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 static bool ip6gre_tunnel_match(struct ip6_tnl *t, int dev_type, int link, 115 int *cand_score, struct ip6_tnl **ret) 116 { 117 int score = 0; 118 119 if (t->dev->type != ARPHRD_IP6GRE && 120 t->dev->type != dev_type) 121 return false; 122 123 if (t->parms.link != link) 124 score |= 1; 125 if (t->dev->type != dev_type) 126 score |= 2; 127 if (score == 0) { 128 *ret = t; 129 return true; 130 } 131 132 if (score < *cand_score) { 133 *ret = t; 134 *cand_score = score; 135 } 136 return false; 137 } 138 139 /* Given src, dst and key, find appropriate for input tunnel. */ 140 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 141 const struct in6_addr *remote, const struct in6_addr *local, 142 __be32 key, __be16 gre_proto) 143 { 144 struct net *net = dev_net(dev); 145 int link = dev->ifindex; 146 unsigned int h0 = HASH_ADDR(remote); 147 unsigned int h1 = HASH_KEY(key); 148 struct ip6_tnl *t, *cand = NULL; 149 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 150 int dev_type = (gre_proto == htons(ETH_P_TEB) || 151 gre_proto == htons(ETH_P_ERSPAN) || 152 gre_proto == htons(ETH_P_ERSPAN2)) ? 153 ARPHRD_ETHER : ARPHRD_IP6GRE; 154 struct net_device *ndev; 155 int cand_score = 4; 156 157 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 158 if (!ipv6_addr_equal(local, &t->parms.laddr) || 159 !ipv6_addr_equal(remote, &t->parms.raddr) || 160 key != t->parms.i_key || 161 !(t->dev->flags & IFF_UP)) 162 continue; 163 164 if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand)) 165 return cand; 166 } 167 168 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 169 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 170 key != t->parms.i_key || 171 !(t->dev->flags & IFF_UP)) 172 continue; 173 174 if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand)) 175 return cand; 176 } 177 178 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 179 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 180 (!ipv6_addr_equal(local, &t->parms.raddr) || 181 !ipv6_addr_is_multicast(local))) || 182 key != t->parms.i_key || 183 !(t->dev->flags & IFF_UP)) 184 continue; 185 186 if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand)) 187 return cand; 188 } 189 190 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 191 if (t->parms.i_key != key || 192 !(t->dev->flags & IFF_UP)) 193 continue; 194 195 if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand)) 196 return cand; 197 } 198 199 if (cand) 200 return cand; 201 202 if (gre_proto == htons(ETH_P_ERSPAN) || 203 gre_proto == htons(ETH_P_ERSPAN2)) 204 t = rcu_dereference(ign->collect_md_tun_erspan); 205 else 206 t = rcu_dereference(ign->collect_md_tun); 207 208 if (t && t->dev->flags & IFF_UP) 209 return t; 210 211 ndev = READ_ONCE(ign->fb_tunnel_dev); 212 if (ndev && ndev->flags & IFF_UP) 213 return netdev_priv(ndev); 214 215 return NULL; 216 } 217 218 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 219 const struct __ip6_tnl_parm *p) 220 { 221 const struct in6_addr *remote = &p->raddr; 222 const struct in6_addr *local = &p->laddr; 223 unsigned int h = HASH_KEY(p->i_key); 224 int prio = 0; 225 226 if (!ipv6_addr_any(local)) 227 prio |= 1; 228 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 229 prio |= 2; 230 h ^= HASH_ADDR(remote); 231 } 232 233 return &ign->tunnels[prio][h]; 234 } 235 236 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 237 { 238 if (t->parms.collect_md) 239 rcu_assign_pointer(ign->collect_md_tun, t); 240 } 241 242 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 243 { 244 if (t->parms.collect_md) 245 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 246 } 247 248 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 249 { 250 if (t->parms.collect_md) 251 rcu_assign_pointer(ign->collect_md_tun, NULL); 252 } 253 254 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 255 struct ip6_tnl *t) 256 { 257 if (t->parms.collect_md) 258 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 259 } 260 261 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 262 const struct ip6_tnl *t) 263 { 264 return __ip6gre_bucket(ign, &t->parms); 265 } 266 267 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 268 { 269 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 270 271 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 272 rcu_assign_pointer(*tp, t); 273 } 274 275 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 276 { 277 struct ip6_tnl __rcu **tp; 278 struct ip6_tnl *iter; 279 280 for (tp = ip6gre_bucket(ign, t); 281 (iter = rtnl_dereference(*tp)) != NULL; 282 tp = &iter->next) { 283 if (t == iter) { 284 rcu_assign_pointer(*tp, t->next); 285 break; 286 } 287 } 288 } 289 290 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 291 const struct __ip6_tnl_parm *parms, 292 int type) 293 { 294 const struct in6_addr *remote = &parms->raddr; 295 const struct in6_addr *local = &parms->laddr; 296 __be32 key = parms->i_key; 297 int link = parms->link; 298 struct ip6_tnl *t; 299 struct ip6_tnl __rcu **tp; 300 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 301 302 for (tp = __ip6gre_bucket(ign, parms); 303 (t = rtnl_dereference(*tp)) != NULL; 304 tp = &t->next) 305 if (ipv6_addr_equal(local, &t->parms.laddr) && 306 ipv6_addr_equal(remote, &t->parms.raddr) && 307 key == t->parms.i_key && 308 link == t->parms.link && 309 type == t->dev->type) 310 break; 311 312 return t; 313 } 314 315 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 316 const struct __ip6_tnl_parm *parms, int create) 317 { 318 struct ip6_tnl *t, *nt; 319 struct net_device *dev; 320 char name[IFNAMSIZ]; 321 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 322 323 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 324 if (t && create) 325 return NULL; 326 if (t || !create) 327 return t; 328 329 if (parms->name[0]) { 330 if (!dev_valid_name(parms->name)) 331 return NULL; 332 strscpy(name, parms->name, IFNAMSIZ); 333 } else { 334 strcpy(name, "ip6gre%d"); 335 } 336 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 337 ip6gre_tunnel_setup); 338 if (!dev) 339 return NULL; 340 341 dev_net_set(dev, net); 342 343 nt = netdev_priv(dev); 344 nt->parms = *parms; 345 dev->rtnl_link_ops = &ip6gre_link_ops; 346 347 nt->dev = dev; 348 nt->net = dev_net(dev); 349 350 if (register_netdevice(dev) < 0) 351 goto failed_free; 352 353 ip6gre_tnl_link_config(nt, 1); 354 ip6gre_tunnel_link(ign, nt); 355 return nt; 356 357 failed_free: 358 free_netdev(dev); 359 return NULL; 360 } 361 362 static void ip6erspan_tunnel_uninit(struct net_device *dev) 363 { 364 struct ip6_tnl *t = netdev_priv(dev); 365 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 366 367 ip6erspan_tunnel_unlink_md(ign, t); 368 ip6gre_tunnel_unlink(ign, t); 369 dst_cache_reset(&t->dst_cache); 370 netdev_put(dev, &t->dev_tracker); 371 } 372 373 static void ip6gre_tunnel_uninit(struct net_device *dev) 374 { 375 struct ip6_tnl *t = netdev_priv(dev); 376 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 377 378 ip6gre_tunnel_unlink_md(ign, t); 379 ip6gre_tunnel_unlink(ign, t); 380 if (ign->fb_tunnel_dev == dev) 381 WRITE_ONCE(ign->fb_tunnel_dev, NULL); 382 dst_cache_reset(&t->dst_cache); 383 netdev_put(dev, &t->dev_tracker); 384 } 385 386 387 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 388 u8 type, u8 code, int offset, __be32 info) 389 { 390 struct net *net = dev_net(skb->dev); 391 const struct ipv6hdr *ipv6h; 392 struct tnl_ptk_info tpi; 393 struct ip6_tnl *t; 394 395 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 396 offset) < 0) 397 return -EINVAL; 398 399 ipv6h = (const struct ipv6hdr *)skb->data; 400 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 401 tpi.key, tpi.proto); 402 if (!t) 403 return -ENOENT; 404 405 switch (type) { 406 case ICMPV6_DEST_UNREACH: 407 