| 1 | /* |
| 2 | * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT) |
| 3 | * Linux INET6 implementation |
| 4 | * |
| 5 | * Authors: |
| 6 | * Pedro Roque <roque@di.fc.ul.pt> |
| 7 | * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> |
| 8 | * |
| 9 | * This program is free software; you can redistribute it and/or |
| 10 | * modify it under the terms of the GNU General Public License |
| 11 | * as published by the Free Software Foundation; either version |
| 12 | * 2 of the License, or (at your option) any later version. |
| 13 | * |
| 14 | * Changes: |
| 15 | * Roger Venning <r.venning@telstra.com>: 6to4 support |
| 16 | * Nate Thompson <nate@thebog.net>: 6to4 support |
| 17 | * Fred Templin <fred.l.templin@boeing.com>: isatap support |
| 18 | */ |
| 19 | |
| 20 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
| 21 | |
| 22 | #include <linux/module.h> |
| 23 | #include <linux/capability.h> |
| 24 | #include <linux/errno.h> |
| 25 | #include <linux/types.h> |
| 26 | #include <linux/socket.h> |
| 27 | #include <linux/sockios.h> |
| 28 | #include <linux/net.h> |
| 29 | #include <linux/in6.h> |
| 30 | #include <linux/netdevice.h> |
| 31 | #include <linux/if_arp.h> |
| 32 | #include <linux/icmp.h> |
| 33 | #include <linux/slab.h> |
| 34 | #include <asm/uaccess.h> |
| 35 | #include <linux/init.h> |
| 36 | #include <linux/netfilter_ipv4.h> |
| 37 | #include <linux/if_ether.h> |
| 38 | |
| 39 | #include <net/sock.h> |
| 40 | #include <net/snmp.h> |
| 41 | |
| 42 | #include <net/ipv6.h> |
| 43 | #include <net/protocol.h> |
| 44 | #include <net/transp_v6.h> |
| 45 | #include <net/ip6_fib.h> |
| 46 | #include <net/ip6_route.h> |
| 47 | #include <net/ndisc.h> |
| 48 | #include <net/addrconf.h> |
| 49 | #include <net/ip.h> |
| 50 | #include <net/udp.h> |
| 51 | #include <net/icmp.h> |
| 52 | #include <net/ipip.h> |
| 53 | #include <net/inet_ecn.h> |
| 54 | #include <net/xfrm.h> |
| 55 | #include <net/dsfield.h> |
| 56 | #include <net/net_namespace.h> |
| 57 | #include <net/netns/generic.h> |
| 58 | |
| 59 | /* |
| 60 | This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c |
| 61 | |
| 62 | For comments look at net/ipv4/ip_gre.c --ANK |
| 63 | */ |
| 64 | |
| 65 | #define HASH_SIZE 16 |
| 66 | #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF) |
| 67 | |
| 68 | static int ipip6_tunnel_init(struct net_device *dev); |
| 69 | static void ipip6_tunnel_setup(struct net_device *dev); |
| 70 | static void ipip6_dev_free(struct net_device *dev); |
| 71 | |
| 72 | static int sit_net_id __read_mostly; |
| 73 | struct sit_net { |
| 74 | struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE]; |
| 75 | struct ip_tunnel __rcu *tunnels_r[HASH_SIZE]; |
| 76 | struct ip_tunnel __rcu *tunnels_l[HASH_SIZE]; |
| 77 | struct ip_tunnel __rcu *tunnels_wc[1]; |
| 78 | struct ip_tunnel __rcu **tunnels[4]; |
| 79 | |
| 80 | struct net_device *fb_tunnel_dev; |
| 81 | }; |
| 82 | |
| 83 | /* |
| 84 | * Locking : hash tables are protected by RCU and RTNL |
| 85 | */ |
| 86 | |
| 87 | #define for_each_ip_tunnel_rcu(start) \ |
| 88 | for (t = rcu_dereference(start); t; t = rcu_dereference(t->next)) |
| 89 | |
| 90 | /* often modified stats are per cpu, other are shared (netdev->stats) */ |
| 91 | struct pcpu_tstats { |
| 92 | u64 rx_packets; |
| 93 | u64 rx_bytes; |
| 94 | u64 tx_packets; |
| 95 | u64 tx_bytes; |
| 96 | struct u64_stats_sync syncp; |
| 97 | }; |
| 98 | |
| 99 | static struct rtnl_link_stats64 *ipip6_get_stats64(struct net_device *dev, |
| 100 | struct rtnl_link_stats64 *tot) |
| 101 | { |
| 102 | int i; |
| 103 | |
| 104 | for_each_possible_cpu(i) { |
| 105 | const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i); |
| 106 | u64 rx_packets, rx_bytes, tx_packets, tx_bytes; |
| 107 | unsigned int start; |
| 108 | |
| 109 | do { |
| 110 | start = u64_stats_fetch_begin_bh(&tstats->syncp); |
| 111 | rx_packets = tstats->rx_packets; |
| 112 | tx_packets = tstats->tx_packets; |
| 113 | rx_bytes = tstats->rx_bytes; |
| 114 | tx_bytes = tstats->tx_bytes; |
| 115 | } while (u64_stats_fetch_retry_bh(&tstats->syncp, start)); |
| 116 | |
| 117 | tot->rx_packets += rx_packets; |
| 118 | tot->tx_packets += tx_packets; |
| 119 | tot->rx_bytes += rx_bytes; |
| 120 | tot->tx_bytes += tx_bytes; |
| 121 | } |
| 122 | |
| 123 | tot->rx_errors = dev->stats.rx_errors; |
| 124 | tot->tx_fifo_errors = dev->stats.tx_fifo_errors; |
| 125 | tot->tx_carrier_errors = dev->stats.tx_carrier_errors; |
| 126 | tot->tx_dropped = dev->stats.tx_dropped; |
| 127 | tot->tx_aborted_errors = dev->stats.tx_aborted_errors; |
| 128 | tot->tx_errors = dev->stats.tx_errors; |
| 129 | |
| 130 | return tot; |
| 131 | } |
| 132 | |
| 133 | /* |
| 134 | * Must be invoked with rcu_read_lock |
| 135 | */ |
| 136 | static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net, |
| 137 | struct net_device *dev, __be32 remote, __be32 local) |
| 138 | { |
| 139 | unsigned int h0 = HASH(remote); |
| 140 | unsigned int h1 = HASH(local); |
| 141 | struct ip_tunnel *t; |
| 142 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 143 | |
| 144 | for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) { |
| 145 | if (local == t->parms.iph.saddr && |
