1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Linux NET3: IP/IP protocol decoder modified to support 4 * virtual tunnel interface 5 * 6 * Authors: 7 * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012 8 */ 9 10 /* 11 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c 12 13 For comments look at net/ipv4/ip_gre.c --ANK 14 */ 15 16 17 #include <linux/capability.h> 18 #include <linux/module.h> 19 #include <linux/types.h> 20 #include <linux/kernel.h> 21 #include <linux/uaccess.h> 22 #include <linux/skbuff.h> 23 #include <linux/netdevice.h> 24 #include <linux/in.h> 25 #include <linux/tcp.h> 26 #include <linux/udp.h> 27 #include <linux/if_arp.h> 28 #include <linux/init.h> 29 #include <linux/netfilter_ipv4.h> 30 #include <linux/if_ether.h> 31 #include <linux/icmpv6.h> 32 33 #include <net/sock.h> 34 #include <net/ip.h> 35 #include <net/icmp.h> 36 #include <net/ip_tunnels.h> 37 #include <net/inet_ecn.h> 38 #include <net/xfrm.h> 39 #include <net/net_namespace.h> 40 #include <net/netns/generic.h> 41 42 static struct rtnl_link_ops vti_link_ops __read_mostly; 43 44 static unsigned int vti_net_id __read_mostly; 45 static int vti_tunnel_init(struct net_device *dev); 46 47 static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi, 48 int encap_type, bool update_skb_dev) 49 { 50 struct ip_tunnel *tunnel; 51 const struct iphdr *iph = ip_hdr(skb); 52 struct net *net = dev_net(skb->dev); 53 struct ip_tunnel_net *itn = net_generic(net, vti_net_id); 54 55 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY, 56 iph->saddr, iph->daddr, 0); 57 if (tunnel) { 58 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) 59 goto drop; 60 61 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel; 62 63 if (update_skb_dev) 64 skb->dev = tunnel->dev; 65 66 return xfrm_input(skb, nexthdr, spi, encap_type); 67 } 68 69 return -EINVAL; 70 drop: 71 kfree_skb(skb); 72 return 0; 73 } 74 75 static int vti_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi, 76 int encap_type) 77 { 78 return vti_input(skb, nexthdr, spi, encap_type, false); 79 } 80 81 static int vti_rcv(struct sk_buff *skb, __be32 spi, bool update_skb_dev) 82 { 83 XFRM_SPI_SKB_CB(skb)->family = AF_INET; 84 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct iphdr, daddr); 85 86 return vti_input(skb, ip_hdr(skb)->protocol, spi, 0, update_skb_dev); 87 } 88 89 static int vti_rcv_proto(struct sk_buff *skb) 90 { 91 return vti_rcv(skb, 0, false); 92 } 93 94 static int vti_rcv_tunnel(struct sk_buff *skb) 95 { 96 return vti_rcv(skb, ip_hdr(skb)->saddr, true); 97 } 98 99 static int vti_rcv_cb(struct sk_buff *skb, int err) 100 { 101 unsigned short family; 102 struct net_device *dev; 103 struct pcpu_sw_netstats *tstats; 104 struct xfrm_state *x; 105 const struct xfrm_mode *inner_mode; 106 struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4; 107 u32 orig_mark = skb->mark; 108 int ret; 109 110 if (!tunnel) 111 return 1; 112 113 dev = tunnel->dev; 114 115 if (err) { 116 dev->stats.rx_errors++; 117 dev->stats.rx_dropped++; 118 119 return 0; 120 } 121 122 x = xfrm_input_state(skb); 123 124 inner_mode = &x->inner_mode; 125 126 if (x->sel.family == AF_UNSPEC) { 127 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol); 128 if (inner_mode == NULL) { 129 XFRM_INC_STATS(dev_net(skb->dev), 130 LINUX_MIB_XFRMINSTATEMODEERROR); 131 return -EINVAL; 132 } 133 } 134 135 family = inner_mode->family; 136 137 skb->mark = be32_to_cpu(tunnel->parms.i_key); 138 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family); 139 skb->mark = orig_mark; 