1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * GRE over IPv6 protocol decoder. 4 * 5 * Authors: Dmitry Kozlov (xeb@mail.ru) 6 */ 7 8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 9 10 #include <linux/capability.h> 11 #include <linux/module.h> 12 #include <linux/types.h> 13 #include <linux/kernel.h> 14 #include <linux/slab.h> 15 #include <linux/uaccess.h> 16 #include <linux/skbuff.h> 17 #include <linux/netdevice.h> 18 #include <linux/in.h> 19 #include <linux/tcp.h> 20 #include <linux/udp.h> 21 #include <linux/if_arp.h> 22 #include <linux/init.h> 23 #include <linux/in6.h> 24 #include <linux/inetdevice.h> 25 #include <linux/igmp.h> 26 #include <linux/netfilter_ipv4.h> 27 #include <linux/etherdevice.h> 28 #include <linux/if_ether.h> 29 #include <linux/hash.h> 30 #include <linux/if_tunnel.h> 31 #include <linux/ip6_tunnel.h> 32 33 #include <net/sock.h> 34 #include <net/ip.h> 35 #include <net/ip_tunnels.h> 36 #include <net/icmp.h> 37 #include <net/protocol.h> 38 #include <net/addrconf.h> 39 #include <net/arp.h> 40 #include <net/checksum.h> 41 #include <net/dsfield.h> 42 #include <net/inet_ecn.h> 43 #include <net/xfrm.h> 44 #include <net/net_namespace.h> 45 #include <net/netns/generic.h> 46 #include <net/rtnetlink.h> 47 48 #include <net/ipv6.h> 49 #include <net/ip6_fib.h> 50 #include <net/ip6_route.h> 51 #include <net/ip6_tunnel.h> 52 #include <net/gre.h> 53 #include <net/erspan.h> 54 #include <net/dst_metadata.h> 55 56 57 static bool log_ecn_error = true; 58 module_param(log_ecn_error, bool, 0644); 59 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 60 61 #define IP6_GRE_HASH_SIZE_SHIFT 5 62 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT) 63 64 static unsigned int ip6gre_net_id __read_mostly; 65 struct ip6gre_net { 66 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE]; 67 68 struct ip6_tnl __rcu *collect_md_tun; 69 struct ip6_tnl __rcu *collect_md_tun_erspan; 70 struct net_device *fb_tunnel_dev; 71 }; 72 73 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 74 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 75 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly; 76 static int ip6gre_tunnel_init(struct net_device *dev); 77 static void ip6gre_tunnel_setup(struct net_device *dev); 78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t); 79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu); 80 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu); 81 82 /* Tunnel hash table */ 83 84 /* 85 4 hash tables: 86 87 3: (remote,local) 88 2: (remote,*) 89 1: (*,local) 90 0: (*,*) 91 92 We require exact key match i.e. if a key is present in packet 93 it will match only tunnel with the same key; if it is not present, 94 it will match only keyless tunnel. 95 96 All keysless packets, if not matched configured keyless tunnels 97 will match fallback tunnel. 98 */ 99 100 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 101 static u32 HASH_ADDR(const struct in6_addr *addr) 102 { 103 u32 hash = ipv6_addr_hash(addr); 104 105 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 106 } 107 108 #define tunnels_r_l tunnels[3] 109 #define tunnels_r tunnels[2] 110 #define tunnels_l tunnels[1] 111 #define tunnels_wc tunnels[0] 112 113 /* Given src, dst and key, find appropriate for input tunnel. */ 114 115 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 116 const struct in6_addr *remote, const struct in6_addr *local, 117 __be32 key, __be16 gre_proto) 118 { 119 struct net *net = dev_net(dev); 120 int link = dev->ifindex; 121 unsigned int h0 = HASH_ADDR(remote); 122 unsigned int h1 = HASH_KEY(key); 123 struct ip6_tnl *t, *cand = NULL; 124 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 125 int dev_type = (gre_proto == htons(ETH_P_TEB) || 126 gre_proto == htons(ETH_P_ERSPAN) || 127 gre_proto == htons(ETH_P_ERSPAN2)) ? 128 ARPHRD_ETHER : ARPHRD_IP6GRE; 129 int score, cand_score = 4; 130 struct net_device *ndev; 131 132 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 133 if (!ipv6_addr_equal(local, &t->parms.laddr) || 134 !ipv6_addr_equal(remote, &t->parms.raddr) || 135 key != t->parms.i_key || 136 !(t->dev->flags & IFF_UP)) 137 continue; 138 139 if (t->dev->type != ARPHRD_IP6GRE && 140 t->dev->type != dev_type) 141 continue; 142 143 score = 0; 144 if (t->parms.link != link) 145 score |= 1; 146 if (t->dev->type != dev_type) 147 score |= 2; 148 if (score == 0) 149 return t; 150 151 if (score < cand_score) { 152 cand = t; 153 cand_score = score; 154 } 155 } 156 157 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 158 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 159 key != t->parms.i_key || 160 !(t->dev->flags & IFF_UP)) 161 continue; 162 163 if (t->dev->type != ARPHRD_IP6GRE && 164 t->dev->type != dev_type) 165 continue; 166 167 score = 0; 168 if (t->parms.link != link) 169 score |= 1; 170 if (t->dev->type != dev_type) 171 score |= 2; 172 if (score == 0) 173 return t; 174 175 if (score < cand_score) { 176 cand = t; 177 cand_score = score; 178 } 179 } 180 181 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 182 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 183 (!ipv6_addr_equal(local, &t->parms.raddr) || 184 !ipv6_addr_is_multicast(local))) || 185 key != t->parms.i_key || 186 !(t->dev->flags & IFF_UP)) 187 continue; 188 189 if (t->dev->type != ARPHRD_IP6GRE && 190 t->dev->type != dev_type) 191 continue; 192 193 score = 0; 194 if (t->parms.link != link) 195 score |= 1; 196 if (t->dev->type != dev_type) 197 score |= 2; 198 if (score == 0) 199 return t; 200 201 if (score < cand_score) { 202 cand = t; 203 cand_score = score; 204 } 205 } 206 207 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 208 if (t->parms.i_key != key || 209 !(t->dev->flags & IFF_UP)) 210 continue; 211 212 if (t->dev->type != ARPHRD_IP6GRE && 213 t->dev->type != dev_type) 214 continue; 215 216 score = 0; 217 if (t->parms.link != link) 218 score |= 1; 219 if (t->dev->type != dev_type) 220 score |= 2; 221 if (score == 0) 222 return t; 223 224 if (score < cand_score) { 225 cand = t; 226 cand_score = score; 227 } 228 } 229 230 if (cand) 231 return cand; 232 233 if (gre_proto == htons(ETH_P_ERSPAN) || 234 gre_proto == htons(ETH_P_ERSPAN2)) 235 t = rcu_dereference(ign->collect_md_tun_erspan); 236 else 237 t = rcu_dereference(ign->collect_md_tun); 238 239 if (t && t->dev->flags & IFF_UP) 240 return t; 241 242 ndev = READ_ONCE(ign->fb_tunnel_dev); 243 if (ndev && ndev->flags & IFF_UP) 244 return netdev_priv(ndev); 245 246 return NULL; 247 } 248 249 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 250 const struct __ip6_tnl_parm *p) 251 { 252 const struct in6_addr *remote = &p->raddr; 253 const struct in6_addr *local = &p->laddr; 254 unsigned int h = HASH_KEY(p->i_key); 255 int prio = 0; 256 257 if (!ipv6_addr_any(local)) 258 prio |= 1; 259 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 260 prio |= 2; 261 h ^= HASH_ADDR(remote); 262 } 263 264 return &ign->tunnels[prio][h]; 265 } 266 267 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 268 { 269 if (t->parms.collect_md) 270 rcu_assign_pointer(ign->collect_md_tun, t); 271 } 272 273 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t) 274 { 275 if (t->parms.collect_md) 276 rcu_assign_pointer(ign->collect_md_tun_erspan, t); 277 } 278 279 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t) 280 { 281 if (t->parms.collect_md) 282 rcu_assign_pointer(ign->collect_md_tun, NULL); 283 } 284 285 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign, 286 struct ip6_tnl *t) 287 { 288 if (t->parms.collect_md) 289 rcu_assign_pointer(ign->collect_md_tun_erspan, NULL); 290 } 291 292 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 293 const struct ip6_tnl *t) 294 { 295 return __ip6gre_bucket(ign, &t->parms); 296 } 297 298 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 299 { 300 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 301 302 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 303 rcu_assign_pointer(*tp, t); 304 } 305 306 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 307 { 308 struct ip6_tnl __rcu **tp; 309 struct ip6_tnl *iter; 310 311 for (tp = ip6gre_bucket(ign, t); 312 (iter = rtnl_dereference(*tp)) != NULL; 313 tp = &iter->next) { 314 if (t == iter) { 315 rcu_assign_pointer(*tp, t->next); 316 break; 317 } 