Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | * $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $ |
| 10 | * |
| 11 | * This program is free software; you can redistribute it and/or |
| 12 | * modify it under the terms of the GNU General Public License |
| 13 | * as published by the Free Software Foundation; either version |
| 14 | * 2 of the License, or (at your option) any later version. |
| 15 | * |
| 16 | * Changes: |
| 17 | * Roger Venning <r.venning@telstra.com>: 6to4 support |
| 18 | * Nate Thompson <nate@thebog.net>: 6to4 support |
| 19 | */ |
| 20 | |
| 21 | #include <linux/config.h> |
| 22 | #include <linux/module.h> |
Randy Dunlap | 4fc268d | 2006-01-11 12:17:47 -0800 | [diff] [blame] | 23 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/errno.h> |
| 25 | #include <linux/types.h> |
| 26 | #include <linux/socket.h> |
| 27 | #include <linux/sockios.h> |
| 28 | #include <linux/sched.h> |
| 29 | #include <linux/net.h> |
| 30 | #include <linux/in6.h> |
| 31 | #include <linux/netdevice.h> |
| 32 | #include <linux/if_arp.h> |
| 33 | #include <linux/icmp.h> |
| 34 | #include <asm/uaccess.h> |
| 35 | #include <linux/init.h> |
| 36 | #include <linux/netfilter_ipv4.h> |
Kris Katterjohn | 46f25df | 2006-01-05 16:35:42 -0800 | [diff] [blame] | 37 | #include <linux/if_ether.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | |
| 57 | /* |
| 58 | This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c |
| 59 | |
| 60 | For comments look at net/ipv4/ip_gre.c --ANK |
| 61 | */ |
| 62 | |
| 63 | #define HASH_SIZE 16 |
| 64 | #define HASH(addr) ((addr^(addr>>4))&0xF) |
| 65 | |
| 66 | static int ipip6_fb_tunnel_init(struct net_device *dev); |
| 67 | static int ipip6_tunnel_init(struct net_device *dev); |
| 68 | static void ipip6_tunnel_setup(struct net_device *dev); |
| 69 | |
| 70 | static struct net_device *ipip6_fb_tunnel_dev; |
| 71 | |
| 72 | static struct ip_tunnel *tunnels_r_l[HASH_SIZE]; |
| 73 | static struct ip_tunnel *tunnels_r[HASH_SIZE]; |
| 74 | static struct ip_tunnel *tunnels_l[HASH_SIZE]; |
| 75 | static struct ip_tunnel *tunnels_wc[1]; |
| 76 | static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l }; |
| 77 | |
| 78 | static DEFINE_RWLOCK(ipip6_lock); |
| 79 | |
| 80 | static struct ip_tunnel * ipip6_tunnel_lookup(u32 remote, u32 local) |
| 81 | { |
| 82 | unsigned h0 = HASH(remote); |
| 83 | unsigned h1 = HASH(local); |
| 84 | struct ip_tunnel *t; |
| 85 | |
| 86 | for (t = tunnels_r_l[h0^h1]; t; t = t->next) { |
| 87 | if (local == t->parms.iph.saddr && |
| 88 | remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP)) |
| 89 | return t; |
| 90 | } |
| 91 | for (t = tunnels_r[h0]; t; t = t->next) { |
| 92 | if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP)) |
| 93 | return t; |
| 94 | } |
| 95 | for (t = tunnels_l[h1]; t; t = t->next) { |
| 96 | if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP)) |
| 97 | return t; |
| 98 | } |
| 99 | if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP)) |
| 100 | return t; |
| 101 | return NULL; |
| 102 | } |
| 103 | |
| 104 | static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t) |
| 105 | { |
| 106 | u32 remote = t->parms.iph.daddr; |
| 107 | u32 local = t->parms.iph.saddr; |
| 108 | unsigned h = 0; |
| 109 | int prio = 0; |
| 110 | |
| 111 | if (remote) { |
| 112 | prio |= 2; |
| 113 | h ^= HASH(remote); |
| 114 | } |
| 115 | if (local) { |
| 116 | prio |= 1; |
