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 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 874 { 875 struct ip6_tnl *t = netdev_priv(dev); 876 int encap_limit = -1; 877 struct flowi6 fl6; 878 __u8 dsfield = 0; 879 __u32 mtu; 880 int err; 881 882 if (!t->parms.collect_md && 883 prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit)) 884 return -1; 885 886 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 887 if (err) 888 return err; 889 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol); 890 891 return err; 892 } 893 894 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 895 struct net_device *dev) 896 { 897 struct ip6_tnl *t = netdev_priv(dev); 898 __be16 payload_protocol; 899 int ret; 900 901 if (!pskb_inet_may_pull(skb)) 902 goto tx_err; 903 904 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 905 goto tx_err; 906 907 payload_protocol = skb_protocol(skb, true); 908 switch (payload_protocol) { 909 case htons(ETH_P_IP): 910 ret = ip6gre_xmit_ipv4(skb, dev); 911 break; 912 case htons(ETH_P_IPV6): 913 ret = ip6gre_xmit_ipv6(skb, dev); 914 break; 915 default: 916 ret = ip6gre_xmit_other(skb, dev); 917 break; 918 } 919 920 if (ret < 0) 921 goto tx_err; 922 923 return NETDEV_TX_OK; 924 925 tx_err: 926 if (!t->parms.collect_md || !IS_ERR(skb_tunnel_info_txcheck(skb))) 927 DEV_STATS_INC(dev, tx_errors); 928 DEV_STATS_INC(dev, tx_dropped); 929 kfree_skb(skb); 930 return NETDEV_TX_OK; 931 } 932 933 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb, 934 struct net_device *dev) 935 { 936 struct ip_tunnel_info *tun_info = NULL; 937 struct ip6_tnl *t = netdev_priv(dev); 938 struct dst_entry *dst = skb_dst(skb); 939 bool truncate = false; 940 int encap_limit = -1; 941 __u8 dsfield = false; 942 struct flowi6 fl6; 943 int err = -EINVAL; 944 __be16 proto; 945 __u32 mtu; 946 int nhoff; 947 948 if (!pskb_inet_may_pull(skb)) 949 goto tx_err; 950 951 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 952 goto tx_err; 953 954 if (gre_handle_offloads(skb, false)) 955 goto tx_err; 956 957 if (skb->len > dev->mtu + dev->hard_header_len) { 958 if (pskb_trim(skb, dev->mtu + dev->hard_header_len)) 959 goto tx_err; 960 truncate = true; 961 } 962 963 nhoff = skb_network_offset(skb); 964 if (skb->protocol == htons(ETH_P_IP) && 965 (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff)) 966 truncate = true; 967 968 if (skb->protocol == htons(ETH_P_IPV6)) { 969 int thoff; 970 971 if (skb_transport_header_was_set(skb)) 972 thoff = skb_transport_offset(skb); 973 else 974 thoff = nhoff + sizeof(struct ipv6hdr); 975 if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff) 976 truncate = true; 977 } 978 979 if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen)) 980 goto tx_err; 981 982 t->parms.o_flags &= ~TUNNEL_KEY; 983 IPCB(skb)->flags = 0; 984 985 /* For collect_md mode, derive fl6 from the tunnel key, 986 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}. 987 */ 988 if (t->parms.collect_md) { 989 const struct ip_tunnel_key *key; 990 struct erspan_metadata *md; 991 __be32 tun_id; 992 993 tun_info = skb_tunnel_info_txcheck(skb); 994 if (IS_ERR(tun_info) || 995 unlikely(ip_tunnel_info_af(tun_info) != AF_INET6)) 996 goto tx_err; 997 998 key = &tun_info->key; 999 memset(&fl6, 0, sizeof(fl6)); 1000 fl6.flowi6_proto = IPPROTO_GRE; 1001 fl6.daddr = key->u.ipv6.dst; 1002 fl6.flowlabel = key->label; 1003 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 1004 fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id); 1005 1006 dsfield = key->tos; 1007 if (!(tun_info->key.tun_flags & TUNNEL_ERSPAN_OPT)) 1008 goto tx_err; 1009 if (tun_info->options_len < sizeof(*md)) 1010 goto tx_err; 1011 md = ip_tunnel_info_opts(tun_info); 1012 1013 tun_id = tunnel_id_to_key32(key->tun_id); 1014 if (md->version == 1) { 1015 erspan_build_header(skb, 1016 ntohl(tun_id), 1017 ntohl(md->u.index), truncate, 1018 false); 1019 proto = htons(ETH_P_ERSPAN); 1020 } else if (md->version == 2) { 1021 erspan_build_header_v2(skb, 1022 ntohl(tun_id), 1023 md->u.md2.dir, 1024 get_hwid(&md->u.md2), 1025 truncate, false); 1026 proto = htons(ETH_P_ERSPAN2); 1027 } else { 1028 goto tx_err; 1029 } 1030 } else { 1031 switch (skb->protocol) { 1032 case htons(ETH_P_IP): 1033 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 1034 prepare_ip6gre_xmit_ipv4(skb, dev, &fl6, 1035 &dsfield, &encap_limit); 1036 break; 1037 case htons(ETH_P_IPV6): 1038 if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr)) 1039 goto tx_err; 1040 if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, 1041 &dsfield, &encap_limit)) 1042 goto tx_err; 1043 break; 1044 default: 1045 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 1046 break; 1047 } 1048 1049 if (t->parms.erspan_ver == 1) { 1050 erspan_build_header(skb, ntohl(t->parms.o_key), 1051 t->parms.index, 1052 truncate, false); 1053 proto = htons(ETH_P_ERSPAN); 1054 } else if (t->parms.erspan_ver == 2) { 1055 erspan_build_header_v2(skb, ntohl(t->parms.o_key), 1056 t->parms.dir, 1057 t->parms.hwid, 1058 truncate, false); 1059 proto = htons(ETH_P_ERSPAN2); 1060 } else { 1061 goto tx_err; 1062 } 1063 1064 fl6.daddr = t->parms.raddr; 1065 } 1066 1067 /* Push GRE header. */ 1068 gre_build_header(skb, 8, TUNNEL_SEQ, proto, 0, htonl(atomic_fetch_inc(&t->o_seqno))); 1069 1070 /* TooBig packet may have updated dst->dev's mtu */ 1071 if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu) 1072 dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu, false); 