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/netdev_lock.h> 47 #include <net/rtnetlink.h> 48 49 #include <net/ipv6.h> 50 #include <net/ip6_fib.h> 51 #include <net/ip6_route.h> 52 #include <net/ip6_tunnel.h> 53 #include <net/gre.h> 54 #include <net/erspan.h> 55 #include <net/dst_metadata.h> 56 57 58 static bool log_ecn_error = true; 59 module_param(log_ecn_error, bool, 0644); 60 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 61 62 #define IP6_GRE_HASH_SIZE_SHIFT 5 63 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT) 64 65 static unsigned int ip6gre_net_id __read_mostly; 66 struct ip6gre_net { 67 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE]; 68 69 struct ip6_tnl __rcu *collect_md_tun; 70 struct ip6_tnl __rcu *collect_md_tun_erspan; 71 struct net_device *fb_tunnel_dev; 72 }; 73 74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 75 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 76 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly; 77 static int ip6gre_tunnel_init(struct net_device *dev); 78 static void ip6gre_tunnel_setup(struct net_device *dev); 79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t); 80 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu); 81 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu); 82 83 /* Tunnel hash table */ 84 85 /* 86 4 hash tables: 87 88 3: (remote,local) 89 2: (remote,*) 90 1: (*,local) 91 0: (*,*) 92 93 We require exact key match i.e. if a key is present in packet 94 it will match only tunnel with the same key; if it is not present, 95 it will match only keyless tunnel. 96 97 All keysless packets, if not matched configured keyless tunnels 98 will match fallback tunnel. 99 */ 100 101 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 102 static u32 HASH_ADDR(const struct in6_addr *addr) 103 { 104 u32 hash = ipv6_addr_hash(addr); 105 106 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 107 } 108 109 #define tunnels_r_l tunnels[3] 110 #define tunnels_r tunnels[2] 111 #define tunnels_l tunnels[1] 112 #define tunnels_wc tunnels[0] 113 114 /* Given src, dst and key, find appropriate for input tunnel. */ 115 116 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 117 const struct in6_addr *remote, const struct in6_addr *local, 118 __be32 key, __be16 gre_proto) 119 { 120 struct net *net = dev_net(dev); 121 int link = dev->ifindex; 122 unsigned int h0 = HASH_ADDR(remote); 123 unsigned int h1 = HASH_KEY(key); 124 struct ip6_tnl *t, *cand = NULL; 125 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 126 int dev_type = (gre_proto == htons(ETH_P_TEB) || 127 gre_proto == htons(ETH_P_ERSPAN) || 128 gre_proto == htons(ETH_P_ERSPAN2)) ? 129 ARPHRD_ETHER : ARPHRD_IP6GRE; 130 int score, cand_score = 4; 131 struct net_device *ndev; 132 133 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 134 if (!ipv6_addr_equal(local, &t->parms.laddr) || 135 !ipv6_addr_equal(remote, &t->parms.raddr) || 136 key != t->parms.i_key || 137 !(t->dev->flags & IFF_UP)) 138 continue; 139 140 if (t->dev->type != ARPHRD_IP6GRE && 141 t->dev->type != dev_type) 142 continue; 143 144 score = 0; 145 if (t->parms.link != link) 146 score |= 1; 147 if (t->dev->type != dev_type) 148 score |= 2; 149 if (score == 0) 150 return t; 151 152 if (score < cand_score) { 153 cand = t; 154 cand_score = score; 155 } 156 } 157 158 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 159 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 160 key != t->parms.i_key || 161 !(t->dev->flags & IFF_UP)) 162 continue; 163 164 if (t->dev->type != ARPHRD_IP6GRE && 165 t->dev->type != dev_type) 166 continue; 167 168 score = 0; 169 if (t->parms.link != link) 170 score |= 1; 171 if (t->dev->type != dev_type) 172 score |= 2; 173 if (score == 0) 174 return t; 175 176 if (score < cand_score) { 177 cand = t; 178 cand_score = score; 179 } 180 } 181 182 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 183 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 184 (!ipv6_addr_equal(local, &t->parms.raddr) || 185 !ipv6_addr_is_multicast(local))) || 186 key != t->parms.i_key || 187 !(t->dev->flags & IFF_UP)) 188 continue; 189 190 if (t->dev->type != ARPHRD_IP6GRE && 191 t->dev->type != dev_type) 192 continue; 193 194 score = 0; 195 if (t->parms.link != link) 196 score |= 1; 197 if (t->dev->type != dev_type) 198 score |= 2; 199 if (score == 0) 200 return t; 201 202 if (score < cand_score) { 203 cand = t; 204 cand_score = score; 205 } 206 } 207 208 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 209 if (t->parms.i_key != key || 210 !(t->dev->flags & IFF_UP)) 211 continue; 212 213 if (t->dev->type != ARPHRD_IP6GRE && 214 t->dev->type != dev_type) 215 continue; 216 217 score = 0; 218 if (t->parms.link != link) 219 score |= 1; 220 if (t->dev->type != dev_type) 221 score |= 2; 222 if (score == 0) 223 return t; 224 225 if (score < cand_score) { 226 cand = t; 227 cand_score = score; 228 } 229 } 230 231 if (cand) 232 return cand; 233 234 if (gre_proto == htons(ETH_P_ERSPAN) || 235 gre_proto == htons(ETH_P_ERSPAN2)) 236 t = rcu_dereference(ign->collect_md_tun_erspan); 237 else 238 t = rcu_dereference(ign->collect_md_tun); 239 240 if (t && t->dev->flags & IFF_UP) 241 return t; 242 243 ndev = READ_ONCE(ign->fb_tunnel_dev); 244 if (ndev && ndev->flags & IFF_UP) 245 return netdev_priv(ndev); 246 247 return NULL; 248 } 249 250 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 251 const struct __ip6_tnl_parm *p) 252 { 253 const struct in6_addr *remote = &p->raddr; 254 const struct in6_addr *local = &p->laddr; 255 unsigned int h = HASH_KEY(p->i_key); 256 int prio = 0; 257 258 if (!ipv6_addr_any(local)) 259 prio |= 1; 260 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 261 prio |= 2; 262 h ^= HASH_ADDR(remote); 263 } 264 265 return &ign->tunnels[prio][h]; 266 } 267 268 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 269 { 270 if (t->parms.collect_md) 271 rcu_assign_pointer(ign->collect_md_tun, t); 272 } 273 274 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 275 { 276 if (t->parms.collect_md) 277 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 278 } 279 280 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 281 { 282 if (t->parms.collect_md) 283 rcu_assign_pointer(ign->collect_md_tun, NULL); 284 } 285 286 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 287 struct ip6_tnl *t) 288 { 289 if (t->parms.collect_md) 290 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 291 } 292 293 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 294 const struct ip6_tnl *t) 295 { 296 return __ip6gre_bucket(ign, &t->parms); 297 } 298 299 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 300 { 301 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 302 303 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 304 rcu_assign_pointer(*tp, t); 305 } 306 307 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 308 { 309 struct ip6_tnl __rcu **tp; 310 struct ip6_tnl *iter; 311 312 for (tp = ip6gre_bucket(ign, t); 313 (iter = rtnl_dereference(*tp)) != NULL; 314 tp = &iter->next) { 315 if (t == iter) { 316 rcu_assign_pointer(*tp, t->next); 317 break; 318 } 319 } 320 } 321 322 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 323 const struct __ip6_tnl_parm *parms, 324 int type) 325 { 326 const struct in6_addr *remote = &parms->raddr; 327 const struct in6_addr *local = &parms->laddr; 328 __be32 key = parms->i_key; 329 int link = parms->link; 330 struct ip6_tnl *t; 331 struct ip6_tnl __rcu **tp; 332 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 333 334 for (tp = __ip6gre_bucket(ign, parms); 335 (t = rtnl_dereference(*tp)) != NULL; 336 tp = &t->next) 337 if (ipv6_addr_equal(local, &t->parms.laddr) && 338 ipv6_addr_equal(remote, &t->parms.raddr) && 339 key == t->parms.i_key && 340 link == t->parms.link && 341 type == t->dev->type) 342 break; 343 344 return t; 345 } 346 347 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 348 const struct __ip6_tnl_parm *parms, int create) 349 { 350 struct ip6_tnl *t, *nt; 351 struct net_device *dev; 352 char name[IFNAMSIZ]; 353 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 354 355 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 356 if (t && create) 357 return NULL; 358 if (t || !create) 359 return t; 360 361 if (parms->name[0]) { 362 if (!dev_valid_name(parms->name)) 363 return NULL; 364 strscpy(name, parms->name, IFNAMSIZ); 365 } else { 366 strcpy(name, "ip6gre%d"); 367 } 368 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 369 ip6gre_tunnel_setup); 370 if (!dev) 371 return NULL; 372 373 dev_net_set(dev, net); 374 375 nt = netdev_priv(dev); 376 nt->parms = *parms; 377 dev->rtnl_link_ops = &ip6gre_link_ops; 378 379 nt->dev = dev; 380 nt->net = dev_net(dev); 381 382 if (register_netdevice(dev) < 0) 383 goto failed_free; 384 385 ip6gre_tnl_link_config(nt, 1); 386 ip6gre_tunnel_link(ign, nt); 387 return nt; 388 389 failed_free: 390 free_netdev(dev); 391 return NULL; 392 } 393 394 static void ip6erspan_tunnel_uninit(struct net_device *dev) 395 { 396 struct ip6_tnl *t = netdev_priv(dev); 397 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 398 399 ip6erspan_tunnel_unlink_md(ign, t); 400 ip6gre_tunnel_unlink(ign, t); 401 dst_cache_reset(&t->dst_cache); 402 netdev_put(dev, &t->dev_tracker); 403 } 404 405 static void ip6gre_tunnel_uninit(struct net_device *dev) 406 { 407 struct ip6_tnl *t = netdev_priv(dev); 408 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 409 410 ip6gre_tunnel_unlink_md(ign, t); 411 ip6gre_tunnel_unlink(ign, t); 412 if (ign->fb_tunnel_dev == dev) 413 WRITE_ONCE(ign->fb_tunnel_dev, NULL); 414 dst_cache_reset(&t->dst_cache); 415 netdev_put(dev, &t->dev_tracker); 416 } 417 418 419 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 420 u8 type, u8 code, int offset, __be32 info) 421 { 422 struct net *net = dev_net(skb->dev); 423 const struct ipv6hdr *ipv6h; 424 struct tnl_ptk_info tpi; 425 struct ip6_tnl *t; 426 427 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 428 offset) < 0) 429 return -EINVAL; 430 431 ipv6h = (const struct ipv6hdr *)skb->data; 432 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 433 tpi.key, tpi.proto); 434 if (!t) 435 return -ENOENT; 436 437 switch (type) { 438 case ICMPV6_DEST_UNREACH: 439 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 440 t->parms.name); 441 if (code != ICMPV6_PORT_UNREACH) 442 break; 443 return 0; 444 case ICMPV6_TIME_EXCEED: 445 if (code == ICMPV6_EXC_HOPLIMIT) { 446 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 447 t->parms.name); 448 break; 449 } 450 return 0; 451 case ICMPV6_PARAMPROB: { 452 struct ipv6_tlv_tnl_enc_lim *tel; 453 __u32 teli; 454 455 teli = 0; 456 if (code == ICMPV6_HDR_FIELD) 457 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 458 459 if (teli && teli == be32_to_cpu(info) - 2) { 460 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 461 if (tel->encap_limit == 0) { 462 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 463 t->parms.name); 464 } 465 } else { 466 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 467 t->parms.name); 468 } 469 return 0; 470 } 471 case ICMPV6_PKT_TOOBIG: 472 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 473 return 0; 474 case NDISC_REDIRECT: 475 ip6_redirect(skb, net, skb->dev->ifindex, 0, 476 sock_net_uid(net, NULL)); 477 return 0; 478 } 479 480 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 481 t->err_count++; 482 else 483 t->err_count = 1; 484 t->err_time = jiffies; 485 486 return 0; 487 } 488 489 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 490 { 491 const struct ipv6hdr *ipv6h; 492 struct ip6_tnl *tunnel; 493 494 ipv6h = ipv6_hdr(skb); 495 tunnel = ip6gre_tunnel_lookup(skb->dev, 496 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 497 tpi->proto); 498 if (tunnel) { 499 if (tunnel->parms.collect_md) { 500 IP_TUNNEL_DECLARE_FLAGS(flags); 501 struct metadata_dst *tun_dst; 502 __be64 tun_id; 503 504 ip_tunnel_flags_copy(flags, tpi->flags); 505 tun_id = key32_to_tunnel_id(tpi->key); 506 507 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 508 if (!tun_dst) 509 return PACKET_REJECT; 510 511 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 512 } else { 513 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 514 } 515 516 return PACKET_RCVD; 517 } 518 519 return PACKET_REJECT; 520 } 521 522 static int ip6erspan_rcv(struct sk_buff *skb, 523 struct tnl_ptk_info *tpi, 524 int gre_hdr_len) 525 { 526 struct erspan_base_hdr *ershdr; 527 const struct ipv6hdr *ipv6h; 528 struct erspan_md2 *md2; 529 struct ip6_tnl *tunnel; 530 u8 ver; 531 532 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr)))) 533 return PACKET_REJECT; 534 535 ipv6h = ipv6_hdr(skb); 536 ershdr = (struct erspan_base_hdr *)skb->data; 537 ver = ershdr->ver; 538 539 tunnel = ip6gre_tunnel_lookup(skb->dev, 540 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 541 tpi->proto); 542 if (tunnel) { 543 int len = erspan_hdr_len(ver); 544 545 if (unlikely(!pskb_may_pull(skb, len))) 546 return PACKET_REJECT; 547 548 if (__iptunnel_pull_header(skb, len, 549 htons(ETH_P_TEB), 550 false, false) < 0) 551 return PACKET_REJECT; 552 553 if (tunnel->parms.collect_md) { 554 struct erspan_metadata *pkt_md, *md; 555 IP_TUNNEL_DECLARE_FLAGS(flags); 556 struct metadata_dst *tun_dst; 557 struct ip_tunnel_info *info; 558 unsigned char *gh; 559 __be64 tun_id; 560 561 __set_bit(IP_TUNNEL_KEY_BIT, tpi->flags); 562 ip_tunnel_flags_copy(flags, tpi->flags); 563 tun_id = key32_to_tunnel_id(tpi->key); 564 565 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 566 sizeof(*md)); 567 if (!tun_dst) 568 return PACKET_REJECT; 569 570 /* skb can be uncloned in __iptunnel_pull_header, so 571 * old pkt_md is no longer valid and we need to reset 572 * it 573 */ 574 gh = skb_network_header(skb) + 575 skb_network_header_len(skb); 576 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 577 sizeof(*ershdr)); 578 info = &tun_dst->u.tun_info; 579 md = ip_tunnel_info_opts(info); 580 md->version = ver; 581 md2 = &md->u.md2; 582 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 583 ERSPAN_V2_MDSIZE); 584 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, 585 info->key.tun_flags); 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 int prepare_ip6gre_xmit_other(struct sk_buff *skb, 710 struct net_device *dev, 711 struct flowi6 *fl6, __u8 *dsfield, 712 int *encap_limit) 713 { 714 struct ip6_tnl *t = netdev_priv(dev); 715 716 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 717 *encap_limit = t->parms.encap_limit; 718 719 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 720 721 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 722 *dsfield = 0; 723 else 724 *dsfield = ip6_tclass(t->parms.flowinfo); 725 726 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 727 fl6->flowi6_mark = skb->mark; 728 else 729 fl6->flowi6_mark = t->parms.fwmark; 730 731 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 732 733 return 0; 734 } 735 736 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb) 737 { 738 struct ip_tunnel_info *tun_info; 739 740 tun_info = skb_tunnel_info(skb); 741 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))) 742 return ERR_PTR(-EINVAL); 743 744 return tun_info; 745 } 746 747 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 748 struct net_device *dev, __u8 dsfield, 749 struct flowi6 *fl6, int encap_limit, 750 __u32 *pmtu, __be16 proto) 751 { 752 struct ip6_tnl *tunnel = netdev_priv(dev); 753 IP_TUNNEL_DECLARE_FLAGS(flags); 754 __be16 protocol; 755 756 if (dev->type == ARPHRD_ETHER) 757 IPCB(skb)->flags = 0; 758 759 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 