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 408 t->parms.name); 409 if (code != ICMPV6_PORT_UNREACH) 410 break; 411 return 0; 412 case ICMPV6_TIME_EXCEED: 413 if (code == ICMPV6_EXC_HOPLIMIT) { 414 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 415 t->parms.name); 416 break; 417 } 418 return 0; 419 case ICMPV6_PARAMPROB: { 420 struct ipv6_tlv_tnl_enc_lim *tel; 421 __u32 teli; 422 423 teli = 0; 424 if (code == ICMPV6_HDR_FIELD) 425 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 426 427 if (teli && teli == be32_to_cpu(info) - 2) { 428 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 429 if (tel->encap_limit == 0) { 430 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 431 t->parms.name); 432 } 433 } else { 434 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 435 t->parms.name); 436 } 437 return 0; 438 } 439 case ICMPV6_PKT_TOOBIG: 440 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 441 return 0; 442 case NDISC_REDIRECT: 443 ip6_redirect(skb, net, skb->dev->ifindex, 0, 444 sock_net_uid(net, NULL)); 445 return 0; 446 } 447 448 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 449 t->err_count++; 450 else 451 t->err_count = 1; 452 t->err_time = jiffies; 453 454 return 0; 455 } 456 457 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 458 { 459 const struct ipv6hdr *ipv6h; 460 struct ip6_tnl *tunnel; 461 462 ipv6h = ipv6_hdr(skb); 463 tunnel = ip6gre_tunnel_lookup(skb->dev, 464 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 465 tpi->proto); 466 if (tunnel) { 467 if (tunnel->parms.collect_md) { 468 IP_TUNNEL_DECLARE_FLAGS(flags); 469 struct metadata_dst *tun_dst; 470 __be64 tun_id; 471 472 ip_tunnel_flags_copy(flags, tpi->flags); 473 tun_id = key32_to_tunnel_id(tpi->key); 474 475 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 476 if (!tun_dst) 477 return PACKET_REJECT; 478 479 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 480 } else { 481 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 482 } 483 484 return PACKET_RCVD; 485 } 486 487 return PACKET_REJECT; 488 } 489 490 static int ip6erspan_rcv(struct sk_buff *skb, 491 struct tnl_ptk_info *tpi, 492 int gre_hdr_len) 493 { 494 struct erspan_base_hdr *ershdr; 495 const struct ipv6hdr *ipv6h; 496 struct erspan_md2 *md2; 497 struct ip6_tnl *tunnel; 498 u8 ver; 499 500 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr)))) 501 return PACKET_REJECT; 502 503 ipv6h = ipv6_hdr(skb); 504 ershdr = (struct erspan_base_hdr *)skb->data; 505 ver = ershdr->ver; 506 507 tunnel = ip6gre_tunnel_lookup(skb->dev, 508 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 509 tpi->proto); 510 if (tunnel) { 511 int len = erspan_hdr_len(ver); 512 513 if (unlikely(!pskb_may_pull(skb, len))) 514 return PACKET_REJECT; 515 516 if (__iptunnel_pull_header(skb, len, 517 htons(ETH_P_TEB), 518 false, false) < 0) 519 return PACKET_REJECT; 520 521 if (tunnel->parms.collect_md) { 522 struct erspan_metadata *pkt_md, *md; 523 IP_TUNNEL_DECLARE_FLAGS(flags); 524 struct metadata_dst *tun_dst; 525 struct ip_tunnel_info *info; 526 unsigned char *gh; 527 __be64 tun_id; 528 529 __set_bit(IP_TUNNEL_KEY_BIT, tpi->flags); 530 ip_tunnel_flags_copy(flags, tpi->flags); 531 tun_id = key32_to_tunnel_id(tpi->key); 532 533 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 534 sizeof(*md)); 535 if (!tun_dst) 536 return PACKET_REJECT; 537 538 /* skb can be uncloned in __iptunnel_pull_header, so 539 * old pkt_md is no longer valid and we need to reset 540 * it 541 */ 542 gh = skb_network_header(skb) + 543 skb_network_header_len(skb); 544 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 545 sizeof(*ershdr)); 546 info = &tun_dst->u.tun_info; 547 md = ip_tunnel_info_opts(info); 548 md->version = ver; 549 md2 = &md->u.md2; 550 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 551 ERSPAN_V2_MDSIZE); 552 __set_bit(IP_TUNNEL_ERSPAN_OPT_BIT, 553 info->key.tun_flags); 554 info->options_len = sizeof(*md); 555 556 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 557 558 } else { 559 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 560 } 561 562 return PACKET_RCVD; 563 } 564 565 return PACKET_REJECT; 566 } 567 568 static int gre_rcv(struct sk_buff *skb) 569 { 570 struct tnl_ptk_info tpi; 571 bool csum_err = false; 572 int hdr_len; 573 574 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 575 if (hdr_len < 0) 576 goto drop; 577 578 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 579 goto drop; 580 581 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 582 tpi.proto == htons(ETH_P_ERSPAN2))) { 583 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) 584 return 0; 585 goto out; 586 } 587 588 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 589 return 0; 590 591 out: 592 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 593 drop: 594 kfree_skb(skb); 595 return 0; 596 } 597 598 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 599 { 600 return iptunnel_handle_offloads(skb, 601 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 602 } 603 604 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 605 struct net_device *dev, 606 struct flowi6 *fl6, __u8 *dsfield, 607 int *encap_limit) 608 { 609 const struct iphdr *iph = ip_hdr(skb); 610 struct ip6_tnl *t = netdev_priv(dev); 611 612 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 613 *encap_limit = t->parms.encap_limit; 614 615 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 616 617 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 618 *dsfield = ipv4_get_dsfield(iph); 619 else 620 *dsfield = ip6_tclass(t->parms.flowinfo); 621 622 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 623 fl6->flowi6_mark = skb->mark; 624 else 625 fl6->flowi6_mark = t->parms.fwmark; 626 627 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 628 } 629 630 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 631 struct net_device *dev, 632 struct flowi6 *fl6, __u8 *dsfield, 633 int *encap_limit) 634 { 635 struct ipv6hdr *ipv6h; 636 struct ip6_tnl *t = netdev_priv(dev); 637 __u16 offset; 638 639 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 640 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 641 ipv6h = ipv6_hdr(skb); 642 643 if (offset > 0) { 644 struct ipv6_tlv_tnl_enc_lim *tel; 645 646 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 647 if (tel->encap_limit == 0) { 648 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 649 ICMPV6_HDR_FIELD, offset + 2); 650 return -1; 651 } 652 *encap_limit = tel->encap_limit - 1; 653 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 654 *encap_limit = t->parms.encap_limit; 655 } 656 657 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 658 659 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 660 *dsfield = ipv6_get_dsfield(ipv6h); 661 else 662 *dsfield = ip6_tclass(t->parms.flowinfo); 663 664 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 665 fl6->flowlabel |= ip6_flowlabel(ipv6h); 666 667 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 668 fl6->flowi6_mark = skb->mark; 669 else 670 fl6->flowi6_mark = t->parms.fwmark; 671 672 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 673 674 return 0; 675 } 676 677 static int prepare_ip6gre_xmit_other(struct sk_buff *skb, 678 struct net_device *dev, 679 struct flowi6 *fl6, __u8 *dsfield, 680 int *encap_limit) 681 { 682 struct ip6_tnl *t = netdev_priv(dev); 683 684 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 685 *encap_limit = t->parms.encap_limit; 686 687 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 688 689 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 690 *dsfield = 0; 691 else 692 *dsfield = ip6_tclass(t->parms.flowinfo); 693 694 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 695 fl6->flowi6_mark = skb->mark; 696 else 697 fl6->flowi6_mark = t->parms.fwmark; 698 699 