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 HASH_SIZE_SHIFT 5 65 #define HASH_SIZE (1 << HASH_SIZE_SHIFT) 66 67 static int ip6gre_net_id __read_mostly; 68 struct ip6gre_net { 69 struct ip6_tnl __rcu *tunnels[4][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))&(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, 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 fl6.flowi6_proto = skb->protocol; 652 653 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 654 if (err) 655 return err; 656 657 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 658 659 return err; 660 } 661 662 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 663 struct net_device *dev) 664 { 665 struct ip6_tnl *t = netdev_priv(dev); 666 struct net_device_stats *stats = &t->dev->stats; 667 int ret; 668 669 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 670 goto tx_err; 671 672 switch (skb->protocol) { 673 case htons(ETH_P_IP): 674 ret = ip6gre_xmit_ipv4(skb, dev); 675 break; 676 case htons(ETH_P_IPV6): 677 ret = ip6gre_xmit_ipv6(skb, dev); 678 break; 679 default: 680 ret = ip6gre_xmit_other(skb, dev); 681 break; 682 } 683 684 if (ret < 0) 685 goto tx_err; 686 687 return NETDEV_TX_OK; 688 689 tx_err: 690 stats->tx_errors++; 691 stats->tx_dropped++; 692 kfree_skb(skb); 693 return NETDEV_TX_OK; 694 } 695 696 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 697 { 698 struct net_device *dev = t->dev; 699 struct __ip6_tnl_parm *p = &t->parms; 700 struct flowi6 *fl6 = &t->fl.u.ip6; 701 int t_hlen; 702 703 if (dev->type != ARPHRD_ETHER) { 704 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 705 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 706 } 707 708 /* Set up flowi template */ 709 fl6->saddr = p->laddr; 710 fl6->daddr = p->raddr; 711 fl6->flowi6_oif = p->link; 712 fl6->flowlabel = 0; 713 fl6->flowi6_proto = IPPROTO_GRE; 714 715 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 716 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 717 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 718 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 719 720 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 721 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 722 723 if (p->flags&IP6_TNL_F_CAP_XMIT && 724 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 725 dev->flags |= IFF_POINTOPOINT; 726 else 727 dev->flags &= ~IFF_POINTOPOINT; 728 729 t->tun_hlen = gre_calc_hlen(t->parms.o_flags); 730 731 t->hlen = t->encap_hlen + t->tun_hlen; 732 733 t_hlen = t->hlen + sizeof(struct ipv6hdr); 734 735 if (p->flags & IP6_TNL_F_CAP_XMIT) { 736 int strict = (ipv6_addr_type(&p->raddr) & 737 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 738 739 struct rt6_info *rt = rt6_lookup(t->net, 740 &p->raddr, &p->laddr, 741 p->link, strict); 742 743 if (!rt) 744 return; 745 746 if (rt->dst.dev) { 747 dev->hard_header_len = rt->dst.dev->hard_header_len + 748 t_hlen; 749 750 if (set_mtu) { 751 dev->mtu = rt->dst.dev->mtu - t_hlen; 752 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 753 dev->mtu -= 8; 754 if (dev->type == ARPHRD_ETHER) 755 dev->mtu -= ETH_HLEN; 756 757 if (dev->mtu < IPV6_MIN_MTU) 758 dev->mtu = IPV6_MIN_MTU; 759 } 760 } 761 ip6_rt_put(rt); 762 } 763 } 764 765 static int ip6gre_tnl_change(struct ip6_tnl *t, 766 const struct __ip6_tnl_parm *p, int set_mtu) 767 { 768 t->parms.laddr = p->laddr; 769 t->parms.raddr = p->raddr; 770 t->parms.flags = p->flags; 771 t->parms.hop_limit = p->hop_limit; 772 t->parms.encap_limit = p->encap_limit; 773 t->parms.flowinfo = p->flowinfo; 774 t->parms.link = p->link; 775 t->parms.proto = p->proto; 776 t->parms.i_key = p->i_key; 777 t->parms.o_key = p->o_key; 778 t->parms.i_flags = p->i_flags; 779 t->parms.o_flags = p->o_flags; 780 dst_cache_reset(&t->dst_cache); 