| 1 | /* |
| 2 | * Linux NET3: GRE over IP protocol decoder. |
| 3 | * |
| 4 | * Authors: Alexey Kuznetsov (kuznet@ms2.inr.ac.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 <asm/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/mroute.h> |
| 28 | #include <linux/init.h> |
| 29 | #include <linux/in6.h> |
| 30 | #include <linux/inetdevice.h> |
| 31 | #include <linux/igmp.h> |
| 32 | #include <linux/netfilter_ipv4.h> |
| 33 | #include <linux/etherdevice.h> |
| 34 | #include <linux/if_ether.h> |
| 35 | |
| 36 | #include <net/sock.h> |
| 37 | #include <net/ip.h> |
| 38 | #include <net/icmp.h> |
| 39 | #include <net/protocol.h> |
| 40 | #include <net/ip_tunnels.h> |
| 41 | #include <net/arp.h> |
| 42 | #include <net/checksum.h> |
| 43 | #include <net/dsfield.h> |
| 44 | #include <net/inet_ecn.h> |
| 45 | #include <net/xfrm.h> |
| 46 | #include <net/net_namespace.h> |
| 47 | #include <net/netns/generic.h> |
| 48 | #include <net/rtnetlink.h> |
| 49 | #include <net/gre.h> |
| 50 | |
| 51 | #if IS_ENABLED(CONFIG_IPV6) |
| 52 | #include <net/ipv6.h> |
| 53 | #include <net/ip6_fib.h> |
| 54 | #include <net/ip6_route.h> |
| 55 | #endif |
| 56 | |
| 57 | /* |
| 58 | Problems & solutions |
| 59 | -------------------- |
| 60 | |
| 61 | 1. The most important issue is detecting local dead loops. |
| 62 | They would cause complete host lockup in transmit, which |
| 63 | would be "resolved" by stack overflow or, if queueing is enabled, |
| 64 | with infinite looping in net_bh. |
| 65 | |
| 66 | We cannot track such dead loops during route installation, |
| 67 | it is infeasible task. The most general solutions would be |
| 68 | to keep skb->encapsulation counter (sort of local ttl), |
| 69 | and silently drop packet when it expires. It is a good |
| 70 | solution, but it supposes maintaining new variable in ALL |
| 71 | skb, even if no tunneling is used. |
| 72 | |
| 73 | Current solution: xmit_recursion breaks dead loops. This is a percpu |
| 74 | counter, since when we enter the first ndo_xmit(), cpu migration is |
| 75 | forbidden. We force an exit if this counter reaches RECURSION_LIMIT |
| 76 | |
| 77 | 2. Networking dead loops would not kill routers, but would really |
| 78 | kill network. IP hop limit plays role of "t->recursion" in this case, |
| 79 | if we copy it from packet being encapsulated to upper header. |
| 80 | It is very good solution, but it introduces two problems: |
| 81 | |
| 82 | - Routing protocols, using packets with ttl=1 (OSPF, RIP2), |
| 83 | do not work over tunnels. |
| 84 | - traceroute does not work. I planned to relay ICMP from tunnel, |
| 85 | so that this problem would be solved and traceroute output |
| 86 | would even more informative. This idea appeared to be wrong: |
| 87 | only Linux complies to rfc1812 now (yes, guys, Linux is the only |
| 88 | true router now :-)), all routers (at least, in neighbourhood of mine) |
| 89 | return only 8 bytes of payload. It is the end. |
| 90 | |
| 91 | Hence, if we want that OSPF worked or traceroute said something reasonable, |
| 92 | we should search for another solution. |
| 93 | |
| 94 | One of them is to parse packet trying to detect inner encapsulation |
| 95 | made by our node. It is difficult or even impossible, especially, |
| 96 | taking into account fragmentation. TO be short, ttl is not solution at all. |
| 97 | |
| 98 | Current solution: The solution was UNEXPECTEDLY SIMPLE. |
| 99 | We force DF flag on tunnels with preconfigured hop limit, |
| 100 | that is ALL. :-) Well, it does not remove the problem completely, |
| 101 | but exponential growth of network traffic is changed to linear |
| 102 | (branches, that exceed pmtu are pruned) and tunnel mtu |
| 103 | rapidly degrades to value <68, where looping stops. |
| 104 | Yes, it is not good if there exists a router in the loop, |
| 105 | which does not force DF, even when encapsulating packets have DF set. |
| 106 | But it is not our problem! Nobody could accuse us, we made |
| 107 | all that we could make. Even if it is your gated who injected |
| 108 | fatal route to network, even if it were you who configured |
| 109 | fatal static route: you are innocent. :-) |
| 110 | |
| 111 | Alexey Kuznetsov. |
| 112 | */ |
| 113 | |
| 114 | static bool log_ecn_error = true; |
| 115 | module_param(log_ecn_error, bool, 0644); |
