1 /* 2 * GRE over IPv6 protocol decoder. 3 * 4 * Authors: Dmitry Kozlov (xeb@mail.ru) 5 * 6 * This program is free software; you can redistribute it and/or 7 * modify it under the terms of the GNU General Public License 8 * as published by the Free Software Foundation; either version 9 * 2 of the License, or (at your option) any later version. 10 * 11 */ 12 13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 14 15 #include <linux/capability.h> 16 #include <linux/module.h> 17 #include <linux/types.h> 18 #include <linux/kernel.h> 19 #include <linux/slab.h> 20 #include <linux/uaccess.h> 21 #include <linux/skbuff.h> 22 #include <linux/netdevice.h> 23 #include <linux/in.h> 24 #include <linux/tcp.h> 25 #include <linux/udp.h> 26 #include <linux/if_arp.h> 27 #include <linux/init.h> 28 #include <linux/in6.h> 29 #include <linux/inetdevice.h> 30 #include <linux/igmp.h> 31 #include <linux/netfilter_ipv4.h> 32 #include <linux/etherdevice.h> 33 #include <linux/if_ether.h> 34 #include <linux/hash.h> 35 #include <linux/if_tunnel.h> 36 #include <linux/ip6_tunnel.h> 37 38 #include <net/sock.h> 39 #include <net/ip.h> 40 #include <net/ip_tunnels.h> 41 #include <net/icmp.h> 42 #include <net/protocol.h> 43 #include <net/addrconf.h> 44 #include <net/arp.h> 45 #include <net/checksum.h> 46 #include <net/dsfield.h> 47 #include <net/inet_ecn.h> 48 #include <net/xfrm.h> 49 #include <net/net_namespace.h> 50 #include <net/netns/generic.h> 51 #include <net/rtnetlink.h> 52 53 #include <net/ipv6.h> 54 #include <net/ip6_fib.h> 55 #include <net/ip6_route.h> 56 #include <net/ip6_tunnel.h> 57 #include <net/gre.h> 58 59 60 static bool log_ecn_error = true; 61 module_param(log_ecn_error, bool, 0644); 62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); 63 64 #define IP6_GRE_HASH_SIZE_SHIFT 5 65 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT) 66 67 static int ip6gre_net_id __read_mostly; 68 struct ip6gre_net { 69 struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE]; 70 71 struct net_device *fb_tunnel_dev; 72 }; 73 74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly; 75 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly; 76 static 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 81 /* Tunnel hash table */ 82 83 /* 84 4 hash tables: 85 86 3: (remote,local) 87 2: (remote,*) 88 1: (*,local) 89 0: (*,*) 90 91 We require exact key match i.e. if a key is present in packet 92 it will match only tunnel with the same key; if it is not present, 93 it will match only keyless tunnel. 94 95 All keysless packets, if not matched configured keyless tunnels 96 will match fallback tunnel. 97 */ 98 99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1)) 100 static u32 HASH_ADDR(const struct in6_addr *addr) 101 { 102 u32 hash = ipv6_addr_hash(addr); 103 104 return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT); 105 } 106 107 #define tunnels_r_l tunnels[3] 108 #define tunnels_r tunnels[2] 109 #define tunnels_l tunnels[1] 110 #define tunnels_wc tunnels[0] 111 112 /* Given src, dst and key, find appropriate for input tunnel. */ 113 114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev, 115 const struct in6_addr *remote, const struct in6_addr *local, 116 __be32 key, __be16 gre_proto) 117 { 118 struct net *net = dev_net(dev); 119 int link = dev->ifindex; 120 unsigned int h0 = HASH_ADDR(remote); 121 unsigned int h1 = HASH_KEY(key); 122 struct ip6_tnl *t, *cand = NULL; 123 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 124 int dev_type = (gre_proto == htons(ETH_P_TEB)) ? 125 ARPHRD_ETHER : ARPHRD_IP6GRE; 126 int score, cand_score = 4; 127 128 for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) { 129 if (!ipv6_addr_equal(local, &t->parms.laddr) || 130 !ipv6_addr_equal(remote, &t->parms.raddr) || 131 key != t->parms.i_key || 132 !(t->dev->flags & IFF_UP)) 133 continue; 134 135 if (t->dev->type != ARPHRD_IP6GRE && 136 t->dev->type != dev_type) 137 continue; 138 139 score = 0; 140 if (t->parms.link != link) 141 score |= 1; 142 if (t->dev->type != dev_type) 143 score |= 2; 144 if (score == 0) 145 return t; 146 147 if (score < cand_score) { 148 cand = t; 149 cand_score = score; 150 } 151 } 152 153 for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) { 154 if (!ipv6_addr_equal(remote, &t->parms.raddr) || 155 key != t->parms.i_key || 156 !(t->dev->flags & IFF_UP)) 157 continue; 158 159 if (t->dev->type != ARPHRD_IP6GRE && 160 t->dev->type != dev_type) 161 continue; 162 163 score = 0; 164 if (t->parms.link != link) 165 score |= 1; 166 if (t->dev->type != dev_type) 167 score |= 2; 168 if (score == 0) 169 return t; 170 171 if (score < cand_score) { 172 cand = t; 173 cand_score = score; 174 } 175 } 176 177 for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) { 178 if ((!ipv6_addr_equal(local, &t->parms.laddr) && 179 (!ipv6_addr_equal(local, &t->parms.raddr) || 180 !ipv6_addr_is_multicast(local))) || 181 key != t->parms.i_key || 182 !(t->dev->flags & IFF_UP)) 183 continue; 184 185 if (t->dev->type != ARPHRD_IP6GRE && 186 t->dev->type != dev_type) 187 continue; 188 189 score = 0; 190 if (t->parms.link != link) 191 score |= 1; 192 if (t->dev->type != dev_type) 193 score |= 2; 194 if (score == 0) 195 return t; 196 197 if (score < cand_score) { 198 cand = t; 199 cand_score = score; 200 } 201 } 202 203 for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) { 204 if (t->parms.i_key != key || 205 !