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 ipv6h = ipv6_hdr(skb); 532 ershdr = (struct erspan_base_hdr *)skb->data; 533 ver = ershdr->ver; 534 535 tunnel = ip6gre_tunnel_lookup(skb->dev, 536 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 537 tpi->proto); 538 if (tunnel) { 539 int len = erspan_hdr_len(ver); 540 541 if (unlikely(!pskb_may_pull(skb, len))) 542 return PACKET_REJECT; 543 544 if (__iptunnel_pull_header(skb, len, 545 htons(ETH_P_TEB), 546 false, false) < 0) 547 return PACKET_REJECT; 548 549 if (tunnel->parms.collect_md) { 550 struct erspan_metadata *pkt_md, *md; 551 struct metadata_dst *tun_dst; 552 struct ip_tunnel_info *info; 553 unsigned char *gh; 554 __be64 tun_id; 555 __be16 flags; 556 557 tpi->flags |= TUNNEL_KEY; 558 flags = tpi->flags; 559 tun_id = key32_to_tunnel_id(tpi->key); 560 561 tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 562 sizeof(*md)); 563 if (!tun_dst) 564 return PACKET_REJECT; 565 566 /* skb can be uncloned in __iptunnel_pull_header, so 567 * old pkt_md is no longer valid and we need to reset 568 * it 569 */ 570 gh = skb_network_header(skb) + 571 skb_network_header_len(skb); 572 pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len + 573 sizeof(*ershdr)); 574 info = &tun_dst->u.tun_info; 575 md = ip_tunnel_info_opts(info); 576 md->version = ver; 577 md2 = &md->u.md2; 578 memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE : 579 ERSPAN_V2_MDSIZE); 580 info->key.tun_flags |= TUNNEL_ERSPAN_OPT; 581 info->options_len = sizeof(*md); 582 583 ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error); 584 585 } else { 586 ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error); 587 } 588 589 return PACKET_RCVD; 590 } 591 592 return PACKET_REJECT; 593 } 594 595 static int gre_rcv(struct sk_buff *skb) 596 { 597 struct tnl_ptk_info tpi; 598 bool csum_err = false; 599 int hdr_len; 600 601 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 602 if (hdr_len < 0) 603 goto drop; 604 605 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 606 goto drop; 607 608 if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) || 609 tpi.proto == htons(ETH_P_ERSPAN2))) { 610 if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD) 611 return 0; 612 goto out; 613 } 614 615 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 616 return 0; 617 618 out: 619 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 620 drop: 621 kfree_skb(skb); 622 return 0; 623 } 624 625 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 626 { 627 return iptunnel_handle_offloads(skb, 628 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 629 } 630 631 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb, 632 struct net_device *dev, 633 struct flowi6 *fl6, __u8 *dsfield, 634 int *encap_limit) 635 { 636 const struct iphdr *iph = ip_hdr(skb); 637 struct ip6_tnl *t = netdev_priv(dev); 638 639 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 640 *encap_limit = t->parms.encap_limit; 641 642 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 643 644 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 645 *dsfield = ipv4_get_dsfield(iph); 646 else 647 *dsfield = ip6_tclass(t->parms.flowinfo); 648 649 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 650 fl6->flowi6_mark = skb->mark; 651 else 652 fl6->flowi6_mark = t->parms.fwmark; 653 654 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 655 } 656 657 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb, 658 struct net_device *dev, 659 struct flowi6 *fl6, __u8 *dsfield, 660 int *encap_limit) 661 { 662 struct ipv6hdr *ipv6h; 663 struct ip6_tnl *t = netdev_priv(dev); 664 __u16 offset; 665 666 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 667 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 668 ipv6h = ipv6_hdr(skb); 669 670 if (offset > 0) { 671 struct ipv6_tlv_tnl_enc_lim *tel; 672 673 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 674 if (tel->encap_limit == 0) { 675 icmpv6_ndo_send(skb, ICMPV6_PARAMPROB, 676 ICMPV6_HDR_FIELD, offset + 2); 677 return -1; 678 } 679 *encap_limit = tel->encap_limit - 1; 680 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { 681 *encap_limit = t->parms.encap_limit; 682 } 683 684 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 685 686 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 687 *dsfield = ipv6_get_dsfield(ipv6h); 688 else 689 *dsfield = ip6_tclass(t->parms.flowinfo); 690 691 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 692 fl6->flowlabel |= ip6_flowlabel(ipv6h); 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 int prepare_ip6gre_xmit_other(struct sk_buff *skb, 705 struct net_device *dev, 706 struct flowi6 *fl6, __u8 *dsfield, 707 int *encap_limit) 708 { 709 struct ip6_tnl *t = netdev_priv(dev); 710 711 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 712 *encap_limit = t->parms.encap_limit; 713 714 memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6)); 715 716 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 717 *dsfield = 0; 718 else 719 *dsfield = ip6_tclass(t->parms.flowinfo); 720 721 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 722 fl6->flowi6_mark = skb->mark; 723 else 724 fl6->flowi6_mark = t->parms.fwmark; 725 726 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 727 728 return 0; 729 } 730 731 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb) 732 { 733 struct ip_tunnel_info *tun_info; 734 735 tun_info = skb_tunnel_info(skb); 736 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX))) 737 return ERR_PTR(-EINVAL); 738 739 return tun_info; 740 } 741 742 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 743 struct net_device *dev, __u8 dsfield, 744 struct flowi6 *fl6, int encap_limit, 745 __u32 *pmtu, __be16 proto) 746 { 747 struct ip6_tnl *tunnel = netdev_priv(dev); 748 __be16 protocol; 749 __be16 flags; 750 751 if (dev->type == ARPHRD_ETHER) 752 IPCB(skb)->flags = 0; 753 754 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 755 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 756 else 757 fl6->daddr = tunnel->parms.raddr; 758 759 /* Push GRE header. */ 760 protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto; 761 762 if (tunnel->parms.collect_md) { 763 struct ip_tunnel_info *tun_info; 764 const struct ip_tunnel_key *key; 765 int tun_hlen; 766 767 tun_info = skb_tunnel_info_txcheck(skb); 768 if (IS_ERR(tun_info) || 769 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 770 return -EINVAL; 771 772 key = &tun_info->key; 773 memset(fl6, 0, sizeof(*fl6)); 774 fl6->flowi6_proto = IPPROTO_GRE; 775 fl6->daddr = key->u.ipv6.dst; 776 fl6->flowlabel = key->label; 777 fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL); 778 fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id); 779 780 dsfield = key->tos; 781 flags = key->tun_flags & 782 (TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ); 783 tun_hlen = gre_calc_hlen(flags); 784 785 if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen)) 786 return -ENOMEM; 787 788 gre_build_header(skb, tun_hlen, 789 flags, protocol, 790 tunnel_id_to_key32(tun_info->key.tun_id), 791 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) 792 : 0); 793 794 } else { 795 if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen)) 796 return -ENOMEM; 797 798 flags = tunnel->parms.o_flags; 799 800 gre_build_header(skb, tunnel->tun_hlen, flags, 801 protocol, tunnel->parms.o_key, 802 (flags & TUNNEL_SEQ) ? htonl(atomic_fetch_inc(&tunnel->o_seqno)) 803 : 0); 804 } 805 806 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 807 NEXTHDR_GRE); 808 } 809 810 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 811 { 812 struct ip6_tnl *t = netdev_priv(dev); 813 int encap_limit = -1; 814 struct flowi6 fl6; 815 __u8 dsfield = 0; 816 __u32 mtu; 817 int err; 818 819 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 820 821 if (!t->parms.collect_md) 822 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 823 &dsfield, &encap_limit); 824 825 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 826 if (err) 827 return -1; 828 829 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 830 skb->protocol); 831 if (err != 0) { 832 /* XXX: send ICMP error even if DF is not set. */ 833 if (err == -EMSGSIZE) 834 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 835 htonl(mtu)); 836 return -1; 837 } 838 839 return 0; 840 } 841 842 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 843 { 844 struct ip6_tnl *t = netdev_priv(dev); 845 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 846 int encap_limit = -1; 847 struct flowi6 fl6; 848 __u8 dsfield = 0; 849 __u32 mtu; 850 int err; 851 852 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 853 return -1; 854 855 if (!t->parms.collect_md && 856 prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit)) 857 return -1; 858 859 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 860 return -1; 861 862 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 863 &mtu, skb->protocol); 864 if (err != 0) { 865 if (err == -EMSGSIZE) 866 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 867 return -1; 868 } 869 870 return 0; 871 } 872 873 /** 874 * ip6gre_tnl_addr_conflict - compare packet addresses to tunnel's own 875 * @t: the outgoing tunnel device 876 * @hdr: IPv6 header from the incoming packet 877 * 878 * Description: 879 * Avoid trivial tunneling loop by checking that tunnel exit-point 880 * doesn't match source of incoming packet. 881 * 882 * Return: 883 * 1 if conflict, 884 * 0 else 885 **/ 886 887 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 888 const struct ipv6hdr *hdr) 889 { 890 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 891 } 892 893 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 894 { 895 struct ip6_tnl *t = netdev_priv(dev); 896 int encap_limit = -1; 897 struct flowi6 fl6; 898 __u8 dsfield = 0; 899 __u32 mtu; 900 int err; 901 902 if (!t->parms.collect_md && 903 prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit)) 904 return -1; 905 906 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 907 if (err) 908 return err; 909 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol); 910 911 return err; 912 } 913 914 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 915 struct net_device *dev) 916 { 917 struct ip6_tnl *t = netdev_priv(dev); 918 struct net_device_stats *stats = &t->dev->stats; 919 __be16 payload_protocol; 920 int ret; 921 922 if (!pskb_inet_may_pull(skb)) 923 goto tx_err; 924 925 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 926 goto tx_err; 927 928 payload_protocol = skb_protocol(skb, true); 929 switch (payload_protocol) { 930 case htons(ETH_P_IP): 931 ret = ip6gre_xmit_ipv4(skb, dev); 932 break; 933 case htons(ETH_P_IPV6): 934 ret = ip6gre_xmit_ipv6(skb, dev); 935 break; 936 default: 937 ret = ip6gre_xmit_other(skb, dev); 938 break; 939 } 940 941 if (ret < 0) 942 goto tx_err; 943 944 return NETDEV_TX_OK; 945 946 tx_err: 947 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 948 stats->tx_errors++; 949 stats->tx_dropped++; 950 kfree_skb(skb); 951 return NETDEV_TX_OK; 952 } 953 954 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 955 struct net_device *dev) 956 { 957 struct ip_tunnel_info *tun_info = NULL; 958 struct ip6_tnl *t = netdev_priv(dev); 959 struct dst_entry *dst = skb_dst(skb); 960 struct net_device_stats *stats; 961 bool truncate = false; 962 int encap_limit = -1; 963 __u8 dsfield = false; 964 struct flowi6 fl6; 965 int err = -EINVAL; 966 __be16 proto; 967 __u32 mtu; 968 int nhoff; 969 970 if (!pskb_inet_may_pull(skb)) 971 goto tx_err; 972 973 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 974 goto tx_err; 975 976 if (gre_handle_offloads(skb, false)) 977 goto tx_err; 978 979 if (skb->len > dev->mtu + dev->hard_header_len) { 980 pskb_trim(skb, dev->mtu + dev->hard_header_len); 981 truncate = true; 982 } 983 984 nhoff = skb_network_header(skb) - skb_mac_header(skb); 985 if (skb->protocol == htons(ETH_P_IP) && 986 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 987 truncate = true; 988 989 if (skb->protocol == htons(ETH_P_IPV6)) { 990 int thoff; 991 992 if (skb_transport_header_was_set(skb)) 993 thoff = skb_transport_header(skb) - skb_mac_header(skb); 994 else 995 thoff = nhoff + sizeof(struct ipv6hdr); 996 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff) 997 truncate = true; 998 } 999 1000 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 1001 goto tx_err; 1002 1003 t->parms.o_flags &= ~TUNNEL_KEY; 1004 IPCB(skb)->flags = 0; 1005 1006 /* For collect_md mode, derive fl6 from the tunnel key, 1007 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 1008 */ 1009 if (t->parms.collect_md) { 1010 const struct ip_tunnel_key *key; 1011 struct erspan_metadata *md; 1012 __be32 tun_id; 1013 1014 tun_info = skb_tunnel_info_txcheck(skb); 1015 if (IS_ERR(tun_info) || 1016 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 1017 goto tx_err; 1018 1019 key = &tun_info->key; 1020 memset(&fl6, 0, sizeof(fl6)); 1021 fl6.flowi6_proto = IPPROTO_GRE; 1022 fl6.daddr = key->u.ipv6.dst; 1023 fl6.flowlabel = key->label; 1024 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1025 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 1026 1027 