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 700 701 return 0; 702 } 703 704 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb) 705 { 706 struct ip_tunnel_info *tun_info; 707 708 tun_info = skb_tunnel_info(skb); 709 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))) 710 return ERR_PTR(-EINVAL); 711 712 return tun_info; 713 } 714 715 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 716 struct net_device *dev, __u8 dsfield, 717 struct flowi6 *fl6, int encap_limit, 718 __u32 *pmtu, __be16 proto) 719 { 720 struct ip6_tnl *tunnel = netdev_priv(dev); 721 IP_TUNNEL_DECLARE_FLAGS(flags); 722 __be16 protocol; 723 724 if (dev->type == ARPHRD_ETHER) 725 IPCB(skb)->flags = 0; 726 727 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 728 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 729 else 730 fl6->daddr = tunnel->parms.raddr; 731 732 /* Push GRE header. */ 733 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 734 735 if (tunnel->parms.collect_md) { 736 struct ip_tunnel_info *tun_info; 737 const struct ip_tunnel_key *key; 738 int tun_hlen; 739 740 tun_info = skb_tunnel_info_txcheck(skb); 741 if (IS_ERR(tun_info) || 742 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 743 return -EINVAL; 744 745 key = &tun_info->key; 746 memset(fl6, 0, sizeof(*fl6)); 747 fl6->flowi6_proto = IPPROTO_GRE; 748 fl6->daddr = key->u.ipv6.dst; 749 fl6->flowlabel = key->label; 750 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 751 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id); 752 753 dsfield = key->tos; 754 ip_tunnel_flags_zero(flags); 755 __set_bit(IP_TUNNEL_CSUM_BIT, flags); 756 __set_bit(IP_TUNNEL_KEY_BIT, flags); 757 __set_bit(IP_TUNNEL_SEQ_BIT, flags); 758 ip_tunnel_flags_and(flags, flags, key->tun_flags); 759 tun_hlen = gre_calc_hlen(flags); 760 761 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen)) 762 return -ENOMEM; 763 764 gre_build_header(skb, tun_hlen, 765 flags, protocol, 766 tunnel_id_to_key32(tun_info->key.tun_id), 767 test_bit(IP_TUNNEL_SEQ_BIT, flags) ? 768 htonl(atomic_fetch_inc(&tunnel->o_seqno)) : 769 0); 770 771 } else { 772 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 773 return -ENOMEM; 774 775 ip_tunnel_flags_copy(flags, tunnel->parms.o_flags); 776 777 gre_build_header(skb, tunnel->tun_hlen, flags, 778 protocol, tunnel->parms.o_key, 779 test_bit(IP_TUNNEL_SEQ_BIT, flags) ? 780 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, test_bit(IP_TUNNEL_CSUM_BIT, 804 t->parms.o_flags)); 805 if (err) 806 return -1; 807 808 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 809 skb->protocol); 810 if (err != 0) { 811 /* XXX: send ICMP error even if DF is not set. */ 812 if (err == -EMSGSIZE) 813 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 814 htonl(mtu)); 815 return -1; 816 } 817 818 return 0; 819 } 820 821 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 822 { 823 struct ip6_tnl *t = netdev_priv(dev); 824 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 825 int encap_limit = -1; 826 struct flowi6 fl6; 827 __u8 dsfield = 0; 828 __u32 mtu; 829 int err; 830 831 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 832 return -1; 833 834 if (!t->parms.collect_md && 835 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 836 return -1; 837 838 if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT, 839 t->parms.o_flags))) 840 return -1; 841 842 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 843 &mtu, skb->protocol); 844 if (err != 0) { 845 if (err == -EMSGSIZE) 846 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 847 return -1; 848 } 849 850 return 0; 851 } 852 853 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 854 { 855 struct ip6_tnl *t = netdev_priv(dev); 856 int encap_limit = -1; 857 struct flowi6 fl6; 858 __u8 dsfield = 0; 859 __u32 mtu; 860 int err; 861 862 if (!t->parms.collect_md && 863 prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit)) 864 return -1; 865 866 err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT, 867 t->parms.o_flags)); 868 if (err) 869 return err; 870 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol); 871 872 return err; 873 } 874 875 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 876 struct net_device *dev) 877 { 878 struct ip6_tnl *t = netdev_priv(dev); 879 __be16 payload_protocol; 880 int ret; 881 882 if (!pskb_inet_may_pull(skb)) 883 goto tx_err; 884 885 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 886 goto tx_err; 887 888 payload_protocol = skb_protocol(skb, true); 889 switch (payload_protocol) { 890 case htons(ETH_P_IP): 891 ret = ip6gre_xmit_ipv4(skb, dev); 892 break; 893 case htons(ETH_P_IPV6): 894 ret = ip6gre_xmit_ipv6(skb, dev); 895 break; 896 default: 897 ret = ip6gre_xmit_other(skb, dev); 898 break; 899 } 900 901 if (ret < 0) 902 goto tx_err; 903 904 return NETDEV_TX_OK; 905 906 tx_err: 907 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 908 DEV_STATS_INC(dev, tx_errors); 909 DEV_STATS_INC(dev, tx_dropped); 910 kfree_skb(skb); 911 return NETDEV_TX_OK; 912 } 913 914 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 915 struct net_device *dev) 916 { 917 struct ip_tunnel_info *tun_info = NULL; 918 struct ip6_tnl *t = netdev_priv(dev); 919 struct dst_entry *dst = skb_dst(skb); 920 IP_TUNNEL_DECLARE_FLAGS(flags) = { }; 921 bool truncate = false; 922 int encap_limit = -1; 923 __u8 dsfield = false; 924 struct flowi6 fl6; 925 int err = -EINVAL; 926 __be16 proto; 927 __u32 mtu; 928 int nhoff; 929 930 if (!pskb_inet_may_pull(skb)) 931 goto tx_err; 932 933 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 934 goto tx_err; 935 936 if (gre_handle_offloads(skb, false)) 937 goto tx_err; 938 939 if (skb->len > dev->mtu + dev->hard_header_len) { 940 if (pskb_trim(skb, dev->mtu + dev->hard_header_len)) 941 goto tx_err; 942 truncate = true; 943 } 944 945 nhoff = skb_network_offset(skb); 946 if (skb->protocol == htons(ETH_P_IP) && 947 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 948 truncate = true; 949 950 if (skb->protocol == htons(ETH_P_IPV6)) { 951 int thoff; 952 953 if (skb_transport_header_was_set(skb)) 954 thoff = skb_transport_offset(skb); 955 else 956 thoff = nhoff + sizeof(struct ipv6hdr); 957 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff) 958 truncate = true; 959 } 960 961 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 962 goto tx_err; 963 964 __clear_bit(IP_TUNNEL_KEY_BIT, t->parms.o_flags); 965 IPCB(skb)->flags = 0; 966 967 /* For collect_md mode, derive fl6 from the tunnel key, 968 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 969 */ 970 if (t->parms.collect_md) { 971 const struct ip_tunnel_key *key; 972 struct erspan_metadata *md; 973 __be32 tun_id; 974 975 tun_info = skb_tunnel_info_txcheck(skb); 976 if (IS_ERR(tun_info) || 977 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 978 goto tx_err; 979 980 key = &tun_info->key; 981 memset(&fl6, 0, sizeof(fl6)); 982 fl6.flowi6_proto = IPPROTO_GRE; 983 fl6.daddr = key->u.ipv6.dst; 984 fl6.flowlabel = key->label; 985 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 986 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 987 988 dsfield = key->tos; 989 if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT, 990 tun_info->key.tun_flags)) 991 goto tx_err; 992 if (tun_info->options_len < sizeof(*md)) 993 goto tx_err; 994 md = ip_tunnel_info_opts(tun_info); 995 996 tun_id = tunnel_id_to_key32(key->tun_id); 997 if (md->version == 1) { 998 erspan_build_header(skb, 999 ntohl(tun_id), 1000 ntohl(md->u.index), truncate, 1001 false); 1002 proto = htons(ETH_P_ERSPAN); 1003 } else if (md->version == 2) { 1004 erspan_build_header_v2(skb, 1005 ntohl(tun_id), 1006 md->u.md2.dir, 1007 get_hwid(&md->u.md2), 1008 truncate, false); 1009 proto = htons(ETH_P_ERSPAN2); 1010 } else { 1011 goto tx_err; 1012 } 1013 } else { 1014 switch (skb->protocol) { 1015 case htons(ETH_P_IP): 1016 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1017 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1018 &dsfield, &encap_limit); 1019 break; 1020 case htons(ETH_P_IPV6): 1021 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1022 goto tx_err; 1023 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1024 &dsfield, &encap_limit)) 1025 goto tx_err; 1026 break; 1027 default: 1028 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1029 break; 1030 } 1031 1032 if (t->parms.erspan_ver == 1) { 1033 erspan_build_header(skb, ntohl(t->parms.o_key), 1034 t->parms.index, 1035 truncate, false); 1036 proto = htons(ETH_P_ERSPAN); 1037 } else if (t->parms.erspan_ver == 2) { 1038 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1039 t->parms.dir, 1040 t->parms.hwid, 1041 truncate, false); 1042 proto = htons(ETH_P_ERSPAN2); 1043 } else { 1044 goto tx_err; 1045 } 1046 1047 fl6.daddr = t->parms.raddr; 1048 } 1049 1050 /* Push GRE header. */ 1051 __set_bit(IP_TUNNEL_SEQ_BIT, flags); 1052 gre_build_header(skb, 8, flags, proto, 0, 1053 htonl(atomic_fetch_inc(&t->o_seqno))); 1054 1055 /* TooBig packet may have updated dst->dev's mtu */ 1056 if (!t->parms.collect_md && dst) { 1057 mtu = READ_ONCE(dst_dev(dst)->mtu); 1058 if (dst_mtu(dst) > mtu) 1059 dst->ops->update_pmtu(dst, NULL, skb, mtu, false); 1060 } 1061 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1062 NEXTHDR_GRE); 1063 if (err != 0) { 1064 /* XXX: send ICMP error even if DF is not set. */ 1065 if (err == -EMSGSIZE) { 1066 if (skb->protocol == htons(ETH_P_IP)) 1067 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1068 ICMP_FRAG_NEEDED, htonl(mtu)); 1069 else 1070 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1071 } 1072 1073 goto tx_err; 1074 } 1075 return NETDEV_TX_OK; 1076 1077 tx_err: 1078 if (!IS_ERR(tun_info)) 1079 DEV_STATS_INC(dev, tx_errors); 1080 DEV_STATS_INC(dev, tx_dropped); 1081 kfree_skb(skb); 1082 return NETDEV_TX_OK; 1083 } 1084 1085 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1086 { 1087 struct net_device *dev = t->dev; 1088 struct __ip6_tnl_parm *p = &t->parms; 1089 struct flowi6 *fl6 = &t->fl.u.ip6; 1090 1091 if (dev->type != ARPHRD_ETHER) { 1092 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1093 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1094 } 1095 1096 /* Set up flowi template */ 1097 fl6->saddr = p->laddr; 1098 fl6->daddr = p->raddr; 1099 fl6->flowi6_oif = p->link; 1100 fl6->flowlabel = 0; 1101 fl6->flowi6_proto = IPPROTO_GRE; 1102 fl6->fl6_gre_key = t->parms.o_key; 1103 1104 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1105 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1106 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1107 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1108 1109 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1110 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1111 1112 if (p->flags&IP6_TNL_F_CAP_XMIT && 1113 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1114 dev->flags |= IFF_POINTOPOINT; 1115 else 1116 dev->flags &= ~IFF_POINTOPOINT; 1117 } 1118 1119 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1120 int t_hlen) 1121 { 1122 const struct __ip6_tnl_parm *p = &t->parms; 1123 struct net_device *dev = t->dev; 1124 1125 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1126 int strict = (ipv6_addr_type(&p->raddr) & 1127 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1128 1129 struct rt6_info *rt = rt6_lookup(t->net, 1130 &p->raddr, &p->laddr, 1131 p->link, NULL, strict); 1132 1133 if (!rt) 1134 return; 1135 1136 if (rt->dst.dev) { 1137 unsigned short dst_len = rt->dst.dev->hard_header_len + 1138 t_hlen; 1139 1140 if (t->dev->header_ops) 1141 dev->hard_header_len = dst_len; 1142 else 1143 dev->needed_headroom = dst_len; 1144 1145 if (set_mtu) { 1146 int mtu = rt->dst.dev->mtu - t_hlen; 1147 1148 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1149 mtu -= 8; 1150 if (dev->type == ARPHRD_ETHER) 1151 mtu -= ETH_HLEN; 1152 1153 if (mtu < IPV6_MIN_MTU) 1154 mtu = IPV6_MIN_MTU; 1155 WRITE_ONCE(dev->mtu, mtu); 1156 } 1157 } 1158 ip6_rt_put(rt); 1159 } 1160 } 1161 1162 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1163 { 1164 int t_hlen; 1165 1166 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1167 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1168 1169 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1170 1171 if (tunnel->dev->header_ops) 1172 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1173 else 1174 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1175 1176 return t_hlen; 1177 } 1178 1179 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1180 { 1181 ip6gre_tnl_link_config_common(t); 1182 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1183 } 1184 1185 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1186 const struct __ip6_tnl_parm *p) 1187 { 1188 t->parms.laddr = p->laddr; 1189 t->parms.raddr = p->raddr; 1190 t->parms.flags = p->flags; 1191 t->parms.hop_limit = p->hop_limit; 1192 t->parms.encap_limit = p->encap_limit; 1193 t->parms.flowinfo = p->flowinfo; 1194 t->parms.link = p->link; 1195 t->parms.proto = p->proto; 1196 t->parms.i_key = p->i_key; 1197 t->parms.o_key = p->o_key; 1198 ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags); 1199 ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags); 1200 t->parms.fwmark = p->fwmark; 1201 t->parms.erspan_ver = p->erspan_ver; 1202 t->parms.index = p->index; 1203 t->parms.dir = p->dir; 1204 t->parms.hwid = p->hwid; 1205 dst_cache_reset(&t->dst_cache); 1206 } 1207 1208 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1209 int set_mtu) 1210 { 1211 ip6gre_tnl_copy_tnl_parm(t, p); 1212 ip6gre_tnl_link_config(t, set_mtu); 1213 return 0; 1214 } 1215 1216 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1217 const struct ip6_tnl_parm2 *u) 1218 { 1219 p->laddr = u->laddr; 1220 p->raddr = u->raddr; 1221 p->flags = u->flags; 1222 p->hop_limit = u->hop_limit; 1223 p->encap_limit = u->encap_limit; 1224 p->flowinfo = u->flowinfo; 1225 p->link = u->link; 1226 p->i_key = u->i_key; 1227 p->o_key = u->o_key; 1228 gre_flags_to_tnl_flags(p->i_flags, u->i_flags); 1229 gre_flags_to_tnl_flags(p->o_flags, u->o_flags); 1230 memcpy(p->name, u->name, sizeof(u->name)); 1231 } 1232 1233 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1234 const struct __ip6_tnl_parm *p) 1235 { 1236 u->proto = IPPROTO_GRE; 1237 u->laddr = p->laddr; 1238 u->raddr = p->raddr; 1239 u->flags = p->flags; 1240 u->hop_limit = p->hop_limit; 1241 u->encap_limit = p->encap_limit; 1242 u->flowinfo = p->flowinfo; 1243 u->link = p->link; 1244 u->i_key = p->i_key; 1245 u->o_key = p->o_key; 1246 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1247 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1248 memcpy(u->name, p->name, sizeof(u->name)); 1249 } 1250 1251 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1252 struct ifreq *ifr, void __user *data, 1253 int cmd) 1254 { 1255 int err = 0; 1256 struct ip6_tnl_parm2 p; 1257 struct __ip6_tnl_parm p1; 1258 struct ip6_tnl *t = netdev_priv(dev); 1259 struct net *net = t->net; 1260 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1261 1262 memset(&p1, 0, sizeof(p1)); 1263 1264 switch (cmd) { 1265 case SIOCGETTUNNEL: 1266 if (dev == ign->fb_tunnel_dev) { 1267 if (copy_from_user(&p, data, sizeof(p))) { 1268 err = -EFAULT; 1269 break; 1270 } 1271 ip6gre_tnl_parm_from_user(&p1, &p); 1272 t = ip6gre_tunnel_locate(net, &p1, 0); 1273 if (!t) 1274 t = netdev_priv(dev); 1275 } 1276 memset(&p, 0, sizeof(p)); 1277 ip6gre_tnl_parm_to_user(&p, &t->parms); 1278 if (copy_to_user(data, &p, sizeof(p))) 1279 err = -EFAULT; 1280 break; 1281 1282 case SIOCADDTUNNEL: 1283 case SIOCCHGTUNNEL: 1284 err = -EPERM; 1285 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1286 goto done; 1287 1288 err = -EFAULT; 1289 if (copy_from_user(&p, data, sizeof(p))) 1290 goto done; 1291 1292 err = -EINVAL; 1293 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1294 goto done; 1295 1296 if (!