(t->dev->flags & IFF_UP)) 206 continue; 207 208 if (t->dev->type != ARPHRD_IP6GRE && 209 t->dev->type != dev_type) 210 continue; 211 212 score = 0; 213 if (t->parms.link != link) 214 score |= 1; 215 if (t->dev->type != dev_type) 216 score |= 2; 217 if (score == 0) 218 return t; 219 220 if (score < cand_score) { 221 cand = t; 222 cand_score = score; 223 } 224 } 225 226 if (cand) 227 return cand; 228 229 dev = ign->fb_tunnel_dev; 230 if (dev->flags & IFF_UP) 231 return netdev_priv(dev); 232 233 return NULL; 234 } 235 236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign, 237 const struct __ip6_tnl_parm *p) 238 { 239 const struct in6_addr *remote = &p->raddr; 240 const struct in6_addr *local = &p->laddr; 241 unsigned int h = HASH_KEY(p->i_key); 242 int prio = 0; 243 244 if (!ipv6_addr_any(local)) 245 prio |= 1; 246 if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) { 247 prio |= 2; 248 h ^= HASH_ADDR(remote); 249 } 250 251 return &ign->tunnels[prio][h]; 252 } 253 254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign, 255 const struct ip6_tnl *t) 256 { 257 return __ip6gre_bucket(ign, &t->parms); 258 } 259 260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t) 261 { 262 struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t); 263 264 rcu_assign_pointer(t->next, rtnl_dereference(*tp)); 265 rcu_assign_pointer(*tp, t); 266 } 267 268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t) 269 { 270 struct ip6_tnl __rcu **tp; 271 struct ip6_tnl *iter; 272 273 for (tp = ip6gre_bucket(ign, t); 274 (iter = rtnl_dereference(*tp)) != NULL; 275 tp = &iter->next) { 276 if (t == iter) { 277 rcu_assign_pointer(*tp, t->next); 278 break; 279 } 280 } 281 } 282 283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net, 284 const struct __ip6_tnl_parm *parms, 285 int type) 286 { 287 const struct in6_addr *remote = &parms->raddr; 288 const struct in6_addr *local = &parms->laddr; 289 __be32 key = parms->i_key; 290 int link = parms->link; 291 struct ip6_tnl *t; 292 struct ip6_tnl __rcu **tp; 293 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 294 295 for (tp = __ip6gre_bucket(ign, parms); 296 (t = rtnl_dereference(*tp)) != NULL; 297 tp = &t->next) 298 if (ipv6_addr_equal(local, &t->parms.laddr) && 299 ipv6_addr_equal(remote, &t->parms.raddr) && 300 key == t->parms.i_key && 301 link == t->parms.link && 302 type == t->dev->type) 303 break; 304 305 return t; 306 } 307 308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net, 309 const struct __ip6_tnl_parm *parms, int create) 310 { 311 struct ip6_tnl *t, *nt; 312 struct net_device *dev; 313 char name[IFNAMSIZ]; 314 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 315 316 t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE); 317 if (t && create) 318 return NULL; 319 if (t || !create) 320 return t; 321 322 if (parms->name[0]) 323 strlcpy(name, parms->name, IFNAMSIZ); 324 else 325 strcpy(name, "ip6gre%d"); 326 327 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, 328 ip6gre_tunnel_setup); 329 if (!dev) 330 return NULL; 331 332 dev_net_set(dev, net); 333 334 nt = netdev_priv(dev); 335 nt->parms = *parms; 336 dev->rtnl_link_ops = &ip6gre_link_ops; 337 338 nt->dev = dev; 339 nt->net = dev_net(dev); 340 ip6gre_tnl_link_config(nt, 1); 341 342 if (register_netdevice(dev) < 0) 343 goto failed_free; 344 345 /* Can use a lockless transmit, unless we generate output sequences */ 346 if (!(nt->parms.o_flags & TUNNEL_SEQ)) 347 dev->features |= NETIF_F_LLTX; 348 349 dev_hold(dev); 350 ip6gre_tunnel_link(ign, nt); 351 return nt; 352 353 failed_free: 354 free_netdev(dev); 355 return NULL; 356 } 357 358 static void ip6gre_tunnel_uninit(struct net_device *dev) 359 { 360 struct ip6_tnl *t = netdev_priv(dev); 361 struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id); 362 363 ip6gre_tunnel_unlink(ign, t); 364 dst_cache_reset(&t->dst_cache); 365 dev_put(dev); 366 } 367 368 369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 370 u8 type, u8 code, int offset, __be32 info) 371 { 372 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data; 373 __be16 *p = (__be16 *)(skb->data + offset); 374 int grehlen = offset + 4; 375 struct ip6_tnl *t; 376 __be16 flags; 377 378 flags = p[0]; 379 if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) { 380 if (flags&(GRE_VERSION|GRE_ROUTING)) 381 return; 382 if (flags&GRE_KEY) { 383 grehlen += 4; 384 if (flags&GRE_CSUM) 385 grehlen += 4; 386 } 387 } 388 389 /* If only 8 bytes returned, keyed message will be dropped here */ 390 if (!pskb_may_pull(skb, grehlen)) 391 return; 392 ipv6h = (const struct ipv6hdr *)skb->data; 393 p = (__be16 *)(skb->data + offset); 394 395 t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr, 396 flags & GRE_KEY ? 