| 116 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); |
| 117 | |
| 118 | static struct rtnl_link_ops ipgre_link_ops __read_mostly; |
| 119 | static int ipgre_tunnel_init(struct net_device *dev); |
| 120 | |
| 121 | static int ipgre_net_id __read_mostly; |
| 122 | static int gre_tap_net_id __read_mostly; |
| 123 | |
| 124 | static __sum16 check_checksum(struct sk_buff *skb) |
| 125 | { |
| 126 | __sum16 csum = 0; |
| 127 | |
| 128 | switch (skb->ip_summed) { |
| 129 | case CHECKSUM_COMPLETE: |
| 130 | csum = csum_fold(skb->csum); |
| 131 | |
| 132 | if (!csum) |
| 133 | break; |
| 134 | /* Fall through. */ |
| 135 | |
| 136 | case CHECKSUM_NONE: |
| 137 | skb->csum = 0; |
| 138 | csum = __skb_checksum_complete(skb); |
| 139 | skb->ip_summed = CHECKSUM_COMPLETE; |
| 140 | break; |
| 141 | } |
| 142 | |
| 143 | return csum; |
| 144 | } |
| 145 | |
| 146 | static int ip_gre_calc_hlen(__be16 o_flags) |
| 147 | { |
| 148 | int addend = 4; |
| 149 | |
| 150 | if (o_flags&TUNNEL_CSUM) |
| 151 | addend += 4; |
| 152 | if (o_flags&TUNNEL_KEY) |
| 153 | addend += 4; |
| 154 | if (o_flags&TUNNEL_SEQ) |
| 155 | addend += 4; |
| 156 | return addend; |
| 157 | } |
| 158 | |
| 159 | static int parse_gre_header(struct sk_buff *skb, struct tnl_ptk_info *tpi, |
| 160 | bool *csum_err, int *hdr_len) |
| 161 | { |
| 162 | unsigned int ip_hlen = ip_hdrlen(skb); |
| 163 | const struct gre_base_hdr *greh; |
| 164 | __be32 *options; |
| 165 | |
| 166 | if (unlikely(!pskb_may_pull(skb, sizeof(struct gre_base_hdr)))) |
| 167 | return -EINVAL; |
| 168 | |
| 169 | greh = (struct gre_base_hdr *)(skb_network_header(skb) + ip_hlen); |
| 170 | if (unlikely(greh->flags & (GRE_VERSION | GRE_ROUTING))) |
| 171 | return -EINVAL; |
| 172 | |
| 173 | tpi->flags = gre_flags_to_tnl_flags(greh->flags); |
| 174 | *hdr_len = ip_gre_calc_hlen(tpi->flags); |
| 175 | |
| 176 | if (!pskb_may_pull(skb, *hdr_len)) |
| 177 | return -EINVAL; |
| 178 | |
| 179 | greh = (struct gre_base_hdr *)(skb_network_header(skb) + ip_hlen); |
| 180 | |
| 181 | tpi->proto = greh->protocol; |
| 182 | |
| 183 | options = (__be32 *)(greh + 1); |
| 184 | if (greh->flags & GRE_CSUM) { |
| 185 | if (check_checksum(skb)) { |
| 186 | *csum_err = true; |
| 187 | return -EINVAL; |
| 188 | } |
| 189 | options++; |
| 190 | } |
| 191 | |
| 192 | if (greh->flags & GRE_KEY) { |
| 193 | tpi->key = *options; |
| 194 | options++; |
| 195 | } else |
| 196 | tpi->key = 0; |
| 197 | |
| 198 | if (unlikely(greh->flags & GRE_SEQ)) { |
| 199 | tpi->seq = *options; |
| 200 | options++; |
| 201 | } else |
| 202 | tpi->seq = 0; |
| 203 | |
| 204 | /* WCCP version 1 and 2 protocol decoding. |
| 205 | * - Change protocol to IP |
| 206 | * - When dealing with WCCPv2, Skip extra 4 bytes in GRE header |
| 207 | */ |
| 208 | if (greh->flags == 0 && tpi->proto == htons(ETH_P_WCCP)) { |
| 209 | tpi->proto = htons(ETH_P_IP); |
| 210 | if ((*(u8 *)options & 0xF0) != 0x40) { |
| 211 | *hdr_len += 4; |
| 212 | if (!pskb_may_pull(skb, *hdr_len)) |
| 213 | return -EINVAL; |
| 214 | } |
| 215 | } |
| 216 | |
| 217 | return 0; |
| 218 | } |
| 219 | |
| 220 | static void ipgre_err(struct sk_buff *skb, u32 info) |
| 221 | { |
| 222 | |
| 223 | /* All the routers (except for Linux) return only |
| 224 | 8 bytes of packet payload. It means, that precise relaying of |
| 225 | ICMP in the real Internet is absolutely infeasible. |
| 226 | |
| 227 | Moreover, Cisco "wise men" put GRE key to the third word |
| 228 | in GRE header. It makes impossible maintaining even soft |
| 229 | state for keyed GRE tunnels with enabled checksum. Tell |
| 230 | them "thank you". |
| 231 | |
| 232 | Well, I wonder, rfc1812 was written by Cisco employee, |
| 233 | what the hell these idiots break standards established |
| 234 | by themselves??? |
| 235 | */ |
| 236 | struct net *net = dev_net(skb->dev); |
| 237 | struct ip_tunnel_net *itn; |
| 238 | const struct iphdr *iph; |
| 239 | const int type = icmp_hdr(skb)->type; |
| 240 | const int code = icmp_hdr(skb)->code; |
| 241 | struct ip_tunnel *t; |
| 242 | struct tnl_ptk_info tpi; |
| 243 | int hdr_len; |
| 244 | bool csum_err = false; |
| 245 | |
| 246 | if (parse_gre_header(skb, &tpi, &csum_err, &hdr_len)) { |
| 247 | if (!csum_err) /* ignore csum errors. */ |
| 248 | return; |
| 249 | } |
| 250 | |
| 251 | switch (type) { |
| 252 | default: |