140 141 if (!ret) 142 return -EPERM; 143 144 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev))); 145 skb->dev = dev; 146 147 tstats = this_cpu_ptr(dev->tstats); 148 149 u64_stats_update_begin(&tstats->syncp); 150 tstats->rx_packets++; 151 tstats->rx_bytes += skb->len; 152 u64_stats_update_end(&tstats->syncp); 153 154 return 0; 155 } 156 157 static bool vti_state_check(const struct xfrm_state *x, __be32 dst, __be32 src) 158 { 159 xfrm_address_t *daddr = (xfrm_address_t *)&dst; 160 xfrm_address_t *saddr = (xfrm_address_t *)&src; 161 162 /* if there is no transform then this tunnel is not functional. 163 * Or if the xfrm is not mode tunnel. 164 */ 165 if (!x || x->props.mode != XFRM_MODE_TUNNEL || 166 x->props.family != AF_INET) 167 return false; 168 169 if (!dst) 170 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET); 171 172 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET)) 173 return false; 174 175 return true; 176 } 177 178 static netdev_tx_t vti_xmit(struct sk_buff *skb, struct net_device *dev, 179 struct flowi *fl) 180 { 181 struct ip_tunnel *tunnel = netdev_priv(dev); 182 struct ip_tunnel_parm *parms = &tunnel->parms; 183 struct dst_entry *dst = skb_dst(skb); 184 struct net_device *tdev; /* Device to other host */ 185 int pkt_len = skb->len; 186 int err; 187 int mtu; 188 189 if (!dst) { 190 dev->stats.tx_carrier_errors++; 191 goto tx_error_icmp; 192 } 193 194 dst_hold(dst); 195 dst = xfrm_lookup(tunnel->net, dst, fl, NULL, 0); 196 if (IS_ERR(dst)) { 197 dev->stats.tx_carrier_errors++; 198 goto tx_error_icmp; 199 } 200 201 if (!vti_state_check(dst->xfrm, parms->iph.daddr, parms->iph.saddr)) { 202 dev->stats.tx_carrier_errors++; 203 dst_release(dst); 204 goto tx_error_icmp; 205 } 206 207 tdev = dst->dev; 208 209 if (tdev == dev) { 210 dst_release(dst); 211 dev->stats.collisions++; 212 goto tx_error; 213 } 214 215 mtu = dst_mtu(dst); 216 if (skb->len > mtu) { 217 skb_dst_update_pmtu(skb, mtu); 218 if (skb->protocol == htons(ETH_P_IP)) { 219 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, 220 htonl(mtu)); 221 } else { 222 if (mtu < IPV6_MIN_MTU) 223 mtu = IPV6_MIN_MTU; 224 225 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu); 226 } 227 228 dst_release(dst); 229 goto tx_error; 230 } 231 232 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(dev))); 233 skb_dst_set(skb, dst); 234 skb->dev = skb_dst(skb)->dev; 235 236 err = dst_output(tunnel->net, skb->sk, skb); 237 if (net_xmit_eval(err) == 0) 238 err = pkt_len; 239 iptunnel_xmit_stats(dev, err); 240 return NETDEV_TX_OK; 241 242 tx_error_icmp: 243 dst_link_failure(skb); 244 tx_error: 245 dev->stats.tx_errors++; 246 kfree_skb(skb); 247 return NETDEV_TX_OK; 248 } 249 250 /* This function assumes it is being called from dev_queue_xmit() 251 * and that skb is filled properly by that function. 252 */ 253 static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev) 254 { 255 struct ip_tunnel *tunnel = netdev_priv(dev); 256 struct flowi fl; 257 258 if (!pskb_inet_may_pull(skb)) 259 goto tx_err; 260 261 memset(&fl, 0, sizeof(fl)); 262 263 switch (skb->protocol) { 264 case htons(ETH_P_IP): 265 xfrm_decode_session(skb, &fl, AF_INET); 266 memset(IPCB(skb), 0, sizeof(*IPCB(skb))); 267 break; 268 case htons(ETH_P_IPV6): 269 xfrm_decode_session(skb, &fl, AF_INET6); 270 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb))); 271 break; 272 default: 273 goto tx_err; 274 } 275 276 /* override mark with tunnel output key */ 277 fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key); 278 279 