397 *(((__be32 *)p) + (grehlen / 4) - 1) : 0, 398 p[1]); 399 if (!t) 400 return; 401 402 switch (type) { 403 __u32 teli; 404 struct ipv6_tlv_tnl_enc_lim *tel; 405 __u32 mtu; 406 case ICMPV6_DEST_UNREACH: 407 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n", 408 t->parms.name); 409 break; 410 case ICMPV6_TIME_EXCEED: 411 if (code == ICMPV6_EXC_HOPLIMIT) { 412 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n", 413 t->parms.name); 414 } 415 break; 416 case ICMPV6_PARAMPROB: 417 teli = 0; 418 if (code == ICMPV6_HDR_FIELD) 419 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data); 420 421 if (teli && teli == be32_to_cpu(info) - 2) { 422 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli]; 423 if (tel->encap_limit == 0) { 424 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n", 425 t->parms.name); 426 } 427 } else { 428 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n", 429 t->parms.name); 430 } 431 break; 432 case ICMPV6_PKT_TOOBIG: 433 mtu = be32_to_cpu(info) - offset; 434 if (mtu < IPV6_MIN_MTU) 435 mtu = IPV6_MIN_MTU; 436 t->dev->mtu = mtu; 437 break; 438 } 439 440 if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO)) 441 t->err_count++; 442 else 443 t->err_count = 1; 444 t->err_time = jiffies; 445 } 446 447 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi) 448 { 449 const struct ipv6hdr *ipv6h; 450 struct ip6_tnl *tunnel; 451 452 ipv6h = ipv6_hdr(skb); 453 tunnel = ip6gre_tunnel_lookup(skb->dev, 454 &ipv6h->saddr, &ipv6h->daddr, tpi->key, 455 tpi->proto); 456 if (tunnel) { 457 ip6_tnl_rcv(tunnel, skb, tpi, NULL, false); 458 459 return PACKET_RCVD; 460 } 461 462 return PACKET_REJECT; 463 } 464 465 static int gre_rcv(struct sk_buff *skb) 466 { 467 struct tnl_ptk_info tpi; 468 bool csum_err = false; 469 int hdr_len; 470 471 hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0); 472 if (hdr_len < 0) 473 goto drop; 474 475 if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false)) 476 goto drop; 477 478 if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD) 479 return 0; 480 481 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); 482 drop: 483 kfree_skb(skb); 484 return 0; 485 } 486 487 struct ipv6_tel_txoption { 488 struct ipv6_txoptions ops; 489 __u8 dst_opt[8]; 490 }; 491 492 static int gre_handle_offloads(struct sk_buff *skb, bool csum) 493 { 494 return iptunnel_handle_offloads(skb, 495 csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE); 496 } 497 498 static netdev_tx_t __gre6_xmit(struct sk_buff *skb, 499 struct net_device *dev, __u8 dsfield, 500 struct flowi6 *fl6, int encap_limit, 501 __u32 *pmtu, __be16 proto) 502 { 503 struct ip6_tnl *tunnel = netdev_priv(dev); 504 __be16 protocol = (dev->type == ARPHRD_ETHER) ? 505 htons(ETH_P_TEB) : proto; 506 507 if (dev->type == ARPHRD_ETHER) 508 IPCB(skb)->flags = 0; 509 510 if (dev->header_ops && dev->type == ARPHRD_IP6GRE) 511 fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr; 512 else 513 fl6->daddr = tunnel->parms.raddr; 514 515 if (tunnel->parms.o_flags & TUNNEL_SEQ) 516 tunnel->o_seqno++; 517 518 /* Push GRE header. */ 519 gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags, 520 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno)); 521 522 return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu, 523 NEXTHDR_GRE); 524 } 525 526 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev) 527 { 528 struct ip6_tnl *t = netdev_priv(dev); 529 const struct iphdr *iph = ip_hdr(skb); 530 int encap_limit = -1; 531 struct flowi6 fl6; 532 __u8 dsfield; 533 __u32 mtu; 534 int err; 535 536 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); 537 538 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 539 encap_limit = t->parms.encap_limit; 540 541 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 542 543 dsfield = ipv4_get_dsfield(iph); 544 545 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 546 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT) 547 & IPV6_TCLASS_MASK; 548 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 549 fl6.flowi6_mark = skb->mark; 550 551 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 552 if (err) 553 return -1; 554 555 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 556 skb->protocol); 557 if (err != 0) { 558 /* XXX: send ICMP error even if DF is not set. */ 559 if (err == -EMSGSIZE) 560 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 561 htonl(mtu)); 562 return -1; 563 } 564 565 return 0; 566 } 567 568 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 569 { 570 struct ip6_tnl *t = netdev_priv(dev); 571 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 572 int encap_limit = -1; 573 __u16 offset; 574 struct flowi6 fl6; 575 __u8 dsfield; 576 __u32 mtu; 577 int err; 578 579 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 580 return -1; 581 582 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 583 if (offset > 0) { 584 struct ipv6_tlv_tnl_enc_lim *tel; 585 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 586 if (tel->encap_limit == 0) { 587 icmpv6_send(skb, ICMPV6_PARAMPROB, 588 ICMPV6_HDR_FIELD, offset + 2); 589 return -1; 590 } 591 encap_limit = tel->encap_limit - 1; 592 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 593 encap_limit = t->parms.encap_limit; 594 595 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 596 597 dsfield = ipv6_get_dsfield(ipv6h); 598 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 599 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 600 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 601 fl6.flowlabel |= ip6_flowlabel(ipv6h); 602 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 603 fl6.flowi6_mark = skb->mark; 604 605 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 606 return -1; 607 608 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 609 &mtu, skb->protocol); 610 if (err != 0) { 611 if (err == -EMSGSIZE) 612 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 613 return -1; 614 } 615 616 return 0; 617 } 618 619 /** 620 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 621 * @t: the outgoing tunnel device 622 * @hdr: IPv6 header from the incoming packet 623 * 624 * Description: 625 * Avoid trivial tunneling loop by checking that tunnel exit-point 626 * doesn't match source of incoming packet. 