(p.i_flags&GRE_KEY)) 1297 p.i_key = 0; 1298 if (!(p.o_flags&GRE_KEY)) 1299 p.o_key = 0; 1300 1301 ip6gre_tnl_parm_from_user(&p1, &p); 1302 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1303 1304 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1305 if (t) { 1306 if (t->dev != dev) { 1307 err = -EEXIST; 1308 break; 1309 } 1310 } else { 1311 t = netdev_priv(dev); 1312 1313 ip6gre_tunnel_unlink(ign, t); 1314 synchronize_net(); 1315 ip6gre_tnl_change(t, &p1, 1); 1316 ip6gre_tunnel_link(ign, t); 1317 netdev_state_change(dev); 1318 } 1319 } 1320 1321 if (t) { 1322 err = 0; 1323 1324 memset(&p, 0, sizeof(p)); 1325 ip6gre_tnl_parm_to_user(&p, &t->parms); 1326 if (copy_to_user(data, &p, sizeof(p))) 1327 err = -EFAULT; 1328 } else 1329 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1330 break; 1331 1332 case SIOCDELTUNNEL: 1333 err = -EPERM; 1334 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1335 goto done; 1336 1337 if (dev == ign->fb_tunnel_dev) { 1338 err = -EFAULT; 1339 if (copy_from_user(&p, data, sizeof(p))) 1340 goto done; 1341 err = -ENOENT; 1342 ip6gre_tnl_parm_from_user(&p1, &p); 1343 t = ip6gre_tunnel_locate(net, &p1, 0); 1344 if (!t) 1345 goto done; 1346 err = -EPERM; 1347 if (t == netdev_priv(ign->fb_tunnel_dev)) 1348 goto done; 1349 dev = t->dev; 1350 } 1351 unregister_netdevice(dev); 1352 err = 0; 1353 break; 1354 1355 default: 1356 err = -EINVAL; 1357 } 1358 1359 done: 1360 return err; 1361 } 1362 1363 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1364 unsigned short type, const void *daddr, 1365 const void *saddr, unsigned int len) 1366 { 1367 struct ip6_tnl *t = netdev_priv(dev); 1368 struct ipv6hdr *ipv6h; 1369 __be16 *p; 1370 1371 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1372 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1373 t->fl.u.ip6.flowlabel, 1374 true, &t->fl.u.ip6)); 1375 ipv6h->hop_limit = t->parms.hop_limit; 1376 ipv6h->nexthdr = NEXTHDR_GRE; 1377 ipv6h->saddr = t->parms.laddr; 1378 ipv6h->daddr = t->parms.raddr; 1379 1380 p = (__be16 *)(ipv6h + 1); 1381 p[0] = ip_tunnel_flags_to_be16(t->parms.o_flags); 1382 p[1] = htons(type); 1383 1384 /* 1385 * Set the source hardware address. 1386 */ 1387 1388 if (saddr) 1389 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1390 if (daddr) 1391 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1392 if (!ipv6_addr_any(&ipv6h->daddr)) 1393 return t->hlen; 1394 1395 return -t->hlen; 1396 } 1397 1398 static const struct header_ops ip6gre_header_ops = { 1399 .create = ip6gre_header, 1400 }; 1401 1402 static const struct net_device_ops ip6gre_netdev_ops = { 1403 .ndo_init = ip6gre_tunnel_init, 1404 .ndo_uninit = ip6gre_tunnel_uninit, 1405 .ndo_start_xmit = ip6gre_tunnel_xmit, 1406 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1407 .ndo_change_mtu = ip6_tnl_change_mtu, 1408 .ndo_get_iflink = ip6_tnl_get_iflink, 1409 }; 1410 1411 static void ip6gre_dev_free(struct net_device *dev) 1412 { 1413 struct ip6_tnl *t = netdev_priv(dev); 1414 1415 gro_cells_destroy(&t->gro_cells); 1416 dst_cache_destroy(&t->dst_cache); 1417 } 1418 1419 static void ip6gre_tunnel_setup(struct net_device *dev) 1420 { 1421 dev->netdev_ops = &ip6gre_netdev_ops; 1422 dev->needs_free_netdev = true; 1423 dev->priv_destructor = ip6gre_dev_free; 1424 1425 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 1426 dev->type = ARPHRD_IP6GRE; 1427 1428 dev->flags |= IFF_NOARP; 1429 dev->addr_len = sizeof(struct in6_addr); 1430 netif_keep_dst(dev); 1431 /* This perm addr will be used as interface identifier by IPv6 */ 1432 dev->addr_assign_type = NET_ADDR_RANDOM; 1433 eth_random_addr(dev->perm_addr); 1434 } 1435 1436 #define GRE6_FEATURES (NETIF_F_SG | \ 1437 NETIF_F_FRAGLIST | \ 1438 NETIF_F_HIGHDMA | \ 1439 NETIF_F_HW_CSUM) 1440 1441 static void ip6gre_tnl_init_features(struct net_device *dev) 1442 { 1443 struct ip6_tnl *nt = netdev_priv(dev); 1444 1445 dev->features |= GRE6_FEATURES; 1446 dev->hw_features |= GRE6_FEATURES; 1447 1448 /* TCP offload with GRE SEQ is not supported, nor can we support 2 1449 * levels of outer headers requiring an update. 1450 */ 1451 if (test_bit(IP_TUNNEL_SEQ_BIT, nt->parms.o_flags)) 1452 return; 1453 if (test_bit(IP_TUNNEL_CSUM_BIT, nt->parms.o_flags) && 1454 nt->encap.type != TUNNEL_ENCAP_NONE) 1455 return; 1456 1457 dev->features |= NETIF_F_GSO_SOFTWARE; 1458 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1459 1460 dev->lltx = true; 1461 } 1462 1463 static int ip6gre_tunnel_init_common(struct net_device *dev) 1464 { 1465 struct ip6_tnl *tunnel; 1466 int ret; 1467 int t_hlen; 1468 1469 tunnel = netdev_priv(dev); 1470 1471 tunnel->dev = dev; 1472 strcpy(tunnel->parms.name, dev->name); 1473 1474 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1475 if (ret) 1476 return ret; 1477 1478 ret = gro_cells_init(&tunnel->gro_cells, dev); 1479 if (ret) 1480 goto cleanup_dst_cache_init; 1481 1482 t_hlen = ip6gre_calc_hlen(tunnel); 1483 dev->mtu = ETH_DATA_LEN - t_hlen; 1484 if (dev->type == ARPHRD_ETHER) 1485 dev->mtu -= ETH_HLEN; 1486 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1487 dev->mtu -= 8; 1488 1489 if (tunnel->parms.collect_md) { 1490 netif_keep_dst(dev); 1491 } 1492 ip6gre_tnl_init_features(dev); 1493 1494 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1495 netdev_lockdep_set_classes(dev); 1496 return 0; 1497 1498 cleanup_dst_cache_init: 1499 dst_cache_destroy(&tunnel->dst_cache); 1500 return ret; 1501 } 1502 1503 static int ip6gre_tunnel_init(struct net_device *dev) 1504 { 1505 struct ip6_tnl *tunnel; 1506 int ret; 1507 1508 ret = ip6gre_tunnel_init_common(dev); 1509 if (ret) 1510 return ret; 1511 1512 tunnel = netdev_priv(dev); 1513 1514 if (tunnel->parms.collect_md) 1515 return 0; 1516 1517 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1518 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1519 1520 if (ipv6_addr_any(&tunnel->parms.raddr)) 1521 dev->header_ops = &ip6gre_header_ops; 1522 1523 return 0; 1524 } 1525 1526 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1527 { 1528 struct ip6_tnl *tunnel = netdev_priv(dev); 1529 1530 tunnel->dev = dev; 1531 tunnel->net = dev_net(dev); 1532 strcpy(tunnel->parms.name, dev->name); 1533 1534 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1535 } 1536 1537 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1538 .handler = gre_rcv, 1539 .err_handler = ip6gre_err, 1540 .flags = INET6_PROTO_FINAL, 1541 }; 1542 1543 static void __net_exit ip6gre_exit_rtnl_net(struct net *net, struct list_head *head) 1544 { 1545 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1546 struct net_device *dev, *aux; 1547 int prio; 1548 1549 for_each_netdev_safe(net, dev, aux) 1550 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1551 dev->rtnl_link_ops == &ip6gre_tap_ops || 1552 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1553 unregister_netdevice_queue(dev, head); 1554 1555 for (prio = 0; prio < 4; prio++) { 1556 int h; 1557 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1558 struct ip6_tnl *t; 1559 1560 t = rtnl_net_dereference(net, ign->tunnels[prio][h]); 1561 1562 while (t) { 1563 /* If dev is in the same netns, it has already 1564 * been added to the list by the previous loop. 1565 */ 1566 if (!net_eq(dev_net(t->dev), net)) 1567 unregister_netdevice_queue(t->dev, head); 1568 1569 t = rtnl_net_dereference(net, t->next); 1570 } 1571 } 1572 } 1573 } 1574 1575 static int __net_init ip6gre_init_net(struct net *net) 1576 { 1577 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1578 struct net_device *ndev; 1579 int err; 1580 1581 if (!net_has_fallback_tunnels(net)) 1582 return 0; 1583 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1584 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1585 if (!ndev) { 1586 err = -ENOMEM; 1587 goto err_alloc_dev; 1588 } 1589 ign->fb_tunnel_dev = ndev; 1590 dev_net_set(ign->fb_tunnel_dev, net); 1591 /* FB netdevice is special: we have one, and only one per netns. 