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 unsigned 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 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 552 553 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 554 if (err) 555 return -1; 556 557 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, 558 skb->protocol); 559 if (err != 0) { 560 /* XXX: send ICMP error even if DF is not set. */ 561 if (err == -EMSGSIZE) 562 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 563 htonl(mtu)); 564 return -1; 565 } 566 567 return 0; 568 } 569 570 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev) 571 { 572 struct ip6_tnl *t = netdev_priv(dev); 573 struct ipv6hdr *ipv6h = ipv6_hdr(skb); 574 int encap_limit = -1; 575 __u16 offset; 576 struct flowi6 fl6; 577 __u8 dsfield; 578 __u32 mtu; 579 int err; 580 581 if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr)) 582 return -1; 583 584 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb)); 585 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */ 586 ipv6h = ipv6_hdr(skb); 587 588 if (offset > 0) { 589 struct ipv6_tlv_tnl_enc_lim *tel; 590 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset]; 591 if (tel->encap_limit == 0) { 592 icmpv6_send(skb, ICMPV6_PARAMPROB, 593 ICMPV6_HDR_FIELD, offset + 2); 594 return -1; 595 } 596 encap_limit = tel->encap_limit - 1; 597 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 598 encap_limit = t->parms.encap_limit; 599 600 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 601 602 dsfield = ipv6_get_dsfield(ipv6h); 603 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS) 604 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK); 605 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL) 606 fl6.flowlabel |= ip6_flowlabel(ipv6h); 607 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK) 608 fl6.flowi6_mark = skb->mark; 609 610 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL); 611 612 if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM))) 613 return -1; 614 615 err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, 616 &mtu, skb->protocol); 617 if (err != 0) { 618 if (err == -EMSGSIZE) 619 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 620 return -1; 621 } 622 623 return 0; 624 } 625 626 /** 627 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own 628 * @t: the outgoing tunnel device 629 * @hdr: IPv6 header from the incoming packet 630 * 631 * Description: 632 * Avoid trivial tunneling loop by checking that tunnel exit-point 633 * doesn't match source of incoming packet. 634 * 635 * Return: 636 * 1 if conflict, 637 * 0 else 638 **/ 639 640 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t, 641 const struct ipv6hdr *hdr) 642 { 643 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr); 644 } 645 646 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev) 647 { 648 struct ip6_tnl *t = netdev_priv(dev); 649 int encap_limit = -1; 650 struct flowi6 fl6; 651 __u32 mtu; 652 int err; 653 654 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 655 encap_limit = t->parms.encap_limit; 656 657 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6)); 658 659 err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)); 660 if (err) 661 return err; 662 663 err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol); 664 665 return err; 666 } 667 668 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb, 669 struct net_device *dev) 670 { 671 struct ip6_tnl *t = netdev_priv(dev); 672 struct net_device_stats *stats = &t->dev->stats; 673 int ret; 674 675 if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr)) 676 goto tx_err; 677 678 switch (skb->protocol) { 679 case htons(ETH_P_IP): 680 ret = ip6gre_xmit_ipv4(skb, dev); 681 break; 682 case htons(ETH_P_IPV6): 683 ret = ip6gre_xmit_ipv6(skb, dev); 684 break; 685 default: 686 ret = ip6gre_xmit_other(skb, dev); 687 break; 688 } 689 690 if (ret < 0) 691 goto tx_err; 692 693 return NETDEV_TX_OK; 694 695 tx_err: 696 stats->tx_errors++; 697 stats->tx_dropped++; 698 kfree_skb(skb); 699 return NETDEV_TX_OK; 700 } 701 702 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu) 703 { 704 struct net_device *dev = t->dev; 705 struct __ip6_tnl_parm *p = &t->parms; 706 struct flowi6 *fl6 = &t->fl.u.ip6; 707 int t_hlen; 708 709 if (dev->type != ARPHRD_ETHER) { 710 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr)); 711 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr)); 712 } 713 714 /* Set up flowi template */ 715 fl6->saddr = p->laddr; 716 fl6->daddr = p->raddr; 717 fl6->flowi6_oif = p->link; 718 fl6->flowlabel = 0; 719 fl6->flowi6_proto = IPPROTO_GRE; 720 721 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS)) 722 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo; 723 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL)) 724 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo; 725 726 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET); 727 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr); 728 729 if (p->flags&IP6_TNL_F_CAP_XMIT && 730 p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER) 731 dev->flags |= IFF_POINTOPOINT; 732 else 733 dev->flags &= ~IFF_POINTOPOINT; 734 735 t->tun_hlen = gre_calc_hlen(t->parms.o_flags); 736 737 t->hlen = t->encap_hlen + t->tun_hlen; 738 739 t_hlen = t->hlen + sizeof(struct ipv6hdr); 740 741 if (p->flags & IP6_TNL_F_CAP_XMIT) { 742 int strict = (ipv6_addr_type(&p->raddr) & 743 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)); 744 745 struct rt6_info *rt = rt6_lookup(t->net, 746 &p->raddr, &p->laddr, 747 p->link, strict); 748 749 if (!rt) 750 return; 751 752 if (rt->dst.dev) { 753 dev->hard_header_len = rt->dst.dev->hard_header_len + 754 t_hlen; 755 756 if (set_mtu) { 757 dev->mtu = rt->dst.dev->mtu - t_hlen; 758 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 759 dev->mtu -= 8; 760 if (dev->type == ARPHRD_ETHER) 761 dev->mtu -= ETH_HLEN; 762 763 if (dev->mtu < IPV6_MIN_MTU) 764 dev->mtu = IPV6_MIN_MTU; 765 } 766 } 767 ip6_rt_put(rt); 768 } 769 } 770 771 static int ip6gre_tnl_change(struct ip6_tnl *t, 772 const struct __ip6_tnl_parm *p, int set_mtu) 773 { 774 t->parms.laddr = p->laddr; 775 t->parms.raddr = p->raddr; 776 t->parms.flags = p->flags; 777 t->parms.hop_limit = p->hop_limit; 778 t->parms.encap_limit = p->encap_limit; 779 t->parms.flowinfo = p->flowinfo; 780 t->parms.link = p->link; 781 t->parms.proto = p->proto; 782 t->parms.i_key = p->i_key; 783 t->parms.o_key = p->o_key; 784 t->parms.i_flags = p->i_flags; 785 t->parms.o_flags = p->o_flags; 786 dst_cache_reset(&t->dst_cache); 787 ip6gre_tnl_link_config(t, set_mtu); 788 return 0; 789 } 790 791 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p, 792 const struct ip6_tnl_parm2 *u) 793 { 794 p->laddr = u->laddr; 795 p->raddr = u->raddr; 796 p->flags = u->flags; 797 p->hop_limit = u->hop_limit; 798 p->encap_limit = u->encap_limit; 799 p->flowinfo = u->flowinfo; 800 p->link = u->link; 801 p->i_key = u->i_key; 802 p->o_key = u->o_key; 803 p->i_flags = gre_flags_to_tnl_flags(u->i_flags); 804 p->o_flags = gre_flags_to_tnl_flags(u->o_flags); 805 memcpy(p->name, u->name, sizeof(u->name)); 806 } 807 808 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u, 809 const struct __ip6_tnl_parm *p) 810 { 811 u->proto = IPPROTO_GRE; 812 u->laddr = p->laddr; 813 u->raddr = p->raddr; 814 u->flags = p->flags; 815 u->hop_limit = p->hop_limit; 816 u->encap_limit = p->encap_limit; 817 u->flowinfo = p->flowinfo; 818 u->link = p->link; 819 u->i_key = p->i_key; 820 u->o_key = p->o_key; 821 u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags); 822 u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags); 823 memcpy(u->name, p->name, sizeof(u->name)); 824 } 825 826 static int ip6gre_tunnel_ioctl(struct net_device *dev, 827 struct ifreq *ifr, int cmd) 828 { 829 int err = 0; 830 struct ip6_tnl_parm2 p; 831 struct __ip6_tnl_parm p1; 832 struct ip6_tnl *t = netdev_priv(dev); 833 struct net *net = t->net; 834 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 835 836 memset(&p1, 0, sizeof(p1)); 837 838 switch (cmd) { 839 case SIOCGETTUNNEL: 840 if (dev == ign->fb_tunnel_dev) { 841 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { 842 err = -EFAULT; 843 break; 844 } 845 ip6gre_tnl_parm_from_user(&p1, &p); 846 t = ip6gre_tunnel_locate(net, &p1, 0); 847 if (!t) 848 t = netdev_priv(dev); 849 } 850 memset(&p, 0, sizeof(p)); 851 ip6gre_tnl_parm_to_user(&p, &t->parms); 852 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 853 err = -EFAULT; 854 break; 855 856 case SIOCADDTUNNEL: 857 case SIOCCHGTUNNEL: 858 err = -EPERM; 859 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 860 goto done; 861 862 err = -EFAULT; 863 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 864 goto done; 865 866 err = -EINVAL; 867 if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING)) 868 goto done; 869 870 if (!(p.i_flags&GRE_KEY)) 871 p.i_key = 0; 872 if (!(p.o_flags&GRE_KEY)) 873 p.o_key = 0; 874 875 ip6gre_tnl_parm_from_user(&p1, &p); 876 t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL); 877 878 if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { 879 if (t) { 880 if (t->dev != dev) { 881 err = -EEXIST; 882 break; 883 } 884 } else { 885 t = netdev_priv(dev); 886 887 ip6gre_tunnel_unlink(ign, t); 888 synchronize_net(); 889 ip6gre_tnl_change(t, &p1, 1); 890 ip6gre_tunnel_link(ign, t); 891 netdev_state_change(dev); 892 } 893 } 894 895 if (t) { 896 err = 0; 897 898 memset(&p, 0, sizeof(p)); 899 ip6gre_tnl_parm_to_user(&p, &t->parms); 900 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 901 err = -EFAULT; 902 } else 903 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); 904 break; 905 906 case SIOCDELTUNNEL: 907 err = -EPERM; 908 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 909 goto done; 910 911 if (dev == ign->fb_tunnel_dev) { 912 err = -EFAULT; 913 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 914 goto done; 915 err = -ENOENT; 916 ip6gre_tnl_parm_from_user(&p1, &p); 917 t = ip6gre_tunnel_locate(net, &p1, 0); 918 if (!t) 919 goto done; 920 err = -EPERM; 921 if (t == netdev_priv(ign->fb_tunnel_dev)) 922 goto done; 923 dev = t->dev; 924 } 925 unregister_netdevice(dev); 926 err = 0; 927 break; 928 929 default: 930 err = -EINVAL; 931 } 932 933 done: 934 return err; 935 } 936 937 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev, 938 unsigned short type, 939 const void *daddr, const void *saddr, unsigned int len) 940 { 941 struct ip6_tnl *t = netdev_priv(dev); 942 struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen); 943 __be16 *p = (__be16 *)(ipv6h+1); 944 945 ip6_flow_hdr(ipv6h, 0, 946 ip6_make_flowlabel(dev_net(dev), skb, 947 t->fl.u.ip6.flowlabel, true, 948 &t->fl.u.ip6)); 949 ipv6h->hop_limit = t->parms.hop_limit; 950 ipv6h->nexthdr = NEXTHDR_GRE; 951 ipv6h->saddr = t->parms.laddr; 952 ipv6h->daddr = t->parms.raddr; 953 954 p[0] = t->parms.o_flags; 955 p[1] = htons(type); 956 957 /* 958 * Set the source hardware address. 