return vti_xmit(skb, dev, &fl); 280 281 tx_err: 282 dev->stats.tx_errors++; 283 kfree_skb(skb); 284 return NETDEV_TX_OK; 285 } 286 287 static int vti4_err(struct sk_buff *skb, u32 info) 288 { 289 __be32 spi; 290 __u32 mark; 291 struct xfrm_state *x; 292 struct ip_tunnel *tunnel; 293 struct ip_esp_hdr *esph; 294 struct ip_auth_hdr *ah ; 295 struct ip_comp_hdr *ipch; 296 struct net *net = dev_net(skb->dev); 297 const struct iphdr *iph = (const struct iphdr *)skb->data; 298 int protocol = iph->protocol; 299 struct ip_tunnel_net *itn = net_generic(net, vti_net_id); 300 301 tunnel = ip_tunnel_lookup(itn, skb->dev->ifindex, TUNNEL_NO_KEY, 302 iph->daddr, iph->saddr, 0); 303 if (!tunnel) 304 return -1; 305 306 mark = be32_to_cpu(tunnel->parms.o_key); 307 308 switch (protocol) { 309 case IPPROTO_ESP: 310 esph = (struct ip_esp_hdr *)(skb->data+(iph->ihl<<2)); 311 spi = esph->spi; 312 break; 313 case IPPROTO_AH: 314 ah = (struct ip_auth_hdr *)(skb->data+(iph->ihl<<2)); 315 spi = ah->spi; 316 break; 317 case IPPROTO_COMP: 318 ipch = (struct ip_comp_hdr *)(skb->data+(iph->ihl<<2)); 319 spi = htonl(ntohs(ipch->cpi)); 320 break; 321 default: 322 return 0; 323 } 324 325 switch (icmp_hdr(skb)->type) { 326 case ICMP_DEST_UNREACH: 327 if (icmp_hdr(skb)->code != ICMP_FRAG_NEEDED) 328 return 0; 329 case ICMP_REDIRECT: 330 break; 331 default: 332 return 0; 333 } 334 335 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr, 336 spi, protocol, AF_INET); 337 if (!x) 338 return 0; 339 340 if (icmp_hdr(skb)->type == ICMP_DEST_UNREACH) 341 ipv4_update_pmtu(skb, net, info, 0, protocol); 342 else 343 ipv4_redirect(skb, net, 0, protocol); 344 xfrm_state_put(x); 345 346 return 0; 347 } 348 349 static int 350 vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd) 351 { 352 int err = 0; 353 struct ip_tunnel_parm p; 354 355 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) 356 return -EFAULT; 357 358 if (cmd == SIOCADDTUNNEL || cmd == SIOCCHGTUNNEL) { 359 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP || 360 p.iph.ihl != 5) 361 return -EINVAL; 362 } 363 364 if (!(p.i_flags & GRE_KEY)) 365 p.i_key = 0; 366 if (!(p.o_flags & GRE_KEY)) 367 p.o_key = 0; 368 369 p.i_flags = VTI_ISVTI; 370 371 err = ip_tunnel_ioctl(dev, &p, cmd); 372 if (err) 373 return err; 374 375 if (cmd != SIOCDELTUNNEL) { 376 p.i_flags |= GRE_KEY; 377 p.o_flags |= GRE_KEY; 378 } 379 380 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) 381 return -EFAULT; 382 return 0; 383 } 384 385 static const struct net_device_ops vti_netdev_ops = { 386 .ndo_init = vti_tunnel_init, 387 .ndo_uninit = ip_tunnel_uninit, 388 .ndo_start_xmit = vti_tunnel_xmit, 389 .ndo_do_ioctl = vti_tunnel_ioctl, 390 .ndo_change_mtu = ip_tunnel_change_mtu, 391 .ndo_get_stats64 = ip_tunnel_get_stats64, 392 .ndo_get_iflink = ip_tunnel_get_iflink, 393 }; 394 395 static void vti_tunnel_setup(struct net_device *dev) 396 { 397 dev->netdev_ops = &vti_netdev_ops; 398 dev->type = ARPHRD_TUNNEL; 399 ip_tunnel_setup(dev, vti_net_id); 400 } 401 402 static int vti_tunnel_init(struct net_device *dev) 403 { 404 struct ip_tunnel *tunnel = netdev_priv(dev); 405 struct iphdr *iph = &tunnel->parms.iph; 406 407 memcpy(dev->dev_addr, &iph->saddr, 4); 408 memcpy(dev->broadcast, &iph->daddr, 4); 409 410 dev->flags = IFF_NOARP; 411 dev->addr_len = 4; 412 dev->features |= NETIF_F_LLTX; 413 netif_keep_dst(dev); 414 415 return ip_tunnel_init(dev); 416 } 417 418 static void __net_init vti_fb_tunnel_init(struct net_device *dev) 419 { 420 struct ip_tunnel *tunnel = netdev_priv(dev); 421 