760 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 761 else 762 fl6->daddr = tunnel->parms.raddr; 763 764 /* Push GRE header. */ 765 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 766 767 if (tunnel->parms.collect_md) { 768 struct ip_tunnel_info *tun_info; 769 const struct ip_tunnel_key *key; 770 int tun_hlen; 771 772 tun_info = skb_tunnel_info_txcheck(skb); 773 if (IS_ERR(tun_info) || 774 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 775 return -EINVAL; 776 777 key = &tun_info->key; 778 memset(fl6, 0, sizeof(*fl6)); 779 fl6->flowi6_proto = IPPROTO_GRE; 780 fl6->daddr = key->u.ipv6.dst; 781 fl6->flowlabel = key->label; 782 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 783 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id); 784 785 dsfield = key->tos; 786 ip_tunnel_flags_zero(flags); 787 __set_bit(IP_TUNNEL_CSUM_BIT, flags); 788 __set_bit(IP_TUNNEL_KEY_BIT, flags); 789 __set_bit(IP_TUNNEL_SEQ_BIT, flags); 790 ip_tunnel_flags_and(flags, flags, key->tun_flags); 791 tun_hlen = gre_calc_hlen(flags); 792 793 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen)) 794 return -ENOMEM; 795 796 gre_build_header(skb, tun_hlen, 797 flags, protocol, 798 tunnel_id_to_key32(tun_info->key.tun_id), 799 test_bit(IP_TUNNEL_SEQ_BIT, flags) ? 800 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 801 0); 802 803 } else { 804 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 805 return -ENOMEM; 806 807 ip_tunnel_flags_copy(flags, tunnel->parms.o_flags); 808 809 gre_build_header(skb, tunnel->tun_hlen, flags, 810 protocol, tunnel->parms.o_key, 811 test_bit(IP_TUNNEL_SEQ_BIT, flags) ? 812 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 813 0); 814 } 815 816 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 817 NEXTHDR_GRE); 818 } 819 820 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 821 { 822 struct ip6_tnl *t = netdev_priv(dev); 823 int encap_limit = -1; 824 struct flowi6 fl6; 825 __u8 dsfield = 0; 826 __u32 mtu; 827 int err; 828 829 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 830 831 if (!t->parms.collect_md) 832 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 833 &dsfield, &encap_limit); 834 835 err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT, 836 t->parms.o_flags)); 837 if (err) 838 return -1; 839 840 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 841 skb->protocol); 842 if (err != 0) { 843 /* XXX: send ICMP error even if DF is not set. */ 844 if (err == -EMSGSIZE) 845 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 846 htonl(mtu)); 847 return -1; 848 } 849 850 return 0; 851 } 852 853 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 854 { 855 struct ip6_tnl *t = netdev_priv(dev); 856 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 857 int encap_limit = -1; 858 struct flowi6 fl6; 859 __u8 dsfield = 0; 860 __u32 mtu; 861 int err; 862 863 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 864 return -1; 865 866 if (!t->parms.collect_md && 867 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 868 return -1; 869 870 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT, 871 t->parms.o_flags))) 872 return -1; 873 874 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 875 &mtu, skb->protocol); 876 if (err != 0) { 877 if (err == -EMSGSIZE) 878 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 879 return -1; 880 } 881 882 return 0; 883 } 884 885 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 886 { 887 struct ip6_tnl *t = netdev_priv(dev); 888 int encap_limit = -1; 889 struct flowi6 fl6; 890 __u8 dsfield = 0; 891 __u32 mtu; 892 int err; 893 894 if (!t->parms.collect_md && 895 prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit)) 896 return -1; 897 898 err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT, 899 t->parms.o_flags)); 900 if (err) 901 return err; 902 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol); 903 904 return err; 905 } 906 907 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 908 struct net_device *dev) 909 { 910 struct ip6_tnl *t = netdev_priv(dev); 911 __be16 payload_protocol; 912 int ret; 913 914 if (!pskb_inet_may_pull(skb)) 915 goto tx_err; 916 917 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 918 goto tx_err; 919 920 payload_protocol = skb_protocol(skb, true); 921 switch (payload_protocol) { 922 case htons(ETH_P_IP): 923 ret = ip6gre_xmit_ipv4(skb, dev); 924 break; 925 case htons(ETH_P_IPV6): 926 ret = ip6gre_xmit_ipv6(skb, dev); 927 break; 928 default: 929 ret = ip6gre_xmit_other(skb, dev); 930 break; 931 } 932 933 if (ret < 0) 934 goto tx_err; 935 936 return NETDEV_TX_OK; 937 938 tx_err: 939 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 940 DEV_STATS_INC(dev, tx_errors); 941 DEV_STATS_INC(dev, tx_dropped); 942 kfree_skb(skb); 943 return NETDEV_TX_OK; 944 } 945 946 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 947 struct net_device *dev) 948 { 949 struct ip_tunnel_info *tun_info = NULL; 950 struct ip6_tnl *t = netdev_priv(dev); 951 struct dst_entry *dst = skb_dst(skb); 952 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 953 bool truncate = false; 954 int encap_limit = -1; 955 __u8 dsfield = false; 956 struct flowi6 fl6; 957 int err = -EINVAL; 958 __be16 proto; 959 __u32 mtu; 960 int nhoff; 961 962 if (!pskb_inet_may_pull(skb)) 963 goto tx_err; 964 965 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 966 goto tx_err; 967 968 if (gre_handle_offloads(skb, false)) 969 goto tx_err; 970 971 if (skb->len > dev->mtu + dev->hard_header_len) { 972 if (pskb_trim(skb, dev->mtu + dev->hard_header_len)) 973 goto tx_err; 974 truncate = true; 975 } 976 977 nhoff = skb_network_offset(skb); 978 if (skb->protocol == htons(ETH_P_IP) && 979 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 980 truncate = true; 981 982 if (skb->protocol == htons(ETH_P_IPV6)) { 983 int thoff; 984 985 if (skb_transport_header_was_set(skb)) 986 thoff = skb_transport_offset(skb); 987 else 988 thoff = nhoff + sizeof(struct ipv6hdr); 989 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff) 990 truncate = true; 991 } 992 993 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 994 goto tx_err; 995 996 __clear_bit(IP_TUNNEL_KEY_BIT, t->parms.o_flags); 997 IPCB(skb)->flags = 0; 998 999 /* For collect_md mode, derive fl6 from the tunnel key, 1000 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 1001 */ 1002 if (t->parms.collect_md) { 1003 const struct ip_tunnel_key *key; 1004 struct erspan_metadata *md; 1005 __be32 tun_id; 1006 1007 tun_info = skb_tunnel_info_txcheck(skb); 1008 if (IS_ERR(tun_info) || 1009 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 1010 goto tx_err; 1011 1012 key = &tun_info->key; 1013 memset(&fl6, 0, sizeof(fl6)); 1014 fl6.flowi6_proto = IPPROTO_GRE; 1015 fl6.daddr = key->u.ipv6.dst; 1016 fl6.flowlabel = key->label; 1017 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1018 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 1019 1020 dsfield = key->tos; 1021 if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, 1022 tun_info->key.tun_flags)) 1023 goto tx_err; 1024 if (tun_info->options_len < sizeof(*md)) 1025 goto tx_err; 1026 md = ip_tunnel_info_opts(tun_info); 1027 1028 tun_id = tunnel_id_to_key32(key->tun_id); 1029 if (md->version == 1) { 1030 erspan_build_header(skb, 1031 ntohl(tun_id), 1032 ntohl(md->u.index), truncate, 1033 false); 1034 proto = htons(ETH_P_ERSPAN); 1035 } else if (md->version == 2) { 1036 erspan_build_header_v2(skb, 1037 ntohl(tun_id), 1038 md->u.md2.dir, 1039 get_hwid(&md->u.md2), 1040 truncate, false); 1041 proto = htons(ETH_P_ERSPAN2); 1042 } else { 1043 goto tx_err; 1044 } 1045 } else { 1046 switch (skb->protocol) { 1047 case htons(ETH_P_IP): 1048 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1049 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1050 &dsfield, &encap_limit); 1051 break; 1052 case htons(ETH_P_IPV6): 1053 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1054 goto tx_err; 1055 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1056 &dsfield, &encap_limit)) 1057 goto tx_err; 1058 break; 1059 default: 1060 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1061 break; 1062 } 1063 1064 if (t->parms.erspan_ver == 1) { 1065 erspan_build_header(skb, ntohl(t->parms.o_key), 1066 t->parms.index, 1067 truncate, false); 1068 proto = htons(ETH_P_ERSPAN); 1069 } else if (t->parms.erspan_ver == 2) { 1070 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1071 t->parms.dir, 1072 t->parms.hwid, 1073 truncate, false); 1074 proto = htons(ETH_P_ERSPAN2); 1075 } else { 1076 goto tx_err; 1077 } 1078 1079 fl6.daddr = t->parms.raddr; 1080 } 1081 1082 /* Push GRE header. */ 1083 __set_bit(IP_TUNNEL_SEQ_BIT, flags); 1084 gre_build_header(skb, 8, flags, proto, 0, 1085 htonl(atomic_fetch_inc(&t->o_seqno))); 1086 1087 /* TooBig packet may have updated dst->dev's mtu */ 1088 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1089 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1090 1091 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1092 NEXTHDR_GRE); 1093 if (err != 0) { 1094 /* XXX: send ICMP error even if DF is not set. */ 1095 if (err == -EMSGSIZE) { 1096 if (skb->protocol == htons(ETH_P_IP)) 1097 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1098 ICMP_FRAG_NEEDED, htonl(mtu)); 1099 else 1100 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1101 } 1102 1103 goto tx_err; 1104 } 1105 return NETDEV_TX_OK; 1106 1107 tx_err: 1108 if (!IS_ERR(tun_info)) 1109 DEV_STATS_INC(dev, tx_errors); 1110 DEV_STATS_INC(dev, tx_dropped); 1111 kfree_skb(skb); 1112 return NETDEV_TX_OK; 1113 } 1114 1115 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1116 { 1117 struct net_device *dev = t->dev; 1118 struct __ip6_tnl_parm *p = &t->parms; 1119 struct flowi6 *fl6 = &t->fl.u.ip6; 1120 1121 if (dev->type != ARPHRD_ETHER) { 1122 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1123 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1124 } 1125 1126 /* Set up flowi template */ 1127 fl6->saddr = p->laddr; 1128 fl6->daddr = p->raddr; 1129 fl6->flowi6_oif = p->link; 1130 fl6->flowlabel = 0; 1131 fl6->flowi6_proto = IPPROTO_GRE; 1132 fl6->fl6_gre_key = t->parms.o_key; 1133 1134 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1135 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1136 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1137 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1138 1139 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1140 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1141 1142 if (p->flags&IP6_TNL_F_CAP_XMIT && 1143 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1144 dev->flags |= IFF_POINTOPOINT; 1145 else 1146 dev->flags &= ~IFF_POINTOPOINT; 1147 } 1148 1149 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1150 int t_hlen) 1151 { 1152 const struct __ip6_tnl_parm *p = &t->parms; 1153 struct net_device *dev = t->dev; 1154 1155 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1156 int strict = (ipv6_addr_type(&p->raddr) & 1157 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1158 1159 struct rt6_info *rt = rt6_lookup(t->net, 1160 &p->raddr, &p->laddr, 1161 p->link, NULL, strict); 1162 1163 if (!rt) 1164 return; 1165 1166 if (rt->dst.dev) { 1167 unsigned short dst_len = rt->dst.dev->hard_header_len + 1168 t_hlen; 1169 1170 if (t->dev->header_ops) 1171 dev->hard_header_len = dst_len; 1172 else 1173 dev->needed_headroom = dst_len; 1174 1175 if (set_mtu) { 1176 int mtu = rt->dst.dev->mtu - t_hlen; 1177 1178 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1179 mtu -= 8; 1180 if (dev->type == ARPHRD_ETHER) 1181 mtu -= ETH_HLEN; 1182 1183 if (mtu < IPV6_MIN_MTU) 1184 mtu = IPV6_MIN_MTU; 1185 WRITE_ONCE(dev->mtu, mtu); 1186 } 1187 } 1188 ip6_rt_put(rt); 1189 } 1190 } 1191 1192 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1193 { 1194 int t_hlen; 1195 1196 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1197 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1198 1199 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1200 1201 if (tunnel->dev->header_ops) 1202 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1203 else 1204 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1205 1206 return t_hlen; 1207 } 1208 1209 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1210 { 1211 ip6gre_tnl_link_config_common(t); 1212 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1213 } 1214 1215 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1216 const struct __ip6_tnl_parm *p) 1217 { 1218 t->parms.laddr = p->laddr; 1219 t->parms.raddr = p->raddr; 1220 t->parms.flags = p->flags; 1221 t->parms.hop_limit = p->hop_limit; 1222 t->parms.encap_limit = p->encap_limit; 1223 t->parms.flowinfo = p->flowinfo; 1224 t->parms.link = p->link; 1225 t->parms.proto = p->proto; 1226 t->parms.i_key = p->i_key; 1227 t->parms.o_key = p->o_key; 1228 ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags); 1229 ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags); 1230 t->parms.fwmark = p->fwmark; 1231 t->parms.erspan_ver = p->erspan_ver; 1232 t->parms.index = p->index; 1233 t->parms.dir = p->dir; 1234 t->parms.hwid = p->hwid; 1235 dst_cache_reset(&t->dst_cache); 1236 } 1237 1238 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1239 int set_mtu) 1240 { 1241 ip6gre_tnl_copy_tnl_parm(t, p); 1242 ip6gre_tnl_link_config(t, set_mtu); 1243 return 0; 1244 } 1245 1246 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1247 const struct ip6_tnl_parm2 *u) 1248 { 1249 p->laddr = u->laddr; 1250 p->raddr = u->raddr; 1251 p->flags = u->flags; 1252 p->hop_limit = u->hop_limit; 1253 p->encap_limit = u->encap_limit; 1254 p->flowinfo = u->flowinfo; 1255 p->link = u->link; 1256 p->i_key = u->i_key; 1257 p->o_key = u->o_key; 1258 gre_flags_to_tnl_flags(p->i_flags, u->i_flags); 1259 gre_flags_to_tnl_flags(p->o_flags, u->o_flags); 1260 memcpy(p->name, u->name, sizeof(u->name)); 1261 } 1262 1263 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1264 const struct __ip6_tnl_parm *p) 1265 { 1266 u->proto = IPPROTO_GRE; 1267 u->laddr = p->laddr; 1268 u->raddr = p->raddr; 1269 u->flags = p->flags; 1270 u->hop_limit = p->hop_limit; 1271 u->encap_limit = p->encap_limit; 1272 u->flowinfo = p->flowinfo; 1273 u->link = p->link; 1274 u->i_key = p->i_key; 1275 u->o_key = p->o_key; 1276 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1277 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1278 memcpy(u->name, p->name, sizeof(u->name)); 1279 } 1280 1281 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1282 struct ifreq *ifr, void __user *data, 1283 int cmd) 1284 { 1285 int err = 0; 1286 struct ip6_tnl_parm2 p; 1287 struct __ip6_tnl_parm p1; 1288 struct ip6_tnl *t = netdev_priv(dev); 1289 struct net *net = t->net; 1290 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1291 1292 memset(&p1, 0, sizeof(p1)); 1293 1294 switch (cmd) { 1295 case SIOCGETTUNNEL: 1296 if (dev == ign->fb_tunnel_dev) { 1297 if (copy_from_user(&p, data, sizeof(p))) { 1298 err = -EFAULT; 1299 break; 1300 } 1301 ip6gre_tnl_parm_from_user(&p1, &p); 1302 t = ip6gre_tunnel_locate(net, &p1, 0); 1303 if (!t) 1304 t = netdev_priv(dev); 1305 } 1306 memset(&p, 0, sizeof(p)); 1307 ip6gre_tnl_parm_to_user(&p, &t->parms); 1308 if (copy_to_user(data, &p, sizeof(p))) 1309 err = -EFAULT; 1310 break; 1311 1312 case SIOCADDTUNNEL: 1313 case SIOCCHGTUNNEL: 1314 err = -EPERM; 1315 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1316 goto done; 1317 1318 err = -EFAULT; 1319 if (copy_from_user(&p, data, sizeof(p))) 1320 goto done; 1321 1322 err = -EINVAL; 1323 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1324 goto done; 1325 1326 if (!