| 117 | h ^= HASH(local); |
| 118 | } |
| 119 | return &tunnels[prio][h]; |
| 120 | } |
| 121 | |
| 122 | static void ipip6_tunnel_unlink(struct ip_tunnel *t) |
| 123 | { |
| 124 | struct ip_tunnel **tp; |
| 125 | |
| 126 | for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) { |
| 127 | if (t == *tp) { |
| 128 | write_lock_bh(&ipip6_lock); |
| 129 | *tp = t->next; |
| 130 | write_unlock_bh(&ipip6_lock); |
| 131 | break; |
| 132 | } |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | static void ipip6_tunnel_link(struct ip_tunnel *t) |
| 137 | { |
| 138 | struct ip_tunnel **tp = ipip6_bucket(t); |
| 139 | |
| 140 | t->next = *tp; |
| 141 | write_lock_bh(&ipip6_lock); |
| 142 | *tp = t; |
| 143 | write_unlock_bh(&ipip6_lock); |
| 144 | } |
| 145 | |
| 146 | static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create) |
| 147 | { |
| 148 | u32 remote = parms->iph.daddr; |
| 149 | u32 local = parms->iph.saddr; |
| 150 | struct ip_tunnel *t, **tp, *nt; |
| 151 | struct net_device *dev; |
| 152 | unsigned h = 0; |
| 153 | int prio = 0; |
| 154 | char name[IFNAMSIZ]; |
| 155 | |
| 156 | if (remote) { |
| 157 | prio |= 2; |
| 158 | h ^= HASH(remote); |
| 159 | } |
| 160 | if (local) { |
| 161 | prio |= 1; |
| 162 | h ^= HASH(local); |
| 163 | } |
| 164 | for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) { |
| 165 | if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr) |
| 166 | return t; |
| 167 | } |
| 168 | if (!create) |
| 169 | goto failed; |
| 170 | |
| 171 | if (parms->name[0]) |
| 172 | strlcpy(name, parms->name, IFNAMSIZ); |
| 173 | else { |
| 174 | int i; |
| 175 | for (i=1; i<100; i++) { |
| 176 | sprintf(name, "sit%d", i); |
| 177 | if (__dev_get_by_name(name) == NULL) |
| 178 | break; |
| 179 | } |
| 180 | if (i==100) |
| 181 | goto failed; |
| 182 | } |
| 183 | |
| 184 | dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup); |
| 185 | if (dev == NULL) |
| 186 | return NULL; |
| 187 | |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 188 | nt = netdev_priv(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 189 | dev->init = ipip6_tunnel_init; |
| 190 | nt->parms = *parms; |
| 191 | |
| 192 | if (register_netdevice(dev) < 0) { |
| 193 | free_netdev(dev); |
| 194 | goto failed; |
| 195 | } |
| 196 | |
| 197 | dev_hold(dev); |
| 198 | |
| 199 | ipip6_tunnel_link(nt); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 200 | return nt; |
| 201 | |
| 202 | failed: |
| 203 | return NULL; |
| 204 | } |
| 205 | |
| 206 | static void ipip6_tunnel_uninit(struct net_device *dev) |
| 207 | { |
| 208 | if (dev == ipip6_fb_tunnel_dev) { |
| 209 | write_lock_bh(&ipip6_lock); |
| 210 | tunnels_wc[0] = NULL; |
| 211 | write_unlock_bh(&ipip6_lock); |
| 212 | dev_put(dev); |
| 213 | } else { |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 214 | ipip6_tunnel_unlink(netdev_priv(dev)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | dev_put(dev); |
| 216 | } |
| 217 | } |
| 218 | |
| 219 | |
| 220 | static void ipip6_err(struct sk_buff *skb, u32 info) |
| 221 | { |
| 222 | #ifndef I_WISH_WORLD_WERE_PERFECT |
| 223 | |
| 224 | /* It is not :-( All the routers (except for Linux) return only |
| 225 | 8 bytes of packet payload. It means, that precise relaying of |
| 226 | ICMP in the real Internet is absolutely infeasible. |
| 227 | */ |
| 228 | struct iphdr *iph = (struct iphdr*)skb->data; |
| 229 | int type = skb->h.icmph->type; |
| 230 | int code = skb->h.icmph->code; |