| 146 | remote == t->parms.iph.daddr && |
| 147 | (!dev || !t->parms.link || dev->iflink == t->parms.link) && |
| 148 | (t->dev->flags & IFF_UP)) |
| 149 | return t; |
| 150 | } |
| 151 | for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) { |
| 152 | if (remote == t->parms.iph.daddr && |
| 153 | (!dev || !t->parms.link || dev->iflink == t->parms.link) && |
| 154 | (t->dev->flags & IFF_UP)) |
| 155 | return t; |
| 156 | } |
| 157 | for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) { |
| 158 | if (local == t->parms.iph.saddr && |
| 159 | (!dev || !t->parms.link || dev->iflink == t->parms.link) && |
| 160 | (t->dev->flags & IFF_UP)) |
| 161 | return t; |
| 162 | } |
| 163 | t = rcu_dereference(sitn->tunnels_wc[0]); |
| 164 | if ((t != NULL) && (t->dev->flags & IFF_UP)) |
| 165 | return t; |
| 166 | return NULL; |
| 167 | } |
| 168 | |
| 169 | static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn, |
| 170 | struct ip_tunnel_parm *parms) |
| 171 | { |
| 172 | __be32 remote = parms->iph.daddr; |
| 173 | __be32 local = parms->iph.saddr; |
| 174 | unsigned int h = 0; |
| 175 | int prio = 0; |
| 176 | |
| 177 | if (remote) { |
| 178 | prio |= 2; |
| 179 | h ^= HASH(remote); |
| 180 | } |
| 181 | if (local) { |
| 182 | prio |= 1; |
| 183 | h ^= HASH(local); |
| 184 | } |
| 185 | return &sitn->tunnels[prio][h]; |
| 186 | } |
| 187 | |
| 188 | static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn, |
| 189 | struct ip_tunnel *t) |
| 190 | { |
| 191 | return __ipip6_bucket(sitn, &t->parms); |
| 192 | } |
| 193 | |
| 194 | static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t) |
| 195 | { |
| 196 | struct ip_tunnel __rcu **tp; |
| 197 | struct ip_tunnel *iter; |
| 198 | |
| 199 | for (tp = ipip6_bucket(sitn, t); |
| 200 | (iter = rtnl_dereference(*tp)) != NULL; |
| 201 | tp = &iter->next) { |
| 202 | if (t == iter) { |
| 203 | rcu_assign_pointer(*tp, t->next); |
| 204 | break; |
| 205 | } |
| 206 | } |
| 207 | } |
| 208 | |
| 209 | static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t) |
| 210 | { |
| 211 | struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t); |
| 212 | |
| 213 | rcu_assign_pointer(t->next, rtnl_dereference(*tp)); |
| 214 | rcu_assign_pointer(*tp, t); |
| 215 | } |
| 216 | |
| 217 | static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn) |
| 218 | { |
| 219 | #ifdef CONFIG_IPV6_SIT_6RD |
| 220 | struct ip_tunnel *t = netdev_priv(dev); |
| 221 | |
| 222 | if (t->dev == sitn->fb_tunnel_dev) { |
| 223 | ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0); |
| 224 | t->ip6rd.relay_prefix = 0; |
| 225 | t->ip6rd.prefixlen = 16; |
| 226 | t->ip6rd.relay_prefixlen = 0; |
| 227 | } else { |
| 228 | struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev); |
| 229 | memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd)); |
| 230 | } |
| 231 | #endif |
| 232 | } |
| 233 | |
| 234 | static struct ip_tunnel *ipip6_tunnel_locate(struct net *net, |
| 235 | struct ip_tunnel_parm *parms, int create) |
| 236 | { |
| 237 | __be32 remote = parms->iph.daddr; |
| 238 | __be32 local = parms->iph.saddr; |
| 239 | struct ip_tunnel *t, *nt; |
| 240 | struct ip_tunnel __rcu **tp; |
| 241 | struct net_device *dev; |
| 242 | char name[IFNAMSIZ]; |
| 243 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 244 | |
| 245 | for (tp = __ipip6_bucket(sitn, parms); |
| 246 | (t = rtnl_dereference(*tp)) != NULL; |
| 247 | tp = &t->next) { |
| 248 | if (local == t->parms.iph.saddr && |
| 249 | remote == t->parms.iph.daddr && |
| 250 | parms->link == t->parms.link) { |
| 251 | if (create) |
| 252 | return NULL; |
| 253 | else |
| 254 | return t; |
| 255 | } |
| 256 | } |
| 257 | if (!create) |
| 258 | goto failed; |
| 259 | |
| 260 | if (parms->name[0]) |
| 261 | strlcpy(name, parms->name, IFNAMSIZ); |
| 262 | else |
| 263 | strcpy(name, "sit%d"); |
| 264 | |
| 265 | dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup); |
| 266 | if (dev == NULL) |
| 267 | return NULL; |
| 268 | |
| 269 | dev_net_set(dev, net); |
| 270 | |
| 271 | nt = netdev_priv(dev); |
| 272 | |
| 273 | nt->parms = *parms; |
| 274 | if (ipip6_tunnel_init(dev) < 0) |
| 275 | goto failed_free; |
| 276 | ipip6_tunnel_clone_6rd(dev, sitn); |
| 277 | |
| 278 | if (parms->i_flags & SIT_ISATAP) |
| 279 | dev->priv_flags |= IFF_ISATAP; |
| 280 | |
| 281 | if (register_netdevice(dev) < 0) |
| 282 | goto failed_free; |
| 283 | |
| 284 | strcpy(nt->parms.name, dev->name); |
| 285 | |
| 286 | dev_hold(dev); |
| 287 | |
| 288 | ipip6_tunnel_link(sitn, nt); |
| 289 | return nt; |
| 290 | |
| 291 | failed_free: |
| 292 | ipip6_dev_free(dev); |
| 293 | failed: |
| 294 | return NULL; |
| 295 | } |
| 296 | |
| 297 | #define for_each_prl_rcu(start) \ |
| 298 | for (prl = rcu_dereference(start); \ |
| 299 | prl; \ |
| 300 | prl = rcu_dereference(prl->next)) |
| 301 | |
| 302 | static struct ip_tunnel_prl_entry * |
| 303 | __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr) |
| 304 | { |
| 305 | struct ip_tunnel_prl_entry *prl; |
| 306 | |
| 307 | for_each_prl_rcu(t->prl) |
| 308 | if (prl->addr == addr) |
| 309 | break; |
| 310 | return prl; |
| 311 | |
| 312 | } |
| 313 | |