1073 1074 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 1075 NEXTHDR_GRE); 1076 if (err != 0) { 1077 /* XXX: send ICMP error even if DF is not set. */ 1078 if (err == -EMSGSIZE) { 1079 if (skb->protocol == htons(ETH_P_IP)) 1080 icmp_ndo_send(skb, ICMP_DEST_UNREACH, 1081 ICMP_FRAG_NEEDED, htonl(mtu)); 1082 else 1083 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 1084 } 1085 1086 goto tx_err; 1087 } 1088 return NETDEV_TX_OK; 1089 1090 tx_err: 1091 if (!IS_ERR(tun_info)) 1092 DEV_STATS_INC(dev, tx_errors); 1093 DEV_STATS_INC(dev, tx_dropped); 1094 kfree_skb(skb); 1095 return NETDEV_TX_OK; 1096 } 1097 1098 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t) 1099 { 1100 struct net_device *dev = t->dev; 1101 struct __ip6_tnl_parm *p = &t->parms; 1102 struct flowi6 *fl6 = &t->fl.u.ip6; 1103 1104 if (dev->type != ARPHRD_ETHER) { 1105 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr)); 1106 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 1107 } 1108 1109 /* Set up flowi template */ 1110 fl6->saddr = p->laddr; 1111 fl6->daddr = p->raddr; 1112 fl6->flowi6_oif = p->link; 1113 fl6->flowlabel = 0; 1114 fl6->flowi6_proto = IPPROTO_GRE; 1115 fl6->fl6_gre_key = t->parms.o_key; 1116 1117 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 1118 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 1119 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 1120 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 1121 1122 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 1123 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 1124 1125 if (p->flags&IP6_TNL_F_CAP_XMIT && 1126 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 1127 dev->flags |= IFF_POINTOPOINT; 1128 else 1129 dev->flags &= ~IFF_POINTOPOINT; 1130 } 1131 1132 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu, 1133 int t_hlen) 1134 { 1135 const struct __ip6_tnl_parm *p = &t->parms; 1136 struct net_device *dev = t->dev; 1137 1138 if (p->flags & IP6_TNL_F_CAP_XMIT) { 1139 int strict = (ipv6_addr_type(&p->raddr) & 1140 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 1141 1142 struct rt6_info *rt = rt6_lookup(t->net, 1143 &p->raddr, &p->laddr, 1144 p->link, NULL, strict); 1145 1146 if (!rt) 1147 return; 1148 1149 if (rt->dst.dev) { 1150 unsigned short dst_len = rt->dst.dev->hard_header_len + 1151 t_hlen; 1152 1153 if (t->dev->header_ops) 1154 dev->hard_header_len = dst_len; 1155 else 1156 dev->needed_headroom = dst_len; 1157 1158 if (set_mtu) { 1159 int mtu = rt->dst.dev->mtu - t_hlen; 1160 1161 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1162 mtu -= 8; 1163 if (dev->type == ARPHRD_ETHER) 1164 mtu -= ETH_HLEN; 1165 1166 if (mtu < IPV6_MIN_MTU) 1167 mtu = IPV6_MIN_MTU; 1168 WRITE_ONCE(dev->mtu, mtu); 1169 } 1170 } 1171 ip6_rt_put(rt); 1172 } 1173 } 1174 1175 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel) 1176 { 1177 int t_hlen; 1178 1179 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1180 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1181 1182 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1183 1184 if (tunnel->dev->header_ops) 1185 tunnel->dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1186 else 1187 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1188 1189 return t_hlen; 1190 } 1191 1192 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 1193 { 1194 ip6gre_tnl_link_config_common(t); 1195 ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t)); 1196 } 1197 1198 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t, 1199 const struct __ip6_tnl_parm *p) 1200 { 1201 t->parms.laddr = p->laddr; 1202 t->parms.raddr = p->raddr; 1203 t->parms.flags = p->flags; 1204 t->parms.hop_limit = p->hop_limit; 1205 t->parms.encap_limit = p->encap_limit; 1206 t->parms.flowinfo = p->flowinfo; 1207 t->parms.link = p->link; 1208 t->parms.proto = p->proto; 1209 t->parms.i_key = p->i_key; 1210 t->parms.o_key = p->o_key; 1211 t->parms.i_flags = p->i_flags; 1212 t->parms.o_flags = p->o_flags; 1213 t->parms.fwmark = p->fwmark; 1214 t->parms.erspan_ver = p->erspan_ver; 1215 t->parms.index = p->index; 1216 t->parms.dir = p->dir; 1217 t->parms.hwid = p->hwid; 1218 dst_cache_reset(&t->dst_cache); 1219 } 1220 1221 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p, 1222 int set_mtu) 1223 { 1224 ip6gre_tnl_copy_tnl_parm(t, p); 1225 ip6gre_tnl_link_config(t, set_mtu); 1226 return 0; 1227 } 1228 1229 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 1230 const struct ip6_tnl_parm2 *u) 1231 { 1232 p->laddr = u->laddr; 1233 p->raddr = u->raddr; 1234 p->flags = u->flags; 1235 p->hop_limit = u->hop_limit; 1236 p->encap_limit = u->encap_limit; 1237 p->flowinfo = u->flowinfo; 1238 p->link = u->link; 1239 p->i_key = u->i_key; 1240 p->o_key = u->o_key; 1241 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 1242 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 1243 memcpy(p->name, u->name, sizeof(u->name)); 1244 } 1245 1246 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 1247 const struct __ip6_tnl_parm *p) 1248 { 1249 u->proto = IPPROTO_GRE; 1250 u->laddr = p->laddr; 1251 u->raddr = p->raddr; 1252 u->flags = p->flags; 1253 u->hop_limit = p->hop_limit; 1254 u->encap_limit = p->encap_limit; 1255 u->flowinfo = p->flowinfo; 1256 u->link = p->link; 1257 u->i_key = p->i_key; 1258 u->o_key = p->o_key; 1259 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 1260 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 1261 