| 253 | case ICMP_PARAMETERPROB: |
| 254 | return; |
| 255 | |
| 256 | case ICMP_DEST_UNREACH: |
| 257 | switch (code) { |
| 258 | case ICMP_SR_FAILED: |
| 259 | case ICMP_PORT_UNREACH: |
| 260 | /* Impossible event. */ |
| 261 | return; |
| 262 | default: |
| 263 | /* All others are translated to HOST_UNREACH. |
| 264 | rfc2003 contains "deep thoughts" about NET_UNREACH, |
| 265 | I believe they are just ether pollution. --ANK |
| 266 | */ |
| 267 | break; |
| 268 | } |
| 269 | break; |
| 270 | case ICMP_TIME_EXCEEDED: |
| 271 | if (code != ICMP_EXC_TTL) |
| 272 | return; |
| 273 | break; |
| 274 | |
| 275 | case ICMP_REDIRECT: |
| 276 | break; |
| 277 | } |
| 278 | |
| 279 | if (tpi.proto == htons(ETH_P_TEB)) |
| 280 | itn = net_generic(net, gre_tap_net_id); |
| 281 | else |
| 282 | itn = net_generic(net, ipgre_net_id); |
| 283 | |
| 284 | iph = (const struct iphdr *)skb->data; |
| 285 | t = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi.flags, |
| 286 | iph->daddr, iph->saddr, tpi.key); |
| 287 | |
| 288 | if (t == NULL) |
| 289 | return; |
| 290 | |
| 291 | if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) { |
| 292 | ipv4_update_pmtu(skb, dev_net(skb->dev), info, |
| 293 | t->parms.link, 0, IPPROTO_GRE, 0); |
| 294 | return; |
| 295 | } |
| 296 | if (type == ICMP_REDIRECT) { |
| 297 | ipv4_redirect(skb, dev_net(skb->dev), t->parms.link, 0, |
| 298 | IPPROTO_GRE, 0); |
| 299 | return; |
| 300 | } |
| 301 | if (t->parms.iph.daddr == 0 || |
| 302 | ipv4_is_multicast(t->parms.iph.daddr)) |
| 303 | return; |
| 304 | |
| 305 | if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED) |
| 306 | return; |
| 307 | |
| 308 | if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO)) |
| 309 | t->err_count++; |
| 310 | else |
| 311 | t->err_count = 1; |
| 312 | t->err_time = jiffies; |
| 313 | } |
| 314 | |
| 315 | static int ipgre_rcv(struct sk_buff *skb) |
| 316 | { |
| 317 | struct net *net = dev_net(skb->dev); |
| 318 | struct ip_tunnel_net *itn; |
| 319 | const struct iphdr *iph; |
| 320 | struct ip_tunnel *tunnel; |
| 321 | struct tnl_ptk_info tpi; |
| 322 | int hdr_len; |
| 323 | bool csum_err = false; |
| 324 | |
| 325 | if (parse_gre_header(skb, &tpi, &csum_err, &hdr_len) < 0) |
| 326 | goto drop; |
| 327 | |
| 328 | if (tpi.proto == htons(ETH_P_TEB)) |
| 329 | itn = net_generic(net, gre_tap_net_id); |
| 330 | else |
| 331 | itn = net_generic(net, ipgre_net_id); |
| 332 | |
| 333 | iph = ip_hdr(skb); |
| 334 | tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, tpi.flags, |
| 335 | iph->saddr, iph->daddr, tpi.key); |
| 336 | |
| 337 | if (tunnel) { |
| 338 | skb_pop_mac_header(skb); |
| 339 | ip_tunnel_rcv(tunnel, skb, &tpi, hdr_len, log_ecn_error); |
| 340 | return 0; |
| 341 | } |
| 342 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); |
| 343 | drop: |
| 344 | kfree_skb(skb); |
| 345 | return 0; |
| 346 | } |
| 347 | |
| 348 | static struct sk_buff *handle_offloads(struct ip_tunnel *tunnel, struct sk_buff *skb) |
| 349 | { |
| 350 | int err; |
| 351 | |
| 352 | if (skb_is_gso(skb)) { |
| 353 | err = skb_unclone(skb, GFP_ATOMIC); |
| 354 | if (unlikely(err)) |
| 355 | goto error; |
| 356 | skb_shinfo(skb)->gso_type |= SKB_GSO_GRE; |
| 357 | return skb; |
| 358 | } else if (skb->ip_summed == CHECKSUM_PARTIAL && |
| 359 | tunnel->parms.o_flags&TUNNEL_CSUM) { |
| 360 | err = skb_checksum_help(skb); |
| 361 | if (unlikely(err)) |
| 362 | goto error; |
| 363 | } else if (skb->ip_summed != CHECKSUM_PARTIAL) |
| 364 | skb->ip_summed = CHECKSUM_NONE; |
| 365 | |
| 366 | return skb; |
| 367 | |
| 368 | error: |
| 369 | kfree_skb(skb); |
| 370 | return ERR_PTR(err); |
| 371 | } |
| 372 | |
| 373 | static struct sk_buff *gre_build_header(struct sk_buff *skb, |
| 374 | const struct tnl_ptk_info *tpi, |
| 375 | int hdr_len) |
| 376 | { |
| 377 | struct gre_base_hdr *greh; |
| 378 | |
| 379 | skb_push(skb, hdr_len); |
| 380 | |
| 381 | greh = (struct gre_base_hdr *)skb->data; |
| 382 | greh->flags = tnl_flags_to_gre_flags(tpi->flags); |
| 383 | greh->protocol = tpi->proto; |
| 384 | |
| 385 | if (tpi->flags&(TUNNEL_KEY|TUNNEL_CSUM|TUNNEL_SEQ)) { |
| 386 | __be32 *ptr = (__be32 *)(((u8 *)greh) + hdr_len - 4); |
| 387 | |
| 388 | if (tpi->flags&TUNNEL_SEQ) { |
| 389 | *ptr = tpi->seq; |
| 390 | ptr--; |
| 391 | } |
| 392 | if (tpi->flags&TUNNEL_KEY) { |