struct iphdr *iph = &tunnel->parms.iph; 422 423 iph->version = 4; 424 iph->protocol = IPPROTO_IPIP; 425 iph->ihl = 5; 426 } 427 428 static struct xfrm4_protocol vti_esp4_protocol __read_mostly = { 429 .handler = vti_rcv_proto, 430 .input_handler = vti_input_proto, 431 .cb_handler = vti_rcv_cb, 432 .err_handler = vti4_err, 433 .priority = 100, 434 }; 435 436 static struct xfrm4_protocol vti_ah4_protocol __read_mostly = { 437 .handler = vti_rcv_proto, 438 .input_handler = vti_input_proto, 439 .cb_handler = vti_rcv_cb, 440 .err_handler = vti4_err, 441 .priority = 100, 442 }; 443 444 static struct xfrm4_protocol vti_ipcomp4_protocol __read_mostly = { 445 .handler = vti_rcv_proto, 446 .input_handler = vti_input_proto, 447 .cb_handler = vti_rcv_cb, 448 .err_handler = vti4_err, 449 .priority = 100, 450 }; 451 452 static struct xfrm_tunnel ipip_handler __read_mostly = { 453 .handler = vti_rcv_tunnel, 454 .err_handler = vti4_err, 455 .priority = 0, 456 }; 457 458 static int __net_init vti_init_net(struct net *net) 459 { 460 int err; 461 struct ip_tunnel_net *itn; 462 463 err = ip_tunnel_init_net(net, vti_net_id, &vti_link_ops, "ip_vti0"); 464 if (err) 465 return err; 466 itn = net_generic(net, vti_net_id); 467 if (itn->fb_tunnel_dev) 468 vti_fb_tunnel_init(itn->fb_tunnel_dev); 469 return 0; 470 } 471 472 static void __net_exit vti_exit_batch_net(struct list_head *list_net) 473 { 474 ip_tunnel_delete_nets(list_net, vti_net_id, &vti_link_ops); 475 } 476 477 static struct pernet_operations vti_net_ops = { 478 .init = vti_init_net, 479 .exit_batch = vti_exit_batch_net, 480 .id = &vti_net_id, 481 .size = sizeof(struct ip_tunnel_net), 482 }; 483 484 static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[], 485 struct netlink_ext_ack *extack) 486 { 487 return 0; 488 } 489 490 static void vti_netlink_parms(struct nlattr *data[], 491 struct ip_tunnel_parm *parms, 492 __u32 *fwmark) 493 { 494 memset(parms, 0, sizeof(*parms)); 495 496 parms->iph.protocol = IPPROTO_IPIP; 497 498 if (!data) 499 return; 500 501 parms->i_flags = VTI_ISVTI; 502 503 if (data[IFLA_VTI_LINK]) 504 parms->link = nla_get_u32(data[IFLA_VTI_LINK]); 505 506 if (data[IFLA_VTI_IKEY]) 507 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]); 508 509 if (data[IFLA_VTI_OKEY]) 510 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]); 511 512 if (data[IFLA_VTI_LOCAL]) 513 parms->iph.saddr = nla_get_in_addr(data[IFLA_VTI_LOCAL]); 514 515 if (data[IFLA_VTI_REMOTE]) 516 parms->iph.daddr = nla_get_in_addr(data[IFLA_VTI_REMOTE]); 517 518 if (data[IFLA_VTI_FWMARK]) 519 *fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]); 520 } 521 522 static int vti_newlink(struct net *src_net, struct net_device *dev, 523 struct nlattr *tb[], struct nlattr *data[], 524 struct netlink_ext_ack *extack) 525 { 526 struct ip_tunnel_parm parms; 527 __u32 fwmark = 0; 528 529 vti_netlink_parms(data, &parms, &fwmark); 530 return ip_tunnel_newlink(dev, tb, &parms, fwmark); 531 } 532 533 static int vti_changelink(struct net_device *dev, struct nlattr *tb[], 534 struct nlattr *data[], 535 struct netlink_ext_ack *extack) 536 { 537 struct ip_tunnel *t = netdev_priv(dev); 538 __u32 fwmark = t->fwmark; 539 struct ip_tunnel_parm p; 540 541 vti_netlink_parms(data, &p, &fwmark); 542 return ip_tunnel_changelink(dev, tb, &p, fwmark); 543 } 544 545 static size_t vti_get_size(const struct net_device *dev) 546 { 547 return 548 /* IFLA_VTI_LINK */ 549 nla_total_size(4) + 550 /* IFLA_VTI_IKEY */ 551 nla_total_size(4) + 552 /* IFLA_VTI_OKEY */ 553 nla_total_size(4) + 