(p.i_flags&GRE_KEY)) 1327 p.i_key = 0; 1328 if (!(p.o_flags&GRE_KEY)) 1329 p.o_key = 0; 1330 1331 ip6gre_tnl_parm_from_user(&p1, &p); 1332 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1333 1334 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1335 if (t) { 1336 if (t->dev != dev) { 1337 err = -EEXIST; 1338 break; 1339 } 1340 } else { 1341 t = netdev_priv(dev); 1342 1343 ip6gre_tunnel_unlink(ign, t); 1344 synchronize_net(); 1345 ip6gre_tnl_change(t, &p1, 1); 1346 ip6gre_tunnel_link(ign, t); 1347 netdev_state_change(dev); 1348 } 1349 } 1350 1351 if (t) { 1352 err = 0; 1353 1354 memset(&p, 0, sizeof(p)); 1355 ip6gre_tnl_parm_to_user(&p, &t->parms); 1356 if (copy_to_user(data, &p, sizeof(p))) 1357 err = -EFAULT; 1358 } else 1359 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1360 break; 1361 1362 case SIOCDELTUNNEL: 1363 err = -EPERM; 1364 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1365 goto done; 1366 1367 if (dev == ign->fb_tunnel_dev) { 1368 err = -EFAULT; 1369 if (copy_from_user(&p, data, sizeof(p))) 1370 goto done; 1371 err = -ENOENT; 1372 ip6gre_tnl_parm_from_user(&p1, &p); 1373 t = ip6gre_tunnel_locate(net, &p1, 0); 1374 if (!t) 1375 goto done; 1376 err = -EPERM; 1377 if (t == netdev_priv(ign->fb_tunnel_dev)) 1378 goto done; 1379 dev = t->dev; 1380 } 1381 unregister_netdevice(dev); 1382 err = 0; 1383 break; 1384 1385 default: 1386 err = -EINVAL; 1387 } 1388 1389 done: 1390 return err; 1391 } 1392 1393 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1394 unsigned short type, const void *daddr, 1395 const void *saddr, unsigned int len) 1396 { 1397 struct ip6_tnl *t = netdev_priv(dev); 1398 struct ipv6hdr *ipv6h; 1399 __be16 *p; 1400 1401 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1402 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1403 t->fl.u.ip6.flowlabel, 1404 true, &t->fl.u.ip6)); 1405 ipv6h->hop_limit = t->parms.hop_limit; 1406 ipv6h->nexthdr = NEXTHDR_GRE; 1407 ipv6h->saddr = t->parms.laddr; 1408 ipv6h->daddr = t->parms.raddr; 1409 1410 p = (__be16 *)(ipv6h + 1); 1411 p[0] = ip_tunnel_flags_to_be16(t->parms.o_flags); 1412 p[1] = htons(type); 1413 1414 /* 1415 * Set the source hardware address. 1416 */ 1417 1418 if (saddr) 1419 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1420 if (daddr) 1421 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1422 if (!ipv6_addr_any(&ipv6h->daddr)) 1423 return t->hlen; 1424 1425 return -t->hlen; 1426 } 1427 1428 static const struct header_ops ip6gre_header_ops = { 1429 .create = ip6gre_header, 1430 }; 1431 1432 static const struct net_device_ops ip6gre_netdev_ops = { 1433 .ndo_init = ip6gre_tunnel_init, 1434 .ndo_uninit = ip6gre_tunnel_uninit, 1435 .ndo_start_xmit = ip6gre_tunnel_xmit, 1436 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1437 .ndo_change_mtu = ip6_tnl_change_mtu, 1438 .ndo_get_iflink = ip6_tnl_get_iflink, 1439 }; 1440 1441 static void ip6gre_dev_free(struct net_device *dev) 1442 { 1443 struct ip6_tnl *t = netdev_priv(dev); 1444 1445 gro_cells_destroy(&t->gro_cells); 1446 dst_cache_destroy(&t->dst_cache); 1447 } 1448 1449 static void ip6gre_tunnel_setup(struct net_device *dev) 1450 { 1451 dev->netdev_ops = &ip6gre_netdev_ops; 1452 dev->needs_free_netdev = true; 1453 dev->priv_destructor = ip6gre_dev_free; 1454 1455 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 1456 dev->type = ARPHRD_IP6GRE; 1457 1458 dev->flags |= IFF_NOARP; 1459 dev->addr_len = sizeof(struct in6_addr); 1460 netif_keep_dst(dev); 1461 /* This perm addr will be used as interface identifier by IPv6 */ 1462 dev->addr_assign_type = NET_ADDR_RANDOM; 1463 eth_random_addr(dev->perm_addr); 1464 } 1465 1466 #define GRE6_FEATURES (NETIF_F_SG | \ 1467 NETIF_F_FRAGLIST | \ 1468 NETIF_F_HIGHDMA | \ 1469 NETIF_F_HW_CSUM) 1470 1471 static void ip6gre_tnl_init_features(struct net_device *dev) 1472 { 1473 struct ip6_tnl *nt = netdev_priv(dev); 1474 1475 dev->features |= GRE6_FEATURES; 1476 dev->hw_features |= GRE6_FEATURES; 1477 1478 /* TCP offload with GRE SEQ is not supported, nor can we support 2 1479 * levels of outer headers requiring an update. 1480 */ 1481 if (test_bit(IP_TUNNEL_SEQ_BIT, nt->parms.o_flags)) 1482 return; 1483 if (test_bit(IP_TUNNEL_CSUM_BIT, nt->parms.o_flags) && 1484 nt->encap.type != TUNNEL_ENCAP_NONE) 1485 return; 1486 1487 dev->features |= NETIF_F_GSO_SOFTWARE; 1488 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1489 1490 dev->lltx = true; 1491 } 1492 1493 static int ip6gre_tunnel_init_common(struct net_device *dev) 1494 { 1495 struct ip6_tnl *tunnel; 1496 int ret; 1497 int t_hlen; 1498 1499 tunnel = netdev_priv(dev); 1500 1501 tunnel->dev = dev; 1502 strcpy(tunnel->parms.name, dev->name); 1503 1504 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1505 if (ret) 1506 return ret; 1507 1508 ret = gro_cells_init(&tunnel->gro_cells, dev); 1509 if (ret) 1510 goto cleanup_dst_cache_init; 1511 1512 t_hlen = ip6gre_calc_hlen(tunnel); 1513 dev->mtu = ETH_DATA_LEN - t_hlen; 1514 if (dev->type == ARPHRD_ETHER) 1515 dev->mtu -= ETH_HLEN; 1516 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1517 dev->mtu -= 8; 1518 1519 if (tunnel->parms.collect_md) { 1520 netif_keep_dst(dev); 1521 } 1522 ip6gre_tnl_init_features(dev); 1523 1524 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1525 netdev_lockdep_set_classes(dev); 1526 return 0; 1527 1528 cleanup_dst_cache_init: 1529 dst_cache_destroy(&tunnel->dst_cache); 1530 return ret; 1531 } 1532 1533 static int ip6gre_tunnel_init(struct net_device *dev) 1534 { 1535 struct ip6_tnl *tunnel; 1536 int ret; 1537 1538 ret = ip6gre_tunnel_init_common(dev); 1539 if (ret) 1540 return ret; 1541 1542 tunnel = netdev_priv(dev); 1543 1544 if (tunnel->parms.collect_md) 1545 return 0; 1546 1547 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1548 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1549 1550 if (ipv6_addr_any(&tunnel->parms.raddr)) 1551 dev->header_ops = &ip6gre_header_ops; 1552 1553 return 0; 1554 } 1555 1556 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1557 { 1558 struct ip6_tnl *tunnel = netdev_priv(dev); 1559 1560 tunnel->dev = dev; 1561 tunnel->net = dev_net(dev); 1562 strcpy(tunnel->parms.name, dev->name); 1563 1564 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1565 } 1566 1567 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1568 .handler = gre_rcv, 1569 .err_handler = ip6gre_err, 1570 .flags = INET6_PROTO_FINAL, 1571 }; 1572 1573 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1574 { 1575 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1576 struct net_device *dev, *aux; 1577 int prio; 1578 1579 for_each_netdev_safe(net, dev, aux) 1580 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1581 dev->rtnl_link_ops == &ip6gre_tap_ops || 1582 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1583 unregister_netdevice_queue(dev, head); 1584 1585 for (prio = 0; prio < 4; prio++) { 1586 int h; 1587 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1588 struct ip6_tnl *t; 1589 1590 t = rtnl_dereference(ign->tunnels[prio][h]); 1591 1592 while (t) { 1593 /* If dev is in the same netns, it has already 1594 * been added to the list by the previous loop. 1595 */ 1596 if (!net_eq(dev_net(t->dev), net)) 1597 unregister_netdevice_queue(t->dev, 1598 head); 1599 t = rtnl_dereference(t->next); 1600 } 1601 } 1602 } 1603 } 1604 1605 static int __net_init ip6gre_init_net(struct net *net) 1606 { 1607 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1608 struct net_device *ndev; 1609 int err; 1610 1611 if (!net_has_fallback_tunnels(net)) 1612 return 0; 1613 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1614 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1615 if (!ndev) { 1616 err = -ENOMEM; 1617 goto err_alloc_dev; 1618 } 1619 ign->fb_tunnel_dev = ndev; 1620 dev_net_set(ign->fb_tunnel_dev, net); 1621 /* FB netdevice is special: we have one, and only one per netns. 