| 231 | struct ip_tunnel *t; |
| 232 | |
| 233 | switch (type) { |
| 234 | default: |
| 235 | case ICMP_PARAMETERPROB: |
| 236 | return; |
| 237 | |
| 238 | case ICMP_DEST_UNREACH: |
| 239 | switch (code) { |
| 240 | case ICMP_SR_FAILED: |
| 241 | case ICMP_PORT_UNREACH: |
| 242 | /* Impossible event. */ |
| 243 | return; |
| 244 | case ICMP_FRAG_NEEDED: |
| 245 | /* Soft state for pmtu is maintained by IP core. */ |
| 246 | return; |
| 247 | default: |
| 248 | /* All others are translated to HOST_UNREACH. |
| 249 | rfc2003 contains "deep thoughts" about NET_UNREACH, |
| 250 | I believe they are just ether pollution. --ANK |
| 251 | */ |
| 252 | break; |
| 253 | } |
| 254 | break; |
| 255 | case ICMP_TIME_EXCEEDED: |
| 256 | if (code != ICMP_EXC_TTL) |
| 257 | return; |
| 258 | break; |
| 259 | } |
| 260 | |
| 261 | read_lock(&ipip6_lock); |
| 262 | t = ipip6_tunnel_lookup(iph->daddr, iph->saddr); |
| 263 | if (t == NULL || t->parms.iph.daddr == 0) |
| 264 | goto out; |
| 265 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) |
| 266 | goto out; |
| 267 | |
| 268 | if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO) |
| 269 | t->err_count++; |
| 270 | else |
| 271 | t->err_count = 1; |
| 272 | t->err_time = jiffies; |
| 273 | out: |
| 274 | read_unlock(&ipip6_lock); |
| 275 | return; |
| 276 | #else |
| 277 | struct iphdr *iph = (struct iphdr*)dp; |
| 278 | int hlen = iph->ihl<<2; |
| 279 | struct ipv6hdr *iph6; |
| 280 | int type = skb->h.icmph->type; |
| 281 | int code = skb->h.icmph->code; |
| 282 | int rel_type = 0; |
| 283 | int rel_code = 0; |
| 284 | int rel_info = 0; |
| 285 | struct sk_buff *skb2; |
| 286 | struct rt6_info *rt6i; |
| 287 | |
| 288 | if (len < hlen + sizeof(struct ipv6hdr)) |
| 289 | return; |
| 290 | iph6 = (struct ipv6hdr*)(dp + hlen); |
| 291 | |
| 292 | switch (type) { |
| 293 | default: |
| 294 | return; |
| 295 | case ICMP_PARAMETERPROB: |
| 296 | if (skb->h.icmph->un.gateway < hlen) |
| 297 | return; |
| 298 | |
| 299 | /* So... This guy found something strange INSIDE encapsulated |
| 300 | packet. Well, he is fool, but what can we do ? |
| 301 | */ |
| 302 | rel_type = ICMPV6_PARAMPROB; |
| 303 | rel_info = skb->h.icmph->un.gateway - hlen; |
| 304 | break; |
| 305 | |
| 306 | case ICMP_DEST_UNREACH: |
| 307 | switch (code) { |
| 308 | case ICMP_SR_FAILED: |
| 309 | case ICMP_PORT_UNREACH: |
| 310 | /* Impossible event. */ |
| 311 | return; |
| 312 | case ICMP_FRAG_NEEDED: |
| 313 | /* Too complicated case ... */ |
| 314 | return; |
| 315 | default: |
| 316 | /* All others are translated to HOST_UNREACH. |
| 317 | rfc2003 contains "deep thoughts" about NET_UNREACH, |
| 318 | I believe, it is just ether pollution. --ANK |
| 319 | */ |
| 320 | rel_type = ICMPV6_DEST_UNREACH; |
| 321 | rel_code = ICMPV6_ADDR_UNREACH; |
| 322 | break; |
| 323 | } |
| 324 | break; |
| 325 | case ICMP_TIME_EXCEEDED: |
| 326 | if (code != ICMP_EXC_TTL) |
| 327 | return; |
| 328 | rel_type = ICMPV6_TIME_EXCEED; |
| 329 | rel_code = ICMPV6_EXC_HOPLIMIT; |
| 330 | break; |
| 331 | } |
| 332 | |
| 333 | /* Prepare fake skb to feed it to icmpv6_send */ |
| 334 | skb2 = skb_clone(skb, GFP_ATOMIC); |
| 335 | if (skb2 == NULL) |
| 336 | return; |
| 337 | dst_release(skb2->dst); |
| 338 | skb2->dst = NULL; |
| 339 | skb_pull(skb2, skb->data - (u8*)iph6); |
| 340 | skb2->nh.raw = skb2->data; |
| 341 | |
| 342 | /* Try to guess incoming interface */ |