1592 * Allowing to move it to another netns is clearly unsafe. 1593 */ 1594 ign->fb_tunnel_dev->netns_immutable = true; 1595 1596 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1597 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1598 1599 err = register_netdev(ign->fb_tunnel_dev); 1600 if (err) 1601 goto err_reg_dev; 1602 1603 rcu_assign_pointer(ign->tunnels_wc[0], 1604 netdev_priv(ign->fb_tunnel_dev)); 1605 return 0; 1606 1607 err_reg_dev: 1608 free_netdev(ndev); 1609 err_alloc_dev: 1610 return err; 1611 } 1612 1613 static struct pernet_operations ip6gre_net_ops = { 1614 .init = ip6gre_init_net, 1615 .exit_rtnl = ip6gre_exit_rtnl_net, 1616 .id = &ip6gre_net_id, 1617 .size = sizeof(struct ip6gre_net), 1618 }; 1619 1620 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1621 struct netlink_ext_ack *extack) 1622 { 1623 __be16 flags; 1624 1625 if (!data) 1626 return 0; 1627 1628 flags = 0; 1629 if (data[IFLA_GRE_IFLAGS]) 1630 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1631 if (data[IFLA_GRE_OFLAGS]) 1632 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1633 if (flags & (GRE_VERSION|GRE_ROUTING)) 1634 return -EINVAL; 1635 1636 return 0; 1637 } 1638 1639 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1640 struct netlink_ext_ack *extack) 1641 { 1642 struct in6_addr daddr; 1643 1644 if (tb[IFLA_ADDRESS]) { 1645 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1646 return -EINVAL; 1647 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1648 return -EADDRNOTAVAIL; 1649 } 1650 1651 if (!data) 1652 goto out; 1653 1654 if (data[IFLA_GRE_REMOTE]) { 1655 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1656 if (ipv6_addr_any(&daddr)) 1657 return -EINVAL; 1658 } 1659 1660 out: 1661 return ip6gre_tunnel_validate(tb, data, extack); 1662 } 1663 1664 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1665 struct netlink_ext_ack *extack) 1666 { 1667 __be16 flags = 0; 1668 int ret, ver = 0; 1669 1670 if (!data) 1671 return 0; 1672 1673 ret = ip6gre_tap_validate(tb, data, extack); 1674 if (ret) 1675 return ret; 1676 1677 /* ERSPAN should only have GRE sequence and key flag */ 1678 if (data[IFLA_GRE_OFLAGS]) 1679 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1680 if (data[IFLA_GRE_IFLAGS]) 1681 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1682 if (!data[IFLA_GRE_COLLECT_METADATA] && 1683 flags != (GRE_SEQ | GRE_KEY)) 1684 return -EINVAL; 1685 1686 /* ERSPAN Session ID only has 10-bit. Since we reuse 1687 * 32-bit key field as ID, check it's range. 1688 */ 1689 if (data[IFLA_GRE_IKEY] && 1690 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1691 return -EINVAL; 1692 1693 if (data[IFLA_GRE_OKEY] && 1694 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1695 return -EINVAL; 1696 1697 if (data[IFLA_GRE_ERSPAN_VER]) { 1698 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1699 if (ver != 1 && ver != 2) 1700 return -EINVAL; 1701 } 1702 1703 if (ver == 1) { 1704 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1705 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1706 1707 if (index & ~INDEX_MASK) 1708 return -EINVAL; 1709 } 1710 } else if (ver == 2) { 1711 if (data[IFLA_GRE_ERSPAN_DIR]) { 1712 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1713 1714 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1715 return -EINVAL; 1716 } 1717 1718 if (data[IFLA_GRE_ERSPAN_HWID]) { 1719 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1720 1721 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1722 return -EINVAL; 1723 } 1724 } 1725 1726 return 0; 1727 } 1728 1729 static void ip6erspan_set_version(struct nlattr *data[], 1730 struct __ip6_tnl_parm *parms) 1731 { 1732 if (!data) 1733 return; 1734 1735 parms->erspan_ver = 1; 1736 if (data[IFLA_GRE_ERSPAN_VER]) 1737 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1738 1739 if (parms->erspan_ver == 1) { 1740 if (data[IFLA_GRE_ERSPAN_INDEX]) 1741 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1742 } else if (parms->erspan_ver == 2) { 1743 if (data[IFLA_GRE_ERSPAN_DIR]) 1744 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1745 if (data[IFLA_GRE_ERSPAN_HWID]) 1746 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1747 } 1748 } 1749 1750 static void ip6gre_netlink_parms(struct nlattr *data[], 1751 struct __ip6_tnl_parm *parms) 1752 { 1753 memset(parms, 0, sizeof(*parms)); 1754 1755 if (!data) 1756 return; 1757 1758 if (data[IFLA_GRE_LINK]) 1759 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1760 1761 if (data[IFLA_GRE_IFLAGS]) 1762 gre_flags_to_tnl_flags(parms->i_flags, 1763 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1764 1765 if (data[IFLA_GRE_OFLAGS]) 1766 gre_flags_to_tnl_flags(parms->o_flags, 1767 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1768 1769 if (data[IFLA_GRE_IKEY]) 1770 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1771 1772 if (data[IFLA_GRE_OKEY]) 1773 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1774 1775 if (data[IFLA_GRE_LOCAL]) 1776 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1777 1778 if (data[IFLA_GRE_REMOTE]) 1779 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1780 1781 if (data[IFLA_GRE_TTL]) 1782 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1783 1784 if (data[IFLA_GRE_ENCAP_LIMIT]) 1785 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1786 1787 if (data[IFLA_GRE_FLOWINFO]) 1788 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1789 1790 if (data[IFLA_GRE_FLAGS]) 1791 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1792 1793 if (data[IFLA_GRE_FWMARK]) 1794 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1795 1796 if (data[IFLA_GRE_COLLECT_METADATA]) 1797 parms->collect_md = true; 1798 } 1799 1800 static int ip6gre_tap_init(struct net_device *dev) 1801 { 1802 int ret; 1803 1804 ret = ip6gre_tunnel_init_common(dev); 1805 if (ret) 1806 return ret; 1807 1808 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1809 1810 return 0; 1811 } 1812 1813 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1814 .ndo_init = ip6gre_tap_init, 1815 .ndo_uninit = ip6gre_tunnel_uninit, 1816 .ndo_start_xmit = ip6gre_tunnel_xmit, 1817 .ndo_set_mac_address = eth_mac_addr, 1818 .ndo_validate_addr = eth_validate_addr, 1819 .ndo_change_mtu = ip6_tnl_change_mtu, 1820 .ndo_get_iflink = ip6_tnl_get_iflink, 1821 }; 1822 1823 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1824 { 1825 int t_hlen; 1826 1827 tunnel->tun_hlen = 8; 1828 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1829 erspan_hdr_len(tunnel->parms.erspan_ver); 1830 1831 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1832 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1833 return t_hlen; 1834 } 1835 1836 static int ip6erspan_tap_init(struct net_device *dev) 1837 { 1838 struct ip6_tnl *tunnel; 1839 int t_hlen; 1840 int ret; 1841 1842 tunnel = netdev_priv(dev); 1843 1844 tunnel->dev = dev; 1845 strcpy(tunnel->parms.name, dev->name); 1846 1847 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1848 if (ret) 1849 return ret; 1850 1851 ret = gro_cells_init(&tunnel->gro_cells, dev); 1852 if (ret) 1853 goto cleanup_dst_cache_init; 1854 1855 t_hlen = ip6erspan_calc_hlen(tunnel); 1856 dev->mtu = ETH_DATA_LEN - t_hlen; 1857 if (dev->type == ARPHRD_ETHER) 1858 dev->mtu -= ETH_HLEN; 1859 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1860 dev->mtu -= 8; 1861 1862 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1863 