| 393 | *ptr = tpi->key; |
| 394 | ptr--; |
| 395 | } |
| 396 | if (tpi->flags&TUNNEL_CSUM && |
| 397 | !(skb_shinfo(skb)->gso_type & SKB_GSO_GRE)) { |
| 398 | *(__sum16 *)ptr = 0; |
| 399 | *(__sum16 *)ptr = csum_fold(skb_checksum(skb, 0, |
| 400 | skb->len, 0)); |
| 401 | } |
| 402 | } |
| 403 | |
| 404 | return skb; |
| 405 | } |
| 406 | |
| 407 | static void __gre_xmit(struct sk_buff *skb, struct net_device *dev, |
| 408 | const struct iphdr *tnl_params, |
| 409 | __be16 proto) |
| 410 | { |
| 411 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 412 | struct tnl_ptk_info tpi; |
| 413 | |
| 414 | if (likely(!skb->encapsulation)) { |
| 415 | skb_reset_inner_headers(skb); |
| 416 | skb->encapsulation = 1; |
| 417 | } |
| 418 | |
| 419 | tpi.flags = tunnel->parms.o_flags; |
| 420 | tpi.proto = proto; |
| 421 | tpi.key = tunnel->parms.o_key; |
| 422 | if (tunnel->parms.o_flags & TUNNEL_SEQ) |
| 423 | tunnel->o_seqno++; |
| 424 | tpi.seq = htonl(tunnel->o_seqno); |
| 425 | |
| 426 | /* Push GRE header. */ |
| 427 | skb = gre_build_header(skb, &tpi, tunnel->hlen); |
| 428 | if (unlikely(!skb)) { |
| 429 | dev->stats.tx_dropped++; |
| 430 | return; |
| 431 | } |
| 432 | |
| 433 | ip_tunnel_xmit(skb, dev, tnl_params); |
| 434 | } |
| 435 | |
| 436 | static netdev_tx_t ipgre_xmit(struct sk_buff *skb, |
| 437 | struct net_device *dev) |
| 438 | { |
| 439 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 440 | const struct iphdr *tnl_params; |
| 441 | |
| 442 | skb = handle_offloads(tunnel, skb); |
| 443 | if (IS_ERR(skb)) |
| 444 | goto out; |
| 445 | |
| 446 | if (dev->header_ops) { |
| 447 | /* Need space for new headers */ |
| 448 | if (skb_cow_head(skb, dev->needed_headroom - |
| 449 | (tunnel->hlen + sizeof(struct iphdr)))) |
| 450 | goto free_skb; |
| 451 | |
| 452 | tnl_params = (const struct iphdr *)skb->data; |
| 453 | |
| 454 | /* Pull skb since ip_tunnel_xmit() needs skb->data pointing |
| 455 | * to gre header. |
| 456 | */ |
| 457 | skb_pull(skb, tunnel->hlen + sizeof(struct iphdr)); |
| 458 | } else { |
| 459 | if (skb_cow_head(skb, dev->needed_headroom)) |
| 460 | goto free_skb; |
| 461 | |
| 462 | tnl_params = &tunnel->parms.iph; |
| 463 | } |
| 464 | |
| 465 | __gre_xmit(skb, dev, tnl_params, skb->protocol); |
| 466 | |
| 467 | return NETDEV_TX_OK; |
| 468 | |
| 469 | free_skb: |
| 470 | dev_kfree_skb(skb); |
| 471 | out: |
| 472 | dev->stats.tx_dropped++; |
| 473 | return NETDEV_TX_OK; |
| 474 | } |
| 475 | |
| 476 | static netdev_tx_t gre_tap_xmit(struct sk_buff *skb, |
| 477 | struct net_device *dev) |
| 478 | { |
| 479 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 480 | |
| 481 | skb = handle_offloads(tunnel, skb); |
| 482 | if (IS_ERR(skb)) |
| 483 | goto out; |
| 484 | |
| 485 | if (skb_cow_head(skb, dev->needed_headroom)) |
| 486 | goto free_skb; |
| 487 | |
| 488 | __gre_xmit(skb, dev, &tunnel->parms.iph, htons(ETH_P_TEB)); |
| 489 | |
| 490 | return NETDEV_TX_OK; |
| 491 | |
| 492 | free_skb: |
| 493 | dev_kfree_skb(skb); |
| 494 | out: |
| 495 | dev->stats.tx_dropped++; |
| 496 | return NETDEV_TX_OK; |
| 497 | } |
| 498 | |
| 499 | static int ipgre_tunnel_ioctl(struct net_device *dev, |
| 500 | struct ifreq *ifr, int cmd) |
| 501 | { |
| 502 | int err = 0; |
| 503 | struct ip_tunnel_parm p; |
| 504 | |
| 505 | if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) |
| 506 | return -EFAULT; |
| 507 | if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) { |
| 508 | if (p.iph.version != 4 || p.iph.protocol != IPPROTO_GRE || |
| 509 | p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)) || |
| 510 | ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))) |
| 511 | return -EINVAL; |
| 512 | } |
| 513 | p.i_flags = gre_flags_to_tnl_flags(p.i_flags); |
| 514 | p.o_flags = gre_flags_to_tnl_flags(p.o_flags); |
| 515 | |
| 516 | err = ip_tunnel_ioctl(dev, &p, cmd); |
| 517 | if (err) |
| 518 | return err; |
| 519 | |
| 520 | p.i_flags = tnl_flags_to_gre_flags(p.i_flags); |
| 521 | p.o_flags = tnl_flags_to_gre_flags(p.o_flags); |
| 522 | |
| 523 | if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) |
| 524 | return -EFAULT; |
| 525 | return 0; |
| 526 | } |
| 527 | |
| 528 | /* Nice toy. Unfortunately, useless in real life :-) |
| 529 | It allows to construct virtual multiprotocol broadcast "LAN" |