318 } 319 } 320 321 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 322 const struct __ip6_tnl_parm *parms, 323 int type) 324 { 325 const struct in6_addr *remote = &parms->raddr; 326 const struct in6_addr *local = &parms->laddr; 327 __be32 key = parms->i_key; 328 int link = parms->link; 329 struct ip6_tnl *t; 330 struct ip6_tnl __rcu **tp; 331 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 332 333 for (tp = __ip6gre_bucket(ign, parms); 334 (t = rtnl_dereference(*tp)) != NULL; 335 tp = &t->next) 336 if (ipv6_addr_equal(local, &t->parms.laddr) && 337 ipv6_addr_equal(remote, &t->parms.raddr) && 338 key == t->parms.i_key && 339 link == t->parms.link && 340 type == t->dev->type) 341 break; 342 343 return t; 344 } 345 346 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 347 const struct __ip6_tnl_parm *parms, int create) 348 { 349 struct ip6_tnl *t, *nt; 350 struct net_device *dev; 351 char name[IFNAMSIZ]; 352 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 353 354 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 355 if (t && create) 356 return NULL; 357 if (t || !create) 358 return t; 359 360 if (parms->name[0]) { 361 if (!dev_valid_name(parms->name)) 362 return NULL; 363 strscpy(name, parms->name, IFNAMSIZ); 364 } else { 365 strcpy(name, "ip6gre%d"); 366 } 367 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 368 ip6gre_tunnel_setup); 369 if (!dev) 370 return NULL; 371 372 dev_net_set(dev, net); 373 374 nt = netdev_priv(dev); 375 nt->parms = *parms; 376 dev->rtnl_link_ops = &ip6gre_link_ops; 377 378 nt->dev = dev; 379 nt->net = dev_net(dev); 380 381 if (register_netdevice(dev) < 0) 382 goto failed_free; 383 384 ip6gre_tnl_link_config(nt, 1); 385 ip6gre_tunnel_link(ign, nt); 386 return nt; 387 388 failed_free: 389 free_netdev(dev); 390 return NULL; 391 } 392 393 static void ip6erspan_tunnel_uninit(struct net_device *dev) 394 { 395 struct ip6_tnl *t = netdev_priv(dev); 396 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 397 398 ip6erspan_tunnel_unlink_md(ign, t); 399 ip6gre_tunnel_unlink(ign, t); 400 dst_cache_reset(&t->dst_cache); 401 netdev_put(dev, &t->dev_tracker); 402 } 403 404 static void ip6gre_tunnel_uninit(struct net_device *dev) 405 { 406 struct ip6_tnl *t = netdev_priv(dev); 407 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 408 409 ip6gre_tunnel_unlink_md(ign, t); 410 ip6gre_tunnel_unlink(ign, t); 411 if (ign->fb_tunnel_dev == dev) 412 WRITE_ONCE(ign->fb_tunnel_dev, NULL); 413 dst_cache_reset(&t->dst_cache); 414 netdev_put(dev, &t->dev_tracker); 415 } 416 417 418 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 419 u8 type, u8 code, int offset, __be32 info) 420 { 421 struct net *net = dev_net(skb->dev); 422 const struct ipv6hdr *ipv6h; 423 struct tnl_ptk_info tpi; 424 struct ip6_tnl *t; 425 426 if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6), 427 offset) < 0) 428 return -EINVAL; 429 430 ipv6h = (const struct ipv6hdr *)skb->data; 431 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 432 tpi.key, tpi.proto); 433 if (!t) 434 return -ENOENT; 435 436 switch (type) { 437 case ICMPV6_DEST_UNREACH: 438 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 439 t->parms.name); 440 if (code != ICMPV6_PORT_UNREACH) 441 break; 442 return 0; 443 case ICMPV6_TIME_EXCEED: 444 if (code == ICMPV6_EXC_HOPLIMIT) { 445 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 446 t->parms.name); 447 break; 448 } 449 return 0; 450 case ICMPV6_PARAMPROB: { 451 struct ipv6_tlv_tnl_enc_lim *tel; 452 __u32 teli; 453 454 teli = 0; 455 if (code == ICMPV6_HDR_FIELD) 456 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 457 458 if (teli && teli == be32_to_cpu(info) - 2) { 459 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 460 if (tel->encap_limit == 0) { 461 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 462 t->parms.name); 463 } 464 } else { 465 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 466 t->parms.name); 467 } 468 return 0; 469 } 470 case ICMPV6_PKT_TOOBIG: 471 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL)); 472 return 0; 473 case NDISC_REDIRECT: 474 ip6_redirect(skb, net, skb->dev->ifindex, 0, 475 sock_net_uid(net, NULL)); 476 return 0; 477 } 478 479 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 480 t->err_count++; 481 else 482 t->err_count = 1; 483 t->err_time = jiffies; 484 485 return 0; 486 } 487 488 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 489 { 490 const struct ipv6hdr *ipv6h; 491 struct ip6_tnl *tunnel; 492 493 ipv6h = ipv6_hdr(skb); 494 tunnel = ip6gre_tunnel_lookup(skb->dev, 495 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 496 tpi->proto); 497 if (tunnel) { 498 if (tunnel->parms.collect_md) { 499 struct metadata_dst *tun_dst; 500 __be64 tun_id; 501 __be16 flags; 502 503 flags = tpi->flags; 504 tun_id = key32_to_tunnel_id(tpi->key); 505 506 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0); 507 if (!tun_dst) 508 return PACKET_REJECT; 509 510 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 511 } else { 512 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 513 } 514 515 return PACKET_RCVD; 516 } 517 518 return PACKET_REJECT; 519 } 520 521 static int ip6erspan_rcv(struct sk_buff *skb, 522 struct tnl_ptk_info *tpi, 523 int gre_hdr_len) 524 { 525 struct erspan_base_hdr *ershdr; 526 const struct ipv6hdr *ipv6h; 527 struct erspan_md2 *md2; 528 struct ip6_tnl *tunnel; 529 u8 ver; 530 531 if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr)))) 532 return PACKET_REJECT; 533 534 ipv6h = ipv6_hdr(skb); 535 ershdr = (struct erspan_base_hdr *)skb->data; 536 ver = ershdr->ver; 537 538 tunnel = ip6gre_tunnel_lookup(skb->dev, 539 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 540 tpi->proto); 541 if (tunnel) { 542 int len = erspan_hdr_len(ver); 543 544 if (unlikely(!pskb_may_pull(skb, len))) 545 return PACKET_REJECT; 546 547 if (__iptunnel_pull_header(skb, len, 548 htons(ETH_P_TEB), 549 false, false) < 0) 550 return PACKET_REJECT; 551 552 if (tunnel->parms.collect_md) { 553 struct erspan_metadata *pkt_md, *md; 554 struct metadata_dst *tun_dst; 555 struct ip_tunnel_info *info; 556 unsigned char *gh; 557 __be64 tun_id; 558 __be16 flags; 559 560 tpi->flags |= TUNNEL_KEY; 561 flags = tpi->flags; 562 tun_id = key32_to_tunnel_id(tpi->key); 563 564 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 565 sizeof(*md)); 566 if (!tun_dst) 567 return PACKET_REJECT; 568 569 /* skb can be uncloned in __iptunnel_pull_header, so 570 * old pkt_md is no longer valid and we need to reset 571 * it 572 */ 573 gh = skb_network_header(skb) + 574 skb_network_header_len(skb); 575 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 576 sizeof(*ershdr)); 577 info = &tun_dst->u.tun_info; 578 md = ip_tunnel_info_opts(info); 579 md->version = ver; 580 md2 = &md->u.md2; 581 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 582 ERSPAN_V2_MDSIZE); 583 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 584 info->options_len = sizeof(*md); 585 586 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 587 588 } else { 589 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 590 } 591 592 return PACKET_RCVD; 593 } 594 595 return PACKET_REJECT; 596 } 597 598 static int gre_rcv(struct sk_buff *skb) 599 { 600 struct tnl_ptk_info tpi; 601 bool csum_err = false; 602 int hdr_len; 603 604 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 605 if (hdr_len < 0) 606 goto drop; 607 608 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 609 goto drop; 610 611 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 612 tpi.proto == htons(ETH_P_ERSPAN2))) { 613 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) 614 return 0; 615 goto out; 616 } 617 618 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 619 return 0; 620 621 out: 622 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 623 drop: 624 kfree_skb(skb); 625 return 0; 626 } 627 628 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 629 { 630 return iptunnel_handle_offloads(skb, 631 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 632 } 633 634 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 635 struct net_device *dev, 636 struct flowi6 *fl6, __u8 *dsfield, 637 int *encap_limit) 638 { 639 const struct iphdr *iph = ip_hdr(skb); 640 struct ip6_tnl *t = netdev_priv(dev); 641 642 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 643 *encap_limit = t->parms.encap_limit; 644 645 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 646 647 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 648 *dsfield = ipv4_get_dsfield(iph); 649 else 650 *dsfield = ip6_tclass(t->parms.flowinfo); 651 652 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 653 fl6->flowi6_mark = skb->mark; 654 else 655 fl6->flowi6_mark = t->parms.fwmark; 656 657 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 658 } 659 660 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 661 struct net_device *dev, 662 struct flowi6 *fl6, __u8 *dsfield, 663 int *encap_limit) 664 { 665 struct ipv6hdr *ipv6h; 666 struct ip6_tnl *t = netdev_priv(dev); 667 __u16 offset; 668 669 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 670 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 671 ipv6h = ipv6_hdr(skb); 672 673 if (offset > 0) { 674 struct ipv6_tlv_tnl_enc_lim *tel; 675 676 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 677 if (tel->encap_limit == 0) { 678 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 679 ICMPV6_HDR_FIELD, offset + 2); 680 return -1; 681 } 682 *encap_limit = tel->encap_limit - 1; 683 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 684 *encap_limit = t->parms.encap_limit; 685 } 686 687 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 688 689 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 690 *dsfield = ipv6_get_dsfield(ipv6h); 691 else 692 *dsfield = ip6_tclass(t->parms.flowinfo); 693 694 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 695 fl6->flowlabel |= ip6_flowlabel(ipv6h); 696 697 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 698 fl6->flowi6_mark = skb->mark; 699 else 700 fl6->flowi6_mark = t->parms.fwmark; 701 702 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 703 704 return 0; 705 } 706 707 static int prepare_ip6gre_xmit_other(struct sk_buff *skb, 708 struct net_device *dev, 709 struct flowi6 *fl6, __u8 *dsfield, 710 int *encap_limit) 711 { 712 struct ip6_tnl *t = netdev_priv(dev); 713 714 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 715 *encap_limit = t->parms.encap_limit; 716 717 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 718 719 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 720 *dsfield = 0; 721 else 722 *dsfield = ip6_tclass(t->parms.flowinfo); 723 724 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 725 fl6->flowi6_mark = skb->mark; 726 else 727 fl6->flowi6_mark = t->parms.fwmark; 728 729 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 730 731 return 0; 732 } 733 734 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb) 735 { 736 struct ip_tunnel_info *tun_info; 737 738 tun_info = skb_tunnel_info(skb); 739 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))) 740 return ERR_PTR(-EINVAL); 741 742 return tun_info; 743 } 744 745 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 746 struct net_device *dev, __u8 dsfield, 747 struct flowi6 *fl6, int encap_limit, 748 __u32 *pmtu, __be16 proto) 749 { 750 struct ip6_tnl *tunnel = netdev_priv(dev); 751 __be16 protocol; 752 __be16 flags; 753 754 if (dev->type == ARPHRD_ETHER) 755 IPCB(skb)->flags = 0; 756 757 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 758 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 759 else 760 fl6->daddr = tunnel->parms.raddr; 761 762 /* Push GRE header. */ 763 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 764 765 if (tunnel->parms.collect_md) { 766 struct ip_tunnel_info *tun_info; 767 const struct ip_tunnel_key *key; 768 int tun_hlen; 769 770 tun_info = skb_tunnel_info_txcheck(skb); 771 if (IS_ERR(tun_info) || 772 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 773 return -EINVAL; 774 775 key = &tun_info->key; 776 memset(fl6, 0, sizeof(*fl6)); 777 fl6->flowi6_proto = IPPROTO_GRE; 778 fl6->daddr = key->u.ipv6.dst; 779 fl6->flowlabel = key->label; 780 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 781 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id); 782 783 dsfield = key->tos; 784 flags = key->tun_flags & 785 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 786 tun_hlen = gre_calc_hlen(flags); 787 788 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen)) 789 return -ENOMEM; 790 791 gre_build_header(skb, tun_hlen, 792 flags, protocol, 793 tunnel_id_to_key32(tun_info->key.tun_id), 794 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) 795 : 0); 796 797 } else { 798 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 799 return -ENOMEM; 800 801 flags = tunnel->parms.o_flags; 802 803 gre_build_header(skb, tunnel->tun_hlen, flags, 804 protocol, tunnel->parms.o_key, 805 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) 806 : 0); 807 } 808 809 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 810 NEXTHDR_GRE); 811 } 812 813 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 814 { 815 struct ip6_tnl *t = netdev_priv(dev); 816 int encap_limit = -1; 817 struct flowi6 fl6; 818 __u8 dsfield = 0; 819 __u32 mtu; 820 int err; 821 822 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 823 824 if (!t->parms.collect_md) 825 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 826 &dsfield, &encap_limit); 827 828 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 829 if (err) 830 return -1; 831 832 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 833 skb->protocol); 834 if (err != 0) { 835 /* XXX: send ICMP error even if DF is not set. */ 836 if (err == -EMSGSIZE) 837 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 838 htonl(mtu)); 839 return -1; 840 } 841 842 return 0; 843 } 844 845 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 846 { 847 struct ip6_tnl *t = netdev_priv(dev); 848 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 849 int encap_limit = -1; 850 struct flowi6 fl6; 851 __u8 dsfield = 0; 852 __u32 mtu; 853 int err; 854 855 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 856 return -1; 857 858 if (!t->parms.collect_md && 859 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 860 return -1; 861 862 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 863 return -1; 864 865 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 866 &mtu, skb->protocol); 867 if (err != 0) { 868 if (err == -EMSGSIZE) 869 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 870 return -1; 871 } 872 873 return 0; 874 } 875 876 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 877 { 878 struct ip6_tnl *t = netdev_priv(dev); 879 int encap_limit = -1; 880 struct flowi6 fl6; 881 __u8 dsfield = 0; 882 __u32 mtu; 883 int err; 884 885 if (!t->parms.collect_md && 886 prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit)) 887 return -1; 888 889 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 890 if (err) 891 return err; 892 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol); 893 894 return err; 895 } 896 897 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 898 struct net_device *dev) 899 { 900 struct ip6_tnl *t = netdev_priv(dev); 901 __be16 payload_protocol; 902 int ret; 903 904 if (!pskb_inet_may_pull(skb)) 905 goto tx_err; 906 907 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 908 goto tx_err; 909 910 payload_protocol = skb_protocol(skb, true); 911 switch (payload_protocol) { 912 case htons(ETH_P_IP): 913 ret = ip6gre_xmit_ipv4(skb, dev); 914 break; 915 case htons(ETH_P_IPV6): 916 ret = ip6gre_xmit_ipv6(skb, dev); 917 break; 918 default: 919 ret = ip6gre_xmit_other(skb, dev); 920 break; 921 } 922 923 if (ret < 0) 924 goto tx_err; 925 926 return NETDEV_TX_OK; 927 928 tx_err: 929 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 930 DEV_STATS_INC(dev, tx_errors); 931 DEV_STATS_INC(dev, tx_dropped); 932 kfree_skb(skb); 933 return NETDEV_TX_OK; 934 } 935 936 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 937 struct net_device *dev) 938 { 939 struct ip_tunnel_info *tun_info = NULL; 940 struct ip6_tnl *t = netdev_priv(dev); 941 struct dst_entry *dst = skb_dst(skb); 942 bool truncate = false; 943 int encap_limit = -1; 944 __u8 dsfield = false; 945 struct flowi6 fl6; 946 int err = -EINVAL; 947 __be16 proto; 948 __u32 mtu; 949 int nhoff; 950 951 if (!pskb_inet_may_pull(skb)) 952 goto tx_err; 953 954 