| 314 | static int ipip6_tunnel_get_prl(struct ip_tunnel *t, |
| 315 | struct ip_tunnel_prl __user *a) |
| 316 | { |
| 317 | struct ip_tunnel_prl kprl, *kp; |
| 318 | struct ip_tunnel_prl_entry *prl; |
| 319 | unsigned int cmax, c = 0, ca, len; |
| 320 | int ret = 0; |
| 321 | |
| 322 | if (copy_from_user(&kprl, a, sizeof(kprl))) |
| 323 | return -EFAULT; |
| 324 | cmax = kprl.datalen / sizeof(kprl); |
| 325 | if (cmax > 1 && kprl.addr != htonl(INADDR_ANY)) |
| 326 | cmax = 1; |
| 327 | |
| 328 | /* For simple GET or for root users, |
| 329 | * we try harder to allocate. |
| 330 | */ |
| 331 | kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ? |
| 332 | kcalloc(cmax, sizeof(*kp), GFP_KERNEL) : |
| 333 | NULL; |
| 334 | |
| 335 | rcu_read_lock(); |
| 336 | |
| 337 | ca = t->prl_count < cmax ? t->prl_count : cmax; |
| 338 | |
| 339 | if (!kp) { |
| 340 | /* We don't try hard to allocate much memory for |
| 341 | * non-root users. |
| 342 | * For root users, retry allocating enough memory for |
| 343 | * the answer. |
| 344 | */ |
| 345 | kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC); |
| 346 | if (!kp) { |
| 347 | ret = -ENOMEM; |
| 348 | goto out; |
| 349 | } |
| 350 | } |
| 351 | |
| 352 | c = 0; |
| 353 | for_each_prl_rcu(t->prl) { |
| 354 | if (c >= cmax) |
| 355 | break; |
| 356 | if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr) |
| 357 | continue; |
| 358 | kp[c].addr = prl->addr; |
| 359 | kp[c].flags = prl->flags; |
| 360 | c++; |
| 361 | if (kprl.addr != htonl(INADDR_ANY)) |
| 362 | break; |
| 363 | } |
| 364 | out: |
| 365 | rcu_read_unlock(); |
| 366 | |
| 367 | len = sizeof(*kp) * c; |
| 368 | ret = 0; |
| 369 | if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen)) |
| 370 | ret = -EFAULT; |
| 371 | |
| 372 | kfree(kp); |
| 373 | |
| 374 | return ret; |
| 375 | } |
| 376 | |
| 377 | static int |
| 378 | ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg) |
| 379 | { |
| 380 | struct ip_tunnel_prl_entry *p; |
| 381 | int err = 0; |
| 382 | |
| 383 | if (a->addr == htonl(INADDR_ANY)) |
| 384 | return -EINVAL; |
| 385 | |
| 386 | ASSERT_RTNL(); |
| 387 | |
| 388 | for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) { |
| 389 | if (p->addr == a->addr) { |
| 390 | if (chg) { |
| 391 | p->flags = a->flags; |
| 392 | goto out; |
| 393 | } |
| 394 | err = -EEXIST; |
| 395 | goto out; |
| 396 | } |
| 397 | } |
| 398 | |
| 399 | if (chg) { |
| 400 | err = -ENXIO; |
| 401 | goto out; |
| 402 | } |
| 403 | |
| 404 | p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL); |
| 405 | if (!p) { |
| 406 | err = -ENOBUFS; |
| 407 | goto out; |
| 408 | } |
| 409 | |
| 410 | p->next = t->prl; |
| 411 | p->addr = a->addr; |
| 412 | p->flags = a->flags; |
| 413 | t->prl_count++; |
| 414 | rcu_assign_pointer(t->prl, p); |
| 415 | out: |
| 416 | return err; |
| 417 | } |
| 418 | |
| 419 | static void prl_list_destroy_rcu(struct rcu_head *head) |
| 420 | { |
| 421 | struct ip_tunnel_prl_entry *p, *n; |
| 422 | |
| 423 | p = container_of(head, struct ip_tunnel_prl_entry, rcu_head); |
| 424 | do { |
| 425 | n = rcu_dereference_protected(p->next, 1); |
| 426 | kfree(p); |
| 427 | p = n; |
| 428 | } while (p); |
| 429 | } |
| 430 | |
| 431 | static int |
| 432 | ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a) |
| 433 | { |
| 434 | struct ip_tunnel_prl_entry *x; |
| 435 | struct ip_tunnel_prl_entry __rcu **p; |
| 436 | int err = 0; |
| 437 | |
| 438 | ASSERT_RTNL(); |
| 439 | |
| 440 | if (a && a->addr != htonl(INADDR_ANY)) { |
| 441 | for (p = &t->prl; |
| 442 | (x = rtnl_dereference(*p)) != NULL; |
| 443 | p = &x->next) { |
| 444 | if (x->addr == a->addr) { |
| 445 | *p = x->next; |
| 446 | kfree_rcu(x, rcu_head); |
| 447 | t->prl_count--; |
| 448 | goto out; |
| 449 | } |
| 450 | } |
| 451 | err = -ENXIO; |
| 452 | } else { |
| 453 | x = rtnl_dereference(t->prl); |
| 454 | if (x) { |
| 455 | t->prl_count = 0; |
| 456 | call_rcu(&x->rcu_head, prl_list_destroy_rcu); |
| 457 | t->prl = NULL; |
| 458 | } |
| 459 | } |
| 460 | out: |
| 461 | return err; |
| 462 | } |
| 463 | |
| 464 | static int |
| 465 | isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t) |
| 466 | { |
| 467 | struct ip_tunnel_prl_entry *p; |
| 468 | int ok = 1; |
| 469 | |
| 470 | rcu_read_lock(); |
| 471 | p = __ipip6_tunnel_locate_prl(t, iph->saddr); |
| 472 | if (p) { |
| 473 | if (p->flags & PRL_DEFAULT) |
| 474 | skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT; |
| 475 | else |
| 476 | skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT; |
| 477 | } else { |
| 478 | const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr; |
| 479 | |
| 480 | if (ipv6_addr_is_isatap(addr6) && |
| 481 | (addr6->s6_addr32[3] == iph->saddr) && |
| 482 | ipv6_chk_prefix(addr6, t->dev)) |
| 483 | skb->ndisc_nodetype = NDISC_NODETYPE_HOST; |
| 484 | else |
| 485 | ok = 0; |
| 486 | } |
| 487 | rcu_read_unlock(); |
| 488 | return ok; |
| 489 | } |
| 490 | |
| 491 | static void ipip6_tunnel_uninit(struct net_device *dev) |
| 492 | { |
| 493 | struct net *net = dev_net(dev); |
| 494 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 495 | |
| 496 | if (dev == sitn->fb_tunnel_dev) { |
