net/ipv4/ip_vti.c

   1 /*
   2  *      Linux NET3: IP/IP protocol decoder modified to support
   3  *                  virtual tunnel interface
   4  *
   5  *      Authors:
   6  *              Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  *
  13  */
  14
  15 /*
  16    This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
  17
  18    For comments look at net/ipv4/ip_gre.c --ANK
  19  */
  20
  21
  22 #include <linux/capability.h>
  23 #include <linux/module.h>
  24 #include <linux/types.h>
  25 #include <linux/kernel.h>
  26 #include <linux/uaccess.h>
  27 #include <linux/skbuff.h>
  28 #include <linux/netdevice.h>
  29 #include <linux/in.h>
  30 #include <linux/tcp.h>
  31 #include <linux/udp.h>
  32 #include <linux/if_arp.h>
  33 #include <linux/init.h>
  34 #include <linux/netfilter_ipv4.h>
  35 #include <linux/if_ether.h>
  36 #include <linux/icmpv6.h>
  37
  38 #include <net/sock.h>
  39 #include <net/ip.h>
  40 #include <net/icmp.h>
  41 #include <net/ip_tunnels.h>
  42 #include <net/inet_ecn.h>
  43 #include <net/xfrm.h>
  44 #include <net/net_namespace.h>
  45 #include <net/netns/generic.h>
  46
  47 static struct rtnl_link_ops vti_link_ops __read_mostly;
  48
  49 static int vti_net_id __read_mostly;
  50 static int vti_tunnel_init(struct net_device *dev);
  51
  52 static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
  53                      int encap_type)
  54 {
  55         struct ip_tunnel *tunnel;
  56         const struct iphdr *iph = ip_hdr(skb);
  57         struct net *net = dev_net(skb->dev);
  58         struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
  59
  60         tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
  61                                   iph->saddr, iph->daddr, 0);
  62         if (tunnel) {
  63                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
  64                         goto drop;
  65
  66                 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
  67
  68                 return xfrm_input(skb, nexthdr, spi, encap_type);
  69         }
  70
  71         return -EINVAL;
  72 drop:
  73         kfree_skb(skb);
  74         return 0;
  75 }
  76
  77 static int vti_rcv(struct sk_buff *skb)
  78 {
  79         XFRM_SPI_SKB_CB(skb)->family = AF_INET;
  80         XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr);
  81
  82         return vti_input(skb, ip_hdr(skb)->protocol, 0, 0);
  83 }
  84
  85 static int vti_rcv_cb(struct sk_buff *skb, int err)
  86 {
  87         unsigned short family;
  88         struct net_device *dev;
  89         struct pcpu_sw_netstats *tstats;
  90         struct xfrm_state *x;
  91         struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
  92         u32 orig_mark = skb->mark;
  93         int ret;
  94
  95         if (!tunnel)
  96                 return 1;
  97
  98         dev = tunnel->dev;
  99
 100         if (err) {
 101                 dev->stats.rx_errors++;
 102                 dev->stats.rx_dropped++;
 103
 104                 return 0;
 105         }
 106
 107         x = xfrm_input_state(skb);
 108         family = x->inner_mode->afinfo->family;
 109
 110         skb->mark = be32_to_cpu(tunnel->parms.i_key);
 111         ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
 112         skb->mark = orig_mark;
 113
 114         if (!ret)
 115                 return -EPERM;
 116
 117         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
 118         skb->dev = dev;
 119
 120         tstats = this_cpu_ptr(dev->tstats);
 121
 122         u64_stats_update_begin(&tstats->syncp);
 123         tstats->rx_packets++;
 124         tstats->rx_bytes += skb->len;
 125         u64_stats_update_end(&tstats->syncp);
 126
 127         return 0;
 128 }
 129
 130 static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src)
 131 {
 132         xfrm_address_t *daddr = (xfrm_address_t *)&dst;
 133         xfrm_address_t *saddr = (xfrm_address_t *)&src;
 134
 135         /* if there is no transform then this tunnel is not functional.
 136          * Or if the xfrm is not mode tunnel.