dsfield = key->tos; 1028 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 1029 goto tx_err; 1030 if (tun_info->options_len < sizeof(*md)) 1031 goto tx_err; 1032 md = ip_tunnel_info_opts(tun_info); 1033 1034 tun_id = tunnel_id_to_key32(key->tun_id); 1035 if (md->version == 1) { 1036 erspan_build_header(skb, 1037 ntohl(tun_id), 1038 ntohl(md->u.index), truncate, 1039 false); 1040 } else if (md->version == 2) { 1041 erspan_build_header_v2(skb, 1042 ntohl(tun_id), 1043 md->u.md2.dir, 1044 get_hwid(&md->u.md2), 1045 truncate, false); 1046 } else { 1047 goto tx_err; 1048 } 1049 } else { 1050 switch (skb->protocol) { 1051 case htons(ETH_P_IP): 1052 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1053 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1054 &dsfield, &encap_limit); 1055 break; 1056 case htons(ETH_P_IPV6): 1057 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1058 goto tx_err; 1059 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1060 &dsfield, &encap_limit)) 1061 goto tx_err; 1062 break; 1063 default: 1064 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1065 break; 1066 } 1067 1068 if (t->parms.erspan_ver == 1) 1069 erspan_build_header(skb, ntohl(t->parms.o_key), 1070 t->parms.index, 1071 truncate, false); 1072 else if (t->parms.erspan_ver == 2) 1073 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1074 t->parms.dir, 1075 t->parms.hwid, 1076 truncate, false); 1077 else 1078 goto tx_err; 1079 1080 fl6.daddr = t->parms.raddr; 1081 } 1082 1083 /* Push GRE header. */ 1084 proto = (t->parms.erspan_ver == 1) ? htons(ETH_P_ERSPAN) 1085 : htons(ETH_P_ERSPAN2); 1086 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno))); 1087 1088 /* TooBig packet may have updated dst->dev's mtu */ 1089 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1090 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1091 1092 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1093 NEXTHDR_GRE); 1094 if (err != 0) { 1095 /* XXX: send ICMP error even if DF is not set. */ 1096 if (err == -EMSGSIZE) { 1097 if (skb->protocol == htons(ETH_P_IP)) 1098 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1099 ICMP_FRAG_NEEDED, htonl(mtu)); 1100 else 1101 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1102 } 1103 1104 goto tx_err; 1105 } 1106 return NETDEV_TX_OK; 1107 1108 tx_err: 1109 stats = &t->dev->stats; 1110 if (!IS_ERR(tun_info)) 1111 stats->tx_errors++; 1112 stats->tx_dropped++; 1113 kfree_skb(skb); 1114 return NETDEV_TX_OK; 1115 } 1116 1117 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1118 { 1119 struct net_device *dev = t->dev; 1120 struct __ip6_tnl_parm *p = &t->parms; 1121 struct flowi6 *fl6 = &t->fl.u.ip6; 1122 1123 if (dev->type != ARPHRD_ETHER) { 1124 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1125 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1126 } 1127 1128 /* Set up flowi template */ 1129 fl6->saddr = p->laddr; 1130 fl6->daddr = p->raddr; 1131 fl6->flowi6_oif = p->link; 1132 fl6->flowlabel = 0; 1133 fl6->flowi6_proto = IPPROTO_GRE; 1134 fl6->fl6_gre_key = t->parms.o_key; 1135 1136 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1137 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1138 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1139 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1140 1141 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1142 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1143 1144 if (p->flags&IP6_TNL_F_CAP_XMIT && 1145 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1146 dev->flags |= IFF_POINTOPOINT; 1147 else 1148 dev->flags &= ~IFF_POINTOPOINT; 1149 } 1150 1151 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1152 int t_hlen) 1153 { 1154 const struct __ip6_tnl_parm *p = &t->parms; 1155 struct net_device *dev = t->dev; 1156 1157 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1158 int strict = (ipv6_addr_type(&p->raddr) & 1159 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1160 1161 struct rt6_info *rt = rt6_lookup(t->net, 1162 &p->raddr, &p->laddr, 1163 p->link, NULL, strict); 1164 1165 if (!rt) 1166 return; 1167 1168 if (rt->dst.dev) { 1169 unsigned short dst_len = rt->dst.dev->hard_header_len + 1170 t_hlen; 1171 1172 if (t->dev->header_ops) 1173 dev->hard_header_len = dst_len; 1174 else 1175 dev->needed_headroom = dst_len; 1176 1177 if (set_mtu) { 1178 dev->mtu = rt->dst.dev->mtu - t_hlen; 1179 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1180 dev->mtu -= 8; 1181 if (dev->type == ARPHRD_ETHER) 1182 dev->mtu -= ETH_HLEN; 1183 1184 if (dev->mtu < IPV6_MIN_MTU) 1185 dev->mtu = IPV6_MIN_MTU; 1186 } 1187 } 1188 ip6_rt_put(rt); 1189 } 1190 } 1191 1192 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1193 { 1194 int t_hlen; 1195 1196 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1197 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1198 1199 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1200 1201 if (tunnel->dev->header_ops) 1202 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1203 else 1204 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1205 1206 return t_hlen; 1207 } 1208 1209 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1210 { 1211 ip6gre_tnl_link_config_common(t); 1212 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1213 } 1214 1215 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1216 const struct __ip6_tnl_parm *p) 1217 { 1218 t->parms.laddr = p->laddr; 1219 t->parms.raddr = p->raddr; 1220 t->parms.flags = p->flags; 1221 t->parms.hop_limit = p->hop_limit; 1222 t->parms.encap_limit = p->encap_limit; 1223 t->parms.flowinfo = p->flowinfo; 1224 t->parms.link = p->link; 1225 t->parms.proto = p->proto; 1226 t->parms.i_key = p->i_key; 1227 t->parms.o_key = p->o_key; 1228 t->parms.i_flags = p->i_flags; 1229 t->parms.o_flags = p->o_flags; 1230 t->parms.fwmark = p->fwmark; 1231 t->parms.erspan_ver = p->erspan_ver; 1232 t->parms.index = p->index; 1233 t->parms.dir = p->dir; 1234 t->parms.hwid = p->hwid; 1235 dst_cache_reset(&t->dst_cache); 1236 } 1237 1238 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1239 int set_mtu) 1240 { 1241 ip6gre_tnl_copy_tnl_parm(t, p); 1242 ip6gre_tnl_link_config(t, set_mtu); 1243 return 0; 1244 } 1245 1246 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1247 const struct ip6_tnl_parm2 *u) 1248 { 1249 p->laddr = u->laddr; 1250 p->raddr = u->raddr; 1251 p->flags = u->flags; 1252 p->hop_limit = u->hop_limit; 1253 p->encap_limit = u->encap_limit; 1254 p->flowinfo = u->flowinfo; 1255 p->link = u->link; 1256 p->i_key = u->i_key; 1257 p->o_key = u->o_key; 1258 