memcpy(u->name, p->name, sizeof(u->name)); 1262 } 1263 1264 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev, 1265 struct ifreq *ifr, void __user *data, 1266 int cmd) 1267 { 1268 int err = 0; 1269 struct ip6_tnl_parm2 p; 1270 struct __ip6_tnl_parm p1; 1271 struct ip6_tnl *t = netdev_priv(dev); 1272 struct net *net = t->net; 1273 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1274 1275 memset(&p1, 0, sizeof(p1)); 1276 1277 switch (cmd) { 1278 case SIOCGETTUNNEL: 1279 if (dev == ign->fb_tunnel_dev) { 1280 if (copy_from_user(&p, data, sizeof(p))) { 1281 err = -EFAULT; 1282 break; 1283 } 1284 ip6gre_tnl_parm_from_user(&p1, &p); 1285 t = ip6gre_tunnel_locate(net, &p1, 0); 1286 if (!t) 1287 t = netdev_priv(dev); 1288 } 1289 memset(&p, 0, sizeof(p)); 1290 ip6gre_tnl_parm_to_user(&p, &t->parms); 1291 if (copy_to_user(data, &p, sizeof(p))) 1292 err = -EFAULT; 1293 break; 1294 1295 case SIOCADDTUNNEL: 1296 case SIOCCHGTUNNEL: 1297 err = -EPERM; 1298 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1299 goto done; 1300 1301 err = -EFAULT; 1302 if (copy_from_user(&p, data, sizeof(p))) 1303 goto done; 1304 1305 err = -EINVAL; 1306 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 1307 goto done; 1308 1309 if (!(p.i_flags&GRE_KEY)) 1310 p.i_key = 0; 1311 if (!(p.o_flags&GRE_KEY)) 1312 p.o_key = 0; 1313 1314 ip6gre_tnl_parm_from_user(&p1, &p); 1315 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 1316 1317 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 1318 if (t) { 1319 if (t->dev != dev) { 1320 err = -EEXIST; 1321 break; 1322 } 1323 } else { 1324 t = netdev_priv(dev); 1325 1326 ip6gre_tunnel_unlink(ign, t); 1327 synchronize_net(); 1328 ip6gre_tnl_change(t, &p1, 1); 1329 ip6gre_tunnel_link(ign, t); 1330 netdev_state_change(dev); 1331 } 1332 } 1333 1334 if (t) { 1335 err = 0; 1336 1337 memset(&p, 0, sizeof(p)); 1338 ip6gre_tnl_parm_to_user(&p, &t->parms); 1339 if (copy_to_user(data, &p, sizeof(p))) 1340 err = -EFAULT; 1341 } else 1342 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 1343 break; 1344 1345 case SIOCDELTUNNEL: 1346 err = -EPERM; 1347 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1348 goto done; 1349 1350 if (dev == ign->fb_tunnel_dev) { 1351 err = -EFAULT; 1352 if (copy_from_user(&p, data, sizeof(p))) 1353 goto done; 1354 err = -ENOENT; 1355 ip6gre_tnl_parm_from_user(&p1, &p); 1356 t = ip6gre_tunnel_locate(net, &p1, 0); 1357 if (!t) 1358 goto done; 1359 err = -EPERM; 1360 if (t == netdev_priv(ign->fb_tunnel_dev)) 1361 goto done; 1362 dev = t->dev; 1363 } 1364 unregister_netdevice(dev); 1365 err = 0; 1366 break; 1367 1368 default: 1369 err = -EINVAL; 1370 } 1371 1372 done: 1373 return err; 1374 } 1375 1376 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 1377 unsigned short type, const void *daddr, 1378 const void *saddr, unsigned int len) 1379 { 1380 struct ip6_tnl *t = netdev_priv(dev); 1381 struct ipv6hdr *ipv6h; 1382 __be16 *p; 1383 1384 ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h)); 1385 ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb, 1386 t->fl.u.ip6.flowlabel, 1387 true, &t->fl.u.ip6)); 1388 ipv6h->hop_limit = t->parms.hop_limit; 1389 ipv6h->nexthdr = NEXTHDR_GRE; 1390 ipv6h->saddr = t->parms.laddr; 1391 ipv6h->daddr = t->parms.raddr; 1392 1393 p = (__be16 *)(ipv6h + 1); 1394 p[0] = t->parms.o_flags; 1395 p[1] = htons(type); 1396 1397 /* 1398 * Set the source hardware address. 1399 */ 1400 1401 if (saddr) 1402 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 1403 if (daddr) 1404 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 1405 if (!ipv6_addr_any(&ipv6h->daddr)) 1406 return t->hlen; 1407 1408 return -t->hlen; 1409 } 1410 1411 static const struct header_ops ip6gre_header_ops = { 1412 .create = ip6gre_header, 1413 }; 1414 1415 static const struct net_device_ops ip6gre_netdev_ops = { 1416 .ndo_init = ip6gre_tunnel_init, 1417 .ndo_uninit = ip6gre_tunnel_uninit, 1418 .ndo_start_xmit = ip6gre_tunnel_xmit, 1419 .ndo_siocdevprivate = ip6gre_tunnel_siocdevprivate, 1420 .ndo_change_mtu = ip6_tnl_change_mtu, 1421 .ndo_get_stats64 = dev_get_tstats64, 1422 .ndo_get_iflink = ip6_tnl_get_iflink, 1423 }; 1424 1425 static void ip6gre_dev_free(struct net_device *dev) 1426 { 1427 struct ip6_tnl *t = netdev_priv(dev); 1428 1429 gro_cells_destroy(&t->gro_cells); 1430 dst_cache_destroy(&t->dst_cache); 1431 free_percpu(dev->tstats); 1432 } 1433 1434 static void ip6gre_tunnel_setup(struct net_device *dev) 1435 { 1436 dev->netdev_ops = &ip6gre_netdev_ops; 1437 dev->needs_free_netdev = true; 1438 dev->priv_destructor = ip6gre_dev_free; 1439 1440 dev->type = ARPHRD_IP6GRE; 1441 1442 dev->flags |= IFF_NOARP; 1443 dev->addr_len = sizeof(struct in6_addr); 1444 netif_keep_dst(dev); 1445 /* This perm addr will be used as interface identifier by IPv6 */ 1446 dev->addr_assign_type = NET_ADDR_RANDOM; 1447 eth_random_addr(dev->perm_addr); 1448 } 1449 1450 #define GRE6_FEATURES (NETIF_F_SG | \ 1451 NETIF_F_FRAGLIST | \ 1452 NETIF_F_HIGHDMA | \ 1453 NETIF_F_HW_CSUM) 1454 1455 static void ip6gre_tnl_init_features(struct net_device *dev) 1456 { 1457 struct ip6_tnl *nt = netdev_priv(dev); 1458 __be16 flags; 1459 1460 dev->features |= GRE6_FEATURES | NETIF_F_LLTX; 1461 dev->hw_features |= GRE6_FEATURES; 1462 1463 flags = nt->parms.o_flags; 1464 1465 /* TCP offload with GRE SEQ is not supported, nor can we support 2 1466 * levels of outer headers requiring an update. 