| 343 | rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0); |
| 344 | if (rt6i && rt6i->rt6i_dev) { |
| 345 | skb2->dev = rt6i->rt6i_dev; |
| 346 | |
| 347 | rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0); |
| 348 | |
| 349 | if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) { |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 350 | struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 351 | if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) { |
| 352 | rel_type = ICMPV6_DEST_UNREACH; |
| 353 | rel_code = ICMPV6_ADDR_UNREACH; |
| 354 | } |
| 355 | icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev); |
| 356 | } |
| 357 | } |
| 358 | kfree_skb(skb2); |
| 359 | return; |
| 360 | #endif |
| 361 | } |
| 362 | |
| 363 | static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb) |
| 364 | { |
| 365 | if (INET_ECN_is_ce(iph->tos)) |
| 366 | IP6_ECN_set_ce(skb->nh.ipv6h); |
| 367 | } |
| 368 | |
| 369 | static int ipip6_rcv(struct sk_buff *skb) |
| 370 | { |
| 371 | struct iphdr *iph; |
| 372 | struct ip_tunnel *tunnel; |
| 373 | |
| 374 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) |
| 375 | goto out; |
| 376 | |
| 377 | iph = skb->nh.iph; |
| 378 | |
| 379 | read_lock(&ipip6_lock); |
| 380 | if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) { |
| 381 | secpath_reset(skb); |
| 382 | skb->mac.raw = skb->nh.raw; |
| 383 | skb->nh.raw = skb->data; |
| 384 | memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options)); |
Patrick McHardy | 8cdfab8 | 2006-01-06 23:04:01 -0800 | [diff] [blame] | 385 | IPCB(skb)->flags = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 386 | skb->protocol = htons(ETH_P_IPV6); |
| 387 | skb->pkt_type = PACKET_HOST; |
| 388 | tunnel->stat.rx_packets++; |
| 389 | tunnel->stat.rx_bytes += skb->len; |
| 390 | skb->dev = tunnel->dev; |
| 391 | dst_release(skb->dst); |
| 392 | skb->dst = NULL; |
| 393 | nf_reset(skb); |
| 394 | ipip6_ecn_decapsulate(iph, skb); |
| 395 | netif_rx(skb); |
| 396 | read_unlock(&ipip6_lock); |
| 397 | return 0; |
| 398 | } |
| 399 | |
Herbert Xu | 45af08b | 2006-04-05 22:31:19 -0700 | [diff] [blame] | 400 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 401 | kfree_skb(skb); |
| 402 | read_unlock(&ipip6_lock); |
| 403 | out: |
| 404 | return 0; |
| 405 | } |
| 406 | |
| 407 | /* Returns the embedded IPv4 address if the IPv6 address |
| 408 | comes from 6to4 (RFC 3056) addr space */ |
| 409 | |
| 410 | static inline u32 try_6to4(struct in6_addr *v6dst) |
| 411 | { |
| 412 | u32 dst = 0; |
| 413 | |
| 414 | if (v6dst->s6_addr16[0] == htons(0x2002)) { |
| 415 | /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */ |
| 416 | memcpy(&dst, &v6dst->s6_addr16[1], 4); |
| 417 | } |
| 418 | return dst; |
| 419 | } |
| 420 | |
| 421 | /* |
| 422 | * This function assumes it is being called from dev_queue_xmit() |
| 423 | * and that skb is filled properly by that function. |
| 424 | */ |
| 425 | |
| 426 | static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) |
| 427 | { |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 428 | struct ip_tunnel *tunnel = netdev_priv(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 429 | struct net_device_stats *stats = &tunnel->stat; |
| 430 | struct iphdr *tiph = &tunnel->parms.iph; |
| 431 | struct ipv6hdr *iph6 = skb->nh.ipv6h; |
| 432 | u8 tos = tunnel->parms.iph.tos; |
| 433 | struct rtable *rt; /* Route to the other host */ |
