID]) { 1742 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1743 1744 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1745 return -EINVAL; 1746 } 1747 } 1748 1749 return 0; 1750 } 1751 1752 static void ip6erspan_set_version(struct nlattr *data[], 1753 struct __ip6_tnl_parm *parms) 1754 { 1755 if (!data) 1756 return; 1757 1758 parms->erspan_ver = 1; 1759 if (data[IFLA_GRE_ERSPAN_VER]) 1760 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1761 1762 if (parms->erspan_ver == 1) { 1763 if (data[IFLA_GRE_ERSPAN_INDEX]) 1764 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1765 } else if (parms->erspan_ver == 2) { 1766 if (data[IFLA_GRE_ERSPAN_DIR]) 1767 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1768 if (data[IFLA_GRE_ERSPAN_HWID]) 1769 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1770 } 1771 } 1772 1773 static void ip6gre_netlink_parms(struct nlattr *data[], 1774 struct __ip6_tnl_parm *parms) 1775 { 1776 memset(parms, 0, sizeof(*parms)); 1777 1778 if (!data) 1779 return; 1780 1781 if (data[IFLA_GRE_LINK]) 1782 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1783 1784 if (data[IFLA_GRE_IFLAGS]) 1785 parms->i_flags = gre_flags_to_tnl_flags( 1786 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1787 1788 if (data[IFLA_GRE_OFLAGS]) 1789 parms->o_flags = gre_flags_to_tnl_flags( 1790 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1791 1792 if (data[IFLA_GRE_IKEY]) 1793 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1794 1795 if (data[IFLA_GRE_OKEY]) 1796 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1797 1798 if (data[IFLA_GRE_LOCAL]) 1799 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1800 1801 if (data[IFLA_GRE_REMOTE]) 1802 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1803 1804 if (data[IFLA_GRE_TTL]) 1805 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1806 1807 if (data[IFLA_GRE_ENCAP_LIMIT]) 1808 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1809 1810 if (data[IFLA_GRE_FLOWINFO]) 1811 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1812 1813 if (data[IFLA_GRE_FLAGS]) 1814 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1815 1816 if (data[IFLA_GRE_FWMARK]) 1817 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1818 1819 if (data[IFLA_GRE_COLLECT_METADATA]) 1820 parms->collect_md = true; 1821 } 1822 1823 static int ip6gre_tap_init(struct net_device *dev) 1824 { 1825 int ret; 1826 1827 ret = ip6gre_tunnel_init_common(dev); 1828 if (ret) 1829 return ret; 1830 1831 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1832 1833 return 0; 1834 } 1835 1836 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1837 .ndo_init = ip6gre_tap_init, 1838 .ndo_uninit = ip6gre_tunnel_uninit, 1839 .ndo_start_xmit = ip6gre_tunnel_xmit, 1840 .ndo_set_mac_address = eth_mac_addr, 1841 .ndo_validate_addr = eth_validate_addr, 1842 .ndo_change_mtu = ip6_tnl_change_mtu, 1843 .ndo_get_stats64 = dev_get_tstats64, 1844 .ndo_get_iflink = ip6_tnl_get_iflink, 1845 }; 1846 1847 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1848 { 1849 int t_hlen; 1850 1851 tunnel->tun_hlen = 8; 1852 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1853 erspan_hdr_len(tunnel->parms.erspan_ver); 1854 1855 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1856 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1857 return t_hlen; 1858 } 1859 1860 static int ip6erspan_tap_init(struct net_device *dev) 1861 { 1862 struct ip6_tnl *tunnel; 1863 int t_hlen; 1864 int ret; 1865 1866 tunnel = netdev_priv(dev); 1867 1868 tunnel->dev = dev; 1869 tunnel->net = dev_net(dev); 1870 strcpy(tunnel->parms.name, dev->name); 1871 1872 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1873 if (!dev->tstats) 1874 return -ENOMEM; 1875 1876 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1877 if (ret) 1878 goto cleanup_alloc_pcpu_stats; 1879 1880 ret = gro_cells_init(&tunnel->gro_cells, dev); 1881 