| 530 | over the Internet, provided multicast routing is tuned. |
| 531 | |
| 532 | |
| 533 | I have no idea was this bicycle invented before me, |
| 534 | so that I had to set ARPHRD_IPGRE to a random value. |
| 535 | I have an impression, that Cisco could make something similar, |
| 536 | but this feature is apparently missing in IOS<=11.2(8). |
| 537 | |
| 538 | I set up 10.66.66/24 and fec0:6666:6666::0/96 as virtual networks |
| 539 | with broadcast 224.66.66.66. If you have access to mbone, play with me :-) |
| 540 | |
| 541 | ping -t 255 224.66.66.66 |
| 542 | |
| 543 | If nobody answers, mbone does not work. |
| 544 | |
| 545 | ip tunnel add Universe mode gre remote 224.66.66.66 local <Your_real_addr> ttl 255 |
| 546 | ip addr add 10.66.66.<somewhat>/24 dev Universe |
| 547 | ifconfig Universe up |
| 548 | ifconfig Universe add fe80::<Your_real_addr>/10 |
| 549 | ifconfig Universe add fec0:6666:6666::<Your_real_addr>/96 |
| 550 | ftp 10.66.66.66 |
| 551 | ... |
| 552 | ftp fec0:6666:6666::193.233.7.65 |
| 553 | ... |
| 554 | */ |
| 555 | static int ipgre_header(struct sk_buff *skb, struct net_device *dev, |
| 556 | unsigned short type, |
| 557 | const void *daddr, const void *saddr, unsigned int len) |
| 558 | { |
| 559 | struct ip_tunnel *t = netdev_priv(dev); |
| 560 | struct iphdr *iph; |
| 561 | struct gre_base_hdr *greh; |
| 562 | |
| 563 | iph = (struct iphdr *)skb_push(skb, t->hlen + sizeof(*iph)); |
| 564 | greh = (struct gre_base_hdr *)(iph+1); |
| 565 | greh->flags = tnl_flags_to_gre_flags(t->parms.o_flags); |
| 566 | greh->protocol = htons(type); |
| 567 | |
| 568 | memcpy(iph, &t->parms.iph, sizeof(struct iphdr)); |
| 569 | |
| 570 | /* Set the source hardware address. */ |
| 571 | if (saddr) |
| 572 | memcpy(&iph->saddr, saddr, 4); |
| 573 | if (daddr) |
| 574 | memcpy(&iph->daddr, daddr, 4); |
| 575 | if (iph->daddr) |
| 576 | return t->hlen + sizeof(*iph); |
| 577 | |
| 578 | return -(t->hlen + sizeof(*iph)); |
| 579 | } |
| 580 | |
| 581 | static int ipgre_header_parse(const struct sk_buff *skb, unsigned char *haddr) |
| 582 | { |
| 583 | const struct iphdr *iph = (const struct iphdr *) skb_mac_header(skb); |
| 584 | memcpy(haddr, &iph->saddr, 4); |
| 585 | return 4; |
| 586 | } |
| 587 | |
| 588 | static const struct header_ops ipgre_header_ops = { |
| 589 | .create = ipgre_header, |
| 590 | .parse = ipgre_header_parse, |
| 591 | }; |
| 592 | |
| 593 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
| 594 | static int ipgre_open(struct net_device *dev) |
| 595 | { |
| 596 | struct ip_tunnel *t = netdev_priv(dev); |
| 597 | |
| 598 | if (ipv4_is_multicast(t->parms.iph.daddr)) { |
| 599 | struct flowi4 fl4; |
| 600 | struct rtable *rt; |
| 601 | |
| 602 | rt = ip_route_output_gre(dev_net(dev), &fl4, |
| 603 | t->parms.iph.daddr, |
| 604 | t->parms.iph.saddr, |
| 605 | t->parms.o_key, |
| 606 | RT_TOS(t->parms.iph.tos), |
| 607 | t->parms.link); |
| 608 | if (IS_ERR(rt)) |
| 609 | return -EADDRNOTAVAIL; |
| 610 | dev = rt->dst.dev; |
| 611 | ip_rt_put(rt); |
| 612 | if (__in_dev_get_rtnl(dev) == NULL) |
| 613 | return -EADDRNOTAVAIL; |
| 614 | t->mlink = dev->ifindex; |
| 615 | ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr); |
| 616 | } |
| 617 | return 0; |
| 618 | } |
| 619 | |
| 620 | static int ipgre_close(struct net_device *dev) |
| 621 | { |
| 622 | struct ip_tunnel *t = netdev_priv(dev); |
| 623 | |
| 624 | if (ipv4_is_multicast(t->parms.iph.daddr) && t->mlink) { |
| 625 | struct in_device *in_dev; |
| 626 | in_dev = inetdev_by_index(dev_net(dev), t->mlink); |
| 627 | if (in_dev) |
| 628 | ip_mc_dec_group(in_dev, t->parms.iph.daddr); |
| 629 | } |
| 630 | return 0; |
| 631 | } |
| 632 | #endif |
| 633 | |
| 634 | static const struct net_device_ops ipgre_netdev_ops = { |
| 635 | .ndo_init = ipgre_tunnel_init, |
| 636 | .ndo_uninit = ip_tunnel_uninit, |
| 637 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
| 638 | .ndo_open = ipgre_open, |
| 639 | .ndo_stop = ipgre_close, |
| 640 | #endif |
| 641 | .ndo_start_xmit = ipgre_xmit, |
| 642 | .ndo_do_ioctl = ipgre_tunnel_ioctl, |
| 643 | .ndo_change_mtu = ip_tunnel_change_mtu, |
| 644 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
| 645 | }; |
| 646 | |
| 647 | #define GRE_FEATURES (NETIF_F_SG | \ |
| 648 | NETIF_F_FRAGLIST | \ |
| 649 | NETIF_F_HIGHDMA | \ |
| 650 | NETIF_F_HW_CSUM) |
