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 __be16 flags; 728 729 if (dev->type == ARPHRD_ETHER) 730 IPCB(skb)->flags = 0; 731 732 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 733 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 734 else 735 fl6->daddr = tunnel->parms.raddr; 736 737 /* Push GRE header. */ 738 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 739 740 if (tunnel->parms.collect_md) { 741 struct ip_tunnel_info *tun_info; 742 const struct ip_tunnel_key *key; 743 int tun_hlen; 744 745 tun_info = skb_tunnel_info_txcheck(skb); 746 if (IS_ERR(tun_info) || 747 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 748 return -EINVAL; 749 750 key = &tun_info->key; 751 memset(fl6, 0, sizeof(*fl6)); 752 fl6->flowi6_proto = IPPROTO_GRE; 753 fl6->daddr = key->u.ipv6.dst; 754 fl6->flowlabel = key->label; 755 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 756 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id); 757 758 dsfield = key->tos; 759 flags = key->tun_flags & 760 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 761 tun_hlen = gre_calc_hlen(flags); 762 763 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen)) 764 return -ENOMEM; 765 766 gre_build_header(skb, tun_hlen, 767 flags, protocol, 768 tunnel_id_to_key32(tun_info->key.tun_id), 769 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) 770 : 0); 771 772 } else { 773 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 774 return -ENOMEM; 775 776 flags = tunnel->parms.o_flags; 777 778 gre_build_header(skb, tunnel->tun_hlen, flags, 779 protocol, tunnel->parms.o_key, 780 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) 781 : 0); 782 } 783 784 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 785 NEXTHDR_GRE); 786 } 787 788 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 789 { 790 struct ip6_tnl *t = netdev_priv(dev); 791 int encap_limit = -1; 792 struct flowi6 fl6; 793 __u8 dsfield = 0; 794 __u32 mtu; 795 int err; 796 797 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 798 799 if (!t->parms.collect_md) 800 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 801 &dsfield, &encap_limit); 802 803 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 804 if (err) 805 return -1; 806 807 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 808 skb->protocol); 809 if (err != 0) { 810 /* XXX: send ICMP error even if DF is not set. */ 811 if (err == -EMSGSIZE) 812 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 813 htonl(mtu)); 814 return -1; 815 } 816 817 return 0; 818 } 819 820 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 821 { 822 struct ip6_tnl *t = netdev_priv(dev); 823 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 824 int encap_limit = -1; 825 struct flowi6 fl6; 826 __u8 dsfield = 0; 827 __u32 mtu; 828 int err; 829 830 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 831 return -1; 832 833 if (!t->parms.collect_md && 834 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 835 return -1; 836 837 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 838 return -1; 839 840 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 841 &mtu, skb->protocol); 842 if (err != 0) { 843 if (err == -EMSGSIZE) 844 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 845 return -1; 846 } 847 848 return 0; 849 } 850 851 /** 852 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 853 * @t: the outgoing tunnel device 854 * @hdr: IPv6 header from the incoming packet 855 * 856 * Description: 857 * Avoid trivial tunneling loop by checking that tunnel exit-point 858 * doesn't match source of incoming packet. 859 * 860 * Return: 861 * 1 if conflict, 862 * 0 else 863 **/ 864 865 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 866 const struct ipv6hdr *hdr) 867 { 868 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 869 } 870 871 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 872 { 873 struct ip6_tnl *t = netdev_priv(dev); 874 int encap_limit = -1; 875 struct flowi6 fl6; 876 __u32 mtu; 877 int err; 878 879 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 880 encap_limit = t->parms.encap_limit; 881 882 if (!t->parms.collect_md) 883 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 884 885 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 886 if (err) 887 return err; 888 889 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 890 891 return err; 892 } 893 894 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 895 struct net_device *dev) 896 { 897 struct ip6_tnl *t = netdev_priv(dev); 898 struct net_device_stats *stats = &t->dev->stats; 899 int ret; 900 901 if (!pskb_inet_may_pull(skb)) 902 goto tx_err; 903 904 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 905 goto tx_err; 906 907 switch (skb->protocol) { 908 case htons(ETH_P_IP): 909 ret = ip6gre_xmit_ipv4(skb, dev); 910 break; 911 case htons(ETH_P_IPV6): 912 ret = ip6gre_xmit_ipv6(skb, dev); 913 break; 914 default: 915 ret = ip6gre_xmit_other(skb, dev); 916 break; 917 } 918 919 if (ret < 0) 920 goto tx_err; 921 922 return NETDEV_TX_OK; 923 924 tx_err: 925 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 926 stats->tx_errors++; 927 stats->tx_dropped++; 928 kfree_skb(skb); 929 return NETDEV_TX_OK; 930 } 931 932 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 933 struct net_device *dev) 934 { 935 struct ip_tunnel_info *tun_info = NULL; 936 struct ip6_tnl *t = netdev_priv(dev); 937 struct dst_entry *dst = skb_dst(skb); 938 struct net_device_stats *stats; 939 bool truncate = false; 940 int encap_limit = -1; 941 __u8 dsfield = false; 942 struct flowi6 fl6; 943 int err = -EINVAL; 944 __be16 proto; 945 __u32 mtu; 946 int nhoff; 947 int thoff; 948 949 if (!pskb_inet_may_pull(skb)) 950 goto tx_err; 951 952 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 953 goto tx_err; 954 955 if (gre_handle_offloads(skb, false)) 956 goto tx_err; 957 958 if (skb->len > dev->mtu + dev->hard_header_len) { 959 pskb_trim(skb, dev->mtu + dev->hard_header_len); 960 truncate = true; 961 } 962 963 nhoff = skb_network_header(skb) - skb_mac_header(skb); 964 if (skb->protocol == htons(ETH_P_IP) && 965 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 966 truncate = true; 967 968 thoff = skb_transport_header(skb) - skb_mac_header(skb); 969 if (skb->protocol == htons(ETH_P_IPV6) && 970 (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)) 971 truncate = true; 972 973 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 974 goto tx_err; 975 976 t->parms.o_flags &= ~TUNNEL_KEY; 977 IPCB(skb)->flags = 0; 978 979 /* For collect_md mode, derive fl6 from the tunnel key, 980 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 981 */ 982 if (t->parms.collect_md) { 983 const struct ip_tunnel_key *key; 984 struct erspan_metadata *md; 985 __be32 tun_id; 986 987 tun_info = skb_tunnel_info_txcheck(skb); 988 if (IS_ERR(tun_info) || 989 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 990 goto tx_err; 991 992 key = &tun_info->key; 993 memset(&fl6, 0, sizeof(fl6)); 994 fl6.flowi6_proto = IPPROTO_GRE; 995 fl6.daddr = key->u.ipv6.dst; 996 fl6.flowlabel = key->label; 997 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 998 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 999 1000 dsfield = key->tos; 1001 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 1002 goto tx_err; 1003 if (tun_info->options_len < sizeof(*md)) 1004 goto tx_err; 1005 md = ip_tunnel_info_opts(tun_info); 1006 1007 tun_id = tunnel_id_to_key32(key->tun_id); 1008 if (md->version == 1) { 1009 erspan_build_header(skb, 1010 ntohl(tun_id), 1011 ntohl(md->u.index), truncate, 1012 false); 1013 } else if (md->version == 2) { 1014 erspan_build_header_v2(skb, 1015 ntohl(tun_id), 1016 md->u.md2.dir, 1017 get_hwid(&md->u.md2), 1018 truncate, false); 1019 } else { 1020 goto tx_err; 1021 } 1022 } else { 1023 switch (skb->protocol) { 1024 case htons(ETH_P_IP): 1025 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1026 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1027 &dsfield, &encap_limit); 1028 break; 1029 case htons(ETH_P_IPV6): 1030 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1031 goto tx_err; 1032 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1033 &dsfield, &encap_limit)) 1034 goto tx_err; 1035 break; 1036 default: 1037 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1038 break; 1039 } 1040 1041 if (t->parms.erspan_ver == 1) 1042 erspan_build_header(skb, ntohl(t->parms.o_key), 1043 t->parms.index, 1044 truncate, false); 1045 else if (t->parms.erspan_ver == 2) 1046 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1047 t->parms.dir, 1048 t->parms.hwid, 1049 truncate, false); 1050 else 1051 goto tx_err; 1052 1053 fl6.daddr = t->parms.raddr; 1054 } 1055 1056 /* Push GRE header. */ 1057 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1058 : htons(ETH_P_ERSPAN2); 1059 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno))); 1060 1061 /* TooBig packet may have updated dst->dev's mtu */ 1062 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1063 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1064 1065 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1066 NEXTHDR_GRE); 1067 if (err != 0) { 1068 /* XXX: send ICMP error even if DF is not set. */ 1069 if (err == -EMSGSIZE) { 1070 if (skb->protocol == htons(ETH_P_IP)) 1071 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1072 ICMP_FRAG_NEEDED, htonl(mtu)); 1073 else 1074 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1075 } 1076 1077 goto tx_err; 1078 } 1079 return NETDEV_TX_OK; 1080 1081 tx_err: 1082 stats = &t->dev->stats; 1083 if (!IS_ERR(tun_info)) 1084 stats->tx_errors++; 1085 stats->tx_dropped++; 1086 kfree_skb(skb); 1087 return NETDEV_TX_OK; 1088 } 1089 1090 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1091 { 1092 struct net_device *dev = t->dev; 1093 struct __ip6_tnl_parm *p = &t->parms; 1094 struct flowi6 *fl6 = &t->fl.u.ip6; 1095 1096 if (dev->type != ARPHRD_ETHER) { 1097 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1098 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1099 } 1100 1101 /* Set up flowi template */ 1102 fl6->saddr = p->laddr; 1103 fl6->daddr = p->raddr; 1104 fl6->flowi6_oif = p->link; 1105 fl6->flowlabel = 0; 1106 fl6->flowi6_proto = IPPROTO_GRE; 1107 fl6->fl6_gre_key = t->parms.o_key; 1108 1109 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1110 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1111 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1112 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1113 1114 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1115 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1116 1117 if (p->flags&IP6_TNL_F_CAP_XMIT && 1118 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1119 dev->flags |= IFF_POINTOPOINT; 1120 else 1121 dev->flags &= ~IFF_POINTOPOINT; 1122 } 1123 1124 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1125 int t_hlen) 1126 { 1127 const struct __ip6_tnl_parm *p = &t->parms; 1128 struct net_device *dev = t->dev; 1129 1130 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1131 int strict = (ipv6_addr_type(&p->raddr) & 1132 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1133 1134 struct rt6_info *rt = rt6_lookup(t->net, 1135 &p->raddr, &p->laddr, 1136 p->link, NULL, strict); 1137 1138 if (!rt) 1139 return; 1140 1141 if (rt->dst.dev) { 1142 unsigned short dst_len = rt->dst.dev->hard_header_len + 1143 t_hlen; 1144 1145 if (t->dev->header_ops) 1146 dev->hard_header_len = dst_len; 1147 else 1148 dev->needed_headroom = dst_len; 1149 1150 if (set_mtu) { 1151 dev->mtu = rt->dst.dev->mtu - t_hlen; 1152 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1153 dev->mtu -= 8; 1154 if (dev->type == ARPHRD_ETHER) 1155 dev->mtu -= ETH_HLEN; 1156 1157 if (dev->mtu < IPV6_MIN_MTU) 1158 dev->mtu = IPV6_MIN_MTU; 1159 } 1160 } 1161 ip6_rt_put(rt); 1162 } 1163 } 1164 1165 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1166 { 1167 int t_hlen; 1168 1169 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1170 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1171 1172 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1173 1174 if (tunnel->dev->header_ops) 1175 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1176 else 1177 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1178 1179 return t_hlen; 1180 } 1181 1182 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1183 { 1184 ip6gre_tnl_link_config_common(t); 1185 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1186 } 1187 1188 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1189 const struct __ip6_tnl_parm *p) 1190 { 1191 t->parms.laddr = p->laddr; 1192 t->parms.raddr = p->raddr; 1193 t->parms.flags = p->flags; 1194 t->parms.hop_limit = p->hop_limit; 1195 t->parms.encap_limit = p->encap_limit; 1196 t->parms.flowinfo = p->flowinfo; 1197 t->parms.link = p->link; 1198 t->parms.proto = p->proto; 1199 t->parms.i_key = p->i_key; 1200 t->parms.o_key = p->o_key; 1201 t->parms.i_flags = p->i_flags; 1202 t->parms.o_flags = p->o_flags; 1203 t->parms.fwmark = p->fwmark; 1204 t->parms.erspan_ver = p->erspan_ver; 1205 t->parms.index = p->index; 1206 t->parms.dir = p->dir; 1207 t->parms.hwid = p->hwid; 1208 dst_cache_reset(&t->dst_cache); 1209 } 1210 1211 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1212 int set_mtu) 1213 { 1214 ip6gre_tnl_copy_tnl_parm(t, p); 1215 ip6gre_tnl_link_config(t, set_mtu); 1216 return 0; 1217 } 1218 1219 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1220 const struct ip6_tnl_parm2 *u) 1221 { 1222 p->laddr = u->laddr; 1223 p->raddr = u->raddr; 1224 p->flags = u->flags; 1225 p->hop_limit = u->hop_limit; 1226 p->encap_limit = u->encap_limit; 1227 p->flowinfo = u->flowinfo; 1228 p->link = u->link; 1229 p->i_key = u->i_key; 1230 p->o_key = u->o_key; 1231 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1232 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1233 memcpy(p->name, u->name, sizeof(u->name)); 1234 } 1235 1236 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1237 const struct __ip6_tnl_parm *p) 1238 { 1239 u->proto = IPPROTO_GRE; 1240 u->laddr = p->laddr; 1241 u->raddr = p->raddr; 1242 u->flags = p->flags; 1243 u->hop_limit = p->hop_limit; 1244 u->encap_limit = p->encap_limit; 1245 u->flowinfo = p->flowinfo; 1246 u->link = p->link; 1247 u->i_key = p->i_key; 1248 u->o_key = p->o_key; 1249 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1250 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1251 memcpy(u->name, p->name, sizeof(u->name)); 1252 } 1253 1254 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1255 struct ifreq *ifr, void __user *data, 1256 int cmd) 1257 { 1258 int err = 0; 1259 struct ip6_tnl_parm2 p; 1260 struct __ip6_tnl_parm p1; 1261 struct ip6_tnl *t = netdev_priv(dev); 1262 struct net *net = t->net; 1263 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1264 1265 memset(&p1, 0, sizeof(p1)); 1266 1267 switch (cmd) { 1268 case SIOCGETTUNNEL: 1269 if (dev == ign->fb_tunnel_dev) { 1270 if (copy_from_user(&p, data, sizeof(p))) { 1271 err = -EFAULT; 1272 break; 1273 } 1274 ip6gre_tnl_parm_from_user(&p1, &p); 1275 t = ip6gre_tunnel_locate(net, &p1, 0); 1276 if (!t) 1277 t = netdev_priv(dev); 1278 } 1279 memset(&p, 0, sizeof(p)); 1280 ip6gre_tnl_parm_to_user(&p, &t->parms); 1281 if (copy_to_user(data, &p, sizeof(p))) 1282 err = -EFAULT; 1283 break; 1284 1285 case SIOCADDTUNNEL: 1286 case SIOCCHGTUNNEL: 1287 err = -EPERM; 1288 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1289 goto done; 1290 1291 err = -EFAULT; 1292 if (copy_from_user(&p, data, sizeof(p))) 1293 goto done; 1294 1295 err = -EINVAL; 1296 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1297 goto done; 1298 1299 if (!(p.i_flags&GRE_KEY)) 1300 p.i_key = 0; 1301 if (!(p.o_flags&GRE_KEY)) 1302 p.o_key = 0; 1303 1304 ip6gre_tnl_parm_from_user(&p1, &p); 1305 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1306 1307 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1308 if (t) { 1309 if (t->dev != dev) { 1310 err = -EEXIST; 1311 break; 1312 } 1313 } else { 1314 t = netdev_priv(dev); 1315 1316 ip6gre_tunnel_unlink(ign, t); 1317 synchronize_net(); 1318 ip6gre_tnl_change(t, &p1, 1); 1319 ip6gre_tunnel_link(ign, t); 1320 netdev_state_change(dev); 1321 } 1322 } 1323 1324 if (t) { 1325 err = 0; 1326 1327 memset(&p, 0, sizeof(p)); 1328 ip6gre_tnl_parm_to_user(&p, &t->parms); 1329 if (copy_to_user(data, &p, sizeof(p))) 1330 err = -EFAULT; 1331 } else 1332 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1333 break; 1334 1335 case SIOCDELTUNNEL: 1336 err = -EPERM; 1337 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1338 goto done; 1339 1340 if (dev == ign->fb_tunnel_dev) { 1341 err = -EFAULT; 1342 if (copy_from_user(&p, data, sizeof(p))) 1343 goto done; 1344 err = -ENOENT; 1345 ip6gre_tnl_parm_from_user(&p1, &p); 1346 t = ip6gre_tunnel_locate(net, &p1, 0); 1347 if (!t) 1348 goto done; 1349 err = -EPERM; 1350 if (t == netdev_priv(ign->fb_tunnel_dev)) 1351 goto done; 1352 dev = t->dev; 1353 } 1354 unregister_netdevice(dev); 1355 err = 0; 1356 break; 1357 1358 default: 1359 err = -EINVAL; 1360 } 1361 1362 done: 1363 return err; 1364 } 1365 1366 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1367 unsigned short type, const void *daddr, 1368 const void *saddr, unsigned int len) 1369 { 1370 struct ip6_tnl *t = netdev_priv(dev); 1371 struct ipv6hdr *ipv6h; 1372 __be16 *p; 1373 1374 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1375 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1376 t->fl.u.ip6.flowlabel, 1377 true, &t->fl.u.ip6)); 1378 ipv6h->hop_limit = t->parms.hop_limit; 1379 ipv6h->nexthdr = NEXTHDR_GRE; 1380 ipv6h->saddr = t->parms.laddr; 1381 ipv6h->daddr = t->parms.raddr; 1382 1383 p = (__be16 *)(ipv6h + 1); 1384 p[0] = t->parms.o_flags; 1385 p[1] = htons(type); 1386 1387 /* 1388 * Set the source hardware address. 