1467 */ 1468 if (flags & TUNNEL_SEQ) 1469 return; 1470 if (flags & TUNNEL_CSUM && nt->encap.type != TUNNEL_ENCAP_NONE) 1471 return; 1472 1473 dev->features |= NETIF_F_GSO_SOFTWARE; 1474 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1475 } 1476 1477 static int ip6gre_tunnel_init_common(struct net_device *dev) 1478 { 1479 struct ip6_tnl *tunnel; 1480 int ret; 1481 int t_hlen; 1482 1483 tunnel = netdev_priv(dev); 1484 1485 tunnel->dev = dev; 1486 tunnel->net = dev_net(dev); 1487 strcpy(tunnel->parms.name, dev->name); 1488 1489 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1490 if (!dev->tstats) 1491 return -ENOMEM; 1492 1493 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1494 if (ret) 1495 goto cleanup_alloc_pcpu_stats; 1496 1497 ret = gro_cells_init(&tunnel->gro_cells, dev); 1498 if (ret) 1499 goto cleanup_dst_cache_init; 1500 1501 t_hlen = ip6gre_calc_hlen(tunnel); 1502 dev->mtu = ETH_DATA_LEN - t_hlen; 1503 if (dev->type == ARPHRD_ETHER) 1504 dev->mtu -= ETH_HLEN; 1505 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1506 dev->mtu -= 8; 1507 1508 if (tunnel->parms.collect_md) { 1509 netif_keep_dst(dev); 1510 } 1511 ip6gre_tnl_init_features(dev); 1512 1513 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1514 netdev_lockdep_set_classes(dev); 1515 return 0; 1516 1517 cleanup_dst_cache_init: 1518 dst_cache_destroy(&tunnel->dst_cache); 1519 cleanup_alloc_pcpu_stats: 1520 free_percpu(dev->tstats); 1521 dev->tstats = NULL; 1522 return ret; 1523 } 1524 1525 static int ip6gre_tunnel_init(struct net_device *dev) 1526 { 1527 struct ip6_tnl *tunnel; 1528 int ret; 1529 1530 ret = ip6gre_tunnel_init_common(dev); 1531 if (ret) 1532 return ret; 1533 1534 tunnel = netdev_priv(dev); 1535 1536 if (tunnel->parms.collect_md) 1537 return 0; 1538 1539 __dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1540 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1541 1542 if (ipv6_addr_any(&tunnel->parms.raddr)) 1543 dev->header_ops = &ip6gre_header_ops; 1544 1545 return 0; 1546 } 1547 1548 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1549 { 1550 struct ip6_tnl *tunnel = netdev_priv(dev); 1551 1552 tunnel->dev = dev; 1553 tunnel->net = dev_net(dev); 1554 strcpy(tunnel->parms.name, dev->name); 1555 1556 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1557 } 1558 1559 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1560 .handler = gre_rcv, 1561 .err_handler = ip6gre_err, 1562 .flags = INET6_PROTO_FINAL, 1563 }; 1564 1565 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1566 { 1567 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1568 struct net_device *dev, *aux; 1569 int prio; 1570 1571 for_each_netdev_safe(net, dev, aux) 1572 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1573 dev->rtnl_link_ops == &ip6gre_tap_ops || 1574 dev->rtnl_link_ops == &ip6erspan_tap_ops) 1575 unregister_netdevice_queue(dev, head); 1576 1577 for (prio = 0; prio < 4; prio++) { 1578 int h; 1579 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1580 struct ip6_tnl *t; 1581 1582 t = rtnl_dereference(ign->tunnels[prio][h]); 1583 1584 while (t) { 1585 /* If dev is in the same netns, it has already 1586 * been added to the list by the previous loop. 1587 */ 1588 if (!net_eq(dev_net(t->dev), net)) 1589 unregister_netdevice_queue(t->dev, 1590 head); 1591 t = rtnl_dereference(t->next); 1592 } 1593 } 1594 } 1595 } 1596 1597 static int __net_init ip6gre_init_net(struct net *net) 1598 { 1599 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1600 struct net_device *ndev; 1601 int err; 1602 1603 if (!net_has_fallback_tunnels(net)) 1604 return 0; 1605 ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1606 NET_NAME_UNKNOWN, ip6gre_tunnel_setup); 1607 if (!ndev) { 1608 err = -ENOMEM; 1609 goto err_alloc_dev; 1610 } 1611 ign->fb_tunnel_dev = ndev; 1612 dev_net_set(ign->fb_tunnel_dev, net); 1613 /* FB netdevice is special: we have one, and only one per netns. 1614 * Allowing to move it to another netns is clearly unsafe. 1615 */ 1616 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1617 1618 1619 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1620 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1621 1622 err = register_netdev(ign->fb_tunnel_dev); 1623 if (err) 1624 goto err_reg_dev; 1625 1626 rcu_assign_pointer(ign->tunnels_wc[0], 1627 netdev_priv(ign->fb_tunnel_dev)); 1628 return 0; 1629 1630 err_reg_dev: 1631 free_netdev(ndev); 1632 err_alloc_dev: 1633 return err; 1634 } 1635 1636 static void __net_exit ip6gre_exit_batch_rtnl(struct list_head *net_list, 1637 struct list_head *dev_to_kill) 1638 { 1639 struct net *net; 1640 1641 ASSERT_RTNL(); 1642 list_for_each_entry(net, net_list, exit_list) 1643 ip6gre_destroy_tunnels(net, dev_to_kill); 1644 } 1645 1646 static struct pernet_operations ip6gre_net_ops = { 1647 .init = ip6gre_init_net, 1648 .exit_batch_rtnl = ip6gre_exit_batch_rtnl, 1649 .id = &ip6gre_net_id, 1650 .size = sizeof(struct ip6gre_net), 1651 }; 1652 1653 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 1654 struct netlink_ext_ack *extack) 1655 { 1656 __be16 flags; 1657 1658 if (!data) 1659 return 0; 1660 1661 flags = 0; 1662 if (data[IFLA_GRE_IFLAGS]) 1663 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1664 if (data[IFLA_GRE_OFLAGS]) 1665 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1666 if (flags & (GRE_VERSION|GRE_ROUTING)) 1667 return -EINVAL; 1668 1669 return 0; 1670 } 1671 1672 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1673 struct