 137          */
 138         if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
 139             x->props.family != AF_INET)
 140                 return false;
 141
 142         if (!dst)
 143                 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET);
 144
 145         if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET))
 146                 return false;
 147
 148         return true;
 149 }
 150
 151 static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev,
 152                             struct flowi *fl)
 153 {
 154         struct ip_tunnel *tunnel = netdev_priv(dev);
 155         struct ip_tunnel_parm *parms = &tunnel->parms;
 156         struct dst_entry *dst = skb_dst(skb);
 157         struct net_device *tdev;        /* Device to other host */
 158         int err;
 159
 160         if (!dst) {
 161                 dev->stats.tx_carrier_errors++;
 162                 goto tx_error_icmp;
 163         }
 164
 165         dst_hold(dst);
 166         dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0);
 167         if (IS_ERR(dst)) {
 168                 dev->stats.tx_carrier_errors++;
 169                 goto tx_error_icmp;
 170         }
 171
 172         if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) {
 173                 dev->stats.tx_carrier_errors++;
 174                 dst_release(dst);
 175                 goto tx_error_icmp;
 176         }
 177
 178         tdev = dst->dev;
 179
 180         if (tdev == dev) {
 181                 dst_release(dst);
 182                 dev->stats.collisions++;
 183                 goto tx_error;
 184         }
 185
 186         if (tunnel->err_count > 0) {
 187                 if (time_before(jiffies,
 188                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 189                         tunnel->err_count--;
 190                         dst_link_failure(skb);
 191                 } else
 192                         tunnel->err_count = 0;
 193         }
 194
 195         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev)));
 196         skb_dst_set(skb, dst);
 197         skb->dev = skb_dst(skb)->dev;
 198
 199         err = dst_output(tunnel->net, skb->sk, skb);
 200         if (net_xmit_eval(err) == 0)
 201                 err = skb->len;
 202         iptunnel_xmit_stats(dev, err);
 203         return NETDEV_TX_OK;
 204
 205 tx_error_icmp:
 206         dst_link_failure(skb);
 207 tx_error:
 208         dev->stats.tx_errors++;
 209         kfree_skb(skb);
 210         return NETDEV_TX_OK;
 211 }
 212
 213 /* This function assumes it is being called from dev_queue_xmit()
 214  * and that skb is filled properly by that function.
 215  */
 216 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 217 {
 218         struct ip_tunnel *tunnel = netdev_priv(dev);
 219         struct flowi fl;
 220
 221         memset(&fl, 0, sizeof(fl));
 222
 223         switch (skb->protocol) {
 224         case htons(ETH_P_IP):
 225                 xfrm_decode_session(skb, &fl, AF_INET);
 226                 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 227                 break;
 228         case htons(ETH_P_IPV6):
 229                 xfrm_decode_session(skb, &fl, AF_INET6);
 230                 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
 231                 break;
 232         default:
 233                 dev->stats.tx_errors++;
 234                 dev_kfree_skb(skb);
 235                 return NETDEV_TX_OK;
 236         }
 237
 238         /* override mark with tunnel output key */
 239         fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
 240
 241         return vti_xmit(skb, dev, &fl);
 242 }
 243
 244 static int vti4_err(struct sk_buff *skb, u32 info)
 245 {
 246         __be32 spi;
 247         __u32 mark;
 248         struct xfrm_state *x;
 249         struct ip_tunnel *tunnel;
 250         struct ip_esp_hdr *esph;
 251         struct ip_auth_hdr *ah ;
 252         struct ip_comp_hdr *ipch;
 253         struct net *net = dev_net(skb->dev);
 254         const struct iphdr *iph = (const struct iphdr *)skb->data;
 255         int protocol = iph->protocol;
 256         struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
 257
 258         tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY,
 259                                   iph->daddr, iph->saddr, 0);
 260         if (!tunnel)
 261                 return -1;
 262
 263         mark = be32_to_cpu(tunnel->parms.o_key);
 264
 265         switch (protocol) {
 266         case IPPROTO_ESP:
 267                 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2));
 268                 spi = esph->spi;
 269                 break;
 270         case IPPROTO_AH:
 271                 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2));
 272                 spi = ah->spi;
 273                 break;
 274         case IPPROTO_COMP:
 275                 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2));
 276                 spi = htonl(ntohs(ipch->cpi));
 277                 break;
 278         default:
 279                 return 0;
 280         }
 281
 282         switch (icmp_hdr(skb)->type) {
 283         case ICMP_DEST_UNREACH:
 284                 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED)
 285                         return 0;
 286         case ICMP_REDIRECT:
 287                 break;
 288         default:
 289                 return 0;