959 */ 960 961 if (saddr) 962 memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr)); 963 if (daddr) 964 memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr)); 965 if (!ipv6_addr_any(&ipv6h->daddr)) 966 return t->hlen; 967 968 return -t->hlen; 969 } 970 971 static const struct header_ops ip6gre_header_ops = { 972 .create = ip6gre_header, 973 }; 974 975 static const struct net_device_ops ip6gre_netdev_ops = { 976 .ndo_init = ip6gre_tunnel_init, 977 .ndo_uninit = ip6gre_tunnel_uninit, 978 .ndo_start_xmit = ip6gre_tunnel_xmit, 979 .ndo_do_ioctl = ip6gre_tunnel_ioctl, 980 .ndo_change_mtu = ip6_tnl_change_mtu, 981 .ndo_get_stats64 = ip_tunnel_get_stats64, 982 .ndo_get_iflink = ip6_tnl_get_iflink, 983 }; 984 985 static void ip6gre_dev_free(struct net_device *dev) 986 { 987 struct ip6_tnl *t = netdev_priv(dev); 988 989 dst_cache_destroy(&t->dst_cache); 990 free_percpu(dev->tstats); 991 free_netdev(dev); 992 } 993 994 static void ip6gre_tunnel_setup(struct net_device *dev) 995 { 996 dev->netdev_ops = &ip6gre_netdev_ops; 997 dev->destructor = ip6gre_dev_free; 998 999 dev->type = ARPHRD_IP6GRE; 1000 1001 dev->flags |= IFF_NOARP; 1002 dev->addr_len = sizeof(struct in6_addr); 1003 netif_keep_dst(dev); 1004 } 1005 1006 static int ip6gre_tunnel_init_common(struct net_device *dev) 1007 { 1008 struct ip6_tnl *tunnel; 1009 int ret; 1010 int t_hlen; 1011 1012 tunnel = netdev_priv(dev); 1013 1014 tunnel->dev = dev; 1015 tunnel->net = dev_net(dev); 1016 strcpy(tunnel->parms.name, dev->name); 1017 1018 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); 1019 if (!dev->tstats) 1020 return -ENOMEM; 1021 1022 ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL); 1023 if (ret) { 1024 free_percpu(dev->tstats); 1025 dev->tstats = NULL; 1026 return ret; 1027 } 1028 1029 tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags); 1030 tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen; 1031 t_hlen = tunnel->hlen + sizeof(struct ipv6hdr); 1032 1033 dev->hard_header_len = LL_MAX_HEADER + t_hlen; 1034 dev->mtu = ETH_DATA_LEN - t_hlen; 1035 if (dev->type == ARPHRD_ETHER) 1036 dev->mtu -= ETH_HLEN; 1037 if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) 1038 dev->mtu -= 8; 1039 1040 return 0; 1041 } 1042 1043 static int ip6gre_tunnel_init(struct net_device *dev) 1044 { 1045 struct ip6_tnl *tunnel; 1046 int ret; 1047 1048 ret = ip6gre_tunnel_init_common(dev); 1049 if (ret) 1050 return ret; 1051 1052 tunnel = netdev_priv(dev); 1053 1054 memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr)); 1055 memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr)); 1056 1057 if (ipv6_addr_any(&tunnel->parms.raddr)) 1058 dev->header_ops = &ip6gre_header_ops; 1059 1060 return 0; 1061 } 1062 1063 static void ip6gre_fb_tunnel_init(struct net_device *dev) 1064 { 1065 struct ip6_tnl *tunnel = netdev_priv(dev); 1066 1067 tunnel->dev = dev; 1068 tunnel->net = dev_net(dev); 1069 strcpy(tunnel->parms.name, dev->name); 1070 1071 tunnel->hlen = sizeof(struct ipv6hdr) + 4; 1072 1073 dev_hold(dev); 1074 } 1075 1076 1077 static struct inet6_protocol ip6gre_protocol __read_mostly = { 1078 .handler = gre_rcv, 1079 .err_handler = ip6gre_err, 1080 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, 1081 }; 1082 1083 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head) 1084 { 1085 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1086 struct net_device *dev, *aux; 1087 int prio; 1088 1089 for_each_netdev_safe(net, dev, aux) 1090 if (dev->rtnl_link_ops == &ip6gre_link_ops || 1091 dev->rtnl_link_ops == &ip6gre_tap_ops) 1092 unregister_netdevice_queue(dev, head); 1093 1094 for (prio = 0; prio < 4; prio++) { 1095 int h; 1096 for (h = 0; h < IP6_GRE_HASH_SIZE; h++) { 1097 struct ip6_tnl *t; 1098 1099 t = rtnl_dereference(ign->tunnels[prio][h]); 1100 1101 while (t) { 1102 /* If dev is in the same netns, it has already 1103 * been added to the list by the previous loop. 1104 */ 1105 if (!net_eq(dev_net(t->dev), net)) 1106 unregister_netdevice_queue(t->dev, 1107 head); 1108 t = rtnl_dereference(t->next); 1109 } 1110 } 1111 } 1112 } 1113 1114 static int __net_init ip6gre_init_net(struct net *net) 1115 { 1116 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1117 int err; 1118 1119 ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0", 1120 NET_NAME_UNKNOWN, 1121 ip6gre_tunnel_setup); 1122 if (!ign->fb_tunnel_dev) { 1123 err = -ENOMEM; 1124 goto err_alloc_dev; 1125 } 1126 dev_net_set(ign->fb_tunnel_dev, net); 1127 /* FB netdevice is special: we have one, and only one per netns. 1128 * Allowing to move it to another netns is clearly unsafe. 1129 */ 1130 ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL; 1131 1132 1133 ip6gre_fb_tunnel_init(ign->fb_tunnel_dev); 1134 ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops; 1135 1136 err = register_netdev(ign->fb_tunnel_dev); 1137 if (err) 1138 goto err_reg_dev; 1139 1140 rcu_assign_pointer(ign->tunnels_wc[0], 1141 netdev_priv(ign->fb_tunnel_dev)); 1142 return 0; 1143 1144 err_reg_dev: 1145 ip6gre_dev_free(ign->fb_tunnel_dev); 1146 err_alloc_dev: 1147 return err; 1148 } 1149 1150 static void __net_exit ip6gre_exit_net(struct net *net) 1151 { 1152 LIST_HEAD(list); 1153 1154 rtnl_lock(); 1155 ip6gre_destroy_tunnels(net, &list); 1156 unregister_netdevice_many(&list); 1157 rtnl_unlock(); 1158 } 1159 1160 static struct pernet_operations ip6gre_net_ops = { 1161 .init = ip6gre_init_net, 1162 .exit = ip6gre_exit_net, 1163 .id = &ip6gre_net_id, 1164 .size = sizeof(struct ip6gre_net), 1165 }; 1166 1167 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) 1168 { 1169 __be16 flags; 1170 1171 if (!data) 1172 return 0; 1173 1174 flags = 0; 1175 if (data[IFLA_GRE_IFLAGS]) 1176 flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); 1177 if (data[IFLA_GRE_OFLAGS]) 1178 flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); 1179 if (flags & (GRE_VERSION|GRE_ROUTING)) 1180 return -EINVAL; 1181 1182 return 0; 1183 } 1184 1185 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[]) 1186 { 1187 struct in6_addr daddr; 1188 1189 if (tb[IFLA_ADDRESS]) { 1190 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) 1191 return -EINVAL; 1192 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) 1193 return -EADDRNOTAVAIL; 1194 } 1195 1196 if (!data) 1197 goto out; 1198 1199 if (data[IFLA_GRE_REMOTE]) { 1200 daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1201 if (ipv6_addr_any(&daddr)) 1202 return -EINVAL; 1203 } 1204 1205 out: 1206 return ip6gre_tunnel_validate(tb, data); 1207 } 1208 1209 1210 static void ip6gre_netlink_parms(struct nlattr *data[], 1211 struct __ip6_tnl_parm *parms) 1212 { 1213 memset(parms, 0, sizeof(*parms)); 1214 1215 if (!data) 1216 return; 1217 1218 if (data[IFLA_GRE_LINK]) 1219 parms->link = nla_get_u32(data[IFLA_GRE_LINK]); 1220 1221 if (data[IFLA_GRE_IFLAGS]) 1222 parms->i_flags = gre_flags_to_tnl_flags( 1223 nla_get_be16(data[IFLA_GRE_IFLAGS])); 1224 1225 if (data[IFLA_GRE_OFLAGS]) 1226 parms->o_flags = gre_flags_to_tnl_flags( 1227 nla_get_be16(data[IFLA_GRE_OFLAGS])); 1228 1229 if (data[IFLA_GRE_IKEY]) 1230 parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); 1231 1232 if (data[IFLA_GRE_OKEY]) 1233 parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); 1234 1235 if (data[IFLA_GRE_LOCAL]) 1236 parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]); 