| 434 | struct net_device *tdev; /* Device to other host */ |
| 435 | struct iphdr *iph; /* Our new IP header */ |
| 436 | int max_headroom; /* The extra header space needed */ |
| 437 | u32 dst = tiph->daddr; |
| 438 | int mtu; |
| 439 | struct in6_addr *addr6; |
| 440 | int addr_type; |
| 441 | |
| 442 | if (tunnel->recursion++) { |
| 443 | tunnel->stat.collisions++; |
| 444 | goto tx_error; |
| 445 | } |
| 446 | |
| 447 | if (skb->protocol != htons(ETH_P_IPV6)) |
| 448 | goto tx_error; |
| 449 | |
| 450 | if (!dst) |
| 451 | dst = try_6to4(&iph6->daddr); |
| 452 | |
| 453 | if (!dst) { |
| 454 | struct neighbour *neigh = NULL; |
| 455 | |
| 456 | if (skb->dst) |
| 457 | neigh = skb->dst->neighbour; |
| 458 | |
| 459 | if (neigh == NULL) { |
| 460 | if (net_ratelimit()) |
| 461 | printk(KERN_DEBUG "sit: nexthop == NULL\n"); |
| 462 | goto tx_error; |
| 463 | } |
| 464 | |
| 465 | addr6 = (struct in6_addr*)&neigh->primary_key; |
| 466 | addr_type = ipv6_addr_type(addr6); |
| 467 | |
| 468 | if (addr_type == IPV6_ADDR_ANY) { |
| 469 | addr6 = &skb->nh.ipv6h->daddr; |
| 470 | addr_type = ipv6_addr_type(addr6); |
| 471 | } |
| 472 | |
| 473 | if ((addr_type & IPV6_ADDR_COMPATv4) == 0) |
| 474 | goto tx_error_icmp; |
| 475 | |
| 476 | dst = addr6->s6_addr32[3]; |
| 477 | } |
| 478 | |
| 479 | { |
| 480 | struct flowi fl = { .nl_u = { .ip4_u = |
| 481 | { .daddr = dst, |
| 482 | .saddr = tiph->saddr, |
| 483 | .tos = RT_TOS(tos) } }, |
| 484 | .oif = tunnel->parms.link, |
| 485 | .proto = IPPROTO_IPV6 }; |
| 486 | if (ip_route_output_key(&rt, &fl)) { |
| 487 | tunnel->stat.tx_carrier_errors++; |
| 488 | goto tx_error_icmp; |
| 489 | } |
| 490 | } |
| 491 | if (rt->rt_type != RTN_UNICAST) { |
| 492 | ip_rt_put(rt); |
| 493 | tunnel->stat.tx_carrier_errors++; |
| 494 | goto tx_error_icmp; |
| 495 | } |
| 496 | tdev = rt->u.dst.dev; |
| 497 | |
| 498 | if (tdev == dev) { |
| 499 | ip_rt_put(rt); |
| 500 | tunnel->stat.collisions++; |
| 501 | goto tx_error; |
| 502 | } |
| 503 | |
| 504 | if (tiph->frag_off) |
| 505 | mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr); |
| 506 | else |
| 507 | mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu; |
| 508 | |
| 509 | if (mtu < 68) { |
| 510 | tunnel->stat.collisions++; |
| 511 | ip_rt_put(rt); |
| 512 | goto tx_error; |
| 513 | } |
| 514 | if (mtu < IPV6_MIN_MTU) |
| 515 | mtu = IPV6_MIN_MTU; |
| 516 | if (tunnel->parms.iph.daddr && skb->dst) |
| 517 | skb->dst->ops->update_pmtu(skb->dst, mtu); |
| 518 | |
| 519 | if (skb->len > mtu) { |
| 520 | icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev); |
| 521 | ip_rt_put(rt); |
| 522 | goto tx_error; |
| 523 | } |
| 524 | |
| 525 | if (tunnel->err_count > 0) { |
| 526 | if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) { |
| 527 | tunnel->err_count--; |
| 528 | dst_link_failure(skb); |
| 529 | } else |
| 530 | tunnel->err_count = 0; |
| 531 | } |
| 532 | |
| 533 | /* |
| 534 | * Okay, now see if we can stuff it in the buffer as-is. |
| 535 | */ |
| 536 | max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr); |
| 537 | |
| 538 | if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) { |
| 539 | struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom); |
| 540 | if (!new_skb) { |
| 541 | ip_rt_put(rt); |
| 542 | stats->tx_dropped++; |
| 543 | dev_kfree_skb(skb); |
| 544 | tunnel->recursion--; |
| 545 | return 0; |
| 546 | } |
| 547 | if (skb->sk) |