netlink_ext_ack *extack) 1674 { 1675 struct in6_addr daddr; 1676 1677 if (tb[IFLA_ADDRESS]) { 1678 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1679 return -EINVAL; 1680 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1681 return -EADDRNOTAVAIL; 1682 } 1683 1684 if (!data) 1685 goto out; 1686 1687 if (data[IFLA_GRE_REMOTE]) { 1688 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1689 if (ipv6_addr_any(&daddr)) 1690 return -EINVAL; 1691 } 1692 1693 out: 1694 return ip6gre_tunnel_validate(tb, data, extack); 1695 } 1696 1697 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[], 1698 struct netlink_ext_ack *extack) 1699 { 1700 __be16 flags = 0; 1701 int ret, ver = 0; 1702 1703 if (!data) 1704 return 0; 1705 1706 ret = ip6gre_tap_validate(tb, data, extack); 1707 if (ret) 1708 return ret; 1709 1710 /* ERSPAN should only have GRE sequence and key flag */ 1711 if (data[IFLA_GRE_OFLAGS]) 1712 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1713 if (data[IFLA_GRE_IFLAGS]) 1714 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1715 if (!data[IFLA_GRE_COLLECT_METADATA] && 1716 flags != (GRE_SEQ | GRE_KEY)) 1717 return -EINVAL; 1718 1719 /* ERSPAN Session ID only has 10-bit. Since we reuse 1720 * 32-bit key field as ID, check it's range. 1721 */ 1722 if (data[IFLA_GRE_IKEY] && 1723 (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK)) 1724 return -EINVAL; 1725 1726 if (data[IFLA_GRE_OKEY] && 1727 (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK)) 1728 return -EINVAL; 1729 1730 if (data[IFLA_GRE_ERSPAN_VER]) { 1731 ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1732 if (ver != 1 && ver != 2) 1733 return -EINVAL; 1734 } 1735 1736 if (ver == 1) { 1737 if (data[IFLA_GRE_ERSPAN_INDEX]) { 1738 u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1739 1740 if (index & ~INDEX_MASK) 1741 return -EINVAL; 1742 } 1743 } else if (ver == 2) { 1744 if (data[IFLA_GRE_ERSPAN_DIR]) { 1745 u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1746 1747 if (dir & ~(DIR_MASK >> DIR_OFFSET)) 1748 return -EINVAL; 1749 } 1750 1751 if (data[IFLA_GRE_ERSPAN_HWID]) { 1752 u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1753 1754 if (hwid & ~(HWID_MASK >> HWID_OFFSET)) 1755 return -EINVAL; 1756 } 1757 } 1758 1759 return 0; 1760 } 1761 1762 static void ip6erspan_set_version(struct nlattr *data[], 1763 struct __ip6_tnl_parm *parms) 1764 { 1765 if (!data) 1766 return; 1767 1768 parms->erspan_ver = 1; 1769 if (data[IFLA_GRE_ERSPAN_VER]) 1770 parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]); 1771 1772 if (parms->erspan_ver == 1) { 1773 if (data[IFLA_GRE_ERSPAN_INDEX]) 1774 parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]); 1775 } else if (parms->erspan_ver == 2) { 1776 if (data[IFLA_GRE_ERSPAN_DIR]) 1777 parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]); 1778 if (data[IFLA_GRE_ERSPAN_HWID]) 1779 parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]); 1780 } 1781 } 1782 1783 static void ip6gre_netlink_parms(struct nlattr *data[], 1784 struct __ip6_tnl_parm *parms) 1785 { 1786 memset(parms, 0, sizeof(*parms)); 1787 1788 if (!data) 1789 return; 1790 1791 if (data[IFLA_GRE_LINK]) 1792 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1793 1794 if (data[IFLA_GRE_IFLAGS]) 1795 parms->i_flags = gre_flags_to_tnl_flags( 1796 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1797 1798 if (data[IFLA_GRE_OFLAGS]) 1799 parms->o_flags = gre_flags_to_tnl_flags( 1800 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1801 1802 if (data[IFLA_GRE_IKEY]) 1803 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1804 1805 if (data[IFLA_GRE_OKEY]) 1806 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1807 1808 if (data[IFLA_GRE_LOCAL]) 1809 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1810 1811 if (data[IFLA_GRE_REMOTE]) 1812 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1813 1814 if (data[IFLA_GRE_TTL]) 1815 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1816 1817 if (data[IFLA_GRE_ENCAP_LIMIT]) 1818 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1819 1820 if (data[IFLA_GRE_FLOWINFO]) 1821 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1822 1823 if (data[IFLA_GRE_FLAGS]) 1824 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1825 1826 if (data[IFLA_GRE_FWMARK]) 1827 parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]); 1828 1829 if (data[IFLA_GRE_COLLECT_METADATA]) 1830 parms->collect_md = true; 1831 } 1832 1833 static int ip6gre_tap_init(struct net_device *dev) 1834 { 1835 int ret; 1836 1837 ret = ip6gre_tunnel_init_common(dev); 1838 if (ret) 1839 return ret; 1840 1841 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1842 1843 return 0; 1844 } 1845 1846 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1847 .ndo_init = ip6gre_tap_init, 1848 .ndo_uninit = ip6gre_tunnel_uninit, 1849 .ndo_start_xmit = ip6gre_tunnel_xmit, 1850 .ndo_set_mac_address = eth_mac_addr, 1851 .ndo_validate_addr = eth_validate_addr, 1852 .ndo_change_mtu = ip6_tnl_change_mtu, 1853 .ndo_get_stats64 = dev_get_tstats64, 1854 .ndo_get_iflink = ip6_tnl_get_iflink, 1855 }; 1856 1857 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel) 1858 { 1859 int t_hlen; 1860 1861 tunnel->tun_hlen = 8; 1862 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen + 1863 erspan_hdr_len(tunnel->parms.erspan_ver); 1864 1865 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1866 tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen; 