1622 * Allowing to move it to another netns is clearly unsafe. 1623 */ 1624 ign->fb_tunnel_dev->netns_immutable = true; 1625 1626 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1627 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1628 1629 err = register_netdev(ign->fb_tunnel_dev); 1630 if (err) 1631 goto err_reg_dev; 1632 1633 rcu_assign_pointer(ign->tunnels_wc[0], 1634 netdev_priv(ign->fb_tunnel_dev)); 1635 return 0; 1636 1637 err_reg_dev: 1638 free_netdev(ndev); 1639 err_alloc_dev: 1640 return err; 1641 } 1642 1643 static void __net_exit ip6gre_exit_batch_rtnl(struct list_head *net_list, 1644 struct list_head *dev_to_kill) 1645 { 1646 struct net *net; 1647 1648 ASSERT_RTNL(); 1649 list_for_each_entry(net, net_list, exit_list) 1650 ip6gre_destroy_tunnels(net, dev_to_kill); 1651 } 1652 1653 static struct pernet_operations ip6gre_net_ops = { 1654 .init = ip6gre_init_net, 1655 .exit_batch_rtnl = ip6gre_exit_batch_rtnl, 1656 .id = &ip6gre_net_id, 1657 .size = sizeof(struct ip6gre_net), 1658 }; 1659 1660 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1661 struct netlink_ext_ack *extack) 1662 { 1663 __be16 flags; 1664 1665 if (!data) 1666 return 0; 1667 1668 flags = 0; 1669 if (data[IFLA_GRE_IFLAGS]) 1670 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1671 if (data[IFLA_GRE_OFLAGS]) 1672 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1673 if (flags & (GRE_VERSION|GRE_ROUTING)) 1674 return -EINVAL; 1675 1676 return 0; 1677 } 1678 1679 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1680 struct netlink_ext_ack *extack) 1681 { 1682 struct in6_addr daddr; 1683 1684 if (tb[IFLA_ADDRESS]) { 1685 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1686 return -EINVAL; 1687 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1688 return -EADDRNOTAVAIL; 1689 } 1690 1691 if (!data) 1692 goto out; 1693 1694 if (data[IFLA_GRE_REMOTE]) { 1695 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1696 if (ipv6_addr_any(&daddr)) 1697 return -EINVAL; 1698 } 1699 1700 out: 1701 return ip6gre_tunnel_validate(tb, data, extack); 1702 } 1703 1704 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1705 struct netlink_ext_ack *extack) 1706 { 1707 __be16 flags = 0; 1708 int ret, ver = 0; 1709 1710 if (!data) 1711 return 0; 1712 1713 ret = ip6gre_tap_validate(tb, data, extack); 1714 if (ret) 1715 return ret; 1716 1717 /* ERSPAN should only have GRE sequence and key flag */ 1718 if (data[IFLA_GRE_OFLAGS]) 1719 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1720 if (data[IFLA_GRE_IFLAGS]) 1721 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1722 if (!data[IFLA_GRE_COLLECT_METADATA] && 1723 flags != (GRE_SEQ | GRE_KEY)) 1724 return -EINVAL; 1725 1726 /* ERSPAN Session ID only has 10-bit. Since we reuse 1727 * 32-bit key field as ID, check it's range. 1728 */ 1729 if (data[IFLA_GRE_IKEY] && 1730 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1731 return -EINVAL; 1732 1733 if (data[IFLA_GRE_OKEY] && 1734 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1735 return -EINVAL; 1736 1737 if (data[IFLA_GRE_ERSPAN_VER]) { 1738 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1739 if (ver != 1 && ver != 2) 1740 return -EINVAL; 1741 } 1742 1743 if (ver == 1) { 1744 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1745 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1746 1747 if (index & ~INDEX_MASK) 1748 return -EINVAL; 1749 } 1750 } else if (ver == 2) { 1751 if (data[IFLA_GRE_ERSPAN_DIR]) { 1752 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1753 1754 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1755 return -EINVAL; 1756 } 1757 1758 if (data[IFLA_GRE_ERSPAN_HWID]) { 1759 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1760 1761 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1762 return -EINVAL; 1763 } 1764 } 1765 1766 return 0; 1767 } 1768 1769 static void ip6erspan_set_version(struct nlattr *data[], 1770 struct __ip6_tnl_parm *parms) 1771 { 1772 if (!data) 1773 return; 1774 1775 parms->erspan_ver = 1; 1776 if (data[IFLA_GRE_ERSPAN_VER]) 1777 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1778 1779 if (parms->erspan_ver == 1) { 1780 if (data[IFLA_GRE_ERSPAN_INDEX]) 1781 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1782 } else if (parms->erspan_ver == 2) { 1783 if (data[IFLA_GRE_ERSPAN_DIR]) 1784 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1785 if (data[IFLA_GRE_ERSPAN_HWID]) 1786 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1787 } 1788 } 1789 1790 static void ip6gre_netlink_parms(struct nlattr *data[], 1791 struct __ip6_tnl_parm *parms) 1792 { 1793 memset(parms, 0, sizeof(*parms)); 1794 1795 if (!data) 1796 return; 1797 1798 if (data[IFLA_GRE_LINK]) 1799 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1800 1801 if (data[IFLA_GRE_IFLAGS]) 1802 gre_flags_to_tnl_flags(parms->i_flags, 1803 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1804 1805 if (data[IFLA_GRE_OFLAGS]) 1806 gre_flags_to_tnl_flags(parms->o_flags, 1807 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1808 1809 if (data[IFLA_GRE_IKEY]) 1810 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1811 1812 if (data[IFLA_GRE_OKEY]) 1813 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1814 1815 if (data[IFLA_GRE_LOCAL]) 1816 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1817 1818 if (data[IFLA_GRE_REMOTE]) 1819 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1820 1821 if (data[IFLA_GRE_TTL]) 1822 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1823 1824 if (data[IFLA_GRE_ENCAP_LIMIT]) 1825 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1826 1827 if (data[IFLA_GRE_FLOWINFO]) 1828 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1829 1830 if (data[IFLA_GRE_FLAGS]) 1831 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1832 1833 if (data[IFLA_GRE_FWMARK]) 1834 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1835 1836 if (data[IFLA_GRE_COLLECT_METADATA]) 1837 parms->collect_md = true; 1838 } 1839 1840 static int ip6gre_tap_init(struct net_device *dev) 1841 { 1842 int ret; 1843 1844 ret = ip6gre_tunnel_init_common(dev); 1845 if (ret) 1846 return ret; 1847 1848 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1849 1850 return 0; 1851 } 1852 1853 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1854 .ndo_init = ip6gre_tap_init, 1855 .ndo_uninit = ip6gre_tunnel_uninit, 1856 .ndo_start_xmit = ip6gre_tunnel_xmit, 1857 .ndo_set_mac_address = eth_mac_addr, 1858 .ndo_validate_addr = eth_validate_addr, 1859 .ndo_change_mtu = ip6_tnl_change_mtu, 1860 .ndo_get_iflink = ip6_tnl_get_iflink, 1861 }; 1862 1863 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1864 { 1865 int t_hlen; 1866 1867 tunnel->tun_hlen = 8; 1868 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1869 erspan_hdr_len(tunnel->parms.erspan_ver); 1870 1871 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1872 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1873 return t_hlen; 1874 } 1875 1876 static int ip6erspan_tap_init(struct net_device *dev) 1877 { 1878 struct ip6_tnl *tunnel; 1879 int t_hlen; 1880 int ret; 1881 1882 tunnel = netdev_priv(dev); 1883 1884 tunnel->dev = dev; 1885 strcpy(tunnel->parms.name, dev->name); 1886 1887 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1888 if (ret) 1889 return ret; 1890 1891 ret = gro_cells_init(&tunnel->gro_cells, dev); 1892 if (ret) 1893 goto cleanup_dst_cache_init; 1894 1895 t_hlen = ip6erspan_calc_hlen(tunnel); 1896 dev->mtu = ETH_DATA_LEN - t_hlen; 1897 if (dev->type == ARPHRD_ETHER) 1898 dev->mtu -= ETH_HLEN; 1899 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1900 dev->mtu -= 8; 1901 1902 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1903 