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 955 goto tx_err; 956 957 if (gre_handle_offloads(skb, false)) 958 goto tx_err; 959 960 if (skb->len > dev->mtu + dev->hard_header_len) { 961 if (pskb_trim(skb, dev->mtu + dev->hard_header_len)) 962 goto tx_err; 963 truncate = true; 964 } 965 966 nhoff = skb_network_offset(skb); 967 if (skb->protocol == htons(ETH_P_IP) && 968 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 969 truncate = true; 970 971 if (skb->protocol == htons(ETH_P_IPV6)) { 972 int thoff; 973 974 if (skb_transport_header_was_set(skb)) 975 thoff = skb_transport_offset(skb); 976 else 977 thoff = nhoff + sizeof(struct ipv6hdr); 978 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff) 979 truncate = true; 980 } 981 982 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 983 goto tx_err; 984 985 t->parms.o_flags &= ~TUNNEL_KEY; 986 IPCB(skb)->flags = 0; 987 988 /* For collect_md mode, derive fl6 from the tunnel key, 989 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 990 */ 991 if (t->parms.collect_md) { 992 const struct ip_tunnel_key *key; 993 struct erspan_metadata *md; 994 __be32 tun_id; 995 996 tun_info = skb_tunnel_info_txcheck(skb); 997 if (IS_ERR(tun_info) || 998 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 999 goto tx_err; 1000 1001 key = &tun_info->key; 1002 memset(&fl6, 0, sizeof(fl6)); 1003 fl6.flowi6_proto = IPPROTO_GRE; 1004 fl6.daddr = key->u.ipv6.dst; 1005 fl6.flowlabel = key->label; 1006 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1007 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 1008 1009 dsfield = key->tos; 1010 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 1011 goto tx_err; 1012 if (tun_info->options_len < sizeof(*md)) 1013 goto tx_err; 1014 md = ip_tunnel_info_opts(tun_info); 1015 1016 tun_id = tunnel_id_to_key32(key->tun_id); 1017 if (md->version == 1) { 1018 erspan_build_header(skb, 1019 ntohl(tun_id), 1020 ntohl(md->u.index), truncate, 1021 false); 1022 proto = htons(ETH_P_ERSPAN); 1023 } else if (md->version == 2) { 1024 erspan_build_header_v2(skb, 1025 ntohl(tun_id), 1026 md->u.md2.dir, 1027 get_hwid(&md->u.md2), 1028 truncate, false); 1029 proto = htons(ETH_P_ERSPAN2); 1030 } else { 1031 goto tx_err; 1032 } 1033 } else { 1034 switch (skb->protocol) { 1035 case htons(ETH_P_IP): 1036 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1037 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1038 &dsfield, &encap_limit); 1039 break; 1040 case htons(ETH_P_IPV6): 1041 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1042 goto tx_err; 1043 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1044 &dsfield, &encap_limit)) 1045 goto tx_err; 1046 break; 1047 default: 1048 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1049 break; 1050 } 1051 1052 if (t->parms.erspan_ver == 1) { 1053 erspan_build_header(skb, ntohl(t->parms.o_key), 1054 t->parms.index, 1055 truncate, false); 1056 proto = htons(ETH_P_ERSPAN); 1057 } else if (t->parms.erspan_ver == 2) { 1058 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1059 t->parms.dir, 1060 t->parms.hwid, 1061 truncate, false); 1062 proto = htons(ETH_P_ERSPAN2); 1063 } else { 1064 goto tx_err; 1065 } 1066 1067 fl6.daddr = t->parms.raddr; 1068 } 1069 1070 /* Push GRE header. */ 1071 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno))); 1072 1073 /* TooBig packet may have updated dst->dev's mtu */ 1074 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1075 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1076 1077 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1078 NEXTHDR_GRE); 1079 if (err != 0) { 1080 /* XXX: send ICMP error even if DF is not set. */ 1081 if (err == -EMSGSIZE) { 1082 if (skb->protocol == htons(ETH_P_IP)) 1083 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1084 ICMP_FRAG_NEEDED, htonl(mtu)); 1085 else 1086 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1087 } 1088 1089 goto tx_err; 1090 } 1091 return NETDEV_TX_OK; 1092 1093 tx_err: 1094 if (!IS_ERR(tun_info)) 1095 DEV_STATS_INC(dev, tx_errors); 1096 DEV_STATS_INC(dev, tx_dropped); 1097 kfree_skb(skb); 1098 return NETDEV_TX_OK; 1099 } 1100 1101 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1102 { 1103 struct net_device *dev = t->dev; 1104 struct __ip6_tnl_parm *p = &t->parms; 1105 struct flowi6 *fl6 = &t->fl.u.ip6; 1106 1107 if (dev->type != ARPHRD_ETHER) { 1108 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1109 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1110 } 1111 1112 /* Set up flowi template */ 1113 fl6->saddr = p->laddr; 1114 fl6->daddr = p->raddr; 1115 fl6->flowi6_oif = p->link; 1116 fl6->flowlabel = 0; 1117 fl6->flowi6_proto = IPPROTO_GRE; 1118 fl6->fl6_gre_key = t->parms.o_key; 1119 1120 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1121 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1122 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1123 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1124 1125 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1126 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1127 1128 if (p->flags&IP6_TNL_F_CAP_XMIT && 1129 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1130 dev->flags |= IFF_POINTOPOINT; 1131 else 1132 dev->flags &= ~IFF_POINTOPOINT; 1133 } 1134 1135 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1136 int t_hlen) 1137 { 1138 const struct __ip6_tnl_parm *p = &t->parms; 1139 struct net_device *dev = t->dev; 1140 1141 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1142 int strict = (ipv6_addr_type(&p->raddr) & 1143 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1144 1145 struct rt6_info *rt = rt6_lookup(t->net, 1146 &p->raddr, &p->laddr, 1147 p->link, NULL, strict); 1148 1149 if (!rt) 1150 return; 1151 1152 if (rt->dst.dev) { 1153 unsigned short dst_len = rt->dst.dev->hard_header_len + 1154 t_hlen; 1155 1156 if (t->dev->header_ops) 1157 dev->hard_header_len = dst_len; 1158 else 1159 dev->needed_headroom = dst_len; 1160 1161 if (set_mtu) { 1162 int mtu = rt->dst.dev->mtu - t_hlen; 1163 1164 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1165 mtu -= 8; 1166 if (dev->type == ARPHRD_ETHER) 1167 mtu -= ETH_HLEN; 1168 1169 if (mtu < IPV6_MIN_MTU) 1170 mtu = IPV6_MIN_MTU; 1171 WRITE_ONCE(dev->mtu, mtu); 1172 } 1173 } 1174 ip6_rt_put(rt); 1175 } 1176 } 1177 1178 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1179 { 1180 int t_hlen; 1181 1182 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1183 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1184 1185 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1186 1187 if (tunnel->dev->header_ops) 1188 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1189 else 1190 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1191 1192 return t_hlen; 1193 } 1194 1195 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1196 { 1197 ip6gre_tnl_link_config_common(t); 1198 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1199 } 1200 1201 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1202 const struct __ip6_tnl_parm *p) 1203 { 1204 t->parms.laddr = p->laddr; 1205 t->parms.raddr = p->raddr; 1206 t->parms.flags = p->flags; 1207 t->parms.hop_limit = p->hop_limit; 1208 t->parms.encap_limit = p->encap_limit; 1209 t->parms.flowinfo = p->flowinfo; 1210 t->parms.link = p->link; 1211 t->parms.proto = p->proto; 1212 t->parms.i_key = p->i_key; 1213 t->parms.o_key = p->o_key; 1214 t->parms.i_flags = p->i_flags; 1215 t->parms.o_flags = p->o_flags; 1216 t->parms.fwmark = p->fwmark; 1217 t->parms.erspan_ver = p->erspan_ver; 1218 t->parms.index = p->index; 1219 t->parms.dir = p->dir; 1220 t->parms.hwid = p->hwid; 1221 dst_cache_reset(&t->dst_cache); 1222 } 1223 1224 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1225 int set_mtu) 1226 { 1227 ip6gre_tnl_copy_tnl_parm(t, p); 1228 ip6gre_tnl_link_config(t, set_mtu); 1229 return 0; 1230 } 1231 1232 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1233 const struct ip6_tnl_parm2 *u) 1234 { 1235 p->laddr = u->laddr; 1236 p->raddr = u->raddr; 1237 p->flags = u->flags; 1238 p->hop_limit = u->hop_limit; 1239 p->encap_limit = u->encap_limit; 1240 p->flowinfo = u->flowinfo; 1241 p->link = u->link; 1242 p->i_key = u->i_key; 1243 p->o_key = u->o_key; 1244 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1245 