781 ip6gre_tnl_link_config(t, set_mtu); 782 return 0; 783 } 784 785 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 786 const struct ip6_tnl_parm2 *u) 787 { 788 p->laddr = u->laddr; 789 p->raddr = u->raddr; 790 p->flags = u->flags; 791 p->hop_limit = u->hop_limit; 792 p->encap_limit = u->encap_limit; 793 p->flowinfo = u->flowinfo; 794 p->link = u->link; 795 p->i_key = u->i_key; 796 p->o_key = u->o_key; 797 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 798 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 799 memcpy(p->name, u->name, sizeof(u->name)); 800 } 801 802 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 803 const struct __ip6_tnl_parm *p) 804 { 805 u->proto = IPPROTO_GRE; 806 u->laddr = p->laddr; 807 u->raddr = p->raddr; 808 u->flags = p->flags; 809 u->hop_limit = p->hop_limit; 810 u->encap_limit = p->encap_limit; 811 u->flowinfo = p->flowinfo; 812 u->link = p->link; 813 u->i_key = p->i_key; 814 u->o_key = p->o_key; 815 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 816 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 817 memcpy(u->name, p->name, sizeof(u->name)); 818 } 819 820 static int ip6gre_tunnel_ioctl(struct net_device *dev, 821 struct ifreq *ifr, int cmd) 822 { 823 int err = 0; 824 struct ip6_tnl_parm2 p; 825 struct __ip6_tnl_parm p1; 826 struct ip6_tnl *t = netdev_priv(dev); 827 struct net *net = t->net; 828 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 829 830 memset(&p1, 0, sizeof(p1)); 831 832 switch (cmd) { 833 case SIOCGETTUNNEL: 834 if (dev == ign->fb_tunnel_dev) { 835 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 836 err = -EFAULT; 837 break; 838 } 839 ip6gre_tnl_parm_from_user(&p1, &p); 840 t = ip6gre_tunnel_locate(net, &p1, 0); 841 if (!t) 842 t = netdev_priv(dev); 843 } 844 memset(&p, 0, sizeof(p)); 845 ip6gre_tnl_parm_to_user(&p, &t->parms); 846 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 847 err = -EFAULT; 848 break; 849 850 case SIOCADDTUNNEL: 851 case SIOCCHGTUNNEL: 852 err = -EPERM; 853 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 854 goto done; 855 856 err = -EFAULT; 857 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 858 goto done; 859 860 err = -EINVAL; 861 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 862 goto done; 863 864 if (!(p.i_flags&GRE_KEY)) 865 p.i_key = 0; 866 if (!(p.o_flags&GRE_KEY)) 867 p.o_key = 0; 868 869 ip6gre_tnl_parm_from_user(&p1, &p); 870 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 871 872 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 873 if (t) { 874 if (t->dev != dev) { 875 err = -EEXIST; 876 break; 877 } 878 } else { 879 t = netdev_priv(dev); 880 881 ip6gre_tunnel_unlink(ign, t); 882 synchronize_net(); 883 ip6gre_tnl_change(t, &p1, 1); 884 ip6gre_tunnel_link(ign, t); 885 netdev_state_change(dev); 886 } 887 } 888 889 if (t) { 890 err = 0; 891 892 memset(&p, 0, sizeof(p)); 893 ip6gre_tnl_parm_to_user(&p, &t->parms); 894 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 895 err = -EFAULT; 896 } else 897 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 898 break; 899 900 case SIOCDELTUNNEL: 901 err = -EPERM; 902 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 903 goto done; 904 905 if (dev == ign->fb_tunnel_dev) { 906 err = -EFAULT; 907 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 908 goto done; 909 err = -ENOENT; 910 ip6gre_tnl_parm_from_user(&p1, &p); 911 t = ip6gre_tunnel_locate(net, &p1, 0); 912 if (!t) 913 goto done; 914 err = -EPERM; 915 if (t == netdev_priv(ign->fb_tunnel_dev)) 916 goto done; 917 dev = t->dev; 918 } 919 unregister_netdevice(dev); 920 err = 0; 921 break; 922 923 default: 924 err = -EINVAL; 925 } 926 927 done: 928 return err; 929 } 930 931 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 932 