ip6erspan_tnl_link_config(tunnel, 1); 1904 1905 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1906 netdev_lockdep_set_classes(dev); 1907 return 0; 1908 1909 cleanup_dst_cache_init: 1910 dst_cache_destroy(&tunnel->dst_cache); 1911 return ret; 1912 } 1913 1914 static const struct net_device_ops ip6erspan_netdev_ops = { 1915 .ndo_init = ip6erspan_tap_init, 1916 .ndo_uninit = ip6erspan_tunnel_uninit, 1917 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1918 .ndo_set_mac_address = eth_mac_addr, 1919 .ndo_validate_addr = eth_validate_addr, 1920 .ndo_change_mtu = ip6_tnl_change_mtu, 1921 .ndo_get_iflink = ip6_tnl_get_iflink, 1922 }; 1923 1924 static void ip6gre_tap_setup(struct net_device *dev) 1925 { 1926 1927 ether_setup(dev); 1928 1929 dev->max_mtu = 0; 1930 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1931 dev->needs_free_netdev = true; 1932 dev->priv_destructor = ip6gre_dev_free; 1933 1934 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 1935 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1936 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1937 netif_keep_dst(dev); 1938 } 1939 1940 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1941 struct ip_tunnel_encap *ipencap) 1942 { 1943 bool ret = false; 1944 1945 memset(ipencap, 0, sizeof(*ipencap)); 1946 1947 if (!data) 1948 return ret; 1949 1950 if (data[IFLA_GRE_ENCAP_TYPE]) { 1951 ret = true; 1952 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1953 } 1954 1955 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1956 ret = true; 1957 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1958 } 1959 1960 if (data[IFLA_GRE_ENCAP_SPORT]) { 1961 ret = true; 1962 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1963 } 1964 1965 if (data[IFLA_GRE_ENCAP_DPORT]) { 1966 ret = true; 1967 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1968 } 1969 1970 return ret; 1971 } 1972 1973 static int ip6gre_newlink_common(struct net *link_net, struct net_device *dev, 1974 struct nlattr *tb[], struct nlattr *data[], 1975 struct netlink_ext_ack *extack) 1976 { 1977 struct ip6_tnl *nt; 1978 struct ip_tunnel_encap ipencap; 1979 int err; 1980 1981 nt = netdev_priv(dev); 1982 1983 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1984 int err = ip6_tnl_encap_setup(nt, &ipencap); 1985 1986 if (err < 0) 1987 return err; 1988 } 1989 1990 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1991 eth_hw_addr_random(dev); 1992 1993 nt->dev = dev; 1994 nt->net = link_net; 1995 1996 err = register_netdevice(dev); 1997 if (err) 1998 goto out; 1999 2000 if (tb[IFLA_MTU]) 2001 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2002 2003 out: 2004 return err; 2005 } 2006 2007 static int ip6gre_newlink(struct net_device *dev, 2008 struct rtnl_newlink_params *params, 2009 struct netlink_ext_ack *extack) 2010 { 2011 struct net *net = params->link_net ? : dev_net(dev); 2012 struct ip6_tnl *nt = netdev_priv(dev); 2013 struct nlattr **data = params->data; 2014 struct nlattr **tb = params->tb; 2015 struct ip6gre_net *ign; 2016 int err; 2017 2018 ip6gre_netlink_parms(data, &nt->parms); 2019 ign = net_generic(net, ip6gre_net_id); 2020 2021 if (nt->parms.collect_md) { 2022 if (rtnl_dereference(ign->collect_md_tun)) 2023 return -EEXIST; 2024 } else { 2025 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2026 return -EEXIST; 2027 } 2028 2029 err = ip6gre_newlink_common(net, dev, tb, data, extack); 2030 if (!err) { 2031 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2032 ip6gre_tunnel_link_md(ign, nt); 2033 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2034 } 2035 return err; 2036 } 2037 2038 static struct ip6_tnl * 2039 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2040 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2041 struct netlink_ext_ack *extack) 2042 { 2043 struct ip6_tnl *t, *nt = netdev_priv(dev); 2044 struct net *net = nt->net; 2045 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2046 struct ip_tunnel_encap ipencap; 2047 2048 if (dev == ign->fb_tunnel_dev) 2049 return ERR_PTR(-EINVAL); 2050 2051 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2052 int err = ip6_tnl_encap_setup(nt, &ipencap); 2053 2054 if (err < 0) 2055 return ERR_PTR(err); 2056 } 2057 2058 ip6gre_netlink_parms(data, p_p); 2059 2060 t = ip6gre_tunnel_locate(net, p_p, 0); 2061 2062 if (t) { 2063 if (t->dev != dev) 2064 return ERR_PTR(-EEXIST); 2065 } else { 2066 t = nt; 2067 } 2068 2069 return t; 2070 } 2071 2072 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2073 struct nlattr *data[], 2074 struct netlink_ext_ack *extack) 2075 { 2076 struct ip6_tnl *t = netdev_priv(dev); 2077 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2078 struct __ip6_tnl_parm p; 2079 2080 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2081 if (IS_ERR(t)) 2082 return PTR_ERR(t); 2083 2084 ip6gre_tunnel_unlink_md(ign, t); 2085 ip6gre_tunnel_unlink(ign, t); 2086 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2087 ip6gre_tunnel_link_md(ign, t); 2088 ip6gre_tunnel_link(ign, t); 2089 return 0; 2090 } 2091 2092 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2093 { 2094 struct net *net = dev_net(dev); 2095 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2096 2097 if (dev != ign->fb_tunnel_dev) 2098 unregister_netdevice_queue(dev, head); 2099 } 2100 2101 static size_t ip6gre_get_size(const struct net_device *dev) 2102 { 2103 return 2104 /* IFLA_GRE_LINK */ 2105 nla_total_size(4) + 2106 /* IFLA_GRE_IFLAGS */ 2107 nla_total_size(2) + 2108 /* IFLA_GRE_OFLAGS */ 2109 nla_total_size(2) + 2110 /* IFLA_GRE_IKEY */ 2111 nla_total_size(4) + 2112 /* IFLA_GRE_OKEY */ 2113 nla_total_size(4) + 2114 /* IFLA_GRE_LOCAL */ 2115 nla_total_size(sizeof(struct in6_addr)) + 2116 /* IFLA_GRE_REMOTE */ 2117 nla_total_size(sizeof(struct in6_addr)) + 2118 /* IFLA_GRE_TTL */ 2119 nla_total_size(1) + 2120 /* IFLA_GRE_ENCAP_LIMIT */ 2121 nla_total_size(1) + 2122 /* IFLA_GRE_FLOWINFO */ 2123 nla_total_size(4) + 2124 /* IFLA_GRE_FLAGS */ 2125 nla_total_size(4) + 2126 /* IFLA_GRE_ENCAP_TYPE */ 2127 nla_total_size(2) + 2128 /* IFLA_GRE_ENCAP_FLAGS */ 2129 nla_total_size(2) + 2130 /* IFLA_GRE_ENCAP_SPORT */ 2131 nla_total_size(2) + 2132 /* IFLA_GRE_ENCAP_DPORT */ 2133 nla_total_size(2) + 2134 /* IFLA_GRE_COLLECT_METADATA */ 2135 nla_total_size(0) + 2136 /* IFLA_GRE_FWMARK */ 2137 nla_total_size(4) + 2138 /* IFLA_GRE_ERSPAN_INDEX */ 2139 nla_total_size(4) + 2140 0; 2141 } 2142 2143 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2144 { 2145 struct ip6_tnl *t = netdev_priv(dev); 2146 struct __ip6_tnl_parm *p = &t->parms; 2147 IP_TUNNEL_DECLARE_FLAGS(o_flags); 2148 2149 ip_tunnel_flags_copy(o_flags, p->o_flags); 2150 2151 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2152 if (!p->collect_md) 2153 __set_bit(IP_TUNNEL_KEY_BIT, o_flags); 2154 2155 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2156 goto nla_put_failure; 2157 2158 if (p->erspan_ver == 1) { 2159 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2160 goto nla_put_failure; 2161 } else { 2162 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2163 goto nla_put_failure; 2164 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2165 goto nla_put_failure; 2166 } 2167 } 2168 2169 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2170 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2171 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2172 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2173 gre_tnl_flags_to_gre_flags(o_flags)) || 2174 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2175 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2176 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2177 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2178 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2179 