if (ret) 1882 goto cleanup_dst_cache_init; 1883 1884 t_hlen = ip6erspan_calc_hlen(tunnel); 1885 dev->mtu = ETH_DATA_LEN - t_hlen; 1886 if (dev->type == ARPHRD_ETHER) 1887 dev->mtu -= ETH_HLEN; 1888 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1889 dev->mtu -= 8; 1890 1891 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1892 ip6erspan_tnl_link_config(tunnel, 1); 1893 1894 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL); 1895 return 0; 1896 1897 cleanup_dst_cache_init: 1898 dst_cache_destroy(&tunnel->dst_cache); 1899 cleanup_alloc_pcpu_stats: 1900 free_percpu(dev->tstats); 1901 dev->tstats = NULL; 1902 return ret; 1903 } 1904 1905 static const struct net_device_ops ip6erspan_netdev_ops = { 1906 .ndo_init = ip6erspan_tap_init, 1907 .ndo_uninit = ip6erspan_tunnel_uninit, 1908 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1909 .ndo_set_mac_address = eth_mac_addr, 1910 .ndo_validate_addr = eth_validate_addr, 1911 .ndo_change_mtu = ip6_tnl_change_mtu, 1912 .ndo_get_stats64 = dev_get_tstats64, 1913 .ndo_get_iflink = ip6_tnl_get_iflink, 1914 }; 1915 1916 static void ip6gre_tap_setup(struct net_device *dev) 1917 { 1918 1919 ether_setup(dev); 1920 1921 dev->max_mtu = 0; 1922 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1923 dev->needs_free_netdev = true; 1924 dev->priv_destructor = ip6gre_dev_free; 1925 1926 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1927 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1928 netif_keep_dst(dev); 1929 } 1930 1931 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1932 struct ip_tunnel_encap *ipencap) 1933 { 1934 bool ret = false; 1935 1936 memset(ipencap, 0, sizeof(*ipencap)); 1937 1938 if (!data) 1939 return ret; 1940 1941 if (data[IFLA_GRE_ENCAP_TYPE]) { 1942 ret = true; 1943 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1944 } 1945 1946 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1947 ret = true; 1948 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1949 } 1950 1951 if (data[IFLA_GRE_ENCAP_SPORT]) { 1952 ret = true; 1953 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1954 } 1955 1956 if (data[IFLA_GRE_ENCAP_DPORT]) { 1957 ret = true; 1958 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1959 } 1960 1961 return ret; 1962 } 1963 1964 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1965 struct nlattr *tb[], struct nlattr *data[], 1966 struct netlink_ext_ack *extack) 1967 { 1968 struct ip6_tnl *nt; 1969 struct ip_tunnel_encap ipencap; 1970 int err; 1971 1972 nt = netdev_priv(dev); 1973 1974 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1975 int err = ip6_tnl_encap_setup(nt, &ipencap); 1976 1977 if (err < 0) 1978 return err; 1979 } 1980 1981 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1982 eth_hw_addr_random(dev); 1983 1984 nt->dev = dev; 1985 nt->net = dev_net(dev); 1986 1987 err = register_netdevice(dev); 1988 if (err) 1989 goto out; 1990 1991 if (tb[IFLA_MTU]) 1992 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1993 1994 out: 1995 return err; 1996 } 1997 1998 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1999 struct nlattr *tb[], struct nlattr *data[], 2000 struct netlink_ext_ack *extack) 2001 { 2002 struct ip6_tnl *nt = netdev_priv(dev); 2003 struct net *net = dev_net(dev); 2004 struct ip6gre_net *ign; 2005 int err; 2006 2007 ip6gre_netlink_parms(data, &nt->parms); 2008 ign = net_generic(net, ip6gre_net_id); 2009 2010 if (nt->parms.collect_md) { 2011 if (rtnl_dereference(ign->collect_md_tun)) 2012 return -EEXIST; 2013 } else { 2014 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2015 return -EEXIST; 2016 } 2017 2018 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2019 if (!err) { 2020 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2021 ip6gre_tunnel_link_md(ign, nt); 2022 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2023 } 2024 return err; 2025 } 2026 2027 static struct ip6_tnl * 2028 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2029 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2030 struct netlink_ext_ack *extack) 2031 { 2032 struct ip6_tnl *t, *nt = netdev_priv(dev); 2033 struct net *net = nt->net; 2034 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2035 struct ip_tunnel_encap ipencap; 2036 2037 if (dev == ign->fb_tunnel_dev) 2038 return ERR_PTR(-EINVAL); 2039 2040 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2041 int err = ip6_tnl_encap_setup(nt, &ipencap); 2042 2043 if (err < 0) 2044 return ERR_PTR(err); 2045 } 2046 2047 ip6gre_netlink_parms(data, p_p); 2048 2049 t = ip6gre_tunnel_locate(net, p_p, 0); 2050 2051 if (t) { 2052 if (t->dev != dev) 2053 return ERR_PTR(-EEXIST); 2054 } else { 2055 t = nt; 2056 } 2057 2058 return t; 2059 } 2060 2061 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2062 struct nlattr *data[], 2063 struct netlink_ext_ack *extack) 2064 { 2065 struct ip6_tnl *t = netdev_priv(dev); 2066 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2067 struct __ip6_tnl_parm p; 2068 2069 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2070 if (IS_ERR(t)) 2071 return PTR_ERR(t); 2072 2073 ip6gre_tunnel_unlink_md(ign, t); 2074 ip6gre_tunnel_unlink(ign, t); 2075 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2076 ip6gre_tunnel_link_md(ign, t); 2077 ip6gre_tunnel_link(ign, t); 2078 return 0; 2079 } 2080 2081 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2082 { 2083 struct net *net = dev_net(dev); 2084 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2085 2086 if (dev != ign->fb_tunnel_dev) 2087 unregister_netdevice_queue(dev, head); 2088 } 2089 2090 static size_t ip6gre_get_size(const struct net_device *dev) 2091 { 2092 return 2093 /* IFLA_GRE_LINK */ 2094 nla_total_size(4) + 2095 /* IFLA_GRE_IFLAGS */ 2096 nla_total_size(2) + 2097 /* IFLA_GRE_OFLAGS */ 2098 nla_total_size(2) + 2099 /* IFLA_GRE_IKEY */ 2100 nla_total_size(4) + 2101 /* IFLA_GRE_OKEY */ 2102 nla_total_size(4) + 2103 /* IFLA_GRE_LOCAL */ 2104 nla_total_size(sizeof(struct in6_addr)) + 2105 /* IFLA_GRE_REMOTE */ 2106 nla_total_size(sizeof(struct in6_addr)) + 2107 /* IFLA_GRE_TTL */ 2108 nla_total_size(1) + 2109 /* IFLA_GRE_ENCAP_LIMIT */ 2110 nla_total_size(1) + 2111 /* IFLA_GRE_FLOWINFO */ 2112 nla_total_size(4) + 2113 /* IFLA_GRE_FLAGS */ 2114 nla_total_size(4) + 2115 /* IFLA_GRE_ENCAP_TYPE */ 2116 nla_total_size(2) + 2117 /* IFLA_GRE_ENCAP_FLAGS */ 2118 nla_total_size(2) + 2119 /* IFLA_GRE_ENCAP_SPORT */ 2120 nla_total_size(2) + 2121 /* IFLA_GRE_ENCAP_DPORT */ 2122 nla_total_size(2) + 2123 /* IFLA_GRE_COLLECT_METADATA */ 2124 nla_total_size(0) + 2125 /* IFLA_GRE_FWMARK */ 2126 nla_total_size(4) + 2127 /* IFLA_GRE_ERSPAN_INDEX */ 2128 nla_total_size(4) + 2129 0; 2130 } 2131 2132 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2133 { 2134 struct ip6_tnl *t = netdev_priv(dev); 2135 struct __ip6_tnl_parm *p = &t->parms; 2136 __be16 o_flags = p->o_flags; 2137 2138 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2139 if (!p->collect_md) 2140 o_flags |= TUNNEL_KEY; 2141 2142 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2143 goto nla_put_failure; 2144 2145 if (p->erspan_ver == 1) { 2146 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2147 goto nla_put_failure; 2148 } else { 2149 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2150 goto nla_put_failure; 2151 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2152 goto nla_put_failure; 2153 } 2154 } 2155 2156 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2157 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2158 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2159 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2160 gre_tnl_flags_to_gre_flags(o_flags)) || 2161 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2162 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2163 