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1259 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1260 memcpy(p->name, u->name, sizeof(u->name)); 1261 } 1262 1263 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1264 const struct __ip6_tnl_parm *p) 1265 { 1266 u->proto = IPPROTO_GRE; 1267 u->laddr = p->laddr; 1268 u->raddr = p->raddr; 1269 u->flags = p->flags; 1270 u->hop_limit = p->hop_limit; 1271 u->encap_limit = p->encap_limit; 1272 u->flowinfo = p->flowinfo; 1273 u->link = p->link; 1274 u->i_key = p->i_key; 1275 u->o_key = p->o_key; 1276 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1277 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1278 memcpy(u->name, p->name, sizeof(u->name)); 1279 } 1280 1281 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1282 struct ifreq *ifr, void __user *data, 1283 int cmd) 1284 { 1285 int err = 0; 1286 struct ip6_tnl_parm2 p; 1287 struct __ip6_tnl_parm p1; 1288 struct ip6_tnl *t = netdev_priv(dev); 1289 struct net *net = t->net; 1290 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1291 1292 memset(&p1, 0, sizeof(p1)); 1293 1294 switch (cmd) { 1295 case SIOCGETTUNNEL: 1296 if (dev == ign->fb_tunnel_dev) { 1297 if (copy_from_user(&p, data, sizeof(p))) { 1298 err = -EFAULT; 1299 break; 1300 } 1301 ip6gre_tnl_parm_from_user(&p1, &p); 1302 t = ip6gre_tunnel_locate(net, &p1, 0); 1303 if (!t) 1304 t = netdev_priv(dev); 1305 } 1306 memset(&p, 0, sizeof(p)); 1307 ip6gre_tnl_parm_to_user(&p, &t->parms); 1308 if (copy_to_user(data, &p, sizeof(p))) 1309 err = -EFAULT; 1310 break; 1311 1312 case SIOCADDTUNNEL: 1313 case SIOCCHGTUNNEL: 1314 err = -EPERM; 1315 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1316 goto done; 1317 1318 err = -EFAULT; 1319 if (copy_from_user(&p, data, sizeof(p))) 1320 goto done; 1321 1322 err = -EINVAL; 1323 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1324 goto done; 1325 1326 if (!(p.i_flags&GRE_KEY)) 1327 p.i_key = 0; 1328 if (!(p.o_flags&GRE_KEY)) 1329 p.o_key = 0; 1330 1331 ip6gre_tnl_parm_from_user(&p1, &p); 1332 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1333 1334 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1335 if (t) { 1336 if (t->dev != dev) { 1337 err = -EEXIST; 1338 break; 1339 } 1340 } else { 1341 t = netdev_priv(dev); 1342 1343 ip6gre_tunnel_unlink(ign, t); 1344 synchronize_net(); 1345 ip6gre_tnl_change(t, &p1, 1); 1346 ip6gre_tunnel_link(ign, t); 1347 netdev_state_change(dev); 1348 } 1349 } 1350 1351 if (t) { 1352 err = 0; 1353 1354 memset(&p, 0, sizeof(p)); 1355 ip6gre_tnl_parm_to_user(&p, &t->parms); 1356 if (copy_to_user(data, &p, sizeof(p))) 1357 err = -EFAULT; 1358 } else 1359 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1360 break; 1361 1362 case SIOCDELTUNNEL: 1363 err = -EPERM; 1364 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1365 goto done; 1366 1367 if (dev == ign->fb_tunnel_dev) { 1368 err = -EFAULT; 1369 if (copy_from_user(&p, data, sizeof(p))) 1370 goto done; 1371 err = -ENOENT; 1372 ip6gre_tnl_parm_from_user(&p1, &p); 1373 t = ip6gre_tunnel_locate(net, &p1, 0); 1374 if (!t) 1375 goto done; 1376 err = -EPERM; 1377 if (t == netdev_priv(ign->fb_tunnel_dev)) 1378 goto done; 1379 dev = t->dev; 1380 } 1381 unregister_netdevice(dev); 1382 err = 0; 1383 break; 1384 1385 default: 1386 err = -EINVAL; 1387 } 1388 1389 done: 1390 return err; 1391 } 1392 1393 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1394 unsigned short type, const void *daddr, 1395 const void *saddr, unsigned int len) 1396 { 1397 struct ip6_tnl *t = netdev_priv(dev); 1398 struct ipv6hdr *ipv6h; 1399 __be16 *p; 1400 1401 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1402 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1403 t->fl.u.ip6.flowlabel, 1404 true, &t->fl.u.ip6)); 1405 ipv6h->hop_limit = t->parms.hop_limit; 1406 ipv6h->nexthdr = NEXTHDR_GRE; 1407 ipv6h->saddr = t->parms.laddr; 1408 ipv6h->daddr = t->parms.raddr; 1409 1410 p = (__be16 *)(ipv6h + 1); 1411 p[0] = t->parms.o_flags; 1412 p[1] = htons(type); 1413 1414 /* 1415 * Set the source hardware address. 1416 */ 1417 1418 if (saddr) 1419 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1420 if (daddr) 1421 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1422 if (!ipv6_addr_any(&ipv6h->daddr)) 1423 return t->hlen; 1424 1425 return -t->hlen; 1426 } 1427 1428 static const struct header_ops ip6gre_header_ops = { 1429 .create = ip6gre_header, 1430 }; 1431 1432 static const struct net_device_ops ip6gre_netdev_ops = { 1433 .ndo_init = ip6gre_tunnel_init, 1434 .ndo_uninit = ip6gre_tunnel_uninit, 1435 .ndo_start_xmit = ip6gre_tunnel_xmit, 1436 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1437 .ndo_change_mtu = ip6_tnl_change_mtu, 1438 .ndo_get_stats64 = dev_get_tstats64, 1439 .ndo_get_iflink = ip6_tnl_get_iflink, 1440 }; 1441 1442 static void ip6gre_dev_free(struct net_device *dev) 1443 { 1444 struct ip6_tnl *t = netdev_priv(dev); 1445 1446 gro_cells_destroy(&t->gro_cells); 1447 dst_cache_destroy(&t->dst_cache); 1448 free_percpu(dev->tstats); 1449 } 1450 1451 static void ip6gre_tunnel_setup(struct net_device *dev) 1452 { 1453 dev->netdev_ops = &ip6gre_netdev_ops; 1454 dev->needs_free_netdev = true; 1455 dev->priv_destructor = ip6gre_dev_free; 1456 1457 dev->type = ARPHRD_IP6GRE; 1458 1459 dev->flags |= IFF_NOARP; 1460 dev->addr_len = sizeof(struct in6_addr); 1461 netif_keep_dst(dev); 1462 /* This perm addr will be used as interface identifier by IPv6 */ 1463 dev->addr_assign_type = NET_ADDR_RANDOM; 1464 eth_random_addr(dev->perm_addr); 1465 } 1466 1467 #define GRE6_FEATURES (NETIF_F_SG | \ 1468 NETIF_F_FRAGLIST | \ 1469 NETIF_F_HIGHDMA | \ 1470 NETIF_F_HW_CSUM) 1471 1472 static void ip6gre_tnl_init_features(struct net_device *dev) 1473 { 1474 struct ip6_tnl *nt = netdev_priv(dev); 1475 __be16 flags; 1476 1477 dev->features |= GRE6_FEATURES | NETIF_F_LLTX; 1478 dev->hw_features |= GRE6_FEATURES; 1479 1480 flags = nt->parms.o_flags; 1481 1482 /* TCP offload with GRE SEQ is not supported, nor can we support 2 1483 * levels of outer headers requiring an update. 