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2180 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2181 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2182 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2183 goto nla_put_failure; 2184 2185 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2186 t->encap.type) || 2187 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2188 t->encap.sport) || 2189 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2190 t->encap.dport) || 2191 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2192 t->encap.flags)) 2193 goto nla_put_failure; 2194 2195 if (p->collect_md) { 2196 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2197 goto nla_put_failure; 2198 } 2199 2200 return 0; 2201 2202 nla_put_failure: 2203 return -EMSGSIZE; 2204 } 2205 2206 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2207 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2208 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2209 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2210 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2211 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2212 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2213 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2214 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2215 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2216 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2217 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2218 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2219 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2220 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2221 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2222 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2223 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2224 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2225 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2226 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2227 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2228 }; 2229 2230 static void ip6erspan_tap_setup(struct net_device *dev) 2231 { 2232 ether_setup(dev); 2233 2234 dev->max_mtu = 0; 2235 dev->netdev_ops = &ip6erspan_netdev_ops; 2236 dev->needs_free_netdev = true; 2237 dev->priv_destructor = ip6gre_dev_free; 2238 2239 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 2240 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2241 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2242 netif_keep_dst(dev); 2243 } 2244 2245 static int ip6erspan_newlink(struct net_device *dev, 2246 struct rtnl_newlink_params *params, 2247 struct netlink_ext_ack *extack) 2248 { 2249 struct net *net = params->link_net ? : dev_net(dev); 2250 struct ip6_tnl *nt = netdev_priv(dev); 2251 struct nlattr **data = params->data; 2252 struct nlattr **tb = params->tb; 2253 struct ip6gre_net *ign; 2254 int err; 2255 2256 ip6gre_netlink_parms(data, &nt->parms); 2257 ip6erspan_set_version(data, &nt->parms); 2258 ign = net_generic(net, ip6gre_net_id); 2259 2260 if (nt->parms.collect_md) { 2261 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2262 return -EEXIST; 2263 } else { 2264 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2265 return -EEXIST; 2266 } 2267 2268 err = ip6gre_newlink_common(net, dev, tb, data, extack); 2269 if (!err) { 2270 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2271 ip6erspan_tunnel_link_md(ign, nt); 2272 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2273 } 2274 return err; 2275 } 2276 2277 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2278 { 2279 ip6gre_tnl_link_config_common(t); 2280 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2281 } 2282 2283 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2284 const struct __ip6_tnl_parm *p, int set_mtu) 2285 { 2286 ip6gre_tnl_copy_tnl_parm(t, p); 2287 ip6erspan_tnl_link_config(t, set_mtu); 2288 return 0; 2289 } 2290 2291 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2292 struct nlattr *data[], 2293 struct netlink_ext_ack *extack) 2294 { 2295 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2296 struct __ip6_tnl_parm p; 2297 struct ip6_tnl *t; 2298 2299 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2300 if (IS_ERR(t)) 2301 return PTR_ERR(t); 2302 2303 ip6erspan_set_version(data, &p); 2304 ip6gre_tunnel_unlink_md(ign, t); 2305 ip6gre_tunnel_unlink(ign, t); 2306 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2307 ip6erspan_tunnel_link_md(ign, t); 2308 ip6gre_tunnel_link(ign, t); 2309 return 0; 2310 } 2311 2312 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2313 .kind = "ip6gre", 2314 .maxtype = IFLA_GRE_MAX, 2315 .policy = ip6gre_policy, 2316 .priv_size = sizeof(struct ip6_tnl), 2317 .setup = ip6gre_tunnel_setup, 2318 .validate = ip6gre_tunnel_validate, 2319 .newlink = ip6gre_newlink, 2320 .changelink = ip6gre_changelink, 2321 .dellink = ip6gre_dellink, 2322 .get_size = ip6gre_get_size, 2323 .fill_info = ip6gre_fill_info, 2324 .get_link_net = ip6_tnl_get_link_net, 2325 }; 2326 2327 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2328 .kind = "ip6gretap", 2329 .maxtype = IFLA_GRE_MAX, 2330 .policy = ip6gre_policy, 2331 .priv_size = sizeof(struct ip6_tnl), 2332 .setup = ip6gre_tap_setup, 2333 .validate = ip6gre_tap_validate, 2334 .newlink = ip6gre_newlink, 2335 .changelink = ip6gre_changelink, 2336 .get_size = ip6gre_get_size, 2337 .fill_info = ip6gre_fill_info, 2338 .get_link_net = ip6_tnl_get_link_net, 2339 }; 2340 2341 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2342 .kind = "ip6erspan", 2343 .maxtype = IFLA_GRE_MAX, 2344 .policy = ip6gre_policy, 2345 .priv_size = sizeof(struct ip6_tnl), 2346 .setup = ip6erspan_tap_setup, 2347 .validate = ip6erspan_tap_validate, 2348 .newlink = ip6erspan_newlink, 2349 .changelink = ip6erspan_changelink, 2350 .get_size = ip6gre_get_size, 2351 .fill_info = ip6gre_fill_info, 2352 .get_link_net = ip6_tnl_get_link_net, 2353 }; 2354 2355 /* 2356 * And now the modules code and kernel interface. 2357 */ 2358 2359 static int __init ip6gre_init(void) 2360 { 2361 int err; 2362 2363 pr_info("GRE over IPv6 tunneling driver\n"); 2364 2365 err = register_pernet_device(&ip6gre_net_ops); 2366 if (err < 0) 2367 return err; 2368 2369 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2370 if (err < 0) { 2371 pr_info("%s: can't add protocol\n", __func__); 2372 goto add_proto_failed; 2373 } 2374 2375 err = rtnl_link_register(&ip6gre_link_ops); 2376 if (err < 0) 2377 goto rtnl_link_failed; 2378 2379 err = rtnl_link_register(&ip6gre_tap_ops); 2380 if (err < 0) 2381 goto tap_ops_failed; 2382 2383 err = rtnl_link_register(&ip6erspan_tap_ops); 2384 if (err < 0) 2385 goto erspan_link_failed; 2386 2387 out: 2388 return err; 2389 2390 erspan_link_failed: 2391 rtnl_link_unregister(&ip6gre_tap_ops); 2392 tap_ops_failed: 2393 rtnl_link_unregister(&ip6gre_link_ops); 2394 rtnl_link_failed: 2395 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2396 add_proto_failed: 2397 unregister_pernet_device(&ip6gre_net_ops); 2398 goto out; 2399 } 2400 2401 static void __exit ip6gre_fini(void) 2402 { 2403 rtnl_link_unregister(&ip6gre_tap_ops); 2404 rtnl_link_unregister(&ip6gre_link_ops); 2405 rtnl_link_unregister(&ip6erspan_tap_ops); 2406 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2407 unregister_pernet_device(&ip6gre_net_ops); 2408 } 2409 2410 module_init(ip6gre_init); 2411 module_exit(ip6gre_fini); 2412 MODULE_LICENSE("GPL"); 2413 MODULE_AUTHOR("D. Kozlov <xeb@mail.ru>"); 2414 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2415 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2416 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2417 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2418 MODULE_ALIAS_NETDEV("ip6gre0"); 2419