| 548 | skb_set_owner_w(new_skb, skb->sk); |
| 549 | dev_kfree_skb(skb); |
| 550 | skb = new_skb; |
| 551 | iph6 = skb->nh.ipv6h; |
| 552 | } |
| 553 | |
| 554 | skb->h.raw = skb->nh.raw; |
| 555 | skb->nh.raw = skb_push(skb, sizeof(struct iphdr)); |
| 556 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
Patrick McHardy | 8cdfab8 | 2006-01-06 23:04:01 -0800 | [diff] [blame] | 557 | IPCB(skb)->flags = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 558 | dst_release(skb->dst); |
| 559 | skb->dst = &rt->u.dst; |
| 560 | |
| 561 | /* |
| 562 | * Push down and install the IPIP header. |
| 563 | */ |
| 564 | |
| 565 | iph = skb->nh.iph; |
| 566 | iph->version = 4; |
| 567 | iph->ihl = sizeof(struct iphdr)>>2; |
| 568 | if (mtu > IPV6_MIN_MTU) |
| 569 | iph->frag_off = htons(IP_DF); |
| 570 | else |
| 571 | iph->frag_off = 0; |
| 572 | |
| 573 | iph->protocol = IPPROTO_IPV6; |
| 574 | iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6)); |
| 575 | iph->daddr = rt->rt_dst; |
| 576 | iph->saddr = rt->rt_src; |
| 577 | |
| 578 | if ((iph->ttl = tiph->ttl) == 0) |
| 579 | iph->ttl = iph6->hop_limit; |
| 580 | |
| 581 | nf_reset(skb); |
| 582 | |
| 583 | IPTUNNEL_XMIT(); |
| 584 | tunnel->recursion--; |
| 585 | return 0; |
| 586 | |
| 587 | tx_error_icmp: |
| 588 | dst_link_failure(skb); |
| 589 | tx_error: |
| 590 | stats->tx_errors++; |
| 591 | dev_kfree_skb(skb); |
| 592 | tunnel->recursion--; |
| 593 | return 0; |
| 594 | } |
| 595 | |
| 596 | static int |
| 597 | ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd) |
| 598 | { |
| 599 | int err = 0; |
| 600 | struct ip_tunnel_parm p; |
| 601 | struct ip_tunnel *t; |
| 602 | |
| 603 | switch (cmd) { |
| 604 | case SIOCGETTUNNEL: |
| 605 | t = NULL; |
| 606 | if (dev == ipip6_fb_tunnel_dev) { |
| 607 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) { |
| 608 | err = -EFAULT; |
| 609 | break; |
| 610 | } |
| 611 | t = ipip6_tunnel_locate(&p, 0); |
| 612 | } |
| 613 | if (t == NULL) |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 614 | t = netdev_priv(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 615 | memcpy(&p, &t->parms, sizeof(p)); |
| 616 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) |
| 617 | err = -EFAULT; |
| 618 | break; |
| 619 | |
| 620 | case SIOCADDTUNNEL: |
| 621 | case SIOCCHGTUNNEL: |
| 622 | err = -EPERM; |
| 623 | if (!capable(CAP_NET_ADMIN)) |
| 624 | goto done; |
| 625 | |
| 626 | err = -EFAULT; |
| 627 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
| 628 | goto done; |
| 629 | |
| 630 | err = -EINVAL; |
| 631 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 || |
| 632 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF))) |
| 633 | goto done; |
| 634 | if (p.iph.ttl) |
| 635 | p.iph.frag_off |= htons(IP_DF); |
| 636 | |
| 637 | t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL); |
| 638 | |
| 639 | if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) { |
| 640 | if (t != NULL) { |
| 641 | if (t->dev != dev) { |
| 642 | err = -EEXIST; |
| 643 | break; |
| 644 | } |
| 645 | } else { |
| 646 | if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) || |
| 647 | (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) { |
| 648 | err = -EINVAL; |
| 649 | break; |
| 650 | } |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 651 | t = netdev_priv(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 652 | ipip6_tunnel_unlink(t); |
| 653 | t->parms.iph.saddr = p.iph.saddr; |