1484 */ 1485 if (flags & TUNNEL_SEQ) 1486 return; 1487 if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE) 1488 return; 1489 1490 dev->features |= NETIF_F_GSO_SOFTWARE; 1491 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1492 } 1493 1494 static int ip6gre_tunnel_init_common(struct net_device *dev) 1495 { 1496 struct ip6_tnl *tunnel; 1497 int ret; 1498 int t_hlen; 1499 1500 tunnel = netdev_priv(dev); 1501 1502 tunnel->dev = dev; 1503 tunnel->net = dev_net(dev); 1504 strcpy(tunnel->parms.name, dev->name); 1505 1506 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1507 if (!dev->tstats) 1508 return -ENOMEM; 1509 1510 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1511 if (ret) 1512 goto cleanup_alloc_pcpu_stats; 1513 1514 ret = gro_cells_init(&tunnel->gro_cells, dev); 1515 if (ret) 1516 goto cleanup_dst_cache_init; 1517 1518 t_hlen = ip6gre_calc_hlen(tunnel); 1519 dev->mtu = ETH_DATA_LEN - t_hlen; 1520 if (dev->type == ARPHRD_ETHER) 1521 dev->mtu -= ETH_HLEN; 1522 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1523 dev->mtu -= 8; 1524 1525 if (tunnel->parms.collect_md) { 1526 netif_keep_dst(dev); 1527 } 1528 ip6gre_tnl_init_features(dev); 1529 1530 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1531 return 0; 1532 1533 cleanup_dst_cache_init: 1534 dst_cache_destroy(&tunnel->dst_cache); 1535 cleanup_alloc_pcpu_stats: 1536 free_percpu(dev->tstats); 1537 dev->tstats = NULL; 1538 return ret; 1539 } 1540 1541 static int ip6gre_tunnel_init(struct net_device *dev) 1542 { 1543 struct ip6_tnl *tunnel; 1544 int ret; 1545 1546 ret = ip6gre_tunnel_init_common(dev); 1547 if (ret) 1548 return ret; 1549 1550 tunnel = netdev_priv(dev); 1551 1552 if (tunnel->parms.collect_md) 1553 return 0; 1554 1555 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1556 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1557 1558 if (ipv6_addr_any(&tunnel->parms.raddr)) 1559 dev->header_ops = &ip6gre_header_ops; 1560 1561 return 0; 1562 } 1563 1564 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1565 { 1566 struct ip6_tnl *tunnel = netdev_priv(dev); 1567 1568 tunnel->dev = dev; 1569 tunnel->net = dev_net(dev); 1570 strcpy(tunnel->parms.name, dev->name); 1571 1572 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1573 } 1574 1575 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1576 .handler = gre_rcv, 1577 .err_handler = ip6gre_err, 1578 .flags = INET6_PROTO_FINAL, 1579 }; 1580 1581 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1582 { 1583 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1584 struct net_device *dev, *aux; 1585 int prio; 1586 1587 for_each_netdev_safe(net, dev, aux) 1588 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1589 dev->rtnl_link_ops == &ip6gre_tap_ops || 1590 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1591 unregister_netdevice_queue(dev, head); 1592 1593 for (prio = 0; prio < 4; prio++) { 1594 int h; 1595 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1596 struct ip6_tnl *t; 1597 1598 t = rtnl_dereference(ign->tunnels[prio][h]); 1599 1600 while (t) { 1601 /* If dev is in the same netns, it has already 1602 * been added to the list by the previous loop. 1603 */ 1604 if (!net_eq(dev_net(t->dev), net)) 1605 unregister_netdevice_queue(t->dev, 1606 head); 1607 t = rtnl_dereference(t->next); 1608 } 1609 } 1610 } 1611 } 1612 1613 static int __net_init ip6gre_init_net(struct net *net) 1614 { 1615 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1616 struct net_device *ndev; 1617 int err; 1618 1619 if (!net_has_fallback_tunnels(net)) 1620 return 0; 1621 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1622 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1623 if (!ndev) { 1624 err = -ENOMEM; 1625 goto err_alloc_dev; 1626 } 1627 ign->fb_tunnel_dev = ndev; 1628 dev_net_set(ign->fb_tunnel_dev, net); 1629 /* FB netdevice is special: we have one, and only one per netns. 1630 * Allowing to move it to another netns is clearly unsafe. 1631 */ 1632 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1633 1634 1635 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1636 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1637 1638 err = register_netdev(ign->fb_tunnel_dev); 1639 if (err) 1640 goto err_reg_dev; 1641 1642 rcu_assign_pointer(ign->tunnels_wc[0], 1643 netdev_priv(ign->fb_tunnel_dev)); 1644 return 0; 1645 1646 err_reg_dev: 1647 free_netdev(ndev); 1648 err_alloc_dev: 1649 return err; 1650 } 1651 1652 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list) 1653 { 1654 struct net *net; 1655 LIST_HEAD(list); 1656 1657 rtnl_lock(); 1658 list_for_each_entry(net, net_list, exit_list) 1659 ip6gre_destroy_tunnels(net, &list); 1660 unregister_netdevice_many(&list); 1661 rtnl_unlock(); 1662 } 1663 1664 static struct pernet_operations ip6gre_net_ops = { 1665 .init = ip6gre_init_net, 1666 .exit_batch = ip6gre_exit_batch_net, 1667 .id = &ip6gre_net_id, 1668 .size = sizeof(struct ip6gre_net), 1669 }; 1670 1671 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1672 struct netlink_ext_ack *extack) 1673 { 1674 __be16 flags; 1675 1676 if (!data) 1677 return 0; 1678 1679 flags = 0; 1680 if (data[IFLA_GRE_IFLAGS]) 1681 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1682 if (data[IFLA_GRE_OFLAGS]) 1683 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1684 if (flags & (GRE_VERSION|GRE_ROUTING)) 1685 return -EINVAL; 1686 1687 return 0; 1688 } 1689 1690 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1691 struct netlink_ext_ack *extack) 1692 { 1693 struct in6_addr daddr; 1694 1695 if (tb[IFLA_ADDRESS]) { 1696 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1697 return -EINVAL; 1698 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1699 return -EADDRNOTAVAIL; 1700 } 1701 1702 if (!data) 1703 goto out; 1704 1705 if (data[IFLA_GRE_REMOTE]) { 1706 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1707 if (ipv6_addr_any(&daddr)) 1708 return -EINVAL; 1709 } 1710 1711 out: 1712 return ip6gre_tunnel_validate(tb, data, extack); 1713 } 1714 1715 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1716 struct netlink_ext_ack *extack) 1717 { 1718 __be16 flags = 0; 1719 int ret, ver = 0; 1720 1721 if (!data) 1722 return 0; 1723 1724 ret = ip6gre_tap_validate(tb, data, extack); 1725 if (ret) 1726 return ret; 1727 1728 /* ERSPAN should only have GRE sequence and key flag */ 1729 if (data[IFLA_GRE_OFLAGS]) 1730 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1731 if (data[IFLA_GRE_IFLAGS]) 1732 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1733 if (!data[IFLA_GRE_COLLECT_METADATA] && 1734 flags != (GRE_SEQ | GRE_KEY)) 1735 return -EINVAL; 1736 1737 /* ERSPAN Session ID only has 10-bit. Since we reuse 1738 * 32-bit key field as ID, check it's range. 