| 497 | RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL); |
| 498 | } else { |
| 499 | ipip6_tunnel_unlink(sitn, netdev_priv(dev)); |
| 500 | ipip6_tunnel_del_prl(netdev_priv(dev), NULL); |
| 501 | } |
| 502 | dev_put(dev); |
| 503 | } |
| 504 | |
| 505 | |
| 506 | static int ipip6_err(struct sk_buff *skb, u32 info) |
| 507 | { |
| 508 | |
| 509 | /* All the routers (except for Linux) return only |
| 510 | 8 bytes of packet payload. It means, that precise relaying of |
| 511 | ICMP in the real Internet is absolutely infeasible. |
| 512 | */ |
| 513 | const struct iphdr *iph = (const struct iphdr *)skb->data; |
| 514 | const int type = icmp_hdr(skb)->type; |
| 515 | const int code = icmp_hdr(skb)->code; |
| 516 | struct ip_tunnel *t; |
| 517 | int err; |
| 518 | |
| 519 | switch (type) { |
| 520 | default: |
| 521 | case ICMP_PARAMETERPROB: |
| 522 | return 0; |
| 523 | |
| 524 | case ICMP_DEST_UNREACH: |
| 525 | switch (code) { |
| 526 | case ICMP_SR_FAILED: |
| 527 | case ICMP_PORT_UNREACH: |
| 528 | /* Impossible event. */ |
| 529 | return 0; |
| 530 | default: |
| 531 | /* All others are translated to HOST_UNREACH. |
| 532 | rfc2003 contains "deep thoughts" about NET_UNREACH, |
| 533 | I believe they are just ether pollution. --ANK |
| 534 | */ |
| 535 | break; |
| 536 | } |
| 537 | break; |
| 538 | case ICMP_TIME_EXCEEDED: |
| 539 | if (code != ICMP_EXC_TTL) |
| 540 | return 0; |
| 541 | break; |
| 542 | } |
| 543 | |
| 544 | err = -ENOENT; |
| 545 | |
| 546 | rcu_read_lock(); |
| 547 | t = ipip6_tunnel_lookup(dev_net(skb->dev), |
| 548 | skb->dev, |
| 549 | iph->daddr, |
| 550 | iph->saddr); |
| 551 | if (t == NULL) |
| 552 | goto out; |
| 553 | |
| 554 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { |
| 555 | ipv4_update_pmtu(skb, dev_net(skb->dev), info, |
| 556 | t->dev->ifindex, 0, IPPROTO_IPV6, 0); |
| 557 | err = 0; |
| 558 | goto out; |
| 559 | } |
| 560 | |
| 561 | if (t->parms.iph.daddr == 0) |
| 562 | goto out; |
| 563 | |
| 564 | err = 0; |
| 565 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) |
| 566 | goto out; |
| 567 | |
| 568 | if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) |
| 569 | t->err_count++; |
| 570 | else |
| 571 | t->err_count = 1; |
| 572 | t->err_time = jiffies; |
| 573 | out: |
| 574 | rcu_read_unlock(); |
| 575 | return err; |
| 576 | } |
| 577 | |
| 578 | static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb) |
| 579 | { |
| 580 | if (INET_ECN_is_ce(iph->tos)) |
| 581 | IP6_ECN_set_ce(ipv6_hdr(skb)); |
| 582 | } |
| 583 | |
| 584 | static int ipip6_rcv(struct sk_buff *skb) |
| 585 | { |
| 586 | const struct iphdr *iph; |
| 587 | struct ip_tunnel *tunnel; |
| 588 | |
| 589 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) |
| 590 | goto out; |
| 591 | |
| 592 | iph = ip_hdr(skb); |
| 593 | |
| 594 | rcu_read_lock(); |
| 595 | tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev, |
| 596 | iph->saddr, iph->daddr); |
| 597 | if (tunnel != NULL) { |
| 598 | struct pcpu_tstats *tstats; |
| 599 | |
| 600 | secpath_reset(skb); |
| 601 | skb->mac_header = skb->network_header; |
| 602 | skb_reset_network_header(skb); |
| 603 | IPCB(skb)->flags = 0; |
| 604 | skb->protocol = htons(ETH_P_IPV6); |
| 605 | skb->pkt_type = PACKET_HOST; |
| 606 | |
| 607 | if ((tunnel->dev->priv_flags & IFF_ISATAP) && |
| 608 | !isatap_chksrc(skb, iph, tunnel)) { |
| 609 | tunnel->dev->stats.rx_errors++; |
| 610 | rcu_read_unlock(); |
| 611 | kfree_skb(skb); |
| 612 | return 0; |
| 613 | } |
| 614 | |
| 615 | tstats = this_cpu_ptr(tunnel->dev->tstats); |
| 616 | tstats->rx_packets++; |
| 617 | tstats->rx_bytes += skb->len; |
| 618 | |
| 619 | __skb_tunnel_rx(skb, tunnel->dev); |
| 620 | |
| 621 | ipip6_ecn_decapsulate(iph, skb); |
| 622 | |
| 623 | netif_rx(skb); |
| 624 | |
| 625 | rcu_read_unlock(); |
| 626 | return 0; |
| 627 | } |
| 628 | |
| 629 | /* no tunnel matched, let upstream know, ipsec may handle it */ |
| 630 | rcu_read_unlock(); |
| 631 | return 1; |
| 632 | out: |
| 633 | kfree_skb(skb); |
| 634 | return 0; |
| 635 | } |
| 636 | |
| 637 | /* |
| 638 | * Returns the embedded IPv4 address if the IPv6 address |
| 639 | * comes from 6rd / 6to4 (RFC 3056) addr space. |
| 640 | */ |
| 641 | static inline |
| 642 | __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel) |
| 643 | { |
| 644 | __be32 dst = 0; |
| 645 | |
| 646 | #ifdef CONFIG_IPV6_SIT_6RD |
| 647 | if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix, |
| 648 | tunnel->ip6rd.prefixlen)) { |
| 649 | unsigned int pbw0, pbi0; |
| 650 | int pbi1; |
| 651 | u32 d; |
| 652 | |
| 653 | pbw0 = tunnel->ip6rd.prefixlen >> 5; |
| 654 | pbi0 = tunnel->ip6rd.prefixlen & 0x1f; |
| 655 | |
| 656 | d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >> |
| 657 | tunnel->ip6rd.relay_prefixlen; |
| 658 | |
| 659 | pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen; |
| 660 | if (pbi1 > 0) |
| 661 | d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >> |
| 662 | (32 - pbi1); |
| 663 | |
| 664 | dst = tunnel->ip6rd.relay_prefix | htonl(d); |
| 665 | } |
| 666 | #else |
| 667 | if (v6dst->s6_addr16[0] == htons(0x2002)) { |
| 668 | /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ |
| 669 | memcpy(&dst, &v6dst->s6_addr16[1], 4); |
| 670 | } |
| 671 | #endif |
| 672 | return dst; |
| 673 | } |
| 674 | |
| 675 | /* |
| 676 | * This function assumes it is being called from dev_queue_xmit() |
| 677 | * and that skb is filled properly by that function. |
| 678 | */ |
| 679 | |
| 680 | static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb, |
| 681 | struct net_device *dev) |
| 682 | { |
| 683 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 684 | struct pcpu_tstats *tstats; |
| 685 | const struct iphdr *tiph = &tunnel->parms.iph; |
| 686 | const struct ipv6hdr *iph6 = ipv6_hdr(skb); |
| 687 | u8 tos = tunnel->parms.iph.tos; |
| 688 | __be16 df = tiph->frag_off; |
| 689 | struct rtable *rt; /* Route to the other host */ |
| 690 | struct net_device *tdev; /* Device to other host */ |
| 691 | struct iphdr *iph; /* Our new IP header */ |
| 692 | unsigned int max_headroom; /* The extra header space needed */ |
| 693 | __be32 dst = tiph->daddr; |
| 694 | struct flowi4 fl4; |
| 695 | int mtu; |
| 696 | const struct in6_addr *addr6; |
| 697 | int addr_type; |
| 698 | |
| 699 | if (skb->protocol != htons(ETH_P_IPV6)) |
| 700 | goto tx_error; |
| 701 | |
| 702 | if (tos == 1) |
| 703 | tos = ipv6_get_dsfield(iph6); |
| 704 | |
| 705 | /* ISATAP (RFC4214) - must come before 6to4 */ |
| 706 | if (dev->priv_flags & IFF_ISATAP) { |
| 707 | struct neighbour *neigh = NULL; |
| 708 | bool do_tx_error = false; |
| 709 | |
| 710 | if (skb_dst(skb)) |
| 711 | neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); |
| 712 | |
| 713 | if (neigh == NULL) { |
| 714 | net_dbg_ratelimited("sit: nexthop == NULL\n"); |
| 715 | goto tx_error; |
| 716 | } |
| 717 | |
| 718 | addr6 = (const struct in6_addr *)&neigh->primary_key; |
| 719 | addr_type = ipv6_addr_type(addr6); |
| 720 | |
| 721 | if ((addr_type & IPV6_ADDR_UNICAST) && |
| 722 | ipv6_addr_is_isatap(addr6)) |
| 723 | dst = addr6->s6_addr32[3]; |
| 724 | else |
| 725 | do_tx_error = true; |
| 726 | |
| 727 | neigh_release(neigh); |
| 728 | if (do_tx_error) |
| 729 | goto tx_error; |
| 730 | } |
| 731 | |
| 732 | if (!dst) |
| 733 | dst = try_6rd(&iph6->daddr, tunnel); |
| 734 | |
| 735 | if (!dst) { |
| 736 | struct neighbour *neigh = NULL; |
| 737 | bool do_tx_error = false; |
| 738 | |
| 739 | if (skb_dst(skb)) |
| 740 | neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr); |
| 741 | |
| 742 | if (neigh == NULL) { |
| 743 | net_dbg_ratelimited("sit: nexthop == NULL\n"); |
| 744 | goto tx_error; |
| 745 | } |
| 746 | |
| 747 | addr6 = (const struct in6_addr *)&neigh->primary_key; |
| 748 | addr_type = ipv6_addr_type(addr6); |
| 749 | |
| 750 | if (addr_type == IPV6_ADDR_ANY) { |
| 751 | addr6 = &ipv6_hdr(skb)->daddr; |
| 752 | addr_type = ipv6_addr_type(addr6); |
| 753 | } |
| 754 | |
| 755 | if ((addr_type & IPV6_ADDR_COMPATv4) != 0) |
| 756 | dst = addr6->s6_addr32[3]; |
| 757 | else |
| 758 | do_tx_error = true; |
| 759 | |
| 760 | neigh_release(neigh); |
| 761 | if (do_tx_error) |
| 762 | goto tx_error; |
| 763 | } |
| 764 | |
| 765 | rt = ip_route_output_ports(dev_net(dev), &fl4, NULL, |
| 766 | dst, tiph->saddr, |
| 767 | 0, 0, |
| 768 | IPPROTO_IPV6, RT_TOS(tos), |
| 769 | tunnel->parms.link); |
| 770 | if (IS_ERR(rt)) { |
| 771 | dev->stats.tx_carrier_errors++; |
| 772 | goto tx_error_icmp; |
| 773 | } |
| 774 | if (rt->rt_type != RTN_UNICAST) { |
| 775 | ip_rt_put(rt); |
| 776 | dev->stats.tx_carrier_errors++; |
| 777 | goto tx_error_icmp; |
| 778 | } |
| 779 | tdev = rt->dst.dev; |
| 780 | |
| 781 | if (tdev == dev) { |
| 782 | ip_rt_put(rt); |
| 783 | dev->stats.collisions++; |
| 784 | goto tx_error; |
| 785 | } |
| 786 | |
| 787 | if (df) { |
| 788 | mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr); |
| 789 | |
| 790 | if (mtu < 68) { |
| 791 | dev->stats.collisions++; |
| 792 | ip_rt_put(rt); |
| 793 | goto tx_error; |
| 794 | } |
| 795 | |
| 796 | if (mtu < IPV6_MIN_MTU) { |
| 797 | mtu = IPV6_MIN_MTU; |
| 798 | df = 0; |
| 799 | } |
| 800 | |
| 801 | if (tunnel->parms.iph.daddr && skb_dst(skb)) |
| 802 | skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu); |
| 803 | |
| 804 | if (skb->len > mtu) { |
| 805 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); |
| 806 | ip_rt_put(rt); |
| 807 | goto tx_error; |
| 808 | } |
| 809 | } |
| 810 | |
| 811 | if (tunnel->err_count > 0) { |
| 812 | if (time_before(jiffies, |
| 813 | tunnel->err_time + IPTUNNEL_ERR_TIMEO)) { |
| 814 | tunnel->err_count--; |
| 815 | dst_link_failure(skb); |
| 816 | } else |
| 817 | tunnel->err_count = 0; |
| 818 | } |
| 819 | |
| 820 | /* |
| 821 | * Okay, now see if we can stuff it in the buffer as-is. |
| 822 | */ |
| 823 | max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr); |
| 824 | |
| 825 | if (skb_headroom(skb) < max_headroom || skb_shared(skb) || |
| 826 | (skb_cloned(skb) && !skb_clone_writable(skb, 0))) { |
| 827 | struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); |
| 828 | if (!new_skb) { |
| 829 | ip_rt_put(rt); |
| 830 | dev->stats.tx_dropped++; |
| 831 | dev_kfree_skb(skb); |
| 832 | return NETDEV_TX_OK; |
| 833 | } |
| 834 | if (skb->sk) |
| 835 | skb_set_owner_w(new_skb, skb->sk); |
| 836 | dev_kfree_skb(skb); |