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1246 memcpy(p->name, u->name, sizeof(u->name)); 1247 } 1248 1249 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1250 const struct __ip6_tnl_parm *p) 1251 { 1252 u->proto = IPPROTO_GRE; 1253 u->laddr = p->laddr; 1254 u->raddr = p->raddr; 1255 u->flags = p->flags; 1256 u->hop_limit = p->hop_limit; 1257 u->encap_limit = p->encap_limit; 1258 u->flowinfo = p->flowinfo; 1259 u->link = p->link; 1260 u->i_key = p->i_key; 1261 u->o_key = p->o_key; 1262 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1263 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1264 memcpy(u->name, p->name, sizeof(u->name)); 1265 } 1266 1267 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1268 struct ifreq *ifr, void __user *data, 1269 int cmd) 1270 { 1271 int err = 0; 1272 struct ip6_tnl_parm2 p; 1273 struct __ip6_tnl_parm p1; 1274 struct ip6_tnl *t = netdev_priv(dev); 1275 struct net *net = t->net; 1276 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1277 1278 memset(&p1, 0, sizeof(p1)); 1279 1280 switch (cmd) { 1281 case SIOCGETTUNNEL: 1282 if (dev == ign->fb_tunnel_dev) { 1283 if (copy_from_user(&p, data, sizeof(p))) { 1284 err = -EFAULT; 1285 break; 1286 } 1287 ip6gre_tnl_parm_from_user(&p1, &p); 1288 t = ip6gre_tunnel_locate(net, &p1, 0); 1289 if (!t) 1290 t = netdev_priv(dev); 1291 } 1292 memset(&p, 0, sizeof(p)); 1293 ip6gre_tnl_parm_to_user(&p, &t->parms); 1294 if (copy_to_user(data, &p, sizeof(p))) 1295 err = -EFAULT; 1296 break; 1297 1298 case SIOCADDTUNNEL: 1299 case SIOCCHGTUNNEL: 1300 err = -EPERM; 1301 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1302 goto done; 1303 1304 err = -EFAULT; 1305 if (copy_from_user(&p, data, sizeof(p))) 1306 goto done; 1307 1308 err = -EINVAL; 1309 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1310 goto done; 1311 1312 if (!(p.i_flags&GRE_KEY)) 1313 p.i_key = 0; 1314 if (!(p.o_flags&GRE_KEY)) 1315 p.o_key = 0; 1316 1317 ip6gre_tnl_parm_from_user(&p1, &p); 1318 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1319 1320 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1321 if (t) { 1322 if (t->dev != dev) { 1323 err = -EEXIST; 1324 break; 1325 } 1326 } else { 1327 t = netdev_priv(dev); 1328 1329 ip6gre_tunnel_unlink(ign, t); 1330 synchronize_net(); 1331 ip6gre_tnl_change(t, &p1, 1); 1332 ip6gre_tunnel_link(ign, t); 1333 netdev_state_change(dev); 1334 } 1335 } 1336 1337 if (t) { 1338 err = 0; 1339 1340 memset(&p, 0, sizeof(p)); 1341 ip6gre_tnl_parm_to_user(&p, &t->parms); 1342 if (copy_to_user(data, &p, sizeof(p))) 1343 err = -EFAULT; 1344 } else 1345 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1346 break; 1347 1348 case SIOCDELTUNNEL: 1349 err = -EPERM; 1350 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1351 goto done; 1352 1353 if (dev == ign->fb_tunnel_dev) { 1354 err = -EFAULT; 1355 if (copy_from_user(&p, data, sizeof(p))) 1356 goto done; 1357 err = -ENOENT; 1358 ip6gre_tnl_parm_from_user(&p1, &p); 1359 t = ip6gre_tunnel_locate(net, &p1, 0); 1360 if (!t) 1361 goto done; 1362 err = -EPERM; 1363 if (t == netdev_priv(ign->fb_tunnel_dev)) 1364 goto done; 1365 dev = t->dev; 1366 } 1367 unregister_netdevice(dev); 1368 err = 0; 1369 break; 1370 1371 default: 1372 err = -EINVAL; 1373 } 1374 1375 done: 1376 return err; 1377 } 1378 1379 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1380 unsigned short type, const void *daddr, 1381 const void *saddr, unsigned int len) 1382 { 1383 struct ip6_tnl *t = netdev_priv(dev); 1384 struct ipv6hdr *ipv6h; 1385 __be16 *p; 1386 1387 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1388 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1389 t->fl.u.ip6.flowlabel, 1390 true, &t->fl.u.ip6)); 1391 ipv6h->hop_limit = t->parms.hop_limit; 1392 ipv6h->nexthdr = NEXTHDR_GRE; 1393 ipv6h->saddr = t->parms.laddr; 1394 ipv6h->daddr = t->parms.raddr; 1395 1396 p = (__be16 *)(ipv6h + 1); 1397 p[0] = t->parms.o_flags; 1398 p[1] = htons(type); 1399 1400 /* 1401 * Set the source hardware address. 1402 */ 1403 1404 if (saddr) 1405 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1406 if (daddr) 1407 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1408 if (!ipv6_addr_any(&ipv6h->daddr)) 1409 return t->hlen; 1410 1411 return -t->hlen; 1412 } 1413 1414 static const struct header_ops ip6gre_header_ops = { 1415 .create = ip6gre_header, 1416 }; 1417 1418 static const struct net_device_ops ip6gre_netdev_ops = { 1419 .ndo_init = ip6gre_tunnel_init, 1420 .ndo_uninit = ip6gre_tunnel_uninit, 1421 .ndo_start_xmit = ip6gre_tunnel_xmit, 1422 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1423 .ndo_change_mtu = ip6_tnl_change_mtu, 1424 .ndo_get_stats64 = dev_get_tstats64, 1425 .ndo_get_iflink = ip6_tnl_get_iflink, 1426 }; 1427 1428 static void ip6gre_dev_free(struct net_device *dev) 1429 { 1430 struct ip6_tnl *t = netdev_priv(dev); 1431 1432 gro_cells_destroy(&t->gro_cells); 1433 dst_cache_destroy(&t->dst_cache); 1434 free_percpu(dev->tstats); 1435 } 1436 1437 static void ip6gre_tunnel_setup(struct net_device *dev) 1438 { 1439 dev->netdev_ops = &ip6gre_netdev_ops; 1440 dev->needs_free_netdev = true; 1441 dev->priv_destructor = ip6gre_dev_free; 1442 1443 dev->type = ARPHRD_IP6GRE; 1444 1445 dev->flags |= IFF_NOARP; 1446 dev->addr_len = sizeof(struct in6_addr); 1447 netif_keep_dst(dev); 1448 /* This perm addr will be used as interface identifier by IPv6 */ 1449 dev->addr_assign_type = NET_ADDR_RANDOM; 1450 eth_random_addr(dev->perm_addr); 1451 } 1452 1453 #define GRE6_FEATURES (NETIF_F_SG | \ 1454 NETIF_F_FRAGLIST | \ 1455 NETIF_F_HIGHDMA | \ 1456 NETIF_F_HW_CSUM) 1457 1458 static void ip6gre_tnl_init_features(struct net_device *dev) 1459 { 1460 struct ip6_tnl *nt = netdev_priv(dev); 1461 __be16 flags; 1462 1463 dev->features |= GRE6_FEATURES | NETIF_F_LLTX; 1464 dev->hw_features |= GRE6_FEATURES; 1465 1466 flags = nt->parms.o_flags; 1467 1468 /* TCP offload with GRE SEQ is not supported, nor can we support 2 1469 * levels of outer headers requiring an update. 1470 */ 1471 if (flags & TUNNEL_SEQ) 1472 return; 1473 if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE) 1474 return; 1475 1476 dev->features |= NETIF_F_GSO_SOFTWARE; 1477 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1478 } 1479 1480 static int ip6gre_tunnel_init_common(struct net_device *dev) 1481 { 1482 struct ip6_tnl *tunnel; 1483 int ret; 1484 int t_hlen; 1485 1486 tunnel = netdev_priv(dev); 1487 1488 tunnel->dev = dev; 1489 tunnel->net = dev_net(dev); 1490 strcpy(tunnel->parms.name, dev->name); 1491 1492 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1493 if (!dev->tstats) 1494 return -ENOMEM; 1495 1496 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1497 if (ret) 1498 goto cleanup_alloc_pcpu_stats; 1499 1500 ret = gro_cells_init(&tunnel->gro_cells, dev); 1501 if (ret) 1502 goto cleanup_dst_cache_init; 1503 1504 t_hlen = ip6gre_calc_hlen(tunnel); 1505 dev->mtu = ETH_DATA_LEN - t_hlen; 1506 if (dev->type == ARPHRD_ETHER) 1507 dev->mtu -= ETH_HLEN; 1508 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1509 dev->mtu -= 8; 1510 1511 if (tunnel->parms.collect_md) { 1512 netif_keep_dst(dev); 1513 } 1514 ip6gre_tnl_init_features(dev); 1515 1516 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1517 netdev_lockdep_set_classes(dev); 1518 return 0; 1519 1520 cleanup_dst_cache_init: 1521 dst_cache_destroy(&tunnel->dst_cache); 1522 cleanup_alloc_pcpu_stats: 1523 free_percpu(dev->tstats); 1524 dev->tstats = NULL; 1525 return ret; 1526 } 1527 1528 static int ip6gre_tunnel_init(struct net_device *dev) 1529 { 1530 struct ip6_tnl *tunnel; 1531 int ret; 1532 1533 ret = ip6gre_tunnel_init_common(dev); 1534 if (ret) 1535 return ret; 1536 1537 tunnel = netdev_priv(dev); 1538 1539 if (tunnel->parms.collect_md) 1540 return 0; 1541 1542 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1543 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1544 1545 if (ipv6_addr_any(&tunnel->parms.raddr)) 1546 dev->header_ops = &ip6gre_header_ops; 1547 1548 return 0; 1549 } 1550 1551 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1552 { 1553 struct ip6_tnl *tunnel = netdev_priv(dev); 1554 1555 tunnel->dev = dev; 1556 tunnel->net = dev_net(dev); 1557 strcpy(tunnel->parms.name, dev->name); 