1389 */ 1390 1391 if (saddr) 1392 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1393 if (daddr) 1394 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1395 if (!ipv6_addr_any(&ipv6h->daddr)) 1396 return t->hlen; 1397 1398 return -t->hlen; 1399 } 1400 1401 static const struct header_ops ip6gre_header_ops = { 1402 .create = ip6gre_header, 1403 }; 1404 1405 static const struct net_device_ops ip6gre_netdev_ops = { 1406 .ndo_init = ip6gre_tunnel_init, 1407 .ndo_uninit = ip6gre_tunnel_uninit, 1408 .ndo_start_xmit = ip6gre_tunnel_xmit, 1409 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1410 .ndo_change_mtu = ip6_tnl_change_mtu, 1411 .ndo_get_stats64 = dev_get_tstats64, 1412 .ndo_get_iflink = ip6_tnl_get_iflink, 1413 }; 1414 1415 static void ip6gre_dev_free(struct net_device *dev) 1416 { 1417 struct ip6_tnl *t = netdev_priv(dev); 1418 1419 gro_cells_destroy(&t->gro_cells); 1420 dst_cache_destroy(&t->dst_cache); 1421 free_percpu(dev->tstats); 1422 } 1423 1424 static void ip6gre_tunnel_setup(struct net_device *dev) 1425 { 1426 dev->netdev_ops = &ip6gre_netdev_ops; 1427 dev->needs_free_netdev = true; 1428 dev->priv_destructor = ip6gre_dev_free; 1429 1430 dev->type = ARPHRD_IP6GRE; 1431 1432 dev->flags |= IFF_NOARP; 1433 dev->addr_len = sizeof(struct in6_addr); 1434 netif_keep_dst(dev); 1435 /* This perm addr will be used as interface identifier by IPv6 */ 1436 dev->addr_assign_type = NET_ADDR_RANDOM; 1437 eth_random_addr(dev->perm_addr); 1438 } 1439 1440 #define GRE6_FEATURES (NETIF_F_SG | \ 1441 NETIF_F_FRAGLIST | \ 1442 NETIF_F_HIGHDMA | \ 1443 NETIF_F_HW_CSUM) 1444 1445 static void ip6gre_tnl_init_features(struct net_device *dev) 1446 { 1447 struct ip6_tnl *nt = netdev_priv(dev); 1448 1449 dev->features |= GRE6_FEATURES; 1450 dev->hw_features |= GRE6_FEATURES; 1451 1452 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1453 /* TCP offload with GRE SEQ is not supported, nor 1454 * can we support 2 levels of outer headers requiring 1455 * an update. 1456 */ 1457 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1458 nt->encap.type == TUNNEL_ENCAP_NONE) { 1459 dev->features |= NETIF_F_GSO_SOFTWARE; 1460 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1461 } 1462 1463 /* Can use a lockless transmit, unless we generate 1464 * output sequences 1465 */ 1466 dev->features |= NETIF_F_LLTX; 1467 } 1468 } 1469 1470 static int ip6gre_tunnel_init_common(struct net_device *dev) 1471 { 1472 struct ip6_tnl *tunnel; 1473 int ret; 1474 int t_hlen; 1475 1476 tunnel = netdev_priv(dev); 1477 1478 tunnel->dev = dev; 1479 tunnel->net = dev_net(dev); 1480 strcpy(tunnel->parms.name, dev->name); 1481 1482 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1483 if (!dev->tstats) 1484 return -ENOMEM; 1485 1486 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1487 if (ret) 1488 goto cleanup_alloc_pcpu_stats; 1489 1490 ret = gro_cells_init(&tunnel->gro_cells, dev); 1491 if (ret) 1492 goto cleanup_dst_cache_init; 1493 1494 t_hlen = ip6gre_calc_hlen(tunnel); 1495 dev->mtu = ETH_DATA_LEN - t_hlen; 1496 if (dev->type == ARPHRD_ETHER) 1497 dev->mtu -= ETH_HLEN; 1498 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1499 dev->mtu -= 8; 1500 1501 if (tunnel->parms.collect_md) { 1502 netif_keep_dst(dev); 1503 } 1504 ip6gre_tnl_init_features(dev); 1505 1506 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL); 1507 return 0; 1508 1509 cleanup_dst_cache_init: 1510 dst_cache_destroy(&tunnel->dst_cache); 1511 cleanup_alloc_pcpu_stats: 1512 free_percpu(dev->tstats); 1513 dev->tstats = NULL; 1514 return ret; 1515 } 1516 1517 static int ip6gre_tunnel_init(struct net_device *dev) 1518 { 1519 struct ip6_tnl *tunnel; 1520 int ret; 1521 1522 ret = ip6gre_tunnel_init_common(dev); 1523 if (ret) 1524 return ret; 1525 1526 tunnel = netdev_priv(dev); 1527 1528 if (tunnel->parms.collect_md) 1529 return 0; 1530 1531 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1532 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1533 1534 if (ipv6_addr_any(&tunnel->parms.raddr)) 1535 dev->header_ops = &ip6gre_header_ops; 1536 1537 return 0; 1538 } 1539 1540 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1541 { 1542 struct ip6_tnl *tunnel = netdev_priv(dev); 1543 1544 tunnel->dev = dev; 1545 tunnel->net = dev_net(dev); 1546 strcpy(tunnel->parms.name, dev->name); 1547 1548 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1549 } 1550 1551 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1552 .handler = gre_rcv, 1553 .err_handler = ip6gre_err, 1554 .flags = INET6_PROTO_FINAL, 1555 }; 1556 1557 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1558 { 1559 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1560 struct net_device *dev, *aux; 1561 int prio; 1562 1563 for_each_netdev_safe(net, dev, aux) 1564 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1565 dev->rtnl_link_ops == &ip6gre_tap_ops || 1566 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1567 unregister_netdevice_queue(dev, head); 1568 1569 for (prio = 0; prio < 4; prio++) { 1570 int h; 1571 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1572 struct ip6_tnl *t; 1573 1574 t = rtnl_dereference(ign->tunnels[prio][h]); 1575 1576 while (t) { 1577 /* If dev is in the same netns, it has already 1578 * been added to the list by the previous loop. 1579 */ 1580 if (!net_eq(dev_net(t->dev), net)) 1581 unregister_netdevice_queue(t->dev, 1582 head); 1583 t = rtnl_dereference(t->next); 1584 } 1585 } 1586 } 1587 } 1588 1589 static int __net_init ip6gre_init_net(struct net *net) 1590 { 1591 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1592 struct net_device *ndev; 1593 int err; 1594 1595 if (!net_has_fallback_tunnels(net)) 1596 return 0; 1597 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1598 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1599 if (!ndev) { 1600 err = -ENOMEM; 1601 goto err_alloc_dev; 1602 } 1603 ign->fb_tunnel_dev = ndev; 1604 dev_net_set(ign->fb_tunnel_dev, net); 1605 /* FB netdevice is special: we have one, and only one per netns. 