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2164 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2165 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2166 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2167 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2168 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2169 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2170 goto nla_put_failure; 2171 2172 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2173 t->encap.type) || 2174 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2175 t->encap.sport) || 2176 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2177 t->encap.dport) || 2178 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2179 t->encap.flags)) 2180 goto nla_put_failure; 2181 2182 if (p->collect_md) { 2183 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2184 goto nla_put_failure; 2185 } 2186 2187 return 0; 2188 2189 nla_put_failure: 2190 return -EMSGSIZE; 2191 } 2192 2193 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2194 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2195 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2196 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2197 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2198 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2199 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2200 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2201 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2202 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2203 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2204 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2205 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2206 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2207 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2208 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2209 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2210 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2211 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2212 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2213 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2214 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2215 }; 2216 2217 static void ip6erspan_tap_setup(struct net_device *dev) 2218 { 2219 ether_setup(dev); 2220 2221 dev->max_mtu = 0; 2222 dev->netdev_ops = &ip6erspan_netdev_ops; 2223 dev->needs_free_netdev = true; 2224 dev->priv_destructor = ip6gre_dev_free; 2225 2226 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2227 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2228 netif_keep_dst(dev); 2229 } 2230 2231 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2232 struct nlattr *tb[], struct nlattr *data[], 2233 struct netlink_ext_ack *extack) 2234 { 2235 struct ip6_tnl *nt = netdev_priv(dev); 2236 struct net *net = dev_net(dev); 2237 struct ip6gre_net *ign; 2238 int err; 2239 2240 ip6gre_netlink_parms(data, &nt->parms); 2241 ip6erspan_set_version(data, &nt->parms); 2242 ign = net_generic(net, ip6gre_net_id); 2243 2244 if (nt->parms.collect_md) { 2245 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2246 return -EEXIST; 2247 } else { 2248 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2249 return -EEXIST; 2250 } 2251 2252 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2253 if (!err) { 2254 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2255 ip6erspan_tunnel_link_md(ign, nt); 2256 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2257 } 2258 return err; 2259 } 2260 2261 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2262 { 2263 ip6gre_tnl_link_config_common(t); 2264 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2265 } 2266 2267 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2268 const struct __ip6_tnl_parm *p, int set_mtu) 2269 { 2270 ip6gre_tnl_copy_tnl_parm(t, p); 2271 