1558 1559 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1560 } 1561 1562 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1563 .handler = gre_rcv, 1564 .err_handler = ip6gre_err, 1565 .flags = INET6_PROTO_FINAL, 1566 }; 1567 1568 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1569 { 1570 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1571 struct net_device *dev, *aux; 1572 int prio; 1573 1574 for_each_netdev_safe(net, dev, aux) 1575 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1576 dev->rtnl_link_ops == &ip6gre_tap_ops || 1577 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1578 unregister_netdevice_queue(dev, head); 1579 1580 for (prio = 0; prio < 4; prio++) { 1581 int h; 1582 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1583 struct ip6_tnl *t; 1584 1585 t = rtnl_dereference(ign->tunnels[prio][h]); 1586 1587 while (t) { 1588 /* If dev is in the same netns, it has already 1589 * been added to the list by the previous loop. 1590 */ 1591 if (!net_eq(dev_net(t->dev), net)) 1592 unregister_netdevice_queue(t->dev, 1593 head); 1594 t = rtnl_dereference(t->next); 1595 } 1596 } 1597 } 1598 } 1599 1600 static int __net_init ip6gre_init_net(struct net *net) 1601 { 1602 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1603 struct net_device *ndev; 1604 int err; 1605 1606 if (!net_has_fallback_tunnels(net)) 1607 return 0; 1608 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1609 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1610 if (!ndev) { 1611 err = -ENOMEM; 1612 goto err_alloc_dev; 1613 } 1614 ign->fb_tunnel_dev = ndev; 1615 dev_net_set(ign->fb_tunnel_dev, net); 1616 /* FB netdevice is special: we have one, and only one per netns. 1617 * Allowing to move it to another netns is clearly unsafe. 1618 */ 1619 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1620 1621 1622 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1623 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1624 1625 err = register_netdev(ign->fb_tunnel_dev); 1626 if (err) 1627 goto err_reg_dev; 1628 1629 rcu_assign_pointer(ign->tunnels_wc[0], 1630 netdev_priv(ign->fb_tunnel_dev)); 1631 return 0; 1632 1633 err_reg_dev: 1634 free_netdev(ndev); 1635 err_alloc_dev: 1636 return err; 1637 } 1638 1639 static void __net_exit ip6gre_exit_batch_rtnl(struct list_head *net_list, 1640 struct list_head *dev_to_kill) 1641 { 1642 struct net *net; 1643 1644 ASSERT_RTNL(); 1645 list_for_each_entry(net, net_list, exit_list) 1646 ip6gre_destroy_tunnels(net, dev_to_kill); 1647 } 1648 1649 static struct pernet_operations ip6gre_net_ops = { 1650 .init = ip6gre_init_net, 1651 .exit_batch_rtnl = ip6gre_exit_batch_rtnl, 1652 .id = &ip6gre_net_id, 1653 .size = sizeof(struct ip6gre_net), 1654 }; 1655 1656 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1657 struct netlink_ext_ack *extack) 1658 { 1659 __be16 flags; 1660 1661 if (!data) 1662 return 0; 1663 1664 flags = 0; 1665 if (data[IFLA_GRE_IFLAGS]) 1666 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1667 if (data[IFLA_GRE_OFLAGS]) 1668 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1669 if (flags & (GRE_VERSION|GRE_ROUTING)) 1670 return -EINVAL; 1671 1672 return 0; 1673 } 1674 1675 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1676 struct netlink_ext_ack *extack) 1677 { 1678 struct in6_addr daddr; 1679 1680 if (tb[IFLA_ADDRESS]) { 1681 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1682 return -EINVAL; 1683 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1684 return -EADDRNOTAVAIL; 1685 } 1686 1687 if (!data) 1688 goto out; 1689 1690 if (data[IFLA_GRE_REMOTE]) { 1691 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1692 if (ipv6_addr_any(&daddr)) 1693 return -EINVAL; 1694 } 1695 1696 out: 1697 return ip6gre_tunnel_validate(tb, data, extack); 1698 } 1699 1700 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1701 struct netlink_ext_ack *extack) 1702 { 1703 __be16 flags = 0; 1704 int ret, ver = 0; 1705 1706 if (!data) 1707 return 0; 1708 1709 ret = ip6gre_tap_validate(tb, data, extack); 1710 if (ret) 1711 return ret; 1712 1713 /* ERSPAN should only have GRE sequence and key flag */ 1714 if (data[IFLA_GRE_OFLAGS]) 1715 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1716 if (data[IFLA_GRE_IFLAGS]) 1717 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1718 if (!data[IFLA_GRE_COLLECT_METADATA] && 1719 flags != (GRE_SEQ | GRE_KEY)) 1720 return -EINVAL; 1721 1722 /* ERSPAN Session ID only has 10-bit. Since we reuse 1723 * 32-bit key field as ID, check it's range. 1724 */ 1725 if (data[IFLA_GRE_IKEY] && 1726 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1727 return -EINVAL; 1728 1729 if (data[IFLA_GRE_OKEY] && 1730 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1731 return -EINVAL; 1732 1733 if (data[IFLA_GRE_ERSPAN_VER]) { 1734 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1735 if (ver != 1 && ver != 2) 1736 return -EINVAL; 1737 } 1738 1739 if (ver == 1) { 1740 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1741 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1742 1743 if (index & ~INDEX_MASK) 1744 return -EINVAL; 1745 } 1746 } else if (ver == 2) { 1747 if (data[IFLA_GRE_ERSPAN_DIR]) { 1748 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1749 1750 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1751 return -EINVAL; 1752 } 1753 1754 if (data[IFLA_GRE_ERSPAN_HWID]) { 1755 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1756 1757 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1758 return -EINVAL; 1759 } 1760 } 1761 1762 return 0; 1763 } 1764 1765 static void ip6erspan_set_version(struct nlattr *data[], 1766 struct __ip6_tnl_parm *parms) 1767 { 1768 if (!data) 1769 return; 1770 1771 parms->erspan_ver = 1; 1772 if (data[IFLA_GRE_ERSPAN_VER]) 1773 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1774 1775 if (parms->erspan_ver == 1) { 1776 if (data[IFLA_GRE_ERSPAN_INDEX]) 1777 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1778 } else if (parms->erspan_ver == 2) { 1779 if (data[IFLA_GRE_ERSPAN_DIR]) 1780 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1781 if (data[IFLA_GRE_ERSPAN_HWID]) 1782 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1783 } 1784 } 1785 1786 static void ip6gre_netlink_parms(struct nlattr *data[], 1787 struct __ip6_tnl_parm *parms) 1788 { 1789 memset(parms, 0, sizeof(*parms)); 1790 1791 if (!data) 1792 return; 1793 1794 if (data[IFLA_GRE_LINK]) 1795 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1796 1797 if (data[IFLA_GRE_IFLAGS]) 1798 parms->i_flags = gre_flags_to_tnl_flags( 1799 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1800 1801 if (data[IFLA_GRE_OFLAGS]) 1802 parms->o_flags = gre_flags_to_tnl_flags( 1803 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1804 1805 if (data[IFLA_GRE_IKEY]) 1806 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1807 1808 if (data[IFLA_GRE_OKEY]) 1809 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1810 1811 if (data[IFLA_GRE_LOCAL]) 1812 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1813 1814 if (data[IFLA_GRE_REMOTE]) 1815 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1816 1817 if (data[IFLA_GRE_TTL]) 1818 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1819 1820 if (data[IFLA_GRE_ENCAP_LIMIT]) 1821 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1822 1823 if (data[IFLA_GRE_FLOWINFO]) 1824 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1825 1826 if (data[IFLA_GRE_FLAGS]) 1827 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1828 1829 if (data[IFLA_GRE_FWMARK]) 1830 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1831 1832 if (data[IFLA_GRE_COLLECT_METADATA]) 1833 parms->collect_md = true; 1834 } 1835 1836 static int ip6gre_tap_init(struct net_device *dev) 1837 { 1838 int ret; 1839 1840 ret = ip6gre_tunnel_init_common(dev); 1841 if (ret) 1842 return ret; 1843 1844 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1845 1846 return 0; 1847 } 1848 1849 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1850 .ndo_init = ip6gre_tap_init, 1851 .ndo_uninit = ip6gre_tunnel_uninit, 1852 .ndo_start_xmit = ip6gre_tunnel_xmit, 1853 .ndo_set_mac_address = eth_mac_addr, 1854 .ndo_validate_addr = eth_validate_addr, 1855 .ndo_change_mtu = ip6_tnl_change_mtu, 