| 654 | t->parms.iph.daddr = p.iph.daddr; |
| 655 | memcpy(dev->dev_addr, &p.iph.saddr, 4); |
| 656 | memcpy(dev->broadcast, &p.iph.daddr, 4); |
| 657 | ipip6_tunnel_link(t); |
| 658 | netdev_state_change(dev); |
| 659 | } |
| 660 | } |
| 661 | |
| 662 | if (t) { |
| 663 | err = 0; |
| 664 | if (cmd == SIOCCHGTUNNEL) { |
| 665 | t->parms.iph.ttl = p.iph.ttl; |
| 666 | t->parms.iph.tos = p.iph.tos; |
| 667 | } |
| 668 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p))) |
| 669 | err = -EFAULT; |
| 670 | } else |
| 671 | err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT); |
| 672 | break; |
| 673 | |
| 674 | case SIOCDELTUNNEL: |
| 675 | err = -EPERM; |
| 676 | if (!capable(CAP_NET_ADMIN)) |
| 677 | goto done; |
| 678 | |
| 679 | if (dev == ipip6_fb_tunnel_dev) { |
| 680 | err = -EFAULT; |
| 681 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
| 682 | goto done; |
| 683 | err = -ENOENT; |
| 684 | if ((t = ipip6_tunnel_locate(&p, 0)) == NULL) |
| 685 | goto done; |
| 686 | err = -EPERM; |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 687 | if (t == netdev_priv(ipip6_fb_tunnel_dev)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 688 | goto done; |
| 689 | dev = t->dev; |
| 690 | } |
| 691 | err = unregister_netdevice(dev); |
| 692 | break; |
| 693 | |
| 694 | default: |
| 695 | err = -EINVAL; |
| 696 | } |
| 697 | |
| 698 | done: |
| 699 | return err; |
| 700 | } |
| 701 | |
| 702 | static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev) |
| 703 | { |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 704 | return &(((struct ip_tunnel*)netdev_priv(dev))->stat); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 705 | } |
| 706 | |
| 707 | static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu) |
| 708 | { |
| 709 | if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr)) |
| 710 | return -EINVAL; |
| 711 | dev->mtu = new_mtu; |
| 712 | return 0; |
| 713 | } |
| 714 | |
| 715 | static void ipip6_tunnel_setup(struct net_device *dev) |
| 716 | { |
| 717 | SET_MODULE_OWNER(dev); |
| 718 | dev->uninit = ipip6_tunnel_uninit; |
| 719 | dev->destructor = free_netdev; |
| 720 | dev->hard_start_xmit = ipip6_tunnel_xmit; |
| 721 | dev->get_stats = ipip6_tunnel_get_stats; |
| 722 | dev->do_ioctl = ipip6_tunnel_ioctl; |
| 723 | dev->change_mtu = ipip6_tunnel_change_mtu; |
| 724 | |
| 725 | dev->type = ARPHRD_SIT; |
| 726 | dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr); |
Kris Katterjohn | 46f25df | 2006-01-05 16:35:42 -0800 | [diff] [blame] | 727 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 728 | dev->flags = IFF_NOARP; |
| 729 | dev->iflink = 0; |
| 730 | dev->addr_len = 4; |
| 731 | } |
| 732 | |
| 733 | static int ipip6_tunnel_init(struct net_device *dev) |
| 734 | { |
| 735 | struct net_device *tdev = NULL; |
| 736 | struct ip_tunnel *tunnel; |
| 737 | struct iphdr *iph; |
| 738 | |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 739 | tunnel = netdev_priv(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 740 | iph = &tunnel->parms.iph; |
| 741 | |
| 742 | tunnel->dev = dev; |
| 743 | strcpy(tunnel->parms.name, dev->name); |
| 744 | |
| 745 | memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4); |
| 746 | memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4); |
| 747 | |
| 748 | if (iph->daddr) { |
| 749 | struct flowi fl = { .nl_u = { .ip4_u = |
| 750 | { .daddr = iph->daddr, |