unsigned short type, 933 const void *daddr, const void *saddr, unsigned int len) 934 { 935 struct ip6_tnl *t = netdev_priv(dev); 936 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen); 937 __be16 *p = (__be16 *)(ipv6h+1); 938 939 ip6_flow_hdr(ipv6h, 0, 940 ip6_make_flowlabel(dev_net(dev), skb, 941 t->fl.u.ip6.flowlabel, true, 942 &t->fl.u.ip6)); 943 ipv6h->hop_limit = t->parms.hop_limit; 944 ipv6h->nexthdr = NEXTHDR_GRE; 945 ipv6h->saddr = t->parms.laddr; 946 ipv6h->daddr = t->parms.raddr; 947 948 p[0] = t->parms.o_flags; 949 p[1] = htons(type); 950 951 /* 952 * Set the source hardware address. 953 */ 954 955 if (saddr) 956 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 957 if (daddr) 958 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 959 if (!ipv6_addr_any(&ipv6h->daddr)) 960 return t->hlen; 961 962 return -t->hlen; 963 } 964 965 static const struct header_ops ip6gre_header_ops = { 966 .create = ip6gre_header, 967 }; 968 969 static const struct net_device_ops ip6gre_netdev_ops = { 970 .ndo_init = ip6gre_tunnel_init, 971 .ndo_uninit = ip6gre_tunnel_uninit, 972 .ndo_start_xmit = ip6gre_tunnel_xmit, 973 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 974 .ndo_change_mtu = ip6_tnl_change_mtu, 975 .ndo_get_stats64 = ip_tunnel_get_stats64, 976 .ndo_get_iflink = ip6_tnl_get_iflink, 977 }; 978 979 static void ip6gre_dev_free(struct net_device *dev) 980 { 981 struct ip6_tnl *t = netdev_priv(dev); 982 983 dst_cache_destroy(&t->dst_cache); 984 free_percpu(dev->tstats); 985 free_netdev(dev); 986 } 987 988 static void ip6gre_tunnel_setup(struct net_device *dev) 989 { 990 dev->netdev_ops = &ip6gre_netdev_ops; 991 dev->destructor = ip6gre_dev_free; 992 993 dev->type = ARPHRD_IP6GRE; 994 995 dev->flags |= IFF_NOARP; 996 dev->addr_len = sizeof(struct in6_addr); 997 netif_keep_dst(dev); 998 } 999 1000 static int ip6gre_tunnel_init_common(struct net_device *dev) 1001 { 1002 struct ip6_tnl *tunnel; 1003 int ret; 1004 int t_hlen; 1005 1006 tunnel = netdev_priv(dev); 1007 1008 tunnel->dev = dev; 1009 tunnel->net = dev_net(dev); 1010 strcpy(tunnel->parms.name, dev->name); 1011 1012 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1013 if (!dev->tstats) 1014 return -ENOMEM; 1015 1016 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1017 if (ret) { 1018 free_percpu(dev->tstats); 1019 dev->tstats = NULL; 1020 return ret; 1021 } 1022 1023 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1024 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1025 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1026 1027 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1028 dev->mtu = ETH_DATA_LEN - t_hlen; 1029 if (dev->type == ARPHRD_ETHER) 1030 dev->mtu -= ETH_HLEN; 1031 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1032 dev->mtu -= 8; 1033 1034 return 0; 1035 } 1036 1037 static int ip6gre_tunnel_init(struct net_device *dev) 1038 { 1039 struct ip6_tnl *tunnel; 1040 int ret; 1041 1042 ret = ip6gre_tunnel_init_common(dev); 1043 if (ret) 1044 return ret; 1045 1046 tunnel = netdev_priv(dev); 1047 1048 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1049 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1050 1051 if (ipv6_addr_any(&tunnel->parms.raddr)) 1052 dev->header_ops = &ip6gre_header_ops; 1053 1054 return 0; 1055 } 1056 1057 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1058 { 1059 struct ip6_tnl *tunnel = netdev_priv(dev); 1060 1061 tunnel->dev = dev; 1062 tunnel->net = dev_net(dev); 1063 strcpy(tunnel->parms.name, dev->name); 1064 1065 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1066 1067 dev_hold(dev); 1068 } 1069 1070 1071 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1072 .handler = gre_rcv, 1073 .err_handler = ip6gre_err, 1074 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1075 }; 1076 1077 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1078 { 1079 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1080 struct net_device *dev, *aux; 1081 int prio; 1082 1083 for_each_netdev_safe(net, dev, aux) 1084 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1085 dev->rtnl_link_ops == &ip6gre_tap_ops) 1086 unregister_netdevice_queue(dev, head); 1087 1088 for (prio = 0; prio < 4; prio++) { 1089 int h; 1090 for (h = 0; h < HASH_SIZE; h++) { 1091 struct ip6_tnl *t; 1092 1093 t = rtnl_dereference(ign->tunnels[prio][h]); 1094 1095 while (t) { 1096 /* If dev is in the same netns, it has already 1097 * been added to the list by the previous loop. 1098 */ 1099 if (!net_eq(dev_net(t->dev), net)) 1100 unregister_netdevice_queue(t->dev, 1101 head); 1102 t = rtnl_dereference(t->next); 1103 } 1104 } 1105 } 1106 } 1107 1108 static int __net_init ip6gre_init_net(struct net *net) 1109 { 1110 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1111 int err; 1112 1113 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1114 NET_NAME_UNKNOWN, 1115 ip6gre_tunnel_setup); 1116 if (!ign->fb_tunnel_dev) { 1117 err = -ENOMEM; 1118 goto err_alloc_dev; 1119 } 1120 dev_net_set(ign->fb_tunnel_dev, net); 1121 /* FB netdevice is special: we have one, and only one per netns. 1122 * Allowing to move it to another netns is clearly unsafe. 1123 */ 1124 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1125 1126 1127 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1128 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1129 1130 err = register_netdev(ign->fb_tunnel_dev); 1131 if (err) 1132 goto err_reg_dev; 1133 1134 rcu_assign_pointer(ign->tunnels_wc[0], 1135 netdev_priv(ign->fb_tunnel_dev)); 1136 return 0; 1137 1138 err_reg_dev: 1139 ip6gre_dev_free(ign->fb_tunnel_dev); 1140 err_alloc_dev: 1141 return err; 1142 } 1143 1144 static void __net_exit ip6gre_exit_net(struct net *net) 1145 { 1146 LIST_HEAD(list); 1147 1148 rtnl_lock(); 1149 ip6gre_destroy_tunnels(net, &list); 1150 unregister_netdevice_many(&list); 1151 rtnl_unlock(); 1152 } 1153 1154 static struct pernet_operations ip6gre_net_ops = { 1155 .init = ip6gre_init_net, 1156 .exit = ip6gre_exit_net, 1157 .id = &ip6gre_net_id, 1158 .size = sizeof(struct ip6gre_net), 1159 }; 1160 1161 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) 1162 { 1163 __be16 flags; 1164 1165 if (!data) 1166 return 0; 1167 1168 flags = 0; 1169 if (data[IFLA_GRE_IFLAGS]) 1170 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1171 if (data[IFLA_GRE_OFLAGS]) 1172 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1173 if (flags & (GRE_VERSION|GRE_ROUTING)) 1174 return -EINVAL; 1175 1176 return 0; 1177 } 1178 1179 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[]) 1180 { 1181 struct in6_addr daddr; 1182 1183 if (tb[IFLA_ADDRESS]) { 1184 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1185 return -EINVAL; 1186 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1187 return -EADDRNOTAVAIL; 1188 } 1189 1190 if (!data) 1191 goto out; 1192 1193 if (data[IFLA_GRE_REMOTE]) { 1194 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1195 if (ipv6_addr_any(&daddr)) 1196 return -EINVAL; 1197 } 1198 1199 out: 1200 return ip6gre_tunnel_validate(tb, data); 1201 } 1202 1203 1204 static void ip6gre_netlink_parms(struct nlattr *data[], 1205 struct __ip6_tnl_parm *parms) 1206 { 1207 memset(parms, 0, sizeof(*parms)); 1208 1209 if (!data) 1210 return; 1211 1212 if (data[IFLA_GRE_LINK]) 1213 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1214 1215 if (data[IFLA_GRE_IFLAGS]) 1216 parms->i_flags = gre_flags_to_tnl_flags( 1217 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1218 1219 if (data[IFLA_GRE_OFLAGS]) 1220 parms->o_flags = gre_flags_to_tnl_flags( 