1856 .ndo_get_stats64 = dev_get_tstats64, 1857 .ndo_get_iflink = ip6_tnl_get_iflink, 1858 }; 1859 1860 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1861 { 1862 int t_hlen; 1863 1864 tunnel->tun_hlen = 8; 1865 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1866 erspan_hdr_len(tunnel->parms.erspan_ver); 1867 1868 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1869 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1870 return t_hlen; 1871 } 1872 1873 static int ip6erspan_tap_init(struct net_device *dev) 1874 { 1875 struct ip6_tnl *tunnel; 1876 int t_hlen; 1877 int ret; 1878 1879 tunnel = netdev_priv(dev); 1880 1881 tunnel->dev = dev; 1882 tunnel->net = dev_net(dev); 1883 strcpy(tunnel->parms.name, dev->name); 1884 1885 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1886 if (!dev->tstats) 1887 return -ENOMEM; 1888 1889 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1890 if (ret) 1891 goto cleanup_alloc_pcpu_stats; 1892 1893 ret = gro_cells_init(&tunnel->gro_cells, dev); 1894 if (ret) 1895 goto cleanup_dst_cache_init; 1896 1897 t_hlen = ip6erspan_calc_hlen(tunnel); 1898 dev->mtu = ETH_DATA_LEN - t_hlen; 1899 if (dev->type == ARPHRD_ETHER) 1900 dev->mtu -= ETH_HLEN; 1901 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1902 dev->mtu -= 8; 1903 1904 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1905 ip6erspan_tnl_link_config(tunnel, 1); 1906 1907 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1908 netdev_lockdep_set_classes(dev); 1909 return 0; 1910 1911 cleanup_dst_cache_init: 1912 dst_cache_destroy(&tunnel->dst_cache); 1913 cleanup_alloc_pcpu_stats: 1914 free_percpu(dev->tstats); 1915 dev->tstats = NULL; 1916 return ret; 1917 } 1918 1919 static const struct net_device_ops ip6erspan_netdev_ops = { 1920 .ndo_init = ip6erspan_tap_init, 1921 .ndo_uninit = ip6erspan_tunnel_uninit, 1922 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1923 .ndo_set_mac_address = eth_mac_addr, 1924 .ndo_validate_addr = eth_validate_addr, 1925 .ndo_change_mtu = ip6_tnl_change_mtu, 1926 .ndo_get_stats64 = dev_get_tstats64, 1927 .ndo_get_iflink = ip6_tnl_get_iflink, 1928 }; 1929 1930 static void ip6gre_tap_setup(struct net_device *dev) 1931 { 1932 1933 ether_setup(dev); 1934 1935 dev->max_mtu = 0; 1936 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1937 dev->needs_free_netdev = true; 1938 dev->priv_destructor = ip6gre_dev_free; 1939 1940 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1941 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1942 netif_keep_dst(dev); 1943 } 1944 1945 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1946 struct ip_tunnel_encap *ipencap) 1947 { 1948 bool ret = false; 1949 1950 memset(ipencap, 0, sizeof(*ipencap)); 1951 1952 if (!data) 1953 return ret; 1954 1955 if (data[IFLA_GRE_ENCAP_TYPE]) { 1956 ret = true; 1957 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1958 } 1959 1960 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1961 ret = true; 1962 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1963 } 1964 1965 if (data[IFLA_GRE_ENCAP_SPORT]) { 1966 ret = true; 1967 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1968 } 1969 1970 if (data[IFLA_GRE_ENCAP_DPORT]) { 1971 ret = true; 1972 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1973 } 1974 1975 return ret; 1976 } 1977 1978 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1979 struct nlattr *tb[], struct nlattr *data[], 1980 struct netlink_ext_ack *extack) 1981 { 1982 struct ip6_tnl *nt; 1983 struct ip_tunnel_encap ipencap; 1984 int err; 1985 1986 nt = netdev_priv(dev); 1987 1988 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1989 int err = ip6_tnl_encap_setup(nt, &ipencap); 1990 1991 if (err < 0) 1992 return err; 1993 } 1994 1995 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1996 eth_hw_addr_random(dev); 1997 1998 nt->dev = dev; 1999 nt->net = dev_net(dev); 2000 2001 err = register_netdevice(dev); 2002 if (err) 2003 goto out; 2004 2005 if (tb[IFLA_MTU]) 2006 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2007 2008 out: 2009 return err; 2010 } 2011 2012 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 2013 struct nlattr *tb[], struct nlattr *data[], 2014 struct netlink_ext_ack *extack) 2015 { 2016 struct ip6_tnl *nt = netdev_priv(dev); 2017 struct net *net = dev_net(dev); 2018 struct ip6gre_net *ign; 2019 int err; 2020 2021 ip6gre_netlink_parms(data, &nt->parms); 2022 ign = net_generic(net, ip6gre_net_id); 2023 2024 if (nt->parms.collect_md) { 2025 if (rtnl_dereference(ign->collect_md_tun)) 2026 return -EEXIST; 2027 } else { 2028 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2029 return -EEXIST; 2030 } 2031 2032 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2033 if (!err) { 2034 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2035 ip6gre_tunnel_link_md(ign, nt); 2036 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2037 } 2038 return err; 2039 } 2040 2041 static struct ip6_tnl * 2042 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2043 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2044 struct netlink_ext_ack *extack) 2045 { 2046 struct ip6_tnl *t, *nt = netdev_priv(dev); 2047 struct net *net = nt->net; 2048 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2049 struct ip_tunnel_encap ipencap; 2050 2051 if (dev == ign->fb_tunnel_dev) 2052 return ERR_PTR(-EINVAL); 2053 2054 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2055 int err = ip6_tnl_encap_setup(nt, &ipencap); 2056 2057 if (err < 0) 2058 return ERR_PTR(err); 2059 } 2060 2061 ip6gre_netlink_parms(data, p_p); 2062 2063 t = ip6gre_tunnel_locate(net, p_p, 0); 2064 2065 if (t) { 2066 if (t->dev != dev) 2067 return ERR_PTR(-EEXIST); 2068 } else { 2069 t = nt; 2070 } 2071 2072 return t; 2073 } 2074 2075 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2076 struct nlattr *data[], 2077 struct netlink_ext_ack *extack) 2078 { 2079 struct ip6_tnl *t = netdev_priv(dev); 2080 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2081 struct __ip6_tnl_parm p; 2082 2083 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2084 if (IS_ERR(t)) 2085 return PTR_ERR(t); 2086 2087 ip6gre_tunnel_unlink_md(ign, t); 2088 ip6gre_tunnel_unlink(ign, t); 2089 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2090 ip6gre_tunnel_link_md(ign, t); 2091 ip6gre_tunnel_link(ign, t); 2092 return 0; 2093 } 2094 2095 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2096 { 2097 struct net *net = dev_net(dev); 2098 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2099 2100 if (dev != ign->fb_tunnel_dev) 2101 unregister_netdevice_queue(dev, head); 2102 } 2103 2104 static size_t ip6gre_get_size(const struct net_device *dev) 2105 { 2106 return 2107 /* IFLA_GRE_LINK */ 2108 nla_total_size(4) + 2109 /* IFLA_GRE_IFLAGS */ 2110 nla_total_size(2) + 2111 /* IFLA_GRE_OFLAGS */ 2112 nla_total_size(2) + 2113 /* IFLA_GRE_IKEY */ 2114 nla_total_size(4) + 2115 /* IFLA_GRE_OKEY */ 2116 nla_total_size(4) + 2117 /* IFLA_GRE_LOCAL */ 2118 nla_total_size(sizeof(struct in6_addr)) + 2119 /* IFLA_GRE_REMOTE */ 2120 nla_total_size(sizeof(struct in6_addr)) + 2121 /* IFLA_GRE_TTL */ 2122 nla_total_size(1) + 2123 /* IFLA_GRE_ENCAP_LIMIT */ 2124 nla_total_size(1) + 2125 /* IFLA_GRE_FLOWINFO */ 2126 nla_total_size(4) + 2127 /* IFLA_GRE_FLAGS */ 2128 nla_total_size(4) + 2129 /* IFLA_GRE_ENCAP_TYPE */ 2130 nla_total_size(2) + 2131 /* IFLA_GRE_ENCAP_FLAGS */ 2132 nla_total_size(2) + 2133 /* IFLA_GRE_ENCAP_SPORT */ 2134 nla_total_size(2) + 2135 /* IFLA_GRE_ENCAP_DPORT */ 2136 nla_total_size(2) + 2137 /* IFLA_GRE_COLLECT_METADATA */ 2138 nla_total_size(0) + 2139 /* IFLA_GRE_FWMARK */ 2140 nla_total_size(4) + 2141 /* IFLA_GRE_ERSPAN_INDEX */ 2142 nla_total_size(4) + 2143 0; 2144 } 2145 2146 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2147 { 2148 struct ip6_tnl *t = netdev_priv(dev); 2149 struct __ip6_tnl_parm *p = &t->parms; 2150 __be16 o_flags = p->o_flags; 2151 2152 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2153 if (!p->collect_md) 2154 o_flags |= TUNNEL_KEY; 2155 2156 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2157 goto nla_put_failure; 2158 2159 if (p->erspan_ver == 1) { 2160 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2161 goto nla_put_failure; 2162 } else { 2163 