ip6erspan_tnl_link_config(tunnel, 1); 1864 1865 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1866 netdev_lockdep_set_classes(dev); 1867 return 0; 1868 1869 cleanup_dst_cache_init: 1870 dst_cache_destroy(&tunnel->dst_cache); 1871 return ret; 1872 } 1873 1874 static const struct net_device_ops ip6erspan_netdev_ops = { 1875 .ndo_init = ip6erspan_tap_init, 1876 .ndo_uninit = ip6erspan_tunnel_uninit, 1877 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1878 .ndo_set_mac_address = eth_mac_addr, 1879 .ndo_validate_addr = eth_validate_addr, 1880 .ndo_change_mtu = ip6_tnl_change_mtu, 1881 .ndo_get_iflink = ip6_tnl_get_iflink, 1882 }; 1883 1884 static void ip6gre_tap_setup(struct net_device *dev) 1885 { 1886 1887 ether_setup(dev); 1888 1889 dev->max_mtu = 0; 1890 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1891 dev->needs_free_netdev = true; 1892 dev->priv_destructor = ip6gre_dev_free; 1893 1894 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 1895 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1896 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1897 netif_keep_dst(dev); 1898 } 1899 1900 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1901 struct ip_tunnel_encap *ipencap) 1902 { 1903 bool ret = false; 1904 1905 memset(ipencap, 0, sizeof(*ipencap)); 1906 1907 if (!data) 1908 return ret; 1909 1910 if (data[IFLA_GRE_ENCAP_TYPE]) { 1911 ret = true; 1912 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1913 } 1914 1915 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1916 ret = true; 1917 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1918 } 1919 1920 if (data[IFLA_GRE_ENCAP_SPORT]) { 1921 ret = true; 1922 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1923 } 1924 1925 if (data[IFLA_GRE_ENCAP_DPORT]) { 1926 ret = true; 1927 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1928 } 1929 1930 return ret; 1931 } 1932 1933 static int ip6gre_newlink_common(struct net *link_net, struct net_device *dev, 1934 struct nlattr *tb[], struct nlattr *data[], 1935 struct netlink_ext_ack *extack) 1936 { 1937 struct ip6_tnl *nt; 1938 struct ip_tunnel_encap ipencap; 1939 int err; 1940 1941 nt = netdev_priv(dev); 1942 1943 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1944 int err = ip6_tnl_encap_setup(nt, &ipencap); 1945 1946 if (err < 0) 1947 return err; 1948 } 1949 1950 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1951 eth_hw_addr_random(dev); 1952 1953 nt->dev = dev; 1954 nt->net = link_net; 1955 1956 err = register_netdevice(dev); 1957 if (err) 1958 goto out; 1959 1960 if (tb[IFLA_MTU]) 1961 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 1962 1963 out: 1964 return err; 1965 } 1966 1967 static int ip6gre_newlink(struct net_device *dev, 1968 struct rtnl_newlink_params *params, 1969 struct netlink_ext_ack *extack) 1970 { 1971 struct net *net = params->link_net ? : dev_net(dev); 1972 struct ip6_tnl *nt = netdev_priv(dev); 1973 struct nlattr **data = params->data; 1974 struct nlattr **tb = params->tb; 1975 struct ip6gre_net *ign; 1976 int err; 1977 1978 ip6gre_netlink_parms(data, &nt->parms); 1979 ign = net_generic(net, ip6gre_net_id); 1980 1981 if (nt->parms.collect_md) { 1982 if (rtnl_dereference(ign->collect_md_tun)) 1983 return -EEXIST; 1984 } else { 1985 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1986 return -EEXIST; 1987 } 1988 1989 err = ip6gre_newlink_common(net, dev, tb, data, extack); 1990 if (!err) { 1991 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1992 ip6gre_tunnel_link_md(ign, nt); 1993 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 1994 } 1995 return err; 1996 } 1997 1998 static struct ip6_tnl * 1999 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2000 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2001 struct netlink_ext_ack *extack) 2002 { 2003 struct ip6_tnl *t, *nt = netdev_priv(dev); 2004 struct net *net = nt->net; 2005 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2006 struct ip_tunnel_encap ipencap; 2007 2008 if (dev == ign->fb_tunnel_dev) 2009 return ERR_PTR(-EINVAL); 2010 2011 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2012 int err = ip6_tnl_encap_setup(nt, &ipencap); 2013 2014 if (err < 0) 2015 return ERR_PTR(err); 2016 } 2017 2018 ip6gre_netlink_parms(data, p_p); 2019 2020 t = ip6gre_tunnel_locate(net, p_p, 0); 2021 2022 if (t) { 2023 if (t->dev != dev) 2024 return ERR_PTR(-EEXIST); 2025 } else { 2026 t = nt; 2027 } 2028 2029 return t; 2030 } 2031 2032 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2033 struct nlattr *data[], 2034 struct netlink_ext_ack *extack) 2035 { 2036 struct ip6_tnl *t = netdev_priv(dev); 2037 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2038 struct __ip6_tnl_parm p; 2039 2040 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2041 if (IS_ERR(t)) 2042 return PTR_ERR(t); 2043 2044 ip6gre_tunnel_unlink_md(ign, t); 2045 ip6gre_tunnel_unlink(ign, t); 2046 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2047 ip6gre_tunnel_link_md(ign, t); 2048 ip6gre_tunnel_link(ign, t); 2049 return 0; 2050 } 2051 2052 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2053 { 2054 struct net *net = dev_net(dev); 2055 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2056 2057 if (dev != ign->fb_tunnel_dev) 2058 unregister_netdevice_queue(dev, head); 2059 } 2060 2061 static size_t ip6gre_get_size(const struct net_device *dev) 2062 { 2063 return 2064 /* IFLA_GRE_LINK */ 2065 nla_total_size(4) + 2066 /* IFLA_GRE_IFLAGS */ 2067 nla_total_size(2) + 2068 /* IFLA_GRE_OFLAGS */ 2069 nla_total_size(2) + 2070 /* IFLA_GRE_IKEY */ 2071 nla_total_size(4) + 2072 /* IFLA_GRE_OKEY */ 2073 nla_total_size(4) + 2074 /* IFLA_GRE_LOCAL */ 2075 nla_total_size(sizeof(struct in6_addr)) + 2076 /* IFLA_GRE_REMOTE */ 2077 nla_total_size(sizeof(struct in6_addr)) + 2078 /* IFLA_GRE_TTL */ 2079 nla_total_size(1) + 2080 /* IFLA_GRE_ENCAP_LIMIT */ 2081 nla_total_size(1) + 2082 /* IFLA_GRE_FLOWINFO */ 2083 nla_total_size(4) + 2084 /* IFLA_GRE_FLAGS */ 2085 nla_total_size(4) + 2086 /* IFLA_GRE_ENCAP_TYPE */ 2087 nla_total_size(2) + 2088 /* IFLA_GRE_ENCAP_FLAGS */ 2089 nla_total_size(2) + 2090 /* IFLA_GRE_ENCAP_SPORT */ 2091 nla_total_size(2) + 2092 /* IFLA_GRE_ENCAP_DPORT */ 2093 nla_total_size(2) + 2094 /* IFLA_GRE_COLLECT_METADATA */ 2095 nla_total_size(0) + 2096 /* IFLA_GRE_FWMARK */ 2097 nla_total_size(4) + 2098 /* IFLA_GRE_ERSPAN_INDEX */ 2099 nla_total_size(4) + 2100 0; 2101 } 2102 2103 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2104 { 2105 struct ip6_tnl *t = netdev_priv(dev); 2106 struct __ip6_tnl_parm *p = &t->parms; 2107 IP_TUNNEL_DECLARE_FLAGS(o_flags); 2108 2109 ip_tunnel_flags_copy(o_flags, p->o_flags); 2110 2111 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2112 if (!p->collect_md) 2113 __set_bit(IP_TUNNEL_KEY_BIT, o_flags); 2114 2115 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2116 goto nla_put_failure; 2117 2118 if (p->erspan_ver == 1) { 2119 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2120 goto nla_put_failure; 2121 } else { 2122 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2123 goto nla_put_failure; 2124 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2125 goto nla_put_failure; 2126 } 2127 } 2128 2129 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2130 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2131 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2132 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2133 gre_tnl_flags_to_gre_flags(o_flags)) || 2134 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2135 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2136 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2137 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2138 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2139 