1221 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1222 1223 if (data[IFLA_GRE_IKEY]) 1224 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1225 1226 if (data[IFLA_GRE_OKEY]) 1227 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1228 1229 if (data[IFLA_GRE_LOCAL]) 1230 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1231 1232 if (data[IFLA_GRE_REMOTE]) 1233 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1234 1235 if (data[IFLA_GRE_TTL]) 1236 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1237 1238 if (data[IFLA_GRE_ENCAP_LIMIT]) 1239 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1240 1241 if (data[IFLA_GRE_FLOWINFO]) 1242 parms->flowinfo = nla_get_u32(data[IFLA_GRE_FLOWINFO]); 1243 1244 if (data[IFLA_GRE_FLAGS]) 1245 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1246 } 1247 1248 static int ip6gre_tap_init(struct net_device *dev) 1249 { 1250 struct ip6_tnl *tunnel; 1251 int ret; 1252 1253 ret = ip6gre_tunnel_init_common(dev); 1254 if (ret) 1255 return ret; 1256 1257 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1258 1259 tunnel = netdev_priv(dev); 1260 1261 ip6gre_tnl_link_config(tunnel, 1); 1262 1263 return 0; 1264 } 1265 1266 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1267 .ndo_init = ip6gre_tap_init, 1268 .ndo_uninit = ip6gre_tunnel_uninit, 1269 .ndo_start_xmit = ip6gre_tunnel_xmit, 1270 .ndo_set_mac_address = eth_mac_addr, 1271 .ndo_validate_addr = eth_validate_addr, 1272 .ndo_change_mtu = ip6_tnl_change_mtu, 1273 .ndo_get_stats64 = ip_tunnel_get_stats64, 1274 .ndo_get_iflink = ip6_tnl_get_iflink, 1275 }; 1276 1277 #define GRE6_FEATURES (NETIF_F_SG | \ 1278 NETIF_F_FRAGLIST | \ 1279 NETIF_F_HIGHDMA | \ 1280 NETIF_F_HW_CSUM) 1281 1282 static void ip6gre_tap_setup(struct net_device *dev) 1283 { 1284 1285 ether_setup(dev); 1286 1287 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1288 dev->destructor = ip6gre_dev_free; 1289 1290 dev->features |= NETIF_F_NETNS_LOCAL; 1291 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1292 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1293 } 1294 1295 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1296 struct ip_tunnel_encap *ipencap) 1297 { 1298 bool ret = false; 1299 1300 memset(ipencap, 0, sizeof(*ipencap)); 1301 1302 if (!data) 1303 return ret; 1304 1305 if (data[IFLA_GRE_ENCAP_TYPE]) { 1306 ret = true; 1307 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1308 } 1309 1310 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1311 ret = true; 1312 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1313 } 1314 1315 if (data[IFLA_GRE_ENCAP_SPORT]) { 1316 ret = true; 1317 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1318 } 1319 1320 if (data[IFLA_GRE_ENCAP_DPORT]) { 1321 ret = true; 1322 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1323 } 1324 1325 return ret; 1326 } 1327 1328 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1329 struct nlattr *tb[], struct nlattr *data[]) 1330 { 1331 struct ip6_tnl *nt; 1332 struct net *net = dev_net(dev); 1333 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1334 struct ip_tunnel_encap ipencap; 1335 int err; 1336 1337 nt = netdev_priv(dev); 1338 1339 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1340 int err = ip6_tnl_encap_setup(nt, &ipencap); 1341 1342 if (err < 0) 1343 return err; 1344 } 1345 1346 ip6gre_netlink_parms(data, &nt->parms); 1347 1348 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1349 return -EEXIST; 1350 1351 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1352 eth_hw_addr_random(dev); 1353 1354 nt->dev = dev; 1355 nt->net = dev_net(dev); 1356 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1357 1358 dev->features |= GRE6_FEATURES; 1359 dev->hw_features |= GRE6_FEATURES; 1360 1361 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1362 /* TCP offload with GRE SEQ is not supported, nor 1363 * can we support 2 levels of outer headers requiring 1364 * an update. 