1606 * Allowing to move it to another netns is clearly unsafe. 1607 */ 1608 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1609 1610 1611 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1612 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1613 1614 err = register_netdev(ign->fb_tunnel_dev); 1615 if (err) 1616 goto err_reg_dev; 1617 1618 rcu_assign_pointer(ign->tunnels_wc[0], 1619 netdev_priv(ign->fb_tunnel_dev)); 1620 return 0; 1621 1622 err_reg_dev: 1623 free_netdev(ndev); 1624 err_alloc_dev: 1625 return err; 1626 } 1627 1628 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1629 { 1630 struct net *net; 1631 LIST_HEAD(list); 1632 1633 rtnl_lock(); 1634 list_for_each_entry(net, net_list, exit_list) 1635 ip6gre_destroy_tunnels(net, &list); 1636 unregister_netdevice_many(&list); 1637 rtnl_unlock(); 1638 } 1639 1640 static struct pernet_operations ip6gre_net_ops = { 1641 .init = ip6gre_init_net, 1642 .exit_batch = ip6gre_exit_batch_net, 1643 .id = &ip6gre_net_id, 1644 .size = sizeof(struct ip6gre_net), 1645 }; 1646 1647 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1648 struct netlink_ext_ack *extack) 1649 { 1650 __be16 flags; 1651 1652 if (!data) 1653 return 0; 1654 1655 flags = 0; 1656 if (data[IFLA_GRE_IFLAGS]) 1657 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1658 if (data[IFLA_GRE_OFLAGS]) 1659 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1660 if (flags & (GRE_VERSION|GRE_ROUTING)) 1661 return -EINVAL; 1662 1663 return 0; 1664 } 1665 1666 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1667 struct netlink_ext_ack *extack) 1668 { 1669 struct in6_addr daddr; 1670 1671 if (tb[IFLA_ADDRESS]) { 1672 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1673 return -EINVAL; 1674 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1675 return -EADDRNOTAVAIL; 1676 } 1677 1678 if (!data) 1679 goto out; 1680 1681 if (data[IFLA_GRE_REMOTE]) { 1682 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1683 if (ipv6_addr_any(&daddr)) 1684 return -EINVAL; 1685 } 1686 1687 out: 1688 return ip6gre_tunnel_validate(tb, data, extack); 1689 } 1690 1691 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1692 struct netlink_ext_ack *extack) 1693 { 1694 __be16 flags = 0; 1695 int ret, ver = 0; 1696 1697 if (!data) 1698 return 0; 1699 1700 ret = ip6gre_tap_validate(tb, data, extack); 1701 if (ret) 1702 return ret; 1703 1704 /* ERSPAN should only have GRE sequence and key flag */ 1705 if (data[IFLA_GRE_OFLAGS]) 1706 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1707 if (data[IFLA_GRE_IFLAGS]) 1708 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1709 if (!data[IFLA_GRE_COLLECT_METADATA] && 1710 flags != (GRE_SEQ | GRE_KEY)) 1711 return -EINVAL; 1712 1713 /* ERSPAN Session ID only has 10-bit. Since we reuse 1714 * 32-bit key field as ID, check it's range. 1715 */ 1716 if (data[IFLA_GRE_IKEY] && 1717 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1718 return -EINVAL; 1719 1720 if (data[IFLA_GRE_OKEY] && 1721 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1722 return -EINVAL; 1723 1724 if (data[IFLA_GRE_ERSPAN_VER]) { 1725 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1726 if (ver != 1 && ver != 2) 1727 return -EINVAL; 1728 } 1729 1730 if (ver == 1) { 1731 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1732 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1733 1734 if (index & ~INDEX_MASK) 1735 return -EINVAL; 1736 } 1737 } else if (ver == 2) { 1738 if (data[IFLA_GRE_ERSPAN_DIR]) { 1739 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1740 1741 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1742 return -EINVAL; 1743 } 1744 1745 if (data[IFLA_GRE_ERSPAN_HWID]) { 1746 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1747 1748 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1749 return -EINVAL; 1750 } 1751 } 1752 1753 return 0; 1754 } 1755 1756 static void ip6erspan_set_version(struct nlattr *data[], 1757 struct __ip6_tnl_parm *parms) 1758 { 1759 if (!data) 1760 return; 1761 1762 parms->erspan_ver = 1; 1763 if (data[IFLA_GRE_ERSPAN_VER]) 1764 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1765 1766 if (parms->erspan_ver == 1) { 1767 if (data[IFLA_GRE_ERSPAN_INDEX]) 1768 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1769 } else if (parms->erspan_ver == 2) { 1770 if (data[IFLA_GRE_ERSPAN_DIR]) 1771 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1772 if (data[IFLA_GRE_ERSPAN_HWID]) 1773 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1774 } 1775 } 1776 1777 static void ip6gre_netlink_parms(struct nlattr *data[], 1778 struct __ip6_tnl_parm *parms) 1779 { 1780 memset(parms, 0, sizeof(*parms)); 1781 1782 if (!data) 1783 return; 1784 1785 if (data[IFLA_GRE_LINK]) 1786 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1787 1788 if (data[IFLA_GRE_IFLAGS]) 1789 parms->i_flags = gre_flags_to_tnl_flags( 1790 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1791 1792 if (data[IFLA_GRE_OFLAGS]) 1793 parms->o_flags = gre_flags_to_tnl_flags( 1794 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1795 1796 if (data[IFLA_GRE_IKEY]) 1797 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1798 1799 if (data[IFLA_GRE_OKEY]) 1800 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1801 1802 if (data[IFLA_GRE_LOCAL]) 1803 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1804 1805 if (data[IFLA_GRE_REMOTE]) 1806 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1807 1808 if (data[IFLA_GRE_TTL]) 1809 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1810 1811 