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2140 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2141 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2142 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2143 goto nla_put_failure; 2144 2145 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2146 t->encap.type) || 2147 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2148 t->encap.sport) || 2149 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2150 t->encap.dport) || 2151 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2152 t->encap.flags)) 2153 goto nla_put_failure; 2154 2155 if (p->collect_md) { 2156 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2157 goto nla_put_failure; 2158 } 2159 2160 return 0; 2161 2162 nla_put_failure: 2163 return -EMSGSIZE; 2164 } 2165 2166 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2167 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2168 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2169 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2170 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2171 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2172 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2173 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2174 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2175 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2176 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2177 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2178 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2179 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2180 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2181 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2182 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2183 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2184 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2185 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2186 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2187 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2188 }; 2189 2190 static void ip6erspan_tap_setup(struct net_device *dev) 2191 { 2192 ether_setup(dev); 2193 2194 dev->max_mtu = 0; 2195 dev->netdev_ops = &ip6erspan_netdev_ops; 2196 dev->needs_free_netdev = true; 2197 dev->priv_destructor = ip6gre_dev_free; 2198 2199 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; 2200 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2201 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2202 netif_keep_dst(dev); 2203 } 2204 2205 static int ip6erspan_newlink(struct net_device *dev, 2206 struct rtnl_newlink_params *params, 2207 struct netlink_ext_ack *extack) 2208 { 2209 struct net *net = params->link_net ? : dev_net(dev); 2210 struct ip6_tnl *nt = netdev_priv(dev); 2211 struct nlattr **data = params->data; 2212 struct nlattr **tb = params->tb; 2213 struct ip6gre_net *ign; 2214 int err; 2215 2216 ip6gre_netlink_parms(data, &nt->parms); 2217 ip6erspan_set_version(data, &nt->parms); 2218 ign = net_generic(net, ip6gre_net_id); 2219 2220 if (nt->parms.collect_md) { 2221 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2222 return -EEXIST; 2223 } else { 2224 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2225 return -EEXIST; 2226 } 2227 2228 err = ip6gre_newlink_common(net, dev, tb, data, extack); 2229 if (!err) { 2230 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2231 ip6erspan_tunnel_link_md(ign, nt); 2232 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2233 } 2234 return err; 2235 } 2236 2237 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2238 { 2239 ip6gre_tnl_link_config_common(t); 2240 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2241 } 2242 2243 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2244 const struct __ip6_tnl_parm *p, int set_mtu) 2245 { 2246 ip6gre_tnl_copy_tnl_parm(t, p); 2247 ip6erspan_tnl_link_config(t, set_mtu); 2248 return 0; 2249 } 2250 2251 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2252 struct nlattr *data[], 2253 struct netlink_ext_ack *extack) 2254 { 2255 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2256 struct __ip6_tnl_parm p; 2257 struct ip6_tnl *t; 2258 2259 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2260 if (IS_ERR(t)) 2261 return PTR_ERR(t); 2262 2263 ip6erspan_set_version(data, &p); 2264 ip6gre_tunnel_unlink_md(ign, t); 2265 ip6gre_tunnel_unlink(ign, t); 2266 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2267 ip6erspan_tunnel_link_md(ign, t); 2268 ip6gre_tunnel_link(ign, t); 2269 return 0; 2270 } 2271 2272 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2273 .kind = "ip6gre", 2274 .maxtype = IFLA_GRE_MAX, 2275 .policy = ip6gre_policy, 2276 .priv_size = sizeof(struct ip6_tnl), 2277 .setup = ip6gre_tunnel_setup, 2278 .validate = ip6gre_tunnel_validate, 2279 .newlink = ip6gre_newlink, 2280 .changelink = ip6gre_changelink, 2281 .dellink = ip6gre_dellink, 2282 .get_size = ip6gre_get_size, 2283 .fill_info = ip6gre_fill_info, 2284 .get_link_net = ip6_tnl_get_link_net, 2285 }; 2286 2287 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2288 .kind = "ip6gretap", 2289 .maxtype = IFLA_GRE_MAX, 2290 .policy = ip6gre_policy, 2291 .priv_size = sizeof(struct ip6_tnl), 2292 .setup = ip6gre_tap_setup, 2293 .validate = ip6gre_tap_validate, 2294 .newlink = ip6gre_newlink, 2295 .changelink = ip6gre_changelink, 2296 .get_size = ip6gre_get_size, 2297 .fill_info = ip6gre_fill_info, 2298 .get_link_net = ip6_tnl_get_link_net, 2299 }; 2300 2301 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2302 .kind = "ip6erspan", 2303 .maxtype = IFLA_GRE_MAX, 2304 .policy = ip6gre_policy, 2305 .priv_size = sizeof(struct ip6_tnl), 2306 .setup = ip6erspan_tap_setup, 2307 .validate = ip6erspan_tap_validate, 2308 .newlink = ip6erspan_newlink, 2309 .changelink = ip6erspan_changelink, 2310 .get_size = ip6gre_get_size, 2311 .fill_info = ip6gre_fill_info, 2312 .get_link_net = ip6_tnl_get_link_net, 2313 }; 2314 2315 /* 2316 * And now the modules code and kernel interface. 2317 */ 2318 2319 static int __init ip6gre_init(void) 2320 { 2321 int err; 2322 2323 pr_info("GRE over IPv6 tunneling driver\n"); 2324 2325 err = register_pernet_device(&ip6gre_net_ops); 2326 if (err < 0) 2327 return err; 2328 2329 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2330 if (err < 0) { 2331 pr_info("%s: can't add protocol\n", __func__); 2332 goto add_proto_failed; 2333 } 2334 2335 err = rtnl_link_register(&ip6gre_link_ops); 2336 if (err < 0) 2337 goto rtnl_link_failed; 2338 2339 err = rtnl_link_register(&ip6gre_tap_ops); 2340 if (err < 0) 2341 goto tap_ops_failed; 2342 2343 err = rtnl_link_register(&ip6erspan_tap_ops); 2344 if (err < 0) 2345 goto erspan_link_failed; 2346 2347 out: 2348 return err; 2349 2350 erspan_link_failed: 2351 rtnl_link_unregister(&ip6gre_tap_ops); 2352 tap_ops_failed: 2353 rtnl_link_unregister(&ip6gre_link_ops); 2354 rtnl_link_failed: 2355 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2356 add_proto_failed: 2357 unregister_pernet_device(&ip6gre_net_ops); 2358 goto out; 2359 } 2360 2361 static void __exit ip6gre_fini(void) 2362 { 2363 rtnl_link_unregister(&ip6gre_tap_ops); 2364 rtnl_link_unregister(&ip6gre_link_ops); 2365 rtnl_link_unregister(&ip6erspan_tap_ops); 2366 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2367 unregister_pernet_device(&ip6gre_net_ops); 2368 } 2369 2370 module_init(ip6gre_init); 2371 module_exit(ip6gre_fini); 2372 MODULE_LICENSE("GPL"); 2373 MODULE_AUTHOR("D. Kozlov <xeb@mail.ru>"); 2374 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2375 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2376 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2377 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2378 MODULE_ALIAS_NETDEV("ip6gre0"); 2379