627 * 628 * Return: 629 * 1 if conflict, 630 * 0 else 631 **/ 632 633 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 634 const struct ipv6hdr *hdr) 635 { 636 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 637 } 638 639 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 640 { 641 struct ip6_tnl *t = netdev_priv(dev); 642 int encap_limit = -1; 643 struct flowi6 fl6; 644 __u32 mtu; 645 int err; 646 647 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 648 encap_limit = t->parms.encap_limit; 649 650 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 651 652 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 653 if (err) 654 return err; 655 656 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 657 658 return err; 659 } 660 661 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 662 struct net_device *dev) 663 { 664 struct ip6_tnl *t = netdev_priv(dev); 665 struct net_device_stats *stats = &t->dev->stats; 666 int ret; 667 668 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 669 goto tx_err; 670 671 switch (skb->protocol) { 672 case htons(ETH_P_IP): 673 ret = ip6gre_xmit_ipv4(skb, dev); 674 break; 675 case htons(ETH_P_IPV6): 676 ret = ip6gre_xmit_ipv6(skb, dev); 677 break; 678 default: 679 ret = ip6gre_xmit_other(skb, dev); 680 break; 681 } 682 683 if (ret < 0) 684 goto tx_err; 685 686 return NETDEV_TX_OK; 687 688 tx_err: 689 stats->tx_errors++; 690 stats->tx_dropped++; 691 kfree_skb(skb); 692 return NETDEV_TX_OK; 693 } 694 695 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 696 { 697 struct net_device *dev = t->dev; 698 struct __ip6_tnl_parm *p = &t->parms; 699 struct flowi6 *fl6 = &t->fl.u.ip6; 700 int t_hlen; 701 702 if (dev->type != ARPHRD_ETHER) { 703 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 704 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 705 } 706 707 /* Set up flowi template */ 708 fl6->saddr = p->laddr; 709 fl6->daddr = p->raddr; 710 fl6->flowi6_oif = p->link; 711 fl6->flowlabel = 0; 712 fl6->flowi6_proto = IPPROTO_GRE; 713 714 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 715 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 716 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 717 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 718 719 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 720 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 721 722 if (p->flags&IP6_TNL_F_CAP_XMIT && 723 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 724 dev->flags |= IFF_POINTOPOINT; 725 else 726 dev->flags &= ~IFF_POINTOPOINT; 727 728 t->tun_hlen = gre_calc_hlen(t->parms.o_flags); 729 730 t->hlen = t->encap_hlen + t->tun_hlen; 731 732 t_hlen = t->hlen + sizeof(struct ipv6hdr); 733 734 if (p->flags & IP6_TNL_F_CAP_XMIT) { 735 int strict = (ipv6_addr_type(&p->raddr) & 736 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 737 738 struct rt6_info *rt = rt6_lookup(t->net, 739 &p->raddr, &p->laddr, 740 p->link, strict); 741 742 if (!rt) 743 return; 744 745 if (rt->dst.dev) { 746 dev->hard_header_len = rt->dst.dev->hard_header_len + 747 t_hlen; 748 749 if (set_mtu) { 750 dev->mtu = rt->dst.dev->mtu - t_hlen; 751 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 752 dev->mtu -= 8; 753 if (dev->type == ARPHRD_ETHER) 754 dev->mtu -= ETH_HLEN; 755 756 if (dev->mtu < IPV6_MIN_MTU) 757 dev->mtu = IPV6_MIN_MTU; 758 } 759 } 760 ip6_rt_put(rt); 761 } 762 } 763 764 static int ip6gre_tnl_change(struct ip6_tnl *t, 765 const struct __ip6_tnl_parm *p, int set_mtu) 766 { 767 t->parms.laddr = p->laddr; 768 t->parms.raddr = p->raddr; 769 t->parms.flags = p->flags; 770 t->parms.hop_limit = p->hop_limit; 771 t->parms.encap_limit = p->encap_limit; 772 t->parms.flowinfo = p->flowinfo; 773 t->parms.link = p->link; 774 t->parms.proto = p->proto; 775 t->parms.i_key = p->i_key; 776 t->parms.o_key = p->o_key; 777 t->parms.i_flags = p->i_flags; 778 t->parms.o_flags = p->o_flags; 779 dst_cache_reset(&t->dst_cache); 780 ip6gre_tnl_link_config(t, set_mtu); 781 return 0; 782 } 783 784 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 785 const struct ip6_tnl_parm2 *u) 786 { 787 p->laddr = u->laddr; 788 p->raddr = u->raddr; 789 p->flags = u->flags; 790 p->hop_limit = u->hop_limit; 791 p->encap_limit = u->encap_limit; 792 p->flowinfo = u->flowinfo; 793 p->link = u->link; 794 p->i_key = u->i_key; 795 p->o_key = u->o_key; 796 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 797 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 798 memcpy(p->name, u->name, sizeof(u->name)); 799 } 800 801 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 802 const struct __ip6_tnl_parm *p) 803 { 804 u->proto = IPPROTO_GRE; 805 u->laddr = p->laddr; 806 u->raddr = p->raddr; 807 u->flags = p->flags; 808 u->hop_limit = p->hop_limit; 809 u->encap_limit = p->encap_limit; 810 u->flowinfo = p->flowinfo; 811 u->link = p->link; 812 u->i_key = p->i_key; 813 u->o_key = p->o_key; 814 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 815 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 816 memcpy(u->name, p->name, sizeof(u->name)); 817 } 818 819 static int