1739 */ 1740 if (data[IFLA_GRE_IKEY] && 1741 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1742 return -EINVAL; 1743 1744 if (data[IFLA_GRE_OKEY] && 1745 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1746 return -EINVAL; 1747 1748 if (data[IFLA_GRE_ERSPAN_VER]) { 1749 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1750 if (ver != 1 && ver != 2) 1751 return -EINVAL; 1752 } 1753 1754 if (ver == 1) { 1755 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1756 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1757 1758 if (index & ~INDEX_MASK) 1759 return -EINVAL; 1760 } 1761 } else if (ver == 2) { 1762 if (data[IFLA_GRE_ERSPAN_DIR]) { 1763 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1764 1765 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1766 return -EINVAL; 1767 } 1768 1769 if (data[IFLA_GRE_ERSPAN_HWID]) { 1770 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1771 1772 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1773 return -EINVAL; 1774 } 1775 } 1776 1777 return 0; 1778 } 1779 1780 static void ip6erspan_set_version(struct nlattr *data[], 1781 struct __ip6_tnl_parm *parms) 1782 { 1783 if (!data) 1784 return; 1785 1786 parms->erspan_ver = 1; 1787 if (data[IFLA_GRE_ERSPAN_VER]) 1788 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1789 1790 if (parms->erspan_ver == 1) { 1791 if (data[IFLA_GRE_ERSPAN_INDEX]) 1792 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1793 } else if (parms->erspan_ver == 2) { 1794 if (data[IFLA_GRE_ERSPAN_DIR]) 1795 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1796 if (data[IFLA_GRE_ERSPAN_HWID]) 1797 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1798 } 1799 } 1800 1801 static void ip6gre_netlink_parms(struct nlattr *data[], 1802 struct __ip6_tnl_parm *parms) 1803 { 1804 memset(parms, 0, sizeof(*parms)); 1805 1806 if (!data) 1807 return; 1808 1809 if (data[IFLA_GRE_LINK]) 1810 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1811 1812 if (data[IFLA_GRE_IFLAGS]) 1813 parms->i_flags = gre_flags_to_tnl_flags( 1814 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1815 1816 if (data[IFLA_GRE_OFLAGS]) 1817 parms->o_flags = gre_flags_to_tnl_flags( 1818 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1819 1820 if (data[IFLA_GRE_IKEY]) 1821 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1822 1823 if (data[IFLA_GRE_OKEY]) 1824 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1825 1826 if (data[IFLA_GRE_LOCAL]) 1827 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1828 1829 if (data[IFLA_GRE_REMOTE]) 1830 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1831 1832 if (data[IFLA_GRE_TTL]) 1833 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1834 1835 if (data[IFLA_GRE_ENCAP_LIMIT]) 1836 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1837 1838 if (data[IFLA_GRE_FLOWINFO]) 1839 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1840 1841 if (data[IFLA_GRE_FLAGS]) 1842 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1843 1844 if (data[IFLA_GRE_FWMARK]) 1845 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1846 1847 if (data[IFLA_GRE_COLLECT_METADATA]) 1848 parms->collect_md = true; 1849 } 1850 1851 static int ip6gre_tap_init(struct net_device *dev) 1852 { 1853 int ret; 1854 1855 ret = ip6gre_tunnel_init_common(dev); 1856 if (ret) 1857 return ret; 1858 1859 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1860 1861 return 0; 1862 } 1863 1864 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1865 .ndo_init = ip6gre_tap_init, 1866 .ndo_uninit = ip6gre_tunnel_uninit, 1867 .ndo_start_xmit = ip6gre_tunnel_xmit, 1868 .ndo_set_mac_address = eth_mac_addr, 1869 .ndo_validate_addr = eth_validate_addr, 1870 .ndo_change_mtu = ip6_tnl_change_mtu, 1871 .ndo_get_stats64 = dev_get_tstats64, 1872 .ndo_get_iflink = ip6_tnl_get_iflink, 1873 }; 1874 1875 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1876 { 1877 int t_hlen; 1878 1879 tunnel->tun_hlen = 8; 1880 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1881 erspan_hdr_len(tunnel->parms.erspan_ver); 1882 1883 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1884 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1885 return t_hlen; 1886 } 1887 1888 static int ip6erspan_tap_init(struct net_device *dev) 1889 { 1890 struct ip6_tnl *tunnel; 1891 int t_hlen; 1892 int ret; 1893 1894 tunnel = netdev_priv(dev); 1895 1896 tunnel->dev = dev; 1897 tunnel->net = dev_net(dev); 1898 strcpy(tunnel->parms.name, dev->name); 1899 1900 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1901 if (!dev->tstats) 1902 return -ENOMEM; 1903 1904 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1905 if (ret) 1906 goto cleanup_alloc_pcpu_stats; 1907 1908 ret = gro_cells_init(&tunnel->gro_cells, dev); 1909 if (ret) 1910 goto cleanup_dst_cache_init; 1911 1912 t_hlen = ip6erspan_calc_hlen(tunnel); 1913 dev->mtu = ETH_DATA_LEN - t_hlen; 1914 if (dev->type == ARPHRD_ETHER) 1915 dev->mtu -= ETH_HLEN; 1916 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1917 dev->mtu -= 8; 1918 1919 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1920 ip6erspan_tnl_link_config(tunnel, 1); 1921 1922 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1923 return 0; 1924 1925 cleanup_dst_cache_init: 1926 dst_cache_destroy(&tunnel->dst_cache); 1927 cleanup_alloc_pcpu_stats: 1928 free_percpu(dev->tstats); 1929 dev->tstats = NULL; 1930 return ret; 1931 } 1932 1933 static const struct net_device_ops ip6erspan_netdev_ops = { 1934 .ndo_init = ip6erspan_tap_init, 1935 .ndo_uninit = ip6erspan_tunnel_uninit, 1936 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1937 .ndo_set_mac_address = eth_mac_addr, 1938 .ndo_validate_addr = eth_validate_addr, 1939 .ndo_change_mtu = ip6_tnl_change_mtu, 1940 .ndo_get_stats64 = dev_get_tstats64, 1941 .ndo_get_iflink = ip6_tnl_get_iflink, 1942 }; 1943 1944 static void ip6gre_tap_setup(struct net_device *dev) 1945 { 1946 1947 ether_setup(dev); 1948 1949 dev->max_mtu = 0; 1950 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1951 dev->needs_free_netdev = true; 1952 dev->priv_destructor = ip6gre_dev_free; 1953 1954 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1955 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1956 netif_keep_dst(dev); 1957 } 1958 1959 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1960 struct ip_tunnel_encap *ipencap) 1961 { 1962 bool ret = false; 1963 1964 memset(ipencap, 0, sizeof(*ipencap)); 1965 1966 if (!data) 1967 return ret; 1968 1969 if (data[IFLA_GRE_ENCAP_TYPE]) { 1970 ret = true; 1971 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1972 } 1973 1974 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1975 ret = true; 1976 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1977 } 1978 1979 if (data[IFLA_GRE_ENCAP_SPORT]) { 1980 ret = true; 1981 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1982 } 1983 1984 if (data[IFLA_GRE_ENCAP_DPORT]) { 1985 ret = true; 1986 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1987 } 1988 