| 837 | skb = new_skb; |
| 838 | iph6 = ipv6_hdr(skb); |
| 839 | } |
| 840 | |
| 841 | skb->transport_header = skb->network_header; |
| 842 | skb_push(skb, sizeof(struct iphdr)); |
| 843 | skb_reset_network_header(skb); |
| 844 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
| 845 | IPCB(skb)->flags = 0; |
| 846 | skb_dst_drop(skb); |
| 847 | skb_dst_set(skb, &rt->dst); |
| 848 | |
| 849 | /* |
| 850 | * Push down and install the IPIP header. |
| 851 | */ |
| 852 | |
| 853 | iph = ip_hdr(skb); |
| 854 | iph->version = 4; |
| 855 | iph->ihl = sizeof(struct iphdr)>>2; |
| 856 | iph->frag_off = df; |
| 857 | iph->protocol = IPPROTO_IPV6; |
| 858 | iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); |
| 859 | iph->daddr = fl4.daddr; |
| 860 | iph->saddr = fl4.saddr; |
| 861 | |
| 862 | if ((iph->ttl = tiph->ttl) == 0) |
| 863 | iph->ttl = iph6->hop_limit; |
| 864 | |
| 865 | nf_reset(skb); |
| 866 | tstats = this_cpu_ptr(dev->tstats); |
| 867 | __IPTUNNEL_XMIT(tstats, &dev->stats); |
| 868 | return NETDEV_TX_OK; |
| 869 | |
| 870 | tx_error_icmp: |
| 871 | dst_link_failure(skb); |
| 872 | tx_error: |
| 873 | dev->stats.tx_errors++; |
| 874 | dev_kfree_skb(skb); |
| 875 | return NETDEV_TX_OK; |
| 876 | } |
| 877 | |
| 878 | static void ipip6_tunnel_bind_dev(struct net_device *dev) |
| 879 | { |
| 880 | struct net_device *tdev = NULL; |
| 881 | struct ip_tunnel *tunnel; |
| 882 | const struct iphdr *iph; |
| 883 | struct flowi4 fl4; |
| 884 | |
| 885 | tunnel = netdev_priv(dev); |
| 886 | iph = &tunnel->parms.iph; |
| 887 | |
| 888 | if (iph->daddr) { |
| 889 | struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL, |
| 890 | iph->daddr, iph->saddr, |
| 891 | 0, 0, |
| 892 | IPPROTO_IPV6, |
| 893 | RT_TOS(iph->tos), |
| 894 | tunnel->parms.link); |
| 895 | |
| 896 | if (!IS_ERR(rt)) { |
| 897 | tdev = rt->dst.dev; |
| 898 | ip_rt_put(rt); |
| 899 | } |
| 900 | dev->flags |= IFF_POINTOPOINT; |
| 901 | } |
| 902 | |
| 903 | if (!tdev && tunnel->parms.link) |
| 904 | tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link); |
| 905 | |
| 906 | if (tdev) { |
| 907 | dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); |
| 908 | dev->mtu = tdev->mtu - sizeof(struct iphdr); |
| 909 | if (dev->mtu < IPV6_MIN_MTU) |
| 910 | dev->mtu = IPV6_MIN_MTU; |
| 911 | } |
| 912 | dev->iflink = tunnel->parms.link; |
| 913 | } |
| 914 | |
| 915 | static int |
| 916 | ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) |
| 917 | { |
| 918 | int err = 0; |
| 919 | struct ip_tunnel_parm p; |
| 920 | struct ip_tunnel_prl prl; |
| 921 | struct ip_tunnel *t; |
| 922 | struct net *net = dev_net(dev); |
| 923 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 924 | #ifdef CONFIG_IPV6_SIT_6RD |
| 925 | struct ip_tunnel_6rd ip6rd; |
| 926 | #endif |
| 927 | |
| 928 | switch (cmd) { |
| 929 | case SIOCGETTUNNEL: |
| 930 | #ifdef CONFIG_IPV6_SIT_6RD |
| 931 | case SIOCGET6RD: |
| 932 | #endif |
| 933 | t = NULL; |
| 934 | if (dev == sitn->fb_tunnel_dev) { |
| 935 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { |
| 936 | err = -EFAULT; |
| 937 | break; |
| 938 | } |
| 939 | t = ipip6_tunnel_locate(net, &p, 0); |
| 940 | } |
| 941 | if (t == NULL) |
| 942 | t = netdev_priv(dev); |
| 943 | |
| 944 | err = -EFAULT; |
| 945 | if (cmd == SIOCGETTUNNEL) { |
| 946 | memcpy(&p, &t->parms, sizeof(p)); |
| 947 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, |
| 948 | sizeof(p))) |
| 949 | goto done; |
| 950 | #ifdef CONFIG_IPV6_SIT_6RD |
| 951 | } else { |
| 952 | ip6rd.prefix = t->ip6rd.prefix; |
| 953 | ip6rd.relay_prefix = t->ip6rd.relay_prefix; |
| 954 | ip6rd.prefixlen = t->ip6rd.prefixlen; |
| 955 | ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen; |
| 956 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd, |
| 957 | sizeof(ip6rd))) |
| 958 | goto done; |
| 959 | #endif |
| 960 | } |
| 961 | err = 0; |
| 962 | break; |
| 963 | |
| 964 | case SIOCADDTUNNEL: |
| 965 | case SIOCCHGTUNNEL: |
| 966 | err = -EPERM; |
| 967 | if (!capable(CAP_NET_ADMIN)) |
| 968 | goto done; |
| 969 | |
| 970 | err = -EFAULT; |
| 971 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
| 972 | goto done; |
| 973 | |
| 974 | err = -EINVAL; |
| 975 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 || |
| 976 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) |
| 977 | goto done; |
| 978 | if (p.iph.ttl) |
| 979 | p.iph.frag_off |= htons(IP_DF); |
| 980 | |
| 981 | t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL); |
| 982 | |
| 983 | if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { |
| 984 | if (t != NULL) { |
| 985 | if (t->dev != dev) { |
| 986 | err = -EEXIST; |
| 987 | break; |
| 988 | } |
| 989 | } else { |
| 990 | if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || |
| 991 | (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { |
| 992 | err = -EINVAL; |
| 993 | break; |
| 994 | } |
| 995 | t = netdev_priv(dev); |
| 996 | ipip6_tunnel_unlink(sitn, t); |
| 997 | synchronize_net(); |
| 998 | t->parms.iph.saddr = p.iph.saddr; |
| 999 | t->parms.iph.daddr = p.iph.daddr; |
| 1000 | memcpy(dev->dev_addr, &p.iph.saddr, 4); |
| 1001 | memcpy(dev->broadcast, &p.iph.daddr, 4); |