| 751 | .saddr = iph->saddr, |
| 752 | .tos = RT_TOS(iph->tos) } }, |
| 753 | .oif = tunnel->parms.link, |
| 754 | .proto = IPPROTO_IPV6 }; |
| 755 | struct rtable *rt; |
| 756 | if (!ip_route_output_key(&rt, &fl)) { |
| 757 | tdev = rt->u.dst.dev; |
| 758 | ip_rt_put(rt); |
| 759 | } |
| 760 | dev->flags |= IFF_POINTOPOINT; |
| 761 | } |
| 762 | |
| 763 | if (!tdev && tunnel->parms.link) |
| 764 | tdev = __dev_get_by_index(tunnel->parms.link); |
| 765 | |
| 766 | if (tdev) { |
| 767 | dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); |
| 768 | dev->mtu = tdev->mtu - sizeof(struct iphdr); |
| 769 | if (dev->mtu < IPV6_MIN_MTU) |
| 770 | dev->mtu = IPV6_MIN_MTU; |
| 771 | } |
| 772 | dev->iflink = tunnel->parms.link; |
| 773 | |
| 774 | return 0; |
| 775 | } |
| 776 | |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 777 | static int __init ipip6_fb_tunnel_init(struct net_device *dev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | { |
Patrick McHardy | 2941a48 | 2006-01-08 22:05:26 -0800 | [diff] [blame] | 779 | struct ip_tunnel *tunnel = netdev_priv(dev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 780 | struct iphdr *iph = &tunnel->parms.iph; |
| 781 | |
| 782 | tunnel->dev = dev; |
| 783 | strcpy(tunnel->parms.name, dev->name); |
| 784 | |
| 785 | iph->version = 4; |
| 786 | iph->protocol = IPPROTO_IPV6; |
| 787 | iph->ihl = 5; |
| 788 | iph->ttl = 64; |
| 789 | |
| 790 | dev_hold(dev); |
| 791 | tunnels_wc[0] = tunnel; |
| 792 | return 0; |
| 793 | } |
| 794 | |
| 795 | static struct net_protocol sit_protocol = { |
| 796 | .handler = ipip6_rcv, |
| 797 | .err_handler = ipip6_err, |
| 798 | }; |
| 799 | |
Alexey Kuznetsov | db44575 | 2005-07-30 17:46:44 -0700 | [diff] [blame] | 800 | static void __exit sit_destroy_tunnels(void) |
| 801 | { |
| 802 | int prio; |
| 803 | |
| 804 | for (prio = 1; prio < 4; prio++) { |
| 805 | int h; |
| 806 | for (h = 0; h < HASH_SIZE; h++) { |
| 807 | struct ip_tunnel *t; |
| 808 | while ((t = tunnels[prio][h]) != NULL) |
| 809 | unregister_netdevice(t->dev); |
| 810 | } |
| 811 | } |
| 812 | } |
| 813 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 814 | void __exit sit_cleanup(void) |
| 815 | { |
| 816 | inet_del_protocol(&sit_protocol, IPPROTO_IPV6); |
Alexey Kuznetsov | db44575 | 2005-07-30 17:46:44 -0700 | [diff] [blame] | 817 | |
| 818 | rtnl_lock(); |
| 819 | sit_destroy_tunnels(); |
| 820 | unregister_netdevice(ipip6_fb_tunnel_dev); |
| 821 | rtnl_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | } |
| 823 | |
| 824 | int __init sit_init(void) |
| 825 | { |
| 826 | int err; |
| 827 | |
| 828 | printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n"); |
| 829 | |
| 830 | if (inet_add_protocol(&sit_protocol, IPPROTO_IPV6) < 0) { |
| 831 | printk(KERN_INFO "sit init: Can't add protocol\n"); |
| 832 | return -EAGAIN; |
| 833 | } |
| 834 | |
| 835 | ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0", |
| 836 | ipip6_tunnel_setup); |
| 837 | if (!ipip6_fb_tunnel_dev) { |
| 838 | err = -ENOMEM; |
| 839 | goto err1; |
| 840 | } |
| 841 | |
| 842 | ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init; |
| 843 | |
| 844 | if ((err = register_netdev(ipip6_fb_tunnel_dev))) |
| 845 | goto err2; |
| 846 | |
| 847 | out: |
| 848 | return err; |
| 849 | err2: |
| 850 | free_netdev(ipip6_fb_tunnel_dev); |
| 851 | err1: |
| 852 | inet_del_protocol(&sit_protocol, IPPROTO_IPV6); |
| 853 | goto out; |
| 854 | } |