| 651 | |
| 652 | static void ipgre_tunnel_setup(struct net_device *dev) |
| 653 | { |
| 654 | dev->netdev_ops = &ipgre_netdev_ops; |
| 655 | dev->type = ARPHRD_IPGRE; |
| 656 | ip_tunnel_setup(dev, ipgre_net_id); |
| 657 | } |
| 658 | |
| 659 | static void __gre_tunnel_init(struct net_device *dev) |
| 660 | { |
| 661 | struct ip_tunnel *tunnel; |
| 662 | |
| 663 | tunnel = netdev_priv(dev); |
| 664 | tunnel->hlen = ip_gre_calc_hlen(tunnel->parms.o_flags); |
| 665 | tunnel->parms.iph.protocol = IPPROTO_GRE; |
| 666 | |
| 667 | dev->needed_headroom = LL_MAX_HEADER + sizeof(struct iphdr) + 4; |
| 668 | dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr) - 4; |
| 669 | |
| 670 | dev->features |= NETIF_F_NETNS_LOCAL | GRE_FEATURES; |
| 671 | dev->hw_features |= GRE_FEATURES; |
| 672 | |
| 673 | if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) { |
| 674 | /* TCP offload with GRE SEQ is not supported. */ |
| 675 | dev->features |= NETIF_F_GSO_SOFTWARE; |
| 676 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; |
| 677 | /* Can use a lockless transmit, unless we generate |
| 678 | * output sequences |
| 679 | */ |
| 680 | dev->features |= NETIF_F_LLTX; |
| 681 | } |
| 682 | } |
| 683 | |
| 684 | static int ipgre_tunnel_init(struct net_device *dev) |
| 685 | { |
| 686 | struct ip_tunnel *tunnel = netdev_priv(dev); |
| 687 | struct iphdr *iph = &tunnel->parms.iph; |
| 688 | |
| 689 | __gre_tunnel_init(dev); |
| 690 | |
| 691 | memcpy(dev->dev_addr, &iph->saddr, 4); |
| 692 | memcpy(dev->broadcast, &iph->daddr, 4); |
| 693 | |
| 694 | dev->flags = IFF_NOARP; |
| 695 | dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
| 696 | dev->addr_len = 4; |
| 697 | |
| 698 | if (iph->daddr) { |
| 699 | #ifdef CONFIG_NET_IPGRE_BROADCAST |
| 700 | if (ipv4_is_multicast(iph->daddr)) { |
| 701 | if (!iph->saddr) |
| 702 | return -EINVAL; |
| 703 | dev->flags = IFF_BROADCAST; |
| 704 | dev->header_ops = &ipgre_header_ops; |
| 705 | } |
| 706 | #endif |
| 707 | } else |
| 708 | dev->header_ops = &ipgre_header_ops; |
| 709 | |
| 710 | return ip_tunnel_init(dev); |
| 711 | } |
| 712 | |
| 713 | static const struct gre_protocol ipgre_protocol = { |
| 714 | .handler = ipgre_rcv, |
| 715 | .err_handler = ipgre_err, |
| 716 | }; |
| 717 | |
| 718 | static int __net_init ipgre_init_net(struct net *net) |
| 719 | { |
| 720 | return ip_tunnel_init_net(net, ipgre_net_id, &ipgre_link_ops, NULL); |
| 721 | } |
| 722 | |
| 723 | static void __net_exit ipgre_exit_net(struct net *net) |
| 724 | { |
| 725 | struct ip_tunnel_net *itn = net_generic(net, ipgre_net_id); |
| 726 | ip_tunnel_delete_net(itn); |
| 727 | } |
| 728 | |
| 729 | static struct pernet_operations ipgre_net_ops = { |
| 730 | .init = ipgre_init_net, |
| 731 | .exit = ipgre_exit_net, |
| 732 | .id = &ipgre_net_id, |
| 733 | .size = sizeof(struct ip_tunnel_net), |
| 734 | }; |
| 735 | |
| 736 | static int ipgre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[]) |
| 737 | { |
| 738 | __be16 flags; |
| 739 | |
| 740 | if (!data) |
| 741 | return 0; |
| 742 | |
| 743 | flags = 0; |
| 744 | if (data[IFLA_GRE_IFLAGS]) |
| 745 | flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]); |
| 746 | if (data[IFLA_GRE_OFLAGS]) |
| 747 | flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]); |
| 748 | if (flags & (GRE_VERSION|GRE_ROUTING)) |
| 749 | return -EINVAL; |
| 750 | |
| 751 | return 0; |
| 752 | } |
| 753 | |
| 754 | static int ipgre_tap_validate(struct nlattr *tb[], struct nlattr *data[]) |
| 755 | { |
| 756 | __be32 daddr; |
| 757 | |
| 758 | if (tb[IFLA_ADDRESS]) { |
| 759 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) |
| 760 | return -EINVAL; |
| 761 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) |
| 762 | return -EADDRNOTAVAIL; |
| 763 | } |
| 764 | |
| 765 | if (!data) |
| 766 | goto out; |
| 767 | |
| 768 | if (data[IFLA_GRE_REMOTE]) { |
| 769 | memcpy(&daddr, nla_data(data[IFLA_GRE_REMOTE]), 4); |
| 770 | if (!daddr) |
| 771 | return -EINVAL; |
| 772 | } |
| 773 | |
| 774 | out: |
| 775 | return ipgre_tunnel_validate(tb, data); |
| 776 | } |
| 777 | |
| 778 | static void ipgre_netlink_parms(struct nlattr *data[], struct nlattr *tb[], |
| 779 | struct ip_tunnel_parm *parms) |
| 780 | { |
| 781 | memset(parms, 0, sizeof(*parms)); |
| 782 | |
| 783 | parms->iph.protocol = IPPROTO_GRE; |