| 1002 | ipip6_tunnel_link(sitn, t); |
| 1003 | netdev_state_change(dev); |
| 1004 | } |
| 1005 | } |
| 1006 | |
| 1007 | if (t) { |
| 1008 | err = 0; |
| 1009 | if (cmd == SIOCCHGTUNNEL) { |
| 1010 | t->parms.iph.ttl = p.iph.ttl; |
| 1011 | t->parms.iph.tos = p.iph.tos; |
| 1012 | if (t->parms.link != p.link) { |
| 1013 | t->parms.link = p.link; |
| 1014 | ipip6_tunnel_bind_dev(dev); |
| 1015 | netdev_state_change(dev); |
| 1016 | } |
| 1017 | } |
| 1018 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) |
| 1019 | err = -EFAULT; |
| 1020 | } else |
| 1021 | err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); |
| 1022 | break; |
| 1023 | |
| 1024 | case SIOCDELTUNNEL: |
| 1025 | err = -EPERM; |
| 1026 | if (!capable(CAP_NET_ADMIN)) |
| 1027 | goto done; |
| 1028 | |
| 1029 | if (dev == sitn->fb_tunnel_dev) { |
| 1030 | err = -EFAULT; |
| 1031 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
| 1032 | goto done; |
| 1033 | err = -ENOENT; |
| 1034 | if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL) |
| 1035 | goto done; |
| 1036 | err = -EPERM; |
| 1037 | if (t == netdev_priv(sitn->fb_tunnel_dev)) |
| 1038 | goto done; |
| 1039 | dev = t->dev; |
| 1040 | } |
| 1041 | unregister_netdevice(dev); |
| 1042 | err = 0; |
| 1043 | break; |
| 1044 | |
| 1045 | case SIOCGETPRL: |
| 1046 | err = -EINVAL; |
| 1047 | if (dev == sitn->fb_tunnel_dev) |
| 1048 | goto done; |
| 1049 | err = -ENOENT; |
| 1050 | if (!(t = netdev_priv(dev))) |
| 1051 | goto done; |
| 1052 | err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data); |
| 1053 | break; |
| 1054 | |
| 1055 | case SIOCADDPRL: |
| 1056 | case SIOCDELPRL: |
| 1057 | case SIOCCHGPRL: |
| 1058 | err = -EPERM; |
| 1059 | if (!capable(CAP_NET_ADMIN)) |
| 1060 | goto done; |
| 1061 | err = -EINVAL; |
| 1062 | if (dev == sitn->fb_tunnel_dev) |
| 1063 | goto done; |
| 1064 | err = -EFAULT; |
| 1065 | if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl))) |
| 1066 | goto done; |
| 1067 | err = -ENOENT; |
| 1068 | if (!(t = netdev_priv(dev))) |
| 1069 | goto done; |
| 1070 | |
| 1071 | switch (cmd) { |
| 1072 | case SIOCDELPRL: |
| 1073 | err = ipip6_tunnel_del_prl(t, &prl); |
| 1074 | break; |
| 1075 | case SIOCADDPRL: |
| 1076 | case SIOCCHGPRL: |
| 1077 | err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL); |
| 1078 | break; |
| 1079 | } |
| 1080 | netdev_state_change(dev); |
| 1081 | break; |
| 1082 | |
| 1083 | #ifdef CONFIG_IPV6_SIT_6RD |
| 1084 | case SIOCADD6RD: |
| 1085 | case SIOCCHG6RD: |
| 1086 | case SIOCDEL6RD: |
| 1087 | err = -EPERM; |
| 1088 | if (!capable(CAP_NET_ADMIN)) |
| 1089 | goto done; |
| 1090 | |
| 1091 | err = -EFAULT; |
| 1092 | if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data, |
| 1093 | sizeof(ip6rd))) |
| 1094 | goto done; |
| 1095 | |
| 1096 | t = netdev_priv(dev); |
| 1097 | |
| 1098 | if (cmd != SIOCDEL6RD) { |
| 1099 | struct in6_addr prefix; |
| 1100 | __be32 relay_prefix; |
| 1101 | |
| 1102 | err = -EINVAL; |
| 1103 | if (ip6rd.relay_prefixlen > 32 || |
| 1104 | ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64) |
| 1105 | goto done; |
| 1106 | |
| 1107 | ipv6_addr_prefix(&prefix, &ip6rd.prefix, |
| 1108 | ip6rd.prefixlen); |
| 1109 | if (!ipv6_addr_equal(&prefix, &ip6rd.prefix)) |
| 1110 | goto done; |
| 1111 | if (ip6rd.relay_prefixlen) |
| 1112 | relay_prefix = ip6rd.relay_prefix & |
| 1113 | htonl(0xffffffffUL << |
| 1114 | (32 - ip6rd.relay_prefixlen)); |
| 1115 | else |
| 1116 | relay_prefix = 0; |
| 1117 | if (relay_prefix != ip6rd.relay_prefix) |
| 1118 | goto done; |
| 1119 | |
| 1120 | t->ip6rd.prefix = prefix; |
| 1121 | t->ip6rd.relay_prefix = relay_prefix; |
| 1122 | t->ip6rd.prefixlen = ip6rd.prefixlen; |
| 1123 | t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen; |
| 1124 | } else |
| 1125 | ipip6_tunnel_clone_6rd(dev, sitn); |
| 1126 | |
| 1127 | err = 0; |
| 1128 | break; |
| 1129 | #endif |
| 1130 | |
| 1131 | default: |
| 1132 | err = -EINVAL; |
| 1133 | } |
| 1134 | |
| 1135 | done: |
| 1136 | return err; |
| 1137 | } |
| 1138 | |
| 1139 | static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu) |
| 1140 | { |
| 1141 | if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr)) |
| 1142 | return -EINVAL; |
| 1143 | dev->mtu = new_mtu; |
| 1144 | return 0; |
| 1145 | } |
| 1146 | |
| 1147 | static const struct net_device_ops ipip6_netdev_ops = { |
| 1148 | .ndo_uninit = ipip6_tunnel_uninit, |
| 1149 | .ndo_start_xmit = ipip6_tunnel_xmit, |
| 1150 | .ndo_do_ioctl = ipip6_tunnel_ioctl, |
| 1151 | .ndo_change_mtu = ipip6_tunnel_change_mtu, |
| 1152 | .ndo_get_stats64= ipip6_get_stats64, |
| 1153 | }; |
| 1154 | |
| 1155 | static void ipip6_dev_free(struct net_device *dev) |
| 1156 | { |
| 1157 | free_percpu(dev->tstats); |
| 1158 | free_netdev(dev); |
| 1159 | } |
| 1160 | |
| 1161 | static void ipip6_tunnel_setup(struct net_device *dev) |
| 1162 | { |
| 1163 | dev->netdev_ops = &ipip6_netdev_ops; |
| 1164 | dev->destructor = ipip6_dev_free; |
| 1165 | |
| 1166 | dev->type = ARPHRD_SIT; |
| 1167 | dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr); |
| 1168 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr); |
| 1169 | dev->flags = IFF_NOARP; |