ip6gre_tunnel_ioctl(struct net_device *dev, 820 struct ifreq *ifr, int cmd) 821 { 822 int err = 0; 823 struct ip6_tnl_parm2 p; 824 struct __ip6_tnl_parm p1; 825 struct ip6_tnl *t = netdev_priv(dev); 826 struct net *net = t->net; 827 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 828 829 memset(&p1, 0, sizeof(p1)); 830 831 switch (cmd) { 832 case SIOCGETTUNNEL: 833 if (dev == ign->fb_tunnel_dev) { 834 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 835 err = -EFAULT; 836 break; 837 } 838 ip6gre_tnl_parm_from_user(&p1, &p); 839 t = ip6gre_tunnel_locate(net, &p1, 0); 840 if (!t) 841 t = netdev_priv(dev); 842 } 843 memset(&p, 0, sizeof(p)); 844 ip6gre_tnl_parm_to_user(&p, &t->parms); 845 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 846 err = -EFAULT; 847 break; 848 849 case SIOCADDTUNNEL: 850 case SIOCCHGTUNNEL: 851 err = -EPERM; 852 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 853 goto done; 854 855 err = -EFAULT; 856 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 857 goto done; 858 859 err = -EINVAL; 860 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 861 goto done; 862 863 if (!(p.i_flags&GRE_KEY)) 864 p.i_key = 0; 865 if (!(p.o_flags&GRE_KEY)) 866 p.o_key = 0; 867 868 ip6gre_tnl_parm_from_user(&p1, &p); 869 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 870 871 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 872 if (t) { 873 if (t->dev != dev) { 874 err = -EEXIST; 875 break; 876 } 877 } else { 878 t = netdev_priv(dev); 879 880 ip6gre_tunnel_unlink(ign, t); 881 synchronize_net(); 882 ip6gre_tnl_change(t, &p1, 1); 883 ip6gre_tunnel_link(ign, t); 884 netdev_state_change(dev); 885 } 886 } 887 888 if (t) { 889 err = 0; 890 891 memset(&p, 0, sizeof(p)); 892 ip6gre_tnl_parm_to_user(&p, &t->parms); 893 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 894 err = -EFAULT; 895 } else 896 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 897 break; 898 899 case SIOCDELTUNNEL: 900 err = -EPERM; 901 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 902 goto done; 903 904 if (dev == ign->fb_tunnel_dev) { 905 err = -EFAULT; 906 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 907 goto done; 908 err = -ENOENT; 909 ip6gre_tnl_parm_from_user(&p1, &p); 910 t = ip6gre_tunnel_locate(net, &p1, 0); 911 if (!t) 912 goto done; 913 err = -EPERM; 914 if (t == netdev_priv(ign->fb_tunnel_dev)) 915 goto done; 916 dev = t->dev; 917 } 918 unregister_netdevice(dev); 919 err = 0; 920 break; 921 922 default: 923 err = -EINVAL; 924 } 925 926 done: 927 return err; 928 } 929 930 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 931 unsigned short type, 932 const void *daddr, const void *saddr, unsigned int len) 933 { 934 struct ip6_tnl *t = netdev_priv(dev); 935 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen); 936 __be16 *p = (__be16 *)(ipv6h+1); 937 938 ip6_flow_hdr(ipv6h, 0, 939 ip6_make_flowlabel(dev_net(dev), skb, 940 t->fl.u.ip6.flowlabel, true, 941 &t->fl.u.ip6)); 942 ipv6h->hop_limit = t->parms.hop_limit; 943 ipv6h->nexthdr = NEXTHDR_GRE; 944 ipv6h->saddr = t->parms.laddr; 945 ipv6h->daddr = t->parms.raddr; 946 947 p[0] = t->parms.o_flags; 948 p[1] = htons(type); 949 950 /* 951 * Set the source hardware address. 952 */ 953 954 if (saddr) 955 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 956 if (daddr) 957 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 958 if (!ipv6_addr_any(&ipv6h->daddr)) 959 return t->hlen; 960 961 return -t->hlen; 962 } 963 964 static const struct header_ops ip6gre_header_ops = { 965 .create = ip6gre_header, 966 }; 967 968 static const struct net_device_ops ip6gre_netdev_ops = { 969 .ndo_init = ip6gre_tunnel_init, 970 .ndo_uninit = ip6gre_tunnel_uninit, 971 .ndo_start_xmit = ip6gre_tunnel_xmit, 972 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 973 .ndo_change_mtu = ip6_tnl_change_mtu, 974 .ndo_get_stats64 = ip_tunnel_get_stats64, 975 .ndo_get_iflink = ip6_tnl_get_iflink, 976 }; 977 978 static void ip6gre_dev_free(struct net_device *dev) 979 { 980 struct ip6_tnl *t = netdev_priv(dev); 981 982 dst_cache_destroy(&t->dst_cache); 983 free_percpu(dev->tstats); 984 free_netdev(dev); 985 } 986 987 static void ip6gre_tunnel_setup(struct net_device *dev) 988 { 989 dev->netdev_ops = &ip6gre_netdev_ops; 990 dev->destructor = ip6gre_dev_free; 991 992 dev->type = ARPHRD_IP6GRE; 993 994 dev->flags |= IFF_NOARP; 995 dev->addr_len = sizeof(struct in6_addr); 996 netif_keep_dst(dev); 997 } 998 999 static int ip6gre_tunnel_init_common(struct net_device *dev) 1000 { 1001 struct ip6_tnl *tunnel; 1002 int ret; 1003 int t_hlen; 1004 1005 tunnel = netdev_priv(dev); 1006 1007 tunnel->dev = dev; 1008 tunnel->net = dev_net(dev); 1009 strcpy(tunnel->parms.name, dev->name); 1010 1011 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1012 if (!dev->tstats) 1013 return -ENOMEM; 1014 1015 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1016 if (ret) { 1017 free_percpu(dev->tstats); 1018 dev->tstats = NULL; 1019 return ret; 1020 } 1021 1022 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1023 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1024 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1025 1026 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1027 dev->mtu = ETH_DATA_LEN - t_hlen; 1028 if (dev->type == ARPHRD_ETHER) 1029 dev->mtu -= ETH_HLEN; 1030 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1031 