1365 */ 1366 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1367 (nt->encap.type == TUNNEL_ENCAP_NONE)) { 1368 dev->features |= NETIF_F_GSO_SOFTWARE; 1369 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1370 } 1371 1372 /* Can use a lockless transmit, unless we generate 1373 * output sequences 1374 */ 1375 dev->features |= NETIF_F_LLTX; 1376 } 1377 1378 err = register_netdevice(dev); 1379 if (err) 1380 goto out; 1381 1382 dev_hold(dev); 1383 ip6gre_tunnel_link(ign, nt); 1384 1385 out: 1386 return err; 1387 } 1388 1389 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 1390 struct nlattr *data[]) 1391 { 1392 struct ip6_tnl *t, *nt = netdev_priv(dev); 1393 struct net *net = nt->net; 1394 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1395 struct __ip6_tnl_parm p; 1396 struct ip_tunnel_encap ipencap; 1397 1398 if (dev == ign->fb_tunnel_dev) 1399 return -EINVAL; 1400 1401 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1402 int err = ip6_tnl_encap_setup(nt, &ipencap); 1403 1404 if (err < 0) 1405 return err; 1406 } 1407 1408 ip6gre_netlink_parms(data, &p); 1409 1410 t = ip6gre_tunnel_locate(net, &p, 0); 1411 1412 if (t) { 1413 if (t->dev != dev) 1414 return -EEXIST; 1415 } else { 1416 t = nt; 1417 } 1418 1419 ip6gre_tunnel_unlink(ign, t); 1420 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 1421 ip6gre_tunnel_link(ign, t); 1422 return 0; 1423 } 1424 1425 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 1426 { 1427 struct net *net = dev_net(dev); 1428 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1429 1430 if (dev != ign->fb_tunnel_dev) 1431 unregister_netdevice_queue(dev, head); 1432 } 1433 1434 static size_t ip6gre_get_size(const struct net_device *dev) 1435 { 1436 return 1437 /* IFLA_GRE_LINK */ 1438 nla_total_size(4) + 1439 /* IFLA_GRE_IFLAGS */ 1440 nla_total_size(2) + 1441 /* IFLA_GRE_OFLAGS */ 1442 nla_total_size(2) + 1443 /* IFLA_GRE_IKEY */ 1444 nla_total_size(4) + 1445 /* IFLA_GRE_OKEY */ 1446 nla_total_size(4) + 1447 /* IFLA_GRE_LOCAL */ 1448 nla_total_size(sizeof(struct in6_addr)) + 1449 /* IFLA_GRE_REMOTE */ 1450 nla_total_size(sizeof(struct in6_addr)) + 1451 /* IFLA_GRE_TTL */ 1452 nla_total_size(1) + 1453 /* IFLA_GRE_ENCAP_LIMIT */ 1454 nla_total_size(1) + 1455 /* IFLA_GRE_FLOWINFO */ 1456 nla_total_size(4) + 1457 /* IFLA_GRE_FLAGS */ 1458 nla_total_size(4) + 1459 /* IFLA_GRE_ENCAP_TYPE */ 1460 nla_total_size(2) + 1461 /* IFLA_GRE_ENCAP_FLAGS */ 1462 nla_total_size(2) + 1463 /* IFLA_GRE_ENCAP_SPORT */ 1464 nla_total_size(2) + 1465 /* IFLA_GRE_ENCAP_DPORT */ 1466 nla_total_size(2) + 1467 0; 1468 } 1469 1470 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 1471 { 1472 struct ip6_tnl *t = netdev_priv(dev); 1473 struct __ip6_tnl_parm *p = &t->parms; 1474 1475 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 1476 nla_put_be16(skb, IFLA_GRE_IFLAGS, 1477 gre_tnl_flags_to_gre_flags(p->i_flags)) || 1478 nla_put_be16(skb, IFLA_GRE_OFLAGS, 1479 gre_tnl_flags_to_gre_flags(p->o_flags)) || 1480 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 1481 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 1482 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 1483 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 1484 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 1485 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 1486 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 1487 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags)) 1488 goto nla_put_failure; 1489 1490 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 1491 t->encap.type) || 1492 