1867 return t_hlen; 1868 } 1869 1870 static int ip6erspan_tap_init(struct net_device *dev) 1871 { 1872 struct ip6_tnl *tunnel; 1873 int t_hlen; 1874 int ret; 1875 1876 tunnel = netdev_priv(dev); 1877 1878 tunnel->dev = dev; 1879 tunnel->net = dev_net(dev); 1880 strcpy(tunnel->parms.name, dev->name); 1881 1882 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1883 if (!dev->tstats) 1884 return -ENOMEM; 1885 1886 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1887 if (ret) 1888 goto cleanup_alloc_pcpu_stats; 1889 1890 ret = gro_cells_init(&tunnel->gro_cells, dev); 1891 if (ret) 1892 goto cleanup_dst_cache_init; 1893 1894 t_hlen = ip6erspan_calc_hlen(tunnel); 1895 dev->mtu = ETH_DATA_LEN - t_hlen; 1896 if (dev->type == ARPHRD_ETHER) 1897 dev->mtu -= ETH_HLEN; 1898 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1899 dev->mtu -= 8; 1900 1901 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1902 ip6erspan_tnl_link_config(tunnel, 1); 1903 1904 netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL); 1905 netdev_lockdep_set_classes(dev); 1906 return 0; 1907 1908 cleanup_dst_cache_init: 1909 dst_cache_destroy(&tunnel->dst_cache); 1910 cleanup_alloc_pcpu_stats: 1911 free_percpu(dev->tstats); 1912 dev->tstats = NULL; 1913 return ret; 1914 } 1915 1916 static const struct net_device_ops ip6erspan_netdev_ops = { 1917 .ndo_init = ip6erspan_tap_init, 1918 .ndo_uninit = ip6erspan_tunnel_uninit, 1919 .ndo_start_xmit = ip6erspan_tunnel_xmit, 1920 .ndo_set_mac_address = eth_mac_addr, 1921 .ndo_validate_addr = eth_validate_addr, 1922 .ndo_change_mtu = ip6_tnl_change_mtu, 1923 .ndo_get_stats64 = dev_get_tstats64, 1924 .ndo_get_iflink = ip6_tnl_get_iflink, 1925 }; 1926 1927 static void ip6gre_tap_setup(struct net_device *dev) 1928 { 1929 1930 ether_setup(dev); 1931 1932 dev->max_mtu = 0; 1933 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1934 dev->needs_free_netdev = true; 1935 dev->priv_destructor = ip6gre_dev_free; 1936 1937 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1938 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1939 netif_keep_dst(dev); 1940 } 1941 1942 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1943 struct ip_tunnel_encap *ipencap) 1944 { 1945 bool ret = false; 1946 1947 memset(ipencap, 0, sizeof(*ipencap)); 1948 1949 if (!data) 1950 return ret; 1951 1952 if (data[IFLA_GRE_ENCAP_TYPE]) { 1953 ret = true; 1954 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1955 } 1956 1957 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1958 ret = true; 1959 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1960 } 1961 1962 if (data[IFLA_GRE_ENCAP_SPORT]) { 1963 ret = true; 1964 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1965 } 1966 1967 if (data[IFLA_GRE_ENCAP_DPORT]) { 1968 ret = true; 1969 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1970 } 1971 1972 return ret; 1973 } 1974 1975 static int ip6gre_newlink_common(struct net *src_net, struct net_device *dev, 1976 struct nlattr *tb[], struct nlattr *data[], 1977 struct netlink_ext_ack *extack) 1978 { 1979 struct ip6_tnl *nt; 1980 struct ip_tunnel_encap ipencap; 1981 int err; 1982 1983 nt = netdev_priv(dev); 1984 1985 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1986 int err = ip6_tnl_encap_setup(nt, &ipencap); 1987 1988 if (err < 0) 1989 return err; 1990 } 1991 1992 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1993 eth_hw_addr_random(dev); 1994 1995 nt->dev = dev; 1996 nt->net = dev_net(dev); 1997 1998 err = register_netdevice(dev); 1999 if (err) 2000 goto out; 2001 2002 if (tb[IFLA_MTU]) 2003 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU])); 2004 2005 out: 2006 return err; 2007 } 2008 2009 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 2010 struct nlattr *tb[], struct nlattr *data[], 2011 struct netlink_ext_ack *extack) 2012 { 2013 struct ip6_tnl *nt = netdev_priv(dev); 2014 struct net *net = dev_net(dev); 2015 struct ip6gre_net *ign; 2016 int err; 2017 2018 ip6gre_netlink_parms(data, &nt->parms); 2019 ign = net_generic(net, ip6gre_net_id); 2020 2021 if (nt->parms.collect_md) { 2022 if (rtnl_dereference(ign->collect_md_tun)) 2023 return -EEXIST; 2024 } else { 2025 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2026 return -EEXIST; 2027 } 2028 2029 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2030 if (!err) { 2031 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 2032 ip6gre_tunnel_link_md(ign, nt); 2033 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2034 } 2035 return err; 2036 } 2037 2038 static struct ip6_tnl * 2039 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[], 2040 struct nlattr *data[], struct __ip6_tnl_parm *p_p, 2041 struct netlink_ext_ack *extack) 2042 { 2043 struct ip6_tnl *t, *nt = netdev_priv(dev); 2044 struct net *net = nt->net; 2045 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2046 struct ip_tunnel_encap ipencap; 2047 2048 if (dev == ign->fb_tunnel_dev) 2049 return ERR_PTR(-EINVAL); 2050 2051 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 2052 int err = ip6_tnl_encap_setup(nt, &ipencap); 2053 2054 if (err < 0) 2055 return ERR_PTR(err); 2056 } 2057 2058 ip6gre_netlink_parms(data, p_p); 2059 2060 t = ip6gre_tunnel_locate(net, p_p, 0); 2061 2062 if (t) { 2063 if (t->dev != dev) 2064 return ERR_PTR(-EEXIST); 2065 } else { 2066 t = nt; 2067 } 2068 2069 return t; 2070 } 2071 2072 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 2073 struct nlattr *data[], 