dev->mtu -= 8; 1032 1033 return 0; 1034 } 1035 1036 static int ip6gre_tunnel_init(struct net_device *dev) 1037 { 1038 struct ip6_tnl *tunnel; 1039 int ret; 1040 1041 ret = ip6gre_tunnel_init_common(dev); 1042 if (ret) 1043 return ret; 1044 1045 tunnel = netdev_priv(dev); 1046 1047 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1048 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1049 1050 if (ipv6_addr_any(&tunnel->parms.raddr)) 1051 dev->header_ops = &ip6gre_header_ops; 1052 1053 return 0; 1054 } 1055 1056 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1057 { 1058 struct ip6_tnl *tunnel = netdev_priv(dev); 1059 1060 tunnel->dev = dev; 1061 tunnel->net = dev_net(dev); 1062 strcpy(tunnel->parms.name, dev->name); 1063 1064 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1065 1066 dev_hold(dev); 1067 } 1068 1069 1070 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1071 .handler = gre_rcv, 1072 .err_handler = ip6gre_err, 1073 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1074 }; 1075 1076 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1077 { 1078 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1079 struct net_device *dev, *aux; 1080 int prio; 1081 1082 for_each_netdev_safe(net, dev, aux) 1083 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1084 dev->rtnl_link_ops == &ip6gre_tap_ops) 1085 unregister_netdevice_queue(dev, head); 1086 1087 for (prio = 0; prio < 4; prio++) { 1088 int h; 1089 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1090 struct ip6_tnl *t; 1091 1092 t = rtnl_dereference(ign->tunnels[prio][h]); 1093 1094 while (t) { 1095 /* If dev is in the same netns, it has already 1096 * been added to the list by the previous loop. 1097 */ 1098 if (!net_eq(dev_net(t->dev), net)) 1099 unregister_netdevice_queue(t->dev, 1100 head); 1101 t = rtnl_dereference(t->next); 1102 } 1103 } 1104 } 1105 } 1106 1107 static int __net_init ip6gre_init_net(struct net *net) 1108 { 1109 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1110 int err; 1111 1112 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1113 NET_NAME_UNKNOWN, 1114 ip6gre_tunnel_setup); 1115 if (!ign->fb_tunnel_dev) { 1116 err = -ENOMEM; 1117 goto err_alloc_dev; 1118 } 1119 dev_net_set(ign->fb_tunnel_dev, net); 1120 /* FB netdevice is special: we have one, and only one per netns. 1121 * Allowing to move it to another netns is clearly unsafe. 1122 */ 1123 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1124 1125 1126 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1127 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1128 1129 err = register_netdev(ign->fb_tunnel_dev); 1130 if (err) 1131 goto err_reg_dev; 1132 1133 rcu_assign_pointer(ign->tunnels_wc[0], 1134 netdev_priv(ign->fb_tunnel_dev)); 1135 return 0; 1136 1137 err_reg_dev: 1138 ip6gre_dev_free(ign->fb_tunnel_dev); 1139 err_alloc_dev: 1140 return err; 1141 } 1142 1143 static void __net_exit ip6gre_exit_net(struct net *net) 1144 { 1145 LIST_HEAD(list); 1146 1147 rtnl_lock(); 1148 ip6gre_destroy_tunnels(net, &list); 1149 unregister_netdevice_many(&list); 1150 rtnl_unlock(); 1151 } 1152 1153 static struct pernet_operations ip6gre_net_ops = { 1154 .init = ip6gre_init_net, 1155 .exit = ip6gre_exit_net, 1156 .id = &ip6gre_net_id, 1157 .size = sizeof(struct ip6gre_net), 1158 }; 1159 1160 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) 1161 { 1162 __be16 flags; 1163 1164 if (!data) 1165 return 0; 1166 1167 flags = 0; 1168 if (data[IFLA_GRE_IFLAGS]) 1169 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1170 if (data[IFLA_GRE_OFLAGS]) 1171 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1172 if (flags & (GRE_VERSION|GRE_ROUTING)) 1173 return -EINVAL; 1174 1175 return 0; 1176 } 1177 1178 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[]) 1179 { 1180 struct in6_addr daddr; 1181 1182 if (tb[IFLA_ADDRESS]) { 1183 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1184 return -EINVAL; 1185 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1186 return -EADDRNOTAVAIL; 1187 } 1188 1189 if (!data) 1190 goto out; 1191 1192 if (data[IFLA_GRE_REMOTE]) { 1193 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1194 if (ipv6_addr_any(&daddr)) 1195 return -EINVAL; 1196 } 1197 1198 out: 1199 return ip6gre_tunnel_validate(tb, data); 1200 } 1201 1202 1203 static void ip6gre_netlink_parms(struct nlattr *data[], 1204 struct __ip6_tnl_parm *parms) 1205 { 1206 memset(parms, 0, sizeof(*parms)); 1207 1208 if (!data) 1209 return; 1210 1211 if (data[IFLA_GRE_LINK]) 1212 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1213 1214 if (data[IFLA_GRE_IFLAGS]) 1215 parms->i_flags = gre_flags_to_tnl_flags( 1216 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1217 1218 if (data[IFLA_GRE_OFLAGS]) 1219 parms->o_flags = gre_flags_to_tnl_flags( 1220 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1221 1222 if (data[IFLA_GRE_IKEY]) 1223 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1224 1225 if (data[IFLA_GRE_OKEY]) 1226 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1227 1228 if (data[IFLA_GRE_LOCAL]) 1229 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1230 1231 if (data[IFLA_GRE_REMOTE]) 1232 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1233 1234 if (data[IFLA_GRE_TTL]) 1235 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1236 1237 if (data[IFLA_GRE_ENCAP_LIMIT]) 