1237 1238 if (data[IFLA_GRE_REMOTE]) 1239 parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]); 1240 1241 if (data[IFLA_GRE_TTL]) 1242 parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]); 1243 1244 if (data[IFLA_GRE_ENCAP_LIMIT]) 1245 parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]); 1246 1247 if (data[IFLA_GRE_FLOWINFO]) 1248 parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]); 1249 1250 if (data[IFLA_GRE_FLAGS]) 1251 parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]); 1252 } 1253 1254 static int ip6gre_tap_init(struct net_device *dev) 1255 { 1256 struct ip6_tnl *tunnel; 1257 int ret; 1258 1259 ret = ip6gre_tunnel_init_common(dev); 1260 if (ret) 1261 return ret; 1262 1263 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1264 1265 tunnel = netdev_priv(dev); 1266 1267 ip6gre_tnl_link_config(tunnel, 1); 1268 1269 return 0; 1270 } 1271 1272 static const struct net_device_ops ip6gre_tap_netdev_ops = { 1273 .ndo_init = ip6gre_tap_init, 1274 .ndo_uninit = ip6gre_tunnel_uninit, 1275 .ndo_start_xmit = ip6gre_tunnel_xmit, 1276 .ndo_set_mac_address = eth_mac_addr, 1277 .ndo_validate_addr = eth_validate_addr, 1278 .ndo_change_mtu = ip6_tnl_change_mtu, 1279 .ndo_get_stats64 = ip_tunnel_get_stats64, 1280 .ndo_get_iflink = ip6_tnl_get_iflink, 1281 }; 1282 1283 #define GRE6_FEATURES (NETIF_F_SG | \ 1284 NETIF_F_FRAGLIST | \ 1285 NETIF_F_HIGHDMA | \ 1286 NETIF_F_HW_CSUM) 1287 1288 static void ip6gre_tap_setup(struct net_device *dev) 1289 { 1290 1291 ether_setup(dev); 1292 1293 dev->netdev_ops = &ip6gre_tap_netdev_ops; 1294 dev->destructor = ip6gre_dev_free; 1295 1296 dev->features |= NETIF_F_NETNS_LOCAL; 1297 dev->priv_flags &= ~IFF_TX_SKB_SHARING; 1298 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; 1299 } 1300 1301 static bool ip6gre_netlink_encap_parms(struct nlattr *data[], 1302 struct ip_tunnel_encap *ipencap) 1303 { 1304 bool ret = false; 1305 1306 memset(ipencap, 0, sizeof(*ipencap)); 1307 1308 if (!data) 1309 return ret; 1310 1311 if (data[IFLA_GRE_ENCAP_TYPE]) { 1312 ret = true; 1313 ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]); 1314 } 1315 1316 if (data[IFLA_GRE_ENCAP_FLAGS]) { 1317 ret = true; 1318 ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]); 1319 } 1320 1321 if (data[IFLA_GRE_ENCAP_SPORT]) { 1322 ret = true; 1323 ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]); 1324 } 1325 1326 if (data[IFLA_GRE_ENCAP_DPORT]) { 1327 ret = true; 1328 ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]); 1329 } 1330 1331 return ret; 1332 } 1333 1334 static int ip6gre_newlink(struct net *src_net, struct net_device *dev, 1335 struct nlattr *tb[], struct nlattr *data[]) 1336 { 1337 struct ip6_tnl *nt; 1338 struct net *net = dev_net(dev); 1339 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1340 struct ip_tunnel_encap ipencap; 1341 int err; 1342 1343 nt = netdev_priv(dev); 1344 1345 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1346 int err = ip6_tnl_encap_setup(nt, &ipencap); 1347 1348 if (err < 0) 1349 return err; 1350 } 1351 1352 ip6gre_netlink_parms(data, &nt->parms); 1353 1354 if (ip6gre_tunnel_find(net, &nt->parms, dev->type)) 1355 return -EEXIST; 1356 1357 if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS]) 1358 eth_hw_addr_random(dev); 1359 1360 nt->dev = dev; 1361 nt->net = dev_net(dev); 1362 ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]); 1363 1364 dev->features |= GRE6_FEATURES; 1365 dev->hw_features |= GRE6_FEATURES; 1366 1367 if (!(nt->parms.o_flags & TUNNEL_SEQ)) { 1368 /* TCP offload with GRE SEQ is not supported, nor 1369 * can we support 2 levels of outer headers requiring 1370 * an update. 