| 784 | |
| 785 | if (!data) |
| 786 | return; |
| 787 | |
| 788 | if (data[IFLA_GRE_LINK]) |
| 789 | parms->link = nla_get_u32(data[IFLA_GRE_LINK]); |
| 790 | |
| 791 | if (data[IFLA_GRE_IFLAGS]) |
| 792 | parms->i_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_IFLAGS])); |
| 793 | |
| 794 | if (data[IFLA_GRE_OFLAGS]) |
| 795 | parms->o_flags = gre_flags_to_tnl_flags(nla_get_be16(data[IFLA_GRE_OFLAGS])); |
| 796 | |
| 797 | if (data[IFLA_GRE_IKEY]) |
| 798 | parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]); |
| 799 | |
| 800 | if (data[IFLA_GRE_OKEY]) |
| 801 | parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]); |
| 802 | |
| 803 | if (data[IFLA_GRE_LOCAL]) |
| 804 | parms->iph.saddr = nla_get_be32(data[IFLA_GRE_LOCAL]); |
| 805 | |
| 806 | if (data[IFLA_GRE_REMOTE]) |
| 807 | parms->iph.daddr = nla_get_be32(data[IFLA_GRE_REMOTE]); |
| 808 | |
| 809 | if (data[IFLA_GRE_TTL]) |
| 810 | parms->iph.ttl = nla_get_u8(data[IFLA_GRE_TTL]); |
| 811 | |
| 812 | if (data[IFLA_GRE_TOS]) |
| 813 | parms->iph.tos = nla_get_u8(data[IFLA_GRE_TOS]); |
| 814 | |
| 815 | if (!data[IFLA_GRE_PMTUDISC] || nla_get_u8(data[IFLA_GRE_PMTUDISC])) |
| 816 | parms->iph.frag_off = htons(IP_DF); |
| 817 | } |
| 818 | |
| 819 | static int gre_tap_init(struct net_device *dev) |
| 820 | { |
| 821 | __gre_tunnel_init(dev); |
| 822 | |
| 823 | return ip_tunnel_init(dev); |
| 824 | } |
| 825 | |
| 826 | static const struct net_device_ops gre_tap_netdev_ops = { |
| 827 | .ndo_init = gre_tap_init, |
| 828 | .ndo_uninit = ip_tunnel_uninit, |
| 829 | .ndo_start_xmit = gre_tap_xmit, |
| 830 | .ndo_set_mac_address = eth_mac_addr, |
| 831 | .ndo_validate_addr = eth_validate_addr, |
| 832 | .ndo_change_mtu = ip_tunnel_change_mtu, |
| 833 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
| 834 | }; |
| 835 | |
| 836 | static void ipgre_tap_setup(struct net_device *dev) |
| 837 | { |
| 838 | ether_setup(dev); |
| 839 | dev->netdev_ops = &gre_tap_netdev_ops; |
| 840 | ip_tunnel_setup(dev, gre_tap_net_id); |
| 841 | } |
| 842 | |
| 843 | static int ipgre_newlink(struct net *src_net, struct net_device *dev, |
| 844 | struct nlattr *tb[], struct nlattr *data[]) |
| 845 | { |
| 846 | struct ip_tunnel_parm p; |
| 847 | |
| 848 | ipgre_netlink_parms(data, tb, &p); |
| 849 | return ip_tunnel_newlink(dev, tb, &p); |
| 850 | } |
| 851 | |
| 852 | static int ipgre_changelink(struct net_device *dev, struct nlattr *tb[], |
| 853 | struct nlattr *data[]) |
| 854 | { |
| 855 | struct ip_tunnel_parm p; |
| 856 | |
| 857 | ipgre_netlink_parms(data, tb, &p); |
| 858 | return ip_tunnel_changelink(dev, tb, &p); |
| 859 | } |
| 860 | |
| 861 | static size_t ipgre_get_size(const struct net_device *dev) |
| 862 | { |
| 863 | return |
| 864 | /* IFLA_GRE_LINK */ |
| 865 | nla_total_size(4) + |
| 866 | /* IFLA_GRE_IFLAGS */ |
| 867 | nla_total_size(2) + |
| 868 | /* IFLA_GRE_OFLAGS */ |
| 869 | nla_total_size(2) + |
| 870 | /* IFLA_GRE_IKEY */ |
| 871 | nla_total_size(4) + |
| 872 | /* IFLA_GRE_OKEY */ |
| 873 | nla_total_size(4) + |
| 874 | /* IFLA_GRE_LOCAL */ |
| 875 | nla_total_size(4) + |
| 876 | /* IFLA_GRE_REMOTE */ |
| 877 | nla_total_size(4) + |
| 878 | /* IFLA_GRE_TTL */ |
| 879 | nla_total_size(1) + |
| 880 | /* IFLA_GRE_TOS */ |
| 881 | nla_total_size(1) + |
| 882 | /* IFLA_GRE_PMTUDISC */ |
| 883 | nla_total_size(1) + |
| 884 | 0; |
| 885 | } |
| 886 | |
| 887 | static int ipgre_fill_info(struct sk_buff *skb, const struct net_device *dev) |
| 888 | { |
| 889 | struct ip_tunnel *t = netdev_priv(dev); |
| 890 | struct ip_tunnel_parm *p = &t->parms; |
| 891 | |
| 892 | if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) || |
| 893 | nla_put_be16(skb, IFLA_GRE_IFLAGS, tnl_flags_to_gre_flags(p->i_flags)) || |
| 894 | nla_put_be16(skb, IFLA_GRE_OFLAGS, tnl_flags_to_gre_flags(p->o_flags)) || |
| 895 | nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) || |
| 896 | nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) || |
| 897 | nla_put_be32(skb, IFLA_GRE_LOCAL, p->iph.saddr) || |
| 898 | nla_put_be32(skb, IFLA_GRE_REMOTE, p->iph.daddr) || |
| 899 | nla_put_u8(skb, IFLA_GRE_TTL, p->iph.ttl) || |
| 900 | nla_put_u8(skb, IFLA_GRE_TOS, p->iph.tos) || |
| 901 | nla_put_u8(skb, IFLA_GRE_PMTUDISC, |
| 902 | !!(p->iph.frag_off & htons(IP_DF)))) |
| 903 | goto nla_put_failure; |
| 904 | return 0; |