 290         }
 291
 292         x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
 293                               spi, protocol, AF_INET);
 294         if (!x)
 295                 return 0;
 296
 297         if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH)
 298                 ipv4_update_pmtu(skb, net, info, 0, 0, protocol, 0);
 299         else
 300                 ipv4_redirect(skb, net, 0, 0, protocol, 0);
 301         xfrm_state_put(x);
 302
 303         return 0;
 304 }
 305
 306 static int
 307 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 308 {
 309         int err = 0;
 310         struct ip_tunnel_parm p;
 311
 312         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 313                 return -EFAULT;
 314
 315         if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) {
 316                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
 317                     p.iph.ihl != 5)
 318                         return -EINVAL;
 319         }
 320
 321         if (!(p.i_flags & GRE_KEY))
 322                 p.i_key = 0;
 323         if (!(p.o_flags & GRE_KEY))
 324                 p.o_key = 0;
 325
 326         p.i_flags = VTI_ISVTI;
 327
 328         err = ip_tunnel_ioctl(dev, &p, cmd);
 329         if (err)
 330                 return err;
 331
 332         if (cmd != SIOCDELTUNNEL) {
 333                 p.i_flags |= GRE_KEY;
 334                 p.o_flags |= GRE_KEY;
 335         }
 336
 337         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 338                 return -EFAULT;
 339         return 0;
 340 }
 341
 342 static const struct net_device_ops vti_netdev_ops = {
 343         .ndo_init       = vti_tunnel_init,
 344         .ndo_uninit     = ip_tunnel_uninit,
 345         .ndo_start_xmit = vti_tunnel_xmit,
 346         .ndo_do_ioctl   = vti_tunnel_ioctl,
 347         .ndo_change_mtu = ip_tunnel_change_mtu,
 348         .ndo_get_stats64 = ip_tunnel_get_stats64,
 349         .ndo_get_iflink = ip_tunnel_get_iflink,
 350 };
 351
 352 static void vti_tunnel_setup(struct net_device *dev)
 353 {
 354         dev->netdev_ops         = &vti_netdev_ops;
 355         dev->type               = ARPHRD_TUNNEL;
 356         ip_tunnel_setup(dev, vti_net_id);
 357 }
 358
 359 static int vti_tunnel_init(struct net_device *dev)
 360 {
 361         struct ip_tunnel *tunnel = netdev_priv(dev);
 362         struct iphdr *iph = &tunnel->parms.iph;
 363
 364         memcpy(dev->dev_addr, &iph->saddr, 4);
 365         memcpy(dev->broadcast, &iph->daddr, 4);
 366
 367         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
 368         dev->mtu                = ETH_DATA_LEN;
 369         dev->flags              = IFF_NOARP;
 370         dev->addr_len           = 4;
 371         dev->features           |= NETIF_F_LLTX;
 372         netif_keep_dst(dev);
 373
 374         return ip_tunnel_init(dev);
 375 }
 376
 377 static void __net_init vti_fb_tunnel_init(struct net_device *dev)
 378 {
 379         struct ip_tunnel *tunnel = netdev_priv(dev);
 380         struct iphdr *iph = &tunnel->parms.iph;
 381
 382         iph->version            = 4;
 383         iph->protocol           = IPPROTO_IPIP;
 384         iph->ihl                = 5;
 385 }
 386
 387 static struct xfrm4_protocol vti_esp4_protocol __read_mostly = {
 388         .handler        =       vti_rcv,
 389         .input_handler  =       vti_input,
 390         .cb_handler     =       vti_rcv_cb,
 391         .err_handler    =       vti4_err,
 392         .priority       =       100,
 393 };
 394
 395 static struct xfrm4_protocol vti_ah4_protocol __read_mostly = {
 396         .handler        =       vti_rcv,
 397         .input_handler  =       vti_input,
 398         .cb_handler     =       vti_rcv_cb,
 399         .err_handler    =       vti4_err,
 400         .priority       =       100,
 401 };
 402
 403 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = {
 404         .handler        =       vti_rcv,
 405         .input_handler  =       vti_input,
 406         .cb_handler     =       vti_rcv_cb,
 407         .err_handler    =       vti4_err,
 408         .priority       =       100,
 409 };
 410
 411 static int __net_init vti_init_net(struct net *net)
 412 {
 413         int err;
 414         struct ip_tunnel_net *itn;
 415
 416         err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0");
 417         if (err)
 418                 return err;
 419         itn = net_generic(net, vti_net_id);
 420         vti_fb_tunnel_init(itn->fb_tunnel_dev);
 421         return 0;
 422 }
 423
 424 static void __net_exit vti_exit_net(struct net *net)
 425 {
 426         struct ip_tunnel_net *itn = net_generic(net, vti_net_id);
 427         ip_tunnel_delete_net(itn, &vti_link_ops);
 428 }
 429
 430 static struct pernet_operations vti_net_ops = {
 431         .init = vti_init_net,
 432         .exit = vti_exit_net,
 433         .id   = &vti_net_id,
 434         .size = sizeof(struct ip_tunnel_net),
 435 };
 436
 437 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
 438 {
 439         return 0;
 440 }
 441
 442 static void vti_netlink_parms(struct nlattr *data[],
 443                               struct ip_tunnel_parm *parms)
 444 {
 445         memset(parms, 0, sizeof(*parms));
 446
 447         parms->iph.protocol = IPPROTO_IPIP;
 448
 449         if (!data)
 450                 return;
 451
 452         parms->i_flags = VTI_ISVTI;