1371 */ 1372 if (!(nt->parms.o_flags & TUNNEL_CSUM) || 1373 (nt->encap.type == TUNNEL_ENCAP_NONE)) { 1374 dev->features |= NETIF_F_GSO_SOFTWARE; 1375 dev->hw_features |= NETIF_F_GSO_SOFTWARE; 1376 } 1377 1378 /* Can use a lockless transmit, unless we generate 1379 * output sequences 1380 */ 1381 dev->features |= NETIF_F_LLTX; 1382 } 1383 1384 err = register_netdevice(dev); 1385 if (err) 1386 goto out; 1387 1388 dev_hold(dev); 1389 ip6gre_tunnel_link(ign, nt); 1390 1391 out: 1392 return err; 1393 } 1394 1395 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[], 1396 struct nlattr *data[]) 1397 { 1398 struct ip6_tnl *t, *nt = netdev_priv(dev); 1399 struct net *net = nt->net; 1400 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1401 struct __ip6_tnl_parm p; 1402 struct ip_tunnel_encap ipencap; 1403 1404 if (dev == ign->fb_tunnel_dev) 1405 return -EINVAL; 1406 1407 if (ip6gre_netlink_encap_parms(data, &ipencap)) { 1408 int err = ip6_tnl_encap_setup(nt, &ipencap); 1409 1410 if (err < 0) 1411 return err; 1412 } 1413 1414 ip6gre_netlink_parms(data, &p); 1415 1416 t = ip6gre_tunnel_locate(net, &p, 0); 1417 1418 if (t) { 1419 if (t->dev != dev) 1420 return -EEXIST; 1421 } else { 1422 t = nt; 1423 } 1424 1425 ip6gre_tunnel_unlink(ign, t); 1426 ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]); 1427 ip6gre_tunnel_link(ign, t); 1428 return 0; 1429 } 1430 1431 static void ip6gre_dellink(struct net_device *dev, struct list_head *head) 1432 { 1433 struct net *net = dev_net(dev); 1434 struct ip6gre_net *ign = net_generic(net, ip6gre_net_id); 1435 1436 if (dev != ign->fb_tunnel_dev) 1437 unregister_netdevice_queue(dev, head); 1438 } 1439 1440 static size_t ip6gre_get_size(const struct net_device *dev) 1441 { 1442 return 1443 /* IFLA_GRE_LINK */ 1444 nla_total_size(4) + 1445 /* IFLA_GRE_IFLAGS */ 1446 nla_total_size(2) + 1447 /* IFLA_GRE_OFLAGS */ 1448 nla_total_size(2) + 1449 /* IFLA_GRE_IKEY */ 1450 nla_total_size(4) + 1451 /* IFLA_GRE_OKEY */ 1452 nla_total_size(4) + 1453 /* IFLA_GRE_LOCAL */ 1454 nla_total_size(sizeof(struct in6_addr)) + 1455 /* IFLA_GRE_REMOTE */ 1456 nla_total_size(sizeof(struct in6_addr)) + 1457 /* IFLA_GRE_TTL */ 1458 nla_total_size(1) + 1459 /* IFLA_GRE_ENCAP_LIMIT */ 1460 nla_total_size(1) + 1461 /* IFLA_GRE_FLOWINFO */ 1462 nla_total_size(4) + 1463 /* IFLA_GRE_FLAGS */ 1464 nla_total_size(4) + 1465 /* IFLA_GRE_ENCAP_TYPE */ 1466 nla_total_size(2) + 1467 /* IFLA_GRE_ENCAP_FLAGS */ 1468 nla_total_size(2) + 1469 /* IFLA_GRE_ENCAP_SPORT */ 1470 nla_total_size(2) + 1471 /* IFLA_GRE_ENCAP_DPORT */ 1472 nla_total_size(2) + 1473 0; 1474 } 1475 1476 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev) 1477 { 1478 struct ip6_tnl *t = netdev_priv(dev); 1479 struct __ip6_tnl_parm *p = &t->parms; 1480 1481 if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || 1482 nla_put_be16(skb, IFLA_GRE_IFLAGS, 1483 gre_tnl_flags_to_gre_flags(p->i_flags)) || 1484 nla_put_be16(skb, IFLA_GRE_OFLAGS, 1485 gre_tnl_flags_to_gre_flags(p->o_flags)) || 1486 nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || 1487 nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || 1488 nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) || 1489 nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) || 1490 nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) || 1491 nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) || 1492 nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) || 1493 nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags)) 1494 goto nla_put_failure; 1495 1496 if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE, 1497 t->encap.type) || 1498 nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT, 