ip6erspan_tnl_link_config(t, set_mtu); 2272 return 0; 2273 } 2274 2275 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2276 struct nlattr *data[], 2277 struct netlink_ext_ack *extack) 2278 { 2279 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2280 struct __ip6_tnl_parm p; 2281 struct ip6_tnl *t; 2282 2283 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2284 if (IS_ERR(t)) 2285 return PTR_ERR(t); 2286 2287 ip6erspan_set_version(data, &p); 2288 ip6gre_tunnel_unlink_md(ign, t); 2289 ip6gre_tunnel_unlink(ign, t); 2290 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2291 ip6erspan_tunnel_link_md(ign, t); 2292 ip6gre_tunnel_link(ign, t); 2293 return 0; 2294 } 2295 2296 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2297 .kind = "ip6gre", 2298 .maxtype = IFLA_GRE_MAX, 2299 .policy = ip6gre_policy, 2300 .priv_size = sizeof(struct ip6_tnl), 2301 .setup = ip6gre_tunnel_setup, 2302 .validate = ip6gre_tunnel_validate, 2303 .newlink = ip6gre_newlink, 2304 .changelink = ip6gre_changelink, 2305 .dellink = ip6gre_dellink, 2306 .get_size = ip6gre_get_size, 2307 .fill_info = ip6gre_fill_info, 2308 .get_link_net = ip6_tnl_get_link_net, 2309 }; 2310 2311 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2312 .kind = "ip6gretap", 2313 .maxtype = IFLA_GRE_MAX, 2314 .policy = ip6gre_policy, 2315 .priv_size = sizeof(struct ip6_tnl), 2316 .setup = ip6gre_tap_setup, 2317 .validate = ip6gre_tap_validate, 2318 .newlink = ip6gre_newlink, 2319 .changelink = ip6gre_changelink, 2320 .get_size = ip6gre_get_size, 2321 .fill_info = ip6gre_fill_info, 2322 .get_link_net = ip6_tnl_get_link_net, 2323 }; 2324 2325 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2326 .kind = "ip6erspan", 2327 .maxtype = IFLA_GRE_MAX, 2328 .policy = ip6gre_policy, 2329 .priv_size = sizeof(struct ip6_tnl), 2330 .setup = ip6erspan_tap_setup, 2331 .validate = ip6erspan_tap_validate, 2332 .newlink = ip6erspan_newlink, 2333 .changelink = ip6erspan_changelink, 2334 .get_size = ip6gre_get_size, 2335 .fill_info = ip6gre_fill_info, 2336 .get_link_net = ip6_tnl_get_link_net, 2337 }; 2338 2339 /* 2340 * And now the modules code and kernel interface. 2341 */ 2342 2343 static int __init ip6gre_init(void) 2344 { 2345 int err; 2346 2347 pr_info("GRE over IPv6 tunneling driver\n"); 2348 2349 err = register_pernet_device(&ip6gre_net_ops); 2350 if (err < 0) 2351 return err; 2352 2353 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2354 if (err < 0) { 2355 pr_info("%s: can't add protocol\n", __func__); 2356 goto add_proto_failed; 2357 } 2358 2359 err = rtnl_link_register(&ip6gre_link_ops); 2360 if (err < 0) 2361 goto rtnl_link_failed; 2362 2363 err = rtnl_link_register(&ip6gre_tap_ops); 2364 if (err < 0) 2365 goto tap_ops_failed; 2366 2367 err = rtnl_link_register(&ip6erspan_tap_ops); 2368 if (err < 0) 2369 goto erspan_link_failed; 2370 2371 out: 2372 return err; 2373 2374 erspan_link_failed: 2375 rtnl_link_unregister(&ip6gre_tap_ops); 2376 tap_ops_failed: 2377 rtnl_link_unregister(&ip6gre_link_ops); 2378 rtnl_link_failed: 2379 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2380 add_proto_failed: 2381 unregister_pernet_device(&ip6gre_net_ops); 2382 goto out; 2383 } 2384 2385 static void __exit ip6gre_fini(void) 2386 { 2387 rtnl_link_unregister(&ip6gre_tap_ops); 2388 rtnl_link_unregister(&ip6gre_link_ops); 2389 rtnl_link_unregister(&ip6erspan_tap_ops); 2390 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2391 unregister_pernet_device(&ip6gre_net_ops); 2392 } 2393 2394 module_init(ip6gre_init); 2395 module_exit(ip6gre_fini); 2396 MODULE_LICENSE("GPL"); 2397 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2398 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2399 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2400 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2401 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2402 MODULE_ALIAS_NETDEV("ip6gre0"); 2403