554 /* IFLA_VTI_LOCAL */ 555 nla_total_size(4) + 556 /* IFLA_VTI_REMOTE */ 557 nla_total_size(4) + 558 /* IFLA_VTI_FWMARK */ 559 nla_total_size(4) + 560 0; 561 } 562 563 static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev) 564 { 565 struct ip_tunnel *t = netdev_priv(dev); 566 struct ip_tunnel_parm *p = &t->parms; 567 568 if (nla_put_u32(skb, IFLA_VTI_LINK, p->link) || 569 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key) || 570 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key) || 571 nla_put_in_addr(skb, IFLA_VTI_LOCAL, p->iph.saddr) || 572 nla_put_in_addr(skb, IFLA_VTI_REMOTE, p->iph.daddr) || 573 nla_put_u32(skb, IFLA_VTI_FWMARK, t->fwmark)) 574 return -EMSGSIZE; 575 576 return 0; 577 } 578 579 static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = { 580 [IFLA_VTI_LINK] = { .type = NLA_U32 }, 581 [IFLA_VTI_IKEY] = { .type = NLA_U32 }, 582 [IFLA_VTI_OKEY] = { .type = NLA_U32 }, 583 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, 584 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, 585 [IFLA_VTI_FWMARK] = { .type = NLA_U32 }, 586 }; 587 588 static struct rtnl_link_ops vti_link_ops __read_mostly = { 589 .kind = "vti", 590 .maxtype = IFLA_VTI_MAX, 591 .policy = vti_policy, 592 .priv_size = sizeof(struct ip_tunnel), 593 .setup = vti_tunnel_setup, 594 .validate = vti_tunnel_validate, 595 .newlink = vti_newlink, 596 .changelink = vti_changelink, 597 .dellink = ip_tunnel_dellink, 598 .get_size = vti_get_size, 599 .fill_info = vti_fill_info, 600 .get_link_net = ip_tunnel_get_link_net, 601 }; 602 603 static int __init vti_init(void) 604 { 605 const char *msg; 606 int err; 607 608 pr_info("IPv4 over IPsec tunneling driver\n"); 609 610 msg = "tunnel device"; 611 err = register_pernet_device(&vti_net_ops); 612 if (err < 0) 613 goto pernet_dev_failed; 614 615 msg = "tunnel protocols"; 616 err = xfrm4_protocol_register(&vti_esp4_protocol, IPPROTO_ESP); 617 if (err < 0) 618 goto xfrm_proto_esp_failed; 619 err = xfrm4_protocol_register(&vti_ah4_protocol, IPPROTO_AH); 620 if (err < 0) 621 goto xfrm_proto_ah_failed; 622 err = xfrm4_protocol_register(&vti_ipcomp4_protocol, IPPROTO_COMP); 623 if (err < 0) 624 goto xfrm_proto_comp_failed; 625 626 msg = "ipip tunnel"; 627 err = xfrm4_tunnel_register(&ipip_handler, AF_INET); 628 if (err < 0) 629 goto xfrm_tunnel_failed; 630 631 msg = "netlink interface"; 632 err = rtnl_link_register(&vti_link_ops); 633 if (err < 0) 634 goto rtnl_link_failed; 635 636 return err; 637 638 rtnl_link_failed: 639 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 640 xfrm_tunnel_failed: 641 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP); 642 xfrm_proto_comp_failed: 643 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH); 644 xfrm_proto_ah_failed: 645 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP); 646 xfrm_proto_esp_failed: 647 unregister_pernet_device(&vti_net_ops); 648 pernet_dev_failed: 649 pr_err("vti init: failed to register %s\n", msg); 650 return err; 651 } 652 653 static void __exit vti_fini(void) 654 { 655 rtnl_link_unregister(&vti_link_ops); 656 xfrm4_tunnel_deregister(&ipip_handler, AF_INET); 657 xfrm4_protocol_deregister(&vti_ipcomp4_protocol, IPPROTO_COMP); 658 xfrm4_protocol_deregister(&vti_ah4_protocol, IPPROTO_AH); 659 xfrm4_protocol_deregister(&vti_esp4_protocol, IPPROTO_ESP); 660 unregister_pernet_device(&vti_net_ops); 661 } 662 663 module_init(vti_init); 664 module_exit(vti_fini); 665 MODULE_LICENSE("GPL"); 666 MODULE_ALIAS_RTNL_LINK("vti"); 667 MODULE_ALIAS_NETDEV("ip_vti0"); 668