1989 return ret; 1990 } 1991 1992 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1993 struct nlattr *tb[], struct nlattr *data[], 1994 struct netlink_ext_ack *extack) 1995 { 1996 struct ip6_tnl *nt; 1997 struct ip_tunnel_encap ipencap; 1998 int err; 1999 2000 nt = netdev_priv(dev); 2001 2002 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2003 int err = ip6_tnl_encap_setup(nt, &ipencap); 2004 2005 if (err < 0) 2006 return err; 2007 } 2008 2009 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 2010 eth_hw_addr_random(dev); 2011 2012 nt->dev = dev; 2013 nt->net = dev_net(dev); 2014 2015 err = register_netdevice(dev); 2016 if (err) 2017 goto out; 2018 2019 if (tb[IFLA_MTU]) 2020 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2021 2022 out: 2023 return err; 2024 } 2025 2026 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 2027 struct nlattr *tb[], struct nlattr *data[], 2028 struct netlink_ext_ack *extack) 2029 { 2030 struct ip6_tnl *nt = netdev_priv(dev); 2031 struct net *net = dev_net(dev); 2032 struct ip6gre_net *ign; 2033 int err; 2034 2035 ip6gre_netlink_parms(data, &nt->parms); 2036 ign = net_generic(net, ip6gre_net_id); 2037 2038 if (nt->parms.collect_md) { 2039 if (rtnl_dereference(ign->collect_md_tun)) 2040 return -EEXIST; 2041 } else { 2042 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2043 return -EEXIST; 2044 } 2045 2046 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2047 if (!err) { 2048 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2049 ip6gre_tunnel_link_md(ign, nt); 2050 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2051 } 2052 return err; 2053 } 2054 2055 static struct ip6_tnl * 2056 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2057 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2058 struct netlink_ext_ack *extack) 2059 { 2060 struct ip6_tnl *t, *nt = netdev_priv(dev); 2061 struct net *net = nt->net; 2062 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2063 struct ip_tunnel_encap ipencap; 2064 2065 if (dev == ign->fb_tunnel_dev) 2066 return ERR_PTR(-EINVAL); 2067 2068 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2069 int err = ip6_tnl_encap_setup(nt, &ipencap); 2070 2071 if (err < 0) 2072 return ERR_PTR(err); 2073 } 2074 2075 ip6gre_netlink_parms(data, p_p); 2076 2077 t = ip6gre_tunnel_locate(net, p_p, 0); 2078 2079 if (t) { 2080 if (t->dev != dev) 2081 return ERR_PTR(-EEXIST); 2082 } else { 2083 t = nt; 2084 } 2085 2086 return t; 2087 } 2088 2089 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2090 struct nlattr *data[], 2091 struct netlink_ext_ack *extack) 2092 { 2093 struct ip6_tnl *t = netdev_priv(dev); 2094 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2095 struct __ip6_tnl_parm p; 2096 2097 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2098 if (IS_ERR(t)) 2099 return PTR_ERR(t); 2100 2101 ip6gre_tunnel_unlink_md(ign, t); 2102 ip6gre_tunnel_unlink(ign, t); 2103 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2104 ip6gre_tunnel_link_md(ign, t); 2105 ip6gre_tunnel_link(ign, t); 2106 return 0; 2107 } 2108 2109 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2110 { 2111 struct net *net = dev_net(dev); 2112 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2113 2114 if (dev != ign->fb_tunnel_dev) 2115 unregister_netdevice_queue(dev, head); 2116 } 2117 2118 static size_t ip6gre_get_size(const struct net_device *dev) 2119 { 2120 return 2121 /* IFLA_GRE_LINK */ 2122 nla_total_size(4) + 2123 /* IFLA_GRE_IFLAGS */ 2124 nla_total_size(2) + 2125 /* IFLA_GRE_OFLAGS */ 2126 nla_total_size(2) + 2127 /* IFLA_GRE_IKEY */ 2128 nla_total_size(4) + 2129 /* IFLA_GRE_OKEY */ 2130 nla_total_size(4) + 2131 /* IFLA_GRE_LOCAL */ 2132 nla_total_size(sizeof(struct in6_addr)) + 2133 /* IFLA_GRE_REMOTE */ 2134 nla_total_size(sizeof(struct in6_addr)) + 2135 /* IFLA_GRE_TTL */ 2136 nla_total_size(1) + 2137 /* IFLA_GRE_ENCAP_LIMIT */ 2138 nla_total_size(1) + 2139 /* IFLA_GRE_FLOWINFO */ 2140 nla_total_size(4) + 2141 /* IFLA_GRE_FLAGS */ 2142 nla_total_size(4) + 2143 /* IFLA_GRE_ENCAP_TYPE */ 2144 nla_total_size(2) + 2145 /* IFLA_GRE_ENCAP_FLAGS */ 2146 nla_total_size(2) + 2147 /* IFLA_GRE_ENCAP_SPORT */ 2148 nla_total_size(2) + 2149 /* IFLA_GRE_ENCAP_DPORT */ 2150 nla_total_size(2) + 2151 /* IFLA_GRE_COLLECT_METADATA */ 2152 nla_total_size(0) + 2153 /* IFLA_GRE_FWMARK */ 2154 nla_total_size(4) + 2155 /* IFLA_GRE_ERSPAN_INDEX */ 2156 nla_total_size(4) + 2157 0; 2158 } 2159 2160 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2161 { 2162 struct ip6_tnl *t = netdev_priv(dev); 2163 struct __ip6_tnl_parm *p = &t->parms; 2164 __be16 o_flags = p->o_flags; 2165 2166 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2167 if (!p->collect_md) 2168 o_flags |= TUNNEL_KEY; 2169 2170 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2171 goto nla_put_failure; 2172 2173 if (p->erspan_ver == 1) { 2174 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2175 goto nla_put_failure; 2176 } else { 2177 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2178 goto nla_put_failure; 2179 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2180 goto nla_put_failure; 2181 } 2182 } 2183 2184 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2185 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2186 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2187 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2188 gre_tnl_flags_to_gre_flags(o_flags)) || 2189 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2190 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2191 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2192 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2193 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2194 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2195 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2196 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2197 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2198 goto nla_put_failure; 2199 2200 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2201 t->encap.type) || 2202 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2203 t->encap.sport) || 2204 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2205 t->encap.dport) || 2206 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2207 t->encap.flags)) 2208 goto nla_put_failure; 2209 2210 if (p->collect_md) { 2211 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2212 goto nla_put_failure; 2213 } 2214 2215 return 0; 2216 2217 nla_put_failure: 2218 return -EMSGSIZE; 2219 } 2220 2221 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2222 