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2164 goto nla_put_failure; 2165 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2166 goto nla_put_failure; 2167 } 2168 } 2169 2170 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2171 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2172 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2173 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2174 gre_tnl_flags_to_gre_flags(o_flags)) || 2175 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2176 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2177 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2178 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2179 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2180 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2181 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2182 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2183 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2184 goto nla_put_failure; 2185 2186 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2187 t->encap.type) || 2188 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2189 t->encap.sport) || 2190 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2191 t->encap.dport) || 2192 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2193 t->encap.flags)) 2194 goto nla_put_failure; 2195 2196 if (p->collect_md) { 2197 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2198 goto nla_put_failure; 2199 } 2200 2201 return 0; 2202 2203 nla_put_failure: 2204 return -EMSGSIZE; 2205 } 2206 2207 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2208 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2209 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2210 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2211 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2212 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2213 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2214 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2215 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2216 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2217 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2218 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2219 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2220 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2221 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2222 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2223 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2224 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2225 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2226 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2227 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2228 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2229 }; 2230 2231 static void ip6erspan_tap_setup(struct net_device *dev) 2232 { 2233 ether_setup(dev); 2234 2235 dev->max_mtu = 0; 2236 dev->netdev_ops = &ip6erspan_netdev_ops; 2237 dev->needs_free_netdev = true; 2238 dev->priv_destructor = ip6gre_dev_free; 2239 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 *src_net, struct net_device *dev, 2246 struct nlattr *tb[], struct nlattr *data[], 2247 struct netlink_ext_ack *extack) 2248 { 2249 struct ip6_tnl *nt = netdev_priv(dev); 2250 struct net *net = dev_net(dev); 2251 struct ip6gre_net *ign; 2252 int err; 2253 2254 ip6gre_netlink_parms(data, &nt->parms); 2255 ip6erspan_set_version(data, &nt->parms); 2256 ign = net_generic(net, ip6gre_net_id); 2257 2258 if (nt->parms.collect_md) { 2259 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2260 return -EEXIST; 2261 } else { 2262 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2263 return -EEXIST; 2264 } 2265 2266 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2267 if (!err) { 2268 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2269 ip6erspan_tunnel_link_md(ign, nt); 2270 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2271 } 2272 return err; 2273 } 2274 2275 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2276 { 2277 ip6gre_tnl_link_config_common(t); 2278 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2279 } 2280 2281 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2282 const struct __ip6_tnl_parm *p, int set_mtu) 2283 { 2284 ip6gre_tnl_copy_tnl_parm(t, p); 2285 ip6erspan_tnl_link_config(t, set_mtu); 2286 return 0; 2287 } 2288 2289 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2290 struct nlattr *data[], 2291 struct netlink_ext_ack *extack) 2292 { 2293 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2294 struct __ip6_tnl_parm p; 2295 struct ip6_tnl *t; 2296 2297 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2298 if (IS_ERR(t)) 2299 return PTR_ERR(t); 2300 2301 ip6erspan_set_version(data, &p); 2302 ip6gre_tunnel_unlink_md(ign, t); 2303 ip6gre_tunnel_unlink(ign, t); 2304 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2305 ip6erspan_tunnel_link_md(ign, t); 2306 ip6gre_tunnel_link(ign, t); 2307 return 0; 2308 } 2309 2310 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2311 .kind = "ip6gre", 2312 .maxtype = IFLA_GRE_MAX, 2313 .policy = ip6gre_policy, 2314 .priv_size = sizeof(struct ip6_tnl), 2315 .setup = ip6gre_tunnel_setup, 2316 .validate = ip6gre_tunnel_validate, 2317 .newlink = ip6gre_newlink, 2318 .changelink = ip6gre_changelink, 2319 .dellink = ip6gre_dellink, 2320 .get_size = ip6gre_get_size, 2321 .fill_info = ip6gre_fill_info, 2322 .get_link_net = ip6_tnl_get_link_net, 2323 }; 2324 2325 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2326 .kind = "ip6gretap", 2327 .maxtype = IFLA_GRE_MAX, 2328 .policy = ip6gre_policy, 2329 .priv_size = sizeof(struct ip6_tnl), 2330 .setup = ip6gre_tap_setup, 2331 .validate = ip6gre_tap_validate, 2332 .newlink = ip6gre_newlink, 2333 .changelink = ip6gre_changelink, 2334 .get_size = ip6gre_get_size, 2335 .fill_info = ip6gre_fill_info, 2336 .get_link_net = ip6_tnl_get_link_net, 2337 }; 2338 2339 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2340 .kind = "ip6erspan", 2341 .maxtype = IFLA_GRE_MAX, 2342 .policy = ip6gre_policy, 2343 .priv_size = sizeof(struct ip6_tnl), 2344 .setup = ip6erspan_tap_setup, 2345 .validate = ip6erspan_tap_validate, 2346 .newlink = ip6erspan_newlink, 2347 .changelink = ip6erspan_changelink, 2348 .get_size = ip6gre_get_size, 2349 .fill_info = ip6gre_fill_info, 2350 .get_link_net = ip6_tnl_get_link_net, 2351 }; 2352 2353 /* 2354 * And now the modules code and kernel interface. 2355 */ 2356 2357 static int __init ip6gre_init(void) 2358 { 2359 int err; 2360 2361 pr_info("GRE over IPv6 tunneling driver\n"); 2362 2363 err = register_pernet_device(&ip6gre_net_ops); 2364 if (err < 0) 2365 return err; 2366 2367 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2368 if (err < 0) { 2369 pr_info("%s: can't add protocol\n", __func__); 2370 goto add_proto_failed; 2371 } 2372 2373 err = rtnl_link_register(&ip6gre_link_ops); 2374 if (err < 0) 2375 goto rtnl_link_failed; 2376 2377 err = rtnl_link_register(&ip6gre_tap_ops); 2378 if (err < 0) 2379 goto tap_ops_failed; 2380 2381 err = rtnl_link_register(&ip6erspan_tap_ops); 2382 if (err < 0) 2383 goto erspan_link_failed; 2384 2385 out: 2386 return err; 2387 2388 erspan_link_failed: 2389 rtnl_link_unregister(&ip6gre_tap_ops); 2390 tap_ops_failed: 2391 rtnl_link_unregister(&ip6gre_link_ops); 2392 rtnl_link_failed: 2393 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2394 add_proto_failed: 2395 unregister_pernet_device(&ip6gre_net_ops); 2396 goto out; 2397 } 2398 2399 static void __exit ip6gre_fini(void) 2400 { 2401 rtnl_link_unregister(&ip6gre_tap_ops); 2402 rtnl_link_unregister(&ip6gre_link_ops); 2403 rtnl_link_unregister(&ip6erspan_tap_ops); 2404 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2405 unregister_pernet_device(&ip6gre_net_ops); 2406 } 2407 2408 module_init(ip6gre_init); 2409 module_exit(ip6gre_fini); 2410 MODULE_LICENSE("GPL"); 2411 MODULE_AUTHOR("D. Kozlov <xeb@mail.ru>"); 2412 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2413 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2414 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2415 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2416 MODULE_ALIAS_NETDEV("ip6gre0"); 2417