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 1493 t->encap.sport) || 1494 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 1495 t->encap.dport) || 1496 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 1497 t->encap.flags)) 1498 goto nla_put_failure; 1499 1500 return 0; 1501 1502 nla_put_failure: 1503 return -EMSGSIZE; 1504 } 1505 1506 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 1507 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 1508 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 1509 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 1510 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 1511 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 1512 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 1513 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 1514 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 1515 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 1516 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 1517 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 1518 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 1519 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 1520 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 1521 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 1522 }; 1523 1524 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 1525 .kind = "ip6gre", 1526 .maxtype = IFLA_GRE_MAX, 1527 .policy = ip6gre_policy, 1528 .priv_size = sizeof(struct ip6_tnl), 1529 .setup = ip6gre_tunnel_setup, 1530 .validate = ip6gre_tunnel_validate, 1531 .newlink = ip6gre_newlink, 1532 .changelink = ip6gre_changelink, 1533 .dellink = ip6gre_dellink, 1534 .get_size = ip6gre_get_size, 1535 .fill_info = ip6gre_fill_info, 1536 .get_link_net = ip6_tnl_get_link_net, 1537 }; 1538 1539 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 1540 .kind = "ip6gretap", 1541 .maxtype = IFLA_GRE_MAX, 1542 .policy = ip6gre_policy, 1543 .priv_size = sizeof(struct ip6_tnl), 1544 .setup = ip6gre_tap_setup, 1545 .validate = ip6gre_tap_validate, 1546 .newlink = ip6gre_newlink, 1547 .changelink = ip6gre_changelink, 1548 .get_size = ip6gre_get_size, 1549 .fill_info = ip6gre_fill_info, 1550 .get_link_net = ip6_tnl_get_link_net, 1551 }; 1552 1553 /* 1554 * And now the modules code and kernel interface. 1555 */ 1556 1557 static int __init ip6gre_init(void) 1558 { 1559 int err; 1560 1561 pr_info("GRE over IPv6 tunneling driver\n"); 1562 1563 err = register_pernet_device(&ip6gre_net_ops); 1564 if (err < 0) 1565 return err; 1566 1567 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 1568 if (err < 0) { 1569 pr_info("%s: can't add protocol\n", __func__); 1570 goto add_proto_failed; 1571 } 1572 1573 err = rtnl_link_register(&ip6gre_link_ops); 1574 if (err < 0) 1575 goto rtnl_link_failed; 1576 1577 err = rtnl_link_register(&ip6gre_tap_ops); 1578 if (err < 0) 1579 goto tap_ops_failed; 1580 1581 out: 1582 return err; 1583 1584 tap_ops_failed: 1585 rtnl_link_unregister(&ip6gre_link_ops); 1586 rtnl_link_failed: 1587 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 1588 add_proto_failed: 1589 unregister_pernet_device(&ip6gre_net_ops); 1590 goto out; 1591 } 1592 1593 static void __exit ip6gre_fini(void) 1594 { 1595 rtnl_link_unregister(&ip6gre_tap_ops); 1596 rtnl_link_unregister(&ip6gre_link_ops); 1597 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 1598 unregister_pernet_device(&ip6gre_net_ops); 1599 } 1600 1601 module_init(ip6gre_init); 1602 module_exit(ip6gre_fini); 1603 MODULE_LICENSE("GPL"); 1604 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 1605 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 1606 MODULE_ALIAS_RTNL_LINK("ip6gre"); 1607 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 1608 MODULE_ALIAS_NETDEV("ip6gre0"); 1609