1238 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1239 1240 if (data[IFLA_GRE_FLOWINFO]) 1241 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1242 1243 if (data[IFLA_GRE_FLAGS]) 1244 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1245 } 1246 1247 static int ip6gre_tap_init(struct net_device *dev) 1248 { 1249 struct ip6_tnl *tunnel; 1250 int ret; 1251 1252 ret = ip6gre_tunnel_init_common(dev); 1253 if (ret) 1254 return ret; 1255 1256 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1257 1258 tunnel = netdev_priv(dev); 1259 1260 ip6gre_tnl_link_config(tunnel, 1); 1261 1262 return 0; 1263 } 1264 1265 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1266 .ndo_init = ip6gre_tap_init, 1267 .ndo_uninit = ip6gre_tunnel_uninit, 1268 .ndo_start_xmit = ip6gre_tunnel_xmit, 1269 .ndo_set_mac_address = eth_mac_addr, 1270 .ndo_validate_addr = eth_validate_addr, 1271 .ndo_change_mtu = ip6_tnl_change_mtu, 1272 .ndo_get_stats64 = ip_tunnel_get_stats64, 1273 .ndo_get_iflink = ip6_tnl_get_iflink, 1274 }; 1275 1276 #define GRE6_FEATURES (NETIF_F_SG | \ 1277 NETIF_F_FRAGLIST | \ 1278 NETIF_F_HIGHDMA | \ 1279 NETIF_F_HW_CSUM) 1280 1281 static void ip6gre_tap_setup(struct net_device *dev) 1282 { 1283 1284 ether_setup(dev); 1285 1286 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1287 dev->destructor = ip6gre_dev_free; 1288 1289 dev->features |= NETIF_F_NETNS_LOCAL; 1290 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1291 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1292 } 1293 1294 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1295 struct ip_tunnel_encap *ipencap) 1296 { 1297 bool ret = false; 1298 1299 memset(ipencap, 0, sizeof(*ipencap)); 1300 1301 if (!data) 1302 return ret; 1303 1304 if (data[IFLA_GRE_ENCAP_TYPE]) { 1305 ret = true; 1306 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1307 } 1308 1309 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1310 ret = true; 1311 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1312 } 1313 1314 if (data[IFLA_GRE_ENCAP_SPORT]) { 1315 ret = true; 1316 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1317 } 1318 1319 if (data[IFLA_GRE_ENCAP_DPORT]) { 1320 ret = true; 1321 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1322 } 1323 1324 return ret; 1325 } 1326 1327 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1328 struct nlattr *tb[], struct nlattr *data[]) 1329 { 1330 struct ip6_tnl *nt; 1331 struct net *net = dev_net(dev); 1332 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1333 struct ip_tunnel_encap ipencap; 1334 int err; 1335 1336 nt = netdev_priv(dev); 1337 1338 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1339 int err = ip6_tnl_encap_setup(nt, &ipencap); 1340 1341 if (err < 0) 1342 return err; 1343 } 1344 1345 ip6gre_netlink_parms(data, &nt->parms); 1346 1347 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1348 return -EEXIST; 1349 1350 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1351 eth_hw_addr_random(dev); 1352 1353 nt->dev = dev; 1354 nt->net = dev_net(dev); 1355 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1356 1357 dev->features |= GRE6_FEATURES; 1358 dev->hw_features |= GRE6_FEATURES; 1359 1360 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1361 /* TCP offload with GRE SEQ is not supported, nor 1362 * can we support 2 levels of outer headers requiring 1363 * an update. 1364 */ 1365 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1366 (nt->encap.type == TUNNEL_ENCAP_NONE)) { 1367 dev->features |= NETIF_F_GSO_SOFTWARE; 1368 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1369 } 1370 1371 /* Can use a lockless transmit, unless we generate 1372 * output sequences 1373 */ 1374 dev->features |= NETIF_F_LLTX; 1375 } 1376 1377 err = register_netdevice(dev); 1378 if (err) 1379 goto out; 1380 1381 dev_hold(dev); 1382 ip6gre_tunnel_link(ign, nt); 1383 1384 out: 1385 return err; 1386 } 1387 1388 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 1389 struct nlattr *data[]) 1390 { 1391 struct ip6_tnl *t, *nt = netdev_priv(dev); 1392 struct net *net = nt->net; 1393 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1394 struct __ip6_tnl_parm p; 1395 struct ip_tunnel_encap ipencap; 1396 1397 if (dev == ign->fb_tunnel_dev) 1398 return -EINVAL; 1399 1400 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1401 int err = ip6_tnl_encap_setup(nt, &ipencap); 1402 1403 if (err < 0) 1404 return err; 1405 } 1406 1407 ip6gre_netlink_parms(data, &p); 1408 1409 t = ip6gre_tunnel_locate(net, &p, 0); 1410 1411 if (t) { 1412 if (t->dev != dev) 1413 return -EEXIST; 1414 } else { 1415 t = nt; 1416 } 1417 1418 ip6gre_tunnel_unlink(ign, t); 1419 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 1420 ip6gre_tunnel_link(ign, t); 1421 return 0; 1422 } 1423 1424 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 1425 { 1426 struct net *net = dev_net(dev); 1427 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1428 1429 if (dev != ign->fb_tunnel_dev) 1430 unregister_netdevice_queue(dev, head); 1431 } 1432 1433 static size_t ip6gre_get_size(const struct net_device *dev) 1434 { 1435 return 1436 /* IFLA_GRE_LINK */ 1437 nla_total_size(4) + 1438 /* IFLA_GRE_IFLAGS */ 1439 nla_total_size(2) + 1440 /* IFLA_GRE_OFLAGS */ 1441 nla_total_size(2) + 1442 /* IFLA_GRE_IKEY */ 1443 nla_total_size(4) + 1444 /* IFLA_GRE_OKEY */ 1445 nla_total_size(4) + 