2074 struct netlink_ext_ack *extack) 2075 { 2076 struct ip6_tnl *t = netdev_priv(dev); 2077 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 2078 struct __ip6_tnl_parm p; 2079 2080 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2081 if (IS_ERR(t)) 2082 return PTR_ERR(t); 2083 2084 ip6gre_tunnel_unlink_md(ign, t); 2085 ip6gre_tunnel_unlink(ign, t); 2086 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 2087 ip6gre_tunnel_link_md(ign, t); 2088 ip6gre_tunnel_link(ign, t); 2089 return 0; 2090 } 2091 2092 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 2093 { 2094 struct net *net = dev_net(dev); 2095 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 2096 2097 if (dev != ign->fb_tunnel_dev) 2098 unregister_netdevice_queue(dev, head); 2099 } 2100 2101 static size_t ip6gre_get_size(const struct net_device *dev) 2102 { 2103 return 2104 /* IFLA_GRE_LINK */ 2105 nla_total_size(4) + 2106 /* IFLA_GRE_IFLAGS */ 2107 nla_total_size(2) + 2108 /* IFLA_GRE_OFLAGS */ 2109 nla_total_size(2) + 2110 /* IFLA_GRE_IKEY */ 2111 nla_total_size(4) + 2112 /* IFLA_GRE_OKEY */ 2113 nla_total_size(4) + 2114 /* IFLA_GRE_LOCAL */ 2115 nla_total_size(sizeof(struct in6_addr)) + 2116 /* IFLA_GRE_REMOTE */ 2117 nla_total_size(sizeof(struct in6_addr)) + 2118 /* IFLA_GRE_TTL */ 2119 nla_total_size(1) + 2120 /* IFLA_GRE_ENCAP_LIMIT */ 2121 nla_total_size(1) + 2122 /* IFLA_GRE_FLOWINFO */ 2123 nla_total_size(4) + 2124 /* IFLA_GRE_FLAGS */ 2125 nla_total_size(4) + 2126 /* IFLA_GRE_ENCAP_TYPE */ 2127 nla_total_size(2) + 2128 /* IFLA_GRE_ENCAP_FLAGS */ 2129 nla_total_size(2) + 2130 /* IFLA_GRE_ENCAP_SPORT */ 2131 nla_total_size(2) + 2132 /* IFLA_GRE_ENCAP_DPORT */ 2133 nla_total_size(2) + 2134 /* IFLA_GRE_COLLECT_METADATA */ 2135 nla_total_size(0) + 2136 /* IFLA_GRE_FWMARK */ 2137 nla_total_size(4) + 2138 /* IFLA_GRE_ERSPAN_INDEX */ 2139 nla_total_size(4) + 2140 0; 2141 } 2142 2143 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 2144 { 2145 struct ip6_tnl *t = netdev_priv(dev); 2146 struct __ip6_tnl_parm *p = &t->parms; 2147 __be16 o_flags = p->o_flags; 2148 2149 if (p->erspan_ver == 1 || p->erspan_ver == 2) { 2150 if (!p->collect_md) 2151 o_flags |= TUNNEL_KEY; 2152 2153 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver)) 2154 goto nla_put_failure; 2155 2156 if (p->erspan_ver == 1) { 2157 if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index)) 2158 goto nla_put_failure; 2159 } else { 2160 if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir)) 2161 goto nla_put_failure; 2162 if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid)) 2163 goto nla_put_failure; 2164 } 2165 } 2166 2167 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 2168 nla_put_be16(skb, IFLA_GRE_IFLAGS, 2169 gre_tnl_flags_to_gre_flags(p->i_flags)) || 2170 nla_put_be16(skb, IFLA_GRE_OFLAGS, 2171 gre_tnl_flags_to_gre_flags(o_flags)) || 2172 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 2173 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 2174 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 2175 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 2176 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 2177 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 2178 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 2179 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) || 2180 nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark)) 2181 goto nla_put_failure; 2182 2183 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 2184 t->encap.type) || 2185 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 2186 t->encap.sport) || 2187 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 2188 t->encap.dport) || 2189 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 2190 t->encap.flags)) 2191 goto nla_put_failure; 2192 2193 if (p->collect_md) { 2194 if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA)) 2195 goto nla_put_failure; 2196 } 2197 2198 return 0; 2199 2200 nla_put_failure: 2201 return -EMSGSIZE; 2202 } 2203 2204 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 2205 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 2206 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 2207 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 2208 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 2209 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 2210 [IFLA_GRE_LOCAL] = { .len = sizeof_field(struct ipv6hdr, saddr) }, 2211 [IFLA_GRE_REMOTE] = { .len = sizeof_field(struct ipv6hdr, daddr) }, 2212 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 2213 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 2214 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 2215 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 2216 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 2217 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 2218 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 2219 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 2220 [IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG }, 2221 [IFLA_GRE_FWMARK] = { .type = NLA_U32 }, 2222 [IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 }, 2223 [IFLA_GRE_ERSPAN_VER] = { .type = NLA_U8 }, 2224 [IFLA_GRE_ERSPAN_DIR] = { .type = NLA_U8 }, 2225 [IFLA_GRE_ERSPAN_HWID] = { .type = NLA_U16 }, 2226 }; 2227 2228 static void ip6erspan_tap_setup(struct net_device *dev) 2229 { 2230 ether_setup(dev); 2231 2232 dev->max_mtu = 0; 2233 dev->netdev_ops = &ip6erspan_netdev_ops; 2234 dev->needs_free_netdev = true; 2235 dev->priv_destructor = ip6gre_dev_free; 2236 2237 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 2238 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 2239 netif_keep_dst(dev); 2240 } 2241 2242 static int ip6erspan_newlink(struct net *src_net, struct net_device *dev, 2243 struct nlattr *tb[], struct nlattr *data[], 2244 struct netlink_ext_ack *extack) 2245 { 2246 struct ip6_tnl *nt = netdev_priv(dev); 2247 struct net *net = dev_net(dev); 2248 struct ip6gre_net *ign; 2249 int err; 2250 2251 ip6gre_netlink_parms(data, &nt->parms); 2252 ip6erspan_set_version(data, &nt->parms); 2253 ign = net_generic(net, ip6gre_net_id); 2254 2255 if (nt->parms.collect_md) { 2256 if (rtnl_dereference(ign->collect_md_tun_erspan)) 2257 return -EEXIST; 2258 } else { 2259 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 2260 return -EEXIST; 2261 } 2262 2263 err = ip6gre_newlink_common(src_net, dev, tb, data, extack); 2264 if (!err) { 2265 ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]); 2266 ip6erspan_tunnel_link_md(ign, nt); 2267 ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt); 2268 } 2269 return err; 2270 } 2271 2272 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu) 2273 { 2274 ip6gre_tnl_link_config_common(t); 2275 ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t)); 2276 } 2277 2278 static int ip6erspan_tnl_change(struct ip6_tnl *t, 2279 const struct __ip6_tnl_parm *p, int set_mtu) 2280 { 2281 ip6gre_tnl_copy_tnl_parm(t, p); 2282 ip6erspan_tnl_link_config(t, set_mtu); 2283 return 0; 2284 } 2285 2286 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[], 2287 struct nlattr *data[], 2288 struct netlink_ext_ack *extack) 2289 { 2290 struct ip6gre_net *ign = net_generic(dev_net(dev), ip6gre_net_id); 2291 struct __ip6_tnl_parm p; 2292 struct ip6_tnl *t; 2293 2294 t = ip6gre_changelink_common(dev, tb, data, &p, extack); 2295 if (IS_ERR(t)) 2296 return PTR_ERR(t); 2297 2298 ip6erspan_set_version(data, &p); 2299 ip6gre_tunnel_unlink_md(ign, t); 2300 ip6gre_tunnel_unlink(ign, t); 2301 ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]); 2302 ip6erspan_tunnel_link_md(ign, t); 2303 ip6gre_tunnel_link(ign, t); 2304 return 0; 2305 } 2306 2307 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 2308 .kind = "ip6gre", 2309 .maxtype = IFLA_GRE_MAX, 2310 .policy = ip6gre_policy, 2311 .priv_size = sizeof(struct ip6_tnl), 2312 .setup = ip6gre_tunnel_setup, 2313 .validate = ip6gre_tunnel_validate, 2314 .newlink = ip6gre_newlink, 2315 .changelink = ip6gre_changelink, 2316 .dellink = ip6gre_dellink, 2317 .get_size = ip6gre_get_size, 2318 .fill_info = ip6gre_fill_info, 2319 .get_link_net = ip6_tnl_get_link_net, 2320 }; 2321 2322 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 2323 .kind = "ip6gretap", 2324 .maxtype = IFLA_GRE_MAX, 2325 .policy = ip6gre_policy, 2326 .priv_size = sizeof(struct ip6_tnl), 2327 .setup = ip6gre_tap_setup, 2328 .validate = ip6gre_tap_validate, 2329 .newlink = ip6gre_newlink, 2330 .changelink = ip6gre_changelink, 2331 .get_size = ip6gre_get_size, 2332 .fill_info = ip6gre_fill_info, 2333 .get_link_net = ip6_tnl_get_link_net, 2334 }; 2335 2336 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = { 2337 .kind = "ip6erspan", 2338 .maxtype = IFLA_GRE_MAX, 2339 .policy = ip6gre_policy, 2340 .priv_size = sizeof(struct ip6_tnl), 2341 .setup = ip6erspan_tap_setup, 2342 .validate = ip6erspan_tap_validate, 2343 .newlink = ip6erspan_newlink, 2344 .changelink = ip6erspan_changelink, 2345 .get_size = ip6gre_get_size, 2346 .fill_info = ip6gre_fill_info, 2347 .get_link_net = ip6_tnl_get_link_net, 2348 }; 2349 2350 /* 2351 * And now the modules code and kernel interface. 2352 */ 2353 2354 static int __init ip6gre_init(void) 2355 { 2356 int err; 2357 2358 pr_info("GRE over IPv6 tunneling driver\n"); 2359 2360 err = register_pernet_device(&ip6gre_net_ops); 2361 if (err < 0) 2362 return err; 2363 2364 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 2365 if (err < 0) { 2366 pr_info("%s: can't add protocol\n", __func__); 2367 goto add_proto_failed; 2368 } 2369 2370 err = rtnl_link_register(&ip6gre_link_ops); 2371 if (err < 0) 2372 goto rtnl_link_failed; 2373 2374 err = rtnl_link_register(&ip6gre_tap_ops); 2375 if (err < 0) 2376 goto tap_ops_failed; 2377 2378 err = rtnl_link_register(&ip6erspan_tap_ops); 2379 if (err < 0) 2380 goto erspan_link_failed; 2381 2382 out: 2383 return err; 2384 2385 erspan_link_failed: 2386 rtnl_link_unregister(&ip6gre_tap_ops); 2387 tap_ops_failed: 2388 rtnl_link_unregister(&ip6gre_link_ops); 2389 rtnl_link_failed: 2390 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2391 add_proto_failed: 2392 unregister_pernet_device(&ip6gre_net_ops); 2393 goto out; 2394 } 2395 2396 static void __exit ip6gre_fini(void) 2397 { 2398 rtnl_link_unregister(&ip6gre_tap_ops); 2399 rtnl_link_unregister(&ip6gre_link_ops); 2400 rtnl_link_unregister(&ip6erspan_tap_ops); 2401 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 2402 unregister_pernet_device(&ip6gre_net_ops); 2403 } 2404 2405 module_init(ip6gre_init); 2406 module_exit(ip6gre_fini); 2407 MODULE_LICENSE("GPL"); 2408 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 2409 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 2410 MODULE_ALIAS_RTNL_LINK("ip6gre"); 2411 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 2412 MODULE_ALIAS_RTNL_LINK("ip6erspan"); 2413 MODULE_ALIAS_NETDEV("ip6gre0"); 2414