if (data[IFLA_GRE_ENCAP_LIMIT]) 1812 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1813 1814 if (data[IFLA_GRE_FLOWINFO]) 1815 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1816 1817 if (data[IFLA_GRE_FLAGS]) 1818 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1819 1820 if (data[IFLA_GRE_FWMARK]) 1821 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1822 1823 if (data[IFLA_GRE_COLLECT_METADATA]) 1824 parms->collect_md = true; 1825 } 1826 1827 static int ip6gre_tap_init(struct net_device *dev) 1828 { 1829 int ret; 1830 1831 ret = ip6gre_tunnel_init_common(dev); 1832 if (ret) 1833 return ret; 1834 1835 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1836 1837 return 0; 1838 } 1839 1840 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1841 .ndo_init = ip6gre_tap_init, 1842 .ndo_uninit = ip6gre_tunnel_uninit, 1843 .ndo_start_xmit = ip6gre_tunnel_xmit, 1844 .ndo_set_mac_address = eth_mac_addr, 1845 .ndo_validate_addr = eth_validate_addr, 1846 .ndo_change_mtu = ip6_tnl_change_mtu, 1847 .ndo_get_stats64 = dev_get_tstats64, 1848 .ndo_get_iflink = ip6_tnl_get_iflink, 1849 }; 1850 1851 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1852 { 1853 int t_hlen; 1854 1855 tunnel->tun_hlen = 8; 1856 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1857 erspan_hdr_len(tunnel->parms.erspan_ver); 1858 1859 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1860 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1861 return t_hlen; 1862 } 1863 1864 static int ip6erspan_tap_init(struct net_device *dev) 1865 { 1866 struct ip6_tnl *tunnel; 1867 int t_hlen; 1868 int ret; 1869 1870 tunnel = netdev_priv(dev); 1871 1872 tunnel->dev = dev; 1873 tunnel->net = dev_net(dev); 1874 strcpy(tunnel->parms.name, dev->name); 1875 1876 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1877 if (!dev->tstats) 1878 return -ENOMEM; 1879 1880 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1881 if (ret) 1882 goto cleanup_alloc_pcpu_stats; 1883 1884 ret = gro_cells_init(&tunnel->gro_cells, dev); 1885 if (ret) 1886 goto cleanup_dst_cache_init; 1887 1888 t_hlen = ip6erspan_calc_hlen(tunnel); 1889 dev->mtu = ETH_DATA_LEN - t_hlen; 1890 if (dev->type == ARPHRD_ETHER) 1891 dev->mtu -= ETH_HLEN; 1892 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1893 dev->mtu -= 8; 1894 1895 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1896 ip6erspan_tnl_link_config(tunnel, 1); 1897 1898 dev_hold_track(dev, &tunnel->dev_tracker, GFP_KERNEL); 1899 return 0; 1900 1901 cleanup_dst_cache_init: 1902 dst_cache_destroy(&tunnel->dst_cache); 1903 cleanup_alloc_pcpu_stats: 1904 free_percpu(dev->tstats); 1905 dev->tstats = NULL; 1906 return ret; 1907 } 1908 1909 static const struct net_device_ops ip6erspan_netdev_ops = { 1910 .ndo_init = ip6erspan_tap_init, 1911 .ndo_uninit = ip6erspan_tunnel_uninit, 1912 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1913 .ndo_set_mac_address = eth_mac_addr, 1914 .ndo_validate_addr = eth_validate_addr, 1915 .ndo_change_mtu = ip6_tnl_change_mtu, 1916 .ndo_get_stats64 = dev_get_tstats64, 1917 .ndo_get_iflink = ip6_tnl_get_iflink, 1918 }; 1919 1920 static void ip6gre_tap_setup(struct net_device *dev) 1921 { 1922 1923 ether_setup(dev); 1924 1925 dev->max_mtu = 0; 1926 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1927 dev->needs_free_netdev = true; 1928 dev->priv_destructor = ip6gre_dev_free; 1929 1930 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1931 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1932 netif_keep_dst(dev); 1933 } 1934 1935 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1936 struct ip_tunnel_encap *ipencap) 1937 { 1938 bool ret = false; 1939 1940 memset(ipencap, 0, sizeof(*ipencap)); 1941 1942 if (!data) 1943 return ret; 1944 1945 if (data[IFLA_GRE_ENCAP_TYPE]) { 1946 ret = true; 1947 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1948 } 1949 1950 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1951 ret = true; 1952 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1953 } 1954 1955 if (data[IFLA_GRE_ENCAP_SPORT]) { 1956 ret = true; 1957 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1958 } 1959 1960 if (data[IFLA_GRE_ENCAP_DPORT]) { 1961 ret = true; 1962 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1963 } 1964 1965 return ret; 1966 } 1967 1968 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1969 struct nlattr *tb[], struct nlattr *data[], 1970 struct netlink_ext_ack *extack) 1971 { 1972 struct ip6_tnl *nt; 1973 struct ip_tunnel_encap ipencap; 1974 int err; 1975 1976 nt = netdev_priv(dev); 1977 1978 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1979 int err = ip6_tnl_encap_setup(nt, &ipencap); 1980 1981 if (err < 0) 1982 return err; 1983 } 1984 1985 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1986 eth_hw_addr_random(dev); 1987 1988 nt->dev = dev; 1989 nt->net = dev_net(dev); 1990 1991 err = register_netdevice(dev); 1992 if (err) 1993 goto out; 1994 1995 if (tb[IFLA_MTU]) 1996 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1997 1998 out: 1999 return err; 2000 } 2001 2002 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 2003 struct nlattr *tb[], struct nlattr *data[], 2004 struct netlink_ext_ack *extack) 2005 { 2006 struct ip6_tnl *nt = netdev_priv(dev); 2007 struct net *net = dev_net(dev); 2008 struct ip6gre_net *ign; 2009 int err; 2010 2011 ip6gre_netlink_parms(data, &nt->parms); 2012 ign = net_generic(net, ip6gre_net_id); 2013 2014 if (nt->parms.collect_md) { 2015 if (rtnl_dereference(ign->collect_md_tun)) 2016 return -EEXIST; 2017 } else { 2018 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2019 return -EEXIST; 2020 } 2021 2022 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2023 if (!err) { 2024 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2025 ip6gre_tunnel_link_md(ign, nt); 2026 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2027 } 2028 return err; 2029 } 2030 2031 static struct ip6_tnl * 2032 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2033 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2034 struct netlink_ext_ack *extack) 2035 { 2036 struct ip6_tnl *t, *nt = netdev_priv(dev); 2037 struct net *net = nt->net; 2038 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2039 struct ip_tunnel_encap ipencap; 2040 2041 if (dev == ign->fb_tunnel_dev) 2042 return ERR_PTR(-EINVAL); 2043 2044 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2045 int err = ip6_tnl_encap_setup(nt, &ipencap); 2046 2047 if (err < 0) 2048 return ERR_PTR(err); 2049 } 2050 2051 ip6gre_netlink_parms(data, p_p); 2052 2053 t = ip6gre_tunnel_locate(net, p_p, 0); 2054 2055 if (t) { 2056 if (t->dev != dev) 2057 return ERR_PTR(-EEXIST); 2058 } else { 2059 t = nt; 2060 } 2061 2062 return t; 2063 } 2064 2065 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2066 struct nlattr *data[], 2067 struct netlink_ext_ack *extack) 2068 { 2069 struct ip6_tnl *t = netdev_priv(dev); 2070 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2071 struct __ip6_tnl_parm p; 2072 2073 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2074 if (IS_ERR(t)) 2075 return PTR_ERR(t); 2076 2077 ip6gre_tunnel_unlink_md(ign, t); 2078 ip6gre_tunnel_unlink(ign, t); 2079 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2080 ip6gre_tunnel_link_md(ign, t); 2081 ip6gre_tunnel_link(ign, t); 2082 return 0; 2083 } 2084 2085 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2086 { 2087 struct net *net = dev_net(dev); 2088 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2089 2090 if (dev != ign->fb_tunnel_dev) 2091 unregister_netdevice_queue(dev, head); 2092 } 2093 2094 static size_t ip6gre_get_size(const struct net_device *dev) 2095 { 2096 return 2097 /* IFLA_GRE_LINK */ 2098 nla_total_size(4) + 2099 /* IFLA_GRE_IFLAGS */ 2100 nla_total_size(2) + 2101 /* IFLA_GRE_OFLAGS */ 2102 nla_total_size(2) + 2103 /* IFLA_GRE_IKEY */ 2104 nla_total_size(4) + 2105 /* IFLA_GRE_OKEY */ 2106 nla_total_size(4) + 2107 /* IFLA_GRE_LOCAL */ 2108 nla_total_size(sizeof(struct in6_addr)) + 2109 /* IFLA_GRE_REMOTE */ 2110 nla_total_size(sizeof(struct in6_addr)) + 2111 /* IFLA_GRE_TTL */ 2112 nla_total_size(1) + 2113 /* IFLA_GRE_ENCAP_LIMIT */ 2114 nla_total_size(1) + 2115 /* IFLA_GRE_FLOWINFO */ 2116 nla_total_size(4) + 2117 /* IFLA_GRE_FLAGS */ 2118 nla_total_size(4) + 2119 /* IFLA_GRE_ENCAP_TYPE */ 2120 nla_total_size(2) + 2121 /* IFLA_GRE_ENCAP_FLAGS */ 2122 nla_total_size(2) + 2123 /* IFLA_GRE_ENCAP_SPORT */ 2124 nla_total_size(2) + 2125 /* IFLA_GRE_ENCAP_DPORT */ 2126 nla_total_size(2) + 2127 /* IFLA_GRE_COLLECT_METADATA */ 2128 nla_total_size(0) + 2129 /* IFLA_GRE_FWMARK */ 2130 nla_total_size(4) + 2131 /* IFLA_GRE_ERSPAN_INDEX */ 2132 nla_total_size(4) + 2133 0; 2134 } 2135 2136 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2137 { 2138 struct ip6_tnl *t = netdev_priv(dev); 2139 struct __ip6_tnl_parm *p = &t->parms; 2140 __be16 o_flags = p->o_flags; 2141 2142 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2143 if (!p->collect_md) 2144 o_flags |= TUNNEL_KEY; 2145 2146 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2147 goto nla_put_failure; 2148 2149 if (p->erspan_ver == 1) { 2150 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2151 goto nla_put_failure; 2152 } else { 2153 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2154 goto nla_put_failure; 2155 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2156 goto nla_put_failure; 2157 } 2158 } 2159 2160 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2161 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2162 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2163 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2164 gre_tnl_flags_to_gre_flags(o_flags)) || 2165 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2166 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2167 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2168 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2169 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2170 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2171 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2172 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2173 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2174 goto nla_put_failure; 2175 2176 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2177 t->encap.type) || 2178 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2179 t->encap.sport) || 2180 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2181 t->encap.dport) || 2182 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2183 t->encap.flags)) 2184 goto nla_put_failure; 2185 2186 if (p->collect_md) { 2187 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2188 goto nla_put_failure; 2189 } 2190 2191 return 0; 2192 2193 nla_put_failure: 2194 return -EMSGSIZE; 2195 } 2196 2197 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2198 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2199 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2200 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2201 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2202 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2203 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2204 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2205 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2206 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2207 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2208 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2209 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2210 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2211 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2212 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2213 