1446 /* IFLA_GRE_LOCAL */ 1447 nla_total_size(sizeof(struct in6_addr)) + 1448 /* IFLA_GRE_REMOTE */ 1449 nla_total_size(sizeof(struct in6_addr)) + 1450 /* IFLA_GRE_TTL */ 1451 nla_total_size(1) + 1452 /* IFLA_GRE_ENCAP_LIMIT */ 1453 nla_total_size(1) + 1454 /* IFLA_GRE_FLOWINFO */ 1455 nla_total_size(4) + 1456 /* IFLA_GRE_FLAGS */ 1457 nla_total_size(4) + 1458 /* IFLA_GRE_ENCAP_TYPE */ 1459 nla_total_size(2) + 1460 /* IFLA_GRE_ENCAP_FLAGS */ 1461 nla_total_size(2) + 1462 /* IFLA_GRE_ENCAP_SPORT */ 1463 nla_total_size(2) + 1464 /* IFLA_GRE_ENCAP_DPORT */ 1465 nla_total_size(2) + 1466 0; 1467 } 1468 1469 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 1470 { 1471 struct ip6_tnl *t = netdev_priv(dev); 1472 struct __ip6_tnl_parm *p = &t->parms; 1473 1474 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 1475 nla_put_be16(skb, IFLA_GRE_IFLAGS, 1476 gre_tnl_flags_to_gre_flags(p->i_flags)) || 1477 nla_put_be16(skb, IFLA_GRE_OFLAGS, 1478 gre_tnl_flags_to_gre_flags(p->o_flags)) || 1479 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 1480 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 1481 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 1482 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 1483 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 1484 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 1485 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 1486 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags)) 1487 goto nla_put_failure; 1488 1489 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 1490 t->encap.type) || 1491 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 1492 t->encap.sport) || 1493 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 1494 t->encap.dport) || 1495 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 1496 t->encap.flags)) 1497 goto nla_put_failure; 1498 1499 return 0; 1500 1501 nla_put_failure: 1502 return -EMSGSIZE; 1503 } 1504 1505 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 1506 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 1507 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 1508 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 1509 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 1510 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 1511 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 1512 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 1513 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 1514 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 1515 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 1516 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 1517 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 1518 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 1519 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 1520 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 1521 }; 1522 1523 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 1524 .kind = "ip6gre", 1525 .maxtype = IFLA_GRE_MAX, 1526 .policy = ip6gre_policy, 1527 .priv_size = sizeof(struct ip6_tnl), 1528 .setup = ip6gre_tunnel_setup, 1529 .validate = ip6gre_tunnel_validate, 1530 .newlink = ip6gre_newlink, 1531 .changelink = ip6gre_changelink, 1532 .dellink = ip6gre_dellink, 1533 .get_size = ip6gre_get_size, 1534 .fill_info = ip6gre_fill_info, 1535 .get_link_net = ip6_tnl_get_link_net, 1536 }; 1537 1538 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 1539 .kind = "ip6gretap", 1540 .maxtype = IFLA_GRE_MAX, 1541 .policy = ip6gre_policy, 1542 .priv_size = sizeof(struct ip6_tnl), 1543 .setup = ip6gre_tap_setup, 1544 .validate = ip6gre_tap_validate, 1545 .newlink = ip6gre_newlink, 1546 .changelink = ip6gre_changelink, 1547 .get_size = ip6gre_get_size, 1548 .fill_info = ip6gre_fill_info, 1549 .get_link_net = ip6_tnl_get_link_net, 1550 }; 1551 1552 /* 1553 * And now the modules code and kernel interface. 1554 */ 1555 1556 static int __init ip6gre_init(void) 1557 { 1558 int err; 1559 1560 pr_info("GRE over IPv6 tunneling driver\n"); 1561 1562 err = register_pernet_device(&ip6gre_net_ops); 1563 if (err < 0) 1564 return err; 1565 1566 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 1567 if (err < 0) { 1568 pr_info("%s: can't add protocol\n", __func__); 1569 goto add_proto_failed; 1570 } 1571 1572 err = rtnl_link_register(&ip6gre_link_ops); 1573 if (err < 0) 1574 goto rtnl_link_failed; 1575 1576 err = rtnl_link_register(&ip6gre_tap_ops); 1577 if (err < 0) 1578 goto tap_ops_failed; 1579 1580 out: 1581 return err; 1582 1583 tap_ops_failed: 1584 rtnl_link_unregister(&ip6gre_link_ops); 1585 rtnl_link_failed: 1586 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 1587 add_proto_failed: 1588 unregister_pernet_device(&ip6gre_net_ops); 1589 goto out; 1590 } 1591 1592 static void __exit ip6gre_fini(void) 1593 { 1594 rtnl_link_unregister(&ip6gre_tap_ops); 1595 rtnl_link_unregister(&ip6gre_link_ops); 1596 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 1597 unregister_pernet_device(&ip6gre_net_ops); 1598 } 1599 1600 module_init(ip6gre_init); 1601 module_exit(ip6gre_fini); 1602 MODULE_LICENSE("GPL"); 1603 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 1604 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 1605 MODULE_ALIAS_RTNL_LINK("ip6gre"); 1606 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 1607 MODULE_ALIAS_NETDEV("ip6gre0"); 1608