 453
 454         if (data[IFLA_VTI_LINK])
 455                 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
 456
 457         if (data[IFLA_VTI_IKEY])
 458                 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
 459
 460         if (data[IFLA_VTI_OKEY])
 461                 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
 462
 463         if (data[IFLA_VTI_LOCAL])
 464                 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]);
 465
 466         if (data[IFLA_VTI_REMOTE])
 467                 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]);
 468
 469 }
 470
 471 static int vti_newlink(struct net *src_net, struct net_device *dev,
 472                        struct nlattr *tb[], struct nlattr *data[])
 473 {
 474         struct ip_tunnel_parm parms;
 475
 476         vti_netlink_parms(data, &parms);
 477         return ip_tunnel_newlink(dev, tb, &parms);
 478 }
 479
 480 static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
 481                           struct nlattr *data[])
 482 {
 483         struct ip_tunnel_parm p;
 484
 485         vti_netlink_parms(data, &p);
 486         return ip_tunnel_changelink(dev, tb, &p);
 487 }
 488
 489 static size_t vti_get_size(const struct net_device *dev)
 490 {
 491         return
 492                 /* IFLA_VTI_LINK */
 493                 nla_total_size(4) +
 494                 /* IFLA_VTI_IKEY */
 495                 nla_total_size(4) +
 496                 /* IFLA_VTI_OKEY */
 497                 nla_total_size(4) +
 498                 /* IFLA_VTI_LOCAL */
 499                 nla_total_size(4) +
 500                 /* IFLA_VTI_REMOTE */
 501                 nla_total_size(4) +
 502                 0;
 503 }
 504
 505 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
 506 {
 507         struct ip_tunnel *t = netdev_priv(dev);
 508         struct ip_tunnel_parm *p = &t->parms;
 509
 510         nla_put_u32(skb, IFLA_VTI_LINK, p->link);
 511         nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key);
 512         nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key);
 513         nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr);
 514         nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr);
 515
 516         return 0;
 517 }
 518
 519 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
 520         [IFLA_VTI_LINK]         = { .type = NLA_U32 },
 521         [IFLA_VTI_IKEY]         = { .type = NLA_U32 },
 522         [IFLA_VTI_OKEY]         = { .type = NLA_U32 },
 523         [IFLA_VTI_LOCAL]        = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
 524         [IFLA_VTI_REMOTE]       = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
 525 };
 526
 527 static struct rtnl_link_ops vti_link_ops __read_mostly = {
 528         .kind           = "vti",
 529         .maxtype        = IFLA_VTI_MAX,
 530         .policy         = vti_policy,
 531         .priv_size      = sizeof(struct ip_tunnel),
 532         .setup          = vti_tunnel_setup,
 533         .validate       = vti_tunnel_validate,
 534         .newlink        = vti_newlink,
 535         .changelink     = vti_changelink,
 536         .dellink        = ip_tunnel_dellink,
 537         .get_size       = vti_get_size,
 538         .fill_info      = vti_fill_info,
 539         .get_link_net   = ip_tunnel_get_link_net,
 540 };
 541
 542 static int __init vti_init(void)
 543 {
 544         const char *msg;
 545         int err;
 546
 547         pr_info("IPv4 over IPsec tunneling driver\n");
 548
 549         msg = "tunnel device";
 550         err = register_pernet_device(&vti_net_ops);
 551         if (err < 0)
 552                 goto pernet_dev_failed;
 553
 554         msg = "tunnel protocols";
 555         err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP);
 556         if (err < 0)
 557                 goto xfrm_proto_esp_failed;
 558         err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH);
 559         if (err < 0)
 560                 goto xfrm_proto_ah_failed;
 561         err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP);
 562         if (err < 0)
 563                 goto xfrm_proto_comp_failed;
 564
 565         msg = "netlink interface";
 566         err = rtnl_link_register(&vti_link_ops);
 567         if (err < 0)
 568                 goto rtnl_link_failed;
 569
 570         return err;
 571
 572 rtnl_link_failed:
 573         xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
 574 xfrm_proto_comp_failed:
 575         xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
 576 xfrm_proto_ah_failed:
 577         xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
 578 xfrm_proto_esp_failed:
 579         unregister_pernet_device(&vti_net_ops);
 580 pernet_dev_failed:
 581         pr_err("vti init: failed to register %s\n", msg);
 582         return err;
 583 }
 584
 585 static void __exit vti_fini(void)
 586 {
 587         rtnl_link_unregister(&vti_link_ops);
 588         xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP);
 589         xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH);
 590         xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP);
 591         unregister_pernet_device(&vti_net_ops);
 592 }
 593
 594 module_init(vti_init);
 595 module_exit(vti_fini);
 596 MODULE_LICENSE("GPL");
 597 MODULE_ALIAS_RTNL_LINK("vti");
 598 MODULE_ALIAS_NETDEV("ip_vti0");