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2223 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2224 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2225 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2226 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2227 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2228 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2229 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2230 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2231 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2232 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2233 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2234 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2235 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2236 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2237 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2238 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2239 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2240 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2241 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2242 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2243 }; 2244 2245 static void ip6erspan_tap_setup(struct net_device *dev) 2246 { 2247 ether_setup(dev); 2248 2249 dev->max_mtu = 0; 2250 dev->netdev_ops = &ip6erspan_netdev_ops; 2251 dev->needs_free_netdev = true; 2252 dev->priv_destructor = ip6gre_dev_free; 2253 2254 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2255 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2256 netif_keep_dst(dev); 2257 } 2258 2259 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2260 struct nlattr *tb[], struct nlattr *data[], 2261 struct netlink_ext_ack *extack) 2262 { 2263 struct ip6_tnl *nt = netdev_priv(dev); 2264 struct net *net = dev_net(dev); 2265 struct ip6gre_net *ign; 2266 int err; 2267 2268 ip6gre_netlink_parms(data, &nt->parms); 2269 ip6erspan_set_version(data, &nt->parms); 2270 ign = net_generic(net, ip6gre_net_id); 2271 2272 if (nt->parms.collect_md) { 2273 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2274 return -EEXIST; 2275 } else { 2276 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2277 return -EEXIST; 2278 } 2279 2280 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2281 if (!err) { 2282 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2283 ip6erspan_tunnel_link_md(ign, nt); 2284 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2285 } 2286 return err; 2287 } 2288 2289 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2290 { 2291 ip6gre_tnl_link_config_common(t); 2292 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2293 } 2294 2295 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2296 const struct __ip6_tnl_parm *p, int set_mtu) 2297 { 2298 ip6gre_tnl_copy_tnl_parm(t, p); 2299 ip6erspan_tnl_link_config(t, set_mtu); 2300 return 0; 2301 } 2302 2303 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2304 struct nlattr *data[], 2305 struct netlink_ext_ack *extack) 2306 { 2307 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2308 struct __ip6_tnl_parm p; 2309 struct ip6_tnl *t; 2310 2311 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2312 if (IS_ERR(t)) 2313 return PTR_ERR(t); 2314 2315 ip6erspan_set_version(data, &p); 2316 ip6gre_tunnel_unlink_md(ign, t); 2317 ip6gre_tunnel_unlink(ign, t); 2318 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2319 ip6erspan_tunnel_link_md(ign, t); 2320 ip6gre_tunnel_link(ign, t); 2321 return 0; 2322 } 2323 2324 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2325 .kind = "ip6gre", 2326 .maxtype = IFLA_GRE_MAX, 2327 .policy = ip6gre_policy, 2328 .priv_size = sizeof(struct ip6_tnl), 2329 .setup = ip6gre_tunnel_setup, 2330 .validate = ip6gre_tunnel_validate, 2331 .newlink = ip6gre_newlink, 2332 .changelink = ip6gre_changelink, 2333 .dellink = ip6gre_dellink, 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 ip6gre_tap_ops __read_mostly = { 2340 .kind = "ip6gretap", 2341 .maxtype = IFLA_GRE_MAX, 2342 .policy = ip6gre_policy, 2343 .priv_size = sizeof(struct ip6_tnl), 2344 .setup = ip6gre_tap_setup, 2345 .validate = ip6gre_tap_validate, 2346 .newlink = ip6gre_newlink, 2347 .changelink = ip6gre_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 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2354 .kind = "ip6erspan", 2355 .maxtype = IFLA_GRE_MAX, 2356 .policy = ip6gre_policy, 2357 .priv_size = sizeof(struct ip6_tnl), 2358 .setup = ip6erspan_tap_setup, 2359 .validate = ip6erspan_tap_validate, 2360 .newlink = ip6erspan_newlink, 2361 .changelink = ip6erspan_changelink, 2362 .get_size = ip6gre_get_size, 2363 .fill_info = ip6gre_fill_info, 2364 .get_link_net = ip6_tnl_get_link_net, 2365 }; 2366 2367 /* 2368 * And now the modules code and kernel interface. 2369 */ 2370 2371 static int __init ip6gre_init(void) 2372 { 2373 int err; 2374 2375 pr_info("GRE over IPv6 tunneling driver\n"); 2376 2377 err = register_pernet_device(&ip6gre_net_ops); 2378 if (err < 0) 2379 return err; 2380 2381 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2382 if (err < 0) { 2383 pr_info("%s: can't add protocol\n", __func__); 2384 goto add_proto_failed; 2385 } 2386 2387 err = rtnl_link_register(&ip6gre_link_ops); 2388 if (err < 0) 2389 goto rtnl_link_failed; 2390 2391 err = rtnl_link_register(&ip6gre_tap_ops); 2392 if (err < 0) 2393 goto tap_ops_failed; 2394 2395 err = rtnl_link_register(&ip6erspan_tap_ops); 2396 if (err < 0) 2397 goto erspan_link_failed; 2398 2399 out: 2400 return err; 2401 2402 erspan_link_failed: 2403 rtnl_link_unregister(&ip6gre_tap_ops); 2404 tap_ops_failed: 2405 rtnl_link_unregister(&ip6gre_link_ops); 2406 rtnl_link_failed: 2407 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2408 add_proto_failed: 2409 unregister_pernet_device(&ip6gre_net_ops); 2410 goto out; 2411 } 2412 2413 static void __exit ip6gre_fini(void) 2414 { 2415 rtnl_link_unregister(&ip6gre_tap_ops); 2416 rtnl_link_unregister(&ip6gre_link_ops); 2417 rtnl_link_unregister(&ip6erspan_tap_ops); 2418 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2419 unregister_pernet_device(&ip6gre_net_ops); 2420 } 2421 2422 module_init(ip6gre_init); 2423 module_exit(ip6gre_fini); 2424 MODULE_LICENSE("GPL"); 2425 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2426 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2427 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2428 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2429 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2430 MODULE_ALIAS_NETDEV("ip6gre0"); 2431