| 905 | |
| 906 | nla_put_failure: |
| 907 | return -EMSGSIZE; |
| 908 | } |
| 909 | |
| 910 | static const struct nla_policy ipgre_policy[IFLA_GRE_MAX + 1] = { |
| 911 | [IFLA_GRE_LINK] = { .type = NLA_U32 }, |
| 912 | [IFLA_GRE_IFLAGS] = { .type = NLA_U16 }, |
| 913 | [IFLA_GRE_OFLAGS] = { .type = NLA_U16 }, |
| 914 | [IFLA_GRE_IKEY] = { .type = NLA_U32 }, |
| 915 | [IFLA_GRE_OKEY] = { .type = NLA_U32 }, |
| 916 | [IFLA_GRE_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, |
| 917 | [IFLA_GRE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, |
| 918 | [IFLA_GRE_TTL] = { .type = NLA_U8 }, |
| 919 | [IFLA_GRE_TOS] = { .type = NLA_U8 }, |
| 920 | [IFLA_GRE_PMTUDISC] = { .type = NLA_U8 }, |
| 921 | }; |
| 922 | |
| 923 | static struct rtnl_link_ops ipgre_link_ops __read_mostly = { |
| 924 | .kind = "gre", |
| 925 | .maxtype = IFLA_GRE_MAX, |
| 926 | .policy = ipgre_policy, |
| 927 | .priv_size = sizeof(struct ip_tunnel), |
| 928 | .setup = ipgre_tunnel_setup, |
| 929 | .validate = ipgre_tunnel_validate, |
| 930 | .newlink = ipgre_newlink, |
| 931 | .changelink = ipgre_changelink, |
| 932 | .dellink = ip_tunnel_dellink, |
| 933 | .get_size = ipgre_get_size, |
| 934 | .fill_info = ipgre_fill_info, |
| 935 | }; |
| 936 | |
| 937 | static struct rtnl_link_ops ipgre_tap_ops __read_mostly = { |
| 938 | .kind = "gretap", |
| 939 | .maxtype = IFLA_GRE_MAX, |
| 940 | .policy = ipgre_policy, |
| 941 | .priv_size = sizeof(struct ip_tunnel), |
| 942 | .setup = ipgre_tap_setup, |
| 943 | .validate = ipgre_tap_validate, |
| 944 | .newlink = ipgre_newlink, |
| 945 | .changelink = ipgre_changelink, |
| 946 | .dellink = ip_tunnel_dellink, |
| 947 | .get_size = ipgre_get_size, |
| 948 | .fill_info = ipgre_fill_info, |
| 949 | }; |
| 950 | |
| 951 | static int __net_init ipgre_tap_init_net(struct net *net) |
| 952 | { |
| 953 | return ip_tunnel_init_net(net, gre_tap_net_id, &ipgre_tap_ops, NULL); |
| 954 | } |
| 955 | |
| 956 | static void __net_exit ipgre_tap_exit_net(struct net *net) |
| 957 | { |
| 958 | struct ip_tunnel_net *itn = net_generic(net, gre_tap_net_id); |
| 959 | ip_tunnel_delete_net(itn); |
| 960 | } |
| 961 | |
| 962 | static struct pernet_operations ipgre_tap_net_ops = { |
| 963 | .init = ipgre_tap_init_net, |
| 964 | .exit = ipgre_tap_exit_net, |
| 965 | .id = &gre_tap_net_id, |
| 966 | .size = sizeof(struct ip_tunnel_net), |
| 967 | }; |
| 968 | |
| 969 | static int __init ipgre_init(void) |
| 970 | { |
| 971 | int err; |
| 972 | |
| 973 | pr_info("GRE over IPv4 tunneling driver\n"); |
| 974 | |
| 975 | err = register_pernet_device(&ipgre_net_ops); |
| 976 | if (err < 0) |
| 977 | return err; |
| 978 | |
| 979 | err = register_pernet_device(&ipgre_tap_net_ops); |
| 980 | if (err < 0) |
| 981 | goto pnet_tap_faied; |
| 982 | |
| 983 | err = gre_add_protocol(&ipgre_protocol, GREPROTO_CISCO); |
| 984 | if (err < 0) { |
| 985 | pr_info("%s: can't add protocol\n", __func__); |
| 986 | goto add_proto_failed; |
| 987 | } |
| 988 | |
| 989 | err = rtnl_link_register(&ipgre_link_ops); |
| 990 | if (err < 0) |
| 991 | goto rtnl_link_failed; |
| 992 | |
| 993 | err = rtnl_link_register(&ipgre_tap_ops); |
| 994 | if (err < 0) |
| 995 | goto tap_ops_failed; |
| 996 | |
| 997 | return 0; |
| 998 | |
| 999 | tap_ops_failed: |
| 1000 | rtnl_link_unregister(&ipgre_link_ops); |
| 1001 | rtnl_link_failed: |
| 1002 | gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO); |
| 1003 | add_proto_failed: |
| 1004 | unregister_pernet_device(&ipgre_tap_net_ops); |
| 1005 | pnet_tap_faied: |
| 1006 | unregister_pernet_device(&ipgre_net_ops); |
| 1007 | return err; |
| 1008 | } |
| 1009 | |
| 1010 | static void __exit ipgre_fini(void) |
| 1011 | { |
| 1012 | rtnl_link_unregister(&ipgre_tap_ops); |
| 1013 | rtnl_link_unregister(&ipgre_link_ops); |
| 1014 | if (gre_del_protocol(&ipgre_protocol, GREPROTO_CISCO) < 0) |
| 1015 | pr_info("%s: can't remove protocol\n", __func__); |
| 1016 | unregister_pernet_device(&ipgre_tap_net_ops); |
| 1017 | unregister_pernet_device(&ipgre_net_ops); |
| 1018 | } |
| 1019 | |
| 1020 | module_init(ipgre_init); |
| 1021 | module_exit(ipgre_fini); |
| 1022 | MODULE_LICENSE("GPL"); |
| 1023 | MODULE_ALIAS_RTNL_LINK("gre"); |
| 1024 | MODULE_ALIAS_RTNL_LINK("gretap"); |
| 1025 | MODULE_ALIAS_NETDEV("gre0"); |
| 1026 | MODULE_ALIAS_NETDEV("gretap0"); |