| 1170 | dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
| 1171 | dev->iflink = 0; |
| 1172 | dev->addr_len = 4; |
| 1173 | dev->features |= NETIF_F_NETNS_LOCAL; |
| 1174 | dev->features |= NETIF_F_LLTX; |
| 1175 | } |
| 1176 | |
| 1177 | static int ipip6_tunnel_init(struct net_device *dev) |
| 1178 | { |
| 1179 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 1180 | |
| 1181 | tunnel->dev = dev; |
| 1182 | |
| 1183 | memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); |
| 1184 | memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); |
| 1185 | |
| 1186 | ipip6_tunnel_bind_dev(dev); |
| 1187 | dev->tstats = alloc_percpu(struct pcpu_tstats); |
| 1188 | if (!dev->tstats) |
| 1189 | return -ENOMEM; |
| 1190 | |
| 1191 | return 0; |
| 1192 | } |
| 1193 | |
| 1194 | static int __net_init ipip6_fb_tunnel_init(struct net_device *dev) |
| 1195 | { |
| 1196 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 1197 | struct iphdr *iph = &tunnel->parms.iph; |
| 1198 | struct net *net = dev_net(dev); |
| 1199 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 1200 | |
| 1201 | tunnel->dev = dev; |
| 1202 | strcpy(tunnel->parms.name, dev->name); |
| 1203 | |
| 1204 | iph->version = 4; |
| 1205 | iph->protocol = IPPROTO_IPV6; |
| 1206 | iph->ihl = 5; |
| 1207 | iph->ttl = 64; |
| 1208 | |
| 1209 | dev->tstats = alloc_percpu(struct pcpu_tstats); |
| 1210 | if (!dev->tstats) |
| 1211 | return -ENOMEM; |
| 1212 | dev_hold(dev); |
| 1213 | rcu_assign_pointer(sitn->tunnels_wc[0], tunnel); |
| 1214 | return 0; |
| 1215 | } |
| 1216 | |
| 1217 | static struct xfrm_tunnel sit_handler __read_mostly = { |
| 1218 | .handler = ipip6_rcv, |
| 1219 | .err_handler = ipip6_err, |
| 1220 | .priority = 1, |
| 1221 | }; |
| 1222 | |
| 1223 | static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head) |
| 1224 | { |
| 1225 | int prio; |
| 1226 | |
| 1227 | for (prio = 1; prio < 4; prio++) { |
| 1228 | int h; |
| 1229 | for (h = 0; h < HASH_SIZE; h++) { |
| 1230 | struct ip_tunnel *t; |
| 1231 | |
| 1232 | t = rtnl_dereference(sitn->tunnels[prio][h]); |
| 1233 | while (t != NULL) { |
| 1234 | unregister_netdevice_queue(t->dev, head); |
| 1235 | t = rtnl_dereference(t->next); |
| 1236 | } |
| 1237 | } |
| 1238 | } |
| 1239 | } |
| 1240 | |
| 1241 | static int __net_init sit_init_net(struct net *net) |
| 1242 | { |
| 1243 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 1244 | struct ip_tunnel *t; |
| 1245 | int err; |
| 1246 | |
| 1247 | sitn->tunnels[0] = sitn->tunnels_wc; |
| 1248 | sitn->tunnels[1] = sitn->tunnels_l; |
| 1249 | sitn->tunnels[2] = sitn->tunnels_r; |
| 1250 | sitn->tunnels[3] = sitn->tunnels_r_l; |
| 1251 | |
| 1252 | sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", |
| 1253 | ipip6_tunnel_setup); |
| 1254 | if (!sitn->fb_tunnel_dev) { |
| 1255 | err = -ENOMEM; |
| 1256 | goto err_alloc_dev; |
| 1257 | } |
| 1258 | dev_net_set(sitn->fb_tunnel_dev, net); |
| 1259 | |
| 1260 | err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev); |
| 1261 | if (err) |
| 1262 | goto err_dev_free; |
| 1263 | |
| 1264 | ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn); |
| 1265 | |
| 1266 | if ((err = register_netdev(sitn->fb_tunnel_dev))) |
| 1267 | goto err_reg_dev; |
| 1268 | |
| 1269 | t = netdev_priv(sitn->fb_tunnel_dev); |
| 1270 | |
| 1271 | strcpy(t->parms.name, sitn->fb_tunnel_dev->name); |
| 1272 | return 0; |
| 1273 | |
| 1274 | err_reg_dev: |
| 1275 | dev_put(sitn->fb_tunnel_dev); |
| 1276 | err_dev_free: |
| 1277 | ipip6_dev_free(sitn->fb_tunnel_dev); |
| 1278 | err_alloc_dev: |
| 1279 | return err; |
| 1280 | } |
| 1281 | |
| 1282 | static void __net_exit sit_exit_net(struct net *net) |
| 1283 | { |
| 1284 | struct sit_net *sitn = net_generic(net, sit_net_id); |
| 1285 | LIST_HEAD(list); |
| 1286 | |
| 1287 | rtnl_lock(); |
| 1288 | sit_destroy_tunnels(sitn, &list); |
| 1289 | unregister_netdevice_queue(sitn->fb_tunnel_dev, &list); |
| 1290 | unregister_netdevice_many(&list); |
| 1291 | rtnl_unlock(); |
| 1292 | } |
| 1293 | |
| 1294 | static struct pernet_operations sit_net_ops = { |
| 1295 | .init = sit_init_net, |
| 1296 | .exit = sit_exit_net, |
| 1297 | .id = &sit_net_id, |
| 1298 | .size = sizeof(struct sit_net), |
| 1299 | }; |
| 1300 | |
| 1301 | static void __exit sit_cleanup(void) |
| 1302 | { |
| 1303 | xfrm4_tunnel_deregister(&sit_handler, AF_INET6); |
| 1304 | |
| 1305 | unregister_pernet_device(&sit_net_ops); |
| 1306 | rcu_barrier(); /* Wait for completion of call_rcu()'s */ |
| 1307 | } |
| 1308 | |
| 1309 | static int __init sit_init(void) |
| 1310 | { |
| 1311 | int err; |
| 1312 | |
| 1313 | pr_info("IPv6 over IPv4 tunneling driver\n"); |
| 1314 | |
| 1315 | err = register_pernet_device(&sit_net_ops); |
| 1316 | if (err < 0) |
| 1317 | return err; |
| 1318 | err = xfrm4_tunnel_register(&sit_handler, AF_INET6); |
| 1319 | if (err < 0) { |
| 1320 | unregister_pernet_device(&sit_net_ops); |
| 1321 | pr_info("%s: can't add protocol\n", __func__); |
| 1322 | } |
| 1323 | return err; |
| 1324 | } |
| 1325 | |
| 1326 | module_init(sit_init); |
| 1327 | module_exit(sit_cleanup); |
| 1328 | MODULE_LICENSE("GPL"); |
| 1329 | MODULE_ALIAS_NETDEV("sit0"); |