1499 t->encap.sport) || 1500 nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT, 1501 t->encap.dport) || 1502 nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS, 1503 t->encap.flags)) 1504 goto nla_put_failure; 1505 1506 return 0; 1507 1508 nla_put_failure: 1509 return -EMSGSIZE; 1510 } 1511 1512 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = { 1513 [IFLA_GRE_LINK] = { .type = NLA_U32 }, 1514 [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, 1515 [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, 1516 [IFLA_GRE_IKEY] = { .type = NLA_U32 }, 1517 [IFLA_GRE_OKEY] = { .type = NLA_U32 }, 1518 [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) }, 1519 [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) }, 1520 [IFLA_GRE_TTL] = { .type = NLA_U8 }, 1521 [IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 }, 1522 [IFLA_GRE_FLOWINFO] = { .type = NLA_U32 }, 1523 [IFLA_GRE_FLAGS] = { .type = NLA_U32 }, 1524 [IFLA_GRE_ENCAP_TYPE] = { .type = NLA_U16 }, 1525 [IFLA_GRE_ENCAP_FLAGS] = { .type = NLA_U16 }, 1526 [IFLA_GRE_ENCAP_SPORT] = { .type = NLA_U16 }, 1527 [IFLA_GRE_ENCAP_DPORT] = { .type = NLA_U16 }, 1528 }; 1529 1530 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = { 1531 .kind = "ip6gre", 1532 .maxtype = IFLA_GRE_MAX, 1533 .policy = ip6gre_policy, 1534 .priv_size = sizeof(struct ip6_tnl), 1535 .setup = ip6gre_tunnel_setup, 1536 .validate = ip6gre_tunnel_validate, 1537 .newlink = ip6gre_newlink, 1538 .changelink = ip6gre_changelink, 1539 .dellink = ip6gre_dellink, 1540 .get_size = ip6gre_get_size, 1541 .fill_info = ip6gre_fill_info, 1542 .get_link_net = ip6_tnl_get_link_net, 1543 }; 1544 1545 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = { 1546 .kind = "ip6gretap", 1547 .maxtype = IFLA_GRE_MAX, 1548 .policy = ip6gre_policy, 1549 .priv_size = sizeof(struct ip6_tnl), 1550 .setup = ip6gre_tap_setup, 1551 .validate = ip6gre_tap_validate, 1552 .newlink = ip6gre_newlink, 1553 .changelink = ip6gre_changelink, 1554 .get_size = ip6gre_get_size, 1555 .fill_info = ip6gre_fill_info, 1556 .get_link_net = ip6_tnl_get_link_net, 1557 }; 1558 1559 /* 1560 * And now the modules code and kernel interface. 1561 */ 1562 1563 static int __init ip6gre_init(void) 1564 { 1565 int err; 1566 1567 pr_info("GRE over IPv6 tunneling driver\n"); 1568 1569 err = register_pernet_device(&ip6gre_net_ops); 1570 if (err < 0) 1571 return err; 1572 1573 err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE); 1574 if (err < 0) { 1575 pr_info("%s: can't add protocol\n", __func__); 1576 goto add_proto_failed; 1577 } 1578 1579 err = rtnl_link_register(&ip6gre_link_ops); 1580 if (err < 0) 1581 goto rtnl_link_failed; 1582 1583 err = rtnl_link_register(&ip6gre_tap_ops); 1584 if (err < 0) 1585 goto tap_ops_failed; 1586 1587 out: 1588 return err; 1589 1590 tap_ops_failed: 1591 rtnl_link_unregister(&ip6gre_link_ops); 1592 rtnl_link_failed: 1593 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 1594 add_proto_failed: 1595 unregister_pernet_device(&ip6gre_net_ops); 1596 goto out; 1597 } 1598 1599 static void __exit ip6gre_fini(void) 1600 { 1601 rtnl_link_unregister(&ip6gre_tap_ops); 1602 rtnl_link_unregister(&ip6gre_link_ops); 1603 inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE); 1604 unregister_pernet_device(&ip6gre_net_ops); 1605 } 1606 1607 module_init(ip6gre_init); 1608 module_exit(ip6gre_fini); 1609 MODULE_LICENSE("GPL"); 1610 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)"); 1611 MODULE_DESCRIPTION("GRE over IPv6 tunneling device"); 1612 MODULE_ALIAS_RTNL_LINK("ip6gre"); 1613 MODULE_ALIAS_RTNL_LINK("ip6gretap"); 1614 MODULE_ALIAS_NETDEV("ip6gre0"); 1615