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2214 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2215 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2216 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2217 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2218 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2219 }; 2220 2221 static void ip6erspan_tap_setup(struct net_device *dev) 2222 { 2223 ether_setup(dev); 2224 2225 dev->max_mtu = 0; 2226 dev->netdev_ops = &ip6erspan_netdev_ops; 2227 dev->needs_free_netdev = true; 2228 dev->priv_destructor = ip6gre_dev_free; 2229 2230 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2231 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2232 netif_keep_dst(dev); 2233 } 2234 2235 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2236 struct nlattr *tb[], struct nlattr *data[], 2237 struct netlink_ext_ack *extack) 2238 { 2239 struct ip6_tnl *nt = netdev_priv(dev); 2240 struct net *net = dev_net(dev); 2241 struct ip6gre_net *ign; 2242 int err; 2243 2244 ip6gre_netlink_parms(data, &nt->parms); 2245 ip6erspan_set_version(data, &nt->parms); 2246 ign = net_generic(net, ip6gre_net_id); 2247 2248 if (nt->parms.collect_md) { 2249 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2250 return -EEXIST; 2251 } else { 2252 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2253 return -EEXIST; 2254 } 2255 2256 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2257 if (!err) { 2258 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2259 ip6erspan_tunnel_link_md(ign, nt); 2260 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2261 } 2262 return err; 2263 } 2264 2265 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2266 { 2267 ip6gre_tnl_link_config_common(t); 2268 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2269 } 2270 2271 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2272 const struct __ip6_tnl_parm *p, int set_mtu) 2273 { 2274 ip6gre_tnl_copy_tnl_parm(t, p); 2275 ip6erspan_tnl_link_config(t, set_mtu); 2276 return 0; 2277 } 2278 2279 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2280 struct nlattr *data[], 2281 struct netlink_ext_ack *extack) 2282 { 2283 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2284 struct __ip6_tnl_parm p; 2285 struct ip6_tnl *t; 2286 2287 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2288 if (IS_ERR(t)) 2289 return PTR_ERR(t); 2290 2291 ip6erspan_set_version(data, &p); 2292 ip6gre_tunnel_unlink_md(ign, t); 2293 ip6gre_tunnel_unlink(ign, t); 2294 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2295 ip6erspan_tunnel_link_md(ign, t); 2296 ip6gre_tunnel_link(ign, t); 2297 return 0; 2298 } 2299 2300 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2301 .kind = "ip6gre", 2302 .maxtype = IFLA_GRE_MAX, 2303 .policy = ip6gre_policy, 2304 .priv_size = sizeof(struct ip6_tnl), 2305 .setup = ip6gre_tunnel_setup, 2306 .validate = ip6gre_tunnel_validate, 2307 .newlink = ip6gre_newlink, 2308 .changelink = ip6gre_changelink, 2309 .dellink = ip6gre_dellink, 2310 .get_size = ip6gre_get_size, 2311 .fill_info = ip6gre_fill_info, 2312 .get_link_net = ip6_tnl_get_link_net, 2313 }; 2314 2315 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2316 .kind = "ip6gretap", 2317 .maxtype = IFLA_GRE_MAX, 2318 .policy = ip6gre_policy, 2319 .priv_size = sizeof(struct ip6_tnl), 2320 .setup = ip6gre_tap_setup, 2321 .validate = ip6gre_tap_validate, 2322 .newlink = ip6gre_newlink, 2323 .changelink = ip6gre_changelink, 2324 .get_size = ip6gre_get_size, 2325 .fill_info = ip6gre_fill_info, 2326 .get_link_net = ip6_tnl_get_link_net, 2327 }; 2328 2329 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2330 .kind = "ip6erspan", 2331 .maxtype = IFLA_GRE_MAX, 2332 .policy = ip6gre_policy, 2333 .priv_size = sizeof(struct ip6_tnl), 2334 .setup = ip6erspan_tap_setup, 2335 .validate = ip6erspan_tap_validate, 2336 .newlink = ip6erspan_newlink, 2337 .changelink = ip6erspan_changelink, 2338 .get_size = ip6gre_get_size, 2339 .fill_info = ip6gre_fill_info, 2340 .get_link_net = ip6_tnl_get_link_net, 2341 }; 2342 2343 /* 2344 * And now the modules code and kernel interface. 2345 */ 2346 2347 static int __init ip6gre_init(void) 2348 { 2349 int err; 2350 2351 pr_info("GRE over IPv6 tunneling driver\n"); 2352 2353 err = register_pernet_device(&ip6gre_net_ops); 2354 if (err < 0) 2355 return err; 2356 2357 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2358 if (err < 0) { 2359 pr_info("%s: can't add protocol\n", __func__); 2360 goto add_proto_failed; 2361 } 2362 2363 err = rtnl_link_register(&ip6gre_link_ops); 2364 if (err < 0) 2365 goto rtnl_link_failed; 2366 2367 err = rtnl_link_register(&ip6gre_tap_ops); 2368 if (err < 0) 2369 goto tap_ops_failed; 2370 2371 err = rtnl_link_register(&ip6erspan_tap_ops); 2372 if (err < 0) 2373 goto erspan_link_failed; 2374 2375 out: 2376 return err; 2377 2378 erspan_link_failed: 2379 rtnl_link_unregister(&ip6gre_tap_ops); 2380 tap_ops_failed: 2381 rtnl_link_unregister(&ip6gre_link_ops); 2382 rtnl_link_failed: 2383 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2384 add_proto_failed: 2385 unregister_pernet_device(&ip6gre_net_ops); 2386 goto out; 2387 } 2388 2389 static void __exit ip6gre_fini(void) 2390 { 2391 rtnl_link_unregister(&ip6gre_tap_ops); 2392 rtnl_link_unregister(&ip6gre_link_ops); 2393 rtnl_link_unregister(&ip6erspan_tap_ops); 2394 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2395 unregister_pernet_device(&ip6gre_net_ops); 2396 } 2397 2398 module_init(ip6gre_init); 2399 module_exit(ip6gre_fini); 2400 MODULE_LICENSE("GPL"); 2401 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2402 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2403 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2404 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2405 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2406 MODULE_ALIAS_NETDEV("ip6gre0"); 2407